Environmental and Agricultural Modeling: Integrated Approaches for Policy Impact Assessment

Environmental and Agricultural Modelling Floor M. Brouwer Martin van Ittersum ● Editors Environmental and Agricu...

Author: Floor M. Brouwer | Martin van Ittersum

273 downloads 1183 Views 6MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

Environmental and Agricultural Modelling

Floor M. Brouwer Martin van Ittersum ●

Editors

Environmental and Agricultural Modelling Integrated Approaches for Policy Impact Assessment

Editors Floor M. Brouwer LEI Wageningen UR P.O. Box 29703 2502 LS The Hague The Netherlands [email protected]

Martin van Ittersum Wageningen University Plant Production Systems Group P.O. Box 430 6700 AK Wageningen The Netherlands [email protected]

ISBN 978-90-481-3618-6 e-ISBN 978-90-481-3619-3 DOI 10.1007/978-90-481-3619-3 Springer Dordrecht Heidelberg London New York Library of Congress Control Number: 2009943042 © Springer Science+Business Media B.V. 2010 No part of this work may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, microfilming, recording or otherwise, without written permission from the Publisher, with the exception of any material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work. Printed on acid-free paper Springer is part of Springer Science+Business Media (www.springer.com)

Preface

Agriculture faces increasingly the challenge of balancing its multiple functions in a sustainable way. Integrated assessment and modelling (IAM) can provide insight into the potential impacts of policy changes. However, concepts to address the wide range of issues and functions typical for agriculture are still scarce. The volume reviews and presents our current understanding of integrated and working tools to assess and compute, ex-ante, alternative agricultural and environmental policy options, allowing: 1. Analysis at the full range of scales (farm to European Union and global), whilst focusing on the most important issues emerging at each scale 2. Analysis of the environmental, economic and social contributions of agricultural systems towards sustainable rural development and rural viability 3. Analysis of a broad range of issues, such as climate change, environmental policies, rural development options, effects of an enlarging EU, international competition and effects on developing countries This volume has a strong ‘lessons to be learnt’ emphasis, to facilitate and promote the use and further development of integrated assessment tools to support policies promoting agricultural development in support of sustainable development. The book is an effort from many contributors whose input is much appreciated. The work presented in this volume has been (co-)funded by the SEAMLESS integrated project (January 2005–March 2009), EU 6th Framework Programme for Research Technological Development and Demonstration, Priority 1.1.6.3 Global Change and Ecosystems (European Commission, DG RTD, contract no. 010036-2); we gratefully acknowledge this support. We also appreciate the assistance provided by Mrs. Eline Bazen (LEI Wageningen UR) who took responsibility for guiding the publication process and preparing the chapters of the book. July 2009

Floor Brouwer Martin van Ittersum

v

Contents

1 Introduction................................................................................................ Martin van Ittersum and Floor Brouwer

1

Part I Indicators and Concepts in Practice 2 Assessment of Multifunctionality and Jointness of Production............. Nadine Turpin, Lee Stapleton, Eric Perret, C. Martijn van der Heide, Guy Garrod, Floor Brouwer, Vaclav Voltr, and Dominique Cairol

11

3 The Institutional Dimension in Policy Assessment................................. Insa Theesfeld, Christian Schleyer, Konrad Hagedorn, Jean-Marc Callois, Olivier Aznar, and Johanna Alkan Olsson

37

Part II Advancements in Modelling, Data and Software 4 A Component-Based Framework for Simulating Agricultural Production and Externalities.................................................................... Marcello Donatelli, Graham Russell, Andrea Emilio Rizzoli, Marco Acutis, Myriam Adam, Ioannis N. Athanasiadis, Matteo Balderacchi, Luca Bechini, Hatem Belhouchette, Gianni Bellocchi, Jacques-Eric Bergez, Marco Botta, Erik Braudeau, Simone Bregaglio, Laura Carlini, Eric Casellas, Florian Celette, Enrico Ceotto, Marie Hélène Charron-Moirez, Roberto Confalonieri, Marc Corbeels, Luca Criscuolo, Pablo Cruz, Andrea di Guardo, Domenico Ditto, Christian Dupraz, Michel Duru, Diego Fiorani, Antonella Gentile, Frank Ewert, Christian Gary, Ephrem Habyarimana, Claire Jouany, Kamel Kansou, Rob Knapen, Giovanni Lanza Filippi, Peter A. Leffelaar, Luisa Manici, Guillaume Martin, Pierre Martin, Eelco Meuter, Nora Mugueta, Rachmat Mulia, Meine van Noordwijk, Roelof Oomen, Alexandra Rosenmund, Vittorio Rossi, Francesca Salinari, Ariel Serrano, Andrea Sorce, Grégoire Vincent, Jean-Pierre Theau, Olivier Thérond, Marco Trevisan, Patrizia Trevisiol, Frits K. van Evert, Daniel Wallach, Jacques Wery, and Arezki Zerourou

63

vii

viii

Contents

5 A Generic Farming System Simulator................................................... 109 Kamel Louhichi, Sander Janssen, Argyris Kanellopoulos, Hongtao Li, Nina Borkowski, Guillermo Flichman, Huib Hengsdijk, Peter Zander, Maria Blanco Fonseca, Grete Stokstad, Ioannis N. Athanasiadis, Andrea E. Rizzoli, David Huber, Thomas Heckelei, and Martin van Ittersum 6 Visualising Changes in Agricultural Landscapes..................................... 133 Sébastien Griffon, Daniel Auclair, and Amélie Nespoulous 7 A Biophysical Typology in Agri-environmental Modelling.................. 159 Gerard Hazeu, Berien Elbersen, Erling Andersen, Bettina Baruth, Kees van Diepen, and Marc Metzger 8 The Use of Regional Typologies in the Assessment of Farms’ Performance............................................................................ 189 Ida J. Terluin, David Verhoog, and Frans E. Godeschalk 9 A Web-Based Software System for Model Integration in Impact Assessments of Agricultural and Environmental Policies..................................................................... 207 Jan-Erik Wien, Andrea Emilio Rizzoli, Rob Knapen, Ioannis Athanasiadis, Sander Janssen, Lorenzo Ruinelli, Ferdinando Villa, Mats Svensson, Patrik Wallman, Benny Jonsson, and Martin van Ittersum Part III Use and Extensions of Integrated Assessment Modelling 10 Evaluating Integrated Assessment Tools for Policy Support............... 237 Jacques-Eric Bergez, Marijke Kuiper, Olivier Thérond, Marie Taverne, Hatem Belhouchette, and Jacques Wery 11 A Comparison of CAPRI and SEAMLESS-IF as Integrated Modelling Systems............................................................ 257 Wolfgang Britz, Ignacio Pérez Domínguez, and Thomas Heckelei 12 Science–Policy Interfaces in Impact Assessment Procedures.............. 275 Ann-Katrin Bäcklund, Jean Paul Bousset, Sara Brogaard, Catherine Macombe, Marie Taverne, and Martin van Ittersum 13 Economic Principles of Monetary Valuation in Evaluation Studies............................................................................... 295 C. Martijn van der Heide, Neil A. Powe, and Ståle Navrud Index.................................................................................................................. 319

Contributors

Marco Acutis UNIMI – University of Milan, Department of Crop Science, Milan, Italy [email protected] Myriam Adam Plant Production Systems Group, Wageningen University P.O. Box 430, 6700 AK Wageningen, The Netherlands [email protected] Johanna Alkan Olsson Lund University, Centre for Sustainability Studies, Box 170, 221 00 Lund, Sweden [email protected] Erling Andersen University of Copenhagen, Faculty of Life Sciences, Forest & Landscape, Rolighedsvej 23, DK-1958 Frederiksberg C, Denmark [email protected] Ioannis N. Athanasiadis IDSIA – Istituto Dalle Molle di Studi sull’Intelligenza Artificiale, Galleria 2, CH-6928 Manno, Switzerland [email protected] Daniel Auclair INRA, UMR AMAP, Montpellier, F-34000, France [email protected] Olivier Aznar Cemagref – UMR Métafort Campus universitaire des Cézeaux, 24, avenue des Landais – BP 50085, 63172 Aubiére Cedex, France [email protected] Ann-Katrin Bäcklund Lund University, Department of Social and Economic Geography, Sölvegatan 12, S-223 61 Lund, Sweden [email protected]

ix

x

Contributors

Matteo Balderacchi UNICATT – Catholic University of Piacenza, Institute for Agricultural and Environmental Chemistry, Piacenza, Italy [email protected] Bettina Baruth Joint Research Centre, Institute for the Protection and Security of the Citizen, Monitoring Agricultural Resources Unit (MARS), AGRI4CAST TP 483, 21027 Ispra (VA), Italy [email protected] Luca Bechini UNIMI – University of Milan, Department of Crop Science, Milan, Italy [email protected] Hatem Belhouchette INRA – UMR SYSTEM (Systèmes de Culture Tropicaux et Méditerranéens) Montpellier, France [email protected] Gianni Bellocchi CRA-CIN – Agriculture Research Council, Via di Corticella, 133 – Bologna, Italy [email protected] Jacques-Eric Bergez INRA, UMR 1248 AGIR, BP 52627, F-31326 Castanet Tolosan, France [email protected] Maria Blanco Fonseca IAMM, 3191 route de Mende, 34090 Montpellier, France [email protected] Nina Borkowski Institute of Socio-Economics, Leibniz Centre for Agricultural Landscape Research (ZALF), Eberswalder Strasse, 84, D-15374, Müncheberg, Germany Present Address: Institute for Food and Resource Economics, Chair for Economic and Agricultural Policy, University of Bonn, Nussallee, 21, D-53115, Bonn, Germany [email protected] Marco Botta CRA-CIN – Agriculture Research Council, Via di Corticella, 133 – Bologna, Italy [email protected] Jean-Paul Bousset Cemagref, Dynamics and Functions of Rural Areas, 24 avenue des Landais, BP 50085, 63172 Aubière Cedex, France [email protected]

Contributors

xi

Erik Braudeau IRD – French Agricultural Research Centre for International Development, Performance of Tropical Production and Processing Systems Department, Montpellier, France [email protected] Simone Bregaglio CRA-CIN – Agriculture Research Council, Via di Corticella, 133 – Bologna, Italy [email protected] Wolfgang Britz University of Bonn, Institute for Food and Resource Economics, Chair for Economic and Agricultural Policy, Nussallee 21, D-53115 Bonn, Germany [email protected] Sara Brogaard Lund University, Centre for Sustainability Studies, Box 170, S-221 00 Lund, Sweden [email protected] Floor Brouwer LEI Wageningen UR, P.O. Box 29703, 2502 LS The Hague, The Netherlands [email protected] Dominique Cairol Cemagref, Direction Générale, GT – DTDG, Parc de Tourvoie, BP 44, 92163 ANTONY Cedex, France [email protected] Jean-Marc Callois Cemagref – UMR Métafort Campus universitaire des Cézeaux, 24, avenue des Landais – BP 50085, 63172 Aubiére Cedex, France [email protected] Laura Carlini CRA-CIN – Agriculture Research Council, Via di Corticella, 133 – Bologna, Italy [email protected] Eric Casellas INRA – UMR SYSTEM (Systèmes de Culture Tropicaux et Méditerranéens) Montpellier, France [email protected] Florian Celette INRA – UMR SYSTEM (Systèmes de Culture Tropicaux et Méditerranéens) Montpellier, France [email protected]

xii

Contributors

Enrico Ceotto CRA-CIN – Agriculture Research Council, Via di Corticella, 133 – Bologna, Italy [email protected] Marie Hélène Charron-Moirez INRA, UMR 1248 AGIR, BP 52627 , F-31326 Castanet Tolosan, France [email protected] Roberto Confalonieri JRC – Joint Research Centre, Institute for the Protection and Security of the Citizen, Agriculture Unit, Agri4cast Action, Ispra, Italy [email protected] Marc Corbeels CIRAD – Agricultural Research Centre for International Development, UMR System #1230, CIRAD-INRA-SupAgro, Montpellier, France [email protected] Luca Criscuolo CRA-CIN – Agriculture Research Council, Via di Corticella, 133 – Bologna, Italy [email protected] Pablo Cruz INRA, UMR 1248 AGIR, BP 52627, F-31326 Castanet Tolosan, France [email protected] Domenico Ditto UNIMI – University of Milan, Department of Crop Science, Milan, Italy [email protected] Marcello Donatelli CRA – Agriculture Research Council, Bologna, Italy JRC – Joint Research Centre, Institute for the Protection and Security of the Citizen, Agriculture Unit, Agri4cast Action, Ispra, Italy [email protected] Christian Dupraz INRA – UMR SYSTEM (Systèmes de Culture Tropicaux et Méditerranéens) Montpellier, France [email protected] Michel Duru INRA, UMR 1248 AGIR, BP 52627, F-31326 Castanet Tolosan, France [email protected] Berien Elbersen Wageningen UR, Alterra, Landscape Centre, Droevendaalsesteeg 3, 6708 PB Wageningen, The Netherlands [email protected]

Contributors

xiii

Frank Ewert Plant Production Systems group, Wageningen University, P.O. Box 430, 6700 AK Wageningen, The Netherlands Present Address: Institute of Crop Science and Resource Conservation (INRES), University of Bonn, Katzenburgweg 5, D-53115 Bonn, Germany [email protected] Diego Fiorani CRA-CIN – Agriculture Research Council, Via di Corticella, 133 – Bologna, Italy [email protected] Guillermo Flichman CIHEAM-Institut de Montpellier, 3191 route de Mende, 34093 Montpellier Cedex 05, France [email protected] Guy Garrod Centre for Rural Economy, School of Agriculture, Food and Rural Development, Newcastle University, Newcastle-upon-Tyne, NE1 7RU, United Kingdom [email protected] Christian Gary INRA – UMR SYSTEM (Systèmes de Culture Tropicaux et Méditerranéens), Montpellier, France [email protected] Antonella Gentile UNIMI – University of Milan, Department of Crop Science, Milan, Italy [email protected] Frans E. Godeschalk LEI Wageningen UR, P.O. Box 29703, 2502 LS The Hague, The Netherlands [email protected] Sébastien Griffon CIRAD, UMR AMAP, Montpellier, F-34000, France [email protected] Andrea di Guardo CRA-CIN – Agriculture Research Council, Via di Corticella, 133 – Bologna, Italy [email protected] Ephrem Habyarimana CRA-CIN – Agriculture Research Council, Via di Corticella, 133 – Bologna, Italy [email protected] Konrad Hagedorn Humboldt University of Berlin, Department of Agricultural Economics, Philippstrasse 13, Building 12; 10099 Berlin, Germany [email protected]

xiv

Contributors

Gerard Hazeu Wageningen UR, Alterra, Centre for Geo-Information, Droevendaalsesteeg 3, 6708 PB Wageningen, The Netherlands [email protected] Thomas Heckelei University of Bonn, Institute for Food and Resource Economics, Chair for Economic and Agricultural Policy, Nussallee 21, D-53115 Bonn, Germany [email protected] Huib Hengsdijk Plant Research International, Wageningen UR, Postbus 16, 6700 AA Wageningen, The Netherlands [email protected] David Huber AntOptima, Via Fusoni 4, 6900 Lugano, Switzerland [email protected] Sander Janssen Plant Production Systems group, Wageningen University, P.O. Box 430, 6700 AK Wageningen, The Netherlands Pressent Address: Center for Geo-Information, Alterra Wageningen UR, P.O. Box 47, 6700 AA Wageningen, The Netherlands [email protected] Benny Jonsson Lund University, Centre for Sustainability Studies, P.O. Box 170, SE-221 00 Lund, Sweden [email protected] Claire Jouany INRA, UMR 1248 AGIR, BP 52627, F-31326 Castanet Tolosan, France [email protected] Argyris Kanellopoulos Plant Production Systems Group, Wageningen University, P.O. Box 430, 6700 AK, Wageningen, The Netherlands; Wageningen University, Business Economics group, P.O. Box 8130, 6700 EW Wageningen, The Netherlands [email protected] Kamel Kansou INRA – UMR SYSTEM (Systèmes de Culture Tropicaux et Méditerranéens) Montpellier, France [email protected]

Contributors

xv

Rob Knapen Alterra Wageningen UR, Center for Geo-Information. P.O. Box 47, 6700 AA Wageningen, The Netherlands [email protected] Marijke Kuiper LEI Wageningen UR, P.O. Box 29703, 2502 LS The Hague, The Netherlands [email protected] Giovanni Lanza Filippi UNIMI – University of Milan, Department of Crop Science, Milan, Italy [email protected] Peter A. Leffelaar Plant Production Systems Group, Wageningen University, P.O. Box 430, 6700 AK Wageningen, The Netherlands [email protected] Hongtao Li IDSIA – Istituto Dalle Molle di Studi sull’Intelligenza Artificiale, Galleria 2, CH-6928 Manno, Switzerland [email protected] Kamel Louhichi INRA-AgroParisTech, UMR économie publique, Avenue Lucien Brétignières, 78850 Thiverval Grignon, France [email protected] Catherine Macombe Cemagref, 24 avenue des Landais, BP 50085, 63172 Aubière Cedex, France [email protected] Luisa Manici CRA-CIN – Agriculture Research Council, Via di Corticella, 133 – Bologna, Italy [email protected] Guillaume Martin INRA, UMR 1248 AGIR, BP 52627 , F-31326 Castanet Tolosan, France [email protected] Pierre Martin LIRMM – Montpellier Laboratory of Computer Science, Robotics, and Microelectronics, University of Montpellier II, Montpellier, France [email protected] Marc Metzger School of GeoSciences, Centre for Environmental Change and Sustainability, The University of Edinburgh, Edinburgh, United Kingdom [email protected]

xvi

Contributors

Eelco Meuter Plant Production Systems Group, Wageningen University, P.O. Box 430, 6700 AK Wageningen, The Netherlands [email protected] Nora Mugueta CRA-CIN – Agriculture Research Council, Via di Corticella, 133 – Bologna, Italy [email protected] Rachmat Mulia ICRAF SEA – World Agroforestry Centre, Jalan CIFOR, Situ Gede, Sindang Barang, Jawa Barat, Indonesia [email protected] Ståle Navrud Department of Economics and Resource Management, Norwegian University of Life Sciences (UMB), P.O. Box 5033, NO-1432 Aas, Norway [email protected] Amélie Nespoulous CNRS, UMR CEFE, Montpellier, F-34000 France [email protected] Roelof Oomen Plant Production Systems group, Wageningen University, P.O. Box 430, 6700 AK Wageningen, The Netherlands Pressent Address: Institute of Crop Science and Resource Conservation (INRES) University of Bonn Katzenburgweg 5, D-53115, Bonn, Germany [email protected] Ignacio Pérez Domínguez Public Issues Division, Agricultural Economics Research Institute (LEI), Alexanderveld 5, 2585 DB, The Hague [email protected] Eric Perret Cemagref, UMR 1273 Métafort, Campus universitaire des Cézeaux, 24, avenue des Landais, BP 50085, 63172 Aubiere Cedex, France [email protected] Neil A. Powe School of Architecture Planning and Landscape, Claremont Tower, Newcastle University, Newcastle-upon-Tyne, NE1 7RU, United Kingdom [email protected] Andrea E. Rizzoli IDSIA – Istituto Dalle Molle di Studi sull’Intelligenza Artificiale, Galleria 2, CH-6928 Manno, Switzerland [email protected]

Contributors

xvii

Alexandra Rosenmund JRC – Joint Research Centre, Institute for the Protection and Security of the Citizen, Agriculture Unit, Agri4cast Action, Ispra, Italy [email protected] Vittorio Rossi UNICATT – Catholic University of Piacenza, Institute for Entomology and Plant Pathology, Piacenza, Italy [email protected] Lorenzo Ruinelli AntOptima, Via Fusoni 4, 6900 Lugano, Switzerland [email protected] Graham Russell UNIEDI – School of GeoSciences, Centre for Environmental Change and Sustainability, The University of Edinburgh, Edinburgh, United Kingdom [email protected] Francesca Salinari CRA-CIN – Agriculture Research Council, Via di Corticella, 133 – Bologna, Italy [email protected] Christian Schleyer Humboldt University of Berlin, Department of Agricultural Economics, Division of Resource Economics, Philippstrasse 13, Building 12; 10099 Berlin, Germany [email protected] Ariel Serrano CRA-CIN – Agriculture Research Council, Via di Corticella, 133 – Bologna, Italy [email protected] Andrea Sorce CRA-CIN – Agriculture Research Council, Via di Corticella, 133 – Bologna, Italy [email protected] Lee Stapleton Centre for Rural Economy, School of Agriculture, Food and Rural Development, Newcastle University, Newcastle-upon-Tyne, NE1 7RU, United Kingdom [email protected] Grete Stokstad Norwegian Forest and Landscape Institute, P.O. Box 115, NO-1431 Aas, Norway [email protected] Mats Svensson Lund University, Centre for Sustainability Studies, P.O. Box 170, SE-221 00 Lund, Sweden [email protected]

xviii

Contributors

Marie Taverne Cemagref, Dynamics and Functions of Rural Areas, 24 avenue des Landais, BP 50085, 63172 Aubière Cedex, France [email protected] Ida J. Terluin LEI Wageningen UR, P.O. Box 29703, 2502 LS The Hague, The Netherlands [email protected] Jean-Pierre Theau INRA, UMR 1248 AGIR, BP 52627, F-31326 Castanet Tolosan, France [email protected] Insa Theesfeld Leibniz Institute of Agricultural Development in Central and Eastern Europe (IAMO), Theodor-Lieser-Straße 2, 06120 Halle (Saale), Germany [email protected] Olivier Thérond INRA, UMR 1248 AGIR, BP 52627, F-31326 Castanet Tolosan, France [email protected] Marco Trevisan UNICATT – Catholic University of Piacenza, Institute for Agricultural and Environmental Chemistry, Piacenza, Italy [email protected] Patrizia Trevisiol UNIMI – University of Milan, Department of Crop Science, Milan, Italy [email protected] Nadine Turpin Cemagref, UMR 1273 Métafort, Campus Universitaire des Cézeaux, 24, avenue des Landais, BP 50085, 63172 Aubiere Cedex, France [email protected] C. Martijn van der Heide LEI Wageningen UR, P.O. Box 29703, 2502 LS The Hague, The Netherlands [email protected] Kees van Diepen Wageningen UR, Alterra, Centre for Geo-Information, Droevendaalsesteeg 3, 6708 PB Wageningen, The Netherlands [email protected] Frits K. van Evert Plant Research International, Wageningen UR, P.O. Box 616, 6700 AP Wageningen, The Netherlands [email protected]

Contributors

Martin van Ittersum Wageningen University, Plant Production Systems Group, P.O. Box 430, 6700 AK Wageningen, The Netherlands [email protected] Meine van Noordwijk INRA – UMR SYSTEM (Systèmes de Culture Tropicaux et Méditerranéens) Montpellier, France David Verhoog LEI Wageningen UR, P.O. Box 29703, 2502 LS The Hague, The Netherlands [email protected] Ferdinando Villa Gund Institute for Ecological Economics, University of Vermont, 617 Main Street, Burlington, Vermont VT 05405, USA [email protected] Grégoire Vincent IRD – French Agricultural Research Centre for International Development, Performance of Tropical Production and Processing Systems Department, Montpellier, France Vaclav Voltr Institute of Agricultural Economics and Information, Slezska 7 Prague 2, 120 56, Czech Republic [email protected] Daniel Wallach INRA, UMR 1248 AGIR, BP 52627, F-31326 Castanet Tolosan, France [email protected] Patrik Wallman Lund University, Centre for Sustainability Studies, P.O. Box 170, SE-221 00 Lund, Sweden [email protected] Jacques Wery INRA – UMR SYSTEM (Systèmes de Culture Tropicaux et Méditerranéens) Montpellier, France [email protected] Jan-Erik Wien Alterra Wageningen UR, Center for Geo-Information, Group Systems. P.O. Box 47, 6700 AA Wageningen, The Netherlands [email protected]

xix

xx

Contributors

Peter Zander Leibniz Centre for Agricultural Landscape Research (ZALF), Institute of Socio-Economics, Eberswalder Strasse 84, 15374 Müncheberg, Germany [email protected] Arezki Zerourou INRA, UMR 1248 AGIR, BP 52627 , F-31326 Castanet Tolosan, France [email protected]

Acronyms and Abbreviations

AAAT ACG ACP countries AEnZ AMF APES API ATS BRME C CA CAP CAPRI model CBA CEH CGE model CIA CLC CM CMO COPI CropMP CRUD CVM DBC DBMS DEM DG DNDC model DOT.force DSM DSS DTM

Average Annual Accumulated Temperature AgroManagement Configuration Generator African, Caribbean and Pacific countries Agri-Environmental Zonation Action Message Format Agricultural Production and Externalities Simulator Application Programming Interface Austrian Shilling Biogeographical Regions Map of Europe Carbon Current activity Common Agricultural Policy Common Agricultural Policy Regional Impact model Cost-Benefit Analysis Centre for Ecology and Hydrology Computable General Equilibrium model Crucial institutional aspect Corine Land Cover Choice modelling Common Market Organization Cost of Policy Inaction Crop Model Library Create, retrieve, update, delete Contingent valuation method Design-by-Contract DataBase Management System Digital elevation model Directorate-General DeNitrification-DeComposition model Data and Ontology Task force Digital surface model Decision Support System Digital terrain model xxi

xxii

DTO EAA EC ECU EDIM EEA E-FAST EFORWOOD ELPEN EnS EnZ ESA ESB ESDB ESU EU FADN FAO FAST FSS FSSIM FSSIM-AM FSSIM-MP GAEZ GAMS GAP GDD GDP GFP GHCN GIS GUI GVA HPM IA IAM IIASA IM GUI IO analysis IPCC ITE JRC LFA LHS LP

Acronyms and Abbreviations

Data Transfer Object Economic Accounts for Agriculture European Commission European Currency Unit European Dairy Industry Model European Environment Agency Extended FAST Sustainability Impact Assessment of the Forestry-Wood Chain European Livestock Policy Evaluation Network Environmental Stratification Environmental Zone Environmentally Sensitive Area European Soil Bureau European Soil Database European Size Unit European Union Farm Accountancy Data Network Food and Agriculture Organization of the United Nations Fourier Amplitude Sensitivity Test Farm Structure Survey Farm System Simulator Agricultural management part of FSSIM Mathematical programming part of FSSIM Global Agro-Ecological Zonation General Algebraic Modeling System Good Agricultural Practice Graphic Data Display Gross domestic product Good Farming Practice Global Historical Climatology Network Geographic information system Graphical User Interface Gross value added Hedonic pricing method Integrated Assessment Integrated Assessment and Modelling International Institute for Applied Systems Analysis Integrative modeller GUI Input-Output analysis Intergovernmental Panel on Climate Change Institute of Terrestrial Ecology Joint Research Centre Less Favoured Area Latin Hybercube Sampling Linear programming

Acronyms and Abbreviations

MARS MCE MFA MPE MUSLE N NLP model NOAA NUTS OCTOP OECD OOP OPEN MI ORM PAD PAR PCA PE PE GUI PEG PICA PLA PMP PTF PTG PTRDB QGIS RDF REST RIA RUE RUSLE SA SEAMCAP SEAMLESS

xxiii

Monitoring of Agriculture with Remote Sensing Model Component Explorer Multifunctional agriculture Model Parameter Editor Modified Universal Soil Loss Equation Nitrogen Non-linear programming model National Oceanic and Atmospheric Administration Nomenclature des Unités Territoriales Statistiques Organic carbon topsoil Organisation for Economic Co-operation and Development Object Oriented Programming Open Modelling Interface Object Relational Mapping Percent Absolute Deviation Photosynthetic Active Radiation Principal Component Analysis Parameter Estimator Policy expert GUI Production Enterprise Generator Procedure for Institutional Compatibility Assessment Protected Landscape Area Positive Mathematical Programming PedoTransfer Function Production Technique Generator Pedo Transfer Rules Database Quantum GIS Resource Description Framework Representational State Transfer Rich Internet Application Radiation Use Efficiency Revised Universal Soil Loss Equation Sensitivity Analysis SEAMLESS version of CAPRI System for Environmental and Agricultural Modelling; Linking European Science and Society SEAMLESS-IF SEAMLESS-Integrated Framework SENSOR Sustainability Impact Assessment Tools for Environmental, Social and Economic Effects of Multifunctional Land Use in Europe SGDBE Soil Geographical Database of Eurasia SINFO Soil Information SLE SEAMLESS Landscape Explorer SOE Simulation Output Evaluator SOM Soil Organic Matter SRTM Shuttle Radar Topographic Mission

xxiv

STU TCG TCM TRIS UCD UF UNCSD USGS WEPP WTO WTP

Acronyms and Abbreviations

Soil Typological Unit Technical Coefficient Generator Travel Cost Method Temperature in paddy-RIce Simulation User Centred Design User Forum Commission on Sustainable Development of the United Nations United States Geological Survey Water Erosion Prediction Project World Trade Organisation Willingness to pay

Chapter 1

Introduction Martin van Ittersum and Floor Brouwer

Context of Integrated Assessment: Policy and Research Agriculture and rural areas face rapid changes in response to agreements to liberalize international trade, the introduction of novel agro-technologies, and climate change. Food production also faces new perspectives as a consequence of competition between food, feed and fuel. Efficient agricultural and environmental policies are needed to support a sustainable development of agriculture in Europe and elsewhere. Increasingly proposed policies go through an assessment procedure before decision making. The European Commission, for instance, has introduced mandatory Impact Assessment regulations since 2003, that aim to reveal strengths and weaknesses of policy proposals. The research community aims at developing relevant tools that can provide better information for performing such impact assessments. Integrated assessment and modelling has been proposed as a key approach to enhance management of complex systems and provide objective information on policy options for the decision makers. Integrated assessment and modelling (IAM) combines the assessment of biophysical, economic, social aspects of a system using computerized tools and aims at involving stakeholders in the assessment. By using models relatively cheap experimentation and quantification of different policy alternatives is possible. Results from IAM complement other sources of information in the participatory Impact Assessment process, which, for instance in the European Union, may take ~2 years.

M. van Ittersum (*) Plant Production Systems Group, Wageningen University, P.O. Box 430, 6700 AK, Wageningen, The Netherlands e-mail: [email protected] F. Brouwer LEI Wageningen UR, P.O. Box 29703, 2502 LS, The Hague, The Netherlands e-mail: [email protected]

F.M. Brouwer and M. van Ittersum (eds.), Environmental and Agricultural Modelling: Integrated Approaches for Policy Impact Assessment, DOI 10.1007/978-90-481-3619-3_1, © Springer Science+Business Media B.V. 2010

1

2

M. van Ittersum and F. Brouwer

The integrated research project SEAMLESS (System for Environmental and Agricultural Modelling; Linking European Science and Society) was funded by the European Union’s Sixth Framework Programme for Research Technological Development and Demonstration. It was one of several projects aimed at developing research methods and tools to support Impact Assessment. Thirty research institutions and over 150 scientists from Europe, United States and Mali participated in the project. The SEAMLESS Integrated Framework (SEAMLESS-IF) was developed to enable a largely quantitative analysis and experimentation of the impacts of future polices in the domain of agriculture and environment (Van Ittersum et al. 2008). At the same time SEAMLESS-IF is accounting for technical innovations and interactions of the policies with other major trends such as population growth, economic growth and increased demand for bio-fuel crops. Analyses with SEAMLESS-IF can be performed at multiple scales and with varying time horizons, while focusing on the most important issues emerging at each scale. The linked models range from a bio-physical field model to a farm model and to an agricultural sector model for the EU. In addition, the effectiveness of a policy in its institutional context is assessed by applying more qualitative procedures. The interlinked pan-European database provides the relevant data needed at different scales. Three distinct phases are followed in the integrated assessment procedure adopted in SEAMLESS-IF, i.e. a pre-modelling, modelling and post-modelling phase (Van Ittersum et al. 2008; Ewert et al. 2009). In the pre-modelling phase the policy question is defined and translated into a baseline and policy scenarios to be assessed through a set of indicators. In the modelling phase a set of models linked into a model chain compute the variables that can be used in the post-modelling phase to assess the selected indicators. In the post-modelling phase also institutional compatibility of proposed policy measure can be assessed. The pre- and post-modelling phase require strong interactions with users and stakeholders, while the modelling phase is run mostly by integrative modellers. The modular set-up of SEAMLESS-IF, using stand-alone knowledge components, makes it flexible such that a broad range of questions can be addressed and makes it relatively easy to extend the framework with new or alternative components. The use of individual components has also advantages in terms of maintenance, transparency and documentation. The components are linked and integrated through an advanced software infrastructure. The stand-alone components used in SEAMLESS-IF are models, a database and indicators. The models include existing and newly developed models simulating crop growth (APES), farm behaviour (FSSIM) and agricultural markets (CAPRI). The project has also developed or used other models and tools, which have not been integrated so far. The models simulate different aspects (indicators) of the system at different levels of organization and scale. This book volume aims at presenting the different components, data and data processing, indicators and software architecture as developed in the SEAMLESS project, as an example of advanced integrated assessment methods for agricultural systems. SEAMLESS-IF and several components are maintained, disseminated and further developed under the coordination of a SEAMLESS Association (www.seamlessassociation.org).

1 Introduction

3

Model Components and Database SEAMLESS-IF integrates relationships and processes across disciplines and scales which are conceptualized following the paradigm of hierarchy theory (Ewert et al. 2009). The relationships and processes at different levels of organization are modelled as separate components. Figure 1.1 presents an overview of the various model components for different systems levels and disciplinary domains. These components include a modular, bio-physical simulation model calculating agricultural production and externalities at field level (APES); a bio-economic farm model quantifying the integrated agricultural, environmental and socio-economic aspects of farming systems (FSSIM); and an agricultural sector model (CAPRI) providing information on supply-demand relationships.

Globe

GTAP

Earth System

Country/ Continent

CAPRI

LABOUR

Structural change Region EXPAMOD Landscape

Landscape Evaluation

SLE

Farm

FSSIM-AM

Field

APES

Biophysical

PICA

FSSIM-MP

Bio-Economic

Social/ Institutional

Fig. 1.1 Schematic representation of model components developed or used in SEAMLESS for the different systems levels (from field to global economy) and disciplinary domains (biophysical, bio-economic and social/institutional): APES – cropping system model; FSSIM – bio-economic farm model; SLE – Landscape visualisation tool; EXPAMOD – econometric model to link farm and market analysis; Structural change – model to assess dynamics in farm structure; CAPRI – agricultural sector model; GTAP – computable general equilibrium model; Landscape Evaluation: procedure to assess social value of landscape; PICA: procedure to assess institutional compatibility; Labour: model to assess agricultural employment. Lines stand for scaling procedures, feedback mechanisms or links (Figure also used in Van Ittersum 2009)

4

M. van Ittersum and F. Brouwer

Various scaling methods have been used to link information from one level to another, or to simulate the feedbacks between levels of organisation and processes. For instance, EXPAMOD is used for upscaling the outputs from FSSIM to the European scale, in the form of price-supply relationships (Pérez Domínguez et al. 2009). EXPAMOD uses an econometric approach to estimate a meta-supply response function, that depends on price variations, farm characteristics, and corresponding soil and climate conditions. The marginal effects of prices are extrapolated to those farm types and regions not covered by FSSIM models. Finally, price-supply elasticities are calculated and aggregated to match the product categories distinguished in CAPRI. APES, FSSIM, CAPRI and EXPAMOD have been integrated in SEAMLESS-IF. The other components (Fig. 1.1) are available as standalone, but have not been integrated in SEAMLESS-IF so far. The framework uses an integrated European database containing data on soils, weather, farming systems, agro-management, prices, trade and policies, as well as a library containing indicators for economic, environmental, social aspects organised in an indicator framework (Janssen et al. 2009). Although the database in SEAMLESS was primarily developed to feed the core models and store their results, it also has stand-alone value. The data in this database include farm data from the EU wide dataset Farm Accountancy Data Network (FADN) organized in different farm types per region (NUTS1/2). Many variables are available for each farm type and region referring to their economic, production, size, labour and structural characteristics. For a selection of regions farm activity data for the main farm types provide a detailed understanding of the farming practices and main environmental characteristics. European wide environmental data on climate, soil and topography at different spatial levels from grid to region characterise the biophysical environment; an agri-environmental zonation is used to capture variation in farming systems and biophysical conditions across the EU.

Software Architecture and Ontologies The linkage of models designed for different scales and from biophysical and socio-economic domains requires appropriate software architecture, and a design and technical implementation of models that allows this. The software backbone of SEAMLESS-IF, SeamFrame, serves that purpose (Chapter 9, this volume). SeamFrame has been developed to facilitate re-use and linkage of models for integrative purposes. Its core runs on a server and provides the services that can be used by the several SEAMLESS client components and applications. SeamFrame is composed of a set of software tools and components such as the modelling environment, project manager, processing environment and the domain manager. The SeamFrame server interacts with the SEAMLESS database and knowledgebase. The SEAMLESS ontology plays a central role in SEAMLESS-IF for harmonising and linking concepts from a wide range of knowledge domains: from models, to indicators, to source data formats, etc. The use of ontologies to semantically annotate

1 Introduction

5

the component models allows, among other things, the match between sources in terms of linking the proper output variables of a component to the input variables of a second component to be checked for consistency. Ontologies have been very instrumental in realizing interdisciplinary collaboration within the project. Modellers can develop their models using different modelling environments, such as MODCOM for the biophysical models and GAMS for farm economic and market models. The Open Modelling Interface (OpenMI; www.openmi.org) provides a standardised interface to define, describe and transfer data between software components running simultaneously or sequentially.

Key Objectives and Organisation of the Book The volume reviews and presents our current understanding of integrated concepts and working tools to assess and compute, ex-ante, alternative agricultural and environmental policy options, allowing: 1. Analysis at the full range of scales (farm to European Union and global), whilst focusing on the most important issues emerging at each scale 2. Analysis of the environmental, economic and social contributions of agricultural systems towards sustainable rural development and rural viability 3. Analysis of a broad range of issues and options, such as environmental policies, effects of an enlarging EU, international competition and climate change This volume has a strong ‘lessons to be learnt’ emphasis, to facilitate and promote the use and further development of integrated assessment tools to support policies promoting agricultural development in support of sustainable development. Part I explores recent advancements on indicators and concepts in practice. Chapter 2 guides the development of multifunctionality indicators that are based on the concept of joint production. Nadine Turpin and her co-authors offer empirical evidence that multifunctionality in agriculture is far from being negligible. The work builds on a framework of analysis based on the concept of jointness of production. In addition to the economic, social and environmental domains of sustainable development, the concept of institutional analysis is increasingly considered to be a critical component for ex-ante impact assessment. Chapter 3 provides a procedure for institutional compatibility assessment. Insa Theesfeld and her co-authors present four steps that are critical to assess the compatibility of policy options with the institutional context. The authors elaborate the steps and illustrate the procedure with the EU Nitrate Directive. Advancements in modelling, data and software are presented in Part II. To start with, Chapter 4 offers a modular simulation system that is aimed at estimating the biophysical behaviour of agricultural production systems in response to the interaction of weather, soils and agro-technical management. Marcello Donatelli and his co-authors present a component-based framework called the agricultural production and externalities

6

M. van Ittersum and F. Brouwer

simulator (APES). APES is a framework with components that offer simulation options for different processes of relevance to agricultural production systems. A bio-economic modelling framework is presented in Chapter 5. Kamel Louhichi and co-authors present a generic farm system simulator that is sufficiently flexible and generic to be applicable for relevant farming systems across Europe. The system – FSSIM – is tested for a set of farms representing the arable farming systems in parts of the Netherlands and France, allowing to analyse current conditions and anticipate the impact of alternative policy scenarios. In addition to advancements made in biophysical and bio-economic modelling, SEAMLESS also made advancements in other modelling approaches. Chapter 6, for example, introduces achievements made on a landscape visualisation component to allow for exploration of landscape changes. Sébastien Griffon and his co-authors present a tool a allowing the visualisation of changes in the spatial configuration of the landscape, as for instance derived from a bio-economic farm model. The use of typologies in impact assessment approaches is presented in Chapter 7 and Chapter 8. Chapter 7 makes use of a biophysical typology in agri-environmental modelling. Gerard Hazeu and co-authors provide an environmental stratification of Europe building largely on climate and altitude characteristics. Environmental zones are combined with soil data and data that represent major obstacles for farming to derive an agri-environmental zonation. In addition, Chapter 8 presents a regional typo logy based on socio-economic data. Ida Terluin and her co-authors capture regional characteristics on the performance of farming based on economic development. The authors argue in favour of including regional characteristics into integrated assessments of farm performance and agricultural policy. Chapter 9 presents how conceptual and technical integration has been achieved. Jan Erik Wien and his co-authors present the major functionalities of SeamFrame, the software architecture and implementation for the integrated assessment framework. Ontologies and the Open Modelling Interface are cornerstones of the architecture. Part III deals with the dissemination and use of integrated assessment modelling. Experiences to test and evaluate integrated assessment tools for policy support are presented in Chapter 10. Jacques-Eric Bergez and his co-authors identify the steps taken from the evaluation of test cases towards the involvement of final end-users in the evaluation process. The authors identify lessons learned from the evaluation process and they conclude that transparency in the development and evaluation process is key to the successful development and use of integrated assessment tools for policy support. Chapter 11 compares SEAMLESS-IF and CAPRI. Wolfgang Britz and co-authors observe differences resulting from SEAMLESS-IF having a stronger focus on field and farm level components stressing bio-economic and technological innovations, whereas CAPRI adopts a more market-oriented perspective with full coverage of EU policies. Major parts of CAPRI are treated as one component in SEAMLESS-IF. Chapter 12 describes the interaction between science and policy in impact assessment procedures. Ann-Katrin Bäcklund and her co-authors share their experience obtained from such interactions and conclude that scientists have an important role in the impact assessment process. The critical contribution of science is to demonstrate

1 Introduction

7

when beliefs by policy makers are supported by evidence and when conventional wisdom is simply wrong. To arrive at meaningful tools that will actually be adopted by the policy domain, policy makers need to be engaged in the time consuming process of tool development. The chapter shares the results of the interactive development process and the impacts it had on the development of the research tools. Finally, Chapter 13 explores economic principles of monetary valuation in evaluation studies. Martijn van der Heide and his co-authors review techniques and concepts of monetary valuation and the contribution of monetary valuation in agricultural policy impact assessment. They conclude that monetary valuation can play a supportive role to the quantitative models in SEAMLESS-IF.

References Ewert, F., van Ittersum, M.K., Bezlepkina, I., Therond, O., Andersen, E., Belhouchette, H., Bockstaller, C., Brouwer, F., Heckelei, T., Janssen, S., Knapen, R., Kuiper, M., Louhichi, K., Alkan Olsson, J., Turpin, N., Wery, J., Wien, J.-E., & Wolf, J. (2009). A methodology for enhanced flexibility of integrated assessment in agriculture. Environmental Science and Policy, 12(5), 546–561. Janssen, S., Andersen, E., Athanasiadis, I., & van Ittersum, M.K. (2009). A database for integrated assessment of European agricultural systems. Environmental Science and Policy, 12(5), 573–587. Pérez Domínguez, I., Bezlepkina, I., Heckelei, T., Romstad, E., Oude Lansink, A., & Kanellopoulos, A. (2009). Capturing market impacts of farm level policies: a statistical extrapolation approach using biophysical characteristics and farm resources. Environmental Science and Policy, 12 (5), 588–600. Van Ittersum, M.K. (2009). Integration across disciplines: The lessons learnt from the integrated project SEAMLESS. Aspects of Applied Biology, 93, 55–60. Van Ittersum, M. K., Ewert, F., Heckelei, T., Wery, J., Alkan Olsson, J. A., Andersen, E., et al. (2008). Integrated assessment of agricultural systems – A component-based framework for the European Union (SEAMLESS). Agricultural Systems, 96, 150–165.

Part I

Indicators and Concepts in Practice

Chapter 2

Assessment of Multifunctionality and Jointness of Production Nadine Turpin, Lee Stapleton, Eric Perret, C. Martijn van der Heide, Guy Garrod, Floor Brouwer, Vaclav Voltr, and Dominique Cairol

Introduction The concept of multifunctional agriculture (MFA) arose after implementation of commitments under the Uruguay Round trade negotiations that started in 1995. Countries that advocated the concept of multifunctional agriculture were Switzerland, Norway, Japan, Korea as well as the European Union (EU). Dissenters included a coalition of countries known as the Cairns Group (a 17 strong alliance composed of Australia, New Zealand, Canada and 14 less developed countries). Usually a distinction is made between the multifunctionality of agriculture (MFA), which has been extensively propounded in Europe (both on political and scientific grounds), and multifunctionality of the rural space. This chapter focuses on the former and addresses multifunctionality of agriculture both at the farm and regional levels.

N. Turpin (*) and E. Perret UMR Métafort, Cemagref, BP 50085, F-63172, Aubiere, France e-mails: [email protected]; [email protected] L. Stapleton and G. Garrod Centre for Rural Economy, School of Agriculture, Food and Rural Development, Newcastle University, Newcastle-upon-Tyne, NE1 7RU, UK e-mails: [email protected]; [email protected] C.M. van der Heide and F. Brouwer LEI Wageningen UR, P.O. Box 29703, 2502 LS The Hague, The Netherlands e-mails: [email protected]; [email protected] V. Voltr Institute of Agricultural Economics and Information, Slezska 7 Prague 2, 120 56, Czech Republic e-mail: [email protected] D. Cairol Cemagref, Direction Générale, GT – DTDG, Parc de Tourvoie, BP 44, 92163, ANTONY Cedex, France e-mail: [email protected] F.M. Brouwer and M. van Ittersum (eds.), Environmental and Agricultural Modelling: Integrated Approaches for Policy Impact Assessment, DOI 10.1007/978-90-481-3619-3_2, © Springer Science+Business Media B.V. 2010

11

12

N. Turpin et al.

Considering multifunctionality as a paradigm for rural development, which emphasises the increasing importance of multifunctional enterprises that link the rural with the urban (OECD 2006), goes beyond the objective of the SEAMLESS project. There are many definitions of MFA in circulation. The original definition, which was espoused by the advocate countries mentioned above, refers to the joint production of commodity and non-commodity outputs (Sakuyama 2005): agriculture, beyond the production of food and fibre (i.e. commodities), provides important social, environmental and economic functions to society. But these functions manifest themselves in products and services that are usually not marketable (i.e. non-commodities) because most often, they exhibit public good or quasi-public good characteristics whereby no individual or organisation can easily control the use of or access to these products and services (Stapleton et al. 2004). However, as pointed out by Vatn (2002), it does not mean that the notion of multifunctionality only includes a mix of private goods and various public goods. Strictly speaking, it also comprises public bads: effects that may have negative consequences for welfare. Furthermore, and most importantly, the production of commodity and non-commodity outputs are mutually dependent to some degree so that the provision of the latter cannot be decoupled or considered in isolation from the former without risking sub-optimal provision. Most of the multifunctionality literature is theoretical rather than empirical; refer for example to the articles provided in Brouwer (2004) as well as Van Huylenbroeck and Durand (2003). The lack of empirical evidence that supports the theoretical ideas and insights has already been clearly stated by the OECD (2001a, 2003). Nevertheless, there are, of course, exceptions. Quantitative work on multifunctionality which has focused on the issue of joint production of commodity and non-commodity outputs can be found in Belletti et al. (2003), Bontems et al. (2005a, b), OECD (2001b) and Wiggering et al. (2006). Quantitative work on multifunctionality has also focused, for example, on how tourists value the contribution to landscape made by farmers (Vanslembrouck and van Huylenbroeck 2003) and analysis of Dutch farmers’ motivation for multifunctionality (Jongeneel and Slangen 2004). Moreover, Van der Heide et al. (2007) use a land use change model – with field survey data as an input – to simulate multifunctional land use by applying it to a case-study area in the Netherlands. But most of the time, the quantitative work focuses on the assessment of several “functions” from farming activities (see Waarts 2005 for recent examples). In this chapter, we propose an alternative approach: instead of assessing environmental or social “functions” from the farming activities, we assume that agriculture provides non-commodity outputs and we aim at designing indicators that measure the degree of multifunctionality involved in the co-production of commodity and non-commodity outputs by farms. It is important that decision makers can measure the sustainable development implications of a given policy intervention in terms of how this affects the multifunctional attributes of a given area. Policy formulation that aims at supplying commodity and non-commodity outputs separately will lead to higher implementation costs than when the policy considers multifunctionality and encourages farmers to supply these outputs jointly (Brunstad et al. 2005; OECD 2001a). More specifically, measuring the degree of jointness will provide insights

2 Assessment of Multifunctionality and Jointness of Production

13

into the possible ramifications of any potential decrease in the use of public funds to support agriculture and its associated amenities in the context of the Common Agricultural Policy (CAP). By placing multifunctionality in this prescriptive or normative context – because it prescribes multifunctionality as an alternative strategy to land use segregation – the focus is on the demand for and supply of the multiple functions of agriculture at both the individual farm and regional level. Multifunctionality has strong links to the concept of sustainability. For example, different functions of the rural landscape – in the sense of different types of land use and related land covers – can be of mutual benefit, for example agro-biodiversity, and generate economic sustainability among rural entrepreneurs and promote and support ecological sustainability in the local area. Multifunctionality is then an important element in the paradigm of sustainability. On the other hand, the various functions can also be conflicting, such as in the case of intensive agriculture versus water storage. We return to this point. The relationship between multifunctionality and sustainability is not always clear-cut or without conflict. However, SEAMLESS attempts to assess contributions of agriculture to sustainable development of the rural area and therefore is inherently driven by concerns of multifunctionality. Therein lies the rationale and motivation for the work presented below whereby we present a theoretical framework for joint supply based on the assumption that the degree of jointness has consequences both in terms of commodity production costs and non-commodity production. Based on this framework, we next derive a design for indicators of multifunctionality based on three sequential stages with assessment of this approach undertaken on data from Auvergne, France: a Nomenclature of Territorial Units for Statistics (NUTS) two region according to Eurostat before, finally, drawing some conclusions from this analysis.

Joint Supply, Theoretical Framework As noted above, we are focusing our analysis of multifunctionality in terms of the joint supply of commodity and non-commodity outputs. Methodologically this is perhaps the most challenging definition to analyse and operationalise in terms of the formalised economics it requires. However couching multifunctionality in different terms such as the independent coexistence of commodity and non-commodity outputs would be largely self-evident and as such not require the level of analysis presented here. Etymologically, the word ‘joint’ was originally used to describe where two bones meet and move in contact with each other (from Old French joint and Latin junctus). This is a reasonable metaphor for describing how this word is now being used to emphasise the relationship between commodity and non-commodity outputs in agriculture in the sense that their production is interdependent. Indeed, the notion of jointness is not unique to agriculture and prior to its emergence in this field it had already been used to explain the existence of firms selling multiple products (Baumol et al. 1981).

14

N. Turpin et al.

The nature of commodity outputs is unambiguous in the sense that they are economic in nature with values determined by the market. Key agricultural commodities include crops, livestock, flowers and raw materials for secondary production. Non-commodities refer to environmental and social characteristics which are usually public goods; society deems non-commodity outputs as desirable but markets do not exist to ensure a convergence between supply and demand.1 Here De Groot’s (2006) classification is of interest because when there is a market, this one is related with the services and goods provided, not to the specific functions involved in the supply of these goods and services. A recurrent problem associated with multifunctional attributes is that providing non-commodities (or, non-food services) is complicated by the fact that multiple and non-unilateral links exist between the different functions fulfilled by agriculture, the ecosystem processes and components involved in fulfilling these functions and the sets of goods and services provided. To the best of our knowledge, only a very recent addition to the literature makes a clear distinction between these very different things: functions, processes and components involved one the one side can be distinguished from the goods and services provided, on the other side (see De Groot 2006). This distinction is important: usually a given component of the landscape (a vegetation root matrix for example) can provide several functions (in our case: soil retention; water regulation and filtering; nutrient regulation) and it can be involved in the supply of several non-commodity outputs and so even the simple description of the relationships between components, functions and services provided is often confused. But the classification proposed by De Groot enables simplifying such a description because each function provides only one type of good and service. Of course, De Groot’s classification was elaborated first for natural and semi-natural landscapes and requires some modification for the man-made components of landscapes. If we take the example of rural landscapes, such landscapes have an important role to play in the maintenance of regulation functions2 from De Groot (2006). One of these concerns is climate regulation; fulfilling this function involves biologically mediated processes that influence climate. Furthermore, fulfilling this function provides services, such as the maintenance of a favourable climate for human habitation, health and cultivation. The most important thing to recognise from De

1 This definition of jointness is not universal: other definitions of jointness couch non-commodity outputs solely in terms of environmental characteristics, excluding the social dimension (e.g. Nowicki 2004) which could in part reflect the fact that assessment of social non-commodity outputs is hampered by a lack of available data. 2 Regulation functions: ‘this group of functions relates to the capacity of natural and semi-natural ecosystems to regulate essential ecological processes and life support systems through bio-geochemical cycles and other biospheric processes. Regulation functions maintain a “healthy” ecosystem at different scale levels and, at the biosphere level, provide and maintain the conditions for life on Earth’ (De Groot 2006).

2 Assessment of Multifunctionality and Jointness of Production

15

Groot’s classification, from our perspective at least, is that the analysis of jointness should be performed only at coherent levels: jointness between functions fulfilled, or between processes involved, or between goods and services provided. Trying to identify relationships between the supply of a commodity and the provision of several non-market functions will lead to conceptual difficulties that will hamper analysis and risk producing results that are spurious. One important point of the analysis of multifunctionality is the connection of the various activities that take place in a landscape. In the context of this chapter, we consider that multifunctional landscapes support different activities in the same plot of land (e.g. both agriculture and tourism). This position is different from the literature of rural sociologists (Van der Ploeg and Roep 2003, e.g.) who consider that several activities can take place at landscape scales but are distinct on each plot of land (for us, this would be pluri-activity). To elaborate, a given area can be devoted to diverse but single-function land use types, like croplands, woodlands, recreational areas; in this area, several activities are possible, like farming, wood production or tourism (in the recreation areas). But the large paths that cross the productive woods are built for trucks where tourism would prefer networks of small lanes. In the farmland area, the fields are fenced, the old lanes have been ploughed and the hedges are not part of the productive system and thus poorly kept. The recreational areas look like tourists would derive amenity value from them but are not related to the cultural heritage of the area and their access is controlled. The different activities are not connected to each other: this is pluri-activity. By contrast, a multifunctional landscape may supply farm outputs along with tourism services on multifunctional farms. Of course, tourism services can also be supplied by activities other than farming, wood and genetic biodiversity in multifunctional forests and have a cultural value because of the scenery that includes fields amidst diverse forests, along with the existence of animal species that need forests margins close to open fields. Thus, we focused our analysis on multifunctionality (and not on pluri-activity) and we argue that the analysis of jointness has to be performed either between functions, or processes or goods and services produced but without crossing these different levels. As a consequence, the analysis of jointness at the farm level will consider jointness between the supply of commodity and non-commodity outputs.

Towards Indicators of Multifunctionality Determining indicators of multifunctionality (based on the jointness definition) includes three important and sequential steps: –– Step 1: identification of jointness –– Step 2: qualitative assessment of jointness –– Step 3: quantitative assessment of jointness

16

N. Turpin et al.

We need to identify the existence of jointness to assess whether the observed levels of commodity outputs are produced along with joint non-commodity outputs or according to a more classical profit function (with separable commodity and non-commodity outputs). This identification can be performed in two stages: –– Identification of multifunctionality at the appropriate scale: does the farm, or the area, fulfil several functions in the three pillars of sustainable development? The presence of multifunctionality can be identified through existing indicators along the economic, environmental and social dimensions of sustainable development. –– In the case where multifunctionality is identified, there is a need to assess whether there is a joint supply of commodity and non-commodity outputs, or whether jointness occurs between processes (e.g., the agricultural activity can mobilise biota for storage and recycling of nutrients, along with the conversion of solar energy to edible plants and provide a large variety in landscapes with potential recreational and cultural value). The qualitative assessment of jointness is necessary to determine if the degree of jointness is strong or weak and to determine whether the provision of non-commodity output increases with the commodity output. Finally, the quantitative assessment should provide the functions describing the relationship between commodity and non-commodity outputs at the farm level. This understanding leads to three main areas of investigation: –– Does the evaluated policy improve sustainability through a development of multifunctional features of agriculture in the targeted areas? In other words is the ex-ante impact of policies positive along the three dimensions of sustainable development (economy, environment, social)? –– Does the degree of multifunctionality increase? A regulator may wish to sustain the development of the area she manages through the increase of multifunctional agriculture. Thus we have to estimate the extent to which this multifunctionality increases, or to provide measurements: in places where the multifunctionality increases (or simply develops) is it possible to determine the extent of this increase (i.e. can we at least rank different policies according to their multifunctionality credentials?). –– Does the nature of multifunctionality evolve because of the evaluated policy? The non-commodity outputs of agriculture are non conventional products, but intuitively according to the concept of joint production, most of them result from specific aspects of the agricultural production process. Some of these outputs are closely tied to agricultural production and others3 compete with agricultural production for land or other resources. In this sense, any policy dealing with agricultural production is likely to modify the nature of the multifunctional features in a region.

Like the establishment or restoration of wetlands, or the creation of wildlife habitat on farmland.

3

2 Assessment of Multifunctionality and Jointness of Production

17

Step 1: Identification of Jointness De Groot’s (2006) classification of functions from natural and semi-natural landscapes distinguishes five categories of functions: regulation, habitat, production, information and carrier functions.4 What is particularly appealing in this approach is that: –– The function categories are independent from each other. The functions provided by the farms can be depicted in a coordinate system with categories of function as axes; because these categories of functions are independent, the axes are orthogonal, and thus distances can be measured between farms in this coordinate system. –– There is a clear distinction between the functions, the ecosystems and components involved in these functions, and the goods and services provided by these functions (such a clear distinction is often ignored in the literature).

Jointness at the Farm Gate At the farm level, starting from De Groot’s work to identify the existence of multifunctionality, the identification of jointness focuses on the supply of goods and services at the farm level. We relied on studies that aim at assessing functions from Farm Accountancy Data Network (FADN) data and expertise (Perret 2006). Using this expertise, identification of multifunctionality between an array of different economic, environmental and social outputs can be systematically undertaken (Table 2.1). The identification of jointness between various economic, environmental and social characteristics associated with agriculture enables a basic comparison between farms in terms of whether or not they exhibit multifunctionality characteristics. Comparison and ranking of farms based on these attributes can therefore serve as an indicator of multifunctionality. However, this binary approach does not provide information as to whether particular characteristics are simply not exhibited or are negatively affected by a particular production activity. Additionally, this system cannot be used to identify whether characteristics are met and not met at different levels of commodity production. This system of identification could be extended to acknowledge these possibilities (Table 2.2). To elaborate, a production activity yielding high quality agricultural products (commodity outputs) could be associated with positive externalities like landscape quality and negative externalities like the emission of air pollutants which reduces environmental quality and the magnitude of these externalities could differ depending on the magnitude of the commodity output. Although the essence of the multifunctionality concept concerns joint production of goods and not bads, it is worthwhile

4 De Groot (2006) proposes to ‘translate the ecological complexity into a more limited number of ecosystem functions. These functions, in turn, provide the goods and services that are valued by humans. (...) ecosystem functions are defined as “the capacity of natural processes and components to provide goods and services that satisfy human needs, directly or indirectly”’.

18

N. Turpin et al. Table 2.1 Hypothetical example of the identification of jointness at the farm scale using a binary system to denote the presence (1) or absence (0) of desirable economic, environmental and social characteristics Jointness at farm level Farm 1 Farm 2 Economy

Environment

Social

Quality of products Diversity of products Non-agricultural activitiesa Services Water conservation Soil conservation Agricultural landscape Contribution to air quality Use of renewable energy Supply of renewable energy Biodiversity Contribution to employment Contribution to rural viability Animal welfare Cultural heritage Provision of recreational areas

1 1 0 1 1 1 1 0 0 0 0 1 0 0 0 0

0 0 0 0 0 0 1 1 0 1 1 0 1 1 1 1

a “Non-agricultural activities:” are activities that take place on the farm area but are not direct farming activities. Examples of such activities are hosting tourists in on-farm bed-and-breakfast, on-farm restaurants, pedagogic activities (hosting pupils), animals (horses, dogs) hosting, etc.

Table 2.2 Hypothetical example of the identification of jointness at the farm scale denoting the presence (1), absence (0), decline (−1), presence and decline (1/−1) of desirable economic, environmental and social characteristics Jointness at farm level Farm 1 Farm 2 Economy

Environment

Social

Quality of products Diversity of products Non-agricultural activities Services Water conservation Soil conservation Agricultural landscape Contribution to air quality Use of renewable energy Supply of renewable energy Biodiversity Contribution to employment Contribution to rural viability Animal welfare Cultural heritage Provision of recreational areas

1 1 0 1 1 1 1 −1 0 0 0 1 0 1/−1 0 0

0 0 1/−1 0 −1 1/−1 1 1 0 1 1 −1 1 1 1 1

2 Assessment of Multifunctionality and Jointness of Production

19

to emphasise that the latter is also possible because, as already mentioned in the introduction of the chapter, multifunctionality comprises not only public goods, but also public bads. Such a system could be used in either an ex-post or ex-ante context i.e. based on available data or forecasting potential implications of a given policy intervention (such as liberalization of European agricultural markets). Of course, assessment here relies still only on expertise. Jointness at the Regional Level Preliminary experiments in five zones of the Rhône-Alpes region (France) emphasized that aggregating individual data concerning multifunctionality from the farm level to a larger area is far from being evident (Gillette et al. 2005a). Aggregating individual information about jointness enabled the authors to design a representation of the multifunctionality of agriculture in a larger zone (groups of municipalities or a small region). This representation was consistent with the common knowledge of the history of the different areas but when compared with local surveys of the relevant areas, it was incomplete because synergies between various farms enhanced the regional multifunctionality (Gillette et al. 2005b). At the regional level, multifunctionality of agriculture relies on two different concepts: the variety of combinations of commodity and non-commodity outputs in the farms and the synergies and antagonisms between these combinations. As such, the farms can combine the provision of commodity and non-commodity outputs in very different ways. For example, some farms can breed sheep and maintain pastures that contribute to the local cultural heritage but other farms in the same area can choose to restore traditional buildings for hosting tourists and contribute in a different manner to this local cultural heritage. Other examples involve the provision of specific landscape patterns that favour the persistence of particular animals or plants; these specific patterns may be the consequence of maintenance of some hedges by small cattle farms that need trees for their animals, along with mosaics of pastures and cereals fields by mixed farms and specific phytosanitary protection practices in large intensive farms. Moreover, identifying the various ways of combining commodity and non-commodity outputs in a population of farms is of importance when the ecological or social processes exhibit threshold effects,5 because there is a need to determine whether the farms that produce them jointly are numerous enough for the global provision of the ecological (or social service) or whether their efforts are lost because they are too few to do so. The second concept is related to the fact that the various ways of combining commodity and non-commodity outputs in a given region can exhibit synergies or be competitors. Intensive farms selling high quality products at the farm-gate along with improvement of the surrounding scenery may benefit from the preservation of hedges by neighbouring low-intensity dairy farms and at the same time compete Threshold effects on ecological discontinuities have been defined by Muradian (2001) as sudden modifications of a given system property, resulting from the soft and continuous variation of an independent variable.

5

20

N. Turpin et al.

with them for land because the latter need large areas to be profitable. Moreover, the average water quality in the area may be good only because the two types of farms are present and the intensive farms may need that the less-intensive ones keep their activity despite competition (otherwise they would be forced to lower their own pressure on water quality).

Step 2: Qualitative Assessment of Jointness Identification of jointness, discussed above, is an important first step but as an indicator of multifunctionality it is limited in terms of the information it presents to stakeholders. Qualitative assessment of jointness can extend such an analysis by providing answers to two different questions which are not addressed above: –– Is the degree of jointness strong or weak? –– What is the origin of jointness? Jointness can be due to: technical interdependencies in the production process; the existence of a non-allocable production factor i.e. when different products are obtained from a single input such as the case of wheat and straw production or of ovine meat and wool from sheep; outputs competing for an allocable and fixed input so that any increase in the production of one output reduces the quantity of the fixed input available for the production of the other product (OECD 2001a). Furthermore, a contribution to qualitative assessment of jointness may result from surveys aiming at measuring the public demand for different functions. In an interesting work on public demand for rural landscapes, Hall et al. (2004) have tried to capture information about public preferences for goods and services that are provided by agriculture and the countryside. Rather than looking at the production side, at the services and goods that a system offers, the authors decided to look at the consumption side, at the goods and services that are demanded and valued by consumers. The methods they distinguish as being mostly used to measure consumer preferences consist basically of three types of survey instruments: –– Pools and surveys conducted by conservation organisations, government departments and the EU –– More rigorous surveys trying to quantify public preferences through structured trade-off methods using willingness-to-pay approaches; and –– Deliberative survey methods as a compromise between polls and valuation methods All three methods have their drawbacks which they suggest can possibly be (partially) overcome by a combination of multicriteria analysis and choice experiments. Interviews with specific key stakeholders could also be applied as a complement to the other methods. Another approach would be based on spatial analysis of the land use pattern. The basic idea is to visualize the land cover and use of a certain territory and to associate

2 Assessment of Multifunctionality and Jointness of Production

21

each class or combination of classes with functions provided, or the related goods and services. Spatial analysis can help to create the maps of land cover and use. Expert knowledge (participative), surveys or interviews and field visits can then help to ‘fill in’ these maps with the functions that can be found. It is possible to combine or overlap several maps at different levels covering, for example: land cover, land use and landscape as well as maps revealing more socio-economic data such as population density, activities in certain sectors and fluxes of people and goods and services (networks). This approach is suggested but not tested by Brandt and Vejre (2004) and Vejre et al. (2006). Knickel et al. (2004) refer to multifunctionality schemes as a way to directly ‘map’ the interrelated functions associated with a certain territory or activity. A promising method is interactive mapping that can be useful as a complementary method to (open) interviews and field visits, or other participatory research methods – questions as to the goods and services provided are directly related to the maps with the representation of the land cover mosaic. Qualitative assessment of jointness could be particularly important to ex-ante analyses: for example, if a particular commodity and non-commodity output are competing for an allocable fixed input then as one output increases, the other necessarily declines and it is unlikely that any change in policy would alter this underlying property.

Step 3: Quantitative Assessment of Jointness The final, and most difficult, stage is quantitative assessment of jointness which involves specifying the magnitude of the coefficient(s) in each particular jointness function. Graphical examples of these different types of jointness including their most basic underlying equations (functional forms) are provided; the magnitudes of these coefficients are illustrative to reflect the fact that jointness can be strong or weak (Fig. 2.1). This is an extension of work currently available in the literature on jointness which has tended not to explore these underlying equations or a range of potential functional forms. The case of positive jointness (Fig. 2.1a) could be associated with rural development policies in the EU, specifically the introduction of the second pillar to the CAP; this theoretical relationship has been articulated by Belletti et al. (2003) for example. Where increases in commodity outputs are associated with increases in non-commodity outputs (Fig. 2.1a) this could be denoted as true multifunctionality. The policy implication of such positive jointness, ceteris paribus, is, simply, dual maximisation of both commodity and non-commodity outputs. As an alternative, Van Huylenbroeck (2003) suggests that agriculture is in most cases a necessary condition to obtain the non-commodity output, but the yield in itself is not as important. This could be termed as static jointness or static multifunctionality and it is possible to express Van Huylenbroek’s idea graphically (Fig. 2.1b). Although increases in the commodity output do not increase levels of the non-commodity output, neither do they decrease this output. Therefore, the policy implication of such static jointness, ceteris paribus, is maximisation of the commodity output.

22

N. Turpin et al.

40 35 30 25 20 15 10 5 0 0

5 10 15 20 25 30 35 40 45 50 Commodity Output (Arbitrary Units) NCO=0.3*CO NCO=1.1*CO

80 70 60 50 40 30 20 10 0 0

5

10

15

20

25

30

35

40

45

Static jointness: NCO=b IF CO>0

60 50 40 30 20 10 0

50

10 8 6 4 2 0

0

100 80 60 40 20 0

0

5 10 15 20 25 30 35 40 45 50 Commodity Output (Arbitrary Units) NCO=80+(–0.09*CO)+(0.02*CO^2) NCO=80+(–2.1*CO)+(0.05*CO^2)

f Non-Commodity Output (Arbitrary Units)

Non-Commodity Output (Arbitrary Units)

120

5 10 15 20 25 30 35 40 45 Commodity Output (Arbitrary Units)

50

NCO=0.6*CO+(–0.01*CO^2) NCO=0.8*CO+(–0.015*CO^2)

NCO=80+(–0.2*CO) NCO=80+(–0.8*CO)

Negative and positive (quadratic) jointness: NCO=b+(–a*CO)+(c*CO^2)

2 3 4 5 6 7 8 9 10 11 Commodity Output (Arbitrary Units) NCO=20 IF CO>0 NCO=50 IF CO>0 Positive and negative (quadratic) jointness: NCO=a*CO+(-c*CO^2)

12

Commodity Output (Arbitrary Units)

e

1

d

Negative jointness: NCO=b+(–a*CO)

90

Non-Commodity Output (Arbitrary Units)

45

c Non-Commodity Ouput (Arbitary Units)

b

Positive jointness: NCO=a*CO

50

Non-Commodity Output (Arbitrary Units)

Non-Commodity Output (Arbitrary Units)

a

Negative jointness breaching an ecological threshold at <60 NCO (Arbitray Units): NCO=b+(–a*CO) IF NCO >x; NCO=0 IF NCO < x 90

80 70 60 50 40 30 20 10 0 0

5 10 15 20 25 30 35 40 45 Commodity Output (Arbitrary Units)

50

NCO=80+(–0.8*CO) IFNCO>60 NCO=80+(–2.1*CO) IFNCO>60

Fig. 2.1 Hypothetical strong and weak jointness between commodity and non-commodity outputs: (a) positive; (b) static; (c) negative; (d) positive and negative; (e) negative and positive; (f ) negative breaching an ecological threshold

2 Assessment of Multifunctionality and Jointness of Production

23

Negative jointness (Fig. 2.1c) is associated with an agricultural modernisation agenda where non-commodity outputs are secondary considerations compared to agricultural production (Belletti et al. 2003); in the context of EU agriculture this could tally with the historical emphasis on supporting agricultural production through price support measures. However, in the case of negative jointness the production of a commodity output, forsaking a given amount of the non-commodity output could illustrate (a) the existence of a negative externality i.e. the non-commodity output is a non-market good which is not internalised in production decisions, (b) the noncommodity output is wholly or partially internalised by the market but its marginal value is lower than the commodity output resulting in a rational trade-off. Determining the optimal production of the commodity output versus the non-commodity output where the latter is an externality requires some assessment of the value of the noncommodity output, perhaps using a ‘stated preference’ valuation technique like surveying individuals to determine their willingness to pay for the non-commodity output. The case of positive and negative jointness (Fig. 2.1d) has been articulated graphically in OECD (2001b). Going further, we can note that if the marginal value of the commodity and non commodity output are equal then the optimal solution is the turning point of the function. Mathematically:

dNCO = a + (−2c*CO ) = 0 dCO −a CO = −2c d 2 NCO = −2c dCO2

(2.1)

In the case of negative to positive jointness (Fig. 2.1e) dual maximisation of commodity and non-commodity outputs is only a viable option if the reductions in non-commodity output at low levels of commodity output, before the turning point is reached and both outputs start to rise together, is not associated with the breaching of any ecological thresholds i.e. we need a priori knowledge that environmental degradation and pollution are occurring in a linear fashion. In reality, many important environmental non-commodity outputs have critical ranges and thresholds which, if breached, could result in drastic and uncontrollable loss of that noncommodity output and thus, potentially, the commodity depending on the interaction between the two outputs (Fig. 2.1f). An ex-post empirical investigation of the magnitudes of non-commodity outputs at different levels of commodity output could elucidate a number of relationships; from an ex-ante perspective it is important to understand whether a particular policy will: –– Shift the commodity/non-commodity relationship from one trajectory to another (alter the underlying functional form of the relationship) e.g. from negative jointness (Fig. 2.1c) to positive jointness (Fig. 2.1a)

24

N. Turpin et al.

–– Shift the commodity/non-commodity relationship off one trajectory into a mutual decline because of the breaching of an ecological threshold (relevant to Fig. 2.1c and e) –– Make no alteration to the commodity/non-commodity trajectory but alter the position on that trajectory e.g. concomitant increases in commodity and noncommodity output in Fig. 2.1a –– Alter the strength of the association between commodity and non-commodity outputs e.g. from weak positive jointness to strong positive jointness An ex-ante analysis such as this obviously depends critically on a case-by-case ex-post empirical investigation of an array of different commodity and non-commodity output relationships i.e. it’s important to understand past relationships between commodity and non-commodity outputs in order to predict future relationships in response to policy changes. Furthermore, the origin of jointness as discussed above is particularly important to aiding an ex-ante analysis. For example, if particular commodity and non-commodity outputs compete for an allocable fixed input, an increase in one output could only be achieved by a decline of the other. In the short run, farms will adapt to policy changes by altering their position on a trajectory. In the long run, they can alter that linkage, for example by achieving a more rational use of labour through the adoption of labour saving technologies but this would require time and investment.

Testing Against Data in Europe Identification of Jointness Jointness at the Farm Gate Some drivers of farm income are drawn from the FADN database and the variables selected include outputs supplied, inputs used, as well as compensatory payments for participation in agri-environment programmes and other farm-support programmes. These data concern only values for commodity outputs and their drivers, because no data is available from FADN for the provision of non-commodity outputs. Nevertheless, these data enable a rough comparison of the evolution of the production structure of the farms and provide insights into how they combine their different activities. We analyse the drivers for farm income on a per hectare basis to enable consistent comparison between farms with quite different sizes. Table 2.3 depicts the regression coefficients for Eq. (2.2): farm income = b0 + b1 O + b2 I + b3 IC + b4 S + b5 ES + b6 ES.O + b7 ES.I + b8 ES.IC + b9 ES.S + e (2.2)

Table 2.3 Regression results for the farms’ income…. for several years from the European FADN (all the coefficients are significant at 1% level, except ES.output for the year 2002). Note ES.output is the crossed-effect of environmental subsidies on the marginal effect of outputs on farm income Estimated Driver Unit coefficient 1990 1995 2000 2001 2002 2003 2004 Constant – −21.02 −62.68 −60.13 −31.51 −22.67 −14.93 11.72 b0 1.02 1.00 0.99 0.99 0.99 1.04 0.99 Output (O) Euros/ha b1 Input (I) Euros/ha b2 −0.01 0.06 0.03 0.02 0.07 0.13 0.01 Intermediate consumption (IC) Euros/ha b3 −1.02 −1.07 −1.03 −1.01 −1.10 −1.26 −0.99 Non-environmental subsidies (S) Euros/ha b4 0.99 1.00 1.08 0.97 1.03 1.07 0.99 Environmental subsidies (ES) Euros/ha b5 0 2.20 1.76 1.50 1.45 1.38 1.13 ES.output/1000 b6 0 0.45 −0.03 −0.05 (ns) 0.19 0.20 ES.input/1000 b7 0 −0.58 −0.54 −0.10 −0.40 −2.86 −0.43 ES.intcons/1000 b8 0 0.00 0.91 0.19 0.61 4.20 0.46 0 −1.38 −1.28 −0.40 −0.56 −0.55 −0.62 ES.subsidies/1000 b9

2 Assessment of Multifunctionality and Jointness of Production 25

26

N. Turpin et al.

where O is the outputs (in Euros/hectare), I the inputs used, IC the intermediate consumption, S the amount of subsidies received (excluding environmental subsidies), ES the amount of environmental subsidies and ES.O, ES.I, ES.IC, ES.S are interaction terms (environmental subsidies crossed with outputs, inputs, intermediate consumptions and subsidies respectively). Data have been analysed on a yearly basis because the series are not stationary and thus a panel analysis would lead to unrealistic conclusions. Moreover we assumed that the relations between the various drivers of farm income evolve over time. This assumption is underlined by the sign of some estimators that vary from 1 year to the other. In particular, the constant is negative for most years but becomes positive for year 2004. Similarly, the marginal effect of the environmental subsidies on the output is negative or positive depending on the year considered. Analysis of regression (2) suggests that: –– The effect of output and subsidies on farm income does not vary much from 1 year to the other, although the output price and the producer support subsidies rates are totally different from the early 1990s to the mid-2000s: in the early 1990s, agricultural products have rather high prices that were drastically decreased in 1992 and then in 2000, while subsidies greatly increased (producer support shifted from price support to direct income support). –– There are significant differences for the effect of the inputs and intermediate consumptions drivers between the years. This suggests that there has been an evolution of the farms’ combination of means and we can assume a modification in the jointness rates between commodity and non-commodity outputs within the farms. –– Some environmental outputs are subsidised, but the influence of the agrienvironmental support programmes reduced since the mid-1990s, although the payments increased during that period. This suggest that despite environment is more and more subsidised along time (on our sample data, the environmental payments increased by a factor of 2.3 between 1995 and 2004), the reorganisation of the other drivers compensate their relative effect on farm income over time. –– More important is the effect of the environmental subsidies on the marginal effect of output, input, intermediate consumption and non-environmental subsidies on the farm income: the environmental subsidies moderate the marginal effect of the input, intermediate consumption and non-environmental ones on the farm income. Moreover, the environmental subsidies moderate the marginal effect of the outputs on the farm income in the early 2000s, but this effect reverses in 2003 and 2004, just as if after several years of environmental programmes, some of the farms have been able to reorganise themselves to take advantage of a higher degree of jointness between commodity and non-commodity outputs. Even restricted to economic data, it is possible to conclude that the EU environmental policies have modified the way the farms combine their intermediate consumptions and the inputs they use in their profit function: there are links between the provision of environmental outputs and the evolution of the farm profit function.

2 Assessment of Multifunctionality and Jointness of Production

27

The characterisation of these links deserves further investigation that can be performed at the regional level only because it requires a description of the non-commodity outputs supplied (and thus more precision in the data used). Jointness at the Farm Gate in a Region From a set of FADN farms in the Auvergne NUTS two region, we used the method developed by Perret (2006) to assess whether farms fulfil different functions. Once the existence of multifunctionality was determined, we analysed the origin of jointness in these farms and performed a qualitative assessment of jointness between commodity and non-commodity outputs at the farm scale. Using only expertise and FADN data it is was not possible to analyse all the functions depicted in Table 2.3 at the farm scale. These functions were approached by a set of indicators: –– F2: farm-gate food provider: –– ind 02 This indicator has the value one if the farm processes agricultural products into ready-to-sell food (and zero otherwise). –– F4: contribution to rural viability: –– ind 7–1 Contribution to the maintenance of rural viability when the farm is located in a low density rural area. This indicator has the value one when the number of on-farm households is greater than two and the farm is located in a low-density area and the value zero when the number of on-farm households is less than two or if the information is not precise enough so that this indicator cannot be assessed. –– ind 7–2 Contribution to the maintenance of rural viability when the farm is located in an area frequented by tourists. In this case, the role of farms is more important to the maintenance of scenery rather than in terms of providing employment (that already exist through tourism). This indicator has the value one when the farm, located in an area frequented by tourists, breeds cattle and the value zero if this is not the case or if the information is not precise enough so that this indicator cannot be assessed. –– F5: employment: –– ind 9 Family employment. This indicator has the value one if total family employment on the farm exceeds one unit of farm employment, and the value zero if this is not the case. –– ind 10 Farm employment. This indicator has the value one when paid labour minus seasonal employment is greater than 0.5 units of farm employment, and the value zero if this is not the case. –– ind 11 Seasonal employment. This indicator has the value one when seasonal employment exceeds 0.2 units of farm employment and the value zero if this is not the case.

28

N. Turpin et al.

–– F6: landscape: –– ind 12–1 Contribution to open space. This indicator is a function of several variables, and has the value one when the share of on-farm forests, moors and low productive areas in total on-farm area and grazed land is below 5%. –– ind12–3 Contribution to landscape patchwork. This indicator is derived using a scoring mechanism: the main crops are grouped together (cereals, corn, oil seeds, etc.) and each group is given a value of one when its area is greater than 10% of the total area, and a percentage of points for lower areas (for example, a group sharing 5% of the total area is assessed a score of 0.5 point). This indicator has the value one when the total score is greater than two and the value zero if less than two. –– F7: water quality: –– ind 13 Water quality. The farm is considered to be contributing to good water quality when the total amount of organic nitrogen spread is lower than 70 kg N/ha and either (a) the share of agricultural area with bare soils in winter if lower than 30%, or (b) for a share of bare soils in winter comprised between 30% and 60%, the farmer joins a programme of nitrate or pesticides management, or (c) the share of organic farming is greater than 75%, or (d) for a share of bare soils greater than 60%, the share of organic farming is greater than 75%. When one of these conditions is met this indicator is given a value of one, otherwise a value of zero. –– F8: biodiversity: –– ind 16 Management of ecologically rich habitats. This indicator has the value one when the farmer’s share of low productive grassland is greater than 12.5% (when he uses collective pastures) and 25% when he does not use collective pastures. –– ind 17 Diversity of crops. This indicator is derived using a scoring mechanism: each crop is assessed a score (see ind12–3 for details of calculation). When the total score is greater than three, ind 17 equals one. Table 2.4 depicts the assessment of the above indicators on a subset of farms, in a presentation close to Table 2.1. Put differently, this corresponds to the identification of jointness which we regard as the precursor to the qualitative and quantitative assessment of jointness that will be explored below. The Auvergne set of farms consists of a total of 354 individuals. A preliminary exploration of data, performed using principal component analysis (PCA), is depicted in Fig. 2.2. The analysis suggests that jointness exists between the supply of commodity outputs (variable PBRTO: total gross product) and the supply of non commodity outputs (indicators designed above). Moreover, the farms in this area are liable to sign a contract, named on-farm territorial contract, according to which the farmers are subsidised for the provision of non-commodity outputs: but, because the variable that depicts this contract, CTEXP, is orthogonal to the two first axis of the principal components analysis, we can consider that the signature of the contract has no influence on the degree of jointness in our set of farms.

Table 2.4 Identification of jointness for a subset of Auvergne farms Social Economics Rural viability Employment IND02 IND71 IND72 IND09 IND10 0 0 1 1 0 0 0 0 1 0 1 0 0 1 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 1 0 0 0 1 0 1 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 IND11 0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 0

Environment Landscape IND121 1 0 0 0 0 0 1 0 0 1 0 1 0 0 1 0 IND123 1 1 1 1 1 1 1 1 1 1 1 1 0 1 1 1

Biodiversity IND16 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0

IND17 1 1 1 0 0 1 1 1 1 1 1 1 0 1 1 1

Water IND13 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0

2 Assessment of Multifunctionality and Jointness of Production 29

30

N. Turpin et al. Variables (axes F1 et F2 : 32,14 %) 1 IND71

0,75

IND09

F2 (14,41 %)

0,5

IND72

SAUTI

IND121

0,25

PBRTO

OTE11 IND17 IND02 IND16

0

CTEXP IND11 IND123

IND10

–0,25 IND13

–0,5 –0,75 –1

–1

–0,75

–0,5

–0,25

0 0,25 F1 (17,73 %)

0,5

0,75

1

Fig. 2.2 Principal component analysis performed on the Auvergne farms set (axes 1 and 2)

Qualitative Assessment of Jointness More sophisticated procedures are required to proceed with steps two and three of our framework (qualitative and quantitative assessment of jointness), so we turned to SEAMLESS-IF. SEAMLESS-IF does not directly assess the jointness in the production system between the supply of commodity and non-commodity outputs. The FSSIM-MP component represents the farmer’s behaviour with regard to the possible relations between the production factors inside the production process of commodity outputs. Environmental non-commodity outputs are modelled as externalities (their supply is non-volunteer) or as constraints in the production process of commodity outputs (see Chapter 5 for more details). The first simulations for some farms in four regions already suggest interesting insights about jointness at the farm level despite the lack of operational indicators. Figure 2.3 illustrates the modelled nitrate (NO3−1) leaching with regard to farms income (per hectare) for some SEAMLESS-IF crop farms in Auvergne. When the sample of farms contains more than a few individuals, the relationship between both variables is not that simple. Figure 2.4 depicts the marginal effect of environmental subsidies on farm income for FADN farms in the Auvergne region, for 2000 and 2004. The estimated coefficient is similar to coefficient b5 in Table 2.3, but estimated for each individual farm (in Table 2.3 it is estimated on the whole sample of FADN farms). It is obvious from the Figure that this marginal effect is not the same for the 2 years, but we can notice an increased effect for some farms and a decreased effect for others. Aggregating these individual

2 Assessment of Multifunctionality and Jointness of Production

31

Leaching (kg N-NO3/ha) 70 60 50 40 30 20 10 farm income (euros/ha) 0 0

50

100

150

200

250

300

350

400

Fig. 2.3 Simulated N leaching depending on the farm income for some SEAMLESS-IF farms

marginal effect of env. subs. on income 2004 100

80

60

40

20

0 –10

0 –20

10

20

30

40

50

60

70

marginal effect of env. subs. on income 2000

Fig. 2.4 Total marginal effect of environmental subsidies on farm income for FADN farm groups in Auvergne (the size of the dots represent the share of the total agricultural area for each group of farms)

32

N. Turpin et al.

variations at the regional level would require a calibrated procedure to assess in which sense positive and negative variations compensate (or not) each other.

Quantitative Assessment of Jointness This section is a first attempt to quantitatively assess indicators of multifunctionality using SEAMLESS-IF. We opted for an estimation technique that is based on the assumption that was already derived from the outcomes presented in Table 2.4 which states that environmental subsidies are an estimator for the value of the subsidised non-commodity outputs. Unfortunately, such a proxy could not be found for noncommodity outputs that are not subsidized because of a lack of simulation results. The assumption that environmental subsidies are an estimator for the value of the subsidized non-commodity outputs enabled us to measure the share of modelled income that is due to the environmental subsidies (direct effect plus cross-effects on output, input and intermediate consumptions), for the years 2000 and 2004. Figure 2.5 plots the difference of the 2000 and 2004 values for this indicator. Analysing Figure 2.5 emphasises that regions that receive relatively high environmental subsidies on a per hectare basis (Finland, Austria, East-Germany)

Fig. 2.5 Evolution of the total effect of environmental subsidies of the average farm income in each region between 2000 and 2004 (in blue decreasing effect and in green increasing one)

2 Assessment of Multifunctionality and Jointness of Production

33

experience a relative decrease between 2000 and 2004 in the way these subsidies share their farm income, because of cross-effects with outputs, and mostly inputs and intermediate consumptions.

Conclusions Although the concepts associated with the multifunctionality of agriculture are far from being consensual (see Cairol et al. 2006 for a description of the main concepts), we focussed in this chapter on one concept only, the joint supply of commodity and non-commodity outputs by farms. This is the definition of multifunctionality generally espoused by the EU; many other definitions of multifunctionality are self evident to some degree. Based on this jointness definition, our novel assessment of indicators of multifunctionality relied on three sequential stages: identification of jointness, qualitative assessment of jointness, quantitative assessment of jointness. Identification of jointness was carried out at the level of the farm gate for both an EU sample and using a regional case study of Auvergne, France. FADN data for the EU sample covered only values for commodity outputs and their drivers and thus enabled only a rough comparison of the evolution of the production structure of the farms including insights on how they combine their different activities. By comparison, using FADN data for the regional case study permitted the assessment of which functions different farms fulfil using De Groot’s (2006) classification of functions fulfilled by natural and semi-natural ecosystems. Qualitative assessment of jointness examined the relationship between farm income and nitrogen leaching for a subset of farms in four regions but problems with data availability make it difficult to authoritatively comment on the degree of jointness. However, for the Auvergne case-study it was possible to show how the marginal effect of environmental subsidies on farm income is not constant over time. Moving to assessing the final stage in our framework, quantitative assessment of jointness extended the approach used to identify jointness at the farm gate level for an EU sample, constrained by problems of data availability already noted with this sample. However, under the assumption that environmental subsidies are an estimator for the value of subsidised non-commodity outputs it was possible to show that regions that receive relatively high environmental subsidies on a per hectare basis experience a relative decrease between 2000 and 2004 in the way these subsidies share their farm income, because of cross-effects with outputs, and mostly inputs and intermediate consumptions. Although the results outlined in this chapter serve to illustrate that multifunctionality of agriculture is far from being negligible, a comprehensive assessment of this phenomenon in terms of both functions and spatial scales covered by such an analysis is still beyond the bounds of possibility because of problems with data availability and model capabilities. Given the prominence placed on the multifunctionality concept by the European Commission, it is imperative that this situation improves in order to more fully and robustly assess this concept. Nevertheless, and

34

N. Turpin et al.

as noted in the introduction to this chapter, because the current state-of-the-art in terms of multifunctionality research is limited, from a quantitative point of view the work presented here is an important contributor to this research area.

References Baumol, W., Panzar, J., & Willig, R. (1981). Contestable markets and the theory of market structure. New York: Harcout, Brace & Jovanovich. Belletti, G., Brunori, G., Marescotti, A., & Rossi, A. (2003). Multifunctionality and rural development: A multilevel approach. In G. van Huylenbroeck & G. Durand (Eds.), Multifunctional agriculture: A new paradigm for European agriculture and rural development (pp. 55–80). Hampshire: Ashgate. Bontems, P., Rotillon, G., & Turpin, N. (2005a). Acceptable reforms of agri-environmental policies. Paper presented at the American Agricultural Economics Association Annual Meeting, Providence, RI, 24–27 July 2005. Bontems, P., Rotillon, G., & Turpin, N. (2005b). Self-selecting agri-environmental policies with an application to the Don watershed. Environmental and Resource Economics, 31, 275–301. Brandt, J., & Vejre, H. (2004). Multifunctional landscapes – motives, concepts and perspectives. In J. Brandt & H. Vejre (Eds.), Multifunctional landscapes – theory, values and history (pp. 3–31). London: WitPress. Brouwer, F. (ed). (2004). Sustaining agriculture and the rural environment: Governance, policy and multifunctionality. Cheltenham: Edward Elgar. Brunstad, R. J., Gaasland, I., & Vårdal, E. (2005). Multifunctionality of agriculture: An inquiry into the complementarity between landscape preservation and food security. European Review of Agricultural Economics, 32(4), 469–488. Cairol, D., Perret, E., & Turpin, N. (2006). A report on results of the Multagri project concerning indicators of multifunctionality and their relevance for SEAMLESS-IF, PD 2.3.1, SEAMLESS integrated project. EU 6th Framework Programme (contract no. 010036-2), www. SEAMLESS-IP.org, 80 p. Retrieved from http://www.seamless-ip.org De Groot, R. S. (2006). Function-analysis and valuation as a tool to assess land use conflicts in planning for sustainable, multi-functional landscapes. Landscape and Urban Planning, 75(3–4), 175–186. Gillette, C., Merchez, L., & Perret, E. (2005a). Indicateurs territoriaux de multifonctionnalité agricole pour un développement durable plus fonctionnel. Paper presented at the Colloque international Indicateurs territoriaux du développement durable, Aix-en-Provence, December 1 and 2, 2005. Gillette, C., Merchez, L., & Perret, E. (2005b). Agricultures multifonctionnelles et constructions territoriales en Rhöne-Alpes: Regards croisés entre statistiques et dires d’acteurs. Paper presented at the Symposium International Territoires et enjeux du développement régional: Résultats de recherche en partenariat avec cinq régions, Lyon, March 9–11, 2005. Hall, C., McVittie, A., & Moran, D. (2004). What does the public want from agriculture and the countryside? A review of evidence and methods. Journal of Rural Studies, 20, 211–225. Jongeneel, R. A., & Slangen, L. (2004). Multifunctionality in agriculture and the contestable public domain in the Netherlands. In F. Brouwer (Ed.), Sustaining agriculture and the rural environment: Governance, policy and multifunctionality (pp. 183–203). Cheltenham: Edward Elgar. Knickel, K., Renting, H., & van der Ploeg, J. D. (2004). Multifunctionality in European agriculture. In F. Brouwer (Ed.), Sustaining agriculture and the rural environment: Governance, policy and multifunctionality (pp. 81–103). Cheltenham: Edward Elgar. Muradian, R. (2001). Ecological thresholds: A survey. Ecological Economics, 38(1), 7–24. Nowicki, P. L. (2004). Jointness of production as a market concept. In F. Brouwer (Ed.), Sustaining agriculture and the rural environment: Governance, policy and multifunctionality (pp. 36–55). Cheltenham: Edward Elgar.

2 Assessment of Multifunctionality and Jointness of Production

35

OECD. (2001a). Multifunctionality: Towards an analytical framework. Paris: Organisation for Economic Co-operation and Development. OECD. (2001b). Multifunctionality: Applying the OECD analytical framework. Guiding policy design. Paris: Organisation for Economic Co-operation and Development. OECD. (2003). Multifunctionality: The policy implications. Paris: Organisation for Economic Co-operation and Development. OECD. (2006). The new rural paradigm, policies and governance. Paris: Organisation for Economic Co-operation and Development. Perret, E. (2006). La multifonctionnalité de l’agriculture: regards sur les exploitations agricoles de Rhône-Alpes. Agreste Rhône-Alpes – Coup d’oeil, 87, 4. Sakuyama, T. (2005). A decade of debate over non-trade concerns and agricultural trade liberalisation: convergences, remaining conflicts and a way forward. International Journal of Agricultural Resources, Governance and Ecology, 4, 203–215. Stapleton, L. M., Young, S. D., & Crout, N. M. J. (2004). Have missing markets for ecological goods and services affected modelling of terrestrial C and N fluxes? Ecological Modelling, 179, 569–574. Van der Heide, C. M., Overmars, K. P., & Jongeneel, R. A. (2007). Land use modelling for sustaining multiple functions in the rural countryside with an application in the Achterhoek region, the Netherlands. In Ü. Mander, H. Wiggering & K. Helming (Eds.), Multifunctional land use; meeting future demands for landscape goods and services (pp. 251–268). Berlin/ Heidelberg: Springer. Van der Ploeg, J. D., & Roep, D. (2003). Multifunctionality and rural development, the actual situation in Europe. In G. van Huylenbroeck & G. Durand (Eds.), Multifunctional agriculture: A new paradigm for European agriculture and rural development (pp. 37–55). Hampshire: Ashgate. Van Huylenbroeck, G. (2003). Multifunctional agriculture: How to provide incentives to farmers? 13th International IFMA Congress of Farm Management, August 2003, Australia. Van Huylenbroeck, G., & Durand, G. (eds). (2003). Multifunctional agriculture: A new paradigm for European agriculture and rural development. Hampshire: Ashgate. Vanslembrouck, I., & van Huylenbroeck, G. (2003). The demand for landscape amenities by rural tourists. In G. van Huylenbroeck & G. Durand (Eds.), Multifunctional agriculture: A new paradigm for European agriculture and rural development (pp. 83–100). Hampshire: Ashgate. Vatn, A. (2002). Multifunctional agriculture: Some consequences for international trade regimes. European Review of Agricultural Economics, 29(3), 309–327. Vejre, H., Abildtrup, J., Andersen, J., Andersen, P. S., Brandt, J., Busck, A., et al. (2006). Multifunctional agriculture and multifunctional landscapes – land use as interface. In Ü. Mander, H. Wiggering & K. Helming (Eds.), Multifunctional land use – meeting future demands for landscape goods and services (pp. 93–104). Berlin: Springer. Waarts, Y. (2005). Indicators for the quantification of multifunctionality impacts. Research report MEASCOPE, deliverable 2.4, European Centre for Nature Conservation, 107 p. Wiggering, H., Dalchow, C., Glemmitz, M., Helming, K., Müller, K., Schultz, A., et al. (2006). Indicators for multifunctional land use – linking socio-economic requirements with landscape potentials. Ecological Indicators, 6, 238–249.

Chapter 3

The Institutional Dimension in Policy Assessment Insa Theesfeld, Christian Schleyer, Konrad Hagedorn, Jean-Marc Callois, Olivier Aznar, and Johanna Alkan Olsson

Introduction There is an urgent need for scientifically well-founded ex-ante policy assessment from an institutional perspective. Currently, policy analysis focuses mainly on ex-post policy impact studies to evaluate past policy performance. While there is a vast amount of institutional ex-post case studies and indicator databanks, there are no standardised procedures in state-of-the-art institutional economics using this information for making predictions of the institutional feasibility of policies. Similarly, there is a need for standardised procedures that can easily be linked to

I. Theesfeld (*) Leibniz Institute of Agricultural Development in Central and Eastern Europe (IAMO), Theodor-Lieser-Straße 2, 06120, Halle (Saale), Germany e-mail: [email protected] C. Schleyer Ecosystem Services Research Group, Berlin-Brandenburg Academy of Sciences and Humanities, Jägerstrasse 22/23, 10117, Berlin, Germany e-mail: [email protected] K. Hagedorn Department of Agricultural Economics, Humboldt University of Berlin, Philippstrasse 13, Building 12, 10099, Berlin, Germany e-mail: [email protected] J.-M. Callois and O. Aznar Cemagref – UMR Métafort Campus universitaire des Cézeaux, 24, avenue des Landais, BP 50085, 63172, Aubiére Cedex, France e-mail: [email protected]; e-mail: [email protected] J. Alkan Olsson Centre for Sustainability Studies, Lund University, Box 170, 221 00, Lund, Sweden e-mail: [email protected] F.M. Brouwer and M. van Ittersum (eds.), Environmental and Agricultural Modelling: Integrated Approaches for Policy Impact Assessment, DOI 10.1007/978-90-481-3619-3_3, © Springer Science+Business Media B.V. 2010

37

38

I. Theesfeld et al.

environmental and agricultural models widely used for policy impact assessment. Both issues point at unsolved theoretical and methodological challenges concerning standardised ex-ante institutional analysis of policies. In this chapter, we introduce a formalised procedure for modelling – ex-ante – institutional aspects for policy implementation: the Procedure for Institutional Compatibility Assessment (PICA). It has recently been developed in the frame of the SEAMLESS project that conducts research in the field of agri-environmental policy impact assessment. In this project, an ambitious integrative modelling framework for ex-ante assessment of policy impacts on sustainable development has been created. Here, the ‘SEAMLESS-Integrated Framework’ has been designed not only to assess the policies’ likely impacts on environmental, economic, and social systems, but it also has to provide indications on whether a policy under scrutiny is feasible from an institutional perspective and, thus, can be expected to become effective (van Ittersum et al. 2008). In this context, PICA has been developed as an explorative and flexible, yet formalised methodology to assess the compatibility between policy options1 and various institutional contexts. Further theoretical and methodological development of the PICA approach should also provide a vision of how to cope with a number of unsolved problems inherent in the analysis of institutions for sustainable resource use. Such problematic issues may encompass examinations of the incentive structure faced by individuals in different decision-making contexts, bounded rational behaviour, informal institutions as an important part of the institutional environment, and the complexity of transactions related to nature. PICA is still work in progress; it has been tested in the Auvergne and in Midi-Pyrénées (France) to gain more insights for modifying and refining the method (Amblard et al. 2008a, b; Schleyer et al. 2007b). Yet, PICA needs to be further developed and the resulting concept tested as a validated and innovative means to cope with these problems and to serve the theoretical and methodological needs. Following an overview about prominent approaches for policy assessment, we outline the basic assumptions leading to the concept of institutional compatibility we use in this chapter. We elaborate on all four distinct steps of PICA, while in the subsequent section we focus on PICA Step 1: the classification system to derive distinct policy types. PICA Step 1 is a crucial and the most generic step within the procedure determining the focus of the subsequent steps. Then, we illustrate different modes of action of the procedure using particular elements of the EU Nitrate Directive as a policy example. In the concluding section, we discuss the functions of PICA within the SEAMLESS-Integrated Framework.

1 We conceive policy options as (sets of) policy instruments that a policy maker intends to implement to reach a (set of) policy objectives; i.e., the policy instruments are not implemented at the time of the assessment.

3 The Institutional Dimension in Policy Assessment

39

Overview of Approaches for Policy Assessment Policy analysis guides the process of selecting appropriate policy options to be put into practice. The analysis is commonly subdivided into two categories: ex-ante and ex-post analysis. Ex-post policy analysis is designed to evaluate past policy performance, e.g., in terms of effectiveness, transparency, and distributional fairness to reach policy objectives and includes a wide range of methods, such as surveys, case studies, etc. The capability of these approaches, however, is limited since they do not provide for a way for evaluating the effects of policies prior to their implementation. In contrast, experiences with ex-ante evaluations are still rare (Blazek and Vozab 2006; Todd and Wolpin 2006). In the early phase of the policy life cycle, the ‘Cost of Policy Inaction (COPI)’ method is often used as an ex-ante evaluation tool. In particular, COPI supports the policy recognition phase of the policy life cycle when the emphasis is on identifying problems, warning, communicating the need for policy action, and sketching the urgency of the policy problem relative to other issues. COPI is used to identify and quantify roughly the environmental damage occurring if no new policy is designed to address the underlying (environmental) problem or if the existing policies are not revised accordingly. The purpose of COPI is to highlight the need for action, prior to the specific development and appraisal of policy instruments. COPI is not suitable, however, for comparing and choosing between different policy options, or for judging on the efficiency of policies (Bakkes et al. 2006). Methods that support a later phase of the policy life cycle – the selection of policy options – are subsumed under the notion of ‘ex-ante impact assessment’. They usually comprise some form of simulation where potential actions are pretested in an artificial setting in order to gather information about possible consequences (Becker 2001: 315). There are two main forms of ex-ante impact assessment: (a) Environmental Impact Assessment that is applied to assess planned projects, and (b) Strategic Environmental Assessment that is used for the ex-ante impact assessment of policies. The various forms of ex-ante – mainly environmental – impact assessments that are conducted are often accompanied by social impact assessment (ibid: 312). Technology assessments as well as economic and fiscal impact assessments are often combined with social impact assessments, too. In the first step of a social impact assessment, scenarios are designed to sketch out possible future contexts for the actor system and the target system. Thereafter, strategies are designed that might be able to mitigate or even eliminate the problem. Here, various economic models are used in forecasting these strategies, i.e., the effects of a project or policy. For instance, Capello and Spairani (2004) use scenario building methodology to estimate growth and spatial distribution of the Gross Domestic Product in alternative scenarios for communication and infrastructure policies. Another example is provided by Todd and Wolpin (2006) who employ a dynamic behavioural model of schooling and fertility to forecast the effects of a program on school and work choices and on family fertility. In contrast, ex-ante impact assessment incorporating the institutional

40

I. Theesfeld et al.

perspective of policy implementation has hardly been an issue in economic analysis. Further, while political science and sociology sometimes address institutional aspects in ex-ante impact assessments they do not focus on the effects institutions have on the (economic) decisions of individuals (e.g., North 1991). In cases where it is possible to quantify costs and benefits in monetary terms, a major tool for ex-ante impact assessment is the Cost-Benefit Analysis (CBA). In contrast to COPI, CBA has a narrower and more concrete focus and tends to work with more specific data. Despite its widespread use, it has many practical and conceptual difficulties associated with monetising costs and, in particular, benefits of a proposed policy. This is particularly true in developing and transition countries where methods of quantification are generally underdeveloped. Further, those countries are rather unfamiliar with systematic assessments of the benefits and costs of new regulations (Kirkpatrick et al. 2003). Due to the methodological difficulties to monetise costs and benefits, CBA is hardly objective and is slanted in various ideological directions. Thus, the role of CBA within a political context is often that of political argument, not scientific evidence (Bickers and Williams 2001; Kirkpatrick et al. 2003: 15). There are other supporting valuation methods that try to capture likely policy impacts in general and the problem of monetising environmental benefits and costs in particular. These are: (a) the Contingent Valuation Method, i.e., a stated preference method where respondents value changes in environmental services and goods on hypothetical markets (Wagner 2000), (b) the Choice Modelling, i.e., a method that is derived from Contingent Valuation, but where respondents’ choices of their preferred alternatives demonstrate their willingness to trade-off one attribute against another (Morrison and Bennet 2004), (c) the Travel Cost Method, i.e., the valuation of benefits of an environmental asset (usually a recreational area) by using the costs of consumption of the ecosystem services, including travel costs, entry fees, on-site expenditures, and outlay on capital equipment necessary (Hanley and Splash 1993), and (d) the Hedonic Pricing, i.e., deriving values of environmental services, such as clean air and water, biodiversity or landscape, from observed differences in prices of affiliated market goods, such as houses or labour (Freeman 1993). These methods can be – and actually are often – embedded into CBA in particular as they provide a basis to monetise public goods that do not have a market value. An alternative to CBA is the multi-criteria analysis (Figueira et al. 2005), in particular social multi-criteria evaluation (Munda 2004). This method tries to introduce more realistic assumptions in their models. Parts of institutional constraints of policy implementation, which are related to hidden interests, influence of lobbying groups, power relations, social participation, ecological awareness, cultural constraints that can be expressed in terms of different actors` values and preferences, can be incorporated into these models. The difficulty with multi-criteria analysis is to assess information about preferences, yet it represents an appropriate promising framework which at least offers the possibility to incorporate institutional aspects by the way the problem is structured. Still, from an institutional perspective, costs and frictions of policy design and implementation are not addressed by these different methods; not the least because they are difficult to estimate and quantify ex-ante.

3 The Institutional Dimension in Policy Assessment

41

Another way to come to ex-ante predictions of the likely impacts of policies or projects is to implement alternative versions of the policy in an experimental situation and to compare their relative impacts and effects. Despite the fact that such an experimental approach is often too costly and time consuming to be feasible for policy design purposes, in some cases experimental data has been used successfully to validate forecasting model outputs (e.g., Todd and Wolpin 2006). A particular form of these experiments is the ‘natural experiment’ (Rosenzweig and Wolpin 2000), a method where treatments are purposively randomised to overcome the problem of self-selection that often leads to misinterpretations. This brief overview shows that there is a lack of methods and procedures of institutional ex-ante evaluation of policies, let alone, reliable and good indicators. However, effectiveness and cost-effectiveness (including transaction costs for design and implementation) of a particular policy depend, among other things, on the institutional environment and the institutional arrangements in place. There will be high transaction costs of implementation if the institutional context does not ‘fit’. Given the strengths and weaknesses of the different approaches, there is a need to triangulate methods and to complement the tool box for ex-ante policy analysis from an institutional perspective.

The Concept of Institutional Analysis for Ex-ante Policy Assessment In this section, we outline the basic assumptions underlying the concept of institutional analysis for ex-ante policy assessment. After sketching out briefly our understanding of institutions as the fourth dimension of sustainability we particularly highlight the importance of compatibility between policy instruments and the respective institutional contexts.

Institutions for Sustainability Institutions are defined as the formal and informal rules of a society or of organisations that facilitate co-ordination among people by helping them form expectations. They also function as constraints that shape human interaction and the enforcement characteristics of these constraints (North 1990: 3).2 In the course of evaluating the progress of implementing the United Nations Program of Action ‘Agenda 21’, the Commission on Sustainable Development of the United Nations (UNCSD) defined sustainability as having four dimensions. Besides the economic, social, and environmental dimension, institutions are defined as the

Institutions do also define certain organisations, but these organisations are best thought of as not being institutions, but as being defined by institutions (Bromley 1989: 43).

2

42

I. Theesfeld et al. Institutions

Social

Environment

Sustainable Development

Economy

Source: own figure.

Fig. 3.1 Institutions for sustainability – the fourth dimension

fourth dimension of sustainability (Spangenberg and Bonniot 1998; Spangenberg et al. 2002). Likewise, in the SEAMLESS project, the fourth dimension represents an important challenge to fully integrate economic, social, and environmental sustainability objectives. Institutions for sustainability are defined as the necessary institutional structure capable of delivering economic, social, and environmental sustainability objectives that are set when choosing a policy option (Fig. 3.1). Thus, a policy option ‘integrating’ all three dimensions can only be effective if proper institutional arrangements are in place in the respective countries or regions.3

Institutional Compatibility The institutional analysis of policy options follows the concept of institutional compatibility. The latter refers to the compatibility between policy instruments and the respective institutional context to assess the effectiveness and costeffectiveness of policies. Effectiveness and cost-effectiveness of policies depend on the institutional arrangements (property rights and governance structures4) in place.

3 It should be noted that there is no single or universal institutional arrangement that is linked to a specific policy and that enables a specific policy option to become effective regardless where it is (supposed to be) implemented. Rather, the concrete set of institutions most conducive for policy implementation is likely to vary among countries and regions. 4 Governance structures are the organisational solutions for making rules (institutions) effective, i.e., they are necessary for guaranteeing the rights and duties and their use in co-ordinating transactions (e.g., Ostrom 1990).

3 The Institutional Dimension in Policy Assessment

43

The institutional analysis is based on the assumption that policies will affect certain areas of reality, which are already subject to valid and (more or less) effective institutions. An understanding of the prevailing institutional context in which a policy is to be implemented is necessary to assess intended and unintended consequences (Aligica 2005; Esty et al. 2005: 11; Bickers and Williams 2001: 235). Thus, on the one hand, appropriate institutions increase the likelihood of actually achieving the policy objectives, i.e., they increase the likelihood of actors’ compliance and (intended) change of behaviour. On the other hand, appropriate institutions ensure that these policy objectives are achieved at reasonable costs. Policy instruments that have proven to be very cost-effective in one specific institutional context might perform rather poorly in another, i.e., they might be not effective at all, or they might induce higher costs to become effective. For example, a regulatory or command-and-control policy that puts a ceiling on the allowed amount of pesticides used per hectare and year might be ineffective if there is no authority in place to monitor and sanction farmers’ non-compliance. Here, effectiveness could be increased by establishing such an institutional mechanism; yet, the costs for establishing it might be substantial, thus, reducing the cost-effectiveness. The justifiable costs to be borne by society to make the policy effective cannot be defined by scientists; it depends upon public opinion and political will. However, the role of scientists can be to identify and to specify those transaction costs in a more transparent manner. This information would enable policy makers to design better policies and to make their choices on a more solid basis. In particular if agricultural, environmental, and rural policies are concerned, suitable governance structures have to address the specificities of nature-related transactions, constraints in accessing information, and the prevailing interdependencies of the actors, i.e., the fact that the choice of one actor may influence the choices other actors make. These characteristics are often overlooked in conventional economics which assumes that agents are independent (Paavola and Adger 2005) and largely ignores the complexity of nature-related transactions (Hagedorn et al. 2002; Hagedorn 2008). Political jurisdictions targeted by a policy have to match, in an appropriate manner, with the range of physical, economic, social and, in particular, institutional linkages found in the rural areas and in the agricultural sector. If carefully designed, governance structures can facilitate communication and co-ordination among diverse networks of stakeholders in EU agricultural, environmental, and rural policy making, thus, making effective policy implementation more likely. The institutional analysis within the SEAMLESS project is being conceptualised to reveal where (in which country or region) a policy option in the implementation phase of the policy cycle would be compatible with the existing institutional structures, and where an institutional misfit that is likely to hamper policy implementation can be expected. Thus, not the institutional performance, such as transparency and distributional fairness, to reach policy objectives is analysed, but whether policy characteristics fit or misfit the institutional arrangements in place.

44

I. Theesfeld et al.

The Procedure for Institutional Compatibility Assessment (PICA) Policy analysis has to be linked to a thoughtful examination of the institutional conditions in which the policies will be implemented, i.e., the institutional contexts in which individuals and groups are seeking to act on their preferences and shared understandings (Bickers and Williams 2001: 234). PICA is based on the assumption that the effectiveness of a policy and the cost-effectiveness of its implementation depend to a large extent on the degree of compatibility between concrete policy instrument(s) and the respective institutional context in a country or region. Following Aligica (2005), an adequate and correct understanding of the institutional configuration and of the situational logic of the environment in which a policy is to be implemented has to be produced as a necessary precondition for assessing the balance between the intended and unintended consequences of that policy. To minimise unexpected and possibly disastrous outcomes, it is important that those who craft and modify rules do understand how particular combinations of rules affect actions and outcomes in a particular ecological and cultural environment (Ostrom 2005: 3). This is particularly important for policy makers at higher administrative levels who often have no direct relation to the problems on the ground. According to Boettke and Coyne (2005), models of human interaction based on economic theory often have their problems and limitations in real social settings. Similarly, although aware of the oversimplification, most agri-environmental models used for policy analysis assume that with the implementation of a new policy the institutional arrangements conducive for that policy will be perfectly in place, or that a sub-optimal institutional arrangement will change automatically towards ‘perfection’ at once and with no costs. In addition, it is often assumed that the actors will comply with the policy. PICA as an innovative approach provides a method that relaxes or ‘corrects’ these assumptions to narrow the gap between theory and the ‘real-world’. The approach combines both an explorative procedure to identify those institutional incompatibilities that are likely to hamper or foster policy implementation, and an analytical framework that enables to reveal the causes underlying the incompatibility of policies and institutions, thus, providing the background for institutional innovations to overcome such problems. The current design of PICA comprises four distinct working steps: –– Step one: The policy options are clustered according to (a) type of intervention (regulatory, economic, and advisory), (b) area of intervention (hierarchy/bureaucracy, market, and self-organised network), (c) possibly induced property rights changes, and (d) the attributes of the natural resource(s) addressed (Hagedorn et al. 2002). This classification allows identifying the generic structure of a policy option. –– Step two: Each policy type is characterised by a specific set of crucial institutional aspects (CIA).5 An initial list of 40 crucial institutional aspects linked to common policy types in agriculture, environment, and rural development has been compiled in the frame of the SEAMLESS project. In this chapter only selected crucial institutional aspects will be introduced. The complete list can be found in Schleyer et al. (2007a).

5

3 The Institutional Dimension in Policy Assessment

45

–– Step three: Indicators help to evaluate the potential of respective CIA to constrain or foster the implementation of a policy option. The institutional indicators6 are selected from existing indicator lists, perhaps modified, or new proxies are elaborated.7 Further, concrete assumptions on links and relationships between a CIA and the respective set of indicators are made. –– Step four: Combinations of the identified CIA and assessment of their relative explanatory power lead to thematic statements about an institutional fit or misfit between policy options and institutional contexts. PICA outputs – which are mainly qualitative in character – are grouped in thematic categories of institutional compatibility and, thus, allow for drawing conclusions about an institutional fit or misfit between policy options and institutional contexts. Accordingly, this result of the PICA procedure is functioning as an early warning system as it informs the policy maker very early of potential institutional incompatibilities that may prevent the proposed policies from being actually implemented or that make them less effective. The result of the PICA procedure can, thus, also serve as a starting point for a subsequent analysis of the causalities of the institutional incompatibilities foreseen and for imploring possibilities to change policies and/or institutions to overcome these incompatibilities. Stakeholders at national and regional level and scientific experts are important sources of information throughout the whole PICA procedure. In particular, the PICA expert team8 that is commissioned to conduct the assessment will consult systematically stakeholders and scientific experts to discuss the appropriateness and completeness of the suggested CIA for the region or country, the selection and evaluation of the institutional indicators and their respective values, and the evaluation and categorisation of the selected CIA. Methods used for this interaction may include semi-structured interviews, group interviews, and focus groups (Amblard et al. 2008a, b; Schleyer et al. 2007b). Further, PICA allows for a close interaction with the policy maker(s) who commission the institutional compatibility assessment. While this is particularly important in PICA Step one when the policy option has to be described in all necessary detail and in PICA Step four when the PICA results are presented and discussed, the policy maker has also the opportunity to interact with the PICA expert team in all other phases of the procedure.

Focussing PICA Step One: Deriving Policy Types In this section, the general classification system which is used in PICA Step one to identify the generic structure of a policy option is presented. The policy types Institutional indicators are here defined as variables and proxies that are used as input to the institutional analysis within PICA. They do not represent the results and output of the institutional analysis. 7 About 100 institutional indicators have been compiled in the frame of the SEAMLESS project (Schleyer et al. 2007a). 8 The PICA expert team is part of the SEAMLESS expert team that is carrying out the policy assessment – on behalf of the policy maker – using SEAMLESS-IF. 6

46

I. Theesfeld et al.

introduced here offer a systematic way to classify policy options linked to agriculture, environment, or rural development that a policy maker might wish to assess. The particular type of intervention together with the area of intervention provide the basic information to describe a certain policy type. Additional dimensions used to classify policy options include possibly induced property rights changes and the attributes of the natural resource(s) addressed, such as excludability, rivalry, and complexity. The objective of this specification of policy types is to provide a suitable, yet formalised, structure to identify crucial institutional aspects that are of particular importance for the policy option under scrutiny. It is assumed that the policy type, as represented by distinct boxes in the matrix of Table 3.1, is decisive for the range and kind of crucial institutional aspects that can be expected to be conducive or detrimental to the implementation of this policy option. Practically, this typology does allow limiting the number of CIA that needs to be reviewed when evaluating the policy to be implemented. In the absence of this classification or ‘filter’, all identified CIA relevant for agricultural, environmental, and rural development policies would have to be processed every time a policy option is to be assessed. At the same time, there may be policy options that cannot be categorised without any doubt in one single box; perhaps because the policy option is described rather poorly. While this would indeed extend the range of relevant CIA to be considered in the subsequent analysis it does not jeopardise the overall assessment. Not the least because the initial list of CIA that is derived directly from the categorisation of the policy option is discussed with stakeholders and other experts in PICA Step two and may be revised accordingly (Schleyer et al. 2007b: 15ff.). In the following, the dimensions of the classification system will be explained in more detail. The types of intervention, i.e., the policy instruments are inscribed in the respective rows of the matrix in Table 3.1. They describe how and by which means the impact of a policy shall be reached9: –– Regulatory or command-and-control instruments (compulsory): e.g., laws, regulations, specific protection targets, and designations of areas for protected habitats or species –– Economic instruments often using financial (dis)incentives: e.g., taxes, subsidies, grants and loans, and tradable pollution permits –– Advisory/voluntary10 instruments: e.g., codes of good practice, extension services and other informative measures, and environmental audits Please note that in general all types of intervention can be induced by public as well as private actors. However, while typical applications of PICA would rather address policy options to be induced by public actors, institutional compatibility assessments using PICA could also be carried out for policies induced by private actors. 10 Of course, some economic policies, such as agri-environmental schemes, are also voluntary in character since farmers can choose to participate in those schemes, or not. In contrast, in this category, the term ‘voluntary’ refers to policies that motivate voluntary actions or behavioural changes of actors without direct financial incentives or regulations, i.e., for example, by convincing actors using various kinds of information materials. 9

Example: Implementing new European statutes for cooperatives

Policies that intervene at self-organised networks using regulatory (commandand-control) instruments

Self-organised network

Water

Land/Soil

Biodiversity

Forestry

Example: Budget cuts for (regional) Example: Subsidising organic Example: Providing funds for administrative bodies milk and non-till farming LEADER-Local Action practices Groups Policies that intervene at Policies that intervene at markets Policies that intervene at selfhierarchies/bureaucracies using advisory/voluntary organised networks using using advisory/voluntary instruments advisory/voluntary instruments instruments Example: Providing training Example: Providing information Example: Providing information material on efficient brochures on health and brochures with Best Practicemanagement structures and organic food to consumers; examples; facilitating administrative procedures providing training on knowledge transfer between (Best Practice) environmental friendly networks farming

Policies that intervene at markets Policies that intervene at using economic instruments self-organised networks using economic instruments

Example: Restrictions on nitrate use

Example: Establishing the European Food Safety Authority and Nature Reserves

Policies that intervene at hierarchies/bureaucracies using economic instruments

Policies that intervene at markets using regulatory (command-and-control) instruments

Market

Policies that intervene at hierarchies/bureaucracies using regulatory (commandand-control) instruments

Hierarchy/Bureaucracy

Area of intervention (governance structures)

Complex Resource Systems

Policies that induce changes in property rights for farmers regarding the natural resources they need for production using regulatory instruments Policies that induce changes in property rights for farmers regarding the natural resources they need for production using economic instruments Policies that induce changes in property rights for farmers regarding the natural resources they need for production using advisory/voluntary

Property rights change

Example: The core element of the EU Nitrate Directive that is used as an example to illustrate PICA in Section 6 is allocated to the grey box ‘Regulatory on market’, as uncompensated restrictions do have an impact on the production costs of the farmers.

Natural resource addressed

Advisory/ Voluntary

Economic on hierarchy/ bureaucracy

Type Regulatory of intervention

Table 3.1 Policy type matrix

48

I. Theesfeld et al.

This classification is based on the work of Stone (2002) who distinguishes between five general mechanisms for changing or coordinating behaviour of actors. These are (1) ‘inducements’, i.e., changing people’s behaviour with, often financial, rewards and punishments, here named economic instruments, (2) ‘rules’, i.e., commands to act or not in certain ways, or determining permissions and entitlements, (3) ‘rights’, i.e., strategies that allow individuals, groups, or organisations to invoke government power on their behalf, (4) ‘powers’, i.e., shifting the power of decision making to different people, the last three are here subsumed under regulatory instruments, and, (5) ‘facts’, i.e., strategies that rely principally on persuasion, here named advisory/voluntary instruments. Stone (2002) also stresses that these instruments are ideal types and that no policy option ever relies purely on one type of instrument. A similar distinction is made by Moskowitz (1978: 65ff.) who analyses a wide range of alternative policy options that have the common objective to redirect financial investments from the private sector to ensure neighbourhood preservation. Here, Moskowitz distinguishes between three types of interventions: (a) regulatory policies for mandatory investments, (b) direct subsidies, such as tax benefits to change the final profit estimation, and (c) persuasion by providing facts, figures, and experience to demonstrate that the private sector could realistically expect profits from these investments. This also corresponds with similar distinctions made by environmental economists (e.g., Stavins 2004). The area of intervention points to the governance structures a policy is supposed to have an impact on. More precisely, a policy aims at influencing real-world transactions (e.g., use of pesticides, protection of species, etc.) by changing existing or creating new governance structures that co-ordinate these transactions in such a way that, e.g., their results are internalised by the actors. The differentiation used in PICA follows to a large extent the widely used categories of governance structures (hierarchies, markets, and hybrids) suggested by Williamson (2004). However, first, it can be assumed that almost every governance structure in the real world can indeed be seen as some hybrid form between the polar cases market and hierarchy.11 Thus, in the respective columns of Table 3.1 those areas of intervention that are closer to either market or hierarchy are subsumed. Second, with specifying the third column self-organised network, the attention is directed to a specific (hybrid) form of governance structures that is of particular interest if pursuing agricultural, environmental, and rural development policy objectives (Hagedorn et al. 2002). The column property rights change is the third dimension to describe a policy type. It accounts for changes in private and collective property rights likely to be induced by the policy option, in particular, on natural resources. It covers an important institutional specificity of environmental policies. Undoubtedly, most

While in markets (repeated) economic exchange is based on voluntary bilateral agreements between individuals (e.g., auctions, stock markets, etc.), an authority on a higher level compulsorily selects economic action in hierarchies (e.g., state agencies, but also within private firms). 11

3 The Institutional Dimension in Policy Assessment

49

policy options will imply some changes in property rights. However, here it is defined in a more narrow sense pointing to changes in the property rights of farmers on natural resources needed for production, such as land and water. For example, most environmental policies, such as the EU Flora-Fauna-Habitat Directive or the EU Nitrate Directive, reduce directly farmers’ property rights. Restrictions on land use, like the prohibition to spread manure on the field during winter months, have direct impacts on the individual production decisions of farmers. Thus, these environmental policies, according to the matrix, would address the governance structure ‘market’ as ‘area of intervention’ since restrictions in land use or farming practices are likely to affect the production function of the farmer resulting in higher production costs and, hence, less profit. Yet, these restrictions are also resulting in severe changes in and constraints on (private) property rights of farmers with respect to the (natural) production factor land. In contrast, policies demanding specific health and quality standards of a farmer’s produce to be kept when entering the market would also affect his production function; yet, no direct changes in property rights would be involved. An additional dimension that is complementing this matrix accounts for specificities of the natural resource(s) addressed by the policy option, i.e., water, land/soil, biodiversity, forestry, or complex resource systems.12 Some crucial institutional aspects stem from the fact that the characteristics of a natural resource addressed or the attributes of a transaction related to nature might call for specific institutional arrangements to make a policy option effective (Hagedorn et al. 2002; Hagedorn 2008). For instance, addressing water quality often has to deal with non-point pollution from agriculture that constitutes challenges for adequate forms of monitoring and sanctioning. Further, policies for the protection of biodiversity or specific rare species face particular incentive problems, not the least because the future value of these rare species is uncertain and the benefits of protection cannot only be reaped by the one protecting it. In addition, the geographical dimensions (local, national, or global) of resources can also be important. Thus, distinct institutional aspects for each of the natural resources addressed can be expected. To sum up, the four dimensions necessary to describe a policy type comprehensively are illustrated as a four-dimensional graphic in Fig. 3.2. The x-axis describes the area of intervention, the y-axis the type of intervention, and the z-axis the natural resource addressed. The colour of the cuboid reflects the dimension ‘property rights change’: dark grey, if a property rights change is involved, and light grey, if not. Each cuboid in the space represents a certain policy type. For illustration, AgriEnvironmental Schemes focussing on reducing diffuse pollution of nitrates from agriculture can be assigned to the policy type of the light grey cuboid.

The category complex resource system refers to resource systems with many externalities involved (e.g., wetlands). Here, a policy is targeting the ‘performance’ of the resource system as a whole, rather than single components or resources. 12

50

I. Theesfeld et al. Property Rights Change Type of Intervention

Property Rights

Natural Resources

Forestry

Advisory/ Voluntary

Biodiversity

Water Hierarchy

Market Economic

No Property Rights Change

er

at

W

Regulatory il

Soil Regulatory

So

Area of Intervention Market

Source: own figure.

Fig. 3.2 Four dimensions of a policy type

Institutional Compatibility of the EU Nitrate Directive The EU Nitrate Directive (Council Directive 91/676/EEC) (EC 1991) that was adopted in 1991 can be seen as a prominent and typical example of an EU environmental policy addressing water pollution. We take one of the core elements of this Directive as an example to illustrate PICA: when implementing the EU Nitrate Directive Member States have to draw up and implement action programmes in vulnerable zones designated before that shall consist of mandatory rules. These rules determine, e.g., periods when the application of certain types of fertiliser is prohibited, and limitations of the application rates of fertilisers taking into account the characteristics of the zones concerned, in particular soil conditions, soil type, slope, land use, and agricultural practices (see Annex III of the Directive). Furthermore, Member States have to establish suitable monitoring and enforcement systems to ensure actors’ compliance with the rules. Being aware that the EU Nitrate Directive comprises more and different policy elements that can be combined in diversified ways, to illustrate PICA we only refer to the element of uncompensated and mandatory production restrictions in previously defined vulnerable zones. We focus on this element of the EU Nitrate Directive and treat it as a single policy instrument and, therewith, abstract from distorting effects due to the other – certainly interdependent – policy elements that would also be implemented when the Directive were to be introduced.

3 The Institutional Dimension in Policy Assessment

51

PICA Step One: Classification of the Policy Option Using all available information on the concrete form and content of the policy option provided by the policy maker the PICA expert team categorises this element of the EU Nitrate Directive – according to the matrix of policy types (see Table 3.1) – as a regulatory type of policy having effects on markets. As described above, it demands from the Member States that action programmes are to be implemented that shall consist, among other things, of clearly defined mandatory measures determined in Annex III. Effectively, only the national regulations determine the precise limits of restrictions in time and space. Further, it is assumed that no compensations are paid covering the costs induced by these restrictions.13 These uncompensated restrictions have an impact on the production costs of farmers (e.g., because yields decrease due to restrictions in fertiliser use) and, thus, on their position at the market. More precisely, farmers might be forced to offer their products at a higher price resulting in a decrease in demand for those products or they might keep the price and accept reduced profits.

PICA Step Two: Crucial Institutional Aspects Related to the Policy Option According to the identification of the policy type in the previous step, in PICA Step 2, only those CIA related to regulatory policy instruments intervening in markets have to be considered. Within the SEAMLESS project, an extensive literature review has been carried out to identify CIA that are typically linked with respective policy types (see Schleyer et al. 2007a: Appendix 2). Based on this compiled ‘library of crucial institutional aspects’, those CIA are extracted that potentially hamper or foster the effective implementation of policies of the type ‘regulatory on market’, thus, accruing also to the selected core element of the EU Nitrate Directive, in particular, to the implementation of restrictions in fertiliser use. During the application of PICA, relevant national and regional stakeholders and scientific experts are consulted by the PICA expert team discussing the relevance of every identified CIA for the policy option under scrutiny. Here, some of the CIA extracted from the initial literature review might be regarded as relevant for a policy type in general, but not be considered as crucial for the specific policy option to be assessed. Thus, the PICA expert team can decide to skip some CIA at this stage. In turn, additional CIA that have not yet been covered by the literature reviewed may be included in the assessment of the policy option under scrutiny and may also be added to the library

However, national laws to implement the EU Nitrate Directive may be complemented with various forms of compensation schemes that ease the burden for some stakeholders in general, or in selected areas, for limitations in nitrate application that are beyond what is legally required.

13

52

I. Theesfeld et al.

of crucial institutional aspects. As a result of the consultation process, the following CIA likely to constrain the implementation of the EU Nitrate Directive were suggested and are presented here for illustrative purposes14: –– Strong bargaining power of farmers’ associations: Implementation of mandatory measures restricting the use of fertilisers in designated vulnerable zones affects directly the production costs of farmers in these zones, often leading to income losses. Yet, the (degree of the) concrete restrictions is determined by the respective Member States or regions. Here, it is assumed that a strong agricultural lobby might be able to weaken these mandatory restrictions, or to obtain exception clauses. Thus, strong farmers’ associations might hamper the effective implementation of the EU Nitrate Directive. –– Information asymmetry state versus firm and high level of opportunism: Information asymmetries between public administrations (state) and agricultural producers can be conceived as the result of problems on part of the state to control and monitor the activities of firms. These problems depend, among other things, on the ability (technical/knowledge/human resources) or even willingness of the administration in charge to monitor and, if applicable, sanction actors’ behaviour, but also on the characteristics of the resources (and the related activities to be monitored) concerned. Mandatory measures to reduce water pollution by nitrates are difficult – or very costly – to observe and to measure, e.g., the exact amount of nitrates applied per hectare. Thus, farmers’ non-compliance with prescribed restrictions is not easy to detect and/or non-compliance cannot be associated clearly with single farmers since nitrates diffuse slowly into often large groundwater basins. Furthermore, it is assumed that high levels of opportunism on part of the farmers concerned are likely to exacerbate the problem leading to high costs for controlling necessary to deter actors from cheating.

PICA Step Three: Linking Crucial Institutional Aspects to Institutional Indicators As a result of PICA Step two, the PICA expert team suggests a restricted list of CIA that is considered to be of particular importance for assessing the effectiveness and costeffectiveness of implementing the selected core element of the EU Nitrate Directive.

We do not claim that the CIA presented here are indeed the most relevant crucial institutional aspects related to the selected core element of the EU Nitrate Directive. Yet, we deem them to be reasonably relevant and sufficiently plausible since they are based on the extensive literature review mentioned above and on discussions within the PICA group. 14

3 The Institutional Dimension in Policy Assessment

53

Those CIA that are selected from the library of CIA are linked with at least one institutional indicator from the available portfolio that can help to evaluate the respective CIA, eventually leading to statements about the effectiveness of policy implementation in PICA Step four. For further processing, only those indicators are selected that are considered to have some explanatory power with respect to the policy option under scrutiny. At this stage, the PICA expert team has to interact with other members of the SEAMLESS expert team, in particular with modelling and data base experts. Here, the availability, quality, and geographical scope of quantitative data need to be discussed. Further, the precise forms and scopes of suggested qualitative assessments need to be decided on. For illustration, Tables 3.2, 3.3, and 3.4 contain examples of institutional indicators that might be used as proxies for assessing the extent of the selected CIA. Table 3.2 Institutional indicators for assessing the CIA ‘Bargaining power of farmers’ associations’ Institutional Description/data Data sources/ Expert assumptions on links indicator databases between indicator and CIA15 High percentages indicate National Number of Memberships a strong bargaining Statistical farmers that in farmers’ power of farmers’ Databases; are member associations associations Assessment in a farmers’ by expert association/ group Number of farms*100 Number of farmers’ National Statistical High numbers indicate a Fragmentation associations Databases relatively weak (total) of farmers’ bargaining power of associations farmers’ associations Data assembled by A high number indicates a Proximity between (Number of) expert group high influence on the farmers’ farmers’ political decision making associations associations process at EU level and (of a country) and EU strong bargaining power with official authorities representatives in Brussels SEAMLESS A low ratio may indicate a Structure of Ratio = Number of Databases farming system dominated farming system farms/Number of by large farms (latifundium people employed system) and, thus, a high in the farming influence on the political sector decision making process at national level

Please note that this column will contain specific assumptions on links between indicator, CIA, and policy option when actually running PICA. It will be filled by the PICA expert team after discussing the relevance and sufficiency of available indicators for evaluating the identified CIA with respect to the concrete policy option. This process is also likely to produce a restricted (smaller) list of those institutional indicators related to a respective CIA that can be linked meaningfully with the policy option under scrutiny. 15

54

I. Theesfeld et al.

Table 3.3 Institutional indicators for assessing the CIA ‘Information asymmetry state vs. firm’ Methodology to identify information asymmetry: (1) Identify potential sources of information asymmetry related to the policy under scrutiny; (2) Evaluate the impact of this information asymmetry on the efficiency of this policy; (3) Assess the additional controlling and monitoring costs necessary to reduce the level of information asymmetry to an ‘acceptable’ level Degree of affinity of the government of a country towards devolution

Qualitative assessment by expert group

High additional controlling and monitoring costs necessary to reach an ‘acceptable’ level of information asymmetry indicate a high constraint

Qualitative assessment by expert group

Farm density

Average number of farms per 100 ha

SEAMLESS Databases

Rule of Law

Composite indicator of the extent to which agents have confidence in and abide by the rules of society, and in particular the quality of contract enforcement, the police, and the courts, as well as the likelihood of crime and violence

World Bank

Low degrees indicate high information asymmetries since centralised control and monitoring is more costly High numbers indicate higher controlling and monitoring costs, thus, likely higher information asymmetries Low measures indicate an ineffective/ inefficient existing controlling and monitoring system causing information asymmetries

Information asymmetry

Affinity of governments towards devolution

PICA Step Four: Aggregating Information on Crucial Institutional Aspects of the Policy Option In this final step of PICA, the expert team that runs PICA with the help of external scientific experts and stakeholders is using the information provided by the institutional indicators for a qualitative assessment of the restricted list of CIA.

3 The Institutional Dimension in Policy Assessment

55

Table 3.4 Institutional indicators for assessing the CIA ‘High level of opportunism’ Infringement cases

Number of infringement cases in a country brought before the Court of Justice

National Statistical Databases

Rule of Law

Composite indicator of the extent to which agents have confidence in and abide by the rules of society, and in particular the quality of contract enforcement, the police, and the courts, as well as the likelihood of crime and violence Assessment of popular observance of the law (Part of composite indicator ‘Rule of Law’)

World Bank

Order

World Bank

High numbers of infringement cases indicate high levels of opportunism Low measures indicate high levels of opportunism

Low measures indicate high levels of opportunism

This includes, first, combining the various indicator information available for every single CIA of the restricted list to arrive at a qualitative statement about the relative extent of this CIA in all countries and/or regions. For example, the level of corruption can be determined for every country where the policy option is to be implemented, thus, providing insights in the relative – country-wise – likelihoods for ineffective policy implementation. Second, the PICA expert team is defining thematic categories of institutional compatibility to group the CIA and the respective qualitative statements. While it is certainly helpful to use science-driven categories, such as property rights compatibility, embeddedness compatibility, etc., policy makers who commissioned the assessment might prefer different or additional categories. Each thematic category draws on information from at least one CIA. For the selected core element of the EU Nitrate Directive the PICA expert team suggests to group the information according to the following two thematic categories: 1. Communication capacity

Bargaining power of farmers’ associations

2. Governance structures compatibility

Information asymmetry state versus firm (including high levels of opportunism)

Finally, these categorised region- or country-wise qualitative statements on the compatibility of the policy option will be presented to the policy maker who has commissioned the policy assessment with SEAMLESS-IF. Here, an interactive form of communication is preferred since this provides the opportunity to discuss the results and, perhaps, the introduction of complementary policy instruments in countries or regions where – according to the PICA results – implementation is likely to be substantially hampered. Figure 2.3 summarises the four steps of PICA.

56

I. Theesfeld et al.

Society

Policy Option e.g., EU Nitrate Directive

Share of agricultural votes

Number of farms Type of Intervention

Natural Resource

Area of Intervention

Number of farmers that are members Number of people in a farmers’ association employed in the farming sector

Property Rights Change

Step1: Identify policytype (e.g., regulatory / market) Step 2: Extract crucial institutional aspects (e.g., Bargaining power of farmers’ associations)

Step 3: Select indicators (e.g., Membership in farmers’ associations) Use existing indicators or elaborate new proxies that indicate the extent of the crucial institutional aspect

Step 4: Conclude on e.g., Communication capacity

Fig. 3.3 PICA scheme

Functions of PICA Within SEAMLESS-IF PICA can play an important role within integrated modelling frameworks – like the SEAMLESS-Integrated Framework – that have been developed for an ex-ante assessment of policy impacts on environmental, economic, and social systems. Here, the economic and environmental models often assume that appropriate and required institutions are in place for resource governance towards sustainability, or that those institutions can be implemented with no costs. PICA can be seen as a method that qualifies those underlying modelling assumptions to narrow the gap between theory and the real-world. Thereby, the institutional assessment can strengthen the modelling approaches in the pre- and post-modelling phase. If PICA is applied in the pre-modelling phase, it can provide hints on whether institutional constraints in some or many countries or regions are likely to be prohibitively high and the policy option will hardly become effective there. As a result, it could be recommended – and discussed with the policy makers – to modify the policy option or to carry out additional in-depth institutional pre-studies before running the other models. Similarly, the results can be used to select and modify policy scenarios that are constructed as input for the modelling tools. For example, an application of PICA may reveal that in some countries the implementation of the EU Nitrate Directive is very likely to be hampered due to high levels of opportunism on part of the farmers and very low levels of trust between public

3 The Institutional Dimension in Policy Assessment

57

authorities and farmers. To mitigate these possibly constraining factors and to still reach the policy objectives scenarios could be constructed that include complementary policy measures, such as the (parallel) introduction of voluntary compensation schemes that pay premiums for farmers limiting fertiliser application even beyond the legal restrictions to increase compliance. Alternatively, modified scenarios may encompass programs to either gain the trust of farmers (e.g., using participatory methods for determining the precise restrictions for farmers in a particular region), or additional funding for those public authorities that are supposed to monitor and sanction farmers’ non-compliance. The results of modelling these modified scenarios may reveal whether complementary policy measures could indeed increase the predicted policy performance in those ‘critical’ countries. When applying PICA in the post-modelling phase, it allows for putting the mainly quantitative model results and calculated impact indicators into (institutional) context. This contributes to the validation of the model results on policy impacts. PICA is still work in progress. It has been tested in two study areas in the Auvergne and at the regional level in Midi-Pyrénées (France) with the policy option EU Nitrate Directive to gain more insights for modifying and refining the procedure (Amblard et al. 2008a, b; Schleyer et al. 2007b). The results clearly show that, despite being an explorative tool, all PICA steps can build already on a solid and useful basis derived from theoretical insights and empirical institutional analysis (see Schleyer et al. 2007a). However, neither the current library of CIA as a whole nor the lists of CIA linked to a particular policy type can be seen as static, but need to be revised and complemented continually to improve the accuracy of the predictions. Therefore, it is essential that the experiences made and insights gained during every application of PICA are used systematically and carefully to make the empirical basis of PICA more comprehensive. Thus, the library of CIA can be seen as an ever-growing source of information. The same applies to the library of institutional indicators used in PICA Step three that would need constant revision. Finally, PICA has been designed as a flexible methodological framework for a systematic and rigorous ex-ante assessment of the institutional dimension of policy implementation; a procedure that can and needs to be adapted to the respective policy option under scrutiny.

References Aligica, P. (2005). Institutional analysis and economic development policy: notes on the applied agenda of the Bloomington School. Extending Peter Boettke and Christopher Coyne’s Outline of the Research Program of the Workshop in Political Theory and Policy Analysis. Journal of Economic Behavior and Organization, 57(2), 159–165. Amblard, L., Aznar, O., Mann, C., Schleyer, C., Theesfeld, I., & Hagedorn, K. (2008a). Evaluation and suggestions for improvement of the Procedure for Institutional Compatibility Assessment (PICA) and integration of PICA into the third prototype of SEAMLESS-IF. PD6.5.5.2, SEAMLESS integrated project, EU 6th Framework Programme, contract no. 010036-2, www. SEAMLESS-IP.org

58

I. Theesfeld et al.

Amblard, L., Mann, C., Lemeilleur, S., Thérond O., Schleyer, C., Theesfeld, I., & Hagedorn, K. (2008b). Application of the Procedure for Institutional Compatibility Assessment (PICA) to the implementation of the EU Nitrate Directive in Midi-Pyrénées. Evaluation and suggestions for further improvement and integration into the final version of SEAMLESS-IF. PD6.6.6.1, SEAMLESS integrated project, EU 6th Framework Programme, contract no. 010036-2, www. SEAMLESS-IP.org Bakkes, J. A., Bräuer, I., ten Brink, P., Görlach, B., Kuik, O. J., & Medhurst, J. (2006). Cost of policy inaction. Scoping study for DG Environment. Bilthoven, The Netherlands: Netherlands Environmental Assessment Agency. Becker, H. (2001). Social impact assessment. European Journal of Operational Research, 128, 311–321. Bickers, K. N., & Williams, J. T. (eds). (2001). Public policy analysis: A political economy approach. Boston/New York: Houghton Mifflin. Blazek, J., & Vozab, J. (2006). Ex-ante evaluation in the new member states: The case of the Czech Republic. Regional Studies, 40(2), 237–248. Boettke, P., & Coyne, Ch. (2005). Methodological individualism, spontaneous order and the research program of the Workshop in Political Theory and Policy Analysis. Journal of Economic Behavior and Organization, 57(2), 145–158. Bromley, D. (1989). Economic interests and institutions: The conceptual foundations of public policy. Oxford/Cambridge: Basil Backwell. Capello, R., & Spairani, A. (2004, August). Ex-ante evaluation of European ICTs policies: Efficiency vs. cohesion scenarios. Paper presented at the 44th European Congress of the European Regional Science Association, University of Porto, Portugal. EC (European Commission) (1991). Nitrate Directive. European Council Directive of 12 December 1991 concerning the protection of waters against pollution by nitrates from agricultural sources. (91/676/EEC). Official Journal, L375, 31/12/1991, 0001–0008. Esty, D. C., Levy, M., Srebotnjak, T., & de Sherbinin, A. (2005). Environmental sustainability index: Benchmarking national environmental stewardship. New Haven, CT: Yale Center for Environmental Law and Policy. Figueira, J., Greco, S., & Ehrgott, M. (2005). Multiple criteria decision analysis: State of the art survey. New York: Springer. Freeman, A. M., III. (1993). The measurement of environmental and resource values: Theory and methods. Washington, DC: Resources for the Future. Hagedorn, K. (2008). Particular requirements of institutional analysis in nature-related sectors. European Review of Agricultural Economics, 35(3), 357–384. Hagedorn, K., Arzt, A., & Peters, U. (2002). Institutional arrangements for environmental cooperatives: A conceptual framework. In K. Hagedorn (Ed.), Environmental co-operation and institutional change: Theories and policies for European agriculture (pp. 3–25). Cheltenham: Edward Elgar. Hanley, N., & Splash, C. L. (1993). Cost-benefit analysis and the environment. Cheltenham: Edward Elgar. Kirkpatrick, C., Parker, D., & Zhang, Y.-F. (2003, November). Regulatory impact assessment in developing and transition economies: A survey of current practice and recommendations for further development. Paper presented at the Regulatory Impact Assessment Conference, CRC, University of Manchester. Morrison, M., & Bennet, J. (2004). Valuing New South Wales rivers for use in benefit transfer. Australian Journal of Agricultural and Resource Economics, 48(1), 591–612. Moskowitz, E. (1978). Neighborhood preservation: An analysis of policy maps and policy options. In J. V. May & A. B. Wildavsky (Eds.), The policy cycle (pp. 65–87). Beverly Hills, CA: Sage. Munda, G. (2004). Social multi-criteria evaluation: Methodological foundations and operational consequences. European Journal of Operational Research, 158, 662–677. North, D. C. (1990). Institutions, institutional change and economic performance. Cambridge: Cambride University Press. North, D. C. (1991). Institutions. Journal of Economic Perspectives, 5(1), 97–112.

3 The Institutional Dimension in Policy Assessment

59

Ostrom, E. (1990). Governing the commons. The evolution of institutions for collective action. Cambridge: Cambridge University Press. Ostrom, E. (2005). Understanding institutional diversity. Princeton, NJ: Princeton University Press. Paavola, J., & Adger, N. (2005). Institutional ecological economics. Ecological Economics, 53, 353–368. Rosenzweig, M., & Wolpin, K. I. (2000). Natural “natural experiment” in economics. Journal of Economic Literature, 38(4), 827–874. Schleyer, C., Theesfeld, I., Hagedorn, K., Amblard, L., Aznar, O., & Mann, C. (2007b). First evaluation and suggestion for improvement of the Procedure for Institutional Compatibility Assessment (PICA) and suggestions for its integration into the third prototype of SEAMLESS-IF. PD 6.5.5.1, SEAMLESS Integrated Project, EU 6th Framework Programme, contract no. 010036-2, www.SEAMLESS-IP.org. Schleyer, C., Theesfeld, I., Hagedorn, K., Aznar, O., Callois, J.-M., et al. (2007a). Approach towards an operational tool to apply institutional analysis for the assessment of policy feasibility within SEAMLESS-IF. SEAMLESS Report No.29, SEAMLESS Integrated Project, EU 6th Framework Programme, contract no. 010036-2, www.SEAMLESS-IP.org. Spangenberg, J. H., & Bonniot, O. (1998). Sustainability indicators – a compass on the road towards sustainability (Wuppertal Paper No. 81). Cologne/Berlin, Germany: Wuppertal Institute. Spangenberg, J. H., Pfahl, S., & Deller, K. (2002). Towards indicators for institutional sustainability: Lessons from an analysis of Agenda 21. Ecological Indicators, 2, 61–77. Stavins, R. N. (2004). Environmental economics. In L. Blume & S. Durlauf (Eds.), The new Palgrave Dictionary of Economics (2nd ed.). London: Palgrave Macmillan. Stone, D. (2002). Policy paradox. The art of political decision making. New York/London: W.W. Norton. Todd, P., & Wolpin, K. (2006). Ex-ante evaluation of social programs. PIER (Penn Institute for Economic Research) Working Paper 06-022. Van Ittersum, M. K., Ewert, F., Heckelei, T., Wery, J., Alkan Olsson, J., Andersen, E., et al. (2008). Integrated assessment of agricultural systems – a component-based framework for the European Union (SEAMLESS). Agricultural Systems, 96(1–3), 150–165. Wagner, R. (2000). Monetäre Umweltbewertung mit der Contingent Valuation-Methode. Frankfurt, Germany: Peter Lang. Williamson, O. E. (2004). Transaction cost economics and agriculture: An excursion. In G. van Huylenbroeck, W. Verbeke & L. Lauwers (Eds.), The role of institutions in rural policies and agricultural markets (pp. 19–39). Amsterdam: Elsevier.

Part II

Advancements in Modelling, Data and Software

Chapter 4

A Component-Based Framework for Simulating Agricultural Production and Externalities Marcello Donatelli, Graham Russell, Andrea Emilio Rizzoli, Marco Acutis, Myriam Adam, Ioannis N. Athanasiadis, Matteo Balderacchi, Luca Bechini, Hatem Belhouchette, Gianni Bellocchi, Jacques-Eric Bergez, Marco Botta, Erik Braudeau, Simone Bregaglio, Laura Carlini, Eric Casellas, Florian Celette, Enrico Ceotto, Marie Hélène Charron-Moirez, Roberto Confalonieri, Marc Corbeels, Luca Criscuolo, Pablo Cruz, Andrea di Guardo, Domenico Ditto, Christian Dupraz, Michel Duru, Diego Fiorani, Antonella Gentile, Frank Ewert, Christian Gary, Ephrem Habyarimana, Claire Jouany, Kamel Kansou, Rob Knapen, Giovanni Lanza Filippi, Peter A. Leffelaar, Luisa Manici, Guillaume Martin, Pierre Martin, Eelco Meuter, Nora Mugueta, Rachmat Mulia, Meine van Noordwijk, Roelof Oomen, Alexandra Rosenmund, Vittorio Rossi, Francesca Salinari, Ariel Serrano, Andrea Sorce, Grégoire Vincent, Jean-Pierre Theau, Olivier Thérond, Marco Trevisan, Patrizia Trevisiol, Frits K. van Evert, Daniel Wallach, Jacques Wery, and Arezki Zerourou

Introduction Several simulation tools allow the impact of agricultural management on production activities in specific environments to be studied (e.g. Brisson et al. 2003; Keating et al. 2003; Jones et al. 2003; Stockle et al. 2003; Van Ittersum et al. 2003). Such tools are specialized, to different extents, to one or more specific production activities: arable crops or cropping systems, grassland, orchards, agro-forestry, livestock etc. Their outputs often only include estimates of a restricted range of system externalities which may have a negative environmental impact; these may include, for example, nitrogen leaching or the fate of pesticides. Very often, the structure of such systems neither allows for an easy plug-in of models for new agricultural production activities, nor the use of different approaches for the simulation of processes via alternate formulations. Furthermore, documentation of such tools may not be up-to-date, and may not follow a single standard, which makes it difficult to access information.

M. Donatelli () CRA-CIN-Agriculture Research Council, Via di Corticella 133, 40128 Bologna, Italy e-mail: [email protected]. F.M. Brouwer and M. van Ittersum (eds.), Environmental and Agricultural Modelling: Integrated Approaches for Policy Impact Assessment, DOI 10.1007/978-90-481-3619-3_4, © Springer Science+Business Media B.V. 2010

63

64

M. Donatelli et al.

Finally, when such systems are proprietary systems of either research groups or projects, it may not be possible for third parties to re-use the system for further development. A basic requirement of any biophysical model is that it must be able to simulate the processes which influence significantly the behaviour of a system, and particularly those aspects that relate to the purpose of the model. An obvious example is that the model must not be restricted to potential production if its intended use is to study water-limited agriculture This example, however, is at a “high” level, meaning that simulating water-limited production may require: • The simulation of a different number of approaches; and • Even different approaches for the simulation of the same process, when environmental conditions change As an example of the first point, studying the impact of mulching requires soil evaporation models which react to soil cover beyond that given by canopy cover, and the fate of the mulching material must also be simulated. An example of the second point can be simulating the water budget of conditions typified by peak evapotranspiration of 5 mm day−1 on a deep soil compared with conditions where the peak evapotranspiration is 12 mm day−1 on a shallow, cracking soil. The former case can be simulated with simpler, yet still adequate, approaches compared to the latter. Moreover, some approaches may demand inputs which may not be easily available, thus compromising its operational use. Also, as peer reviewed publications may propose alternative options for modelling processes with the same assumptions; tests need to be carried out to assess performance and reliability throughout the range of operational conditions. Finally, effective simulation of a biophysical system, no matter what level of simplification is chosen to simulate its behaviour, requires expertise in different domains. This is a demanding task that requires a multi-team effort for system analysis and model development. All these reasons, argue for a flexible and modular simulation system, and provide, in effect, a specification for the simulation system described in this chapter. The advent of component-based software engineering has enabled the development of scalable, robust, large-scale applications in a variety of domains, including agro-ecological modelling. In systems analysis, it is common to deal with the complexity of an entire system by considering it to consist of linked sub-systems. This leads naturally to thinking of models as being made of sub-models. Such a conceptual model can be implemented as a computer model composed of connected component models. This type of implementation has at least two major advantages. First, new models can be constructed by connecting existing component models of known and guaranteed quality to new component models. This has the potential to increase the speed of development. Secondly, the predictive capabilities of two different component models can be compared, as opposed to only comparing whole simulation systems. Further, common and frequently used functionalities, such as numerical integration, visualization and statistical ex-post analysis, can be implemented as generic tools which are developed once and shared by all the model developers.

4 A Component-Based Framework for Simulating Agricultural Production

65

In the past decade there has been an increasing demand for modularity and replaceability in biophysical models (e.g. Jones et al. 2001; David et al. 2002; Donatelli et al., 2003, 2004, 2006a), aimed both at improving the efficiency of use of resources and at fostering a higher quality of modelling through specialization of model builders in their specific domain. The modular approach developed in the software industry is based on the concept of encapsulating the solution of a modelling problem in discrete, replaceable, and interchangeable software units called components. A software component can be defined as “a unit of composition with contractually specified interfaces and explicit context dependencies only. A software component can be deployed independently and is subject by composition by third parties” (Szypersky et al. 2002). Component-oriented designs actually represent a natural choice for building scalable, robust, large-scale applications, and to maximize the ease of maintenance in a variety of domains, including agro-ecological modelling (Argent 2004). This concept has been applied to biophysical simulation and has led to the development of modelling frameworks such as Simile, MODCOM, IMA, TIME, OpenMI, SME, and OMS (Argent and Rizzoli 2004; Rizzoli et al. 2004), which allow use to be made of components by linking them either directly or through a simulation engine. Three major aspects of model implementation are specific to the modelling platform, demand consistent development resources, and are real barriers to reusability. These are: • Data input/output procedures (e.g. input/output data handling and file management) • Common services (e.g. state variable integrator, simulation event handler); and • Graphical user interfaces (GUI) Modelling frameworks can play a key role in addressing these issues. First, the framework allows the application-specific parts of simulations to be segregated from the code employed to accomplish common tasks, thus greatly enhancing code re-use (Hillyer et al. 2003). Secondly, by defining those elements of the framework that actually contain the model implementation and how the elements are used, a designer can be presented with a clear path from conceptual model to simulation (Hillyer et al. 2003). Furthermore, by avoiding the need to re-implement common services, resources can be concentrated on the development of simulation components. Developing a simulation system based on the component-oriented paradigm poses specific challenges in terms of both biophysical model linkages and implementation architecture. Component-based architectures demand the definition and implementation of sub-systems which minimize the need for links to other components, and the need for repeated communication between components. However, even when a system to be simulated is divided into sub-systems with little need for communication between them, data exchange prior to integration within a time step is needed, thus an articulated interface is needed that allows such calls. Although being potentially prone to mix and match “everything” is often suggested as an intrinsic weakness of component-based systems, this problem can be overcome by shifting the focus to the components themselves using semantically rich interfaces which ensure that the linked variables are appropriate. To illustrate the concept, if a component makes available a variable characterized by units, range of use, type and description, and

66

M. Donatelli et al.

another component requires the same variable as an input, the link can be considered correct if a check of the variable attributes show that these are identical, whereas the correctness of the variable as an input must be investigated within the component producing the output. The principle of applying “parsimony” is of course still valid in model building. For instance, there is no point in coupling two components in which strong assumptions (and thus the limitations) of one impose an unnecessary burden on the modelling capabilities of the other. This, however, applies both to monolithic and component-based system development. As always, the choice of model should be conditioned by both the intended application of the model and a comprehensive system analysis, and this is totally independent of the type of implementation. The SEAMLESS project has developed a framework to integrate analyses of impacts on a wide range of aspects of sustainability and multi-functionality (Van Ittersum et al. 2008). This requires the evaluation of the agricultural outputs and system externalities for a wide range of production systems and environments. Although some indicators of system performance can be provided using static models derived from existing databases, estimating system behaviour for novel techniques or existing techniques applied to new environments requires process based simulation. Also, even for known systems, some of the externalities due to agricultural production are only available as observational data for a very small number of experimental sites. The analysis of the biophysical components of agricultural systems thus requires a simulation framework which can be extended and updated by research teams, which allows easy incorporation of research results into operational tools, and which is transparent with respect to its contents and its functionality. The problems and requirements outlined in the previous paragraph have formed the basis of the design of the Agricultural Production and Externalities Simulator (APES) which offers flexibility in being an open modelling environment that allows an extensible set of modelling choices. The emphasis in APES has been to provide a transparent and flexible modelling platform that can be adapted to different modelling goals. This is a quite different rationale from the specific biophysical modelling solutions that are currently implemented.

APES: The Agricultural Production and Externalities Simulator APES is a simulation model system for estimating the biophysical behaviour of agricultural production systems in response to the interaction of weather, soil and agro-technical management options. The system allows the incorporation, at a later date, of other modules which might be needed to simulate processes not included in the existing version, such as the impact of plant pests (see also Fig. 4.1). Biophysical processes are simulated in APES using deterministic approaches which are mostly based on mechanistic representations. The criteria to select modelling approaches were the need to: • Account for specific processes to simulate soil-land use interactions • Input data to run simulations, which may be a constraint at the European scale

4 A Component-Based Framework for Simulating Agricultural Production

67

Fig. 4.1 Main typologies of models and outputs of APES. The details of both are described in the text

• Simulate all agricultural production activities of interest (e.g. crops, grasses, vineyards, agro-forestry); and • Simulate agro-management and its impact on the system

Component Based Structure There is no single solution to the problem of splitting complex systems into components. However, some divisions are more effective than others. The criteria used for doing this in APES were: • Consistency with knowledge about the organization of the real system • Consistency with the goal of encapsulating a useful/reusable set of modelling solutions relevant to the specific domain; and • Minimization of the need for communication between components within a time step This has led to components being developed with different model granularities (from the whole system perspective) as one of the possible solutions to modularization of agricultural production systems. Figure 4.2 shows the APES model components included in the December 2008 release.

68

M. Donatelli et al.

APES - Agricultural Production and Externalities Simulator Production Enterprises CLIMA libraries

Weather reader

Pedo Transfer Functions Diseases

Crop

CropML

Tree

Grass

Light interception

Root distribution

Water uptake

Field manager

Soil water

Soil N

Soil erosion

Agro chemicals

Soil water2

Soil C-N

Soil temperature

Soil reader

APES Engine APES engines are MODCOM applications

Agro management

Abiotic damages

Soil

Mass balance

Fig. 4.2 The APES “coarse” component diagram. Note that there are alternate options for simulating soil water, soil nitrogen, and crops; also, within each of the components there can be alternate approaches for simulating processes

Model Components APES is composed of two main groups of software units: the simulation engine which uses the modelling framework MODCOM (Hillyer et al. 2003), and the model components, which include a cross-component unit to compute mass balance. Model components can be grouped into agricultural management, soil components, production enterprise components, and weather. The description of the models, implemented is available in the help files of each component (see “Web resources”). Help files are in general divided into two sections: “models”, which contain the model description, targeted at model users, and “design and use” which contain component information targeted at developers. The components on public release also include a code documentation file and sample applications in a software development kit. All models use a daily time step for integration and communication across modules, although calculations can be carried out with a shorter time step within a component. Each component contains one or more existing models which simulate the constituent processes. The relevant references are listed in the documentation of each component. A brief summary description of each component follows; the teams which have developed each component are detailed (refer to authors’ affiliation for explanation of the acronyms). Components are grouped with reference to Fig. 4.2.

4 A Component-Based Framework for Simulating Agricultural Production

69

Agro-management Components AgroManagement: Rule Based Modelling of Technical Management The AgroManagement component was developed by CRA and is designed to implement field management actions during simulation. An agricultural activity is defined, in this context, as a production enterprise such as a crop rotation, i.e. an assemblage of crops, an orchard etc., associated with a production system characterized in terms of outputs and inputs such as high input, high output (e.g., irrigated, high nitrogen fertilization, minimum tillage). Such an integrated system must be implemented in a way that imitates as closely as possible farmers’ behaviour. Limiting the drivers of the decision making process to the biophysical system implies that each action must be triggered at run time via a set of rules which can be based on the state of the system, on constraints of resource availability, or on the physical characteristics of the system. However, simulating management in a component-based system poses challenges in defining a re-usable framework which is able to account for the complete range of agricultural management technologies applied to particular enterprises. Finally, the implementation of management must allow different approaches to be used for modelling its impact on different model components. The AgroManagement component formalizes the decision making process in models called rules, and the drivers of the implementation of the impact on the biophysical system as a set of parameters encapsulated in data-types called impacts. Rules and impacts are both easily extendable, thus allowing a wide range of modelling approaches to be used. Furthermore, the information on the biophysical system is passed through a data-type called states, which can also be extended in case new rules require additional variables. The outputs from the management actions, applied as a result of rules evaluated at run-time, needed to provide the simulation output (output to a text file, an XML file or a database are all currently available) can be fully customized by the user and added to without recompiling the component. Currently, the management actions which can be implemented are nitrogen fertilization (mineral and organic), tillage, irrigation, pesticide application, and crop, tree, and grassland operations. The software implementation is such that new agromanagement typologies, and new actions within the typologies, can be easily added. The rule-based model is characterized by three main sections: –– Inputs: states of the system and time –– Parameters specific for each rule (values are compared to states of the system via the rule model) –– A model which returns a true/false output Rules, which are based on relative date or on a set of state variables, are implemented as a class encapsulating their parameter declarations and tests of pre-conditions (this also allows management configuration files to be validated using pre-condition tests). One feature of interest is that implementing the rule approach allows the formalization of what is generically referred to as “expert knowledge”.

70

M. Donatelli et al.

For example, expert knowledge which suggests that in a specific environment, a farmer will “plant maize on a date later than April first, if it has not rained for the last three days, and when average air temperature has been above five °C for seven days continuously” can be formalized and used in simulations. The italicized words are the parameters of the rule to be compared with system states/exogenous variables at run-time (e.g. the condition “no rain for the last seven days” is tested against the values of rain at run time starting from April first as in this example). The possible uses of such formalization include building a consistent quantitative database of agricultural management across Europe, optimising parameters in climate change scenarios as an adaptation strategy and using such metrics in climate change impact assessments, and improving technical management in current conditions through rule-parameter optimization. Parameters are needed by model components to implement the impact of management actions. Some are common to many management events (e.g. management type) while others apply to a specific management event (e.g. amount of water for irrigation, tillage depth for tillage). Other parameters are needed by specific modelling approaches and generally differ even within specific management event types (e.g. implement type and an associated set of eight parameters is needed for modelling tillage according to the WEPP (Water Erosion Prediction Project) approach, Alberts et al. 1995, as opposed to other approaches which do not need such information). All model components reference the AgroManagement data-types to trigger management impact models at runtime. An example of the graphical representation of a management configuration for a 3-year rotation is shown in Fig. 4.3.

Fig. 4.3 Agro-management scheduled actions in a 3 year rotation. For simulations longer than this, the sequence is repeated. Vertical bars in the upper section of the graph are actions scheduled at a relative (to year) date; horizontal bars are actions scheduled in a time window, if other conditions are met; horizontal bars with a shading gradient are actions scheduled with an end date but associated with a phenological event (the width of gradient boxes is arbitrarily fixed as 30 days in this graphical representation)

4 A Component-Based Framework for Simulating Agricultural Production

71

Production Enterprise Components AbioticDamage: Damage on Plants by Abiotic Factors The AbioticDamage component was developed jointly by JRC and CRA. It implements several approaches for the simulation of abiotic damage to crops. Models are implemented with a fine granularity. The constituent models currently belong to five categories: lodging, frost, cold-induced spikelet sterility, ozone, and salinity. –– Lodging implements the approach proposed by Baker et al. (1998), modified by Acutis et al. (2008), assuming that the dominant parameter that affects lodging is the wind-induced bending moment at the stem base. –– Frost (Ritchie 1991) calculates crown temperature, hardening and de-hardening index, a killing temperature, the possible reduction in leaf area index, and evaluates if the crop has been killed by the frost. –– SpikeletSterility implements two different approaches, proposed by Confalonieri et al. (2006) and Shimono et al. (2005). An option model allows an automatic choice between them according to input availability. The Confalonieri approach is based on the computation of hourly stresses which are summed to compute the daily stress. The Shimono approach computes daily stress directly, but it requires the calibration of empirical parameters. The different susceptibility to sterility in the period between spikelet initiation and heading is accounted for by both models. –– Ozone contains a complex model for the simulation of the damage due to ozone. It models leaf aerodynamic and boundary layer resistance (Spiker et al. 1992), calculates average leaf conductance using the method of Georgiadis et al. (1995), and calculates the fractional reduction of plant production as a function of the ozone flux through the stomata and the leaf conductance of water using the approach of Sitch et al. (2007). – – Salinity implements two different approaches, proposed respectively by Ferrer-Alegre and Stockle (1999) and by Karlberg et al. (2006). The FerrerAlegre approach is based on the calculation of plant conductance and then of a function for the estimation of salinity stress in different layers of the vegetation. The Karlberg approach calculates the reduction of nutrients partitioned to the leaves due to salinity stress on the roots. AgroChemicals: Pesticide Fate The Agrochemicals component was developed by UNICATT, and is a one-dimension model that simulates the pesticide fate at field scale with a daily time step for communication with other modules; this component was developed by using new knowledge (Jantunen et al. 2005; Balderacchi et al. 2007) to modify earlier models (Carsel et al. 1988; Tiktak et al. 2001). The model considers five environmental compartments where the pesticide can be stored:

72

M. Donatelli et al.

–– Canopy surface: represents the pesticide deposit on branches, leaves, fruits, shoots and green parts (hence the outer part of the plant) –– Plant: represents the agrochemical stored inside the plant –– Soil available fraction: the pesticide quantity which can move and can be transformed by biotic processes –– Soil aged fraction: the pesticide trapped within soil micro pores and organic matter and not available for transformation –– Soil bound fraction: the pesticide fraction that cannot be extracted from the soil without altering its physical-chemical structure and therefore not available for transformation The models were implemented in four modules, representing environmental compartments: –– Air considers the processes that happen before the product reaches the soil including drift and plant interception. –– Crop simulates only the plant mass balance, although in this first prototype plant is only a sink of pesticide. –– Canopy simulates the processes that happen on the leaf surface. –– Soil describes the pesticide flow through the soil profile. Each soil profile has to be split in numerical layers; the equations which describe the fate of pesticides differ for the top and bottom layers, because there are different boundary conditions. The processes that redistribute the pesticide into the system connect two compartments and are: –– –– –– –– –– ––

Penetration: from canopy surface to plant Wash off; from canopy surface to soil available fraction Ageing: from soil available fraction to soil aged fraction and vice versa Binding: from soil available fraction to soil bound fraction Plant uptake: from soil available fraction to plant Transport in liquid and gaseous phase: between the soil available fraction compartments of different layers

The output variables of greatest interest due to their importance for environmental pollution are: amount of pesticide lost due to drift; amount of pesticide volatilized, amount of pesticide lost due to run off, amount of pesticide lost to the drain system when present, amount of pesticide leached, and amount of pesticide remaining in the soil profile. Crop: Crop Development and Growth The crop component was developed by PPS and CIRAD. It simulates crop growth and development for the major crops of Europe. Tropical crops are currently being added. Crop growth is based on the interception of radiation by green plant parts and its conversion into dry matter. Crop development goes through vegetative

4 A Component-Based Framework for Simulating Agricultural Production

73

and, for some crops, reproductive phenological stages, at a rate that depends on physiological time, expressed as a temperature sum. The timing of readiness for harvest is also simulated. Potential crop growth is defined by temperature, radiation and crop phenology. The crop growth and development model simulates both potential growth and attainable growth, which is limited by water and nitrogen availability. The crop component uses a generic crop simulator in which parameters and modelling approaches can differ according to the crop simulated. The crop component has been based on the concept of light interception and utilization from the Lintul model. However, modifications and additions have been introduced to extend the list of crop types for which the model can be used. These changes include the implementation of alternative modelling approaches for each of the main crop physiological processes, such as: –– –– –– –– –– ––

Leaf area expansion Biomass accumulation (Monteith 1977) Biomass partitioning (van Keulen and Seligman 1987) Phenology (van Keulen and Seligman 1987; Streck et al. 2003; Hearn 1994) Senescence N dynamics (Shibu et al., 2009)

The model set-up allows new approaches for modelling these processes to be included easily. Parameter sets for 19 crops are currently available, including cereals, legumes, roots and tuber crops, and comprising determinate and indeterminate, and winter and spring crop types. The crop list can be extended not only by adding new crops but also varieties suitable for a particular environment, by editing existing parameter sets of the relevant crop type. In SEAMLESS-IF, these parameters are fine-tuned for regional applications by defining two correction factors that adjust crop cycle duration and crop radiation use efficiency. CropML: Crop Development and Growth The CropML (Crop Model Library) was developed by JRC, CRA, and UNIMI. The component implements alternative modelling solutions from those in the Crop component using different generic and crop-specific crop models. The architecture adopted allows easy extension of the component through the incorporation of other models. In fact, the fine granularity used for coding the different processes related to crop growth and development allow the re-use of the same strategies for other modelling approaches where common algorithms are present. CropML is implemented as two separate components, CropML and CropML. Interfaces. The first includes all the algorithms (the models), the second interfaces, domain classes, and information about crop model parameters. Three versions of the component were developed: CropML, CropML.WaterLimited, and CropML. NitrogenLimited. The last two extend, respectively, CropML – CropML.Interfaces and CropML.WaterLimited – CropML.WaterLimited.Interfaces. The models currently implemented are the plant growth and development approaches of CropSyst (Stockle et al. 2003), WOFOST (van Keulen and Wolf 1986), and WARM

74

M. Donatelli et al.

(Confalonieri et al. 2006). The first two models are generic crop simulators, based respectively on the simulation of net and gross photosynthesis; the latter is a model specifically for rice simulations. CropSyst simulates net daily biomass accumulation using two approaches: the first is based on the concept of radiation use efficiency (RUE); the second is a vapour pressure – corrected Transpiration Use Efficiency approach. Each day, the minimum of the two biomasses is taken. CropSyst simulates daily biomass partitioning as a function of Specific Leaf Area at emergence, cumulated biomass, and an empiric parameter representing the partitioning of biomass between stems and leaves. WOFOST simulates the daily fixation of CO2 (gross photosynthesis), the growth and maintenance respirations and a dynamic partitioning between leaves, stems and storage organs. WARM is based on the RUE approach, accounting for limitations to RUE due to temperature, senescence, saturation of the enzymatic chains and diseases. A dynamic approach for biomass partitioning of assimilates into stems, leaves and storage organs is also adopted in this case, driven by a single input parameter. The CropML – WARM component uses a micrometeorological component, TRIS (Temperature in paddy-RIce Simulation). The TRIS component simulates the floodwater effect on the vertical soil thermal profile in paddy rice fields. TRIS is particularly important for rice simulations in temperate environments, where there is a significant effect of floodwater on temperature (one of the main driving variables in cropping systems models). Two alternative models were implemented, a mechanistic and an empirical one, for use according to data availability. The first is based on the solution of surface energy balance equations and estimates the temperature of floodwater, of each 10 cm canopy layer from the air–water interface to the top of the canopy, of the meristematic apex, and of the canopy. The model has an hourly time step. Context strategies allow also the generation or estimation of canopy and meteorological variables according to their availability in the domain classes. If needed, hourly inputs can be generated using the CLIMA libraries (Fig. 4.2). Maximum and minimum daily temperatures of floodwater, meristematic apex, and mid-canopy are calculated. The empirical model is based on modified Gaussian filters which reproduce the smoothing effect of water on daily thermal extremes, and the water heat storage capacity. The component can be extended through the implementation of alternative approaches, e.g. for the simulation of meteorological variables into the canopy profile. Diseases: Air-Borne Plant Diseases The Diseases component was developed by CRA and UNICATT. It allows the impact of plant disease epidemics on plant growth and yield to be estimated. It consists of four modules providing a generic frame to simulate disease development: –– –– –– ––

Disease progress Inoculum pressure (initial conditions) Impact on plants Agricultural management impact on pathogen populations

4 A Component-Based Framework for Simulating Agricultural Production

75

The Disease progress module simulates the epidemics of a generic air-borne fungal pathogen, considering the following components of the infection process: infection (Analytis 1977; Magarey et al. 2005), incubation, latency, infectiousness (Blaise and Gessler 1992; Wadia and Butler 1994), sporulation (Analytis 1977), and spores dispersal (Aylor 1982; Waggoner 1973; Waggoner and Horsfall 1969). These processes, which are driven by weather conditions and interactions with the host plant (Zadoks and Schein 1979), are modelled as a function of meteorological variables, temperature, air relative humidity, vapour pressure deficit, leaf wetness duration, rain, and wind speed – hourly values estimated/generated from the CLIMA libraries (Fig. 4.2), and parameters specific for each host-pathogen combination. As output, the Disease progress module returns the proportion of host tissue affected compared to the total host tissue. The initial conditions for infections are derived from a pool of models which use information about the preceding crops, site-specific potential, and a random component obtained by sampling from a distribution, either provided by default or fitted from historical data. Impact on plants is currently implemented as a reduction of the photosynthesizing host tissue according to the Bastiaans’ model (1991), but it will be extended to a more direct interaction with plant simulation as some air-borne pathogens such as rusts inhibit the conversion of solar radiation to dry matter. Finally, agromanagement is accounted for through the impact of fungicide applications and other disease control actions on the fungal population. Prototype study applications have been made for vineyard diseases and powdery mildew of wheat (Blumeria graminis f.sp. tritici). The component also contains a model to simulate the pathogen rice blast (Pyricularia orize) and its impact on the rice crop, and a generic model for potential infection (Magarey et al. 2005), with parameters for more than 80 diseases. Grasses: Grassland Growth and Quality The grassland component was developed by INRA. It allows a wide diversity of grassland types to be simulated: (i) sown grass species including tall fescue (Festuca arundinacea), perennial ryegrass (Lolium perenne) and cocksfoot (Dactylis glomerata), sown legumes such as alfalfa (Medicago sativa), permanent grasslands ranging from plant communities growing under nutrient-poor to those from nutrient-rich conditions, and mixtures of grasses and alfalfa or white clover (Trifolium repens). For species-rich permanent grasslands the approach is based on plant functional traits (Lavorel and Garnier 2002; Duru et al. 2009). The grass growth module is similar to that of the crops component except that additional functionalities are included: –– The calculation of the herbage feeding value: (i) protein content using the standing herbage mass and the crop nitrogen index; (ii) digestibility from herbage age, nitrogen index and plant type (Duru 2008; Duru et al. 2008). –– A detailed phenological sub-model (McCall and Bishop-Hurley 2003), for which parameters are specific to vegetation type in order to simulate a large range of defoliation regimes (cutting, grazing, short and long regrowth periods), over the vegetative or the reproductive phase.

76

M. Donatelli et al.

In comparison to crops, some simplifications have been made: –– Leaf area is calculated only as a function of environmental factors and species type and not of dry matter accumulation. –– Over the reproductive phase, stem biomass accumulation is calculated as a fraction of the above-ground herbage mass (Calvière and Duru 1999). –– The root mass is considered constant over the growing season (as in many herbage growth models, e.g. Schapendonk et al. 1998; Barrett et al. 2005); i.e. root growth rate is assumed to be the same as the root death rate. In grass-legume mixtures, the competition between the two components is not simulated. An a priori sward composition is defined that depends on the nitrogen and defoliation management.

FieldManager: Spatial Information for Multiple Plant Species The FieldManager component, which was developed by INRA, provides field dimension information for heterogeneous stands such as agroforestry plots or vineyards where rows of trees or other woody perennials separate cultivated strips of either crop or grassland. Field configuration is defined by three parameters: WidthIntraRow (distance between trees in a row), WidthInterRow (distance between rows), WidthCultivatedStrip (width of cultivated strip between each row). Three field configurations are currently available: for crop alone, for a continuous row of grapevine with a cultivated strip of Crop or Grassland, for a row of timber trees with a cultivated strip of crop.

LightInterception: Light Interception and Competition by Canopies The LightInterception component, which was developed by INRA, implements models that estimate the daily interception of solar radiation partitioned between one or two types of plant and the soil. From the daily solar radiation, a simple field description (from FieldManager) and a small number of plant parameters that characterize canopy dimensions (LAI, crown dimension), this component estimates the fraction and amount of PAR (Photosynthetic Active Radiation) intercepted by the tree canopy, any crop or grassland canopy below, and the soil. Three models are available in the current version, all derived from the geometrical model of Pronk et al. (2003), but corresponding to different field configurations (continuous rows, rows of cubes, or rows of cuboids). These models assume homogeneity in the light transmission properties of leaves and uniform canopies (this assumption is needed to permit the one-dimension simplification of canopy structure in other parts of the model). In the case of a single crop the model follows the Beer-Lambert law (Monteith and Unsworth 1990) for light interception.

4 A Component-Based Framework for Simulating Agricultural Production

77

MassBalance: A Library of Mass Balance Tests The MassBalance component was developed by CRA. It implements equations of balance in discrete code units (strategies). The water and the nitrogen balance are currently implemented. At each time step the component computes the active balances, and outputs the values. Each balance has a threshold parameter as maximum departure from zero allowed; this threshold can activate a flag at run time. The component, which can be extended, is designed to be used either in the test phase of a new modelling solution, or permanently as a check on correct functioning. RootDistribution: Roots Growth and Distribution The RootDistribution component, which was developed by INRA, estimates the partitioning of fine roots between layers in the soil profile. Only one model is currently available, derived from Hi-sAFe 3D model (Root Voxel Automaton, Mulia 2005; Mulia and Dupraz 2006). From the daily fine root growth and death, the soil layer thickness and the water extraction of the previous day, the allocation of root length and biomass to the different soil layers is estimated using an opportunistic growth paradigm. Parameters can be fitted to adjust the water and distance to collar sensitivity so that root profiles of most species can be simulated. It is assumed that the horizontal distribution of crop roots is homogeneous, given that the root distribution is a one-dimensional simplification of the system. Tree: Woody Plant Growth and Quality The Tree component was developed by INRA. It is a generic woody plant model designed to simulate grapevines, fruit trees and timber trees; it is currently parameterized for grapevines. It simulates crop growth, production and product quality. Its basic features are similar to those implemented in the Crop component: temperature drives crop development, and intercepted radiation determines the potential growth rate that can be reduced by water or nitrogen limitations. Dry matter and nitrogen are partitioned between the growing organs on the basis of partitioning tables (Nendel and Kersebaum 2004; Vivin et al. 2002; Wermelinger and Koblet 1990). Some features are specific to woody crops. Where these crops are planted in rows, as is generally the case, their canopy is heterogeneous and the Tree component converts the daily carbon increment into an increase in crown dimensions (needed in the Light interception component, Pronk et al. 2003). Woody crops are perennial so carbon and nitrogen can be stored during the annual crop cycle and used during the next cycle (particularly at bud break) (Castelan-Estrada 2001). The Tree component calculates variables that define fruit crop quality, which is important for crops such as grapevines or apple trees. For grapevines, the dynamics of fruit water and sugar contents depend on the thermal time during the phase of grape growth after the onset of ripening (veraison) (Ollat et al. 2002).

78

M. Donatelli et al.

The tree component does not use fixed partitioning tables for timber trees because this would prevent the modelling of the impact of branch or root pruning on the tree growth. The tree component includes a dynamic module of C allocation for timber trees, governed by two types of rule: –– Teleonomic allocation rules based on allometric equations define the relative sizes of above-ground sub-compartments, e.g. the relationship between Diameter at Breast Height (DBH) and tree height. Such allometric relationships capture internal constraints not explicitly dealt with in the model (e.g. architectural model and structural stability constraints limiting amount of leaf biomass for a given amount of wood biomass) in relation to the tree dimensions. –– An optimal allocation assumption (‘functional balance’) between above- and below-ground biomass mediated through stress indices, which assume that a plant allocates its biomass so as to maximise its growth rate under the given environmental conditions. This approach has been extended to the ratio between coarse and fine roots, with a dynamic allocation procedure that avoids the need for fixed partitioning tables. WaterUptake: Plant Water Uptake The WaterUptake component was developed by INRA. It allows daily plant water extraction to be estimated from each of the soil layers and for each plant species (one or two possible) in the field from the plant water demand, the soil description, and the root distribution in the soil. Roots are assumed to be homogeneously distributed horizontally. Two types of model have been implemented. The simple model for cases with up to two plant species, assumes that the water demand is met as long as water is available in the soil layers containing roots (water content above wilting point). The second model is the more complex one used in the His-AFe model for mixed vegetation. This model estimates the amount of water that each plant can extract from the soil by integrating the matric flux potential for each plant in each rooted layer.

Soil Components SoilCarbonNitrogen: Soil Carbon and Nitrogen Dynamics The SoilCarbonNitrogen component was developed by CIRAD and INRA. The models implemented describe N mineralization-immobilization turnover and the interactions between C and N dynamics in decomposing plant residues and soil organic matter (SOM). It includes above- and below-ground plant residue pools and three soil organic matter pools (microbial biomass, young and old SOM) with different turnover times (Fig. 4.4). Rates of decomposition are modified by temperature, moisture, lignin content of the residues and N availability. Stabilization of SOM is simulated by transferring fractions of decomposed microbial biomass and young

4 A Component-Based Framework for Simulating Agricultural Production

a

Foliar and fine root litter MP

CO2

LCP L

HCP CO2

Woody litter WP L

LCP L

HCP

WP L

CO2 CO2

CO2

Microbial biomass

Inorganic flows Mineralisation

Sm

Old SOM

MP

Woody litter

Inorganic N

CO2

Sy

Foliar and fine root litter

CO2

CO2

Microbial biomass

CO2

b

79

Young SOM

Immobilisation

Young SOM

Organic flow Stabilisation/ humification

Old SOM

Fig. 4.4 Pools and fluxes of (a) carbon and (b) nitrogen in the new decomposition model. MP: metabolic pool; HCP: holocellulosic pool; LCP: ligno-cellulosic pool; L: lignin; SOM: soil organic matter; Sm: stabilisation coefficient for microbial biomass; Sy: stabilisation coefficient for young SOM (From Corbeels et al. 2005)

SOM into more recalcitrant forms (respectively into young and old SOM). Nitrogen is mineralized to, or immobilized from, the soil inorganic N pool to maintain the N:C ratio of decomposing microbial biomass within a specified range. Balancing potential microbial N demand against inorganic N availability determines whether the activity of decomposers is limited by N. If so, then simulated decomposition fluxes are reduced. The maximum rate of microbial N uptake is proportional to soil inorganic N content. Lignin incorporation in the young SOM pool results in additional N immobilization in the young SOM pool, which simulates the process of chemical N immobilization. SoilErosion: Water Runoff and Soil Erosion by Water The runoff-erosion component was developed by UNIMI. It simulates surface runoff and erosion, and handles irrigation events. The same soil description as in the SoilW component can be used. Runoff and erosion can be simulated daily when only daily rainfall is available, or for shorter time periods, if hourly or more frequent data are available. The following four processes are simulated, allowing for easy interchangeability and extension of options: • Interception of rain by vegetation; two approaches (von Hoyningen-Huene 1981; Brisson et al. 1988) are available, both of which calculate interception as a function of Leaf Area Index. • Interception of rain by mulch. • Runoff using either the Curve Number approach, which is suitable for daily rain data (SCS 1972) or the kinematic wave approach, when hourly or more detailed

80

M. Donatelli et al.

data are available. Infiltration of water in soil is simulated using either Smith and Parlange (1978) equation or Green and Ampt (1914) equation. Peak runoff is estimated with an empirical equation from EPIC (Williams et al. 1989), or is intrinsic in the Kinematic Wave approach. • Erosion is estimated using the MUSLE (Modified Universal Soil Loss Equation, Williams and Berndt 1977) which is suitable for single rain events. The EUROSEM (Morgan et al. 1998) and Kineros (Woolhiser et al. 1990) models use a differential dynamic balance between splash erosion, shear stress of runoff water, carrying capacity of runoff water and deposition. RUSLE (Revised Universal Soil Loss Equation) using the adaptation of Cooley (1980) for single storm events is currently being developed for SoilErosion. The SoilErosion component can also be used to simulate small hydrological basins because each simulation unit can accept as input runoff from an adjacent unit of simulation and can be either a plane or a channel. SoilNitrogen: Soil Nitrogen Dynamics This component was developed by UNIMI and is an implementation of SOILN (Johnsson et al. 1987), simulates the transformation of organic carbon and of organic and inorganic nitrogen in the soil. The model uses three pools to represent organic C and N: one is slow cycling (humus), and two are labile (litter and manure). Dead roots and incorporated crop residues are added to the litter pool, while animal faeces are added to the manure pool. Each input of organic matter is characterised by a specific N:C ratio and humification and ammonification coefficients, and is assigned to a “litter” or “manure” category. Inorganic N is represented by two pools, ammonium and nitrate. All transformations of C and N (except denitrification) are simulated with first-order kinetics, using environmental controls (soil temperature and water content) to modify decomposition rate constants. Denitrification is simulated with a zero-order kinetic. Potential decomposition of organic matter is simulated by calculating C flows from litter and manure to humus and from all pools to CO2. Soil microbial biomass is implicitly represented as part of the two labile pools, which therefore represent the association of added organic materials with their decomposers. The following sources and sinks of ammonium and nitrate are simulated by the component: urea hydrolysis, nitrification, denitrification, atmospheric deposition, nitrate leaching, crop uptake, and ammonia volatilisation. Although denitrification and ammonia volatilization are implemented following the strategy pattern, there are no alternative approaches currently available. SoilTemperature: Simulation of Temperature in the Soil Profile The soil temperature component was developed by UNIMI. It allows soil temperature to be simulated down a one dimensional profile. The following processes are simulated:

4 A Component-Based Framework for Simulating Agricultural Production

81

–– Surface temperature: The empirical approach of Parton (2004) and the mechanistic approach of Campbell (1985) based on the energy balance at the soil surface are both available. –– Transmission of heat in the profile: The Campbell (1985) approach based on a one-dimensional differential equation of heat transfer is available. The empirical approach of SWAT (Neitsch et al. 2002) is being developed. SoilReader: Accessing Soil Data at Initialization The SoilReader component was developed by CRA and UNIMI. It has four functions: 1. To load data (soil parameters, soil initial conditions, water table presence) 2. To estimate parameters which are either missing or which need to be estimated using pedo-transfer functions 3. To create soil layering from soil horizon data 4. To create daily values of water table depth The component uses the PedoTransferFunctions (PTF) component (Fig. 4.2) to make estimates both of soil hydrological properties and of soil parameters needed by soil water retention curve models from the available soil information. The PTF component is implemented using the same design as other APES dynamic components. SoilWater: Soil Water and Hydrologic Characteristics Dynamics The Soil water component was developed by UNIMI (Acutis et al. 2007). It allows one dimensional water redistribution in the soil to be simulated, and the changes in soil physical characteristic after a soil tillage operation. A soil profile is represented as a series of superimposed horizontal layers. For each layer, hydrological properties are provided by specifying the parameters of the appropriate hydraulic functions. Alternatively, the HYPRESS pedotransfer functions (Wosten et al. 1999) can be used to calculate hydrologic parameters from soil texture, bulk density, and SOM, or the PTF component (see SoilReader) that includes a large collection of Pedotransfer functions can be used to provide estimates. In addition, it is possible to provide the soil water contents corresponding to field capacity and wilting point. When, for numerical reasons (i.e. a finite difference approach for water dynamics simulation), a finer soil layer definition is needed, a method is available to split existing pedological horizons into thinner soil layers. The following processes are simulated, allowing for selection of alternate approaches: –– Soil water distribution: three approaches are available, an empirical cascading model, a cascading model with travel time taken allowing for water contents greater than field capacity but preserving the speed of calculation of the cascading method itself, and a finite difference solution of the Richards’ equation. –– Water evaporation: two approaches, CropSyst (Stockle et al. 2003) and Ritchie (1972) have been adopted.

82

M. Donatelli et al.

–– Water uptake by roots: five approaches have been used, CropSyst (Stockle et al. 2003), Ceres (Ritchie and Otter 1985), Swap (van Dam et al. 1997), EPIC (Williams et al. 1989), and “control” where water uptake is an external input, as a total amount or by layers. “Control” only checks if the external water requirement is consistent with the actual soil water content. –– Tillage: the WEPP approach (Alberts et al. 1995) is implemented. The tillage model simulates the effect of tillage and the successive soil settling using a database of over 80 implements. According to the characteristics of the different tillage implements, SoilW reacts to the “Tillage” event by: (i) redistributing soil particles if layers with different textural composition are included in the tillage depth; (ii) redistributing organic matter; (iii) burying crop residues; (iv) redistributing water; (v) calculating a new soil bulk density and consequent changes in layer thickness; (vi) changing the retention and conductivity functions, and field capacity and wilting point; (vii) simulating the time course of soil settling with its effect on soil characteristics according to the amount of rain and time elapsed after the tillage event.

SoilWater 2: Soil Water and Hydrologic Characteristics Dynamics The SoilWater2 component was developed by CIRAD and LIRMM. It mainly differs from SoilWater by taking account of preferential water flow through the soil profile. The model considers three structural levels within the soil profile, the pedostructure, the primary peds and the primary particles (Braudeau and Mohtar 2009). The clayey plasma of the primary peds, micro-porosity, and the inter-ped pore space, macro-porosity, are represented by two compartments which are in contact through a transitional zone at the surface of the primary peds. Water is distributed between these two compartments whose volume varies with water content. Water fluxes between compartments and from one pedostructure unit to the other result from an alteration in the hydrostatic equilibrium due to water supply (rain, irrigation) and evapotranspiration. Fluxes are simulated using Richards equations. The functionality of the pedostructure is quantitatively described by equations that originate in the measurement of four soil characteristic curves: the shrinkage curve, the swelling curve (Braudeau and Mohtar 2006), the conductivity curve, and the soil water potential curve (Braudeau 2006; Martin et al. 2006; Braudeau and Mohtar 2009). Those four equations are described using 15 parameters which can be estimated using specific pedotransfer functions gathered by Saxton and Rawls (2006) and available in the PTF component, thus allowing model runs to be carried out with the same soil information used by conventional soil-water models. The SoilWater 2 component simulates the dynamics and interactions of soil structure and soil water. The profile consists of a surface layer and underlying horizons. The impacts of technical practices like tillage, or the effect of a soil surface crust, are on water infiltration and evaporation. Surface hydraulic conductivity, layer thickness and maximum surface storage are the three principal factors that are modified.

4 A Component-Based Framework for Simulating Agricultural Production

83

Each horizon has a pedo-structure, a homogeneous zone in terms of structure and organization of particles. The soil is divided into homogeneous layers. The equations used allow the uniformity of the layer’s depth in each horizon and the differences between horizons to be represented (Braudeau and Mohtar 2009).

Weather Data Components WeatherReader: Accessing Weather Data and Estimating Missing Values The WeatherReader component was developed by CRA. It has two functions: (1) to estimate missing weather data, and (2) to provide access at run time to location and weather data. The component is associated with components of an application provided as a separate tool (CLIMA), which allow missing weather variables to be estimated and weather to be generated from a rich library of alternative models implemented in six categories: AirTemperature, Evapotranspiration, LeafWetness, SolarRadiation, Rainfall, and Wind. The CLIMA application is described in the following paragraphs. Some estimation capabilities from the CLIMA components which are encapsulated in the WeatherReader and are active at run time, estimate weather data which are never available from weather data records (e.g. vapour pressure deficit, day length, extraterrestrial radiation)

The Intended Use of APES Version 1.0 of APES allows rotations of crops and vineyards to be simulated for water- and nitrogen-limited conditions. The current modelling solutions allow one dimensional fluxes to be estimated. It is primarily a prototype for evaluating the adequacy of APES in terms of: –– The model framework, and in particular its ability to link operationally different model constructs within the simulation tool. Such an evaluation includes conceptual evaluation (criteria and needs for combining models implemented in discrete units), and a technical evaluation (adequacy of the modelling framework for linking components). –– The Graphical User Interfaces. This evaluation involves seeking feed-back from APES users on the stand-alone application. Users of the SEAMLESS integrated framework with a biophysical background are also able to test APES using a specialized user interface. The system allows water, nitrogen, and pesticide dynamics to be simulated at the field scale in response to agro-management in the range of environments (soil-weather combinations) characteristic of the agricultural parts of Europe. The choice of spatial scale has been a direct consequence of the goals of the

84

M. Donatelli et al.

simulation, namely to estimate production and system externalities in response to detailed agricultural management applied in specific soil-weather combinations. Modelling approaches selected and implemented in APES were mostly developed at field scale. Simulation outputs at this scale have been used in the literature to provide outputs at the regional scale by linking to Geographical Information Systems holding information on the spatial distribution of soils and weather. In such cases the most frequent recommendation is to use simulation outputs to make relative comparisons between different agro-management options. Other options are to use simulation outputs at the field scale as “cell” data to be integrated in spatially explicit models, as in some catchment models. In this case, the increased number of inputs needed generally limits the use of these models to case studies. All uses at scales other than the field scale involve additional assumptions that may be difficult to justify. Moving across scales is being addressed in SEAMLESS with specific actions, but it is outside the modelling domain of APES. The optimum temporal scale is still a matter of debate, as opinions differ about the significance of possible drift in multi-year simulations without the re-initialization of state variables. However, this use is both a given and implicit in the simulation of multi-year crop rotations and is accepted in peer-reviewed publications on the use of tools like APES. In any case, the issue is not about APES itself, but about all model tools built with modelling approaches similar to APES. As the simulation tool has been developed with a focus on modularity, APES versions including different modelling engines and components (modelling solutions) can be made available as “closed” modelling solutions to be used for situations where the assumptions made by their developers (modellers) apply. A set of options may be made accessible (e.g. to simulate reference evapotranspiration using different approaches), but in order to protect system integrity, APES users will not be able to access model composition (in their role of model users). However, APES is an open system so that the same individual, with a different role, may access model building, in this case taking the responsibility for the choices made. This is the expected use beyond the end of the SEAMLESS project. Simulations can be run using long series of either generated or observational weather data, to account for the stochastic variability of weather. Outputs can be evaluated as means and deriving measures of variability.

Inputs Whenever alternative options are available to simulate a given process and such options perform almost equally well, the less demanding model in terms of parameters and inputs should be selected. However, data availability and quality cannot be allowed to limit the implementation of model capabilities when it prevents the achievement of the goals of SEAMLESS. APES releases minimize data requirements and use options, such as pedo-transfer functions and weather generators, to estimate missing variables and parameters from the available data.

4 A Component-Based Framework for Simulating Agricultural Production

85

The operational use of APES will highlight data gaps that need to be filled if it is to be used throughout the EU (both in absolute terms and in specific areas/countries). Such an analysis can be considered a side-product of the APES development action, but of specific importance. For instance, the formalization of agro-management using a common framework for current agricultural practices is a need which goes beyond its use in APES. Current APES inputs can be grouped in six file types (the detail is provided in the documentation): • Site data (e.g. latitude, elevation, …) • Daily weather • Soil data (e.g. clay, silt, and organic carbon percentages, horizon thickness) by soil horizon; slope, field length – unique values • Soil initialization data (initial conditions for state variables; if missing, default values are used) • Soil water table (if missing, the assumption is that no water table affects the root zone) • Planned agro-management (see AgroManagement component) Site and weather data are loaded at run-time from the WeatherReader component, which allows missing data to be estimated using CLIMA. Soil data are loaded at run time from the SoilReader component, which allows missing data and hydraulic parameters to be estimated using the PedoTransferFunctions component. Agro-management has proved to be the most challenging problem, because the lack of a common formalism to store information beyond its use in agricultural statistics (i.e. a static, summary description) makes it difficult to develop rules to simulate the dynamic part of farmers’ decision making processes based on bio-physical drivers. APES development, and specifically the AgroManagement component has, however, provided a framework to formalize such data, making them of use for simulation of current and alternative agricultural management at field level.

Parameters Parameters are defined as quantities that do not change value during either the whole simulation or parts of it. For example, crop parameters change when a crop is changed during the simulation, but their values do not change during the time a given crop is simulated. A simulation system which allows the use of more than one modelling approach cannot define a constant set of parameters for two reasons. First, some simulation approaches may need to model a parameter which then becomes a variable. For instance, a simulation system which does not model impact of tillage on soil physical properties will probably consider soil bulk density as a parameter; whereas bulk density will be modelled as a variable if the goal of simulation is to estimate the impact of tillage on soil hydrological characteristics. Secondly,

86

M. Donatelli et al.

modelling approaches may differ in the number and type of parameters required. Model components encapsulate the knowledge of their own parameters and handle them in response to either initialization or agro-management events. Sets of parameters can be edited via a dynamic, generic parameter editor (see the paragraph about the application Model Parameter Editor). Default values are part of the definition of the parameter together with minimum and maximum values, as is the case for variables.

Software Architecture One critical issue in model linking to assemble model components into a composite model, is the difficulty of finding a component design that satisfies the requirement of ‘third-party composition’. In order to integrate my work with yours, my component must be compatible with your component, but frequently this is not the case: components are designed for a specific architecture or framework, and they are not usable outside it. Component design choices, rather than being peculiar to a specific architecture, should promote re-usability by including design traits which represent a compromise between level re-usability and complexity of the design chosen to maximize adaptability of components. Using a pragmatic approach, simplification can be obtained if the target use of components is within a specific knowledge domain. This has an impact not only on simplifying the design of components, but on clearly defining the scope of the knowledge domain which is embedded in the modelling exercise. Yet, restricting to a particular knowledge domain has often also meant restricting to a particular framework, where implementations of model components strongly depend on the modelling framework core. Targeting model component design to match a specific interface requested by a modelling framework decreases its re-usability. This explains why modelling frameworks, although in theory a great advance with respect to traditional model code development, are rarely adopted by groups other than the ones developing them (Rizzoli et al. 2005). One way to overcome this problem is to adopt a component design which targets intrinsic re-usability and interchangeability of model components (e.g. Carlini et al., 2006; Donatelli et al., 2009; Donatelli and Rizzoli 2008). Such components can be used in a specific modelling framework by encapsulating them using dedicated classes called “adapters”. Such classes act as bridges between the framework and the component interface. The disadvantage of this solution is the creation of another “layer” in the implementation, which adds to the already implemented machinery in the framework. However, if components are correctly designed, there is a negligible, if any, penalty in performance, and the adapter does not add complexity. A further argument in favour of framework-independent components is that they allow modelling knowledge to be shared in a form which makes it easily re-usable. Model components developed in SEAMLESS for use in APES are based on the above design paradigm: framework-independent components which can be linked to different modelling frameworks. A proof-of-concept of this claim has been

4 A Component-Based Framework for Simulating Agricultural Production

87

shown by Argent and Rizzoli (2004). MODCOM (Hillyer et al. 2003) is the linking framework used in the current release of APES.

Component Design Components are discrete software units used for composition. Hence, components cannot be used in isolation. Further, the reason for adopting a component-based paradigm for implementing models as computer programs is to achieve specific functionalities not available with monolithic structures. Consequently, the component architecture and its implementation are crucial in developing a component based system for biophysical simulations that enables solutions to biophysical modelling problems to be implemented using different designs and technologies. Developing a design for a component architecture and selecting a technology for its implementation should be the result of a careful definition of requirements. The following requirements were defined for the APES model components: Functional requirements: –– Estimation and generation of variables from different models –– Estimation of parameters from observational data –– Provision of data at run time, accessing either observational or generated data, and making available model outputs –– Provision of quality checks on imported data –– Provision of quality checks on outputs –– Robust behaviour of the component that degrades gracefully, raising appropriate exceptions –– Traceable component behaviour with traces that are scaleable, i.e. browsable at different debug levels Non-functional requirements: –– Ease of use: the components must be easily usable by clients; this has an impact on technology and on documentation. –– Extensibility: the capability to easily add alternative processing capabilities to the ones of the component from the side of the component user, without needing to recompile the component, and using the same interface. –– Re-usability: the practical possibility of using the component in different software systems; ease of use and solution to a common modelling problem are the pre-requisites. –– Replaceability: the capability for components to be replaced by a different component conforming to the same specification. “Different” here means either a newer version of the same component, or an implementation from a different party. –– Availability of fit-for-purpose documentation of models, software design, and code. –– Successful unit tests: unit tests for each public method, input-output tests should be reported in the documentation.

88

M. Donatelli et al.

Technological requirements: –– The component software implementation must be made using technologies with a widespread adoption. There are several ways of matching the above requirements via software design and implementation. In the next sections we list the major choices made for the design and implementation of static and dynamic components developed for APES.

Ontology A pre-condition of the successful integration and re-use of an existing component is that the modeller understands the meaning of the data elaborated by the model. Understanding means associating a variable with a shared concept, knowing its spatio-temporal extent, its dimension and units and so on. In order to facilitate this, the components must contain information, possibly extracted from a public ontology, which describes the variables/parameters used, and allows checks on input-output links and data quality tests at run time. Information consists of concepts (variables in this case, which can be seen as instances of the concepts) and of attributes for each variable, encapsulated using the VarInfo type implemented in a utility component. VarInfo attributes are: Name, Description, Minimum/Maximum/Default value, Units (Athanasiadis et al. 2006). The properties with respect to data flow are not included among these attributes as they are not an intrinsic attribute of the variable. In other words, a variable can be an input to one model, and an output from another. This information from VarInfo is used in the domain classes described below. The components also contain internal information about parameters and variables, using the same VarInfo type. Such information is defined in the component and used as described in the section on pre- and post-conditions. Collections of variables that are associated with particular domains define the Domain Classes (Del Furia et al. 1995). For instance, we could define SoilStates as the collection of all measurements relevant to a specific goal chosen for soil modelling. Such collections can be manually entered by a user or automatically created, using the built-in reasoning features of the ontology. The definition of domain classes in the component interface allows the dependency of the model to be abstracted from the data and the extensibility of models fostered using design patterns (Mesketer 2004; Bishop 2008). The importance of domain classes goes beyond their meaning as software implementation items. In fact they provide a detailed description of the domain of interest. Using domain classes, a modeller can exploit the knowledge structured in the ontology in different modelling frameworks and programming languages. The adoption of an ontology-driven approach for defining a model interface has clear advantages as it enables the reusability of models in an easier way, while common problems related to poor semantics of model interfaces can be effectively tackled. The APES ontology is browsable on the web at a dedicated page (see Web resources) in which each domain class and strategy is described using the VarInfo values of their variables and parameters.

4 A Component-Based Framework for Simulating Agricultural Production

89

Model Granularity in Components One definition of a model is that it is a conceptualization of a process. From a software point of view a model can be implemented in a class that offers a method for computing a variable (or a set of interrelated variables), thus obtaining the desired level of granularity. There might be more than one way to do this. If two different models estimate the same variable A, we can implement both of them as alternative methods for estimating variable A even if they have different input requirements and different parameters. To do this, the two models must be made available as separate units, and their input, parameters and output must be defined. Such units are here called “strategies”, from the related design pattern introduced below. A set of alternative models can be implemented in the same component using the design patterns Strategy and Composite. These designs offer the component user alternative algorithms (strategies) for doing the same thing. When building a biophysical model component, this allows in principle alternative options to be offered for estimating a variable or, more generally, for modelling a process. This often-needed feature in the implementation of biophysical models comes with two very welcome benefits on the software side: (1) it allows easier maintenance of the component, by facilitating the addition of other algorithms, (2) it allows the easy addition of further algorithms from the client side, without the need to recompile the component, while keeping the same interface and the same method signature. In summary, the strategy (a model class) encapsulates a model, the ontology of its parameters and the test of its pre- and post-conditions. It can be used either directly as a strategy (in this case we call it “simple strategy”, where simple indicates that does not use other strategies as part of its implementation), or it can be used as a unit of composition. A composite strategy differs from a simple strategy because it needs other (simple) strategies to provide its output(s). A sequence of calls might be implemented inside a composite class. The list of inputs includes all the inputs of all classes involved (except those which are matched internally). The list of outputs includes all outputs produced by each strategy and the ones specific to the composite class (if any). The list of parameters needed includes those of the classes associated with and the ones (if any) defined in the composite class. When the value of a parameter is set, if the parameter belongs to an associated class, it is set in that class. The test of pre- and post-conditions (see following paragraph) makes use of the methods available in each associated simple strategy class, plus the new tests specified in the composite class. If a violation of pre- and post-conditions occurs in one of the associated classes, the message informs the user not only about the violation that has occurred, but also in what class it has occurred. Composite strategies do not differ in their use compared to simple strategies. A different type of composite strategy is the context strategy. Such classes implement an internal model to select the appropriate strategy (either simple or composite) based on the context, that is, on the inputs received at each call. An example of a context strategy is the one which estimates reference evapotranspiration by the Penman-Monteith, Priestley-Taylor, or Hargreaves method depending on the inputs available.

90

M. Donatelli et al. IDomainClass

ComponentAPI

property + Description() : char + URL() : char + StrategyUsed() : List

+ Estimate(IStrategyComponent, DomainClass*) : void + Estimate(IStrategyComponent, DomainClass*, char, boolean) : void + Info() : void

1..* «interface» IStrategy

«interface» IStrategyComponent

1..* DomainClass 1..* Instances of domain classes are input-ouput objects to be used in the modelling of the domain

+ + + + +

Estimate(DomainClass*) : void TestPreconditions(DomainClass*) : void TestPostconditions(DomainClass*) : void ResetOutput() : void SetParamatersDefault() : void

2..* Context classes encapsulate logic to select among strategies and composite strategies

CompositeStrategy

Strategy

ContextStrategy

property + TimeStep() : List + IsContext() : boolean + Description() : char + URL() : char

1..* 2..*

Fig. 4.5 Class diagram of a generic component according to the design used in APES components

The sample class diagram of a generic component in Fig. 4.5 shows the different type of strategies and the common interface which allows the design patterns described above to be implemented. The common interface IStrategyComponent and the capability to inherit from DomainClasses to extend them allows extending the component independently by third parties, and still using the original ComponentAPI. The interface used for models is the same for all modelling solutions in the component and can be seen as an implementation of the Façade design pattern (Bishop 2008) to hide the complexity of model solutions based on composite strategies. This leads to their being a single signature for internal and extended models. An example of simple and composite strategies is given by Villa et al. (2006). Composite strategies too can be added to the components without requiring a recompilation of their code, thus providing a way to extend component models fully autonomously by third parties. Composite strategies are solutions to modelling problems at a coarser granularity (in principle) than that of simple strategies. As an example, a composite strategy may be built to simulate “crop potential production” and be developed by putting together simple strategies such as “light interception”, “crop development”, “leaf area expansion”, etc. In other terms, a composite strategy is a “closed” solution which makes use of selected models of finer granularity as units of composition. Such a closed solution is not proposed as the unique solution for a specific modelling problem in a component as other options can co-exist or be added by third parties. Referring to the example above, two composite strategies may use different simple strategies to simulate “light interception” depending on whether they target the simulation of homogeneous canopies or wide-row crops. Even though such

4 A Component-Based Framework for Simulating Agricultural Production

91

differences in light interception models may not lead to noticeable differences in simulated potential yield, they may lead to sharp differences when simulating waterlimited production, particularly in arid environments. Further, two alternative approaches to modelling light interception, say for “homogeneous canopies”, could be implemented as two composite strategies, and this would allow modelling approaches to be compared at finer granularity. This kind of composite model provides a sound foundation for selecting modelling approaches to be used operationally. Formalizing models in basic units of composition (simple strategies) and in aggregated units (composite strategies), with the same interface, and decoupling interfaces and data from modelling equations provides the design infrastructure to link and populate a knowledge base. The use of semantically rich interfaces fosters safe re-usability of components as discussed in the introduction. Finally, both simple and composite strategies are discrete units of code which can be used either to build components, or even as “full” simulation models to be used in stand-alone mode, in the latter case still preserving the benefits of a modular system as described in the introduction. By imposing the same interface on simple, composite and context strategies, the components obtain the full benefits of the Strategy and Composite design patterns, making re-use simpler and allowing full extensibility.

Test of Pre- and Post-conditions Implementing tests of pre- and post-conditions is the central idea of the Design-byContract approach (Meyer 1991). In DBC software, entities have obligations to other entities based upon formalized rules between them. A functional specification, or ‘contract’, is created for each module. Program execution is then viewed as the interaction between the various modules bound by these contracts. In general, routines have explicit pre-conditions that the caller must satisfy before calling them, and explicit post-conditions that describe the conditions that the routine will guarantee to be true after the routine finishes. When building biophysical models, the DBC approach not only ensures the correct functionality of the software, but also specifies the limits of valid use of the model, which is knowledge about the model itself. It also allows data of uncertain quality to be used: if an input (either an exogenous variable or the output of another component) is out of the expected range, an exception can be raised, both informing the user of the problem and allowing for exception handling. The DBC approach is implemented in APES via a utility component developed for the purpose, called Preconditions.

Exception Handling Exception handling is a programming language construct designed to handle runtime errors or other problems (exceptions) which occur during the execution of a computer program. Handling exceptions is of crucial importance in a component

92

M. Donatelli et al.

based system as it allows users of the applications and subsystems using the components to know precisely the source of the error and thus to choose what to do in response, preventing hard-crashes of the whole application. Components raise and propagate exceptions and provide a customised message informing users which component and class are the source of the error. If an unhandled exception occurs, an informative message describes the error, and model and component source of the exception, allowing for continuing execution of the client if the user chooses.

Tracing The traceability of component behaviour is implemented in.NET versions using the TraceSource class in an implementation that allows the client to set receivers of the messages called listeners. Various levels of tracing (critical, error, warning, information, and verbose) can be pooled in one or more listeners with all the traces from other components and from the client. Traceability is used in components to provide a log of execution shown at run-time in the APES stand-alone application.

Unit Tests In computer programming, a unit test is a procedure used to verify that a particular module of source code is working properly. The principle underlying unit tests is to write test cases for all functions and methods so that whenever a change causes a deterioration, the cause can be quickly identified and fixed. The goal of unit testing is to isolate each part of the program and show that the individual parts are correct. Unit testing provides a strict, written contract that the piece of code must satisfy. Beyond the general benefits which derive from unit test implementation in software development, implementing unit tests to test model implementation and make available the relevant input-outputs in the documentation allows the user of the components to have sample application results for the specific model.

Model and Software Design and User Documentation Each component requires a help file which contains detailed documentation about the models implemented, and information about the design and use of the component. The test of documentation adequacy is that it should allow re-implementation of all the models that comprise the component, although the characteristics of re-usability of the component make it much easier to use it again rather than to duplicate it. The code of each component should also be thoroughly documented, so that

4 A Component-Based Framework for Simulating Agricultural Production

93

automated documentation processing utilities, such as those available for the .NET development environment, can generate the documentation reports.

Component Public Interface One of the key elements for component adoption by third parties is simplicity of default usage cases via the Application Programming Interface (API). The usage model for component-oriented design follows a pattern of instantiating a type (a class) with a default or relatively simple constructor, setting some instance properties, and finally calling some simple instance method. This “Create-Set-Call” pattern (Cwalina and Abrams 2006) has been implemented in the APES components. The component API contains one method for each of the strategies using the same signature (see the paragraph Model granularity in components). Each of the strategies uses the same signature. Domain classes and strategy inputs, parameters, and outputs can be found using the Model Component Explorer application described below.

Technology Used The technology used is based on the Object Oriented Programming (OOP) paradigm as implemented in the Microsoft .NET 2.0 framework. However, the object model of.NET allows easy migration to the Sun Java platform. Such migration has been realized for some of the components referenced. Most of the components have been made available as discrete units inclusive of a software development kit with example projects in which to use the component by desktop clients, by web services and applications, so that components can be extended independently of their source code.

Model Component Diagram Each of the model components represented in Fig. 4.2 is actually a package of discrete units. Figure 4.6 shows the component diagram of a generic APES model component. If a component implements models to simulate agro-management, then it must have a dependency on the components CRA.AgroManagement and CRA. AgroManagement.Impacts. A component may have dependencies also on other components, such as numerical or statistical libraries, but must have no dependency on any modelling framework. Descriptions of model architecture are available in

IVarlnfoClass

IStrategy

IRule

ModelComponent

IManagement

CRA.Core.Preconditions

ModelComponent.Interfaces

«XML» Parameters

IDomainClass

M. Donatelli et al. IStrategyComponent

94

CRA.AgroManagement CRA.AgroManagement.Impacts IManagementPruning

IManagementFertilizers

IManagementPesticides

IManagementManure

IManagementNitrogen

IManagementCrop

IManagementTillage

IManagementIrrigation

Fig. 4.6 Model component diagram. The MODCOM adapter component is not represented in the diagram (see Fig. 4.7)

the documentation of each component. Note that in APES, model components are encapsulated in an adapter class inheriting from a MODCOM class so that the MODCOM framework can be used, as described in the following section.

The MODCOM Engine Components exchange data via the modelling-framework MODCOM (Van Evert and Lamaker 2007), which was developed by PRI. MODCOM is a software framework that facilitates the assembly of simulation models from previously and independently developed component models. It offers connectivity, time and state events, and numerical integration. APES components are registered via adapters to the MODCOM application which serves as the model engine, implementing the Adapter pattern as shown in Fig. 4.7.

4 A Component-Based Framework for Simulating Agricultural Production

«library» Framework

95

«library» Framework Application

Framework Component Framework Independent Component ComponentAdapter «library» Component

Fig. 4.7 Linkage of a generic component via the Adapter pattern in APES applications using MODCOM

Data Exchange Between Components Data are exchanged at a time step of 1 day, but within each time step components can communicate up to three times. Within a time step, components can communicate, for instance to meet supply and demand computed by two different components, thus allowing estimation of actual rates besides the potential ones which do not need to match supply (e.g. for water and nitrogen). The multiple calls within the time step also permit arbitration of a source between two or more sinks. Details are provided in the MODCOM documentation. The fine granularity, strategy-based discretization of models has been shown to be adequate for accommodating multiple calls for different purposes within a time step, and allowing each component to be developed without any dependency on other components or on the modelling framework itself.

The APES Stand-Alone Application The APES stand-alone application allows different instances of the modelling engine to be run using the following conventions: –– APES modelling solutions using alternative components (e.g. SoilWater or SoilWater2) are kept separate and can be loaded by the user. –– Once an APES modelling solution is loaded, a user interface page is dynamically built allowing modelling options to be set within components (e.g. potential growth or water limited growth options in the Crop component; the curve number or the kinematic wave approach for runoff).

96

M. Donatelli et al. APES plug-ins

APES Graphical User Interface

APES Models

Plug-in

Model components

Plug-in

CLIMA Weather Generator application and components «component» Data Quality

«component» Data IO

«application» Graphica Data Display

«application» Simulation Output Evaluator

«application» AgroManagement Configuration Generator

«component» IRENE

«component» AgroManagement Files Display

«application» Model Parameter Editor

«application» Model Component Explorer

NMath Statistical libraries

Fig. 4.8 The main components and packages of the APES stand-alone application. Connections represent dependencies. The package “APES Models” represents different realizations (modelling solutions) developed using the items of Fig. 4.2. All application tools (third row from the top) are reusable independently by third parties in custom developed applications. The component “Data Quality” is the component Preconditions cited in the paragraph Ontology

The APES stand-alone allows new applications to be added as supporting tools (plug-ins) independently by third parties. An example is provided inclusive of source code. The application is also built using the component-oriented paradigm; single components can be re-used independently of APES. Figure 4.8 shows the main components of the December 2008 version.

The AgroManagement Configuration Generator The AgroManagement Configuration Generator is an application developed by CRA to build XML files containing a set of planned agricultural management actions, and to visualize configurations. Such files can be used by the CRA. AgroManagement component at run-time to simulate the decision making process to implement agro-management actions in a field. As described in the documentation for the AgroManagement component, agro-management is simulated by parameterizing rules (classes implementing the interface IRule)

4 A Component-Based Framework for Simulating Agricultural Production

97

and testing them against the state of the system. If a rule is satisfied, the coupled impact (a class implementing the interface IManagement) is published. The ACG uses rules and impacts to build sets of rule-impact couples. ACG allows also files to be merged and uses the component CRA.AgroManagement.AFD to display planned agro-management actions using a graphical metaphor (see Fig. 4.3).

The Model Parameter Editor Developing and maintaining a simulation system implies, among other things, that the parameters used can change. Composite models are made of simpler models, which can be often replaced by alternative formulations. This means that the development and management of a simulation system requires the ability to deal with changing the number and type of the parameters of the composite model when a sub-model is substituted. If the system consists of interchangeable components, the need for dealing with different sets of parameters is an inherent feature of the system; an alternative component may model the same domain variables, but its approaches may demand different, model-specific, parameters. The need for changing parameters has a primary impact on the Graphical User Interfaces developed for the system: such user interfaces must be easily maintainable, and ideally present the same look and feel to the user when different sets of para meters are in use. Moreover, there should be a facility to check the accuracy of all parameter values. A parameter editor with these features must allow the parameters to be edited to be changed without changing the code, hence without a need for re-compilation of the editor. The Model Parameter Editor (MPE) is an application developed by CRA (Di Guardo et al., 2007) which allows a dedicated user interface to be generated for each available parameter definition. It groups interfaces in different tabs either according to user criteria, or according to the model components which originate the parameter definitions. The application allows selection of parameter definitions, or it loads automatically parameter definitions from a folder of choice. MPE can be used as a stand-alone application, but it is meant to be used primarily in a simulation system like APES, becoming one object of the Graphical User Interface.

The Graphic Data Display Component Providing data views from Graphical User Interfaces is a common need for applications built to make use of models. If model output is generated by a modular system in which model components are interchangeable, output variables may change. Thus, maintaining GUIs can be challenging and resource demanding. A tool which can load datasets with various schemas and which helps the user to visualize data in a range of ways speeds up application development, allowing the user to focus

98

M. Donatelli et al.

on the models, rather than on the user interfaces. Whenever flexibility of use is important, providing domain specific views of data adds value to such a tool, both in operational usage and in model development. The component GDD (Graphic Data Display) is a Microsoft .NET component developed by CRA (Di Guardo et al., 2007), which has the specific purpose of retrieving a set of output variables and allowing values to be displayed either as textual tables or as graphs. GDD can be used as a stand-alone tool or as a component inside an application. In the former case it provides access to a file dialogue that allows the user to select a file, whereas in the latter case it can be opened inside a modelling framework to directly load the current dataset. GDD accepts inputs in three different formats: XML, MS Excel, and the more compact and faster binary form (another available component also allows I/O operations with the binary format). Readers can however be extended by third parties implementing the proper interface. Each variable can be either a table column, or an entire table of the dataset, depending on whether it is either only time-variant or time and one-dimensional space-variant (the latter are variables that vary down soil profiles, such as soil temperature). GDD has seven tab pages supporting data views such as tabular views (which can be saved using the Microsoft Excel format), scatter graphs, time courses, histograms, soil profiles (water, temperature, nitrates, agrochemicals), ‘Micale’ graphics, frequency histograms, and probability of excedence. Also, GDD allows showing reference data against simulation outputs via configurations which can be saved. GDD can read APES GUI output files, in both XML and binary formats.

The Simulation Output Evaluator The Simulation Output Evaluator (SOE) is a data analysis tool developed by CRA that provides easy access to model evaluation techniques. As the literature gives neither a standard theory on model evaluation, nor a standard “box of tools”, the emphasis is on statistical techniques for comparing estimates either with actual measurements, or two series of estimates, making use of an extensible library called IRENE (the.NET 2.0 version). Non-replicated estimates are mostly compared with the non-replicated measurements. The program also allows comparison of individual estimates with replicated measurements (or vice versa) and replicated estimates with replicated measurements. The program provides extensive statistical capabilities with tools for a variety of needs. Ready-to-use procedures handle a wide range of statistical indices and test statistics. Basic statistics allow a preliminary check of data quality. The evaluation of model performance is based on either the model residuals or on the correlation coefficient. In addition, model evaluation by probability distributions (i.e., probability of excedence), residual analysis (i.e., pattern indices), and fuzzy-based aggregation statistics are allowed both for indices produced internally by the component and for external numerical values. The fuzzy aggregation model is saved as an XML file. Graphics are included in most analytical

4 A Component-Based Framework for Simulating Agricultural Production

99

tasks and the user can request many types of graphs directly. SOE can read APES GUI output files, both in XML and binary format.

The CLIMA Weather Generator CLIMA is a Windows application developed by CRA (Donatelli et al. 2009b) which generates and estimates daily and hourly values of weather and weather-related variables using several alternative models (Carlini et al. 2006; Donatelli et al. 2006a and 2006b; Donatelli et al. 2009). It exports data in a format readable by APES, and custom writers can be added. CLIMA was built following the component-oriented paradigm and it is an example of re-use of components developed for APES. The component architecture is the same as other APES components, hence allowing both for component extensibility autonomously by third parties and for their re-use in custom developed applications (Fig. 4.9). CLIMA allows for composition of models into a composite model which is then saved as a discrete DLL. Such DLLs can be used either for data generation within CLIMA, or separately in a custom application.

«» «component» Data IO

«application» Graphic Data Display

LeafWetness

«component» EvapoTranspiration

«application» CLIMA «component» AirTemperature «component» Climlndices

«component» ClimParameters

«component» ClimTimeSeries

«component» Rain

«component» Data quality

«component» Wind

«component» SolarRadiation

Fig. 4.9 CLIMA component diagram (main components shown)

«component» NMath

100

M. Donatelli et al.

The Model Component Explorer The Model Component Explorer (MCE) is a Windows application developed by CRA for inspecting model components to reveal interfaces, domain classes and VarInfo values for each variable, simple and composite strategies and their parameters, inputs, outputs, and associated strategies. Given the component architecture implemented, the tool can be used to show the input and output variables within the domain classes of any model. The tool also allows XML files to be saved with schemas for each domain class and strategy, allowing them to be uploaded into a shared ontology (see paragraph Web resources).

APES Tools for Integration in Broader Modelling Systems Given its component-based structure, APES can be run not only from a user interface, but also using a command line procedure. This allows the model to also be called from applications developed using languages which have no binary compatibility with .NET (e.g. Java) provided these applications are on a machine running Windows. APES is used in the software system SeamFrame (Wien et al., this volume) as an external component. As SeamFrame is implemented in Java, it requires APES to run as executable files. Support for integration can also be provided if the application which acts like a client is a web application, provided that such an application exposes web services and includes rich clients. In this case, some of the applications implementing a user interface can also be used. APES includes tools for integration.

The Parameter Estimator Parameter estimation is a major aspect of crop modelling. Together with the functional forms of the equations, it is a major determinant of prediction quality. Parameter estimation is a difficult and time-consuming exercise and requires expertise not always available in a modelling project. The purpose of the Parameter Estimator (PE), developed by PRI and INRA, is to provide software to automate model parameter estimation. The Parameter Estimator consists of functions in the R statistical computing the language (R Development Core Team 2007). Models such as APES are coupled to the PE through specific functions. APES is written in the C# computer language and is, for the purposes of the PE, exposed as a Microsoft COM object. This makes it possible to use the R-to-COM bridge (Baier 2007) to execute an APES simulation run from within R statements. The R software requires the following information: the observed data, the name of the model to be run, the paths to the input files for the contexts of the data (for example climate, soil and management files for each context), the list of model parameters and indicators as to which are to be estimated and finally information related to the correspondence between the data and the model output. The R routine sets the parameter values of APES to the current values at each iteration, executes the model for each context,

4 A Component-Based Framework for Simulating Agricultural Production

101

retrieves the results, calculates the objective function to be minimized and determines the parameter values for the next iteration. The current version of the PE has five different algorithms for parameter estimation. The different parameter estimation methods correspond to different hypotheses about model error.

The Sensitivity Analysis Component Sensitivity analysis (SA) is a fundamental tool in the building, use and understanding of all types of mathematical model. SA provides information regarding the behaviour of the underlying simulated system. This information ranges from the identification of relevant model factors (parameters or inputs) to model reduction or simplification, better understanding of the model structure for given components of a system, model quality assurance, and model building in general. Among the most commonly used methods, it is possible to identify three classes: screening methods, regression-based methods, variance-based methods. The most used screening method is the one proposed by Morris, which is particularly effective in identifying the few important factors in models with many factors or with high computational requirements. The second class includes the regression methods, which are based on the computation of standard or partial regression coefficients quantifying the effects due to a change in a factor value while the others are kept constant. Within this class, different methods can be used to generate the sample of factors combinations necessary to obtain the model evaluations and therefore to calculate the regression coefficients; here, Latin Hypercube Sampling (LHS), Random, and Quasi-Random LpTau will be used. The last class, variance-based methods, includes the Fourier Amplitude Sensitivity Test (FAST), its evolution Extended FAST (E-FAST), and the method of Sobol’. All the methods belonging to this class compute total sensitivity indices for first and higher order effects and are demanding in terms of computational time because of the high number of model evaluations needed for each model factor. SensitivityAnalysis is a component developed by JRC and CRA (Donatelli et al., 2009c) with the goal of making available the sensitivity analysis models implemented in the SimLab library (Saltelli et al. 2004) via a user friendly application programming interface, in the memory managed environment of the Microsoft. NET platform (the Simlab library is available for C, C++, Matlab and Fortran). The component allows sensitivity analysis to be run on a model of choice using the methods mentioned above. It is implemented using C# under the Microsoft .NET v 3.0 platform. Sample applications inclusive of source code are provided to allow an easy start to SensitivityAnalysis use via different software clients.

Remarks on APES Integration in Larger Systems APES integration, although technically possible at even closer levels than the ones used in the integration into the SEAMLESS integrated framework should, however, be approached with caution, providing users means to access and verify results of

102

M. Donatelli et al.

any operation involving simulations, such as simulation per se or finalizing parameter calibration. This is not because of the component based structure of APES, but rather because complex systems are being simulated with process-based models. Automated optimization procedures in model chains may produce results which are meaningless in biophysical terms. Such simulation anomalies due to either inappropriate input data or even the misuse of the simulation model might be evident working with the biophysical system simulation alone, but a misuse of APES (or of any other process based system) in model chains may be very difficult to spot and could have an unpredictable impact on final results of the analysis. When included in a model chain, it is advisable to link APES simulations to other models asynchronously, allowing for simulation results to be evaluated by an analyst prior to using APES outputs as inputs for further processing. The paragraph above is not meant to suggest that a complex simulation system should not be integrated in model chains such as the SEAMLESS integrated framework. Instead, it is meant to stress the importance of implementing procedures to facilitate the evaluation of intermediate results, both by domain experts and via specific utilities.

Concluding Remarks APES development represents a paradigm shift for two reasons. First, APES is not proposed as “the” model. Instead it stresses the need for broadening modelling approaches and for comparing them at a finer granularity than for whole simulation systems. Secondly, compared to the first modelling frameworks for overcoming the problems of monolithic models, APES moves the focus onto components and their re-usability outside APES itself, even as stand-alone entities. A somewhat surprising result emerged during the initial development of APES. Contrary to past experience when implementation of complex systems has often been the most challenging task, the major difficulty has turned out to be thinking in “modular”, “multiple choices”, “transparent” modelling terms. The goal of making models available as discrete, re-usable units aimed at including ideally one process in each basic model unit has forced us to thoroughly analyze assumptions and the independence of each model from others. In fact, developing model components, even with the requirements listed in the previous paragraphs, is a modest challenge in terms of implementation, but it forces us to formalize modelling knowledge and the problem of model linkage and re-use. Technology is expected to move more and more towards declarative modelling in an operational way. The work carried out in creating fine-granularity, discrete model units, encapsulating a semantically-rich description of interfaces, has involved discussing and advancing understanding of many aspects of model assumptions and structures, and will be of great help in that direction. APES development during the SEAMLESS project has led to an increasing opportunity to concentrate on modelling options by re-using expertise in different domains. APES is offered as a complete simulation tool, but also, and of no less

4 A Component-Based Framework for Simulating Agricultural Production

103

importance, as a loose collection of model objects which allow the modelling knowledge that APES teams have assembled to be shared operationally. Utilities and applications are also available as independent objects for re-use. A third party may want to use a single component or an extended set of them. They will be fully documented and extensible so they can be easily used in custom-developed applications.

Web Resources The APES portal: http://www.apesimulator.org Component and applications documentation pages: http://www.apesimulator.org/ help/ APES ontology: http://www.apesimulator.org/OntologyBrowser.aspx SEAMLESS EU integrated project: http://www.seamless-ip.org, partly continued in the SEAMLESS Association (www.seamlessassociation.org) Acknowledgements The development of APES was partially funded by the EU – DG Research, Sixth Framework Research Programme, Integrated Project SEAMLESS (http://www.seamless-ip.org), contract no. 010036-2

References Acutis, M., Confalonieri, R., Donatelli, M., & Rana, G. (2008). Modellazione dell’allettamento dei cereali a paglia. Proceedings of IX National Congress of Agrometeorology, pp. 84–85. Acutis, M., Trevisiol, P., Gentile, A., Ditto, D., & Bechini, L. (2007). Software components to simulate surface runoff, water, carbon, and nitrogen dynamics in the soil. Proceedings of Farming Systems Design 2007, Catania, Italy, 10–12 September, 2007. Alberts, E.E., Nearing, M.A., Weltz, M.A., Risse, L.M., Pierson, F.B., Zhang, X.C., Laflen, J.M., & Simanton, J.R. (1995). WEPP Model user guide. Chapter 7. Soil component. Analytis, S. (1977). Über die Relation zwischen biologischer Entwicklung und Temperatur bei phytopathogenen Pilzen. Phytopathologische Zeitschrift, 90, 64–76. Argent, R. M. (2004). An overview of model integration for environmental applications – components, frameworks and semantics. Environmental Modelling and Software, 19, 219–234. Argent, R.M., & Rizzoli, A.E. (2004). Development of multi-framework model components. In C. Pahl-Wostl, S. Schmidt, A.E. Rizzoli, & A.J. Jakeman (Eds.), Transactions of the 2nd Biennial Meeting of the International Environmental Modelling & Software Society (Vol. 1, pp. 365–370). Osnabrück, Germany: InternationalEnvironmental Modelling and Software Society (iEMSs). Athanasiadis, I.N., Rizzoli, A.E., Donatelli, M., & Carlini, L. (2006). Enriching software model interfaces using ontology-based tools. iEMSs Congress, Vermont, July 2006, http://www. iemss.org/iemss2006/papers/s5/284_Athanasiadis_1.pdf. Aylor, D. (1982). Modeling spore dispersal in a barley crop. Agricultural Meteorology, 26(3), 215–219. Baier, T. (2007). The rcom package. Available online at http://cran.r-project.org/doc/packages/ rcom.pdf version 1.5-2.2. Retrieved September 17, 2007.

104

M. Donatelli et al.

Baker, C. J., Berry, P. M., Spink, J. H., Sylvester-Bradley, R., Griffin, J. M., Scott, R. K., et al. (1998). A method for the assessment of the risk of wheat lodging. Journal of Theoretical Biology, 194, 587–603. Balderacchi, M., Boccelli, R., & Trevisan, M. (2007). Tools to assess pesticide environmental fate – Agrochemicals/APES, EPRIP 2 and FitoMarche software. Pavia, Italy: La Goliardica Pavese, pp. 142, ISBN 978-88-7830-477-2. Barrett, P. D., Laidlaw, A. S., & Mayne, C. S. (2005). GrazeGro, a European herbage growth model to predict pasture production in perennial ryegrass swards for decision support. European Journal of Agronomy, 23, 37–56. Bastiaans, L. (1991). Ratio between virtual and visual lesion size as a measure to describe reduction in leaf photosynthesis of rice due to leaf blast. Phytopathology, 81(6), 611–615. Bishop, J. (2008). C# 3.0 design patterns (p. 314). Sebastopol, CA: O’Reilly. Blaise, P. H., & Gessler, C. (1992). An extended progeny/parent ratio model. I. Theoretical development. Journal of Phytopathology, 134, 39–52. Braudeau, E. (2006). Le modèle Kamel. DDN.FR.001.390019.000.S.P.2006.000.31500. Paris: Agence pour la Protection des Programmes. Braudeau, E., & Mohtar, R. H. (2006). Modeling the swelling curve for packed soil aggregates using the pedostructure concept. Journal of the Soil Science Society of America, 70, 494–502. Braudeau, E., & Mohtar, R.H. (2009). Modeling the soil system: Bridging the gap between pedology and soil-water physics. Global and Planetary Change 67, 51–61. Brisson, N., Gary, C., Justes, E., Roche, R., Mary, B., Ripoche, D., et al. (2003). An overview of the crop model STICS. European Journal of Agronomy, 18(3–4), 309–332. Brisson, N., Mary, B., Ripoche, D., Jeuffroy, M. H., Ruget, F., Nicoullaud, B., et al. (1988). STICS: A generic model for the simulation of crops and their water and nitrogen balances. I. Theory and parameterization applied to wheat and corn. Agronomie, 18, 311–346. Calvière, I., & Duru, M. (1999). The effect of N and P fertilizer application and botanical composition on the leaf/stem ratio patterns in spring in Pyrenean meadows. Grass and Forage Science, 54, 255–266. Campbell, G. S. (1985). Soil physics with BASIC (p. 150). Amsterdam: Elsevier. Carlini, L., Bellocchi, G., & Donatelli, M. (2006). Rain, a software component to generate synthetic precipitation data. Agronomy Journal, 98, 1312–1317. Carsel, R.F., Imhoff, J.C., Kummel, P.R., Cheplick, J.M., & Donigan, A.S.J. (1988). PRZM-3 A model for predicting pesticide and nitrogen fate in the crop root and unsatured soil zones: User manual for release 3.0. Athens, GA: National Exposure Research Laboratory, Office of Research and Development, U.S. Environmental Protection Agency. Castelan-Estrada, M. (2001). Growth and dry matter allocation in grapevine (Vitis vinifera): Radiation use efficiency and energetic costs. Ph.D. thesis, INA Paris-Grignon, France, 121 p. Confalonieri, R., Acutis, M., Bellocchi, G., Cerrani, L., Tarantola, S., Donatelli, M., et al. (2006). Exploratory sensitivity analysis of CropSyst, WARM and WOFOST: A case-study with rice biomass simulations. Italian Journal of Agrometeorology, 3, 17–25. Confalonieri, R., Bellocchib, G., & Donatelli, M. (2010). A software component to compute agro-meteorological indicators. Environmental Modelling and Software. (in press) Available from http://dx.doi.org/10.1016/j.envsoft.2008.11.007 Cooley, K.R. (1980). Erosivity “R” for individual design storms. In W.G. Knisel (Ed.), CREAMS: A field-scale model for chemicals, runoff, and erosion from agricultural management systems (pp. 386–397) (USDA-SEA Conservation Research Rep. No. 26). Washington, DC: USDA. Corbeels, M., McMurtrie, R. E., Pepper, P. A., & O’Connell, A. M. (2005). A process-based model of nitrogen cycling in forest plantations. Part I. Structure, calibration and analysis of the decomposition model. Ecological Modelling, 187(4), 426–448. Cwalina, K., & Abrams, B. (2006). Aggregate components. In Framework design guidelines: Conventions, idioms, and patterns for reusable .NET libraries (pp. 235–271). Westford, MA: Addison-Wesley. David, O., Markstrom, S. L., Rojas, K. W., Ahuja, L. R., & Schneider, W. (2002). The object modelling system. In L. R. Ahuja, L. Ma & T. A. Howell (Eds.), Agricultural system models in field research and technology transfer (pp. 317–344). Boca Raton, FL: Lewis.

4 A Component-Based Framework for Simulating Agricultural Production

105

Del Furia, L., Rizzoli, A., & Arditi, R. (1995). Lakemaker: A general object-oriented software tool for modelling the eutrophication process in lakes. Environmental Software, 10(1), 43–64. Di Guardo, A., Donatelli, M., & Botta, M. (2007). Two framework components to simulate biophysical systems. In Proceedings of Farming Systems Design 2007, Catania, Italy, 10–12 September 2007. Donatelli M., Acutis M., Bregaglio S., Rosenmund A., & Casellas E. (2009). A framework and a software component to simulate agricultural management. Environmental Modelling and Software (submitted). Donatelli, M., Bellocchi, G., & Carlini, L. (2006a). A software component for estimating solar radiation. Environmental Modelling and Software, 21(3), 411–416. Donatelli, M., Bellocchi, G., & Carlini, L. (2006b). Sharing knowledge via software components: Models on reference evapotranspiration. European Journal of Agronomy, 24(2), 186–192. Donatelli, M., Bolte, J., van Evert, F., & Wang, W. (2003). Which software designs for evolution. In M.K. van Ittersum & M. Donatelli (Eds.), Modelling cropping systems: Science, software and applications. European Journal of Agronomy, 18, 193–195. Donatelli M., Bellocchi G., Habyarimana E., Confalonieri R., & Micale F. (2009). An extensible model library for generating wind speed data. Computers and Electronics in Agriculture, 69 (2009) 165–170. Donatelli, M., Bellocchi G., Habyarimana E., Bregaglio S., Confalonieri R. & Baruth B., (2009b). CLIMA: a weather generator framewor. In Anderssen, R.S., R.D. Braddock and L.T.H. Newham (eds) 18th World IMACS Congress and MODSIM09 International Congress on Modelling and Simulation. Modelling and Simulation Society of Australia and New Zealand and International Association for Mathematics and Computers in Simulation, July 2009, pp. 2377–2383. ISBN: 978-0-9758400-7-8. http://www.mssanz.org.au/modsim09/C3/donatelli_C3a.pdf Donatelli, M., Confalonieri R., Cerrani I., Fanchini D., Acutis M., Tarantola S. & Baruth B., (2009c). LUISA (Library User Interface for Sensitivity Analysis): a generic software component for sensitivity analysis of bio-physical models. In Anderssen, R.S., R.D. Braddock and L.T.H. Newham (eds) 18th World IMACS Congress and MODSIM09 International Congress on Modelling and Simulation. Modelling and Simulation Society of Australia and New Zealand and International Association for Mathematics and Computers in Simulation, July 2009, pp. 2377–2383. ISBN: 978-0-9758400-7-8. http://www.mssanz.org.au/modsim09/C3/donatelli_C3b.pdf Donatelli, M., Omicini, A., Fila, G., & Monti, C. (2004). Targeting reusability and replaceability of simulation models for agricultural systems. In S.E. Jacobsen, C.R. Jensen, & J.R. Porter (Eds.), Proceedings of the 8th European Society for Agronomy Congress (pp. 237–238), 11–15 July, Copenhagen, Denmark. Donatelli, M., & Rizzoli, A. (2008). A design for framework-independent model components of biophysical systems International Congress on Environmental Modelling and Software iEMSs 2008. Proceedings of the iEMSs Fourth Biennial Meeting, Barcelona, Catalonia, 7–10 July 2008, pp. 727–734. Duru, M. (2008). Improvement of time-driven models of lamina cocksfoot digestibility by a process-based model to take account of plant N nutrition and defoliation. Journal of Agronomy and Crop Science, 194(5), 401–412. Duru, M., Adam, M., Cruz, P., Martin, G., Ansquer, P., Ducourtieux, C., Jouany, C., Theau, J.P., & Viegas, J. (2009a). Modelling above-ground herbage mass for a wide range of grassland community types. Ecological Modelling, 220, 209–225. Duru, M., Al Haj Khaled, R., Ducourtieux, C., Theau, J.P., Quadros, F., & Cruz, C. (2009). Do plant functional types based on leaf dry matter content allow characterizing native grass species and grasslands for herbage growth pattern? Plant Ecology, 201, 421–433. Duru, M., Cruz, P., Haj Khaled, R., Ducourtieux, C., & Theau, J. P. (2008). Relevance of plant functional types based on leaf dry matter content for assessing digestibility of native grass species and species-rich grassland communities in spring. Agronomy Journal, 100, 1622–1630. Ferrer-Alegre, F., & Stockle, C. O. (1999). A model for assessing crop response to salinity. Irrigation Science, 19, 15–23. Georgiadis, T., Rossi, S., & Nerozzi, F. (1995). Inferring ozone deposition on agricultural surfaces: An application to herbaceous and tree canopies. Water, Air, and Soil Pollution, 84, 117–128.

106

M. Donatelli et al.

Green, W.H., & Ampt, G.A. (1914). Studies on soil physics. Journal of Agricultural Science, 4(1), 1–24. Hydrology. Transactions American Society Agricultural Engineering, 20, 1100–1104. Hearn, A.B. (1994). The principles of cotton water relations and their application in management. In G.A. Constable & N.W. Forrester (Eds.), Proceedings World Cotton Research Conference, 1st, Brisbane, Australia (pp. 66–92), 14–17 February 1994. Melbourne, Australia: CSIRO. Hillyer, C., Bolte, J., van Evert, F., & Lamaker, A. (2003). The MODCOM modular simulation system. European Journal of Agronomy, 18(3–4), 333–343. Jantunen, A. P. K., Trevisan, M., & Capri, E. (2005). Computer models for characterizing the fate of chemicals in soil: Pesticide leaching models and their practical applications. In J. AlvarezBendì & R. Muñoz-Carpena (Eds.), Soil-Water-Solute process characterisation: An integrate approach (pp. 715–756). Boca Raton, FL: CRC Press. Johnsson, H., Bergström, L., Jansson, P. E., & Paustian, K. (1987). Simulated nitrogen dynamics and losses in a layered agricultural soil. Agriculture, Ecosystems, and Environment, 18, 333–356. Jones, J. W., Hoogenboom, G., Porter, C. H., Boote, K. J., Batchelor, W. D., Hunt, L. A., et al. (2003). The DSSAT cropping system model. European Journal of Agronomy, 18(3–4), 235–265. Jones, J. W., Keating, B. A., & Porter, C. H. (2001). Approaches to modular model development. Agricultural Systems, 70, 421–443. Karlberg, L., Ben-Gal, A., Jansson, P.-E., & Shani, U. (2006). Modelling transpiration and growth in salinity-stressed tomato under different climatic conditions. Ecological Modelling, 190, 15–40. Keating, B. A., Carberry, P. S., Hammer, G. L., Probert, M. E., Robertson, M. J., Holzworth, D., et al. (2003). An overview of APSIM, a model designed for farming systems simulation. European Journal of Agronomy, 18(3–4), 267–288. Lavorel, S., & Garnier, E. (2002). Predicting changes in community composition and ecosystem functioning from plant traits, revisiting the Holy Grail. Functional Ecology, 16, 545–556. Magarey, R. D., Sutton, T. B., & Thayer, C. L. (2005). A simple generic infection model for foliar fungal plant pathogens. Phytopathology, 95(1), 92–100. Makowski, D., Hillier, J., Wallach, D., Andrieu, B., & Jeuffroy, M. H. (2006). Parameter estimation for crop models. In D. Wallach, D. Makowski & J. W. Jones (Eds.), Working with dynamic crop models (pp. 101–150). Amsterdam: Elsevier. Martin, P., Mohtar, R. H., Clouvel, P., & Braudeau, E. (2006, July). Modeling soil-water dynamics for diverse environmental needs. Vermont: iEMSs Congress. McCall, D. G., & Bishop-Hurley, G. J. (2003). A pasture growth model for use in a whole-farm dairy production model. Agricultural Systems, 76, 1183–1205. Mesketer, S. J. (2004). Design patterns in C#. Boston: Addison-Wesley. Meyer, B. (1991). Design by contract. In D. Mandrioli & B. Meyer (Eds.), Advances in object-oriented software engineering. Englewnod Cliffs, NJ: Prentice Hall. Monteith, J. L. (1977). Climate and the efficiency of crop production in Britain. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences, 281, 277–294. Monteith, J. L., & Unsworth, M. H. (1990). Principles of environmental physics. Woburn, MA: Butterworth-Heinemann. Morgan, R. P. C., Quinton, J. N., Smith, R. E., Govers, G., Poesen, J. W. A., Auerswald, K., et al. (1998). The European soil erosion model (EUROSEM): A dynamic approach for predicting sediment transport from fields and small catchments. Earth Surface Processes and Landforms, 23, 527–544. Mulia, R. (2005). Modélisation tri-dimensionnelle de la croissance du système racinaire des plantes en milieu hétérogène avec l’approche de l’automate voxellaire. Ph.D. thesis, USTL, Montpellier 2, p. 86. Mulia, R., & Dupraz, C. (2006). Unusual fine root distributions of two deciduous tree species in southern France: What consequences for modelling of tree root dynamics? Plant and Soil, 281(1/2), 71–85. Neitsch, S.L., Arnold, J.G., Kiniry, J.R., Williams, J.R., & King, K.W. (2002). Soil and water assessment tool. Theoretical Documentation. Temple, TX: Grassland, Soil and Water Research Laboratory, p. 506.

4 A Component-Based Framework for Simulating Agricultural Production

107

Nendel, C., & Kersebaum, K. C. (2004). A simple model approach to simulate nitrogen dynamics in vineyard soils. Ecological Modelling, 177, 1–15. Ollat, N., Diakou-Verdin, P., Garde, J. P., Barrieu, F., Gaudillière, J. P., & Moing, A. (2002). Grape berry development: A review. Journal International des Sciences de la Vigne et du Vin, 36(3), 109–131. Parton, W. J. (2004). Predicting soil temperatures in a shortgrass steppe. Soil Science, 138, 93–101. Pronk, A., Goudriaan, J., Stilma, E., & Challa, H. (2003). A simple method to estimate radiation interception by nursery stock conifers: A case study of eastern white cedar. Netherlands Journal of Agricultural Science, 51, 279–295. R Development Core Team. (2007). R: A language and environment for statistical computing. Vienna: R Foundation for Statistical Computing. Ritchie, J. T. (1972). Model for predicting evaporation from a row crop with incomplete cover. Water Resources Research, 85, 1204–1211. Ritchie, J.T. (1991). Wheat phasic development. In J. Hanks & J.T. Ritchie (Eds.), Modelling plant and soil systems. Agronomy Monographs 31 (pp. 31–54). Madison, WI: ASA, CSSSA, SSSA. Ritchie, J.T., & Otter, S. (1985). Description and performance of CERES-Wheat: A user oriented wheat yield model. In ARS Wheat Yield Project. ARS-28 (pp. 159–175). Springfield, VA: National Technology Information Service. Rizzoli, A. E., Donatelli, M., Athanasiadis, I., Villa, F., Muetzelfeldt, R., & Huber, D. (2005). Semantic links in integrated modelling frameworks. Mathematics and Computers in Simulation, 78, 412–423. Rizzoli, A.E., Donatelli, M., Muetzelfeldt, R., Otjens, T., Svennson, M.G.E., van Evert, F., Villa, F., & Bolte, J. (2004). SEAMFRAME, a proposal for an integrated modelling framework for agricultural systems. In S.E. Jacobsen, C.R. Jensen, & J.R. Porter (Eds.), Proceedings of the 8th European Society for Agronomy Congress (pp. 331–332), 11–15 July, Copenhagen, Denmark. Saltelli, A.,Tarantola, S., Campolongo, F. & Ratto, M. (2004). Sensitivity analysis in practice: a guide to assessing scientific models. Chichester, England: John Wiley & Sons Ltd. Saxton, K. E., & Rawls, W. J. (2006). Soil water characteristic estimates by texture and organic matter for hydrologic solutions. Journal of the Soil Science Society of America, 70, 1569–1578. Schapendonk, A. H. C. M., Stol, W., Van Kraalingen, D. W. G., & Bouman, B. A. M. (1998). LINGRA, a sink/source model to simulate grassland productivity in Europe. European Journal of Agronomy, 9, 87–100. M.E. Shibu, P.A. Leffelaar, H. van Keulen, P.K. Aggarwal (2009). LINTUL3, a simulation model for nitrogen-limited situations – application to rice European Journal of Agronomy (submitted). Shimono, H., Hasegawa, T., Moriyama, M., Fujimura, S., & Nagata, T. (2005). Modeling spikelet sterility induced by low temperature in rice. Agronomy Journal, 97, 1524–1536. Sitch, S., Cox, P. M., Collins, W. J., & Huntigford, C. (2007). Indirect radiative forcing of climate change through ozone effects on the land-carbon sink. Nature, 448, 791–795. Smith, R. E., & Parlange, J. Y. (1978). A parameter-efficient hydrologic infiltration model. Water Resources Research, 14, 533–538. Soil Conservation Service. (1972). Section 4: Hydrology. In National Engineering Handbook. SCS. Spiker, E. C., Hosker, R. P., Comer, V. J., White, J. R., Werre, R. W., Jr., Harmon, F. L., et al. (1992). Environmental chamber for study of the deposition flux of gaseous pollutants to material surfaces. Atmospheric Environment, 26, 2885–2892. Stockle, C. O., Donatelli, M., & Nelson, R. (2003). CropSyst, a cropping systems simulation model. European Journal of Agronomy, 18(3–4), 289–307. Streck, N. A., Weiss, A., Xue, Q., & Baenziger, P. S. (2003). Improving predictions of developmental stages in winter wheat: A modified Wang and Engel model. Agricultural and Forest Meteorology, 115, 139–150.

108

M. Donatelli et al.

Szypersky, C., Gruntz, D., & Murer, S. (2002). Component software - beyond object-oriented programming (2nd ed.). London: Addison-Wesley. Tiktak, A., Van den Berg, F., Boesten, J.J.T.I., Van Kraalingen, D., Leistra, M., & Van der Linden, A.M.A. (2001). Manual of FOCUS PEARL v 1.1.1. RIVM Report 711401008, Alterra Report 28 (p. 144). Bilthoven, The Netherlands: RIVM. Trevisan, M., Sorce, A., Balderacchi, M., & Di Guardo, A. (2007). A software component to simulate agro-chemicals fate. In Proceedings of Farming Systems Design 2007, Catania, Italy, 10–12 September 2007. Van Dam, J.C., Huygen, J., Wesseling, J.G., Feddes, R.A., Kabat, P., Van Walsum, P.E.V., Groenendijk, P., & Van Diepen, C.A. (1997). Theory of SWAP version 2.0. Report 71. Wageningen, The Netherlands: Department of Water Resources, WAU. Van Evert, F., & Lamaker, A. (2007). The MODCOM framework for component-based simulation. In proceedings of Farming Systems Design 2007, Catania, Italy, 10–12 September, 2007. Van Ittersum, M. K., Ewert, F., Heckelei, T., Wery, J., Alkan Olsson, J., Andersen, E., et al. (2008). Integrated assessment of agricultural systems – A component-based framework for the European Union (SEAMLESS). Agricultural Systems, 96(1–3), 150–165. Van Ittersum, M. K., Leffelaar, P. A., Van Keulen, H., Kropff, M. J., Bastiaans, L., & Goudriaan, J. (2003). On approaches and applications of the Wageningen crop models. European Journal of Agronomy, 18(3–4), 187–393. Van Keulen, H., & Seligman, N.G. (1987). Simulation of water use, nitrogen nutrition and growth of a spring wheat crop. Simulation Monographs. Wageningen, The Netherlands: Pudoc. Van Keulen, H., & Wolf, J. (1986). Modelling of agricultural production: Weather soils and crops. Simulation Monographs. Wageningen, The Netherlands: Pudoc. Villa F., Donatelli, M., Rizzoli, A., Krause, P., Kralisch, S., & Van Evert, F.K. (2006, July). Declarative modelling for architecture independence and data/model integration: A case study. iEMSs congress, Vermont. Vivin, Ph, Castelan, M., & Gaudillère, J. P. (2002). A source/sink model to simulate seasonal allocation of carbon in grapevine. Acta Horticulturae, 584, 43–56. Von Hoyningen-Huene, J. (1981). Die Interzeption des Niederschlags in landwirtschaftlichen Pjanzenbestanden. Arbeitsbericht Deutscher Verband fur Wasserwirtschaft und Kulturbau, DVWK, Braunschweig. Wadia, K. D. R., & Butler, D. R. (1994). Relationship between temperature and latent periods of rust and leaf-spot diseases of groundnut. Plant Pathology, 43, 121–129. Waggoner, P. E. (1973). The removal of Helminthosporium maydis spores by wind. Phytopathology, 63(10), 1252–1255. Waggoner, P. E., & Horsfall, J. G. (1969). EPIDEM. A simulator of plant disease written for a computer. Bulletin of the Connecticut Agricultural Experiment Station, 698, 80. Wermelinger, B., & Koblet, W. (1990). Seasonal growth and nitrogen distribution in grapevine leaves, shoots and grape. Vitis, 29, 15–26. Williams, J.R., & Berndt, H.D. (1977). Sediment Yield Prediction Based on Watershed Hydrology. Transactions of the American Society of Agricultural Engineers, 20, 1100–1104. Williams, J. R., Jones, C. A., Kiniry, J. R., & Spanel, D. A. (1989). The EPIC crop growth model. Transactions of the American Society of Agricultural Engineering, 32, 497–511. Woolhiser, D.A., Smith, R.E., & Goodrich, D.C. (1990). KINEROS, a kinematic runoff and erosion model: Documentation and user manual. United States Department of Agriculture, ARS-77. Wosten, J. H. M., Lilly, A., Nemes, A., & Le Bas, C. (1999). Development and use of a database of hydraulic properties of European soils. Geoderma, 90, 169–185. Zadoks, J.C., & Schein, R.D. (1979). Epidemiology and plant disease management (p. 427). London: Oxford University Press.

Chapter 5

A Generic Farming System Simulator Kamel Louhichi, Sander Janssen, Argyris Kanellopoulos, Hongtao Li, Nina Borkowski, Guillermo Flichman, Huib Hengsdijk, Peter Zander, Maria Blanco Fonseca, Grete Stokstad, Ioannis N. Athanasiadis, Andrea E. Rizzoli, David Huber, Thomas Heckelei, and Martin van Ittersum

Introduction The prime decision making unit in agriculture is the farm. It is the unit where agroecological innovations start and where agricultural and agri-environmental policies trigger changes in land use, production and externalities (e.g. nitrate leaching, soil erosion and pesticide use). The European Union (27 member states) counts ca. 15 million farms with a wide variation in endowments, specialisation and land use (Eurostat 2007). As a consequence of these differences and the diversity in entrepreneurship and personal or household aims, responses to a specific policy or innovation may differ across the farming community. This seriously complicates the devise and selection of effective and efficient policies, i.e. what may be an effective (realizing the desired effect with respect to for instance the environment) and efficient (realizing a desired effect at low cost for a farm or community) for one type of farms may not be so for another type. Evaluation of present policies can be done based on empirical data, for instance using systematic data collected for a sample of farms throughout a region, nation or continent. The Farm Accountancy Data Network (FADN) provides such a source of information for the European Union. This is indeed useful to evaluate effectiveness of policies in terms of some indicators, particularly economic ones. However, such sources generally lack information on agricultural management and environmental issues. Moreover, these two data gaps are interrelated: the lack of agricultural management makes the application of for instance crop simulation models to assess environmental issues largely impossible. Hence, only FADN data complemented with detailed surveys and measurements enable the full ex-post evaluation of policy measures. For ex-ante assessment of policies, i.e. assessment of policies before their introduction, there is little empirical basis. Here, mathematical modelling can K. Louhichi (*) CIHEAM-IAMM, 3191, Route de Mende, 34093 Cedex 5, Montpellier, France. Current address: UMR économie publique, INRA-AgroParisTech, Avenue Lucien Brétignières, 78850, Thiverval Grignon, France e-mail: [email protected] F.M. Brouwer and M. van Ittersum (eds.), Environmental and Agricultural Modelling: Integrated Approaches for Policy Impact Assessment, DOI 10.1007/978-90-481-3619-3_5, © Springer Science+Business Media B.V. 2010

109

110

K. Louhichi et al.

potentially be an important source of information for the assessment. But for larger areas and systems, it requires a consistent and efficient application of such models to the great variation in prevailing farm types. Recently, there has been increasing interest in so-called bio-economic farm models (Thompson 1982; Deybe and Flichman 1991; Wossink et al. 1992; Janssen and Van Ittersum 2007). These models link formulations describing farmers’ resource management decisions to formulations that describe current and alternative production possibilities in terms of required inputs to achieve certain outputs and associated externalities (Kruseman and Bade 1998; Janssen and Van Ittersum 2007). One of their applications is to assess farm responses to policies and how these may differ across various farm types. More precisely for the European Union (EU), such applications might focus on assessing supply responses of farms across the EU and their effect on markets, and on more detailed regional assessments of policies in terms of economic, environmental and landscape issues. For application of a bio-economic farm model across the European Union, the model must be generic and flexible enough to capture for instance the range of conditions from North to South in biophysical terms and from West to East in socio-economic aspects. Application of one consistent bio-economic farm model to a broad range of farm types differing in size, intensity, specialisation and land use (Andersen et al. 2007) in our view requires a modular set-up. The aim of this chapter is to present a bio-economic farm model, FSSIM (Farm System Simulator) with a modular set-up, which can be used as a standalone model and as a model within the framework for integrated assessment, i.e. SEAMLESS-IF (Van Ittersum et al. 2008). This farm model includes a data module for agricultural management (FSSIM-AM) and a mathematical programming model (FSSIM-MP). It offers a structure to flexibly apply it to farm types that may differ in: soils and climate, resource endowments, agricultural activities and their management options and utility functions, and that may be subject to a broad range of agricultural and agri-environmental policies (Fig. 5.1). The chapter starts with a brief description of the farm typology that is used as a basis to simulate European farms. We present the mathematical programming part of FSSIM (FSSIM-MP), in which information on farm activities, resource constraints, policies and utility function of the farm is integrated. The following section presents the agricultural management part (FSSIM-AM) and its optional link to biophysical simulation models. The software implementation of FSSIM is presented and an application is provided at the end of this chapter.

Farm Typology Aim of the Farm Typology Modelling all individual farms within the EU is not feasible because of the large number of farms, and the existing diversity among different farming systems. For that reason it was decided to develop a farm typology that captures the heterogeneity

5 A Generic Farming System Simulator

FSSIM-AM DATA BASE ON SOIL, CLIMATE...

DATA BASE ON FARM RESOURCES

DATA BASE ON CROPS, ANIMALS, PRODUCTION TECHNIQUES, AGRONOMIC RULES...

AGRICULTURAL ACTIVITY GENERATOR PRODUCTION ENTERPRISE GENERATOR

111

PRODUCTION TECHNIQUE GENERATOR

SET OF CURRENT AND ALTERNATIVE AGRICULTURAL ACTIVITIES

Land

Water

Labour

Yield, externalities

DATA ON AGRICULTURAL POLICIES AND SOCIOECONOMC ENVIRONMENT

TECHNICAL COEFFICIENT GENERATOR

Costs, prices, premiums, labour and machine need

INPUT/OUTPUT COEFFICIENTS

y

...

FSSIM-MP LIVESTOCK MODULE

PERENNIAL MODULE

TREND MODULE

Production enterprise with specified production technique (level, type... of input)

Biophysical model "APES"

Machiner

CROPS MODULE

COMMON MODULE

POLICY MODULE

PMP CALIBRATION MODULE

PREMIUM MODULE

RISK MODULE

Maximize: Utility function Subject to: agronomic, technical, economic, institutional, feeding ... constraints

FSSIM-OUTPUT - Farm income - Positive/negative externalities - Agricultural activity levels ...

Fig. 5.1 FSSIM and its components

in farming systems. Based on FADN and Farm Structure Survey (FSS), this farm typology provides for each region in the EU (so called NUTS2 regions) a set of typical, well defined farm types in terms of size (i.e. total available agricultural land), intensity of production (i.e. output per hectare), land use and specialisation. The thresholds that are used to allocate farms to a specific farm type are the same for all Europe. A spatial allocation procedure was also used for allocating the farm types to spatial units, with more homogenous bio-physical endowments (Chapter 7, this Volume). The aim was to enable the aggregation of farm types to both natural (territorial) and administrative regions. Each farm type of the SEAMLESS farm typology is linked to a number of agri-environmental zones defined as a combination of climatic conditions, soil characteristics and other geographical attributes.

Representation of the Farm Type: Average Versus Typical Farm The farm type consists of a group of farms with similar socio-economic and agri-environmental characteristics. To simulate the behaviour of a certain farm type with a farm model (like FSSIM) it is important to select/construct a farm that represents adequately the whole group of farms that are classified in the same farm type.

112

K. Louhichi et al.

Table 5.1 Advantages and disadvantages of the average and the typical farm Average farm Typical farm • More observed production activities • Representative for a group of holdings that meet at the regional and market scale the group criteria • Lower additional data requirements • Only the production activities of a specific farm are (compared with FADN) observed • The average farm does not exist • Farm scale data are required which are not always available • The typical farm is observed in reality • More constraints and non-zero terms are needed for calibration • Less constraints and non-zero terms are needed for calibration

Representation of the farm type could be achieved either by the average or the typical farm. The average farm could be defined as a virtual (not observed in reality) farm which is derived by averaging historical data from farms that are grouped in the same farm type. A typical farm is an existing (observed) farm with representative, for a certain farm type, properties and characteristics. Different approaches could be used when trying to identify a representative typical farm (e.g. selecting the farm that is close to the average farm or the one with the median profit). The advantages and disadvantages of using one of the two approaches are summarized in Table 5.1. The average farm was selected to represent all farms that belong to the same farm type in a certain region. The advantage of using the average instead of the typical farm is that this enables upscaling of farm type analyses to regional or even EU level. Generally the simulated farm plans of an average farm are less specialized (i.e. include a broader range of activities) than the results of a typical farm. This is mainly because production activities of all farms that belong to the same farm type will be represented in the base year data which is used for calibration. Even less common activities that are not so interesting to include in a farm level analysis but which may be important at the aggregated regional or EU level analysis will be represented in the base year data. In contrast, the base year data of a typical farm are restricted to the few production activities that are observed in the farm plan of a single farm. One problem associated with the use of the average farm for representing a farm type is related to the fact that the average farm is not observed in reality. All different activities observed in the farm plans of individual farms of a certain farm type, are included in the farm plan of the average farm. Calibration of FSSIM to reproduce base year data of the average farm becomes a complicated procedure. It is expected, that a much larger number of binding constraints are required for calibration, when simulating the average farm than when simulating an individual farm. To overcome this problem a calibration method which is based on Positive Mathematical Programming (PMP) (Howitt 1995) has been developed and used in FSSIM. The specification of the average farm and the calculation of its resource endowments are obtained by dividing the resources endowments of all the farms that are classified in this group by their numbers (i.e. number of farms of the group). The simple calculation is demonstrated in Table 5.2 with data from three farm types that have been identified in Midi-Pyrénées (France).

5 A Generic Farming System Simulator

113

Table 5.2 Crop areas (ha) of two farm types (average farms) in Midi-Pyrénées Farm type 3201 Farm type 3202 Size Large Large Intensity Medium Medium Specialization – land use Arable/cereals Arable/set-aside Number of farms represented 2,330 991 Total Per farm Total Per farm Barley 9,550 4.1 1,559 1.6 Dry pulses 8,749 3.8 3,593 3.6 Maize (grain) 81,764 35.1 24,820 25 Oil seeds 44,112 18.9 17,509 17.7 Other cereals 7,344 3.2 2,026 2 Peas 8,259 3.5 3,593 3.6 Set-aside 21,780 9.3 18,739 18.9 Soya 6,936 3 3,612 3.6 Sunflower 33,267 14.3 12,509 12.6 Wheat (durum) 40,313 17.3 11,326 11.4 Wheat (soft) 30,570 13.1 12,182 12.3 Permanent crops and 15,034 6.5 10,212 10.3 vineyards Irrigated area 97,375 41.8 30,085 30.4

FSSIM-MP: Mathematical Programming Model Aim of FSSIM-MP Based on mathematical programming, FSSIM-MP seeks to capture resource, socio-economic and policy constraints and the farmer’s major objectives. The use of a mathematical programming approach has the advantage to explicitly model technological and political constraints (set-aside obligations, production quotas and cross-compliance restrictions) under which behavioural functions cannot be derived easily or at all (Heckelei and Wolff 2003). It allows also mixed ecological-economic analysis (Falconer and Hodge 2000; Louhichi et al. 2004). The principal components of FSSIM-MP are: –– A set of decision variables that describe the agricultural activities and state of the system. –– An objective function describing the farmer’s behaviour and goals in particular concerning risk. –– A set of explicit physical, financial, technical, economic and agronomic constraints, representing specifications for system operation. –– A set of policy and environmental measures (price and market support, quota and seta-side obligations, cross-compliance restrictions, etc.) as included in the Common Market Organizations (CMOs) regulations and some specific regulations.

114

K. Louhichi et al.

FSSIM-MP is linked to a data module for agricultural management (FSSIM-AM), which aims to describe or generate current and alternative activities and quantifies their input-output coefficients (both yields and environmental effects) (see the following section). Once the potential activities have been generated, FSSIM-MP chooses those that best fit the farmer’s objectives, given the set of resource, technological and political constraints. The principal outputs generated by the FSSIM-MP model are land use, production, input use, farm income and environmental effects of the farm type for a specific policy. These outputs can be used directly or translated into indicators (simple or composite) to provide measures of the impact of policies.

FSSIM-MP Overview FSSIM-MP is a comparative static programming model with a non-linear objective function representing important elements of a farmer’s behaviour. FSSIM uses exogenous prices that can come from different sources (in the base year they come from Eurostat or/and FADN data and in the simulation they can come from a market model, such as CAPRI). The principal FSSIM-MP specifications are: (i) A mono-periodic model which optimizes an objective function for one period (i.e. 1 year) over which decisions are taken. This implies that it does not explicitly take account of time. Nevertheless, to incorporate some temporal effects, agricultural activities are defined as “crop rotations” and “dressed animal” instead of individual crops and animals. (ii) A risk programming model based on the Mean-Standard deviation method in which expected utility is defined under two arguments: expected income and risk (Hazell and Norton 1986). (iii) An activity based model to enable integrated assessment of new policies which are linked to an activity (i.e. production process). (iv) A primal based model where technology is explicitly represented in order to simulate the switch between production techniques as well as between production systems. (v) A model with discrete activities to integrate easily the engineering production functions generated from biophysical models and to account for positive and negative jointness in outputs (i.e., joint production) associated with the production process. (vi) A positive model in the sense that its empirical applications exploit the observed behaviour of economic agents and where the main objective is to reproduce the observed production situation as precisely as possible. (vii) A generic model designed with the aim to be flexible, re-usable, adaptable and easily extendable to achieve different modelling goals. The mathematical structure of FSSIM-MP is formulated as follows:

Maximise: U = Z - φ σ

(5.1)

5 A Generic Farming System Simulator

115

Subject to: Ax ≤ B; x ≥ 0

(5.2)

Where: U is the variable to be maximised (i.e. utility), Z is the expected income (i.e. the average annual income), x is a (n × 1) vector of agricultural activity levels, A is a (m × n) matrix of technical coefficients, B is a (m × 1) vector of available resource levels, f is a scalar for the risk aversion coefficient and s is the standard deviation of income according to states of nature defined under two different sources of variation: yield (due to climatic conditions) and prices. The expected income (Z) is a non-linear profit function. Using matrix denotation, this gives:

(

Z = ∑ p j q j + ∑ paj, l q aj, l + ∑ si, t − c i, t j

j, l

i, t

y i, t x i  x i  + ∑  d i, t + − ϖL 2  hi i, t 

)hx

i

i

(5.3)

Where: i indexes agricultural activities, j indexes crop products, l indexes quota types (e.g. for sugar beet A and B quota exist), t indexes number of years in a rotation, p is a vector of average product prices, q is a vector of sold production, pa is a vector of additional price that the farmer gets when selling within quota l, qa is a vector of sold production within quota l, s is a vector of subsidies per crop within agricultural activity i (depending on the Common Market Organisations [CMOs]), c is a vector to account for variable cost per crop within agricultural activity i, d is a vector of linear terms used to calibrate the model (depending on the calibration approach), Y is a symmetric, positive (semi-) definite matrix of quadratic terms used to calibrate the model (depending on the calibration approaches), h is a vector representing the length of a rotation within each agricultural activity, v is a scalar for the labour cost and L is the number of hours of rented labour. An agricultural activity is defined in FSSIM as a way of growing a rotation (including mono-crop rotations) taking into account the agri-environmental zone (or soil type), the management practice, and the production orientation. It consists of a combination of one crop rotation, one agri-environmental zone, one production technique (i.e. management type) and one production orientation. Let R denote the set of crop rotations (including mono-crop rotations), S the set of agri-environmental zones, T the set of production techniques and Sys the set of production orientations. The set of agricultural activities i can be defined as follows: i = {i1, i2,…} = {(R1,S1,T1,Sys1), (R2,S1,T1,Sys1),…} ⊆ R × S × T × Sys. Agricultural activities can be based on individual crops if data on crop rotations are not available. The principal technical and socio-economic constraints that are implemented in FSSIM-MP are: arable land per soil type (or agri-environmental zone), irrigable land per soil type, labour and water constraints. The same rule was applied for all

116

K. Louhichi et al.

of these constraints: the sum of the requirements for each resource cannot exceed resource availability. For estimating the risk coefficient to include in FSSIM, three options are proposed in the Risk module to be selected by users: –– Risk neutral: implies that the risk aversion coefficient is equal to zero (f = 0). –– Risk averse: set risk aversion coefficient: implies that the user has to choose the value to attribute to the risk aversion coefficient. The chosen value can vary from 0 to 1.65 (0 < f £ 1.65). –– Risk averse: automatically calibrate the risk aversion coefficient: implies that the model will attribute automatically a value to the risk coefficient which gives the best fit between the model’s predicted crop pattern and the observed values in the base year. This value ranges between 0 and 1.65 (0 < f £ 1.65). FSSIM-MP can be calibrated using any of the following approaches, depending on the application type: (i) The risk approach; (ii) The standard PMP procedure (Howitt 1995); (iii) The Rhöm and Dabbert’s PMP approach (Röhm and Dabbert 2003); and (iv) The approach described in Kanellopoulos et al. (2009). The base year information for which the model is calibrated stems from a 3-year average around 2003 (or any update of this baseyear). In terms of policy representation, FSSIM includes the major policy instruments related to production activities such as price and market support and set-aside schema as well as cross-compliance and agro-environmental measures. The following section gives an overview of the different policy instruments linked to arable crops and how they are considered in FSSIM-MP.

Modelling of Policy Instruments in FSSIM-MP FSSIM is developed to analyse the European agricultural and environmental policies, either proposed or actual, and to enable ex-ante assessments of policy and market changes. To achieve this goal, it is necessary to take into account a wide range of the proposed EU policy instruments. The principal policy instruments that are implemented in FSSIM-MP are the Common Agricultural Policy (CAP) support regime (price and market support, set-aside schema, quota system, etc.) included in the CMOs regulations, as well as cross-compliance and agri-environmental measures included in Horizontal and Rural Development Regulations. These policy instruments are captured in FSSIM-MP either by embedding them in the objective function (e.g. premiums), or by including them as constraints (e.g. set-aside and non-food production must cover set-aside obligations, set-aside is not allowed to exceed more than a certain

5 A Generic Farming System Simulator Table 5.3 Policy instruments implemented in FSSIM-MP Instrument Modelling Linked to agricultural activities CAP compensation payment and included in the objective (including Single Farm function Payment) Milk and sugar Constraints; upper bounds on sales beet quotas Compulsory set-aside Constraints; restrict set-aside to minimum 10% of COP (cereals, oilseeds and protein) crops Voluntary set-aside Constraints; restrict total set-aside to 33% of COP crops Environmental condition/ Constraints; controlled by cross-compliance binary-variables Agri-environmental measures

Constraints; controlled by binary-variables

Modulation of payment

Constraints; controlled by binary-variables Linked to agricultural activities and included in the objective function

Member State (national) compensation payment

117

Data source CMOs

CMOs CMOs

CMOs CMOs + specific national and regional implementation CMOs+ specific national and regional implementation CMOs Specific national and regional implementation

CMO: Common Market Organisation

percentage of COP crops). Table 5.3 gives a brief description of how the different policy instruments are modelled in FSSIM-MP. In case of a non-EU application these policy instruments can be de-activated. Modelling all these instruments was an important challenge for FSSIM-MP, as even if some of them are implemented in an identical way everywhere in the EU25 (e.g. direct payment), others such as environmental measures have quite different national/regional implementations. In addition, the information on the administrative implementation of these specific measures is usually scarce, and often not systematically monitored, not published or even not open to the public. The implementation of these instruments depends on the analysed policy in different scenario assumptions which are the Agenda 2000 for the base year scenario and the 2003 CAP reform for the baseline scenario. Using a time horizon of 2013, the baseline scenario is interpreted as a projection in time covering the most probable future development of the European agricultural policy, with the Luxemburg Agreements on Common Agricultural Policy Reform as the core, and including all future changes already foreseen in the current domestic, EU and international legislation (e.g. sugar market reform). Taken as reference run, the baseline scenario is used for the interpretation and analysis of different policy scenarios.

118

K. Louhichi et al.

Exogenous Assumptions for a Baseline Scenario The baseline scenario should capture the complex interrelations between technological, structural, policy, population and market changes related to agricultural production and commodities world-wide. A number of exogenous assumptions are adopted in FSSIM-MP while building the baseline scenario. Some of these are characteristic for all farm models; others are specific to our system, because there is a need of consistency with the other models in the model chain of SEAMLESS, especially with the regional market model. The key underlying assumptions considered in FSSIM-MP are the following: –– Inflation: an assumed inflation rate of 1.9% per year was adopted. –– Prices: the FSSIM baseline prices are obtained indirectly from the market model CAPRI. It consists to multiply the FSSIM base year prices (coming from the survey or Eurostat) by the relative change of SEAMCAP prices between the base year and baseline scenarios. –– Technical progress: the technical innovation is captured through the set of alternative activities generated and assessed by other components of FSSIM (see next section). This means that in the base year analysis only current activities are considered and in the baseline scenario both current and alternative activities can be included without any trend on yield.

FSSIM-MP Structure: Modular Setup FSSIM-MP has a modular set-up which includes crops, livestock, perennials, premium, Positive Mathematical Programming (PMP), risk, trend and policy modules. These modules are linked indirectly by an integrative module involving the objective function and the common constraints (Fig. 5.2). Each module includes two GAMS files. The first one links the data definition and the module’s equations and the second file contains the module’s equations. Each module generates at least one variable which is used to define the common module’s equations, thus providing a link between the different modules. Thanks to this modularity, FSSIM-MP provides the capabilities to add and delete modules (and their corresponding constraints) following the needs of the simulation, to select one or several calibration approaches (risk, standard PMP, Rhöm and Dabbert PMP approach) and to control the flow of data between database and software tools. FSSIM-MP also has the advantage that it can be run with simple or detailed survey data (i.e. according to the level of detail of the available data). Additionally, it can read input data stored in any relational database, in Excel or in GAMS-include files provided that they are structured in the required format. FSSIM-MP can be applied to individual (i.e. real) or representative farms (i.e. typical or average farms) as well as to natural (territorial) or administrative regions by considering the selected region as a large farm (i.e. if the heterogeneity

5 A Generic Farming System Simulator

119

FSSIM-Mathematical Programming (MP) CROPS MODULE

LIVESTOCK MODULE

Crop income

Animal income PREMIUM MODULE

Premium levels

PMP terms

COMMON MODULE (Objective function)

Yield and price projection TREND MODULE

PMP CALIBRATION MODULE

Policy measures

.......

PERENNIAL MODULE

Standard deviation of farm income RISK MODULE

POLICY MODULE

Fig. 5.2 Relationships between FSSIM-MP components

among farms inside the region is insignificant) or by aggregating the results of individual or representative farms (i.e. assuming the inter-dependencies between farms are minor).

FSSIM-AM: Agricultural Management Aim of FSSIM-AM Aim of the Agricultural Management Module is to describe, generate and quantify production techniques of current and alternative production enterprises which can be evaluated by APES (Chapter 4, this Volume), or other cropping/livestock system models, in terms of production and environmental effects. In this chapter, we focus on annual crop activities to describe FSSIM-AM, although the same methodologies have been used for livestock and perennial activities. The fully quantified activities i.e. the complete sets of agricultural inputs and outputs are assessed in FSSIM-MP on their contribution to the farmer’s and policy goals considered. Alternative activities are new activities or currently not widely practised activities in the study area, and include technological innovations or newly developed cropping or husbandry practices (Van Ittersum and Rabbinge 1997; Hengsdijk and Van Ittersum 2002). Current activities are widely practiced in a sample region and their management operations and some of the associated outputs can be based on observed data and expert knowledge.

120

K. Louhichi et al. Agricultural Management Module (1b) database on agronomic rules

(1a) database on weather, soils, crops and animals

(4c) Expert knowledge (2) Production enterprise generator

(3a) Production technique generator

(4a) Databases with costs, labour and machine needs, etc.

(4b) APES (APES uses (1a) and (1b))

(3b) database on production techniques (Also uses 4a)

(4d) Survey & Database on current activities.

(5) Technical Coefficient Generator

(6) Matrix or relational database for FSSIM

= Algorithm = Database = Information flow Fig. 5.3 Relationships between algorithms and database components in FSSIM-AM

The main components of FSSIM-AM are the Production Enterprise Generator (PEG), Production Technique Generator (PTG), Survey on Current Activities, and the Technical Coefficient Generator (TCG) (Fig. 5.3); FSSIM-AM is linked to a cropping system model, i.e. APES within SEAMLESS-IF. Relational databases are used to collect and store input and output information which is used in different components. FSSIM and APES require information on the following items: yield of products, general management (sowing/harvesting), tillage/crop residue management, nutrient management, water management, weed, pest and disease management and the timing of different management events, while FSSIM-MP additionally requires information on costs of activities, price and yield variability, and different types of policies.

5 A Generic Farming System Simulator

121

Deriving and Quantifying Current Activities A current activity (CA) is an agricultural activity currently in use on farms. The biophysical assessment of these activities requires detailed information on the implements used in field operations and timing of all field operations to produce a product. This detailed management information cannot be extracted from aggregated databases such as FADN. In order to collect those data on current activities, two computer-based surveys were developed at different levels of detail. Detailed Survey A detailed survey includes detailed agronomic information on crop rotations and the field operations of each crop (Zander et al. 2009). This survey is completed by local experts with several years of experience in crop cultivation and knowledge about current agricultural practices. In order to limit the possible rotations, the number of potential crops in the survey is reduced on the basis of the crop distribution information in the FADN. The survey’s Graphical User Interface gives direct access to three windows containing all data entered: (i) The crop rotations; (ii) The field operations from tillage to harvest and their timing per crop; and (iii) The resources related to field operations, differentiated per type of soil and climate. Crops, field operations and resources have to be selected from predefined lists. The crop is defined by the botanical name, the growing period, the plant part used, the product quality aimed at and the production orientation (conventional or organic). Different intended usage and growing procedures for botanically the same plant species can be classified in a consistent way. For crop management, we tried to reduce complexity by identifying 70 different field operations which, in the case of seed and harvest, can have different economic and technical characteristics (e.g. harvest of sugar beet or cereals). In the case of fertilizers and pesticides, the survey offers no brand choice but the possibility to choose a certain type of fertilizer or pesticide – in the latter case only types of treatment are available (e.g. post emergence treatment of grass weeds in rape). The survey is designed to obtain the most complete information set on agricultural activities with a minimum effort. The tool includes a semi-automatic cost calculation procedure, which requires some economic information like machinery costs. Simple Survey The detailed survey requires a detailed knowledge about crop production, which is not easily accessible in a large number of regions. Consequently, a less demanding survey, e.g. Simple Survey was developed and implemented in a larger sample of

122

K. Louhichi et al.

regions. This survey concentrates on economic data for crop farming as well as on livestock and policy variables. Data needs for filling in the Simple Survey are limited, and required data can be readily derived from national or regional publications. The Simple Survey’s structure includes one part for each topic: livestock, crop farming and policies. Livestock is divided into one sheet each for beef cattle, dairy cattle, small ruminants for dairy, small ruminants for meat and grassland. Data for three intensities of livestock production can be entered. In the crop part, there is one variable list that must be filled in for all major crops based on regional FADN data. Crop rotations are entered in a separate survey sheet. The major difference compared with the detailed survey is that simple survey does not contain information on the timing of operations. The policy part consists of three single sub-parts with different structures, referring to CAP compensation payments, cross-compliance and agri-environmental measures as well as national subsidies.

Data Storage and Checking Both surveys are server-based tools for which users only have to install a small application on their own computer. Entered data are directly stored in a PostgreSQL database server. Data from these databases are uploaded into the integrated database of SEAMLESS (Janssen et al. 2009). To facilitate error checking of entered data, there are several overviews provided in the surveys that can be opened from the graphical user interfaces.

Generating Alternative Activities Purpose Two components are used to define management operations of alternative activities: (i) PEG generating rotations; and (ii) PTG generating management operations and associated inputs for rotations.

Production Enterprise Generator PEG is a tool to generate feasible sets of crop rotations of farms based on crop suitability filters, such as soil and climate characteristics and specific agronomic, rotation filters for annual arable crops. For example, timeliness rules avoid the generation of rotations in which crops are sown before the preceding crop is harvested. These pre-screening suitability filters limit the number of crop rotations for which production techniques need to be defined and the number of simulations to be carried out by APES.

5 A Generic Farming System Simulator

123

Based on specific crop requirements (Russell 1990; Reinds and Van Lanen 1992; Wolf et al. 2004; Alterra and INRA 2005), ten crop suitability filters were developed. An example of such a filter is the altitude filter, which excludes crops from areas with unfavourable temperatures for crop production. The PEG contains an adapted version of ROTAT (Dogliotti et al. 2003). ROTAT is a tool to generate feasible crop rotations based on agronomic rotation suitability filters in a flexible and transparent manner. An example of a rotation suitability filter from ROTAT that was re-used in the PEG is the crop frequency filter, which limits the crop frequency in a rotation. Production Technique Generator The PTG is a tool to generate alternative production techniques for a feasible set of crop rotations. A production technique is a complete set of agronomic inputs characterized by type, level, timing and application technique (Van Ittersum and Rabbinge 1997). First, the PTG creates alternative management practices based on user defined parameters and agronomic expert rules. Second, the PTG combines different alternative management practices into production techniques. The complete set of inputs consists of the following management practices (Fig. 5.3): – – General management includes all operations that are mandatory for a successful harvest such as sowing, harvesting, clipping, pruning and field inspection. –– Water management includes rain fed or irrigated production; for irrigation, the method of application (sprinkler, furrow, etc.), timing rules and amounts must be described. –– Nutrient management includes a description of the level of application, type of nutrient, method of application and dose/timing of application of nutrient management. –– Weed pest and disease management includes packages that describe all operations required to achieve a well-defined control level of weeds, pests and diseases for a crop. –– Conservation management includes operations aimed at soil conservation and landscape and biodiversity management.

Technical Coefficient Generator The TCG converts the agronomic input and output coefficients generated by the Surveys on CA, PEG and PTG in APES and FSSIM-MP compatible inputs. The TCG extracts data from the farm typology (Andersen et al. 2007) to define the farm types for FSSIM-MP. The result of the TCG is a fully quantified set of agricultural activities (Technical Coefficient Matrix) that can be transferred to FSSIM-MP.

124

K. Louhichi et al.

The Technical Implementation Through an Integrated Modelling Framework (SeamFrame) FSSIM is a collection of models, which are integrated into the modelling framework SeamFrame (Chapter 9, this Volume), have consistent inputs and outputs through an ontology and implement the OpenMI-standard to exchange data at runtime as components. SeamFrame is the software framework developed within the SEAMLESS project. The models of FSSIM are developed in different programming languages (e.g. C#, Java and GAMS), while data are stored in relational databases. The architecture of SeamFrame is shown in Fig. 5.4. SeamFrame links the models to the data in the database and requires that models adhere to the ontology. The models are left in their original programming language and wrappers translate between the programming languages of the different models, the framework and the database (Fig. 5.4). A model wrapper provides the four functionalities. First, it wraps the model to a processing environment compliant interface and defines the exchange items (model inputs and outputs). Second, it initializes the model as component right after the start of the execution of the workflow. Third, it prepares for each run of the model dynamically the meta-models describing the model specifications (e.g. modules and equations to be used, sets definitions, how selected modules are structured, etc.). Fourth, it prepares the model input data in an exact format the model needs for each run of the model and retrieves model outputs of each run to be stored or communicated with other linkable model components. Although the architecture leaves the models relatively untouched, the models lose their direct link to the database or data-source. The development of the wrappers is a tedious, difficult and time-consuming task. Each wrapper is specific to a model and

MODELLING FRAMEWORK

Integrated database

WRAPPER 1

WRAPPER 2

MODEL 1

MODEL 2

Ontology

OpenMIComponent

Fig. 5.4 The modelling framework SeamFrame with the wrapped models referring to a common ontology and database schema

5 A Generic Farming System Simulator

125

therefore requires updates if the model is updated, which is difficult for maintenance. Such wrappers are not required for the FSSIM-AM models, as these have been developed in Javatm, which is also the programming language of SeamFrame. The Open Modelling Interface and Environment (OpenMI; Moore and Tindall 2005) was used in the SEAMLESS modelling framework to link the models at run time into a model chain. OpenMI is based on a pull-approach in which the last model in the chain pulls its outputs from other models in the chain by calling “getValues()”methods, which means requesting outputs from a model or data source. The model can set its outputs as inputs to other models through a “setValues()”-method. The model components (Fig. 5.4) are developed as OpenMI-components. If a model is wrapped, then the wrapper needs to be developed as an OpenMI-component, which implies that the models are not aware of OpenMI or affected by OpenMI. The definition of data exchanged in setValues() (e.g. inputs) and getValues() (e.g. outputs) forced modellers to be specific about the inputs and outputs of a model and facilitated linking of the models in a model chain. Wrapping the model as an (OpenMI) component facilitated the definition of models independently of each other, of data sources and of the graphical user interface. Model assumptions, interfaces and available data sources need to be clearly and explicitly specified, so that the models can be linked to each other. The FSSIM model interfaces in terms of inputs, outputs, states and parameters have been defined explicitly in an ontology, which is a collection of all concepts and relationships between concepts relevant to the domain (Antoniou and Van Harmelen 2004) and which functions as a dictionary. It sets up clear definitions for loosely integrating models in an open environment, facilitated by the knowledge manger for knowledge processing (such as reasoning and consistency checking) and by the domain manager (such as automatic generation of code templates for models and domain classes, accessing an instance of a domain class at runtime to supply the model component with the appropriate data).

FSSIM Application: Detailed and Simple Applications Through the first application of FSSIM to a few regions it appeared that the data requirements of the models (FSSIM/APES) are too high (i.e. good data on farm management are extremely scarce). For this reason, it was decided to allow two variants of FSSIM: one that uses detailed data an agro-management and one that uses less detailed (simple or summarized) data on agro-management. The simple version can be more easily used for a larger number of applications necessary for up-scaling to the EU level (cf. Pérez Domínguez et al. 2009), whereas the detailed application is useful for application to specific regions. The principal differences between the two versions of FSSIM are summarised in Table 5.4. The purpose of this section is to describe the results of a detailed and simple application of FSSIM to explain the followed procedure for running the model and to assess its capacity to reproduce the current situation and forecast the future.

126

K. Louhichi et al.

Table 5.4 FSSIM application to regions with detailed or summarized data availability FSSIM with detailed data FSSIM with summarized data − Use APES with observed input data − Use APES with generated input data − Use detailed survey − Use simplified survey − Includes current and alternatives activities − Includes current and alternatives activities − Use all FSSIM-MP modules − Use only some FSSIM-MP modules − Use semi-automatic procedure for calibration − Use automated procedure for based on risk and/or Positive Mathematical calibration based on risk or/and Positive Programming Mathematical Programming

Detailed Application of FSSIM FSSIM was tested for a range of detailed applications with the aim to analyse the current situation and to anticipate the impact of new, alternative scenarios and policy changes. In this chapter, results of Midi-Pyrénées (France) are presented as an example of the test application. An overview on the selected components, modules and calibration procedure used in the detailed application as well as the tested scenario is described (Fig. 5.5) below: –– Components: the selected components are: (i) the farm typology; (ii) the detailed computer-based survey for agro-management and FSSIM-AM; (iii) the biophysical model APES; and (vi) the mathematical programming model FSSIM-MP. –– FSSIM-MP modules: the selected modules are the crops, premiums, risk, PMP, perennial, policy and common modules. –– Calibration procedure: the calibration procedure is based on two steps: in the first step, we apply the risk approach in order to calibrate the model, as precisely as possible. The model assigns automatically a value to the risk aversion coefficient1 which gives the best fit between the model’s predicted crop pattern and the observed values. The difference between both values is assessed statistically by using the Percent Absolute Deviation2 (PAD). The aim of this step is to ensure that the model produces acceptable results before going to the second step.

The chosen value can vary from 0 to 1.65, as suggested by the literature. Percent absolute deviation (%):

1 2

n

PAD (%) =

∑ Xˆ i =1

i

− Xi .100

n

∑ Xˆ i =1

i

where Xˆi is the observed value of the variable i and Xi is the simulated value. The best calibration is reached when PAD is close to zero.

5 A Generic Farming System Simulator

127

To do this test, the following assumptions were used: if the PAD is less than 15% the model is acceptable and we can start the second step, if PAD is more than 15%, the model specification must be improved before applying the second step. In the second step, we apply the Positive Mathematical Programming according to the Röhm and Dabbert (2003) approach in order to calibrate the model exactly to the observed situation. –– Tested scenario: The policy test case is the integrated assessment of a trade liberalisation proposal by the so called G20 group of developing countries at the current Doha Round of the World Trade Organisation (WTO) (G20 2005). This proposal was based on the reduction of tariffs for agricultural products and abolition of export subsidises by the EU. This scenario was implemented at the market level (i.e. inside the market model, CAPRI) and the generated prices from CAPRI were used in FSSIM in order to analyse the impact of the price changes due to the liberalisation proposal at farm level. The policy case is illustrated with some economic indicators (farm income, production and premiums) and environmental indicators (nitrate leaching and soil organic matter) (Van Ittersum et al. 2008) (Fig. 5.5).

√

Midi-Pyrénées

Set Farm type and the corresponding data

√ Farm Type 1 Farm Type 2 ... √ Flevoland Crops

√

√ Set modules and linked constraints

√ √

Land Labour Set-aside ...

Perennial

√

Livestock Premium

√ Set calibration procedure

√

Risk approach Risk neutral Automate choose of Risk Aversion Coefficient ...

√ √

PMP approach Standard PMP approach Röhm and Dabbert PMP approach √ Kanellopoulos et al. approach

Set FSSIM scenario

√

Baseyear (2003)

√ √

Baseline (2013) WTO G20 proposal (2013)

Fig. 5.5 Modules and calibration procedure selected in the detailed application

128

K. Louhichi et al.

Table 5.5 Simulation of the WTO G20 proposal (policy scenario) for some economic and environmental aspects of the farm types of the Midi Pyrénées region Farm Type 1 Farm Type 2 WTO G20 Baseline WTO G20 Baseline proposal proposal Base year scenario Base year scenario (2013) (2013) (2013) (2003) (2013) (2003) % change % change to % change % change to baseyear baseline to baseyear to baseline Farm income (k€) 86.3 65.8 62.0 81.9 65.4 63.5 −24.9% −4.3% −20.1% −3.0% Premiums (k€) 39.9 29.9 29.8 35.3 27.4 27.4 −23.3% −0.1% −22.4% −0.1% Nitrate leaching 50.8 43.6 43.9 47.1 43.4 45.3 (kg N-NO3/ha) −14% 1% −8% 5% Soil erosion (t/ha) 2.0 1.9 1.8 2.9 3.3 2.7 −6% −7% 13% −19% 2.0 1.9 2.0 1.9 1.8 Pesticide use (kg/ha) 2.2 −7% −4% −3% −10%

Table 5.5 shows how two different farm types respond to the policy and baseline scenario, in comparison with the base year. Compared to the base year 2003, the farm income decreased in the baseline 2013 for the two farm types respectively with ca. 25% and 20%, mainly because of reduction of premiums. The environmental impacts in terms of nitrate leaching, soil erosion and pesticide consumption (average at farm level weighed by area per crop) seem positive due to the drop in the area devoted to cereals (mainly durum wheat and irrigated maize) and the increase in area of protein crops which are more efficient from an environmental point of view. The policy scenario tested in this example had a modest impact on farm income and nitrate leaching, in comparison with the baseline scenario, due to the limited impact of the policy proposal (G20) on the price of the major arable products as simulated by the market model.

Simple or Summarized Application of FSSIM To enable upscaling of farm type analysis to the EU through the assessment of price-supply relationships (Pérez Domínguez et al. 2009), FSSIM is used with simple or summarized data. The data needs of FSSIM for the simple survey regions should be restricted to what is available in EU wide databases and the simple survey of current activities which was conducted within SEAMLESS to identify the currently used activities and the corresponding technical coefficients. The purpose of this section is to describe a simplified version of FSSIM that can be used for analysis at EU level and to illustrate the type of analysis by presenting some preliminary results from application to Flevoland.

5 A Generic Farming System Simulator

129

An overview on the components, modules and calibration procedure of the summary version of FSSIM-MP used for EU25 level analysis is described below: –– Components: the selected components are: (i) the farm typology; (ii) the simple computer-based survey; and (iii) the mathematical programming model FSSIM-MP. –– FSSIM-MP modules: the selected modules are the crops, premiums, PMP, trend, policy and common modules. –– Calibration procedure: The calibration procedure used in the detailed version of FSSIM-MP is also used here with minor adjustments. –– Tested Scenario: The model was calibrated for the base year and used to predict changes in the baseline scenario. Sensitivity of crop product quantities to prices changes was simulated to assess price supply relationships at higher levels. The price of each crop product was changed iteratively to 60%, 80%, 120% and 140% of the original price keeping the other product prices constant. The effects on supply were assessed in each iteration. Running FSSIM for all farm types of the regions with summary information on agro-management requires some adjustments of FSSIM to restrict the data requirements to what is available in FADN and the simple survey. Those adjustments are: 1. In the first phase of PMP the observed crop levels are used as upper bounds to the added calibration constraints. In FADN there is no information on single crop levels, instead there is information on groups of crops (e.g. fresh vegetables which refers to the area of a number of crops such as area of onion, carrot and cabbage). In the detailed version of FSSIM, expert knowledge is used to transform the observed levels of FADN crop groups to observed levels of single crops. Finding experts in all sampled regions would be a resource demanding process. To avoid this process it was decided to evaluate and calibrate the reduced version of FSSIM directly on FADN crop groups. 2. In some cases, the observed cropping pattern of some farm types included crops that are not part of any rotation identified in the simple survey. This implies that it is not possible to simulate such crops. To avoid this problem it wad decided to treat the area of these crops as fixed land and it was subtracted from the total available farm land. 3. Finally, in order to ensure that there is at least a linear combination of activities that reproduces the observed cropping pattern, we decided to include some mono-crop activities, defined as rotations of a single crop; this is justified because this represents rented land (e.g. from dairy farms) on which specific crops are grown for 1 year. An example of the sensitivity analysis of prices performed for one of the farm types in Flevoland is presented in Table 5.6. Note, that spring wheat substitutes winter wheat when prices of spring wheat are high or those of winter wheat are low. These farm level results are used to estimate price-supply relationships at regional level and subsequently they will be extrapolated with advanced econometric procedures (EXPAMOD) to non-sampled regions (Pérez Domínguez et al. 2009).

130 Table 5.6 Simulated supply response (tonnes per farm) to price changes for Flevoland Supply response (tonnes) Maize Sugar Price (€/t) (silage) Onions Potatoes beet Maize 21 23 671 1,219 628 (silage) 28 29 671 1,217 626 35 35 670 1,216 624 41 41 669 1,215 622 48 47 669 1,214 620 Onions 66 39 526 1,237 660 89 37 598 1,227 642 111 35 670 1,216 624 133 33 742 1,206 606 155 31 814 1,195 588 Potatoes 45 41 684 1,055 672 60 38 677 1,135 648 75 35 670 1,216 624 89 33 663 1,297 600 104 30 655 1,376 585 Sugar beet 27 36 673 1,222 585 36 36 673 1,222 585 46 35 670 1,216 624 55 33 662 1,201 724 64 30 655 1,187 824 Wheat spring 86 35 670 1,216 624 115 35 670 1,216 624 144 35 670 1,216 624 173 35 670 1,216 624 202 33 662 1,201 599 94 38 677 1,231 649 Wheat winter 125 38 677 1,231 649 156 35 670 1,216 624 187 32 661 1,199 594 218 28 651 1,178 585

K. Louhichi et al. farm type 3303 in

Wheat spring

Wheat winter 132 131 131 130 129 143 137 131 124 118 147 139 131 122 113 134 134 131 122 113 131 131 131

119 126 111 111 131 140 148

Conclusions This chapter presented a detailed description of the bio-economic farm model (FSSIM), especially its specifications, structure, model linking and component integration. The original contributions of FSSIM to bio-economic farm modelling are the integrative approach, the modular setup and the generic features. The integrative approach of FSSIM makes the complex relationship between biological processes and economic decisions more transparent and allows a correct integration of technical economic and environmental issues, enabling the simulation of the different type of policies. The multidisciplinary framework facilitates synthesis of scientific knowledge in the domain of agriculture and its environment. The generic features allow the

5 A Generic Farming System Simulator

131

application to the broad variation of farming systems inside and outside the EU. The modular setup provides the possibilities to activate and de-activate modules depending on regions and conditions, to consider different types of policy instruments (subsidies, regulation, taxes, etc.) and to choose different methodological approaches, that are consistent with the data availability for a specific application. This includes different approaches concerning the representation of risk, different calibration approaches and different representations of agricultural activities. FSSIM targets to be applied by different types of users such as (i) researchers with the purpose of testing different approaches; (ii) policy experts having the purpose of making ex-ante assessment of policies; and (iii) other stakeholder groups with the purpose to anticipate the effect of new policies.

References Alterra & INRA. (2005). New soil information for CGMS (Crop Growth Monitoring System) (SINFO). In Alterra – INRA (pp. 219–260). Wageningen: Author. Andersen, E., Elbersen, B., Godeschalk, F., & Verhoog, D. (2007). Farm management indicators and farm typologies as a basis for assessment in a changing policy environment. Journal of Environmental Management, 82, 353–362. Antoniou, G., & van Harmelen, F. (2004). A semantic web primer. Cambridge, MA/London: MIT Press. Deybe, D., & Flichman, G. (1991). A regional agricultural model using a plant growth simulation program as activities generator. Agricultural Systems, 37, 369–385. Dogliotti, S., Rossing, W. A. H., & van Ittersum, M. K. (2003). ROTAT, a tool for systematically generating crop rotations. European Journal of Agronomy, 19, 239–250. Eurostat. (2007). Office statistique des Communautés européennes, juin 2007. Falconer, K., & Hodge, I. (2000). Using economic incentives for pesticide usage reductions: Responsiveness to input taxation and agricultural systems. Agricultural Systems, 63, 175–194. G20. (2005). G20 proposal on market access – October 12, 2005, from http://www.g-20.mre.gov. br/conteudo/proposals_marketaccess.pdf Hazell, P. B. R., & Norton, R. D. (1986). Mathematical programing for economic analysis in agriculture (p. 400). New York: Macmillan. Heckelei, T., & Wolff, H. (2003). Estimation of constrained optimisation models for agricultural supply analysis based on generalised maximum entropy. European Review of Agricultural Economics, 30(1), 27–50. Hengsdijk, H., & van Ittersum, M. K. (2002). A goal-oriented approach to identify and engineer land use systems. Agricultural Systems, 71, 231–247. Howitt, R. E. (1995). A calibration method for agricultural economic production models. Journal of Agricultural Economics, 46(2), 147–159. Janssen, S., Andersen, E., Athanasiadis, I., & Van Ittersum, M.K. (2009). A database for integrated assessment of European agricultural systems. Environmental Science and Policy, 12(5), 573–587. Janssen, S., & Van Ittersum, M. K. (2007). Assessing farm innovations and responses to policies: A review of bio-economic farm models. Agricultural Systems, 94, 622–636. Kanellopoulos, A., Berentsen, P.B.M., Heckelei, T., Van Ittersum, M.K., Oude Lansink, A.G.J.M. (2010). Assessing the forecasting performance of a generic bio-economic farm model calibrated with two different PMP variants. Journal of Agricultural Economics, under review. Kruseman, G., & Bade, J. (1998). Agrarian policies for sustainable land use: Bio-economic modelling to assess the effectiveness of policy instruments. Agricultural Systems, 58, 465–481.

132

K. Louhichi et al.

Louhichi, K., Alary, V., & Grimaud, P. (2004). A dynamic model to analyse the bio-technical and socio-economic interactions in the dairy farming systems on the Réunion Island. Animal Research, 53, 1–19. Moore, R. V., & Tindall, C. I. (2005). An overview of the open modelling interface and environment (the OpenMI). Environmental Science and Policy, 8, 279–286. Pérez Domínguez, I., Bezlepkina, I., Heckelei, T., Romstad, E., Oude Lansink, A., & Kanellopoulos, A. (2009). Capturing market impacts of farm level policies: A statistical extrapolation approach using biophysical characteristics and farm resources. Environmental Science and Policy, 12(5), 588–600. Reinds, G.J., & Van Lanen, H.A.J. (1992). Crop production potential of rural areas within the European Communities. II. A physical land evaluation procedure for annual crops and grass. In Publication of the Scientific Council for Government Policy. The Hague, The Netherlands: Scientific Council for Government Policy. Röhm, O., & Dabbert, S. (2003). Integrating agri-environmental programs into regional production models: An extension of positive mathematical programming. American Journal of Agricultural Economics, 85(1), 254–265. Russell, G. (1990). Barley knowledge base. An agricultural information systems for the European Community. Brussels: Commission of the European Community. Thompson, A. M. M. (1982). A farm-level model to evaluate the impacts of current energy policy options. Canterbury: Lincoln College. Van Ittersum, M. K., Ewert, F., Heckelei, T., Wery, J., Alkan Olsson, J., Andersen, E., et al. (2008). Integrated assessment of agricultural systems – A component-based framework for the European Union (SEAMLESS). Agricultural Systems, 96, 150–165. Van Ittersum, M. K., & Rabbinge, R. (1997). Concepts in production ecology for analysis and quantification of agricultural input-output combinations. Field Crops Research, 52, 197–208. Wolf, J., Boogaard, H.L., & Van Diepen, C.A. (2004). Evaluating crop data in CGMS system. Crop data for winter wheat and spring barley (MARSOP-2 Project Rep. No. 1). Wageningen: Alterra. Wossink, G. A. A., De Koeijer, T. J., & Renkema, J. A. (1992). Environmental-economic policy assessment: A farm economic approach. Agricultural Systems, 39, 421–438. Zander, P., Borkowski, N., Hecker, J.-M., Uthes, S., Stokstad, G., Rørstad, P.K., Bellocchi, G. (2009). Procedure to identify and assess current activities. SEAMLESS deliverable PD3.3.9, SEAMLESS Integrated Project, EU 6th Framework Programme, contract no. 010036-2, www. SEAMLESS-IP.org, 124p.

Chapter 6

Visualising Changes in Agricultural Landscapes Sébastien Griffon, Daniel Auclair, and Amélie Nespoulous

Introduction Rural land managers, foresters and farmers, but also local decision makers, local authorities and members of local governments, are increasingly aware of the necessity to take into account the perception of the landscape by the general public, and to predict the evolution of landscapes according to management decisions (Bergen et al. 1995; Bell 2001). Different management choices can lead to similar, or to very different landscapes. The positioning of woodlots, of fields, and of agroforestry areas, the type of silvicultural management (selective or systematic thinning, artificial pruning, clear-cut or shelterwood systems, reforestation, choice of species, etc.) or agricultural system (rotation, land attribution, crop allocation, etc.) and the balance between forest and agriculture, are susceptible of drastically modifying the visual aspect of the landscape. The public generally considers the landscape as timeless, and resistance to any change is often very great. Forest management can arouse public antipathy and in some cases can lead to outspoken criticism. Conversely, a well thought out management plan can considerably improve the visual aspect of a rural landscape (Savill et al. 1997). Although land managers generally have a good experience of what result can be expected from their decisions, they are often faced with difficulty when trying to communicate the visual impact of a future management option to all the stakeholders (local and regional decision-makers, land managers, landscape planners, and various communities involved in outdoor activities). The perception and aesthetic evaluation S. Griffon (*) CIRAD, UMR AMAP, Montpellier, F-34000, France e-mail: [email protected] D. Auclair INRA, UMR AMAP, Montpellier, F-34000, France e-mail: [email protected] A. Nespoulous CNRS, UMR CEFE, Montpellier, F-34000, France e-mail: [email protected] F.M. Brouwer and M. van Ittersum (eds.), Environmental and Agricultural Modelling: Integrated Approaches for Policy Impact Assessment, DOI 10.1007/978-90-481-3619-3_6, © Springer Science+Business Media B.V. 2010

133

134

S. Griffon et al.

of a landscape can be analysed according to three organisation levels: an objective dimension (formal landscape organisation, proportion, composition rhythm of various biophysical elements), a cultural dimension (related to social groups, to history, to representations of nature), and an individual subjective dimension (Sauget and Depuy 1996). Some socio-economists have undertaken to assign values to the aesthetics of a landscape (Thomas and Price 1999), but this subjective dimension generally leads land managers to seek a consensus between stakeholders, by negotiating around some kind of visual representation (Tyrväinen and Tahvanainen 2000). Representing a landscape has always been a difficult task. As early as the sixteenth century, “bird’s-eye” views sought to combine two-dimensional maps with a representation of perspective. Nowadays, land managers have a number of tools at their disposal (Perrin et al. 2001). –– Maps and plans have often been – and still are – used for their rigour and for the possibility of quantification they offer. –– Their use is now greatly facilitated by geographic information systems (GIS). These computer tools include georeferenced databases, which help to produce two-dimensional maps. They contain large amounts of information, which can help to represent land-use and its evolution. GIS manufacturers provide three-dimensional visualisation techniques, which however remain too restrictive to represent satisfactorily large landscapes: the terrain is generally well represented, but particular landscape features and architectural elements are often simply extruded, and do not represent real volumes. –– Photographs can be used directly, or can be modified with the help of computer-aided imagery. This technique has been used for example by Tress and Tress (2003) to discuss various scenarios with stakeholders. It is however very time-consuming to build different scenarios of future land-use. –– Schematic representations can be produced simply by drawing sketches or diagrams, either by hand or with the help of computer systems. –– Photographs and drawings have no direct link with maps, and only show a limited number of viewpoints. They are purely visual techniques, which cannot easily simulate the reactions of a landscape to human interventions. –– Virtual imagery is a modern tool which can help represent landscapes, in three dimensions, and which can also provide the possibility to include a dynamic aspect. Three-dimensional visualisation of the landscape provides means that are better understood than maps, especially for the general public. With such methods, visual changes of the landscape can be shown very impressively, which can allow for an intuitive assessment of the visual landscape quality. Static, web-based landscape visualisation tools have made considerable progress in recent years, such as for example Google Earth (http://earth.google.com/), covering the entire planet in 3D. The French geographic service proposes Géoportail (http://www.geoportail.fr/), a tool which couples 3D visualisation with a number of additional GIS layers, such as high quality maps, as well as extruded buildings. A number of companies offer facilities for creating and customizing maps and virtual visits, such as the Microsoft product Virtual Earth 3D, a plug-in for 3D

6 Visualising Changes in Agricultural Landscapes

135

visualisation (http://www.microsoft.com/virtualearth/default.aspx) or Skyline Globe http://www.skylineglobe.com. Google Earth offers facilities with Google Earth Outreach (http://earth.google.com/intl/en_uk/outreach/index.html) and SketchUp (Google Inc 2006). A “Google Earth Community” (http://bbs.keyhole. com/ubb/categories.php/Cat/0) and several international societies have appeared (for example the International Society For Digital Earth http://www.digitalearth-isde. org/). Examples are shown below: Fig. 6.1 shows examples of 2D visualisation with the French Geoportail and with Google Earth for two different situations, and Fig. 6.2 shows details for Flevoland in 2D and in 3D. Figure 6.3 shows an example of 3D view of a coastal village, with 3D extruded buildings (Géoportail). The examples show static views based on aerial (satellite) photographs, at a specific date, but they are not dynamic. The challenge in SEAMLESS was to enable visualising future changes in land use, according to scenarios. It is also expected to produce more details than can be seen in Figs. 6.1–6.3. In the last 20 years, research on 3D landscape modelling has increased considerably, mainly for urban planning, but also for rural and forest landscapes (Danahy 1989; Auclair et al. 2001a; b; Bishop et al. 2001; Lovett et al. 2001; Orland et al. 2001; Herwig and Paar 2002; Snyder 2003; Dockerty et al. 2005; Paar and Clasen 2007; Lange et al. 2008). Several international meetings have addressed the issue

Fig. 6.1 (a) Example of a landscape viewed from above: the Restinclières agroforestry estate (France) viewed with the French Geoportail. (b) Example of a landscape viewed from above: the Flevoland area (The Netherlands) viewed with Google Earth

136

S. Griffon et al.

Fig. 6.2 Details of the Flevoland area (The Netherlands) viewed from above in 2D (left) and in 3D (right), using Google Earth. Note the difference in colour for some of the details, which is due to different dates for the aerial photographs

Fig. 6.3 Example of a 3D view of a coastal village (Collioure, France), with 3D extruded buildings (Géoportail)

of visualising landscape, for example “Our Visual Landscape” (Ascona, Switzerland, 1999) or “Futurescapes” (Belfast, UK, 2002), resulting in special issues of the journal Landscape and Urban Planning (Lange and Bishop 2001; Lovett 2005). Specific software have been developed (Perrin et al. 2001; Bishop et al. 2005), which increasingly benefit from recent advances in technology, such as visioning hubs or immersive tools (Sheppard 2006; Salter et al. 2009) or augmented reality (Ghadirian and Bishop 2008). Some freeware and/or open-source software have been developed in addition to those proposed by commercial companies (for example Geomantics in the UK which proposes freeware versions or more complete

6 Visualising Changes in Agricultural Landscapes

137

professional versions, http://www.geomantics.com/ or LandSIM3D® commercialised by Bionatics in France http://www.bionatics.com/). Many are linked to GIS, and have been described in several literature reviews (Ervin and Hasbrouck 2001; Appleton et al. 2002; Pettit et al. 2008). Often specific 3D objects must be included in the scene, such as vegetation (Auclair et al. 2001a; b; Muhar 2001), and several approaches have been described in order to integrate indicators into landscape visualisations (Bishop et al. 2008; Wissen et al. 2008). A particular emphasis has been put on reliability and validation of landscape simulation (Sheppard 1982; Daniel 1992; Lange 2001; Bishop et al. 2005). Several European research projects are involved in the issue of landscape visualisation, such as VisuLands (Wissen et al. 2008), Greenspace (Lange et al. 2008), SENSOR (Helming et al. 2008) and BioScene (Soliva and Hunziker 2009). These projects however address different issues, they are in particular concerned by urban landscapes and communication infrastructures, at country or regional scales. Our aim here was to concentrate on detailed agricultural land-use at the scale of a territory, it therefore appeared necessary to develop separate components, although on a common basis. In the present chapter, we describe a landscape visualisation freeware component, which is part of the SEAMLESS project (Van Ittersum et al. 2008). This component can be launched at the end of a modelling chain destined to simulate a particular agricultural and/or environmental policy, to allow for exploration of landscape changes. Visualisation could have a significant implication for the choice of effective land-use policy, and could be used as a basis for discussion and negotiation within the community. The pressures causing changes in landscape can be simulated by a bio-economic farm model, such as described in Chapter 5 of this volume. This can then be translated into changes in the spatial configuration of the landscape. The mapped results (environmental data such as land cover and land use) will be specific to each individual region, and should be available from a GIS database. They will be used here to compute and visualise a 3D scene. In this chapter we first describe briefly the software methodology and design and the data processing, and in a second phase provide an example of application based on a study of four scenarios in the Pic Saint Loup area of the French Mediterranean region.

Software Methodology and Design To build a system that computes a virtual landscape from accurate spatial data and delivers a reasonable realistic representation of an existing landscape, a large amount of data is necessary. Firstly, we need to build a real-time renderer that is powerful enough to handle large datasets of landscape terrain. Secondly, a matching virtual representation of an existing area has to be constructed, including vegetation and man-made structures. We have built a system using a free graphic engine as a

138

S. Griffon et al.

base for building a virtual environment. To obtain a suitable 3D dataset of an area of interest, users of the system can select an area on a 2D digital map in a GIS and the application sends the matching data to the landscape visualisation component. The generation of the 3D model and its rendering is done by this software component, that we have called here SLE. Thanks to this virtual environment, a post-model analysis can be conducted to address issues specific of the landscape, with an objective of participative planning and negotiation with the various stakeholders. The challenge of the SLE software is, without proprietary tools or database preparation, to extract specific data (elevation, imagery and land cover) from a GIS, fuse this data in a procedural manner to enhance its apparent quality, and add vegetation objects to the scene.

Existing Software Tools Today, 3D landscape visualization software such as Vue® (E-on Software, http:// www.e-onsoftware.com/), or World Construction Set® (3D Nature, http://3dnature. com/) are available. They provide the possibility for modelling and rendering landscapes with a high degree of realism, but are specifically designed for infographists. Since 2000, Visual Nature Studio® (3D Nature) is a GIS-compatible version of World Construction Set® and in 2003; an add-on has been released that introduced real-time capability on a lower level of detail. Furthermore, our partner Bionatics (http://www.bionatics.com/), the specialist in 3D plant modelling and landscape design has released LandSIM3D®, a simulation and design software for quickly modelling existing landscape in 3D from GIS data sources. It permits to import a project, visualize its integration in a real site, study its alternatives, its future evolution and decide for the best options. These software tools are actually used for landscape planning. However, these are commercial software and not designed especially for research scientists. That is why we developed a free GIS-compatible and interactive landscape visualisation software, based on available open-source components.

Data Needs The representation of 3D landscape models requires a variety of components and corresponding spatial data types. These include terrain texture (orthoimagery, raster maps), digital height models (DTM, digital terrain model or DEM, digital elevation model, a digital representation of ground surface topography or terrain; or DSM, digital surface model, a topographic model of the reflected surface of the earth that can be manipulated by computer programs), vector-based 2D geo-objects, 3D objects, object textures, animations and hyperlinks. The spectrum of these components ranges from very large spatial objects to large numbers of complex and

6 Visualising Changes in Agricultural Landscapes

139

possibly dynamic 3D objects. These data types have very different characteristics and requirements in terms of management, visualisation and multi-scale representation. Digital Terrain Elevation Data (Raster) At present, available through many databases, the GTOPO30 elevation data (around 1:1,000,000 scale) (http://edc.usgs.gov/products/elevation/gtopo30/gtopo30.html) are not accurate enough. For testing, we have used the NASA Shuttle Radar Topographic Mission (SRTM) data. This data is currently distributed free of charge by USGS and is available for download from the National Map Seamless Data Distribution System, or the USGS ftp site (http://glcfapp.umiacs.umd.edu:8080/ esdi/index.jsp). The SRTM data is available as three arc second (~90 m resolution) DEM. A one arc second data product was also produced, but is not available for all countries. The vertical error of the DEM is reported to be less than 16 m. We need at least these kinds of scale datasets but more accurate data like the IGN® BD ALTI® (50 m resolution, https://professionnels.ign.fr/ficheProduitCMS. do?idDoc = 5323461) gives better results. Land Cover Classification Data (Raster or Vector) For vegetation placement, we can use self-made scientist datasets and/or CORINE Land Cover like datasets (http://dataservice.eea.europa.eu/dataservice/metadetails. asp?id = 822). Land use and land cover maps show areas of land as ‘parcels’ or polygons. Each parcel has attached to it a list of values or attributes, covering such topics as land cover class, parcel area, length of boundary, processing history, knowledgebased correction and identification of the original satellite scene. We need these kinds of data to display actual and simulated landscape, i.e. the environmental impacts of a simulated scenario in SEAMLESS need to be translated into land cover data. Geotypical Textures Library (Raster) For texture splatting (a method for combining different textures) on the terrain, we need a collection of textures for the different land cover classes we could find (e.g. wheat, corn and grassland). A preliminary library of textures is being developed, and if necessary, according to the user requirements, a variety of different textures can be added. Orthophoto (Raster) For rendering more realistic landscape, an orthophoto (an aerial photograph that has been geometrically corrected such that the scale of the photograph is uniform) can be draped over the terrain and blended with the geotypical soil textures.

140

S. Griffon et al.

Vegetation Model Library Vegetation elements, especially trees, can render the landscape more realistically than simple 2D textures. Producing high quality 3D vegetation models is a very time-consuming and costly operation, and it is not yet planned to create such a library for vegetation objects. As a viewer moves away from the object, we need the model to switch from a geometric representation to a cross-polygon model, and to a billboard (one or a combination of several two-dimensional images), such as presented in Fig. 6.4. Building a variety of realistic plant models at multiple levels of detail could be accomplished using a package from our commercial partner (i.e. REALnat® by Bionatics® http://www.bionatics.com/). An example is presented in Fig. 6.5. Another solution is to represent less realistic plants using basic primitives such as cone, cylinder, sphere or convex hull (Fig. 6.5). At present, SLE is available only with a small number of simplified samples of AMAPsim files (full detailed 3D tree models) and free 3D models from 3D-Diggers (http://www.3d-diggers.de). Furthermore, we have planned to build a content library based on existing vegetation textures and models. This library should eventually include some major European species and be linked to the land cover classes. Such a library should also offer the possibility for the user to include additional textures according to land-use and vegetation not accounted for in the first versions.

Fig. 6.4 Two level of detail plant models from REALNat® by Bionatics®. A high level of detail is necessary when visualizing the object from a nearby position (left) and a lower level of detail when visualizing it from afar (right)

6 Visualising Changes in Agricultural Landscapes

141

Fig. 6.5 Examples of a very detailed plant from the Bionatics® library (left) and simple plant forms (right) built under the GEOM module of AMAPmod (Boudon et al. 2001), now integrated as PlantGL in the ALEA platform (Pradal et al. 2004) http://gforge.inria.fr/projects/openalea/

Data Handling and Processing The system described here is able to automatically generate suitable 3D content models from spatial 2D data. SLE users have the option to load geospatially referenced data resulting from different policies and select an area of interest for the visualisation. Ideally, this should eventually be possible anywhere in Europe, if the appropriate data is made available. To manipulate these data we propose the user friendly Open Source Quantum GIS (QGIS, http://www.qgis.org/). A 3D data conversion tool has been developed and integrated into QGIS. It works as an external module of the visualisation one and converts raster layers and shape files into a format the renderer can read. The 3D visualisation module has been developed as a stand-alone model. It can therefore run independently from other modelling components and be launched by the GIS. Currently, the link with bio-economic models is done by database server connection. In Fig. 6.6, we describe the flow of data through this spatial database and SLE. Land-use distribution can be downloaded from the POSTGRES spatial database by QGIS. More accurate data such as field patterns and digital elevation model must be also loaded (from the local user hard drive or from a server) to allow SLE visualisation. The SLE QGIS plug-in can allocate each agricultural parcel to a specific land-use. This will be done by importing the proportion of each land-use class computed by the bio-economic farm model and distributing it on the field pattern according to specific allocation rules. At present the rule implemented in the model is simply random allocation of changes, but this should be improved in future versions. Then the plug-in can export any extent of up to 10 by 10 km selected by users.

142

S. Griffon et al.

POSTGRES Database SERVER Spatial database

Quantum GIS Land-use Additional GIS data : e.g DTM, field shapes, Orthophoto

SLE Exporter Plug-In

Project xml file + Formated data Project configuration file Camera bookmarks file Video path bookmarks file

SLE Editor

Renderer

Textures and 3D objects (plants, trees, houses, etc)

Real-time rendering Screenshots Videos

Fig. 6.6 The SLE diagram. The results from the bio-economic farm model drive the land-use

Data are cropped, fused and formatted to be visualized. An xml project file with geo-reference and land-use information is also written. SLE is then launched and the user can edit the geo-typical configuration (texture and vegetation) for each land-use class. In practice, the land-use geo-typical parameters consist of the following items: • A tileable texture representing typical ground cover. This is used to build the texture that is draped over the terrain when rendering. A number of examples are shown in Fig. 6.7 • A list of possible objects in the land use type defined by the following parameters: –– The type of 3D objects: it is possible also to define objects like rocks, as long as they have a natural distribution. –– The type of spatialization: the object’s distribution over the terrain can be done randomly or by rows.

6 Visualising Changes in Agricultural Landscapes

143

Fig. 6.7 Examples of texture splatting; top: dense and open Mediterranean forest, garrigue and wheat crop; middle: bare soil and different grasslands patches; bottom: orthophoto (left) and virtual image computed according to the land-use map of the same area (right)

–– The level of detail: the object can be displayed in 2D (billboards) or in 3D according the distance to the camera. –– Object density: this is used in conjunction with a random offset to determine the number of objects in a specific area. –– Inter-row/Intra-row spacing: this is used in conjunction with row spatialization to specify the distance between objects. –– Scale variation information: this is used to generate variations in the appearance of objects, and is particularly useful when billboards are used to render trees. Once the configuration has been defined, the terrain mesh, the textures and the vegetation objects can be computed and the user can explore the landscape in real-time. It is possible to bookmark some interesting points of view or camera paths, take screenshots, modify the terrain, add 3D objects manually and modify the rendering. Several “test case landscapes” are now under study. The option which is at present being investigated is to select only a small number of representative

144

S. Griffon et al.

landscapes for which sufficient spatially explicit data are available, and concentrate on these. One important issue concerns the way to disaggregate and make data which have been aggregated for running farm models spatially explicit, by defining the above-mentioned land-use allocation rules.

Data Rendering Technically, a dynamically optimized elevation mesh from the digital terrain elevation raster is first computed using the Geographic MipMaps technique (MipMaps are pre-calculated, optimized collections of bitmap images that accompany a main texture, intended to increase rendering speed and reduce artifacts, de Boer 2000). Then the mesh is textured with the texture splatting technique (Bloom 2000; Tyrväinen and Tahvanainen 2000) and with satellite imageries or thematic maps. Texture splatting means that from a set of appearance parameters, a selection of tilable textures is blended together and then splat onto the surface. In Fig. 6.7, we can see four possible types of ground cover for landscapes (top). These are large scale layer ground covers and were obtained from aerial photographs for two types of Mediterranean forest, “garrigue” (Mediterranean scrubland) and wheat crop. The middle four textures define ground cover on small scale showing bare soil and different grasslands patches. At the bottom, we compare an orthophoto of the Pic Saint Loup (South of France) with a virtual image computed with texture splatting according to the land-use map. The number of 3D objects on the landscape is very important. For example, if the landscape represented is 1 km2 and with an average density of one object every 10 m2 we would have 100,000 objects on the landscape. This is more than many systems can handle in real-time. It is therefore necessary to give some form of organisation to the 3D scene. Thus, we established a fixed grid around the camera to manage the vegetation data for each layer of plants and other natural objects. Each grid cell contains all of the data to render its layer in the physical space it occupies. For each layer, we establish a distance from the camera that the layer needs to generate visuals; this determines the size of our virtual grid. This operation is done in real time and care must be taken to ensure that planting is a fast operation. The different layers of vegetation consist in trees, shrubs, small plants, rocks, and other small objects to complete the illusion of natural complexity. We apply random transforms to vary their size and orientation as we pick our planting points. Some of these can be represented as 2D textures on 3D planes (billboards) just as grass is, but the richness of the environment is enhanced when we mix in an assortment of geometric objects as well. An extension to the “billboard” technique has been developed, the impostor rendering (Day and Willmott, 2005). Impostors are simply dynamic billboards; this means that the texture is updated dynamically according to the viewing angles, so as to reduce the visual error incurred by

6 Visualising Changes in Agricultural Landscapes

145

Fig. 6.8 Impostors of a 3D AMAP plant (Pinus halepensis): the same 3D plant is visualized here from 16 different directions

using a flat object (Fig. 6.8). Finally, the user can activate options of rendering to add realism and immersion to the landscape such as: sky, clouds, sun, haze, shadows, lake and stereo-vision.

Example of Application In this chapter, we present an example of SLE visualisation for a landscape in the South of France, resulting from different scenarios (Nespoulous 2004).

The Situation For centuries, the North of the Mediterranean basin has been characterized by complex and quick changes. These changes mark a break with the progressive variations which took place since the Neolithic and contributed to manufacture the landscape in mosaic which is considered as typical for the North of the Mediterranean basin. The nature of these changes affecting biodiversity of the Mediterranean basin raises the question of the impact of human activities on biodiversity, which is a central stake in the development of territories (Cheylan and Gumuchian 2002). The studied zone is the Pic Saint Loup region, located in the department of Hérault at about 20 km to the Northwest of Montpellier (Fig. 6.9). This territory is characterized by the mountain of the Pic Saint Loup reaching the altitude of 658 m.

146

S. Griffon et al.

Fig. 6.9 Localization of the Pic Saint Loup area (France)

The natural vegetation is dominated by garrigue, the typical Mediterranean shrubland ecosystem characterised by discontinuous bushy associations of calcareous plateaus, often composed of kermes oak, lavender, thyme, and white cistus, with isolated trees (holm oak and juniper). Agriculture is dominated by pasture, mainly in silvopastoral systems, olive groves and vineyards, with occasionally agricultural crops (wheat and barley).

The Scenarios A detailed study of the past and present land-use on the territory of Pic Saint Loup, based on aerial photography in years 1946, 1962, 1981, 1992 and 2002, showed an important impact of peri-urban development on biodiversity (Sirami et al. 2007, 2008). Following these results, four scenarios were set up, as part of a participatory process for planning the regional peri-urban and agricultural policy, in an area dominated by the typical culturally sensitive Mediterranean garrigue shrubland. They can be summarized according to two axes (Fig. 6.10). The abscissa concerns “nature”, which is more or less rich in biodiversity, corresponding to the environmental quality of natural areas, and the ordinate fluctuates between urban and rural, which determines the type of urbanization and way of life favoured in each scenario (Nespoulous 2004). Land-use maps were produced for each scenario and the resulting scenes were processed with SLE. Two representative viewpoints are presented here

6 Visualising Changes in Agricultural Landscapes

147

Nature +

Rural

Urban

Nature -

Fig. 6.10 Four scenarios of the Pic Saint Loup area in 2030

(Figs. 6.11, 6.12, 6.13, and 6.14), one general aerial view of the landscape (top), and one characteristic view dominated by the Pic Saint Loup mountain (bottom). Scenario 1: Biodiversity by Agriculture This first scenario arises from the awareness that high quality agriculture contributes to the preservation of biodiversity and Mediterranean landscapes. This awareness is supported by the collapse of the international transport, which involves a relocation of agricultural production. In this context, the balance between the place of residence and the place of work is declared as a major target for town and country planning. Therefore, the growth of periurban villages is self-restrained and new housing settlements are built close to the existing habitat, to consume as little space as possible. The traditional agricultural activities, of gathering, viticulture and olive-growing reinvest the garrigue. The forest is grazed and cut for wood energy production. As a consequence, the landscape is opened up by agricultural activities, but the diversity of practices leads to the preservation of an environmental mosaic. Scenario 2: A Green City in a Mediterranean Forest In this second scenario, the extension of large cities such as Montpellier reaches increasingly distant villages. However, public policies structure this urban growth

148

S. Griffon et al.

Fig. 6.11 Biodiversity by agriculture – aerial view (top) and close-up (bottom)

and are very sensitive to the value of the natural environment. The natural areas around cities are frequently used for outdoor leisure activities, such as climbing, hiking and mountain biking. Therefore, the expansion of villages takes place close to historical centres and alongside the main highways, in harmony with nature protection, according to environmental models of land management. Agriculture no longer belongs to the garrigue; it is limited to a few remnants of historical vineyards producing high quality wine. Landscapes are thus very wooded, as the natural value put forth is the Mediterranean forest.

6 Visualising Changes in Agricultural Landscapes

149

Fig. 6.12 A green city in Mediterranean forest – aerial view (top) and close-up (bottom)

Scenario 3: Urban Pressure For this third scenario, urban development is constantly increasing, with no regulation by land-use planning and development. Urban growth takes place along the highways and close to the existing urban centres, on a larger scale than for the second scenario. In this context, it is important to develop highways between the main city and its periphery. Leisure activities are very important and recreation areas are created to the detriment of the garrigue. Agriculture is also present and is

150

S. Griffon et al.

Fig. 6.13 Urban pressure – aerial view (top) and close-up (bottom)

based on two different models. On the one hand, there are high income activities, based on the reputation of the high quality local vineyards, and on the other hand there is subsistence agriculture around settlements. Scenario 4: The Garrigue After the Energy Crisis With the collapse of the fossil fuel system, populations have to find alternative, sustainable energy. A direct consequence of this crisis is the increase of transport

6 Visualising Changes in Agricultural Landscapes

151

Fig. 6.14 The garrigue after energy crisis – aerial view (top) and close-up (bottom)

costs and this therefore requires re-concentrating the production at the local level. Thus, the life in the garrigue falls back locally and concentrates on the production of energy as well as subsistence crops. The landscape is then largely agricultural and open, composed of large open fields, to acquire a better financial return. In the hills, the landscape mixes small patchy mosaics with large logging areas. On south facing hills, fields of solar panels are developed, and wind turbines appear on the windy ridges. All of the natural resources are used to produce energy. As this scenario reflects a crisis period, urbanization is slowed down.

152

S. Griffon et al.

Discussion The four scenarios were presented to stakeholders, including professionals and general public, as fixed images as well as fly-over videos. These computer simulations helped to visualize and characterize the landscape, and are one element, among others, which can contribute to policy and decision-making. Such techniques have been used in various similar situations for assessing landscape quality in urban areas (Cartwright 2008), at the urban-rural fringe (Lange et al. 2008) or in more natural landscapes (see, e.g. Appleton and Lovett 2005; Bishop et al. 2005, 2008; Ghadirian and Bishop 2008; Lange et al. 2008; Wissen et al. 2008). Landscape visualization has also been used in scenarios with a certain persuasive approach, in particular concerning climate change (Sheppard 2005, 2006; Dockerty et al. 2005; Mansergh et al. 2008). The scenarios presented here have been built by a small working group, following several discussions with stakeholders, and the visual aspect is part of a more global project. The present work was developed with the objective of being linked to more comprehensive tools for integrated assessment of agri-environmental policies, such as the one developed in the SEAMLESS project (Van Ittersum et al. 2008). The SLE module has up to now not been integrated in the SEAMLESS-IF, but this is planned for the near future. Several issues however remain to be solved. A first issue concerns the landscape that the stakeholder may wish to visualize. Indeed, the optimum resolution for an adequate visualisation is a cadastral map, but sufficiently precisely geospatially referenced land-use data are not commonly available. One way of resolving this problem could be to produce a representative, neutral landscape instead of a real case study. Some research is under way to produce such neutral landscapes (Gardner and Urban 2006; Le Ber et al. 2006b), but these are still insufficiently representative to be used for visualisation purposes. A second issue concerns the way land-use change can be attributed to specific parcels or fields. Indeed, models simulating policy scenarios generally consider statistical data which is most often aggregated at a regional scale (Chapter 7 of this volume). Disaggregating the results remains a difficult task. Le Ber et al. (2006a) suggest the use of a knowledge discovery system based on high-order hidden Markov models for analyzing spatio-temporal data bases. Taking as input an array of discrete data, a land-use being attributed to each land unit, a probability matrix resulting from the scenario assessment modelling chain can then be applied to produce future land-use. A large focus has been put on indicators of sustainable development (see for example Alkan Olsson et al. 2009). Although the outputs of SLE do not aim to produce any indicator of landscape quality, the module can be used to visualize some specific indicators which are relevant at the landscape scale. These indicators can be presented using more abstract 3D views which do not display the landscape in a perfectly realistic way, but rather display abstract symbols that show the dispersion of the indicator over the landscape. In the VisuLands project Wissen et al. (2008) propose several approaches to integrate indicators in landscape visualisation for

6 Visualising Changes in Agricultural Landscapes

153

Fig. 6.15 Spatial patterns of urban area extension (represented in light purple)

participative planning. In the BioScene project Soliva and Hunziker (2009) coupled ideal type narratives with computer-aided photo-editing for participatory landscape planning. In Australia, Bishop et al. (2008) propose agent-based modelling to explore the decision-making process, in a virtual decision environment. In SLE it is possible to visualize an additional GIS layer: for example, in the urban pressure scenario, we computed 3D views with an abstract indicator representing probable urban expansion area (light purple colour in Fig. 6.15). Indeed, the introduction of new large structures such as a city represents visual intrusions which may reduce visual quality. This quality decrease is related to the level of modification such as contrast in size, shape, colour and texture between the structure and the pre-existing landscape. The magnitude of the impact can be considered to depend on the extent of the area affected (Rivas et al. 1997) and indicators of magnitude can be for example the total area from which the new structure can be seen, as in Fig. 6.15. It is also interesting to use GIS maps to display non-visual information such as wildlife migration corridors, water flow and remoteness. There has been much discussion concerning the ethics of using landscape visualization, and several different issues can be raised. One of them concerns the relation between ecology or ecosystem quality and aesthetics: Gobster et al. (2007) argue for instance that future landscape patterns, human experiences, and actions can be devised to create landscapes that are ecologically beneficial and simultaneously elicit aesthetic pleasure. However, the aesthetics of a virtual landscape can be influenced by the degree of realism of the visualization tools (Lange 2001; Daniel 2001; MacFarlane et al. 2005), and it is necessary to define what is the sufficient realism for environmental decision making (Appleton and Lovett 2003). In addition, using virtual imagery can lead to a strong bias due to the selection or highlighting of particular aspects in order to persuade the public on particular environmental

154

S. Griffon et al.

issues (Sheppard 2005, 2006). Sheppard and Cizek (2009) show several examples of misuse of landscape visualization, and suggest a code of ethics, with “the combination of scientific/technical expertise, 3D computer modelling skills, and understanding of social responses to landscape imagery”. Decisions support systems are increasingly being applied in spatial planning, and virtual landscapes become an important part of decision making. Planners recognise realism as an important factor in this type of visualisation (Appleton and Lovett 2005). It is important to define an appropriate level of realism, because the photorealism can have potential negative effects if it is not linked to real-world data. Furthermore, the very fact that we have so much control over the content and style of a visualisation means that everything must be questioned –viewpoint choice, presentation method and addition of auxiliary information should all be considered alongside realism issues when creating images for planning purposes, since they are not subject to the limitations imposed by photo-based or artistic techniques, and they all have the potential to affect the feedback gained from a consultation exercise. The opportunities presented by advancing technology should not be automatically taken, but carefully evaluated and implemented with regard to the needs of the project in question. Only then will computer-generated visualisations form a useful and reliable part of the planning process.

References Alkan Olsson, J., Bockstaller, C., Stapleton, L.M., Ewert, F., Knapen, R., Therond, O., Geniaux, G., Bellon, S., Pinto Correira, T., Turpin, N., & Bezlepkina, I. (2009). A goal oriented indicator framework to support impact assessment of new policies for agri- environmental systems. Environmental Science and Policy, 12(5), 562–572. Appleton, K., & Lovett, A. (2003). GIS-based visualisation of rural landscapes: Defining ‘sufficient’ realism for environmental decision-making. Landscape and Urban Planning, 65, 117–131. Appleton, K., & Lovett, A. (2005). GIS-based visualisation of development proposals: Reactions from planning and related professionals. Computers, Environment and Urban Systems, 29(3), 321–339. Appleton, K., Lovett, A., Sünnenberg, G., & Dockerty, T. (2002). Rural landscape visualisation from FIS databases: A comparison of approaches, options and problems. Computers, Environment and Urban Systems, 26, 141–162. Auclair, D., Barczi, J. F., Borne, F., & Etienne, M. (2001). Landscape visualisation software as a forest management decision support system. In A. Franc, O. Laroussinie & T. Karjalainen (Eds.), Criteria and indicators for sustainable forest management at the forest management unit level (pp. 207–214). Joensuu: European Forest Institute. Auclair, D., Barczi, J.-F., Borne, F., & Étienne, M. (2001). Assessing the visual impact of agroforestry management with landscape design software. Landscape Research, 26, 397–406. Bell, S. (2001). Landscape pattern, perception and visualisation in the visual management of forests. Landscape and Urban Planning, 54, 204–211. Bergen, S. D., Fridley, J. L., Ganter, M. A., & Schiess, P. (1995). Predicting the visual effect of forest operations. Journal of Forestry, 93, 33–37. Bishop, I. D., Hull, R. B., & Stock, C. (2005). Supporting personal world-views in an envisioning system. Environmental Modelling and Software, 20, 1459–1468.

6 Visualising Changes in Agricultural Landscapes

155

Bishop, I. D., Stock, C., & Williams, K. J. (2008). Using virtual environments and agent models in multi-criteria decision-making. Land Use Policy, 26, 87–94. Bishop, I. D., Wherrett, J. R., & Miller, D. R. (2001). Assessment of path choices on a country walk using a virtual environment. Landscape and Urban Planning, 52, 225–237. Bloom, C. (2000). Terrain texture composition by blending in the frame buffer (“Splatting Textures”), from www.cbloom.com/3d/techdocs/splatting.txt Boudon, F., Nouguier, C., & Godin, C. (2001). GEOM module manual. II. Developper guide (p. 600). Document de travail du Programme Modélisation des plantes, DT4-2001. Montpellier, France: CIRAD. Cartwright, W. (2008). Visualising alternative futures. In C. Pettit, W. Cartwright, I. Bishop, K. Lowell, D. Pullar & D. Duncan (Eds.), Landscape analysis and visualisation: Spatial models for natural resource management and planning (pp. 490–507). Berlin: Springer. Cheylan, J.P., & Gumuchian, H. (2002). L’évaluation des impacts des politiques publiques paysagères au sein des Parcs Naturels Régionaux en montagnes méditerranéennes: quelles méthodes? Quels outils? Méthodes et outils pour l’évaluation des impacts des PPP, Numéro spécial: Politiques publiques paysagères et Parcs Naturels Régionaux; pour une évaluation (pp. 13–24). Danahy, J. W. (1989). Irises in a landscape: An experiment in dynamic interaction and teaching design studio. In M. McCullough, W. J. Mitchell & P. Purcell (Eds.), Proceedings of the CAAD Futures’89, The electronic design studio: Architectural knowledge and media in the computer era (pp. 363–376). Cambridge, MA: MIT Press. Daniel, T. C. (1992). Data visualization for decision support in environmental management. Landscape Urban Planning, 21, 261–263. Daniel, T. C. (2001). Wither scenic beauty? Visual landscape quality assessment in the 21st century. Landscape Urban Planning, 54, 267–281. Day, A. M., & Willmott J. (2005). Compound textures for dynamic impostor rendering. Computers & Graphics, 29(1), 109–124. De Boer, W.H. (2000). Fast terrain rendering using geometrical mipmapping, from http://www. flipcode.com/articles/article_geomipmaps.shtml Dockerty, T., Lovett, A., Sünnenberg, G., Appleton, K., & Parry, M. (2005). Visualising the potential impacts of climate change on rural landscapes. Computers, Environment and Urban Systems, 29, 297–320. Ervin, S. M., & Hasbrouck, H. H. (2001). Landscape modelling: Digital techniques for landscape visualization. New York: McGraw-Hill. Gardner, R. H., & Urban, D. L. (2006). Neutral models for testing landscape hypotheses. Landscape Ecology, 22, 15–29. Ghadirian, P., & Bishop, I. D. (2008). Integration of augmented reality and GIS: A new approach to realistic landscape visualisation. Landscape and Urban Planning, 86, 226–232. Gobster, P. H., Nassauer, J. I., Daniel, T. C., & Fry, G. (2007). The shared landscape: What does aesthetics have to do with ecology? Landscape Ecology, 22, 959–972. Google Inc. (2006). Sketchup user’s guide (p. 362), from http://download.sketchup.com/GSU/ pdfs/GSUUsersGuide_WIN.pdf Helming, K., Tscherning, K., König, B., Sieber, S., Wiggering, H., Kuhlman, T., et al. (2008). Ex ante impact assessment of land use changes in European regions—the SENSOR approach. In K. Helming, M. Perez-Soba & P. Tabbush (Eds.), Sustainability impact assessment of land use changes. Berlin: Springer. Herwig, A., & Paar, P. (2002). Game engines: Tools for landscape visualization and planning? In E. Buhmann, U. Nothelfer & M. Pietsch (Eds.), Proceedings at Anhalt University of Applied Sciences, trends in GIS and virtualization in environmental planning and design (pp. 162–171). Heidelberg: Wichmann. Lange, E. (2001). The limits of realism: Perceptions of virtual landscapes. Landscape and Urban Planning, 54, 163–182. Lange, E., & Bishop, I. (2001). Our visual landscape: Analysis, modeling, visualization and protection. Landscape and Urban Planning, 54, 1–3.

156

S. Griffon et al.

Lange, E., Hehl-Lange, S., & Brewer, M. J. (2008). Scenario-visualization for the assessment of perceived green space qualities at the urban-rural fringe. Journal of Environmental Management, 89, 245–256. Le Ber, F., Benoît, M., Schott, C., Mari, J.-F., & Mignolet, C. (2006). Studying crop sequences with CarrotAge, a HMM-based data mining software. Ecological Modelling, 191, 170–185. Le Ber, F., Lavigne, C., Mari, J.F., Adamczyk, K., & Angevin, F. (2006b). GENEXP, un logiciel pour simuler des paysages agricoles en vue de l’étude de la diffusion de transgènes. Paper presented at the SAGEO 2006 International Conference on Spatial Analysis and GEOmatics; Research & Development, Strasbourg, FR. Retrieved from http://hal.archives-ouvertes.fr/ docs/00/10/33/82/PDF/sageo06.pdf Lovett, A. (2005). Futurescapes. Computers, Environment and Urban Systems, 29, 249–253. Lovett, A., Kennaway, R., Sünnenberg, G., Cobb, D., Dolman, P., & O’Riordan, T. (2001). Visualising sustainable agricultural landscapes. In D. Unwin & P. Fisher (Eds.), Virtual reality in geography (pp. 102–130). London: Taylor & Francis. MacFarlane, R., Stagg, H., Turner, K., & Lievesley, M. (2005). Peering through the smoke? Tensions in landscape visualisation. Computers, Environment and Urban Systems, 29, 341–359. Mansergh, I., Lau, A., & Anderson, R. (2008). Geographic landscape visualisation in planning adaptation to climate change in Victoria, Australia. In C. Pettit, W. Cartwright, I. Bishop, K. Lowell, D. Pullar & D. Duncan (Eds.), Landscape analysis and visualisation: Spatial models for natural resource management and planning (pp. 469–487). Berlin: Springer. Muhar, A. (2001). Three-dimensional modelling and visualisation of vegetation for landscape simulation. Landscape and Urban Planning, 54, 5–17. Nespoulous, A. (2004). Relations entre la dynamique de la végétation et la gestion sociale de l’espace: les garrigues du Pic Saint Loup. MA thesis, University of Montpellier III, p. 89. Orland, B., Budthimedhee, K., & Uusitalo, J. (2001). Considering virtual worlds as representations of landscape realities and as tools for landscape planning. Landscape and Urban Planning, 54, 139–148. Paar, P., & Clasen, M. (2007). Earth, landscape, biotope, plant. Interactive visualisation with Biosphere3D. In M. Schrenk, V. V. Popovich & J. Benedikt (Eds.), Real Corp 007 – to plan is not enough: Strategies, plans, concepts, projects and their successful implementation in urban, regional and real estate development. Vienna: CORP. Perrin, L., Beauvais, N., & Puppo, M. (2001). Procedural landscape modeling with geographic information: The IMAGIS approach. Landscape and Urban Planning, 54, 33–47. Pettit, C., Cartwright, W., Bishop, I., Lowell, K., Pullar, D., & Duncan, D. (eds). (2008). Landscape analysis and visualisation: Spatial models for natural resource management and planning (Book Series: Lecture notes in geoinformation and cartography). Berlin: Springer. Pradal, C., Donès, N., Godin, C., Barbier de Reuille, P., Boudon, F., Adam, B., & Sinoquet, H. (2004). ALEA: A software for integrating analysis and simulation tools for 3D architecture and ecophysiology. In C. Godin et al. (Eds.), Proceedings of the 4th International Workshop on Functional-Structural Plant Models, Montpellier, France, 7–11 June 2004 (p. 406). Montpellier: UMR AMAP. Rivas, V., Rix, K., Frances, E., Cendrero, A., & Brunsden, D. (1997). Geomorphological indicators for environmental impact assessment;consumable and non consumable geomorphological resources. Geomorphology, 18(3–4), 169–182. Salter, J.D., Campbell, C., Journeay, M., & Sheppard, S.R.J. (2009). The digital workshop: Exploring the use of interactive and immersive visualisation tools in participatory planning. Journal of Environmental Management, 90, 2090–2101. Sauget, N., & Depuy, M. (1996). Forêt paysanne et paysage: les agriculteurs et le visible. In G. Balent (Ed.), La forêt paysanne dans l’espace rural. Biodiversité, paysages, produits. Étud. Rech. Syst. Agraires Dév., 29 (pp. 245–264). Savill, P., Evans, J., Auclair, D., & Falck, J. (1997). Plantation silviculture in Europe. Oxford: Oxford University Press. 297 p.

6 Visualising Changes in Agricultural Landscapes

157

Sheppard, S.R.J. (1982). Landscape portrayals: Their use, accuracy, and validity in simulating proposed landscape change. Dissertation, University of California, Berkeley, CA. Sheppard, S. R. J. (2005). Landscape visualisation and climate change: The potential for influencing perceptions and behaviour. Environmental Science and Policy, 8, 637–654. Sheppard, S. R. J. (2006). Bridging the sustainability gap with landscape visualisation in community visioning hubs. The Integrated Assessment Journal, 6(4), 79–108. Sheppard, S.R.J., & Cizek, P. (2009). The ethics of Google Earth: Crossing thresholds from spatial data to landscape visualisation. Journal of Environmental Management, 90, 2102–2117. Sirami, C., Brotons, L., & Martin, J. L. (2007). Vegetation and songbird response to land abandonment: From landscape to census plot. Diversity and Distributions, 13, 42–52. Sirami, C., Brotons, L., & Martin, J. L. (2008). Spatial extent of bird species response to landscape changes: Colonisation/extinction dynamics at the community-level in two contrasting habitats. Ecography, 31, 509–518. Snyder, K. (2003). Tools for community design and decision making. In S. Geertman & J. Stilwell (Eds.), Planning support systems in practice (pp. 99–120). Berlin: Springer. Soliva, R., & Hunziker, M. (2009). Beyond the visual dimension: Using ideal type narratives to analyse people’s assessments of landscape scenarios. Land Use Policy, 26, 284–294. Thomas, A. L., & Price, C. (1999). Landscape valuation of farm woodlands. In P. J. Burgess, E. D. R. Brierley, J. Morris & J. Evans (Eds.), Farm woodlands for the future (pp. 69–79). Oxford: BIOS Scientific Publishers. Tress, B., & Tress, G. (2003). Scenario visualisation for participatory landscape planning—a study from Denmark. Landscape and Urban Planning, 64, 161–178. Tyrväinen, L., & Tahvanainen, L. (2000). Landscape visualisation in rural land-use planning. In B. Krishnapillay, et al. (Ed.), Forests and society: The role of research, XXI IUFRO World Congress (Vol. 1, pp. 338–347), Kuala Lumpur, Malaysia. Van Ittersum, M. K., Ewert, F., Heckelei, T., Wery, J., Alkan Olsson, J., Andersen, E., et al. (2008). Integrated assessment of agricultural systems – A component-based framework for the European Union (SEAMLESS). Agricultural Systems, 96, 150–165. Wissen, U., Schroth, O., Lange, E., & Schmid, W. A. (2008). Approaches to integrating indicators into 3D landscape visualisations and their benefits for participative planning situations. Journal of Environmental Management, 89, 184–196.

Chapter 7

A Biophysical Typology in Agri-environmental Modelling Gerard Hazeu, Berien Elbersen, Erling Andersen, Bettina Baruth, Kees van Diepen, and Marc Metzger

Introduction Objective of SEAMLESS Biophysical Typology The reason to develop an Agri-Environmental Zonation (AEnZ) is to establish an agri-environmental framework to be used in the ex-ante integrated assessment framework developed in the SEAMLESS project (Van Ittersum et al. 2008). The framework is needed to assess the impacts of agricultural policies covering the wide biophysical variation in which agricultural activities take place in Europe.

G. Hazeu (*) and K. van Diepen Wageningen UR, Alterra, Centre for Geo-Information, Droevendaalsesteeg 3, 6708 PB, Wageningen, The Netherlands e-mail: [email protected]; [email protected] B. Elbersen Wageningen UR, Alterra, Landscape Centre, Droevendaalsesteeg 3, 6708 PB, Wageningen, The Netherlands e-mail: [email protected] E. Andersen Faculty of Life Sciences, Forest and Landscape, University of Copenhagen, Rolighedsvej 23, DK-1958, Frederiksberg C, Denmark e-mail: [email protected] B. Baruth Joint Research Centre, Institute for the Protection and Security of the Citizen, Monitoring Agricultural Resources Unit (MARS), AGRI4CAST TP 483, 21027, Ispra (VA), Italy e-mail: [email protected] M. Metzger School of GeoSciences, Centre for Environmental Change and Sustainability, The University of Edinburgh, Edinburgh, UK e-mail: [email protected] F.M. Brouwer and M. van Ittersum (eds.), Environmental and Agricultural Modelling: Integrated Approaches for Policy Impact Assessment, DOI 10.1007/978-90-481-3619-3_7, © Springer Science+Business Media B.V. 2010

159

160

G. Hazeu et al.

The main objective of building this AEnZ was therefore to stratify Europe on the main biophysical factors that determine the agronomic production capacity in Europe. Furthermore, the variation in environment also needs to be taken into account when assessing the impacts of agriculture on water, soil, air, climate, biodiversity and landscape. Overall it is clear that we can distinguish zones in Europe, according to a rough division, where agricultural activities are very much limited by climatic, soil and/or other topographic factors, while in other areas the natural factors provide good opportunities for a wide range of agricultural activities or specific types of agriculture. Within these coarse zones a further subdivision can be made in units that have similar biophysical environment and therefore have the same agricultural potential and agri-environmental problems.

Background In the last couple of years there is a larger emphasis on assessing the impacts of agricultural development and agricultural policies especially in relation to environment and biodiversity. This has mainly been caused by the increased pressure on the environment, biodiversity loss and landscape degradation by agriculture in most areas of Europe in the last decades. The impacts on environment, biodiversity and landscape that accompanied the changes in farming have been well documented (e.g. Buckwell and Armstrong-Brown 2004; Wadsworth et al. 2003; Boatman et al. 1999; EEA 1999; MAFF 1998; Pretty 1998; EPA 1999; Campbell and Cooke 1997; Baldock et al. 1996) and they are almost all negative. The main causes for this were a continuous intensification of land use but also land abandonment. The intensification process has been a key response of farmers to changes in markets and policies (Common Agricultural Policy [CAP] support). This agricultural intensification usually goes together with an increase in efficiency of the agricultural production process but also with negative externalities on the environment in terms of habitat loss and fragmentation, loss of landscape connectivity, loss of diversity and the creation of monoculture, loss of habitat quality through pollution of soil, water and air and even direct poisoning and loss of food supplies for certain species (Poiret 1999; Pau Vall and Vidal 1999). At the same time not only intensification shows a heavy impact on farmland biodiversity but also abandonment (EEA 1999; Baldock et al. 1996). This process of polarisation, in which abandonment and an increase in stocking density occur simultaneously in different locations, poses a threat to biodiversity in semi-natural areas created by extensive livestock farming. On the other hand it should also be emphasized that the relationship between farming and environment is not only a negative one. The continuation of extensive farming practices plays a key role in conserving farmland biodiversity and this is the key factor for introducing the concept of High Nature Value farming and targeting these type of systems and associated areas in the most recent Rural Development program of the EC. The assumed positive relationship between extensive farming practices and biodiversity values is the main reason why these farming systems are

7 A Biophysical Typology in Agri-environmental Modelling

161

considered to be of ‘high nature value’ (Beaufoy et al. 1994; Anger et al. 2002; Bignal and McCracken 1996; de Miguel 1999; Nagy 2002 and Andersen et al. 2003). One of the first reports showing a more or less EU wide overview was “The Nature of Farming” published in 1994. It discusses case studies of livestock, cereal, permanent crop and mixed systems which were of significance for nature conservation (Beaufoy et al. 1994). This proceeded by ancillary studies and interpretative papers (Beaufoy et al. 1994; Bignal and McCracken 1996, 2000 and Andersen et al. 2003). The continuation of the extensive grassland management is very important for the maintenance of associated biodiversity value (e.g. Anger et al. 2002; Bignal and McCracken 1996; de Miguel 1999; Nagy 2002). It is also estimated that approximately 16% of the habitats in Natura 2000 areas depend on a continuation of extensive farming.1 In response to the major influence agriculture has on the environment, the CAP gradually integrated environmental considerations. The McSharry reforms of the Policy CAP in 1992 led to the implementation of the first Agri-environmental Regulation (EEC 2078/92). Also codes of ‘Good Farming Practices (GFP)’, i.e. agricultural production methods compatible with the requirements of the protection of the environment and the maintenance of the countryside are being promoted though the CAP. The Nitrates Directive 91/676/EC, approved in 1991, requires Member States to identify, specify and encourage farmers to apply so-called ‘Good Agricultural Practices (GAP)’ for use of animal manure and fertilizer. The prospect of enlargement of the European Union (EU) to the Central and Eastern European countries and the continuing pressure for trade liberalisation stimulated even a further reform of the CAP and the further integration of environmental considerations into EU policy. It resulted in the 1996 Cork Declaration which placed sustainable rural development at the top of the EU agenda. Recently, the objectives of the agricultural policies have been changed, which resulted in the further broadening of the CAP towards environment, landscape and rural viability (Commission of the European Communities 2003). The support has been further decoupled from production with the 2003 reform of the CAP. Cross compliance has been introduced as an enforcement mechanism for EU standards for good farming practices and a larger share of the support will be targeted to rural development. Implementation of policies through mechanisms such as the Water Framework Directive, Natura 2000, the Birds and Habitats Directives, and the Nitrates Directive, dictate environmental quality targets. All these policies require different monitoring and evaluation approaches in European agriculture. Therefore, a systematic appraisal of the wide variety of farming activities within Europe’s wide range of environmental conditions (climate, soil, vegetation etc.) is required. The only way to do this is by placing the different farming systems in their wider spatial and environmental context. It is not only the action of the farmer in response

Source: Reporting of Member States in the framework of the Habitats Directive (92/42/EEC); status of July 2006.

1

162

G. Hazeu et al.

to a policy that determines the policy result, but it is the combination of the farmers action within the environmental context. The appraisal needs to be spatially explicit as it is not sufficient for policy makers to know what the impact will be, but it is also important to know where the impacts will be and how and why they vary in different regions in the EU. An agri-environmental stratification should provide a sampling basis for assessing the impacts of European Union policy in the wide variation of combinations of farming activities and environmental endowments. It enables to measure the effect of farming in response to policy across the widely varying environmental conditions. The classifying factors of the typology are based on environmental factors that are relatively stable in time and do not change under influence of antropogenic factors, at least not in a short period of time (Cochran 1997; Bunce et al. 1996a).

Requirements of a Spatial Agri-environmental Framework The AEnZ needs to be an agri-environmental framework that systematically covers the wide range of combinations between farming activities and environments that together determine the effects of farming on the environmental objectives of the CAP. Minimum requirements of such a framework are: –– It should provide a good overview of the agri-environmental diversity in Europe supporting the development of agri-environmental modelling that is applicable to the main agricultural production areas in Europe. It should therefore be based on soil and climate factors, slope and altitude. Land cover and yield data were not incorporated as they can easily change under influence of human interference. They can be used as attributes to describe the AEnZs. –– It should provide a statistically robust classification that can be used as a sampling and up-scaling basis for collection of farm information and (point) modelling of farming activities. –– It should cover the whole of Europe (EU27+, i.e. EU27 and Norway, Switzerland and the Balkan countries) which leads to the requirement for similar input data with a European-wide coverage. –– It should not duplicate any existing environmental classification but rather build on these and further extend these classifications for the specific requirements of the project. The selection of input data for the AEnZ is based on user requirements, experience from former projects and the availability of data at the European scale, such as the European maps of soil (http://eusoils.jrc.it; Jones et al. 2004, 2005a; Baruth et al. 2006b) and land use/land cover (Büttner et al. 2004), the weather data in the Monitoring of Agriculture with Remote Sensing (MARS) climate database (Genovese et al. 2007; http://mars.jrc.it/marsstat), and the statistics from Eurostat’s regional databases.

7 A Biophysical Typology in Agri-environmental Modelling

163

The resulting AEnZ consists of spatial land units that are as homogeneous as possible from an agronomic perspective.

Short Description of the Contents We first provide a short review of existing environmental typologies and shows the need for a environmental typology reflecting the variation in European soil properties. In the next section the databases used to build an Agri-Environmental Zonation are presented with the relevant reference to more in depth descriptions of the datasets. The methodology to develop an Agri-Environmental Zonation is described. An important part is the selection of a soil variable that explain most of the variation within the environmental zones from an agronomic perspective. Also the division of Europe in three zones with different agricultural potential (suited, marginally suited, unsuited) is described (the Agri-mask database). The definition of the AEnZ and the characterization of agriculturally important AEnZ land types are presented. The Seamzones are the basis for the modelling within SEAMLESS and the attachment of a representative soil profile and climate variables to each Seamzone is discussed. The selection of sample regions, the allocation of farms and as input for models used within the integrated assessment framework of SEAMLESS are applications of the Agri-Environmental Zonation and Seamzones that are briefly described. Finally, the Agri-Environmental Zones, Climate Zones, Seamzones and the applications of the biophysical typology are briefly discussed and summarized.

Environmental Typologies and Up-scaling Environmental classifications have been produced by several people. Reasons for developing these classifications were related to the need to monitor changes in the environment, effects of farming or other human interference on biodiversity and landscape or systematically describe different environmental regions. At the International Institute of Applied System Analysis (IIASA) a Global Agro-Ecological Zonation (GAEZ) at 5 min resolution grid cells has been produced in 2000 with a recent update for 2007 (http://www.iiasa.ac.at/Research/LUC/luc07/ Research-AEZ/index.html?sb = 8). However, the dataset is based on global datasets and for the European context far too coarse. Two groups of environmental classifications can be distinguished: • Classifications based on expert judgment; and • Statistically defined classifications. Examples of the first group are the stratification of European landscapes by Meeuws et al. (1990), the Biogeographical Regions Map of Europe (BRME) by the

164

G. Hazeu et al.

European Environmental Agency (EEA 2002), and the Potential Natural Vegetation map by Bohn et al. (2000). All these stratifications were produced by expert judgment and the resulting strata are useful for a general characterization and simplification of the wide diversity in the European environment, but they are not appropriate as a sampling and an up-scaling framework for producing representative assessments of Europe. The need for statistical environmental stratification was first recognized by field ecologists at the Institute of Terrestrial Ecology (ITE) (now Centre for Ecology and Hydrology [CEH]) in the UK in the 1970s. These scientists realized that stratified random sampling was the only feasible way of assessing ecological resources, such as habitats and vegetation, and enabling monitoring schemes to be developed for large, heterogeneous areas (Bunce et al. 1996a). Sheail and Bunce (2003) have recently described the history and development of environmental classification and strategic ecological survey in the UK. Several other countries and regions have also adopted quantitative classifications as the basis for survey, monitoring and management, e.g. Spain (Elena-Rosselló 1997), Austria (Wrbka et al. 1999), New Zealand (Leathwick et al. 2003), and Senegal (Tappan et al. 2004). Two earlier European statistical stratifications have been produced. In the first, Jones and Bunce (1985) defined 11 classes on a 50 × 50 km grid for Europe. More than a decade later, improved data availability, software and computing power allowed the classification of 64 classes on a 0.5° grid (approximately 50 × 50 km) (Bunce et al. 1996b). Although this latter classification was used in a range of studies, the coarse resolution limited its application for ecological sampling. The Environmental Stratification of Europe (EnS; Metzger et al. 2005; Jongman et al. 2006) forms the latest statistical classification of the European environment, distinguishing 84 strata at a 1 km2 resolution. Kappa’s analysis of aggregations of the strata shows they compare well with other European classifications (Metzger et al. 2005). The EnS shows strong statistical correlations with other European environmental datasets, including soil and agronomic variables (Metzger et al. 2005) Fig. 7.1 gives a summary of two agronomic variables for 12 of the 13 Environmental Zones that are considered to be relevant for this study. The EnS has been used in several European projects, e.g. in BioHab as a framework for consistent monitoring of European habitats (Bunce et al. 2008), in ATEAM to summarize outputs from global change impact models (Metzger et al. 2006, 2008) as well as by the European Environment Agency to assess environmentally compatible bioenergy potential from agriculture (EEA 2007) and the distribution of high nature value farmland (EEA 2004). The EnS, which is based mainly on climatic variables, partitions environmental variation across a continental gradient, and therefore does not recognize regional variation in soil properties. As such, the EnS does not provide a suitable stratification framework for agricultural modelling, as at a regional scale soil factors determine the agronomic potential and environmental impact of farming. For this reason it was decided that for SEAMLESS soil information should be integrated with the EnS. This will then result in an AEnZ of Europe suitable for SEAMLESS purposes.

7 A Biophysical Typology in Agri-environmental Modelling

165

Fig. 7.1 Twelve environmental zones and their variation in temperature sums and growing season (days > 5°C)

Data The AEnZ is based on the Environmental Stratification of Europe (EnS), the European Soil Database (ESDBv2), the TOPsoil Organic Carbon (OCTOP) and the Global Digital Elevation Model (GTOPO30). In this section a short description of those databases is presented.

Description and Characterization of the Environmental Zones (EnZ) The EnS consists of 84 strata, which have been aggregated into 13 environmental zones (EnZs) based on divisions of the mean first principal component score of the strata. The only exception is the Mediterranean mountain zone, which was separated based on altitude. The dataset covers a ‘Greater European Window’ (11°W–32°E, 34°N–72°N), encompassing the EU27+ with extensions to the East and South. The EnS has been constructed using tried and tested statistical procedures so that the strata are unambiguously determined and, as far as possible, independent of personal bias. It forms an appropriate stratification for stratified random sampling of ecological resources, the selection of sites for representative studies across the continent and for the provision of strata for modelling exercises and reporting at European scale.

166

G. Hazeu et al.

Twenty of the most relevant available environmental variables were selected for the construction of the stratification, based mainly on those identified by statistical screening (Bunce et al. 1996b). These were: • Climate variables from the Climatic Research Unit (CRU) TS1.2 dataset (Mitchell et al. 2004) • Elevation and slope data from the United States Geological Survey HYDRO1k digital terrain model; and • Indicators for oceanicity and northing. Data were analysed at 1 km2 resolution. Principal component analysis (PCA) was then used to compress 88% of the variation into three dimensions, which were subsequently clustered using an ISODATA clustering routine. The classification procedure is described in detail elsewhere (Metzger et al. 2005). The following 13 EnZs are distinguished: –– –– –– –– –– –– –– –– –– –– –– –– ––

EnZ 1: Alpine North (ALN) EnZ 2: Boreal (BOR) EnZ 3: Nemoral (NEM) EnZ 4: Atlantic North (ATN) EnZ 5: Alpine South (ALS) EnZ 6: Continental (CON) EnZ 7: Atlantic Central (ATC) EnZ 8: Pannonian (PAN) EnZ 9: Lusitanian (LUS) EnZ 10: Anatolian (ANA) EnZ 11: Mediterranean Mountains (MDM) EnZ 12: Mediterranean North (MDN) EnZ 13: Mediterranean South (MDS)

Here only 12 EnZs are important, as Anatolia (Turkey) is not covered by the SEAMLESS spatial framework (EU27+).

Description of European Soil DataBase (ESDBv2) The European Soil Database, version two (ESDBv2) consists of a number of databases of which the following two are used as input for the AEnZ: –– The Soil Geographical Database of Eurasia at scale 1:1,000,000 (SGDBE), which is a digitized European soil map and related attributes (version 4 beta). –– The Pedo Transfer Rules Database (PTRDB) version 2.0, which holds a number of pedotransfer rules which can be applied to the SGDBE; the results of the application of the pedotransfer rules to the SGDBE are delivered as a table with new attributes related to the European soil map.

7 A Biophysical Typology in Agri-environmental Modelling

167

Extensive information of both databases can be found on the website http://eusoils. jrc.it/ESDB_Archive/ESDBv2/fr_intro.htm. A number of soil variables that are important from an agronomic perspective were selected to define the most important ones in explaining the variation in soil properties (see Methods).

Description of TOPsoil Organic Carbon (OCTOP) OCTOP is published by Jones et al. (2004, 2005b). This topsoil (0–30 cm) organic carbon dataset is the result of a novel approach combining a rule-based system (provided by pedo-transfer rules) with detailed thematic spatial data layers. The effects of land use, vegetation and temperature were taken into account in the calculations to estimate the organic carbon contents. Point data extrapolation was not suitable to generate the database as the number of samples for Europe is insufficient, data were insufficiently geo-referenced and the organic carbon contents vary within soil units depending on vegetation and land management (Jones et al. 2005b). The resolution of the dataset is a one by one km grid spacing. This resolution is regarded as appropriate for planning effective soil protection measures at European level. The estimation or determination of the spatial distribution of organic carbon content of soils will have always an element of uncertainty. The data sources used to compile the OCTOP database are: –– European Soil Database version 1.0 (ESDBv1) (Heineke et al. 1998); –– European Land Cover Data (combination of CORINE Land Cover [CLC]) of the EEA and Eurasian land cover (United States Geological Survey [USGS]) (Hiederer 2001); –– Average Annual Accumulated Temperature (AAAT) from the Global Historical Climatology Network (GHCN) (Easterling et al. 1996).

Description of the Global Digital Elevation Model (GTOPO30) GTOPO30 is a global digital elevation model (DEM) with a horizontal grid spacing of 30 arc seconds (~1 km). The global dataset is covering the full extent of latitude from 90° south to 90° north, and the full extent of longitude from 180° west to 180° east. The horizontal coordinate system is decimal degrees of latitude and longitude referenced to the coordinate system WGS84. The vertical units represent elevation in meters above mean sea level. The elevation values range from −407 to 8,752 m. In the digital elevation model, ocean areas have been masked as “no data” and have been assigned a value of −9,999. For more information on the dataset we refer to the website http://edc.usgs.gov/products/elevation/gtopo30/gtopo30.html.

168

G. Hazeu et al.

Methods After the selection of the soil variable that is most important from an agronomic perspective, the EnZ is combined with the selected soil variable. Furthermore, an Agri-mask is added to distinguish zones that are very much limited by climatic, soil and/or other biophysical factors, while in other areas the natural factors provide good opportunities. The different databases have been prepared to fit the SEAMLESS spatial data framework. If necessary they were converted from vector into raster (1 × 1 km resolution) and transformed into the European standard projection ETRS_1989_LAEA. The geographical coverage is limited to EU27 and Norway, Switzerland and Balkan countries (i.e. EU27+) (see also Baruth et al. 2006a). Furthermore, the “empty” SEAMLESS grid cells of EnZ are filled by Euclidean Distance routine.

Selection of Soil Variables As it is clear that the EnZ is not representing the diversity in soil factors, it is needed to determine the soil variable(s) that explains most of the variance from agronomic perspective within Europe. The following soil variables were selected as important from an agronomic perspective: –– –– –– –– –– ––

TOPsoil Organic Carbon (OCTOP) (continuous variable); Available water holding capacity (continuous variable, PTRDB); Rooting depth (classes, SGDBE); Depth of gleyed horizon (classes, PTRDB); Topsoil textural classes (classes, SGDBE); Topsoil Cation Exchange Capacity (classes, PTRDB).

These variables come from different data sources, and have different data properties. For example rooting depth is only available in a few classes, while organic carbon is available as a continuous variable, showing far more local variation. It is therefore necessary to analyze and compare these datasets, and examine how each of them could contribute to the soil stratification, and how much each available variable actually contribute to the explanation of soil differences. There will be much overlap in explained spatial variation between these variables, e.g. a soil with limitation in rooting depth will be associated with certain texture classes, and may also have gleyed horizons. This makes that only a limited number of data layers will be used for the stratification. PCA was used to screen the selected variables. PCA is an effective multivariate technique to reduce the variation of many variables into a limited number of dimensions. The eigenvectors of the principal components explain how much of each component is explained by each variable. In this way it is possible to detect

7 A Biophysical Typology in Agri-environmental Modelling

169

which variables are most important for explaining the variation in agronomic soil properties (i.e. the complete set). Other variables in the complete set will be correlated, or show less regional spatial variation. The following three principal component analyses were performed: –– Only European Soil Bureau (ESB) data (SGDBE and PTRDB data); –– Only continuous variables; –– Combination of all of variables. A principal component analysis is most effective on one type of variables. A mixture of continuous variables and classes is problematic. Furthermore, a restriction in using the ESDB data is that geographical (country) borders have large influences on the class (value) of certain soil variables. The interpretation of certain soil variables differs between countries resulting in a heterogeneous database. Therefore, the ESB data are difficult to use for the determination of the most important soil factor that explains the most variation in agronomic properties of soils within Europe. The PCA on only the continuous data revealed that all variation in the input variables is explained by the topsoil organic carbon content. Disadvantage is the relatively small, but balanced dataset (OCTOP and available water holding capacity). The PCA on all variables revealed that 95% of the input variables is explained by the OCTOP. The overall conclusion of the PCA analysis was to use the Topsoil Organic Carbon content as variable to differentiate between soils in Europe. The OCTOP is a continuous variable which has been grouped into the following six classes (in percentage): –– –– –– –– –– –– ––

Class 1: 0.1–1.23 Class 2: 1.23–2.46 Class 3: 2.46–3.94 Class 4: 3.94–5.66 Class 5: 5.66–8.86 Class 6: 8.86–63.0 Class 9: no data or 0

The class limits of the six classes are established in such a way that each class is covering a European land surface area of approximately the same extent. This was achieved by using the quantiles option in ArcGIS for distribution over six subclasses. Note that in the statistics the “no data” class is dealt with as a seventh class. Despite the fact that the statistical analysis identified the OCTOP as the most suitable data to represent variation in soils properties in Europe, it is important to note that this is likely caused by the limited quality and availability of other high-resolution spatial data for agronomic soil properties. Despite strong correlations with OCTOP, there is likely to be considerable variation in soil properties within each class e.g. caused by variation in soil parent material and soil structure which could affect issues such as rooting depth.

170

G. Hazeu et al.

Agri-mask The Agri-mask was developed by combining several datasets. Datasets used in the development of the Agri-mask were the CORINE Land Cover map for the year 2000 (CLC2000), GTOPO30 and ESDBv2 database. The altitude – latitude relation reflects the area above which no arable agriculture is possible. This relation is based on the highest points in mountainous areas all over Europe where agriculture (non irrigated arable land and pastures) was still found according to the CLC2000 database (see Fig. 7.2). This relation has been applied to GTOPO30 dataset to select all grid cells above this agriculture line. The slope grid database with 8% and 16% thresholds was calculated from the GTOPO30 database. Soil variables rooting depth (<20 cm), alkalinity (>15% exchangeable sodium) and salinity (>15 dS/m) were taken from the European Soil Database (ESDBv2) and Soil Information (SINFO) study (Baruth et al. 2006b). The mentioned thresholds have been based on expert knowledge. An Agri-mask database was developed by combining those separate grid databases resulting in the following three classes: –– Class 0 (suited): areas with no or relatively small constraints to agriculture; –– Class 1 (unsuited): areas where no arable agriculture is possible (mountainous areas above a certain altitude, depending on the latitude, and/or very steep slopes [>16%] and/or limited rooting depth [<20 cm]); –– Class 2 (marginal): strongly naturally handicapped areas where agriculture, if practiced, is heavily constrained and restricted to extensive farming (areas with steep slopes [>8%] and/or high alkalinity and/or salinity [>15 dS/m]). 2000 1800 1600

Altitude (m)

1400 1200 1000 800 600 400 200 0 30

40

50 Latitude (degrees)

60

70

y = –54.51x + 3743.40 R2 = 0.92

Fig. 7.2 The relationship between the altitude (in meters) above which arable farming is not possible anymore and latitude (in degrees)

7 A Biophysical Typology in Agri-environmental Modelling

171

Results Agri-environmental Zonation (AEnZ) Definition The EnZ, OCTOP and Agri-mask data layers are integrated into the AEnZ that covers the EU27+ at a spatial resolution of 1 × 1 km. The AEnZ is presented hierarchically with at the first level the Environmental Zones (only 12 classes are relevant as the Anatolian EnZ is not covered by the EU27+), at the second level the seven organic carbon classes (OCTOP) and at the third level the three Agri-mask classes. The resulting spatial land units, i.e. AEnZ land types, are as homogeneous as possible from an agronomic perspective. The maximum number of possible combinations of the three data layers is 12 × 7 × 3 is 252, representing 216 AEnZ land types, and 36 units for which the carbon data are lacking. The actual map overlay resulted in 238 different AEnZ land types for which 35 units lack carbon data. From those 238 different types 76 land types belong to the mountainous areas (Agri-mask class one) where no arable agriculture is possible. Eighty land types are strongly handicapped by the natural conditions (Agri-mask class two). The remaining 82 land types belong to the Agri-mask class 0 that indicates that there are no or only small constraints to agriculture. From those 82 types 12 relate to environmental zones for which information on the organic carbon content was not available (OCTOP class nine). The number of 70 agri-environmental land types in Agri-mask class zero is only slightly less than the maximum of 72 possible types (12*6). The following table (Table 7.1) gives an overview of AEnZ land types that do not exist in the EU27+. An overview of land shares per EnZ, OCTOP class and Agri-mask class is presented in Table 7.2. The four largest zones (zones EnZ 6, EnZ 2, EnZ 7, and EnZ 12) occupy 52% of EU27+. The remaining eight zones occupy 48%, each of them covering 5.5–7%, except for the smallest zone EnZ nine which covers only 4%. The OCTOP classes share more or less equal portions of land in EU27+ (13–16%), with the exception of OCTOP class two with 20.7% and class nine (no data available) with 5.1%. Characterization The characterization of the AEnZ land types is based on land cover and biophysical variables (mean altitude, mean growing season [number of days with above 5°C], mean temperature range [temperature August – temperature January], mean summer drought [sum of rainfall deficit for May, June and July], mean slope, mean available water holding capacity, rooting depth and medium texture). As indicated in Table 7.2, three-quarter (75.8%) of the EU27+ has no or relatively small constraints to agriculture (Agri-mask class zero), which are large parts of the Nemoral, Atlantic Central and North, Continental, Pannonian, Lusitanian,

172

G. Hazeu et al.

Table 7.1 AEnZ land types not existing in EU27+ Environmental AGRI_ zone (EnZ) OCTOP MASK AEnZ land type 1. Alpine North (ALN) – 2. Boreal (BOR) 1 2 212

3. Nemoral (NEM)

4. Atlantic North (ATN)

5. Alpine South (ALS) 6. Continental (CON) 7. Atlantic Central (ATC) 8. Pannonian (PAN) 9. Lusitanian (LUS) 10. Anatolian (ANA) 11. Mediterranean Mountains (MDM) 12. Mediterranean North (MDN) 13. Mediterranean South (MDS)

1

0, 1, 2

310, 311, 312

2–9

1

5

2

321, 331, 341, 351,361, 391 352

1–9

0, 1, 2

– – – – – All –

Description Combined class EnZ 2, OCTOP 1 and AGR_MASK 2 is missing Entire OCTOP class 1 is missing Entire AGRI_MASK class 1 is missing Combined class EnZ 3, OCTOP 5 and AGR_ MASK 2 is missing

Entire EnZ is missing

– 6

0, 1, 2

1360, 1361, 1362

Entire OCTOP class 6 is missing

Table 7.2 Land share for EnZ, OCTOP and Agri-mask for EU27+ (%). Total surface is million square kilometres Environmental Zone (EnZ) % OCTOP % AGRI_MASK class 1. Alpine North (ALN) 6.9 1. 0.1–1.23 14.1 0. Areas with no or relatively small constraints 2. Boreal (BOR) 13.1 2. 1.23–2.46 20.7 1. Areas where no arable agriculture is possible 3. Nemoral (NEM) 5.7 3. 2.46–3.94 16.0 2. Areas that are strongly handicapped for agriculture 4. Atlantic North (ATN) 6.2 4. 3.94–5.66 15.9 5. Alpine South (ALS) 5.7 5. 5.66–8.86 14.6 6. Continental (CON) 19.4 6. 8.86–63.0 13.6 7. Atlantic Central (ATC) 10.2 9. No data 5.1 8. Pannonian (PAN) 7.1 9. Lusitanian (LUS) 3.9 10. Anatolian (ANA) 11. Mediterranean Mountains 5.5 (MDM) 12. Mediterranean North (MDN) 9.4 13. Mediterranean South (MDS) 7.0

50.69 % 75.8

17.2 7.1

7 A Biophysical Typology in Agri-environmental Modelling

173

Mediterranean North and South EnZs. The areas less favorable for agriculture (Agri-mask classes one and two) are mainly located in the Alpine, Boreal and Mediterranean Mountain EnZs. These EnZs show much lower relative shares of land classified as land with no or relatively small constraints to agriculture (Agri-mask class zero) than the overall EU27+ figure of 75.8%. Furthermore, lower organic carbon contents are overrepresented in the Agri-mask class zero while the higher organic carbon contents (classes five and six) are underrepresented in this class (see Hazeu et al. 2006). The main reason for this overrepresentation is that arable farming which is particularly present in this class results in lower carbon contents in the topsoil. The agricultural most important EnZs are the Atlantic North and Central, Continental and Pannonian EnZs (and to a lesser extent the Lusitanian and Mediterranean EnZs) as they have relatively high shares of land used for agriculture (mainly non-irrigated arable land). The Alpine North and South, Boreal, Nemoral and Mediterranean Moutains EnZs show low proportions, even for the area which is potentially suitable (Agri-mask class zero) for agriculture. Furthermore, the proportion of agricultural land is highest on soils low in carbon and decreases with increasing carbon content for most EnZs. The agricultural land that is potentially suitable within the agricultural important EnZs is mainly situated below 200 m. The southern environmental zones (Atlantic Central, Pannonian, Lusitanian and Mediterranean EnZs) have a longer growing season (>250 days with temperatures above 5°C) than the northern zones (Atlantic North, Continental). The mean annual temperature range (temperature August minus temperature August) for the Atlantic and Lusitanian EnZs is below 15°C indicating a sea climate with relatively small temperature variations. The rainfall minus potential evapotranspiration (ETp) summed over May, June and July (summer drought) is less than 150 mm with as exceptions the Pannonian, Lusitanian and Mediterranean EnZ with deficits from 150 mm up to almost 450 mm. The variation of slope is small as the Agri-mask class zero land is land with slopes below 8%. The available water holding capacity is around 0.15 with as exception the OCTOP five class with relatively high available water holding capacity values. High shares of deep soils (soils deeper than 80 cm) are found in the Atlantic, Continental and Pannonian EnZs. Medium textured soils are most prominent in the agricultural important AEnZ land types as they surpass 60% of the area. For a more elaborate description on the variation between EnZs and within EnZs for all AEnZ land types we refer to the report “Regional typologies of ecological and biophysical context” (Hazeu et al. 2006).

Seamzones Definition For the use in SEAMLESS the agri-environmental zoning described in the previous sections has been used as a basis for delineating spatial units (Seamzones) to be used for modelling. For this purpose it was decided to combine three different layers:

174

G. Hazeu et al.

–– Administrative regions (270 NUTS2 regions); –– Environmental Zones (12 classes); –– Soil types (seven OCTOP classes). A subdivision in suitability classes of land for agriculture as defined by the Agri-mask was not included – mainly to restrict the number of regions to be modelled. Therefore, unsuitable land is not excluded in the Seamzones. The spatial allocation of farm types described in the next section ensures that restrictions on farmed area and types of farming in areas with limitations for farming, such as the areas indicated by the Agri-mask layer, are included. The links between the spatial levels and modelling are described later in this chapter. The top layer of the Seamzones used in SEAMLESS are the NUTS22 regions as defined by Eurostat. This layer includes 270 regions in EU27 plus Liechtenstein, Norway and Switzerland with an average size of 1.7 million hectares. The smallest of these regions is 16,041 ha, while the largest extends to 16.5 million hectares. Adding the layer of environmental zones to the NUTS region layer results in 591 regions corresponding to 2.2 environmental zones per NUTS regions. These combined regions have an average size of 784,711 ha, ranging from 88 to almost 11 million hectares. The subdivision of the EnZ by NUTS2 regions is meant to develop zones which are homogeneous in terms of climatic conditions, so we may call it a climate zone. The NUTS2-specific climatic data represent a major refinement compared to the average climatic data per EnZ, and they can be combined conveniently with the land use and farm statistical data which are available. Within a given NUTS2 region more than one EnZ may be present. The climate zone is further subdivided in Seamzones on the basis of differences in soils. A Seamzone is homogeneous in terms of farming conditions as defined by climate, soil and administrative region. Adding the soil type layer to reach the final Seamzones results in a total of 3,513 regions. These Seamzones have an average size of 132,013 ha, ranging from 1 ha and up to almost 7.6 million hectares. The size distribution of the Seamzones is explored further in Table 7.3. As it can be seen half of the Seamzones are smaller than 29,400 ha, 10% are smaller than 970 ha and 90% is smaller than 334,634 ha. This large variation in the size of the Seamzones reflects the large variation in homogeneity of biophysical characteristics in different parts of the European Union. The Seamzones do not form contiguous areas but consist of several separate areas. This spatial distribution is mainly a consequence of the scattered pattern of Table 7.3 Size distribution of the Seamzones (in hectares) Deciles 10 20 30 40 50 60 70 80 90 100 Size 970 3,366 8,113 15,582 29,400 54,501 95,800 167,646 334,634 7,599,200

NUTS is the abbreviation for Nomenclature of Territorial Units for Statistics. NUTS regions are based on existing national administrative subdivisions. There are three NUTS levels defined. The NUTS2 regions in the Netherlands are equal to provinces.

2

7 A Biophysical Typology in Agri-environmental Modelling

175

soil types, whereas the combinations of NUTS2 regions and Environmental zones generally are continuous areas. The OCTOP used in the definition of the Seamzones, is a function of ESDB soil types, temperature and land cover. The methodology used to compile the OCTOP database resulted in this spatial pattern. In Fig. 7.3 the scattered distribution of a Seamzone (dark) is illustrated for Sicily in Italy. Sicily is a NUTS2 region in the MDM, MDN and MDS EnZs with a total of 16 Seamzones. The Seamzone presented in dark is a combination of the NUTS2 region Sicily, MDS EnZ and OCTOP class two (see Fig. 7.3). The map also illustrates that the processing of data for the Seamzones was done at a one km grid, so that the smallest units are 100 ha (an exemption from this is along coastlines, where the actual coastline is used). For all 3,513 Seamzones the average number of scattered areas is 162. Ten percent of the Seamzones consists of four or less subareas, half of the Seamzones of 58 or less sub areas and 90% of the Seamzones of 398 or less subareas. The maximum number of subareas within a single Seamzone is 4,322 (Table 7.4).

Fig. 7. 3 The spatial pattern of the 16 Seamzones on Sicily, Italy. As an example the Seamzone (NUTS2: region Sicily, EnZ: Mediterranean South, OCTOP class: 1.23–2.46%) indicated with the grey colour includes 692 areas scattered across the island. The other lines in the map are borders between the other Seamzones (white areas) Table 7. 4 Distribution of number of separate areas within the Seamzones Deciles Number of areas

10 4

20 10

30 20

40 35

50 58

60 92

70 144

80 224

90 398

100 4,322

176

G. Hazeu et al.

Climate The climatic data used in the SEAMLESS project are part of the Joint Research Centre Monitoring Agriculture with Remote Sensing (JRC MARS) database of the MARS Crop Yield Forecasting System (Micale and Genovese 2004; http://mars.jrc.it/marsstat; Genovese et al. 2007). This database provides daily values that are needed as input data for the model chain. It contains European wide, quality checked daily data since 1975. The database is based on observed meteorological station data interpolated to a 50 by 50 km grid (MARS-grid). Daily data of roughly 1,300 stations is interpolated to a 50 by 50 km grid (MARS-grid) (Micale and Genovese 2004). In the frame of SEAMLESS a meteorological database has been developed, based on a weighted average of grid weather values for each Climate Zone (NUTS2/EnZ combination), according to the surface covered by the NUTS2/EnZ combination within the MARS-grid. The occurring variability of the parameters within the NUTS2/EnZ combinations, due to the different grids that are averaged according to their spatial weight, is described via minimum, maximum values and standard deviation of the parameters. As a result one Climate Zone is described by the time series of daily rainfall data over the period 1975–2007 for the assumed representative point (pseudo weather station) of the climate zone. To test the aggregated data the long term average for the climate zones was calculated from 1975 until 2004. If the aggregated data at the level of Climate Zones is compared with the patterns coming from the MARS Meteorological Grid database they are well reflected for the majority of the countries, depending on the NUTS2 area variability. The correlation of the Environmental Stratification with the climatic parameters from the JRC MARS Climate database can be assessed for example in using the annual temperature sum as shown by Metzger et al. (2005). It showed a high correlation (R2 of the regression of 0.95). Figure 7.4 shows the overlay of the annual temperature sum (Tbase = 0) from the long term average (1975–2004) with Environmental zones of the EnS highlighting the spatial homogeneity of the zones in terms of temperature sums. Similar patterns are found when using other relevant climatic parameters based on the long term average. Climatic gradients that did not lead to additional zones in the EnS, as climatic factors where only one part of the selection criteria, are partly reflected in the Climate Zones, for example the rainfall gradient from the west to the east coast in Great Britain (see Fig. 7.5). Climate Zones in Great Britain are well suited to highlight the specific distribution of weather parameters throughout the country. This is valid for the majority of the countries and areas, whereas Sweden, Norway, Finland and the Baltic States show a much coarser pattern due to the large administrative units averaging partly different climatic situations within the Climate Zones. Soil For modelling purposes it is needed to provide one soil profile description per Seamzone. This is a complicated task for two reasons. Firstly, the majority of soil

177

Fig. 7. 4 Annual temperature sum (Tbase = 0o) based on the long term average from 1975 until 2004 for the 50 km by 50 km MARS-grids in the EU27, Norway and Switzerland

Fig. 7. 5 Average rainfall of the Climate Zones (NUTS2/EnZ combinations) in November based on the long term average from 1975 until 2004 in the EU27, Norway and Switzerland. The bold lines indicate the EnZs while differences within the EnZs reflect the Climate Zones with different average rainfall for November

178

G. Hazeu et al.

data are of the categorical type, so it is not possible to calculate averages for the Seamzones as can be done with for example climatic data. Instead it is needed to select one soil profile description from the original data to represent the Seamzone. Secondly, the mapping units of the European soil map, i.e. the so-called Soil Mapping Units (SMUs), are described by a number of Soil Typological Units (STUs). The STUs are described by a representative soil profile including information on the share of it within a specific SMU. The SMUs are associations of STUs without a spatial indication where the STUs within a SMU occur. Therefore, an overlay of the Seamzones with the soil map (or SMUs) will not directly deliver the most dominant soil type (STU) per Seamzone. Another problem is that it might be that the most dominant STU actually is not representative for the soil variables that are most important from an agricultural point of view. The following procedure was therefore adopted to select a specific soil profile to represent the Seamzones: 1. The coverage of STUs was calculated by overlaying the Seamzones with the SMUs and using the result for weighting the share of the STUs coverage of the Seamzones. If for example, a STU covers 50% of a Soil Mapping Unit that covers 25% of a Seamzone, this specific STU covers 12.5% of the Seamzone. 2. In a second step it was decided to focus on two variables that are of high importance for agriculture: texture in topsoil and rooting depth. The dominant combination of these two variables was therefore calculated per Seamzone. 3. Finally, the soil dataset of the most dominant STU among the ones with the dominant combination of texture and rooting depth class identified under point 2 was selected as the descriptive set of soil data for the Seamzone. The relation between the parameter used for defining the soil types (carbon content) and the two other variables used in the selection of representative soil profiles can be explored in Tables 7.5 and 7.6. As can be seen from Table 7.5 there is no clear relationship between the different soil types and the rooting depth. The share of the different soil types that is without obstacles for roots to a depth of 80 cm is on average 64% if the soil with no information is excluded. The only tendency that can be seen from the table is that obstacles for roots are more often found for the soil types with low carbon content. Table 7.5 Shares of the area of the different soil types (characterized by organic carbon content in %, first column) that fall in different categories of rooting depth. The last column, 0–80 cm, contains soil types where obstacles do occur, but the exact depth between 0 and 80 cm is not specified OCTOP class No obstacles 60–80 cm 40–60 cm 20–40 cm 0–80 cm 0.1–1.23% 59 6 19 15 0 1.23–2.46% 72 6 7 16 0 2.46–3.94% 66 12 11 10 0 3.94–5.66% 59 6 24 10 1 5.66–8.86% 67 12 15 5 0 8.86–63.0% 63 5 29 3 0 No information 51 4 22 23 0 Total 64 8 17 11 0

7 A Biophysical Typology in Agri-environmental Modelling

179

Table 7.6 Shares of the area of the different soil types (characterized by organic carbon content in %, first column) that fall in different categories of texture of topsoil (%) OCTOP class Coarse Medium Medium fine Fine Very fine No mineral texture 0.1–1.23% 10 63 21 6 0 0 1.23–2.46% 23 43 15 17 1 0 2.46–3.94% 22 44 15 20 0 0 3.94–5.66% 25 58 14 3 0 0 5.66–8.86% 44 47 5 1 0 3 8.86–63.0% 48 39 1 2 0 10 No data 44 44 4 4 0 4 Total 29 48 12 9 0 2

For the texture in the topsoil (see Table 7.6) there is a clear relationship with the soil types (organic carbon content). The soil types with lower carbon content are generally with a lower share of coarse soils and a higher share of medium fine and fine textured soils. However, it is worth noting that for the most widespread texture type, medium, the picture is more complex, with the soil types with 0.1–1.23% and 3.94–5.66% carbon content having the highest share of this texture category. The largest share of non-mineral soils is, not surprisingly, found for the soil types with highest carbon content. It is also worth noting that a high share of the areas with no information on soil type are of the texture classes that are considered poor in agronomic terms, with a very high percentage of medium and coarse textured soils. Overall, looking at the three key variables, it is clear, firstly that almost two thirds of the EU27+ is without identified obstacles. Secondly, the variables carbon content and texture are clearly related. Thirdly, the fact that the information on carbon content is available as continuous data as described above, gives this variable an advantage compared to texture. Finally, as described earlier, the information on carbon content is available in 1 km2 grids, which is more suitable for mapping than the other variables. Overall, it is thus clear that the selection of carbon content to define the soil types and thus also to map the Seamzones seems justified. One final remark on the processing of the data for SEAMLESS is, that the selection of representative soil profiles tends to focus on the most dominant soil profiles (Soil Typological Units). The original dataset with STUs holds information on app. 5,300 sets of soil information. However, only 844 STUs, i.e. the dominant ones with a dominant combination of soil texture and rooting depth per Seamzone, have been chosen using the method described above. On average one set of soil information from the STUs is used to describe 4.2 Seamzones. The distribution is uneven as can be seen in Table 7.7: half of the soil sets is used to describe one or two Seamzones, whereas the 10% of the soil sets that is used most frequently is used to describe ten or more Seamzones. Already in the original data used to describe the soil characteristics the heterogeneity of the soils across the EU is not fully described. The processing of the data to Seamzones has increased this problem focusing on the most dominating soils. However, due to the type of information on soil variables, mainly categorical data, this is unavoidable.

180

G. Hazeu et al.

Table 7.7 The distribution of the number of Seamzones represented per Soil Typological Unit Deciles Number of Seamzones per STU

10 1

20 1

30 1

40 2

50 2

60 3

70 4

80 6

90 10

100 27

Applications This section gives a few examples on how the Agri-Environmental Zonation is applied in the SEAMLESS project. Firstly, the use of Seamzones as a stratification to select so-called sample regions is described. Secondly, it is described how statistical information on farm types and biophysical data are linked using the Seamzones as spatial framework. Finally, an overview of the use of biophysical data as model input in SEAMLESS is given.

Selection of Sample Regions The Seamzones have been used as a framework for selection of sample regions. Sample regions are used to collect detailed information on farm management not available in the European level statistical sources. This again enables detailed modelling at crop and farm type level within these regions. The starting point for the selection of sample regions was that NUTS2 regions should be selected as this is the level for market level agricultural modelling in SEAMLESS. Another starting point was that a total number of regions of ca. 20 was judged to be feasible for modelling purposes. Finally, the selection of the regions should also take into account that the collection of data had to prioritize a limited number of regions, where a more detailed set of data could be collected, taken into account that these regions should still represent the variation in biophysical conditions for farming across EU25. An optimal solution for this would be to select regions that could represent each of the 12 environmental zones. However, some of the environmental zones on the one hand occur in complex patterns (the zones highly influenced by altitude rather than latitude/longitude) and on the other hand are less important from an agricultural point of view. This is the case for the zones Alpine south, Alpine north and Mediterranean mountains. Optimally, there should be one detailed sample region within the nine remaining environmental zones. As a second step it was decided to aim that the remaining sample regions should ensure representation of the variation in farm types within all environmental zones. This was done by selecting regions that together included the most important farm types in terms of area farmed. Typically, this resulted in selection of regions that differed in terms of for example arable versus livestock farms and in terms of small versus large farms.3 At the time of the selection FADN data on farm types were not yet available for the ten new Member States. Therefore information on intensity of farming was not available across the entire territory and was not used as a basis for the selection of sample regions.

3

7 A Biophysical Typology in Agri-environmental Modelling

181

The soil dimension of the Seamzones was not used as such in the selection of the sample regions. On average 5.9 Seamzones are found within the combinations of NUTS2 regions and environmental zones (or sample regions) thus representing 5.9 soil types out of seven possible (six soil types plus no data). It can therefore be concluded that the variation in soil types is represented be default in the selection, at least in terms of the carbon content as this is defining the soil types. In summary, the agri-environmental zonation was used as a basis for the selection of sample regions that ensures a good representation of the variation in conditions for farming across EU25. This again ensures that the different modelling approaches in SEAMLESS can be combined and that scaling of model results is facilitated.

Allocation of Farms The strata of the AEnZ of Europe are used for the agricultural modelling in SEAMLESS. Biophysical data and information on farming activities are required for this modelling. However, from existing European statistical sources, e.g. FADN (Farm Accountancy Data network) 4 and Farm Structure Survey (FSS), land use and the farm information is only available at administrative level. These administrative regions are usually quite large and have a very large variation in environmental characteristics. In most cases they cut through several AEnZ. A spatial allocation approach has therefore been developed in CAPRI-Dynaspat (Dynamic and Spatial dimension of CAPRI) (Kempen et al. 2007) and SEAMLESS to add a spatial dimension to all land uses and farms respectively. This disaggregation approach makes it possible to aggregate the land use and the farms to the Agri-Environmental Zones used in SEAMLESS. The spatially allocated farm type characteristics together with the characteristics of the AEnZ facilitate the modelling of environmental effects of farming. At the same time these combination of farms with AEnZ enable the linking of models that are applied at different scales such as the farm scale (e.g. the bio-economic farm model FSSIM) and the field scale (the cropping system model APES) and models that are covering different domains (e.g. environment, agriculture, markets).

FADN data are the main input data source for the Common Agricultural Policy Regional Impact (CAPRI) model. Farm Accountancy Data Network (FADN) (sample of holdings, representing a large share of agricultural production, with information on costs and revenues, income generated from agriculture, also including subsidies). A database, provided by the Commission of the European Communities. Figures are available on an annual basis for the European Union as a whole, distinguishing FADN regions (NUTS 1 or NUTS 2) representative for the main farm types. FADN is based on a representative sample of holdings.

4

182

G. Hazeu et al.

Model Input in SEAMLESS The Seamzones are used in SEAMLESS as the most spatially detailed specific combination of weather and soil characteristics. These biophysical data for Seamzones are used directly as model input for: –– Biophysical modelling of agricultural activities and environmental externalities. Some key variables here are data on soil thickness, carbon content and texture and time series of daily values on temperature and rainfall (see also Chapter 4 of this book). –– Bio-economic farm type modelling: the biophysical data on soil and climate are used together with rotation restrictions as constraints to define feasible sets of crop rotations within a given region (see also Chapter 5 of this book). –– Meta-modelling to ensure consistency between farm type level and market level modelling. The climate and soil data are used to extrapolate supply responses of farm types from regions, where bio-economic modelling is carried out to other regions within the entire European Union (see also Perez-Dominguez et al. 2009 for the EXPAMOD model). The flow of information between these different levels of modelling is ensured because the models are applied always to Seamzones or to aggregations of Seamzones (climate zones or NUTS2 regions).

Discussion and Conclusions Agri-environmental Zones The AEnZ is a new typology and does provide a dedicated and consistent framework for agri-environmental modelling in Europe. The zonation is detailed enough to cover the wide spectrum of bio-physical environments in Europe and also simple enough to present modelling results. AEnZ is a hierarchical and flexible subdivision of the European landscape into 238 relatively homogeneous units from an agronomic perspective. The AEnZ is based on 12 environmental zones, seven topsoil organic carbon content classes and an Agri-mask that consist of three classes. From the 238 classes of the AEnZ, 82 classes referring to the Agri-mask 0 (75.8% of EU27+) are described in terms of land cover, climate and biophysical parameters. A PCA of several soil parameters revealed that organic carbon content in the topsoil explained most of the variation. Therefore, this variable was selected as the variable to differentiate between soils within EU27+. A review of environmental typologies resulted in the selection of the EnS as a basis for the AEnZ typology. The EnS is a statistically robust environmental classification of sufficient detail, used in several EU projects and published in

7 A Biophysical Typology in Agri-environmental Modelling

183

peer-reviewed journals. Furthermore, it shows high correlations with climate factors. However, the EnS has to be extended with one or more soil variables to come to a spatial framework that covers the wide range of agri-environmental diversity in Europe. The AEnZ typology proved to be a useful agri-environmental modelling framework for the SEAMLESS project. However, some remarks on the limitations of the typology can be made. In general it can be stated that to build a biophysical typology there exists a need for more detailed input data than currently available. More efforts should therefore be invested in the collection of biophysical data and the building of better bio-physical databases that represent more precisely the diversity and spatial detail of the European environment. The quality of the soil information used is well described in the SINFO study (Baruth et al. 2006). One major drawback of the soil data in the European soil database (ESDBv2) is the inconsistency between countries. The statistical procedure to select the soil variable which explains best the variation of soils in Europe shows preferences for selecting homogeneous databases as the OCTOP database with continuous variables. This problem can be overcome only by improving soil databases (detail, homogeneous).

Climate Zones The description of the EnZ by the JRC MARS climate data reflects very well the spatial homogeneity of these EnZ units. It did not lead to additional zones as climate variables were only one factor in the definition of the EnZ. Climatic gradients are partly reflected in the Climate Zones (NUTS2/EnZ units), but this has mainly to do with the size of the administrative regions. Large administrative regions have still different climatic conditions. Furthermore, the level of detail coming from the MARS climate database (50 km2) reflects only regional climatic conditions. A description of Seamzones by climate data does not make sense as soils vary more than climate.

Seamzones For modelling purposes Seamzones are defined as the spatially most detailed combination of administrative NUTS2 regions, weather and soil characteristics. They are by definition a unique combination of a NUTS2 region, an environmental zone and a soil type (topsoil organic carbon class). The Agri-mask of the AEnZ is not taken into account as this would result into too many regions to be modelled. The resulting Seamzones are 270 NUTS2 regions subdivided by environmental zones into 591 regions with a mean of 2.2 environmental zones per NUTS2 region. Combined with the soil type layer, this resulted in 3,513 regions ranging from 1 ha up to more than 7.5 million hectares (average size 132,013 ha, 50% of Seamzones smaller than 29,400 ha). The 3,513 Seamzones do not have a continuous extent in

184

G. Hazeu et al.

space mainly due to the scattered distribution of soil types. On average a Seamzone consists of 162 individual areas. A representative and complete soil profile, and not only a topsoil organic carbon class, is attached to a Seamzone. A representative soil profile is selected using the ESDB dataset, which means the STU that is most present in the Seamzone and that has at the same time the most dominant combination of the soil variables topsoil texture and rooting depth. On average the soil information from a STU is used to describe 4.2 Seamzones. However, this distribution is uneven as 50% of the STUs is used to describe one to two Seamzones. All selected soil profiles per Seamzone do not reflect the heterogeneity of soils across Europe. This originates from the fact that the original data of the ESDB do not fully describe the heterogeneity of the soils (categorical data, level of detail). Furthermore, the selection of dominating soils further homogenize the picture of soils across Europe. This is in contrast with the OCTOP data (soil types) in the AEnZ which show a more detailed/fragmented pattern due to the higher spatial detail and character of the soil variable (dependent among other things on land cover).

Applications The sample regions, which is a NUTS2 region, are selected by stratified sampling to represent the heterogeneity of farming systems, climate and soil conditions across EU25. The AEnZ provides a very important element of the spatial framework for the selection of sample regions within SEAMLESS. The allocated farm type information in combination with the spatial framework of the AEnZ facilitate the modelling within SEAMLESS. Besides the modelling of environmental effects of changes in agriculture, it is a basis for linking different models at different levels of detail and/or domains.

References http://edc.usgs.gov/products/elevation/gtopo30/gtopo30.html http://eusoils.jrc.it/ESDB_Archive/ESDBv2/fr_intro.htm http://www.iiasa.ac.at/Research/LUC/luc07/Research-AEZ/index.html?sb = 8 Andersen, E., Baldock, D., Bennett, H., Beaufoy, G., Bignal, E., Brouwer, F., Elbersen, B., Eiden, G., Godeschalk, F., Jones, G., McCracken, D., Nieuwenhuizen, W., van Eupen, M., Hennekens, S., & Zervas, G. (2003). Developing a high nature value farming area indicator. Internal report for the European Environment Agency. Anger, M., Malcharek, A., & Kuhbach, W. (2002). An evaluation of the fodder values of extensively utilised grasslands in upland areas of Western Germany. I. Botanical composition of the sward and DM yield. Journal of Applied Ecology, 76(1–2), 41–46. Baldock, D., Beaufoy, G., Brouwer, F., & Godeschalk, F. (1996). Farming at the margins; abandonment or redeployment of agricultural land in Europe. London/The Hague: Institute for European Environmental Policy (IEEP) and Agricultural Economics Research Institute (LEI).

7 A Biophysical Typology in Agri-environmental Modelling

185

Baruth, B., Genovese, G., Andersen, E., Diepen, C.A., Elbersen, B.S., & Hazeu, G.W. (2006a). A spatial framework for environmental data. SEAMLESS Deliverable PD4.3.2, SEAMLESS integrated project, EU 6th Framework Programme, contract no. 010036-2 (p. 35), www. SEAMLESS-IP.org Baruth, B., Genovese, G., & Montanarella, L. (Eds.) (2006b). New soil information for the MARS crop yield forecasting system. EUR Report EUR 22499 EN (p. 95), with Annex pp 86. Ispra, Italy: Joint Research Centre of the European Commission. Beaufoy, G., Baldock, D., & Clark, J. (1994). The nature of farming: Low intensity farming systems in nine European countries. London: Institute for European Environmental Policy (IEEP). Bignal, E. M., & McCracken, D. I. (1996). Low-intensity farming systems in the conservation of the countryside. Journal of Applied Ecology, 33, 413–424. Bignal, E. M., & McCracken, D. I. (2000). The nature conservation value of European traditional farming systems. Environmental Reviews, 8, 149–171. Boatman, N., Stoate, C., Gooch, R., Rio Carvalho, C., Borralho, R., De Snoo, G., & Eden, P. (1999). The environmental impact of arable crop production in the European Union. Practical options for improvement. EC-study contract B4-3040/98/000703/MAR/D1. Leicester: Allerton Research and Educational Trust. Bohn, U., Gollub, G., & Hettwer, C. (2000). Karte der natürlichen Vegetation Europas: Maßstab 1:2.500.000. Bad Godesberg, Bonn: Bundesamt für Naturschutz. Buckwell, A. E., & Armstrong-Brown, S. (2004). Changes in farming and future prospects: Technology and policy. IBIS International Journal of Avian Science, 146(S2), 14–21. Bunce, R. G. H., Barr, C. J., Clarke, R. T., Howard, D. C., & Lane, A. M. J. (1996). Land classification for strategic ecological survey. Journal of Environmental Management, 47, 37–60. Bunce, R. G. H., Metzger, M. J., Jongman, R. H. G., Brandt, J., de Blust, G., Elena-Rossello, R., et al. (2008). A standardized procedure for surveillance and monitoring European habitats and provision of spatial data. Landscape Ecology, 23, 11–25. Bunce, R. G. H., Watkins, J. W., Brignall, P., & Orr, J. (1996). A comparison of the environmental variability within the European Union. In R. H. G. Jongman (Ed.), Ecological and landscape consequences of land use change in Europe (pp. 82–90). Tilburg, The Netherlands: European Centre for Nature Conservation. Büttner, G., Feranec, J., Jaffrain, G., Mari, L., Maucha, G., & Soukup, T. (2004). The CORINE Land Cover 2000 Project. EARSeL eProceedings, 3(3), 331–346. Campbell, L. H., & Cooke, A. S. (eds). (1997). The indirect effects of pesticides on birds. Peterborough: Joint Nature Conservation Committee. Cochran, W. G. (1997). Sampling techniques (3rd ed.). New York: Wiley. de Miguel, J. M. (1999). Nature and configuration of the agrosilvopastoral landscape in the conservation of biological diversity in Spain. Revista Chilena de Historia Natural, 72(4), 547–557. Easterling, D. R., Thomas, C. P., & Thomas, R. K. (1996). On the development and use of homogenized climate datasets. Journal of Climate, 9, 1429–1434. EEA. (1999). Environment in the European Union at the turn of the century (Environmental Assessment Rep. No 2). Copenhagen: European Environment Agency, from http://www. eea.eu.int/ EEA. (2002). The biogeographical regions map of Europe. Copenhagen: European Environment Agency. EEA. (2004). High nature value farmland, characteristics, trends and policy challenges (EEA report 1/2004). Copenhagen: European Environment Agency. EEA. (2007). Estimating the environmentally compatible bioenergy potential from agriculture (EEA Tech. Rep. No. 12/2007). Copenhagen: European Environment Agency. Elena-Rosselló, R. (1997). Clasificatión Biogeoclimática de España Peninsular y Balear. Madrid: Ministerio de Agricultura, Pesca y Alimentación. EPA. (1999). Inventory of greenhouse gas emissions and sink 1990–1997. Washington, DC: Environmental Protection Agency.

186

G. Hazeu et al.

Commission of the European Communities, DG-Agriculture. (2003). Reform of the common agricultural policy. Medium-term prospects for agricultural markets and income in the European Union 2003–2010. Brussels: Author. Genovese, G., Baruth, B., Royer, A., & Burger, A. (2007). MARS Stat action of the European commission – Crop and yield monitoring activities. GEOinformatics, 10(4), 20–23. Hazeu, G.W., Elbersen, B.S., van Diepen, C.A., Metzger, M.J., & Baruth, B. (2006). Regional typologies of ecological and biophysical context (SEAMLESS Rep. No. 14). SEAMLESS Integrated Project, EU 6th Framework Programme, contract no. 010036-2, www.SEAMLESS-IP. org, 55 pp., ISBN no. 90-8585-042-8. Heineke, H.J., Eckelmann, W., Thomasson, A.J., Jones, R.J.A., Montanarella, L., & Buckley, B. (Eds.) (1998). Land information systems: Developments for planning the sustainable use of land resources (European Soil Bureau Research Rep. No. 4), EUR 17729 EN. Luxembourg: Office for Official Publications of the European Communities. Hiederer, R. (2001). European Catchment Information System for agri-environmental issues. In Proceedings of Euroconference Link GEO and Water Research, Genoa, Italy, 7–9 February 2002. Jones, H. E., & Bunce, R. G. H. (1985). A preliminary classification of the climate of Europe from temperature and precipitation records. Journal of Environmental Management, 20, 17–29. Jones, R.J.A., Hiederer, R., Rusco, E., Loveland, P.J., & Montanarella, L. (2004). The map of organic carbon in topsoils in Europe, Version 1.2, September 2003: Explanation of Special Publication Ispra 2004 No.72 (S.P.I.04.72) (European Soil Bureau Research Rep. No.17), EUR 21209 EN, 26 pp. and 1 map in ISO B1 format. Luxembourg: Office for Official Publications of the European Communities. Jones, R. J. A., Hiederer, R., Rusco, E., & Montanarella, L. (2005). Estimating organic carbon in the soils of Europe for policy support. European Journal of Soil Science, 56, 655–671. Jones, R.J.A., Houšková, B., Bullock, P., & Montanarella, L. (Eds.) (2005a). Soil resources of Europe (2nd ed.) (European Soil Bureau Research Rep. No. 9), EUR 20559 EN, 420 pp. Luxembourg: Office for Official Publications of the European Communities. Jongman, R., Bunce, R., Metzger, M., Mücher, C., Howard, D., & Mateus, V. (2006). Objectives and applications of a statistical environmental stratification of Europe. Landscape Ecology, 21, 409–419. Kempen, M., Heckelei, T., Britz, W., Leip, A., Koeble, R., & Marchi, G. (2007). Computation of a European agricultural land use map - statistical approach and validation. Technical Paper of Institute for Food and Resource Economics University of Bonn. Leathwick, J. R., Overton, J. M., & McLeod, M. (2003). An environmental domain classification of New Zealand and its use as a tool for biodiversity management. Conservation Biology, 17, 1612–1623. MAFF. (1998). Official Group on OPs: Report to Ministers 1998. London: MAFF. Meeuws, J., Weijermans, M., & Vroom, M. (1990). Agricultural landscapes in Europe and their transformation. Landscape and Urban Planning, 18, 289–352. Metzger, M. J., Bunce, R. G. H., Jongman, R. H. G., Mücher, C. A., & Watkins, J. W. (2005). A climatic stratification of the environment in Europe. Global Ecology and Biogeography, 14(6), 549–563. Metzger, M. J., Rounsevell, M. D. A., Acosta-Michlik, L., Leemans, R., & Schröter, D. (2006). The vulnerability of ecosystem services to land use change. Agriculture, Ecosystems, and Environment, 114, 69–85. Metzger, M.J., Schröter, D., Leemans, R., & Cramer, W. (2008). A spatially explicit and quantitative vulnerability assessment of environmental change in Europe. Regional Environmental Change, 35. Micale, F., & Genovese, G. (Eds.) (2004). Methodology of the MARS Crop Yield Forecasting System (MCYFS) Vol. 1, Meteorological data collection, processing and analysis. EUR 21291 EN/1 (2005), 100 pp. Luxembourg: Office for Official Publications of the European Communities. Mitchell, T.D., Carter, T.R., Jones, P.D., Hulme, M., & New, M. (2004). A comprehensive set of high-resolution grids of monthly climate for Europe and the globe: The observed record

7 A Biophysical Typology in Agri-environmental Modelling

187

(1901–2000) and 16 scenarios (2001–2100) (Tyndall Centre Working Paper No. 55). Norwich: Tyndall Centre for Climate Change Research, University of East Anglia. Nagy, G. (2002). The multifunctionality of grasslands in rural development in a European context. Acta Agronomica Hungarica, 50(2), 209–222. Pau Vall, M., & Vidal, C. (1999). Nitrogen in agriculture. In CEC, agriculture, environment, rural development: Facts and figures – a challenge for agriculture. Perez-Dominguez, I., Bezlepkina, I., Heckelei, T., Romstad, E., Oude Lansink, A., & Kanellopoulos, A. (2009). Capturing market impacts of farm level policies: A statistical extrapolation approach using biophysical characteristics and farm resources. Environmental Science and Policy, 12(5), 588–600. Poiret, M. (1999). Crop trends and environmental impacts. In CEC, agriculture, environment, rural development: Facts and figures – a challenge for agriculture. Pretty, J. N. (1998). The living land: Agriculture, food and community regeneration in rural Europe. London: Earthscan. Sheail, J., & Bunce, R. G. H. (2003). The development and scientific principles of an environmental classification for strategic ecological survey in Great Britain. Environmental Conservation, 30, 147–159. Tappan, G. G., Sall, M., Wood, E. C., & Cushing, M. (2004). Ecoregions and land cover trends in Senegal. Journal of Arid Environments, 59(3), 427–462. Van Ittersum, M. K., Ewert, F., Heckelei, T., Wery, J., Alkan Olsson, J., Andersen, E., et al. (2008). Integrated assessment of agricultural systems - A component-based framework for the European Union (SEAMLESS). Agricultural Systems, 96(1–3), 150–165. Wadsworth, R.A., Carey, P.D., Heard, M.S., Hill, M.O., Hinsley, M.S., Meek, W.R., Panell, D., Ponder, V., Renwick, A., & James, K. (2003). A review of research into the environmental and socio-economic impacts of contemporary and alternative cropping systems. Report to Defra, pp. 85. Wrbka, T., Szerncsits, E., Moser, D., & Reiter, K. (1999). Biodiversity patterns in cultivated landscapes: Experiences and first results from a nationwide Austrian survey. In M.J. Maudsley & E.J.P. Marshall (Eds.), Heterogeneity in landscape ecology: Pattern and scale (pp. 276–286). Lymm: IALE.

Chapter 8

The Use of Regional Typologies in the Assessment of Farms’ Performance Ida J. Terluin, David Verhoog, and Frans E. Godeschalk

Introduction In SEAMLESS a farm typology based on economic size, farming intensity, specialization and land use has been designed (Andersen et al. 2006). It builds upon the EU farm typology, that has already been in use for presenting and assessing the situation in the agricultural sector in the EU since the 1970s (Andersen et al. 2007). This EU farm typology classifies holdings according to their main source of income. As such, its rationale is exclusively economic. As a result of the subsequent CAP reforms (Mac Sharry reform, Agenda 2000 and CAP 2003 reform) and the Health Check (2008), the original scope of market and price support of the CAP has widened to support for the enhancement and protection of natural resources and landscapes, and rural viability. Given these changes, an extension of the EU farm typology by an environmental rationale – like conducted in the SEAMLESS farm typology – could create a typology, that might serve as a tool for the monitoring and assessment of the economic and environmental performance of farms under the reformed CAP. In the SEAMLESS farm typology it is aimed to classify all EU farms in groups that are homogeneous in their economic and environmental performance (Andersen et al. 2006). By doing so, the SEAMLESS farm typology builds upon previous studies, in which alternative typologies have been tested on limited groups of farms, like the project European Livestock Policy Evaluation Network (ELPEN) (Andersen et al. 2004), the project aiming to identify farms managing High Nature Value farmland (Andersen et al. 2003 and EEA/UNEP 2004), and the IRENA indicator project on the development of farming in the EU15 (EEA 2005). For classifying farms, four differentiating dimensions were used in the SEAMLESS farm typology.

I.J. Terluin (*), D. Verhoog, and F.E. Godeschalk LEI Wageningen UR, P.O. Box 29703, 2502 LS, The Hague, The Netherlands e-mail: [email protected]; [email protected]; [email protected] F.M. Brouwer and M. van Ittersum (eds.), Environmental and Agricultural Modelling: Integrated Approaches for Policy Impact Assessment, DOI 10.1007/978-90-481-3619-3_8, © Springer Science+Business Media B.V. 2010

189

190

I.J. Terluin et al.

First, the economic size of the farm measured in European Size Units (ESUs) was used to distinguish small scale, medium scale and large scale farms. It is thought that economic size reflects both economic and social aspects of farming: small and large farms might differ in their response to policy and market changes and in their contribution to the viability of rural areas. Second, farming intensity expressed in terms of agricultural output in euro per hectare was used to classify farms into low intensity, medium intensity and high intensity farms. Farming intensity is presented as a measure for both economic and environmental performance: low intensity farms with a low output per hectare are likely to have a lower pressure on the environment than high intensity farms. Third, specialization in agricultural activities was applied to classify farms in the farm types used in the EU farm typology. However, in the SEAMLESS farm typology only ten farm types are distinguished against 26 in the EU farm typology. Specialization is believed to provide information on the economic performance of the farm and on likely future choices on farm management. Fourth, land use types were defined dependent on thresholds for the use of specific land, like grassland, horticultural crops etc. This resulted in nine land use types. Land use is related to the environmental impact of farming. Usually, land use on livestock farms reflects feeding strategies varying from extensive grassland to highly intensive arable crops, whereas land use on crop farms may vary from rotation and mixed cropping strategies to monocultures and highly specialized intensive cropping. Three size types, three intensity types, ten specialization types and nine land use types combined results in potentially 810 types. In order to reduce the number of types, the specialization and land use dimension were combined in the SEAMLESS farm typology. By doing so, the number of types have been reduced to 189. According to the SEAMLESS farm typology, 12% of the EU15 farms are classified as low intensity farms, over 50% as medium intensity farms and 35% as high intensity farms in 2003 (Andersen et al. 2006). Low intensity farms cultivate 24% of the utilized agricultural area in the EU15, medium intensity farms 62% and high intensity farms 15%. Low intensity farms produced only 3% of the agricultural output in the EU against a share of 60% on high intensity farms. In the SEAMLESS farm typology, the regional context of the farm is not taken into account. This implies that the impact of socio-economic and physical characteristics of the territorial context on the economic and environmental performance of farms is disregarded. It could be wondered whether such regional characteristics could be included in the SEAMLESS farm typology. If we would do so, the number of types in the SEAMLESS farm typology will rise. This would create a rather untransparent typology. Therefore, in this chapter we use regional typologies instead of extended farm typologies for exploring whether the economic and environmental performance of farms is affected by regional characteristics. In particular, we focus on three regional typologies based on rurality, the stage of economic development and the presence of Less Favoured Areas (LFA). In order to avoid too many farm types, we base our analysis on the average regional farm. The economic and environmental performance of this average regional farm is – like in the SEAMLESS farm typology – approached by agricultural output in

8 The Use of Regional Typologies in the Assessment of Farms’ Performance

191

euro per hectare. Agricultural output per hectare depends on a large number of factors, like the type of crops or animals, the input of labour, the use of machinery, fertilizers and chemicals, price ratios and soil conditions. Agricultural output per hectare could be used as an indicator for land productivity (economic performance), and as such it reflects also the intensity of farming (environmental performance). A relatively low land productivity could be said to denote a rather extensive way of farming, whereas a relatively high land productivity reveals a more intensive type of farming. Our analysis covers 100 regions in the EU15 and is based on 2 years: 1990 and 2003. The focus of the chapter is first on the methodological approach of our analysis. Successively we discuss our hypotheses, our definition of farming intensity, the use of data and the design of regional typologies. We will then elaborate on the results of the testing of the hypotheses. In the final section we make some concluding remarks.

Methodological Approach In order to explore whether the intensity of farming is related to socio-economic and physical characteristics of EU regions, we consider a number of hypotheses, that are explained below. Hypothesis 1 Farming in rural regions tends to be less intensive than farming in urban regions. This hypothesis assumes a positive relationship between farming intensity and the degree of urbanization. Von Thünen already put forward this idea in the study ‘Der isolierte Staat in Beziehung auf Landwirtschaft und Nationalökonomie’ in 1826. In this study, Von Thünen tried to explain the location of agricultural production (van den Noort 1980). He assumed a situation with a central city surrounded by a homogeneous plain. Prices for agricultural products depend on transport costs to the city. According to Von Thünen, agricultural production is organized in a series of concentric zones around the city. These concentric zones or rings correspond with the intensity of production. In the first ring around the city, horticulture and milk production take place. In the next rings, cereals are produced. These rings are distinguished by their mode of cultivation: two-crop rotation, pastoral rotation and three-crop rotation. The outer ring is used for extensive cattle breeding. In order to translate Von Thünen’s theory to our hypothesis, we could assume that the city and its first ring correspond to an urban region, and that the subsequent rings are more or less similar to rural regions. A more recent underpinning of the hypothesis was given by Hayami and Ruttan (1985). It could be said that their induced technological innovation model is derived from two premises: in the process of agricultural development an inelastic supply of land may be offset by advances in biological and chemical technology, whereas constraints imposed by an inelastic supply of labour may be offset by advances in

192

I.J. Terluin et al.

mechanical technology. Biological and chemical technology in this context has to be understood as the whole set of advances that increases the crop output per unit of land area or that improves the yield of animal products per unit of feed or per unit of breeding stock, while mechanical technology includes all changes in the use of mechanical equipment designed to bring about larger output per worker by increasing the land areas that can be operated per worker. It might be clear that biological and chemical technology tends to be land-saving and increases land productivity, whereas mechanical technology is rather labour-saving and increases labour productivity. Given these two premises, Hayami and Ruttan (1985) assume in their model that a process of induced technological innovation starts in response to a change in relative factor prices: farmers are induced to search for technical alternatives that save the increasingly scarce factors of production. So whenever the price of labour increases relative to the price of land, induced innovation refers to mechanical technology; on the other hand, whenever the price of land increases relative to the price of labour, induced innovation refers to biological and chemical technology. If we apply this model to urban and rural regions, we could assume that – given the fact that land is more scarce in urban regions than in rural regions – induced processes of land saving innovations tend to emerge in urban regions, resulting in a higher land productivity relative to rural regions. Hypothesis 2 Farming in regions with a relatively high share of employment in agriculture tends to be less intensive than farming in regions with a relatively low share of employment in agriculture. The main idea behind this hypothesis is that economies with a relatively low share of employment in agriculture tend to be in a further stage of economic development with relatively high levels of both land and labour productivity. In a narrow sense, the concept of economic development can be seen as a rise in per capita income, accompanied by a qualitative change in the production structure (Szirmai 1994). This change usually refers to an increase in the share of the industrial and services sectors in the gross domestic product (GDP) and a decrease in the share of the agricultural sector. In a broader concept of economic development, social indicators, like life expectancy, literacy, education level, income distribution, infant mortality, daily calorie intake, the number of hospital beds, doctors and telephones, rationality, planning, an efficient institutional structure, democracy etc. are added to the narrow concept (Szirmai 1994). These broad societal changes combined with the change in sectoral employment, in which abundant agricultural labour is pulled by the industrial and services sector, boost productivity (Maddison 1982). Hypothesis 3 Farming in LFA tends to be less intensive than farming in non-LFA. Natural handicaps in LFA, like altitude, slopes and unfavourable soil conditions, might hamper agricultural production. Labour productivity lags behind as these areas are more difficult to cultivate than non-LFA and land productivity tends to be relatively low due to unfavourable physical conditions.

8 The Use of Regional Typologies in the Assessment of Farms’ Performance

193

Approach for Testing Hypotheses For testing the hypotheses, we have to decide how to measure farming intensity and to design a typology of urban and rural regions, a typology of regions based on the share of agricultural employment and a typology of LFA and non-LFA regions. These issues are discussed below.

Measurement of Farming Intensity In this analysis, farming intensity is interpreted as total agricultural output per hectare in euro. By doing so, we follow the SEAMLESS farm typology. Total agricultural output (FADN code: SE131) is defined as total output of crops and crop products, livestock and livestock products and of other output.1 It includes also sales and use of (crop and livestock) products and livestock, change in stocks of (crop and livestock) products, and change in valuation of livestock (EC 2007). In order to determine faming intensity of a region, we divide the total agricultural output of all farms in a region by the total number of hectares of utilized agricultural area in the region. Data are derived from the Farm Accountancy Data Network (FADN). Farming intensity is measured in 2 years: 1990 and 2003. We use data in current euros, which implies that absolute values cannot be compared between both years. However, the data enable us to explore whether relative differences in farming intensity between urban and rural regions tend to increase or decrease in the course of the years. We restrict our analysis to the EU15 Member States, as these are the only countries for which there is data available in FADN for 1990 and 2003. For 1990 there is, however, no information in FADN for Eastern Germany, Austria, Finland and Sweden. A main question in the analysis of socio-economic indicators across EU regions refers to the benchmark: have values to be compared with the EU average or with the national average? The answer depends on whether the indicator is independent or dependent on national specific factors. In the case of farming intensity, large differences in output per hectare exists among Member States, reflecting national differences in the composition of agricultural production and in the use of biological and chemical technology. Therefore, we will compare the value of the regional farming intensity with the national average.

Typology of Rural and Urban Regions The OECD (Organisation for Economic Co-operation and Development) rural typology of predominantly rural, intermediate rural and predominantly urban 1 Leased land ready for sowing, receipts from occasional letting of fodder areas, agistment, forestry products, contract work for others, hiring out of equipment, etc. (FADN code SE256).

194

I.J. Terluin et al.

regions is often used in comparisons of socio-economic indicators in functional regions. This typology is derived from population density. All three types of regions are territorial entities with villages, cities and agricultural area. The OECD rural typology could be used as a typology of rural and urban regions in this study as well. However, due to the use of FADN data in our analysis, we have to make some adjustments for the regional unit. Whereas the OECD rural typology was designed for a mix of NUTS2 and NUTS3 2 regions (OECD 1996), which covers 599 regions in the EU15, FADN distinguishes only 105 EU15 regions. For some smaller EU Member States (Belgium, Denmark, Ireland, Luxembourg, Netherlands, Austria and Finland), a FADN region coincides with the whole country.3 In order to distinguish FADN regions according to their rurality, we designed a typology of urban and rural FADN regions by using the same differentiating characteristics and method of classifying types as in the OECD rural typology (OECD 1996). The differentiating characteristic refers to population density in local communities. The method of classifying types consists of three steps. First, when population density in local communities is less than 150 inhabitants per square kilometre, the community is classified as ‘rural’; when population exceeds 150 inhabitants per square kilometre as ‘urban’. Second, a deductive method of classification has been applied by distinguishing three types of regions (Fig. 8.1): 1. Most rural regions 2. Intermediate rural regions 3. Most urban regions These three types were created by using the following threshold values: –– When more than 50% of the population of the region lives in rural local communities, the region is classified as ‘most rural’. –– When between 15% and 50% of the population of the region lives in rural local communities, the region is classified as ‘intermediate rural’. –– And when less than 15% of the population of the region lives in rural local communities, the region is classified as ‘most urban’. Finally, when most rural regions include a city of 200,000 inhabitants or more, the region is classified as intermediate rural; when intermediate rural regions include a city of 500,000 inhabitants or more, the region is classified as most urban. NUTS = Nomenclature des Unités Territoriales Statistiques. For linking FADN data with data from other harmonized European regional data sources, such as Eurostat REGIO and the Farm Structure Survey (FSS), we have to take account of the use of different borders of regions. In order to harmonize the regional levels used in FADN, REGIO and FSS, we have designed a so-called HARM1 regional division. For most EU15 Member States, FADN regions coincide with HARM1 regions. However, in some countries the number of FADN regions exceeds that of HARM1 regions: Germany (16 FADN regions and 14 HARM1 regions), Italy (21 FADN regions and 20 HARM1 regions) and Portugal (five FADN regions and four HARM1 regions). In our calculations, we use the HARM1 regional division. As the differences with FADN regions are relatively small, and as the FADN regions are more known than the HARM1 regions, we use for convenience sake the term FADN regions in the rest of the text.

2 3

8 The Use of Regional Typologies in the Assessment of Farms’ Performance

195

Fig. 8.1 Most rural, intermediate rural and most urban FADN regions in the EU15, 2003 (Own calculations based on Eurostat, Luxembourg)

It has to be noted that upscaling of the OECD rural typology to the FADN level has some major drawbacks, as such an upscaling tends to imply a shift from rural regions to urban regions, simply due to the fact of the weighting of big cities and their population density over larger areas. For example, suppose a FADN region with a city of 600,000 inhabitants. This FADN region is composed of three NUTS3 regions, two of them predominantly rural and one predominantly urban according to the OECD rural typology. However, when the FADN region is the unit for the classification, then this FADN region would be classified as predominantly urban. From this, it could be concluded that although upscaling is possible, it results in a relative shift from rural to urban. So it might be clear that a rural typology derived from FADN regions is not the most optimal approach of urban and rural regions.

Typology of Regions Based on the Share of Agricultural Employment In the process of economic development, the structure of the economy changes: employment in agriculture shrinks and employment in the industry and services sector increases. As such, regions with a relatively high share of employment in agriculture

196

I.J. Terluin et al.

still have to go through a process of economic transition. In order to explore trends and conditions in regions with a rather similar share of employment in agriculture, a typology of regions derived from the share of agriculture in total employment has been designed. On the whole, in previous studies on regional employment, population and economic dynamics, regions which reflect the size of a labour market area appeared to be useful territorial units for such a typology (OECD 1996; Terluin and Post 2000; Bollman et al. 2005). Labour market areas are areas where most of the people living there, also work there. However, labour market areas do not always coincide with administrative regions for which data are collected. In order to approximate regions which reflect the size of a labour market area, in previous studies of regional economic dynamics, often a mix of NUTS2 and NUTS3 regions was applied. The reason for using such a mix is that the size of the administrative NUTS2 and NUTS3 regions rather varies among countries. Here again, as in the case of the typology of rural and urban regions, the use of FADN data forces us to design a typology at a higher aggregated regional level. We applied a deductive method by setting – based on expert knowledge – thresholds for five types: FADN regions with a share of agriculture in total employment of <3%; 3–8%; 8–15%; 15–25% and >25%, respectively (Fig. 8.2).

Fig. 8.2 Share of agriculture in total employment in FADN regions in the EU15, 2003 (%) (Own calculations based on Eurostat, Luxembourg)

8 The Use of Regional Typologies in the Assessment of Farms’ Performance

197

Typology of LFA and Non-LFA Regions LFA can be described as areas where agriculture is hampered by slopes or altitude (mountainous areas), by unfavourable production circumstances (other LFA) or by environmental restrictions (LFA with specific handicaps). In the scope of the second pillar of the CAP, farmers in LFA can be compensated for these difficult production circumstances by means of Compensatory Allowances per hectare. The aim of the LFA policy is to ensure continued agricultural land use and, thereby, contributing to the maintenance of a viable rural community, maintenance of the countryside and maintenance and promotion of (environmentally) sustainable farming systems (CEU 1999; CEU 2005). The size of the area designated as LFA varies among Member States. For the design of a LFA typology, we used the share of LFA farms in the total number of farms in a FADN region as differentiating characteristic (Terluin et al. 1994). We distinguish three types (Fig. 8.3): –– Non-LFA regions (less than 30% of all farms are located in LFA); –– Partly LFA regions (30–60% of all farms are located in LFA); –– LFA regions (over 60% of all farms are located in LFA).

Fig. 8.3 Non-LFA, partly-LFA and LFA regions in the EU15, 1990* (Own calculations based on Eurostat, Luxembourg). * 1995 for Finland, Sweeden, Austria and Eastern Germany.

198

I.J. Terluin et al.

The classification is based on the situation of 1990; however, for Finland, Austria, Eastern Germany and Sweden, the situation of 1995 is used.4

Results In this section we discuss the results of our testing of the three hypotheses. Hypothesis 1 Farming in rural regions tends to be less intensive than farming in urban regions. Eight EU15 Member States appear to have regions in different types derived from rurality. The fact that we cannot distinguish groups of most rural, intermediate rural and most urban regions in all EU15 Member States is due to the relatively high aggregation level of the FADN regions: there is only one FADN region in Austria, Belgium, Denmark, Ireland, Luxembourg and the Netherlands, whereas Finland has only most rural regions. As we decided to compare farming intensity to the national average, this implies that we can test the hypothesis in eight EU15 Member States. From these countries, the hypothesis is supported in Germany, Spain Italy, Portugal and the UK and rejected in Greece, France and Sweden, both in 1990 and 2003 (Table 8.1). It appears that differences in farming intensity among rural and urban regions are quite substantial in countries for which the hypothesis is valid: usually farming intensity in most rural regions is about half the national average, whereas that in most urban regions is often substantially above the national average. On the other hand, in the three countries in which the hypothesis is rejected, differences in farming intensity between rural and urban regions are relatively moderate. Obviously, in these countries the degree of rurality does not affect farming intensity to a large extent. It could be wondered whether farming intensity at individual farm types differs from that of the average of the regional farm, that is analysed in Table 8.1. It appears that individual farm types follow the pattern found for the regional average farm. However, horticultural farms seem to be an exception to this pattern: in most countries differences in farming intensity are relatively large among rural, intermediate and urban regions, likely related to the predominantly occurrence of open field horticulture in rural regions and that of greenhouse horticulture in urban regions. Among countries, rather large differences in farming intensity exist. Farming intensity is relatively low in Spain, Ireland, Portugal, Finland, Sweden and the UK (about €800–1,300/ha in 2003) and rather high in Belgium (about €4,000/ha in 2003) and The Netherlands (over €9,700/ha in 2003) (Table 8.1). These differences reflect differences in the composition of agricultural production and in the use of in biological and chemical technology. Finland, Sweden and Austria entered the EU in 1995; for Eastern-Germany no data for 1990 were available. As the share of LFA farms in the total number of farms is rather stable over time, the typology is rather independent from the year on which it is based.

4

8 The Use of Regional Typologies in the Assessment of Farms’ Performance

199

Table 8.1 Farming intensity in FADN regions in the EU15, 1990 and 2003 (output per hectare in current euro) (Own calculations based on FADN) 1990 2003 National National As % of national average average As % of national average average Most Intermediate Most Most Intermediate Most rural rural urban (€/ha) rural rural urban (€/ha) Belgium – – 100 4,130 – – 100 3,985 Denmark – 100 – 2,601 – 100 – 2,750 Germany – 97 114 2,558 51 100 112 2,186 Greece 110 95 2,106 115 93 2,533 Spain 61 98 233 888 53 109 183 1,246 France 95 119 106 1,814 94 127 106 1,649 Ireland – 100 – 761 – 100 – 811 Italy 39 95 161 2,429 40 91 166 3,247 Luxembourg – 100 – 1,877 – 100 – 1,643 The Netherlands – – 100 8,120 – – 100 9,719 Austria n.a. n.a. n.a. n.a. – 100 – 2,012 Portugal 37 147 – 821 47 143 – 925 Finland n.a. n.a. n.a. n.a. 100 – – 1,256 Sweden n.a. n.a. n.a. n.a. 114 96 – 1,300 UK – 69 149 1,050 – 75 132 1,193 EU12/15 79 87 170 1,841 69 92 176 1,980 ‘–’ Denotes that the type does not exist

The use of current prices does not allow for comparisons of the absolute level of farming intensity between 1990 and 2003. As a result of inflation, one might expect that output in current euro per hectare in 2003 exceeds that in 1990. Nevertheless, in a number of countries, like Belgium and Germany, output in current euro per hectare in 2003 is below that in 1990. Likely, the decrease in real prices of agricultural products is not offset by higher physical outputs per hectare in these countries. Hypothesis 2 Farming in regions with a relatively high share of employment in agriculture tends to be less intensive than farming in regions with a relatively low share of employment in agriculture. Nine EU15 Member States appear to have regions in different types derived from the share of agriculture in total employment. For the seven countries tested in 1990 (Finland and Sweden were excluded as they were not a Member of the EU), the hypothesis is supported in Germany, Greece, Italy, Portugal and the UK, and rejected in Spain and France (Table 8.2). In these latter countries, regions with the highest share of agriculture in total employment have the highest farming intensity. In Spain, there is only one region in the group with the highest share of agriculture in employment: Galicia. Agriculture in this region is composed of dairy cattle farming and beef and mixed cattle, which are both characterized by a relatively high farming intensity. In France, the group of regions with the highest

200

I.J. Terluin et al.

Table 8.2 Farming intensity according to the share of agriculture in total employment in FADN regions in the EU12, 1990 (output per hectare in current euro) (Own calculations based on FADN) Farming intensity as% of national intensity National <3% 3–8% 8–15% 15–25% >25% share agriculture National in total farming Type with share employment intensity agriculture in total

employment → Belgium Denmark Germany Greece Spain France Ireland Italy Luxembourg The Netherlands Portugal UK EU12

– – 119 – – 101 – 175 – – – 144 106

100 100 97 – 131 91 118 100 100 184 48 122

– – – 106 112 106 100 90 – – – – 88

– – – – 77 – – 64 – – 89 – 47

– – – 98 295 – – – – – – – 116

(%) 3.2 5.5 4.4 22.4 10.8 5.3 14.5 7.3 3.2 4.8 15.6 2.1 6.0

(€/ha) 4,130 2,601 2,558 2,106 888 1,814 761 2,429 1,877 8,120 821 1,050 1,841

‘–’ Denotes that the type does not exist

share of agriculture in employment consists of about 40% of the French regions. These regions are characterized by a relatively high share of permanent crop farms with a rather high farming intensity. In contrast to the urban-rural typology and the LFA typology, where regions belong to the same type in both 1990 and 2003, in the typology based on the share of employment in agriculture regions could move from one type to another type. This is due to the properties of the differentiating characteristics on which the typologies are based: population density and the share of LFA farms tend to be rather stable over time, whereas the share of agriculture in total employment tends to decline. So in 2003, a shift can be perceived towards more regions in types with a lower share of agriculture in employment relative to 1990. On the whole, this reflects economic development. From the nine countries tested in 2003, it appears that the hypothesis is supported in Germany, Spain, France, Greece, Italy, Portugal and the UK, and rejected in Finland and Sweden (Table 8.3). In Greece and Italy, however, there is an interruption of the general trend due to a type with only one region: Makedonia-Thraki in the type 15–25% in Greece and Calabria in the type 15–25% in Italy. Makedonia-Thraki is characterized by a relatively low share of permanent crops, which explains its low intensity, whereas Calabria has a relatively high share of permanent crops, which explains its high intensity. In contrast to 1990, the hypothesis is supported in Spain and France in 2003. In Spain this is due to the shift of the ‘outlier’ Galicia to a group with many other regions, which average out the specific properties of this region; in France this is due to a shift of a quite large group of regions towards the type <3%, in which the share of permanent and horticultural farms increases from 11% in 1990 to 30% in 2003. The rejection of the hypothesis

8 The Use of Regional Typologies in the Assessment of Farms’ Performance

201

Table 8.3 Farming intensity according to the share of agriculture in total employment in FADN regions in the EU15, 2003 (output per hectare in current euro) (Own calculations based on FADN) Farming intensity as% of national intensity National share National <3% 3–8% 8–15% 15–25% >25% agriculture Type with share farming agriculture in total in total intensity employment → employment (%) (€/ha) Belgium 100 – – – – 2.4 3,985 Denmark – 100 – – – 3.6 2,750 Germany 115 96 – – – 2.5 2,186 Greece – 127 73 115 16.2 2,533 Spain 130 130 91 – – 6.0 1,246 France 115 96 – – – 3.6 1,649 Ireland – 100 – – – 7.0 811 Italy 156 93 60 73 – 4.4 3,247 Luxembourg 100 – – – – 1.4 1,643 The Netherlands 100 – – – – 3.0 9,719 Austria – 100 – – – 4.9 2,012 Portugal – 306 125 56 – 9.6 925 Finland – 94 113 – – 5.3 1,256 Sweden 99 129 – – – 2.4 1,300 UK 104 75 – – – 1.6 1,193 EU15 129 96 62 69 147 4.9 1,980 ‘–’ Denotes that the type does not exist

in Finland and Sweden in 2003 could be related to the much larger share of dairy cattle farms in the regions with a higher share of employment in agriculture. Hypothesis 3 Farming in LFA tends to be less intensive than farming in non-LFA Nine EU15 Member States appear to have regions in different types of LFA regions. In 1990 the hypothesis is supported in all countries with different types of LFA (Finland and Sweden are excluded from the analysis for this year). In 2003 the hypothesis is supported in six countries and rejected in Germany, Finland and Sweden (Table 8.4). The rejection in Finland and Sweden is again due to the relatively high share of dairy cattle in LFA regions compared to that in partly-LFA regions, and this could also be a likely reason for the rejection in Germany in 2003. On the whole, the share of grazing livestock farms in LFA regions tends to exceed that in non-LFA regions, whereas the share of permanent crop farms in LFA regions is usually below that in non-LFA regions. However, in Germany and Italy the composition of farm types in LFA regions and non-LFA regions is rather similar. Especially in Spain, Italy, Portugal and the UK differences in farming intensity in LFA, partly-LFA and non-LFA regions are considerable. In Spain and Portugal, this gap between LFA, partly-LFA and non-LFA regions widened in the period 1990–2003, whereas it decreased in Italy and remained rather stable in the UK. Related to the higher share of grazing livestock in LFA regions, usually the share of grassland in total UAA in LFA regions tends to exceed that in non-LFA regions (Table 8.5). As grassland gives home to many birds and other species, this could

202

I.J. Terluin et al.

Table 8.4 Farming intensity in LFA, partly-LFA and non-LFA regions in the EU15, 2003 (output per hectare in current euro) (Own calculations based on FADN) 1990 2003 National As % of national average average As % of national average LFA Partly-LFA Non-LFA (€/ha) LFA Partly-LFA Non-LFA Belgium – – 100 4,130 – – 100 Denmark – – 100 2,601 – – 100 Germany 97 98 130 2,558 106 91 94 Greece 95 110 – 2,106 93 115 – Spain 90 179 187 888 89 177 282 France 68 93 119 1,814 67 99 119 Ireland 100 – – 761 100 – – Italy 66 108 172 2,429 70 109 140 Luxembourg 100 – – 1,877 100 – – The Netherlands – – 100 8,120 – 100 Austria n.a. n.a. n.a. n.a. 100 – – Portugal 89 184 821 89 – 306 Finland n.a. n.a. n.a. n.a. 119 86 – Sweden n.a. n.a. n.a. n.a. 114 96 – UK 48 – 144 1,050 48 – 136 EU12/15 68 125 140 1,841 72 121 134

1990 and

National average (€/ha) 3,985 2,750 2,186 2,533 1,246 1,649 811 3,247 1,643 9,719 2,012 925 1,256 1,300 1,193 1,980

‘–’ Denotes that the type does not exist Table 8.5 Grassland in LFA, partly-LFA and non-LFA regions in the EU15, 1990 and 2003 (% of utilized agricultural area) (Own calculations based on FADN) 1990 2003 As % of national National As % of national National average average average average Partly- NonPartly- NonLFA LFA LFA (% of UAA) LFA LFA LFA (% of UAA) Belgium – – 100 44 – – 100 43 Denmark – – 100 6 – – 100 5 Germany 100 106 82 35 111 91 63 25 Greece 29 232 – 1 64 181 – 2 Spain 112 7 0 12 106 59 0 20 France 163 36 83 24 177 41 73 22 Ireland 100 – 84 100 – – 84 Italy 185 59 97 20 255 37 23 8 Luxembourg 100 – – 56 100 – – 48 The Netherlands – – 100 50 – 100 48 Austria n.a. n.a. n.a. n.a. 100 – – 35 Portugal 101 89 11 104 – 31 21 Finland n.a. n.a. n.a. n.a. 0 0 – 0 Sweden n.a. n.a. n.a. n.a. 183 77 – 14 UK 141 – 65 59 156 – 61 51 EU12/15 126 45 85 32 135 38 82 27 ‘–’ Denotes that the type does not exist

8 The Use of Regional Typologies in the Assessment of Farms’ Performance

203

justify CAP support as a compensation for the provision of public goods and land management by farmers in LFA regions.

Concluding Remarks In this chapter we explored whether the economic and environmental performance of farms is affected by regional characteristics. We used farm intensity – measured as agricultural output per hectare – as an indicator for the economic and environmental performance of farms. Regional characteristics were captured by three regional typologies. In particular, we considered three hypotheses: 1. Farming in rural regions tends to be less intensive than farming in urban regions. 2. Farming in regions with a relatively high share of employment in agriculture tends to be less intensive than farming in regions with a relatively low share of employment in agriculture. 3. Farming in LFA tends to be less intensive than farming in non-LFA.

Drawbacks of Typologies at a Relatively High Regional Aggregation Level In order to test these hypotheses, we used data on farm intensity from the FADN. A drawback of these data is that they are only available at a relatively high regional aggregation level. As a consequence, our typologies of rural and urban regions, regions with a high and low share of employment in agriculture, and LFA and non-LFA regions were designed also at a relatively high regional aggregation level. It has to be recognized that by doing so less optimal typologies emerge, as socio-economic and natural differences within regions are averaged out. A second drawback of this relatively high regional aggregation level is that we were not able to test the hypotheses in Belgium, Denmark, Ireland, Luxembourg, The Netherlands and Austria, as the FADN regions coincide with the national level in these Member States.

Wide Support for Hypothesis (2) and (3) An overview of the results of the tested hypotheses in the nine EU15 Member States is presented in Table 8.6. From our analysis, it appears that hypotheses

204

I.J. Terluin et al. Table 8.6 Summary: support for the tested hypotheses Hypothesis (1) Hypothesis (2) 1990 2003 1990 2003 Germany Yes Yes Yes Yes Greece No No Yes Yes Spain Yes Yes No Yes France No No No Yes Italy Yes Yes Yes Yes Portugal Yes Yes Yes Yes Finland – – – No Sweden – No – No UK Yes Yes Yes Yes

Hypothesis (3) 1990 2003 Yes No Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes – No – No Yes Yes

(2) and (3) were widely supported, whereas hypothesis (1) was only accepted in five countries. Finland and Sweden are the two countries where all hypotheses were rejected. So it seems that there is a tendency that two of our examined regional characteristics, i.e. the share of agriculture in total employment and the presence of LFA, affect farming intensity, whereas farming intensity is less affected by an urban or rural territorial context. It is, however, not unlikely that our hypothesis on the impact of the degree of rurality on farming intensity shows a better score if a lower level of regional aggregation would have been considered. The studied regional characteristics do not affect farming intensity in Finland and Sweden. Obviously, farming intensity in the Nordic agriculture in these countries is more dependent on other factors, likely related to a short growing season due to climatological conditions.

Regional Characteristics Included in Integrated Assessments Our findings suggests that regional characteristics matter in the economic and environmental performance of farms in Western Europe. In the scope of integrated assessments of farms’ performance and agricultural policy, it could be recommended to include regional characteristics as well in such assessments. A main issue in such an inclusion refers to the regional level for measuring regional characteristics. On the whole, there is not one regional level that fits all purposes. On the contrary, per single indicator a suitable regional level should be strived at by looking for territorial units that are rather homogeneous for the specific indicator at stake. Expert knowledge could be of help in determining a suitable regional level for the measurement of regional indicators. Nevertheless, in reality often a trade-off has to be made between what is a theoretically suitable regional level and the regional level at which data is available. In this chapter, for example, we used the FADN regions as data on farming intensity were available at this level. However, we argued that for the analysis of the studied regional characteristics a lower level of regional aggregation would be more appropriate. Future research on methods for up

8 The Use of Regional Typologies in the Assessment of Farms’ Performance

205

and down scaling of regional data and increased efforts on the collection of data for small regional units could be recommended in order to escape from this trade-off between the most suitable regional level from a theoretically point of view and the regional level at which data is available.

References Andersen, E., Baldock, D., Bennett, H., Beaufoy, G., Bignal, E., Brouwer, F., Elbersen, B., Eiden, G., Godeschalk, F., Jones, G., McCracken, D., Nieuwenhuizen, W., van Eupen, M., Hennekens, S., & Zervas, G. (2003). Developing a high nature value farming area indicator. Report to the European Environment Agency, Copenhagen, Denmark. Andersen, E., Elbersen, B., & Godeschalk, F. (2004, February). Farming and the environment in the European Community – using agricultural statistics to provide farm management indicators. Paper presented at OECD Expert Meeting on Farm Management Indicators, New Zealand. Andersen, E., Elbersen, B., Godeschalk, F., & Verhoog, D. (2007). Farm management indicators and farm typologies as a basis for assessments in a changing policy environment. Journal of Environmental Management, 82, 353–362. Andersen, E., Verhoog, D., Elbersen, B., Godeschalk, F., & Koole, B. (2006, April 14). A multidimensional farming system typology. SEAMLESS PD 4.4.2. Bollman, R., Terluin, I., Godeschalk, F., & Post, J. (2005). Comparative analysis of leading and lagging rural regions in OECD countries in the 1980s and 1990s. Paper for the Congress of the European Regional Science Association ‘Land Use and Water Management in a Sustainable Network Society’, Vrije Universiteit, Amsterdam, 23–27 August 2005. CEU. (2005). Council Regulation (EC) No 1698/2005 of 20 September 2005 on support for rural development by the European Agricultural Fund for Rural Development (EAFRD). Brussels, Official Journal of the European Union, L 277, 21/10/2005. Council of the European Union (CEU). (1999). Council Regulation (EC) No 1257/1999 of 17 May 1999 on support for rural development from the EAGGF and amending and repealing certain Regulations. Brussels, Official Journal of the European Communities, L 160, 26/6/1999. EEA/UNEP. (2004). High nature value farmland. European Environmental Agency and UNEP regional office for Europe. Luxembourg: Office for Official Publications of the European Communities. European Commission (EC). (2007). Definitions and variables used in FADN standard results. Brussels: DG for Agriculture and Rural Development, Community Committee for the Farm Accountancy Data Network (FADN), RI/CC 882 Rev. 8.1. European Environmental Agency (EEA). (2005). Agriculture and environment in EU-15-the IRENA indicator report. EEA Rep. No. 6/2005. Copenhagen, Denmark: EEA. Hayami, Y., & Ruttan, V. W. (1985). Agricultural development; an international perspective. Baltimore/London: The John Hopkins University Press. Maddison, A. (1982). Phases of capitalist development. Utrecht/Antwerpen: Het Spectrum. OECD. (1996). Territorial indicators of employment; focusing on rural development. Paris: Organisation for Economic Co-operation and Development. Szirmai, A. (1994). Ontwikkelingslanden: Dynamiek en stagnatie (Developing countries: Dynamics and stagnation). Groningen: Wolters Noordhof. Terluin, I.J., Godeschalk, F.E., von Meyer, H., Strijker, D., & Post, J.H. (1994). The agricultural income situation in less favoured areas of the EC. Luxembourg/Brussels. Office for Official Publications of the European Communities. Terluin, I. J., & Post, J. H. (eds). (2000). Employment dynamics in rural Europe. Wallingford: CABI. van den Noort, P. C. (1980). Inleiding tot de algemene agrarische economie (Introduction into general agricultural economics). Leiden: Stenfert Kroeze.

Chapter 9

A Web-Based Software System for Model Integration in Impact Assessments of Agricultural and Environmental Policies Jan-Erik Wien, Andrea Emilio Rizzoli, Rob Knapen, Ioannis Athanasiadis, Sander Janssen, Lorenzo Ruinelli, Ferdinando Villa, Mats Svensson, Patrik Wallman, Benny Jonsson, and Martin van Ittersum

Introduction Policy makers are confronted with complex, socially relevant problems (Van Ittersum et al., 2008). When handling these problems, there is a general tendency to increase participation of citizens and other stakeholders. In this way policy makers

J-.E. Wien, R. Knapen (*), and S. Janssen Center for Geo-Information, Alterra Wageningen UR, P.O. Box 47, 6700 AA, Wageningen, The Netherlands e-mail: [email protected]; [email protected] A.E. Rizzoli and I. Athanasiadis IDSIA – Istituto Dalle Molle di Studi sull’Intelligenza Artificiale, Galleria 2, CH-6928, Manno, Switzerland e-mail: [email protected]; [email protected] R. Knapen Center for Geo-Information, Alterra Wageningen UR, P.O. Box 47, 6700 AA, Wageningen, The Netherlands e-mail: [email protected] L. Ruinelli AntOptima, Via Fusoni 4, 6900, Lugano, Switzerland e-mail: [email protected] F. Villa Gund Institute for Ecological Economics, University of Vermont, 617 Main Street, Burlington, Vermont VT, 05405, USA e-mail: [email protected] M. Svensson, P. Wallman, and B. Jonsson Centre for Sustainability Studies, Lund University, P.O. Box 170, SE-221 00, Lund, Sweden e-mail: [email protected]; [email protected]; [email protected] M. van Ittersum and S. Janssen Plant Production Systems Group, Wageningen University, P.O. Box 430, 6700 AK, Wageningen, The Netherlands e-mail: [email protected] F.M. Brouwer and M. van Ittersum (eds.), Environmental and Agricultural Modelling: Integrated Approaches for Policy Impact Assessment, DOI 10.1007/978-90-481-3619-3_9, © Springer Science+Business Media B.V. 2010

207

208

J.-E. Wien et al.

want to improve the quality of their policies and obtain a broader support for and understanding of the proposed policy measures. In this new governance concept, the policy making process is the product of complex interactions between governmental and non-governmental organizations, each seeking to influence the collectively binding decisions that have consequences for their interest. Policy making is more and more a process of cooperation and participation in which the policy maker becomes a facilitator of the process. To account for this new governance, policy measures need to be assessed in an integrated context. Rotmans and van Asselt (1996) defined Integrated Assessment (IA) as “an interdisciplinary and participatory process combining, interpreting and communicating knowledge from diverse scientific disciplines to allow a better understanding of complex phenomena”. In this interdisciplinary and participatory process, information needs to be accessible in the way that all different types of stakeholders achieve a mutual understanding of the problems, objectives and solutions. But this mutual understanding across disciplines is often hindered by jargon, language, past experiences and presumptions of what constitutes persuasive argument, and different outlooks across disciplines or experts of what makes knowledge or information salient for policy makers or policy assessments (Cash et al. 2003). This mutual understanding is essential in integrated assessment studies in which modelling frameworks are built from multiple models, data sources and indicators, which have to be coupled meaningfully. The software architecture of modelling frameworks should be such that it reflects and enables the mutual understanding of the group of researchers. This chapter addresses problems and offers some solutions concerning knowledge integration in integrated assessment studies, first by analysing the theoretical perspective and then discussing the key role played by ontologies to support model integration in the context of the SEAMLESS project. Finally, a web-based software architecture for implementing modelling frameworks in integrated assessment studies is presented and we demonstrate its use to improve mutual understanding within the researchers’ team. Along the way, the paper details the challenges and the processes used to manage and tame the complexity of a large European integrated project (e.g. SEAMLESS) and the use of an ontology to support (linking of) projects, models, indicators and raw data. This chapter concludes this paper by describing the major functionalities offered by SEAMLESS-IF and its expected future developments.

Challenges in Integrated Modelling Interoperability from a Theoretical Perspective Integrated modelling requires interoperability, which is the ability of two or more systems or components to exchange information and to use the information that has been exchanged (IEE 1990). Sølvberg (1998) distinguished between syntactic, structural and semantic interoperability. Syntactic interoperability is defined

9 A Web-Based Software System for Model Integration in Impact Assessments

209

as the ability of two or more systems/components to exchange and share information by marking up data in a similar fashion (e.g. using the extensible markup language, XML). Structural interoperability means that the systems/components share semantic schemas (data models) that enable them to exchange and structure information (e.g. using the resource description framework, RDF). Semantic interoperability is the ability of systems/components to share and understand information at the level of formally defined and mutually accepted domain concepts. Semantic interoperability requires the correct interpretation and mutual understanding of all exchanged information. In order to obtain mutual understanding of exchanged information, the actors have to share a model of what the data stand for. Semantic interoperability is about how to achieve such mutual understanding (Sølvberg 1998). One of the earliest theories dealing with understanding and remedies for misunderstanding is Richards’s “Meaning of Meaning” theory (Ogden and Richards 1923). Instead of focusing on the information that is communicated, Richards wanted to study the meaning of the words. He felt that understanding is the main goal of communication and communication problems result from misunderstanding. One of the ideas behind the Meaning of Meaning Theory is “The Proper Meaning Superstition” (Ogden and Richards 1923). This is the belief that every word has an exact meaning. Richards says that the Proper Meaning Superstition is false because words mean different things to different people in different situations. This misunderstanding causes problems when two people believe they are talking about the same thing, but actually talk about different things. Another concept that Richards uses is the idea of signs and symbols in communication. Words are examples of such symbols. Symbols have no natural connection with the things they describe (Griffen 1997). Words are symbols of something because they have been given meaning. But very often words mean one thing in a certain context and mean another thing in a different context. This is why it is so important to study the context to get a better understanding of the meaning. To come to this better understanding, Richards invented the Semantic Triangle (Fig. 9.1). This triangle shows the relationship between symbols and their referent. Thought or reference

s) on to lati s er re ef l R usa ca

r he

Symbol

t (o

s n) ize atio l bo el m al r y S us ca (a

Stands for (an imputed relation)

Fig. 9.1 Semantic triangle by Richards (Ogden and Richards 1923)

Referent

210

J.-E. Wien et al. Cereal crop, grain

(o th o tio s t la er re ef l R usa ca

er

ns

) es on liz lati o b e m lr Sy usa ca (a

)

Wheat

Stands for (an imputed relation)

Fig. 9.2 Semantic triangle for wheat

One part of the triangle is the symbol, or the word. Another part of the triangle is the thought or reference. These are the words that one would use to describe the referent. The referent, the last part, is the thing that one would picture in his mind. An example in integrated agricultural modelling (Fig. 9.2) would be “wheat”. The words to describe the referent can be for example “cereal crop, grain, flour”. The referent, the last part, is the thing that one would picture in his mind. Understanding that people mean different things when they say the same word is an important finding that is very relevant for the process of integrated modelling, more so when it comes to associate the proper meaning with variables to be exchanged between two models (Wien et al. 2007).

Interoperability in Integrated Modelling If the integrated model, composed by a number of different models pertaining to domains as diverse as economy, agronomy, social sciences, ecology etc., is developed by a single scientist, or even by a closely connected group of scientists, the potential ambiguities and misunderstandings can be solved by internal discussions and interactions. The problem is that, given the scale and the ambition of such integrated models, it is unrealistic to even think that a single group can encompass all the required knowledge to develop such a model. Therefore, different models are developed by different groups, and as a consequence the need arises for structured and organized ways to solve ambiguities and misunderstandings regarding the meaning of the variables to be shared and exchanged by the models. In such a complex integration task, different types of misunderstandings around the meaning of concepts can occur:

9 A Web-Based Software System for Model Integration in Impact Assessments

211

1. As the same concept might be used for different meanings, assuming unspecified information, for example “crop yield” in a model as a single value and “crop yield” in a database as a series of values over space and time; 2. As different concepts might be used, which have the same meaning, for example “derivative” and “rate”; 3. As concepts might be used with an ambiguous meaning, for example the word “scenario” (Schoemaker 1993); 4. As relationships between concepts might be understood in a different way, for example the spatial relationship of inclusion of farm within an “agri-environmental zone”, and the relationship between the same farm and an administrative area. The challenge in integrated modelling is the conceptual integration. To achieve this, we need explicit semantics and a shared conceptualization. For this we need to tackle the different perceptions and interpretations of people involved. Different modelling approaches, different formalisms and last but certainly not least, the different integration requirements and ambitions need to be taken into account. To enable semantic interoperability in integrated modelling, the problem of semantic conflicts or semantic heterogeneity needs to be solved.

The Need for a Common Ontology One of the many definitions of ontology is from Neches et al. (1991) “an ontology defines the basic terms and relations comprising the vocabulary of a topic area as well as the rules for combining terms and relations to define extensions to the vocabulary”. According to Gruber (1993), an ontology is “an explicit and formal specification of a conceptualization”. A conceptualization is explained as an abstract model of a phenomenon, identifying the relevant attributes. A conceptualization is an abstract, simplified view of the world that we wish to represent. Every knowledge base or knowledge-based system is committed to some conceptualization (Gruber 1995). Formal specification refers to the fact that the language semantics are machine readable, and follow a mathematical framework for logical reasoning. Often this is done by use of W3C OWL (Ontology Web Language, Patel-Schneider et al. 2004). A formal specification helps to communicate the definition of terms in a context independent way and formal language semantics allows some automated consistency checks. An ontology helps to formalize the knowledge captured in and/or between models, in order to subsequently facilitate model development, testing and documentation (Scholten et al. 2007) and increase model re-usability and exchangeability (Rizzoli et al. 2008; Villa et al. 2009). Besides this it separates knowledge captured in the model from the actual implementation in a modelling language or in software e.g. Java, FORTRAN, Matlab, STATA, etc. (Gruber 1993; Villa et al. 2006) or from the data in a database (Zander and Kächele 1999).

212

J.-E. Wien et al.

The development of a common ontology by a group of researchers is a complex, challenging and time-consuming task (Farquhar et al. 1995; Gruber 1993), and it still remains a scientific challenge. To achieve ontological commitment, i.e. the agreement by multiple parties to adhere to a common ontology, when these parties do not have the same experiences and theories (Holsapple and Joshi 2002), a collaborative approach is needed (Pennington et al. 2007). Other approaches for ontology development are the inspirational approach, the inductive approach, the deductive approach and the synthetic approach (Holsapple and Joshi 2002). A collaborative approach has the advantages that researchers from different disciplines are diverse in their contributions, which avoids blind spots and which has more chances of getting a wide acceptance (Holsapple and Joshi 2002).

The Role of Ontologies in SEAMLESS The Structure of SEAMLESS-IF Framework The SEAMLESS consortium develops a computerized and integrated framework (SEAMLESS-IF) to assess the impacts on environmental and economic sustainability of a wide range of policies and technological improvements across a number of scales (Van Ittersum et al. 2008). In the SEAMLESS-IF approach different type of models and indicators are linked into a type of scientific workflows (van der Aalst and van der Hee 2002) named model chains, where each model uses the outputs of another model as its inputs and ultimately indicators are calculated. With respect to the models (Fig. 9.3), macro-level economic partial equilibrium models (GTAP and CAPRI; Britz et al. 2007) are linked to micro-level farm optimization models (FSSIM-MP and FSSIM-AM; see chapter 5 of this volume) and field crop growth models (APES; cf. Van Ittersum and Donatelli 2003 - see Chapter 4 of this Volume), using micro-macro up scaling methods (EXPAMOD; Bezlepkina et al. 2007). The macro-level economic partial equilibrium models simulate markets for agricultural commodities and trade between the European Union and other world trading blocks and the micro-level farm optimization problems allocate the farm area to different agricultural activities based on farmer objectives and available farm resources. Examples of agricultural activities are growing a wheat–sugar beet rotation with intensive management or keeping dairy cattle for the production of milk. The agricultural activities for cropping systems can be evaluated by the field crop growth models on their yield and environmental effects (e.g. nitrate leaching, pesticide leaching, soil erosion). Finally the micro-macro up-scaling methods extrapolate on the basis of the micro-level behaviour of the optimization model FSSIM-MP priceelasticities that are an input to the macro level partial equilibrium model CAPRI. These models provide, through their outputs, the basis for the calculation of indicators of interest to the user. Each of these models are derived from different disciplines, operate on different time and spatial scales, are programmed in different programming languages and have a different implementation of scenarios.

9 A Web-Based Software System for Model Integration in Impact Assessments

213

Fig. 9.3 Overview of SEAMLESS-IF models. The arrows indicate the dependencies among models in the chain. The components in lighter colour are not presently integrated in SEAMLESS-IF

Knowledge Management Technologies in SEAMLESS Knowledge management technologies are employed, such as semantic modelling and ontologies for specifying data, models, projects and their relationships. All the commonly shared data types in SEAMLESS were declared in ontology (starting from projects, up to the model exchange items). This is a very important shift in everyday practice that SEAMLESS has achieved: modellers specify the data requirements of their models in a higher level, i.e. that of an ontology. This ontology is then automatically transformed into a relational database model, to which “data collecting” activities need to comply with. Data may originate from third-party, pre-existing databases or may have been directly collected (i.e. through surveys). Either of the two is the case, data collection activities need to facilitate the filling-up of the generated database, which is fit to the model data requirements.

214

J.-E. Wien et al.

The process used in the SEAMLESS project for creating a common ontology for models, indicators and raw data was based on a participatory and collaborative approach. A dedicated group of scientists was formed with participants from all parts of the project. This cross-disciplinary group, named DOT.force (after Data and Ontology Task force), aimed to develop a common ontology that represents a shared conceptualization between the different databases, models and indicators, involved in the SEAMLESS-IF process. The DOT.force listed among its members, knowledge engineers and domain experts. Knowledge engineers have the technical skills and relevant experience in ontology design and conceptual modelling. The domain experts hold knowledge about a specific domain, like a certain model or dataset, a set of indicators, a particular case study or scenario, or even the SEAMLESS-IF process. Domain experts interacted with knowledge engineers for specifying the conceptualizations involved in their niches, and ultimately facilitated the development of specific parts of the ontology (Janssen et al. 2009). The knowledge engineers in the DOT.force worked on a number of actions that lead to a complete and consistent ontology. These actions included: 1. Integrating the different database schemas into a single SEAMLESS database schema; 2. Clarifying the model interface data structures (domain models), while adding relevant metadata; 3. Associating indicators with model result sets or other static data collections; 4. Supplementing the ontology with meta-data on the concepts it holds, like textual descriptions, data sources, other documentation references, units and value ranges; 5. Developing an upper ontology to cover concepts relevant with the SEAMLESS-IF process and execution. Different methods were used to construct the ontology for the different actions. For actions one and two on databases and models, dedicated meetings were organized to develop the ontology, while for action three on indicators a proposal was made by the knowledge engineers, which was then evaluated by relevant domain experts. Action four on metadata was carried out independently by domain experts, once agreement on the common ontology was reached between domain members and knowledge engineers. For action five on project and scenario definition an iterative process was used to develop a document, which was later synchronized with the project ontology after each iteration. The database design is directly mandated by the object structure of the entities and properties described in the ontology and the data can be stored in the database based on an adjunct persistence XML (eXtensible Markup Language) document provided along with each class implementing the ontology structure. Other working groups within the project have been populating the database following the schema generated by the ontology that the DOT.force concluded to. To explain more details about the knowledge management technology, we need first to give an overview of the SEAMLESS-IF architecture.

9 A Web-Based Software System for Model Integration in Impact Assessments

215

An Architecture for Knowledge Integration in Integrated Assessment Studies The SEAMLESS-IF Architecture A keystone of the SEAMLESS project is the software architecture enabling and supporting the development of integrated assessment studies. The global architecture of SEAMLESS-IF is typical for a layered (multi-tier) web based application. The diagram in Fig. 9.4 illustrates this architecture. The principle of dividing a software system into layers reaches back to the early days of computer science (Dahl et al. 1968). Originally the idea was used for the design of operating systems but it applies equally well to other types of software. In the layered software architecture the system’s functionality is usually separated in a persistence layer for storage of domain object state, a domain layer for the domain model and domain logic, a services layer that controls transactions and contains the business logic, an application layer for use-case workflows, syntactic validation and interaction with the services layer and finally a presentation layer for the user interfaces. As a general rule each layer has only dependencies on those below it, not above, limiting the effect of changes and increasing maintainability. In a way each layer acts as a client to the tier below and as a server to the tier above. All layers can be located on a single computer, or they can be divided amongst a number of systems. The latter is the case for SEAMLESS-IF.

Fig. 9.4 The SEAMLESS-IF architecture on the right is composed by three layers (SeamFrame, the core framework; SeamServer, the server side services; SeamGUI the client-based graphical user interface) and by external modules such as OpenMI compliant models, a set of ontologies managed by the Knowledge Manager and the SEAMLESS database

216

J.-E. Wien et al.

The end user interacts with the system by means of SeamGUI, a browser based application providing a Graphical User Interface (GUI) running as client of the web (server) application. The client-server architecture of SEAMLESS-IF allows for future clients to be developed and linked to the existing server, in order to cater for specific needs of different user groups. The currently available SeamGUI client includes two task specific user interfaces: –– PE GUI, the policy expert GUI: through this interface an expert can evaluate the impact of alternative agricultural policies from the different aspects of sustainability; –– IM GUI, the integrative modeller GUI: the module that guides the end-user to manage projects and request the execution of model chains, in order to produce results to be later used by policy experts. SeamGUI requires a user to log in to the system first, and then decides based on the user permissions which task user interface to show. It then interacts with the software services provided by the SeamServer, the web-based application framework built on the core classes provided by SeamFrame. The latter is a modelling framework purposely designed to develop integrated assessment tools, thanks to: – – The Modelling Environment, which is a programming framework that offers a series of facilities to encapsulate and wrap existing models for execution by the processing environment. It allows to deliver model components wrapped by a SeamFrame specific interface, compliant with the Open Modelling Interface (OpenMI) standard (www.openmi.org) (Gijsbers and Gregersen 2005, Gijsbers et al. 2006), so that it can be executed by the Processing Environment (see below). Future versions of the Modelling Environment could incorporate interactive modelling facilities. – – The Processing Environment is both a programming framework and a software application that retrieves requests for the execution of chains of model-components from a queue. This queue is represented by a table in the database, and experiment processing requests can be entered through the SeamGUI. The Processing Environment enables model composition and execution. The actual exchange of data among the models in a chain is based on the OpenMI that provides a standardized interface to define, describe and transfer data between software components that run sequentially. This choice was made based on technical and functional requirements and the possibility offered to re-use legacy models. The use of the central database to store the queue allows for multiple instances of the Processing Environment to run in parallel. External to the layered architecture, but fundamental for SEAMLESS-IF, is the Knowledge Manager, that provides access to and manages persistency of data in the databases. SeamFrame uses the domain model and classes generated for it from the SEAMLESS ontology by the Knowledge Manager. These classes become part of the Modelling Environment. Through a Hibernate-based object relational mapping (www.hibernate.org) the domain model is stored in databases.

9 A Web-Based Software System for Model Integration in Impact Assessments

217

The Role of Software Engineering in the Design of Integrated Assessment Tools Integrated assessment systems tend to be studied and built as part of (large) research projects where the approach is mostly bottom-up: starting from simple models building a larger and more complex model, accounting for the interactions among the simpler processes (Knapen et al. 2007). This emphasizes the linking of the models and perhaps the development or re-using of a framework to support it. Possibilities of the models and the created links between them, and the framework’s flexibility are then put behind a most of the time rather technical looking user interface. The same system could also be thought of from a top-down perspective: start from the larger problem, and try to detail its subcomponents. This is more of a User Centred Design (UCD) (Raskin 2000) approach where usability and business requirements drive the features and technical development. Tell tailing would be that the user interface would be conceptualised first, and its usefulness studied with the intended customers or target customer groups. The bottom-up view serves mostly the modellers and framework builders, and the top-down all the other customers that most likely pay (in some way or another) for the development of the system and only care to a limited level about the intricate internal machinery of connected interdisciplinary models. To them, in essence it is a data intensive software system that has to provide usable information at the right moment, timely for the decision taking process. Data intensive software systems are well known to software engineering, for example consider the large banking and insurance systems. These are commonly called (Business) Information Systems, or Enterprise Applications. If we regard integrated modelling systems from this perspective, would it make sense to apply to them some of the same software design rules – or software architecture, as used for other enterprise applications? The answer is logically positive, and we have thus adopted advanced software development methodologies, especially relying on design patterns to assist our design process, as they provide a mechanism for providing software design advice in a reference format (Gamma et al. 1994). We have been inspired by design patterns in some of our key design choices: –– The layered software architecture, an architectural design pattern (Fowler 2007), with its multi-tier levels, helped us to organise the complexity of software, by clearly delegating different roles to components in the different layers. –– Data Transfer Objects (DTOs) helped us to optimise remote method calls. A data transfer object is an object that holds all the data required for a call to a remote interface. This pattern has proven itself very useful as the amount of data to be transferred between the client and the server can become rather sizeable in integrated assessment studies, and optimisation in this area brought sensible improvements in the tool usability. –– The Service Layer and the Data Access Object design pattern were useful in the definition of SeamServer boundary and its set of available operations from the perspective of interfacing client layers. It encapsulates the business logic, controlling transactions and coordinating responses in the implementation of its operations.

218

J.-E. Wien et al.

–– The Message Service pattern was fundamental in the integration and the exchange of information between the client and the server applications. It provides a publish/subscribe infrastructure that allows client(s) and the server the exchange messages in real time. There are two key components: a message service running in the application server and a client-side API (Application Programming Interface, a set of routines, data structures, object classes and/or protocols provided by libraries and/or operating system services in order to support the building of applications). The message service manages a set of destinations and handles the asynchronous messaging to them. Unlike synchronous messaging this does not rely on direct connections, the sender of the message does not have to wait for a response from the recipient because it can rely on the middleware to ensure delivery of the request and eventually the response. Representational State Transfer (REST; the architectural style of e.g. HTTP) and web services are examples of asynchronous messaging and a good strategy for integration of enterprise applications. They promote a loosely coupled system of components and also encourages design of components with high cohesion (local processing) and low adhesion (remote work). –– The CRUD pattern (create, retrieve, update, delete) has been placed at the base of our data persistence strategy. Almost all applications include some form of persistence storage and have to perform CRUD operations on it (Kilov 1990).

SEAMLESS Server Technologies As described, SEAMLESS-IF is based on a layered, client-server architecture. The processing environment facilities, in particular the model chain executor, are deployed on the SEAMLESS server. The current software stack for the server configuration consists of a web application server (Tomcat 1), data storage (PostgreSQL2), a data access layer (Hibernate) and the SeamServer software (Fig. 9.5). The ontology is used to maintain the consistency across the different data perspectives: it is used to generate the database structure, object-relational mapping files for Hibernate and Java Beans to access the data. The business logic, where the preparation and management of experiments to be run by the processing environment happens, is based on the Spring Framework,3 a Java 4 based solution delivering a full-stack Java/JEE application framework.

http://tomcat.apache.org/ http://www.postgresql.org/ 3 http://www.springsource.org/ 4 http://java.sun.com/ 1 2

9 A Web-Based Software System for Model Integration in Impact Assessments

219

Fig. 9.5 The server technologies adopted for the SEAMLESS server. At the lowest level, we have the Tomcat application server, and the PostgreSQL database. Access to data is made possible through Java Beans using Hibernate ORM (Object Relational Mapping). The business logic is implemented using the Spring framework and remoting based on BlazeDS. The end-user GUI is implemented in Flex, it runs on the client’s web-browser (using the Flash Player) and dialogues with the remoting level

Finally, the remoting technology we adopted is BlazeDS, a server-based Java remoting and web messaging technology that allows the connection to Adobe® Flex® and Adobe AIR™ applications for delivering rich Internet application (RIA). Some of the advantages of this approach are: –– Maintainability: less custom source code to maintain, instead leveraging of popular open source frameworks; –– Stability and security: replacing custom coding with well-known and tested open source frameworks; –– Performance: replacing XML with the binary AMF (Action Message Format). Tests shows that this is at least four times faster; –– Flexibility: use of well-known design patterns to enhance the flexibility, such as dependency injection.5

http://martinfowler.com/articles/injection.html#InversionOfControl

5

220

J.-E. Wien et al.

SEAMLESS Client Technologies The client-server model is a popular design for large and complex applications, not only for mainframe systems but also for remote systems that provide services over the Internet. Think of e-mail clients using mail storage servers or applications running inside a web browser. Confusingly client is used for software as well as hardware. In general clients are classified as fat clients, thin clients, or hybrid clients (Table 9.1). A fat client (also known as a thick client or rich client) is a client that performs the bulk of any data processing operations itself, and does not necessarily rely on the server. A thin client is a minimal sort of client, its functionality being limited to the presentation layer. Thin clients use the resources of the host computer. A thin client’s job is generally just to graphically display pictures provided by an application server, which performs the bulk of any required data processing. A hybrid client is a mixture of the above two client models. Similar to a fat client, it processes locally, but relies on the server for storage data. These are also known as rich clients and implement both presentation and application layers. In designing a multi-tier architecture, there is a decision to be made as to which parts of the task should be done on the client, and which on the server. This decision can crucially affect the cost of clients and servers, the robustness and security of the application as a whole, and the flexibility of the design for later modification, porting and reuse. The SeamServer server application can be accessed by SEAMLESS clients. Basically we can consider two types of clients: system-to-system and human-to-system. A human-to-system client is a graphical user interface that allows using SeamServer functionalities, e.g. to define a project or to visualize results. An initial client (SeamGUI) is developed as part of SEAMLESS-IF. SeamGUI is a hybrid client, it has no local storage but does provide local processing. For example in the case of its SeamPRES component that retrieves calculation result data from SeamServer and processes it locally to create visualisations (tables, charts, map, and so on). A hybrid client currently is mostly referred to as a RIA, software applications running on the user’s computer that rely on a server for core functionality but also have some own logic for an enhanced user experience. A system-to-system client is a programming interface that exposes the SeamServer functionality. In this case, there is no user interaction, rather there are just computers exchanging information (actually the SeamServer application server provides services that the second computer is consuming). The chosen SEAMLESS-IF architecture, using open standards whenever possible, is capable of supporting system-to-system clients. However, this is not a priority within our work, and we do not envision deploying such clients during the duration of the project. Table 9.1 Types of clients Local storage Fat client Yes Hybrid client No Thin client No

Local processing Yes Yes No

9 A Web-Based Software System for Model Integration in Impact Assessments

221

For building SeamGUI and SeamPRES, the two flagship client-applications of SEAMLESS-IF, Adobe Flex6 has been selected as the most usable technology. It offers declarative user interface programming with a rich library of professional and functional components. Code is compiled and runs in a Flash virtual machine (inside a web browser). Protocols for data exchange (pull, server push and data binding) with a server are included. Since a Flex application basically is deployed as a Flash application running within a web browser it is easy to create GUI mock-ups and distribute them to get feedback. This process helped a lot in clarifying the end-user requirements. End-users could easily access the client application and provide with their feedback on design and usability issues. Finally, it is worth mentioning that web based clients use extensively the Cascading Style Sheets (CSS) standard whenever possible to allow future synchronization (or changes) of look and feel.

The SEAMLESS Knowledge Manager The SEAMLESS Knowledge Manager has been developed as part of the Thinklab platform, as an open source project7 for accessing and managing ontologies. It is built upon the Protege-OWL libraries, and has been designed using the Java Plug-in Framework approach (Villa 2005). The SEAMLESS Domain Manager, a plug-in part of the Knowledge Manager, is a tool for delivering Java objects that are used by models and tools for exchanging information. The SEAMLESS Domain Manager uses Thinklab core infrastructure for accessing ontologies and following the domain class principle, it may: 1. Generate domain classes from ontology; 2. Access domain objects at runtime (by using EJB/Hibernate). The Domain Manager plug-in of the Knowledge Manager (Villa 2005; Rizzoli et al. 2007a, b; Athanasiadis and Janssen 2008) provides in this way the service for persistently storing generated classes using the EJB/Hibernate technology (www.hibernate.org). Through Hibernate a relational database may store (and retrieve) populated Java objects. In this sense, the Domain Manager is the ‘knowledge processing’ component of SeamFrame, providing access to and modification of SEAMLESS data, through the ontology-specified interfaces (Fig. 9.6). Fetching data from the Knowledge Base actually means to retrieve data from a relational database (the SEAMLESS database), using information provided by Hibernate.8

http://www.adobe.com/products/flex/ http://www.integratedmodelling.org. Terms of use are defined by the General Public Licence (http://www.gnu.org/gpl). 8 http://www.hibernate.org/ 6 7

222

J.-E. Wien et al.

Fig. 9.6 The Domain Manager component. The Domain Manager is used to automatically generate Java Beans which, using the Hibernate platform, can persist objects used to exchange data among models and the database

The Knowledge Manager only deals with metadata (i.e. information on data structure). Data are delivered by the SEAMLESS database and are accessible by using the generated Java Beans through Hibernate. In the generated source code, Java annotations are used for linking Java Beans with the ontology, like in this example: @ConceptURI(“http://ontologies.seamless-ip.org/crop.owl#Crop”)

public class Crop implements Serializable { ... …@PropertyURI(“http://ontologies.seamless-ip.org/crop.owl#hasCropSoil Requirements”) …public CropSoilRequirements getCropSoilRequirements(){ ... } } Since the annotations are accessible at runtime using Java reflection the ontology information can be used in the future for reasoning, e.g. to validate whether an output of a model can be used as an input for another model. In all practicality it is a fully automatically generated persistence layer for the SEAMLESS system. Higher layers of the system, and the models (or model wrappers) themselves, use it to retrieve, work with and store instances of the concepts as defined in the ontology.

9 A Web-Based Software System for Model Integration in Impact Assessments

223

OpenMI and Model Linking Although the SeamFrame application server operates in Java environment, the model components can be implemented using other languages as long as they can be integrated. This requirement has been translated into the fact that model components should be OpenMI 9 compliant, by implementing the OpenMI interfaces and allow linking to other components (for data exchange). The model component can take care of this by itself or a wrapper or bridge can be programmed. The Open Modelling Interface and Environment (www.openmi.org) has been developed from the need to answer questions related to integrated hydrological management of catchments within the EU fifth framework program project HarmonIT (www.harmonit.org). OpenMI provides a standardized interface for data exchange between software components that run sequentially, based on a pipes and filters architecture (Gregersen et al. 2007). Since the release of the OpenMI in early 2006 the environmental domain adopted the OpenMI within several European projects. This introduced new requirements, which were implemented, resulting in a new OpenMI version (Verweij et al. 2007). Examples of these projects are: –– –– –– ––

SEAMLESS – assess agricultural and agro-environmental policies; SENSOR – assess sustainability impacts of land use related policies; NitroEurope – assess the effects of reactive nitrogen in the environment; EFORWOOD – assess sustainability impacts of European forest wood chains.

These projects have a strong integrated character: environmental, social and economic dimensions are taken into account to perform an ex-ante integrated assessment. An IA can be defined as an interdisciplinary process of combining, interpreting and communicating knowledge from diverse scientific disciplines in such a way that the whole cause–effect chain of a problem can be evaluated from a synoptic perspective with two characteristics: • It should have added value compared to single disciplinary assessment. • It should provide useful information to decision makers (Rotmans and Dowlatabadi 1998). OpenMI provides a standardized interface to define, describe and transfer (numerical) data between software components. The data definition concerns what the data is about (quantity) and where (element set) and when (time) it applies. Each component (LinkableComponent) has a meta data description of its exchangeable data in terms of a quantity and an element set. Each unique exchangeable quantity is registered and published in a so-called ExchangeItem. Connections between ExchangeItems of LinkableComponents are defined by a Link and exist as a separate entity (Fig. 9.7). For more information see http://www.openmi.org. OpenMI is a pull-based system (a pipes and filters architecture). Figure 9.8 shows how the chain of LinkableComponents is triggered (step one) by a successive http://www.openmi.org/

9

224

J.-E. Wien et al. OutputExchangeltem

Exchangeltem InputExchangeltem

1 Quantity LinkableComponent

getValues(link)

input

output

input Link

input

1 ElementSet

output

output

Fig. 9.7 Simplified class diagram of LinkableComponents and Links

3. getValues C

2. getValues

1. trigger

B 4 A 5

Fig. 9.8 Call chain of LinkableComponents. A, B and C are linkable components

call of the LinkableComponent method getValues() (step two and step three) after which values are returned to the original caller (step four and step five). The number of links between two models can vary from one up to hundreds. A link describes one semantic connection concerning what variable is exchanged between the models and where the exchange takes place (Gregersen et al. 2007). To get any number of variables from a model, it needs to be called for each of these

9 A Web-Based Software System for Model Integration in Impact Assessments

225

variables. It is up to the internal intelligence of the model to avoid unnecessary (re) calculations (e.g. by using some caching system). Allowing a link to concern multiple variables simplifies the use of that link within a call for variable values. Links could also be given more functionality, but there should stay a clear distinction between functionality of a link versus that of a LinkableComponent.

Using SEAMLESS-IF for Integrated Assessment Studies In this section we describe how SEAMLESS-IF can be used in an integrated assessment study. First we focus on the pre-modelling phase and the definition of the boundary conditions of the problem. Then, we enter the modelling phase, where the appropriate model chains are selected, their parameters are defined and sets of simulations experiments are executed. Finally we describe the post-modelling stage, where results are analysed and interactively examined by the end user, an activity that may lead to a redefinition of the pre-modelling or modelling stages. The process is iterated until the end user has gained a sufficient insight on the problem and she has obtained a satisfactory result in the analysis process.

Starting a Project in SEAMLESS-IF (Pre-modelling Phase) Through SeamGUI, the user may build up a project that specifies an integrated assessment exercise. A project (Fig. 9.9) is characterized by the definition of the problem it tries to solve or study, and it incorporates at least one experiment configuration, that is, the configuration of the models to be executed during the analysis. An experiment, in turn, is associated with a single model chain and is parameterized by the specification of an outlook, a context and a policy option. Also the indicators are associated with a problem; they are used for quantifying the analysis results. Through a single project, there are several alternatives that can be investigated. Based on the results of the computation, the calculated indicators become available in the framework and can be used for visualizing results in SeamPRES, the visualization client. The project definition is a result of the DOT. Force 10 team (Janssen et al. 2007), and has been specified as an ontology in OWL. Through the SeamGUI client application, the end user may configure a SEAMLESS-IF project, by specifying the narrative descriptions of problem and experiments, context, outlook and policy options, but also, by selecting indicators applicable for the exercise at hand. Figure 9.10 presents the screenshot of the project definition panel.

Seamless Data-Ontology Task Force.

10

Fig. 9.9 Partial view of the Project Ontology

seamproj:Narrative

seamproj:hasPolicyOption

seamproj:hasNarrative

seamproj:PolicyOption

seamproj:hasNarrative

seamproj:Outlook

seamproj:hasPolicyEvaluation

seamproj:PolicyEvaluation

seamproj:hasOutlook

seamproj:hasNarrative

seamproj:Context

seamproj:hasContext

seamproj:BiophysicalSimulation

seamproj:hasBiophysicalSimulation

seamproj: Experiment

seamproj:SpatialScale

seamproj:ComplexityScale

seamproj:hasSpatialScale

seamproj:hasOwners*

seamproj:Owner

seamproj:hasProblem

seamproj:hasComplexityScale

seamproj:hasExperiments*

seamproj:Problem

seamproj:ofProblem

seamproj:Project

226 J.-E. Wien et al.

9 A Web-Based Software System for Model Integration in Impact Assessments

227

Fig. 9.10 Screenshot of the project definition panels in SeamGUI

Narrative Experiments Another important aspect in the pre-modelling phase is the narrative specification of the experiments. The narrative specification is required, so that the narrative of the policy expert can be translated to the integrative modeller (and the specialists of each of the model component, i.e. the agricultural modellers), who will materialize them in turn, as project parameters through the Experiment configuration in the modelling phase. There is a one-to-one relationship between the narrative and the configuration: each context, outlook and policy option has a narrative description that will be accessible to the integrative modeller while detailing the configuration.

Indicator Selector and Fact Sheet Viewer The pre-modelling phase concludes with a choice of indicators for the project. The indicator manager allows the selection of indicators based on the Goal Oriented Framework or directly from the library of available indicators. The aim of using an indicator framework in selecting indicators for an impact assessment is to assist the user in selecting a balanced set of indicators that can help to get a clear picture of how and in what way the assessed future policy may contribute or not to a more sustainable agriculture and society as a whole. The indicator framework can consequently be seen as an indicator “sorting” tool that helps its user to see the impacts of the future policy from different perspectives. Figure 9.11 illustrates the panels for managing the indicators.

228

J.-E. Wien et al.

Fig. 9.11 Indicator manager screenshot

Two types of indicators are available in SEAMLESS-IF. First, the so called endorsed indicators which are fully described by a group of indicator experts and documented with PDF documents that can be reviewed from within SeamGUI. The other type is called model variables, basically the data exchanged between models on the OpenMI level, and documented on a model basis. The selection of indicators for a project triggers the model chain calculation, and if supported by a model it can optimise its calculations based on the requested results. The visualization of results in SeamGUI is also based on this selection of indicators.

Constrained Choice of the Model Chain by Scale Selection (Modelling Phase) The scales that are relevant when performing an integrated assessment are the scales of the problem definition, including policy options, contexts, outlooks and indicators. The scales (spatial and temporal extents/resolution) of the problem are defined by the integrative modeller jointly with the policy expert. The scale of a model is defined by the modeller, taking into account the scales of the available data. The interpretation of the problem scale into the model scale is made by the modeller when the model is integrated in SEAMLESS-IF. The selection of the extent and the resolution of the problem scale in SEAMLESS (in the pre-modelling phase) implies the scale of the models and also the model chain to be used. Specifically, in the final release of SEAMLESS-IF the combinations

9 A Web-Based Software System for Model Integration in Impact Assessments

229

Table 9.2 Problem scales and model chains in SEAMLESS-IF Extent Resolution Model chain 11 Regional AEnZ FSSIM-AM/APES/FSSIM-MP EU25 Farmtype CAPRI/EXPAMOD/FSSIM-MP/FSSIM-AM Farmtype AEnZ FSSIM-AM/APES

reported in Table 9.2 are foreseen, but more can be added, thanks to the extensibility of the proposed approach.

Detailed Specification of Experiments A project may encapsulate several experiments. Each experiment corresponds to a single run of the model chain. All experiments in the same project refer to the same model chain, as the latter is implied by the scales of the problem. An experiment consists of three parts: the Outlook, that defines the trends of the envisioned future, the Context that specifies the boundaries of the problem and is specific to the biophysical subchain, and the Policy option that defines the conditions for the policy assessment subchain. In Fig. 9.12 we display, as example, the policy option panel.

Chain Execution, Server-Side Queuing Model Execution and Client-Side Monitoring When the modeller has completed the configuration of a model chain (that is, of the experiment), s/he can start its execution. This adds the newly configured model chain to the queue of the model chains to be executed on the server. SeamServer provides the queuing facility. The end-user may monitor the queue on his/her client and see which model chain is next to be executed. When an experiment arrives to the top of the queue, it is retrieved by the (one of the) processing environment(s) that executes it on a server. The model chain type is retrieved from the experiment, instantiated, initialized with the experiment and executed. While models of the chain are finishing, intermediate results and indicator values are linked to the experiment. When the chain is finished, it links the final results to the experiment. As the results are attached to the experiment, they can be visualized with SeamPRES, a component of the SeamGUI application. 11 Model name abbreviations: FSSIM – Farm System Simulator, FSSIM-AM – the Agricultural Management part of FSSIM; FSSIM-MP – the Mathematical Programming part of FSSIM, SEAMCAP – a SEAMLESS version of the ‘Common Agricultural Policy Regionalised Impact analysis’ model, APES – Agricultural Production and Externalities Simulator.

230

J.-E. Wien et al.

Fig. 9.12 Detailed experiment configuration in the SeamGUI. The example gives the export subsidies for different commodities in a policy proposal

At present, only one model chain can be executed at a time, but the current design also allows for parallel execution of more model chains at once, by installing multiple (virtual) servers with all required models and a Processing Environment.

The Visualization Environment (Post-modelling Phase) SeamPRES is the visualization tool in SEAMLESS-IF, implemented as a component of the SeamGUI application. It enables the users of SEAMLESS-IF to interactively display impacts, indicators and model outputs and provides clipboard copy and paste integration with other applications. For example it is possibly to copy table data from SeamPRES to the systems clipboard and paste it into another application like Microsoft Excel for further processing. SeamPRES is a tool that can digest and visually display SEAMLESS model outputs in various ways, to improve the analysis and the dissemination of the model results. These model outputs are either available directly, processed (or copied) into indicator results or, when compared based on experiments and expected changes, as impacts. The initial version of SeamPRES can retrieve calculated indicator values and display those in three major ways: tables, graphs and maps. As an integrated component of

9 A Web-Based Software System for Model Integration in Impact Assessments

231

Fig. 9.13 SeamPRES map viewer displaying indicator results per NUTS2 regions

SeamGUI SeamPRES uses all information and functionality from the other components to perform its specific tasks. For example the selection of indicators is used in SeamPRES to decide what values are supposed to be visualized and possible compared to each other, avoiding some of the possible illogical choices that a user could make. SeamPRES uses the same functionality as the rest of SeamGUI to communicate with the SeamServer to retrieve the calculation results. Some specific optimisations have been made to cope with the large amount of data and variety of types of indicators that have to be processed (see also Fig. 9.13).

Conclusions The challenge in integrated modelling is the conceptual integration. To achieve this, we need explicit semantics and a shared conceptualization. A participatory and collaborative approach is a key success factor for the creation of a common ontology for models, indicators and raw data. SEAMLESS has achieved a very important shift in integrated modelling practice, i.e. modellers specify the data requirements of their models in a higher level, i.e. that of an ontology. This ontology is then automatically transformed into a relational database model, to which “data collecting” activities need to comply with. This ensures the match between data collection and data requirement for running the model-chain. It also provides the required semantic

232

J.-E. Wien et al.

interoperability. The use of ontology has proved to be very useful if not essential both for the technical integration of knowledge in the SEAMLESS Integrated Framework and in understanding the meaning of communicated words of the diversity of people within the project. The client and server technologies that were adopted in SEAMLESS-IF have an advantage in the field of stability and security by replacing custom coding with well-known and tested open source frameworks, use of well-known design patterns, and component-oriented programming. By replacing XML with the binary AMF format for exchange of information between clients and server, we improved performance. Tests show that this is at least four times faster. Another advantage and an important requirement for the system is in the field of maintainability and flexibility. There is less custom source code to maintain because of use of popular open source frameworks and standards, and part of the source code is generated from a higher level ontology.

References Athanasiadis, I. N., & Janssen, S. (2008). Semantic mediation for environmental model components integration. Information Technologies in Environmental Engineering, 1, 3–11. Bezlepkina, I., Domínguez, I.P., Heckelei, T., Romstad, E., & Oude Lansink, A.G.J.M. (2007). EXPAMOD: Component to statistically extrapolate from FSSIM models to other farm types and regions including aggregation to NUTS2: Motivation, description and prototype. PD3.6.11.2, SEAMLESS Integrated Project, EU 6th Framework Programme, contract no. 010036-2 (p. 28), from www.SEAMLESS-IP.org Britz, W., Pérez, I., Zimmermann, A., & Heckelei, T. (2007). Definition of the CAPRI core modelling system and interfaces with other components of SEAMLESS-IF. SEAMLESS Rep. No. 26, SEAMLESS Integrated Project, EU 6th Framework Programme, contract no. 010036-2 (p. 116), from www.SEAMLESS-IP.org Cash, D. W., Clark, W. C., Alcock, F., Dickson, N. M., Eckley, N., Guston, D. H., et al. (2003). Science and technology for sustainable development special feature: Knowledge systems for sustainable development. PNAS, 100(14), 8086–8091. Dahl, O.-J., Dijkstra, E. W., & Hoare, C. A. R. (1968). Structured programming. London: Academic. Farquhar, A., Fikes, R., Pratt, W., & Rice, J. (1995). Collaborative ontology construction for information integration (Tech. Rep. No. KSL-95-63). Stanford, CA: Knowledge Systems Laboratory, Department of Computer Science, Stanford University. Fowler, M. (2007). Patterns of enterprise application architecture. Boston, MA: Addison-Wesley. Gamma, E., Helm, R., Johnson, R., & Vlissides, J. M. (1994). Design patterns: Elements of reusable object-oriented software. Upper Saddle River, NJ: Pearson Education. Gijsbers, P.J.A., & Gregersen, J.B. (2005). OpenMI: Glue for model integration. In A. Zerger & R.M. Argent (Eds.), MODSIM 2005 International Congress on Modelling and Simulation (pp. 648–654). Australia/New Zealand: Modelling and Simulation Society. Gijsbers, P.J.A., Wien, J.E., Verweij, P., & Knapen, R. (2006). Advances in the OpenMI. In Proceedings of 7th International Conference on Hydroinformatics, HIC 2006 (pp. 72–81). Gregersen, J. B., Gijsbers, P. J. A., & Westen, S. J. P. (2007). OpenMI: Open modelling interface. Journal of Hydroinformatics, 9(3), 175–191. Griffen, E. M. (1997). A first look at communication theory. New York: McGraw-Hill. Gruber, T. R. (1993). A translation approach to portable ontology specifications. Knowledge Acquisition, 5, 199–220.

9 A Web-Based Software System for Model Integration in Impact Assessments

233

Gruber, T. R. (1995). Toward principles for the design of ontologies used for knowledge sharing. International Journal of Human and Computer Studies, 43(5/6), 907–928. Holsapple, C. W., & Joshi, K. D. (2002). A collaborative approach to ontology design. Communications of the ACM, 45(2), 42–47. IEE (Institute of Electrical and Electronics Engineers). (1990). IEEE standard computer dictionary: A compilation of IEEE standard computer glossaries. New York: IEEE. Janssen, S. J. C., Andersen, E., Athanasiadis, I. N., & van Ittersum, M. K. (2009). A database for integrated assessment of European agricultural systems. Environmental Science and Policy 12(5), 573–587. Janssen, S.J.C., Wien, J.J.F., Li, H., Athanasiadis, I.N., Ewert, F., Knapen, M.J.R., Huber, D., Thérond, O., Rizzoli, A., Belhouchette, H., Svensson, M., & van Ittersum, M.K. (2007). Defining projects and scenarios for integrated assessment modelling using ontology. In L. Oxley & D. Kulasiri (Eds.), MODSIM 2007 International Congress on Modelling and Simulation (pp. 2055–2061). Australia/New Zealand: Modelling and Simulation Society. Kilov, H. (1990). From semantic to object-oriented data modeling. In First International Conference on System Integration (pp. 385–393). Knapen, M.J.R., Verweij, P.J.F.M., & Wien, J.J.F. (2007). Applying enterprise application architectures in integrated modelling. In L. Oxley & D. Kulasiri (Eds.), MODSIM 2007 International Congress on Modelling and Simulation (pp. 798–804). Australia/New Zealand: Modelling and Simulation Society. Neches, R., Fikes, R., Finin, T., Gruber, T., Patil, R., Senator, T., et al. (1991). Enabling technology for knowledge sharing. AI Magazine, 12(3), 36–56. Ogden, C. K., & Richards, I. A. (1923). The meaning of meaning (8th ed.). New York: Harcourt, Brace & World. Patel-Schneider, P.F., Hayes, P., & Horrocks, I. (2004). OWL web ontology language semantics and abstract syntax. W3C Recommendation. Retrieved from http://www.w3.org/TR/owlsemantics/ Pennington, D.D., Madin, J., Villa, F., & Athanasiadis, I.N. (2007). Computer-supported collaborative knowledge modeling in ecology. Workshop on social and collaborative construction of structured knowledge. In 16th International World Wide Web Conference (WWW2007). CEUR Workshop Proceedings, 273. BANFF, Canada published online. Raskin, J. (2000). The humane interface – new directions for designing interactive systems. New York: ACM Press. Rizzoli, A.E., Athanasiadis, I.N., & Villa, F. (2007a, September). Delivering environmental knowledge: A semantic approach. In O. Hryniewicz, J. Studzinski, & M. Romaniuk (Eds.), Proceedings of 21st International Conference on Informatics for Environmental Protection: EnviroInfo 2007 (Vol. 1, pp. 43–50). Germany: Shaker Verlag. Rizzoli, A. E., Donatelli, M., Athanasiadis, I. N., Villa, F., & Huber, D. (2008). Semantic links in integrated modelling frameworks. Mathematics and Computers in Simulation, 78(2–3), 412–423. Rizzoli, A., Li, H., Athanasiadis, I.N., & Marechal, F. (2007b). Semantically rich interfaces for simulation interoperability. 2007 European Simulation Interoperability Workshop (p. 60), Simulation Interoperability Standards Organization, Genoa, Italy. Rotmans, J., & Dowlatabadi, H. (1998). Integrated assessment modeling. In S. Rayner & E.L. Malone (Eds.), Human choice and climate change: Tools for policy analysis (Vol. 3, pp. 291–377). Columbus, OH: Batelle Press. Rotmans, J., & van Asselt, M. (1996). Integrated assessment: A growing child on its way to maturity. Climatic Change, 34(3–4), 327–336. Schoemaker, P. J. H. (1993). Multiple scenario development: Its conceptual and behavioral foundation. Strategic Management Journal, 14(3), 193. Scholten, H., Kassahun, A., Refsgaard, J. C., Kargas, T., Gavardinas, C., & Beulens, A. J. M. (2007). A methodology to support multidisciplinary model-based water management. Environmental Modelling and Software, 22(5), 743–759.

234

J.-E. Wien et al.

Sølvberg, A. (1998). Data and what they refer to. In P. P. Chen (Ed.), Concept modeling: Historical perspectives and future trends. Berlin: Springer. van der Aalst, W. M. P., & van der Hee, K. M. (2002). Workflow management: Models, methods and systems. Cambridge: MIT Press. Van Ittersum, M. K., & Donatelli, M. (2003). Modelling cropping systems [Special issue]. European Journal of Agronomy, 18(3–4), 187–394. Van Ittersum, M. K., Ewert, F., Heckelei, T., Wery, J., Alkan Olsson, J., Andersen, E., et al. (2008). Integrated assessment of agricultural systems - a component based framework for the European Union (SEAMLESS). Agricultural Systems, 96, 150–165. Verweij, P.J.F.M., Knapen, M.J.R., & Wien, J.J.F. (2007). The use of OpenMI in model based integrated assessments. In L. Oxley & D. Kulasiri (Eds.), MODSIM 2007 International Congress on Modelling and Simulation (pp. 1166–1171). Australia/New Zealand: Modelling and Simulation Society. Villa, F. (2005). A semantic model of computation for natural system modelling. In A. Zerger & R.M. Argent (Eds.), MODSIM 2005 International Congress on Modelling and Simulation (pp. 751–757). Australia/New Zealand: Modelling and Simulation Society. Villa, F., Athanasiadis, I. N., & Rizzoli, A. E. (2009). Modelling with knowledge: A review of emerging semantic approaches to environmental modelling. Environmental Modelling and Software, 24(5), 577–674. Villa, F., Donatelli, M., Rizzoli, A.E., Krause, P., Kralisch, S., & van Evert, F.K. (2006). Declarative modelling for architecture independence and data/model integration: A case study. In A. Voinov, A. Jakeman, & A.E. Rizzoli (Eds.), Proceedings of the iEMSs Third Biennial Meeting, “Summit on Environmental Modelling and Software”. Burlington, VT: The International Environmental Modelling and Software Society. Wien, J. J. F., Knapen, M. J. R., Janssen, S. J. C., Verweij, P. J. F. M., Athanasiadis, I. N., Li, H., et al. (2007). Using ontology to harmonize knowledge concepts in data and models. In L. Oxley & D. Kulasiri (Eds.), MODSIM 2007 International Congress on Modelling and Simulation (pp. 1959–1965). Australia/New Zealand: Modelling and Simulation Society. Zander, P., & Kächele, H. (1999). Modelling multiple objectives of land use for sustainable development. Agricultural Systems, 59, 311–325.

Part III

Use and Extensions of Integrated Assessment Modelling

Chapter 10

Evaluating Integrated Assessment Tools for Policy Support Jacques-Eric Bergez, Marijke Kuiper, Olivier Thérond, Marie Taverne, Hatem Belhouchette, and Jacques Wery

Introduction During recent decades Integrated Assessment (IA) tools (van Ittersum et al. 2008) have been developed and used widely in order to provide answers to complex questions arising from policy-makers (Rotmans 1998; Harris 2002; Toth 2003). Success of such tools lay in the fact that they provide a conceptual and operational framework to assess the effectiveness and trade-off among different policy options, while integrating the knowledge from various disciplines and/or stakeholders which is required for dealing with complex system issues and associated interactions and feedbacks (Rotmans 1998; Tol and Vellinga 1998; Toth 2003). Their policy-orientation makes evaluation of IA tools a critical issue. Aiming to support policy decisions, results from the assessment studies need to be robust enough to be useful in the political arena but also sensitive enough to distinguish and choose between different policy proposals. IA tools are however by definition complex and encompass paradigms, methods and tools from different disciplines. Developing an operational IA methodology that provides useful results for policy decisions is already a tall order; evaluating such a system is even more challenging J.-E. Bergez (*) and O. Thérond INRA, UMR 1248 AGIR, BP 52627, F-31326, Castanet Tolosan, France e-mail: [email protected]; [email protected] M. Kuiper LEI Wageningen UR, P.O. Box 29703, 2502 LS, The Hague, The Netherlands e-mail: [email protected] M. Taverne Dynamics and Functions of Rural Areas, Cemagref, 24 avenue des Landais, BP 50085, 63172, Aubière Cedex, France e-mail: [email protected] H. Belhouchette and J. Wery INRA – UMR SYSTEM (Systèmes de Culture Tropicaux et Méditerranéens), Montpellier, France e-mail: [email protected]; [email protected] F.M. Brouwer and M. van Ittersum (eds.), Environmental and Agricultural Modelling: Integrated Approaches for Policy Impact Assessment, DOI 10.1007/978-90-481-3619-3_10, © Springer Science+Business Media B.V. 2010

237

238

J.-E. Bergez et al.

(Lopez 2003). Not only do individual components of the system need to be tested, which could be done in accordance with the evaluation procedures of a specific discipline from which the component originates. In addition the system as a whole needs to be evaluated, assessing whether interactions between different components originating from different disciplines perform properly. Evaluation is then likely to be less dependent on the conventional and classical peer review and history matching and more dependant upon protocols and tests yet to be developed (Parker et al. 2002). In practice development of an operational IA methodology requires extensive resources, leaving limited time and funds for evaluation. This chapter reflects on the evaluation of IA tools, looking at different types of components of an IA methodology as well as their interactions. The aim of this chapter is to derive general lessons and ideas for evaluating IA tools from the practical experiences gained from the SEAMLESS project. We first offer the evaluation approach used in SEAMLESS. It consists of three steps (conceptual evaluation, technical evaluation, system evaluation), each of which is discussed in more detail in subsequent sections. In each of these sections we discuss evaluation of three types of components of an IA methodology: procedures, quantitative tool and graphical user interfaces. The different character of these components gives rise to different evaluation approaches. The last section concludes by summarizing the lessons learned from SEAMLESS which may be useful for others projects aiming at evaluating IA tools.

Deriving a General Approach to Evaluate IA Tools Many methods and tools have been developed to deal with IA objectives and constraints (van Ittersum et al. 2008). Broadly speaking, they can be split into two groups: analytical (embracing models, scenarios and risk analysis) and participatory (including dialogue methods, policy exercises and mutual learning methods) (Rotmans 1998). Among these methods, Integrated Assessment and Modelling (IAM) includes a variety of quantitative models as well as scenario-based approaches (Sharma and Norton 2005). Such tools aim to support managers to control uncertainty when they are making decisions about future options. When making policy decisions, scenarios enable policy makers to anticipate by exploring possible futures and to assess different alternatives according to their potential consequences (van Notten et al. 2003; Börjeson et al. 2006). Scenario-based approaches tell ‘highly detailed, logically consistent stories about the future’ (Sharma and Norton 2005). Model-based approaches aim at describing quantitatively as accurately as possible the causal relationships and interactions between the various components of the system under study, in reaction to external constraints and endogenous changes. Scenario-based approaches can use simulation results from quantitative models. Since IAM tools aim at addressing policy questions, their evaluation needs to look beyond the classical evaluation processes of quantitative models (Oreskes et al. 1994; Rykiel 1996; Sinclair and Seligman 2000). The majority of literature on evaluation deals with verification and validation of Decision Support Systems (DSS), which

10 Evaluating Integrated Assessment Tools for Policy Support

239

are decision-oriented and model-based like IAM tools (Finlay and Wilson 1991; Mosqueira-Rey and Moret-Bonillo 2000; Brunner and Starkl 2004; Jakeman et al. 2006; Sojda 2007). Evaluation of DSS can be separated into verification (‘building the system correctly’) and validation (‘building the right system for a given purpose’) (Boehm 1981 cited by Mosqueira-Rey and Moret-Bonillo 2000). The verification step ensures that the DSS is internally complete, coherent, and logical from a modelling and programming perspective (Sojda, 2007). Validation is less concerned with internal operation of the software and more concerned with its output and its usefulness to the user. It analyzes whether the decision support system addresses the user’s problem (e.g., making better decisions, avoiding bad ones, or helping the user to take these decisions more quickly or with less data, information, and knowledge). This attention for suitability to address the user’s problem sets DSS and IAM tools apart from the classical evaluation of numerical models, like ecological and agronomical models (Parker et al. 2002). For DSS development, iterative development-evaluation processes, such as spiral methodology (Boehm 1988 cited by Mosqueira-Rey and Moret-Bonillo 2000) are advocated. These methods allow incremental development and fast prototyping that are fundamental in the development of an intelligent system. In spiral methodology the final stage of each development cycle is considered as an evaluation step of the quality of the developed product. Successful implementation of DSS or IAM for decision-making relies on the use of three evaluation dimensions (Adelman 1992 cited by Sojda 2007): • Examining the logical consistency of the system’s algorithms (verification); • Empirically testing the predictive accuracy of the system (validation); and • Documenting users’ satisfaction (validation). Finlay and Wilson (1991) make a further distinction inside the testing of the predictive accuracy of the system. They use ‘analytical validation’, for checking each part of the modelling system, and ‘synoptic validation’ for checking that an acceptable output from the whole modelling system is achieved for each set of inputs. Many authors add to this validation one or several additional phases – commonly grouped together under the term evaluation – assessing aspects of the tool beyond the validity of the final solutions. As a result evaluation becomes an endeavour to analyse aspects such as utility, robustness, rapidity, efficiency, extension possibilities, ease of use, credibility, etc. (Mosqueira-Rey and Moret-Bonillo 2000).

Three Phases to Evaluate an Integrated Framework In this chapter we focus on the case of the integrated framework (IF), SEAMLESS-IF. SEAMLESS-IF has been developed for ex-ante assessment of alternative agricultural, economic and environmental policy options and technical innovations for their impacts on the sustainability of agricultural systems at European, national or regional levels (van Ittersum et al. 2008). This framework is a model-based tool to

240

J.-E. Bergez et al.

support policy decisions by translating policies to scenarios which can be analyzed through this tool. In terms of evaluation three major components of SEAMLESS-IF can be distinguished: –– IA procedures; –– Quantitative tools (database, numerical models and indicators); –– And user interface of the computer system. Two main procedures are part of SEAMLESS. The first deals with how SEAMLESS-IF may be used in impact assessments. The second is more concerned with the assessment of the institutional compatibility of proposed policies. The framework is built upon various types of quantitative tools. SEAMLESS-IF includes several databases that not only provide inputs for the model but also have stand-alone value in providing a single point of access to datasets at different levels and in different domains. Reflecting the aim of IA there are models from different domains and ranging from field to global market level which are linked to each other, using a modular software to keep the framework open for further developments. Indicators are the third group of quantitative tools which aim to integrate results from the various models in a set of variables relevant to assessing the sustainability impact of policies. Being designed as a computer-based tool there is an important role for the user interface in bringing together the different components, while allowing for assessment of a wide range of potential policy questions with SEAMLESS-IF. In evaluating SEAMLESS-IF, we thus deal with a mixed bag of components requiring an evaluation methodology that is both flexible and practical (i.e. aimed at getting the system operational for the policy support for which it is designed). In order to influence the design of SEAMLESS-IF, this evaluation took place throughout the project, with evaluation methods reflecting the development of the framework over time (Fig. 10.1). Following different authors (Boehm 1988; Mosqueira-Rey and Moret-Bonillo 2000; Sojda 2007) the participatory development strategy of SEAMLESS-IF was

Individual components

Design structure of framework

Integrate components in framework

Link all components of framework

Conceptual evaluation

Technical evaluation

System evaluation

Integrated framework

Fig. 10.1 Three types of evaluation corresponding to development phases of an integrated framework

10 Evaluating Integrated Assessment Tools for Policy Support

241

based on iterative cycles of design and evaluation of the framework’s prototypes. This involved potential users for some aspects and in any cases, feedbacks between ‘testers’ and developers. Prime users (i.e. Directorate General of European Commission) were involved in these cycles by the way of a User Forum (van Ittersum et al. 2008) and other potential users (regional and national decision makers, stakeholder groups) have been periodically consulted. Starting from individual components the first phase consists of designing the structure of the integrated framework (for example by making mock-ups showing how the final system may look without any functionality). Conceptual evaluation checks that the design of components and the integrated framework will lead to the desired functionalities for end-users. This phase precede the ‘verification phase’ as identified by Sojda (2007) with explicit attention to end-user requirements. It aims at defining and clarifying the operational objectives of the tool development (i.e. main expected capacities) and the future tool shape (e.g.: main functionalities and material form) (Dieste et al. 2003). The second phase consists of building the framework from the individual components. Technical evaluation assess whether the components function in technical terms while keeping the objective of the integrated framework in mind (i.e. are the components designed in such a way that they can be further integrated to perform the foreseen analyses). This corresponds to the ‘verification phase’ as defined by Sojda (2007) which checks that the system is complete, coherent and logical from a modelling and programming perspective. But it is better defined by the ‘analytical validation’ (checking each part of the system) as defined by Finlay and Wilson (1991). While more and more components are integrated during the tool development, the evaluation shifts from testing individual components to combinations of different components, evolving into the system evaluation of whether the framework as a whole gives acceptable results for scientists which are useful for end-users. This phase encompasses both tests of validity of solutions of the framework as a whole (“synoptic validation” as defined by Finlay and Wilson [1991]) as well as analysis of solution time, extension possibilities, ease of using the system, etc. (as identified by Mosqueira-Rey and Moret-Bonillo 2000).

Using Case Studies to Guide Evaluation and Development of the Integrated Framework Case studies played a crucial role in the evaluations of SEAMLESS-IF by providing, real world applications of the integrated framework. Through the evaluations test cases have become an important driver of the framework development. In the case of SEAMLESS we opted for two types of test cases (Belhouchette et al. 2006a, b; 2007). The first deals with trade liberalization consisting of lowering EU tariffs on imports and abolishing EU export subsidies. This case study represents a macro (EU and global level) policy in the economic domain for which the implications at all levels (from field to global level) and in all domains (economic, environmental and social) are assessed. The second group of case studies challenges

242

J.-E. Bergez et al.

the system in the opposite directions. It consists of an assessment, at regional level, of environmental policies (Nitrate Directive) in contrasting EU regions in terms of biophysical conditions and cropping systems (Louhichi et al. 2008). These case studies represent regional and environmental policies where new production technologies play an important role and which are again assessed at different scales (field to region) and in all domains. The test cases used different combinations of models, allowing to test the core modelling chains of SEAMLESS-IF. Apart from serving as different ways of evaluating the framework in a technical sense, the policy relevance of the test cases (and the integrated framework in general) was assured by presenting them to prime users (i.e. Directorate General of European Commission) and other potential users (regional and national decision makers, stakeholder groups) at several points during the project (van Ittersum et al. 2008). These case studies proved to be useful for the development of the framework. The multidisciplinary character of the test cases forced to reach consensus among disciplines to arrive at a single and clear definition of the concepts and procedure (ontology) needed for the framework. Hence after a large number of iterations among SEAMLESS participants with different backgrounds a project assessment has been defined as a set of realistic scenarios, based on economic, policy and technological changes. This project is first defined by the spatial scales (resolution and extent) at which the scenario is assessed (e.g. field-region, region-EU) which determines the appropriate modelling chain for the investigated IA problem, as well as the possible economic, environmental and social indicators with which the scenario should be assessed. After choosing scales and indicators, several experiments can be developed by varying the external driving forces, the biophysical context and the policy parameters (Thérond et al. 2009). In a nutshell, the case studies provided real time applications involving different disciplines which allowed derivation of the general approach to implement the IA problem within SEAMLESS-IF and accordingly the graphical user interface (GUI) functionalities to develop.

Conceptual Evaluation The first type of evaluation consists of assessing whether the foreseen integrated framework is internally consistent and suited for addressing the purpose for which it is designed. The specificity of this evaluation is that it used the specifications of the framework’s components as described in the available project reports while components nor framework were yet available for testing. In the case of SEAMLESS-IF the aim – a computer-based tool for ex-ante assessments of alternative agricultural, economic and environmental policy options and technical innovations – is too general for an evaluation. To create a reference point or yardstick for evaluating the framework the test cases were used. These test cases provided the concrete background of the prototype’s evaluations.

10 Evaluating Integrated Assessment Tools for Policy Support

243

Conceptual Evaluation of the Procedure for Using the Integrated Framework Integrated assessment implies an analysis of policies from different perspectives or disciplines, requiring cooperation between actors with different background. Procedures for implementing policy assessments with the computer-based tools aid this cooperation by structuring the exchange of ideas in such a way that policies can be translated into scenarios that can be assessed by the available modelling chains. The development of a procedure for using the framework is a way to ensure that the integrated framework will be able to address the issues of interest for end-users. It is thus closely linked to the aim of a conceptual evaluation (checking that design of components and the integrated framework will lead to the desired functionalities for end-users). Evaluation criteria can be of very diverse abstraction levels. They can be a priori defined or elaborated during the evaluation process (Darses et al. 2004; Darses 2002). For the conceptual evaluation of the IA procedure only criteria of high abstraction level were pre-defined, like relevance, understanding, etc. Commonly procedures are assessed by comparing users’ organisational situations and working methods to spot possible incompatibilities without involving the users and assuming sufficient knowledge of their working environment. However, to ensure the best possible fit between the procedure, and more generally the development of SEAMLESS-IF, and end-user requirements, potential end-users and modellers from different disciplines were involved in the evaluation of the procedure. Involving the end-users allowed inclusion of more concrete evaluation criteria during the evaluation itself. Similar to the argument made with the test cases, care has to be taken to ensure representativeness of the end-users involved in the evaluation for the targeted end-users of the framework. Else there is a risk of gearing the development of the framework to specific users in such a way that functionalities for other users become limited. SEAMLESS-IF is foreseen to be used within a participatory process involving policy experts who bring the problem to study and integrative modellers who set up the IA within SEAMLESS-IF. PE are either in charge of policy design or interested in impacts of policies designed by others. A specific procedure has been developed in which the framework use is a component of the impact assessment procedures of the end-users (Fig. 10.2). The later can include additional tools before or after the scenario analysis done through SEAMLESS-IF (see, for example, SEC 2005 for the EU procedure). The SEAMLESS-IF procedure consists of three main steps: pre-modelling (describing the problem and defining indicators at levels and for domains relevant to the policy question at hand), modelling (selecting the appropriate models and defining scenarios needed to assess the policy question at hand), and post-modelling (analyzing and presenting the model results) (Thérond et al. 2009). This SEAMLESS-IF procedure had a strong influence on the design of the GUI of SEAMLESS-IF, the evaluation of which will be discussed below.

244

J.-E. Bergez et al.

Impact assessment Procedure Stakeholders

SEAMLESS-IF

Articulation between procedures

Graphical User interface

Pre-modelling

Problem description Indicator selection

Modelling

Post-modelling

Typologies - tools chain selection and run

Results presentation and analysis

SEAMLESS-IF scenarios Other user criteria

Fig. 10.2 The SEAMLESS-IF procedure in relation to the integrated assessment procedures of end-users and the SEAMLESS-IF GUI

For organizational reasons, evaluation of the procedure in conceptual, technical and system terms was done during one session in order to facilitate the interactions with end-users. In these sessions working situations, using graphical representations of the framework, were simulated (the framework was still under development) to facilitate further design (Béguin and Cerf 2004; Pastre 2005). Evaluation was done both through in-situ observations of problem formulation and through users’ opinions, highlighting problems met at each step. Given that the framework was not yet operational, the evaluation focussed on the first steps (pre-modelling – problem definition) of defining the policy question in a form suitable for further analysis with SEAMLESS-IF. The conceptual evaluation of the SEAMLESS-IF procedures was based on a role-game simulating the joint problem definition by policy experts (end-users) and modellers (involved in SEAMLESS-IF). The modellers were provided with several guidelines for defining the problem in a SEAMLESS-IF compatible from which they could choose the most appropriate one. This allowed to test different versions of the procedure. During the problem formulation exercise somebody observed the discussion between policy experts and modellers. At the end of the meeting, policy experts were questioned about points that did not emerged spontaneously but were identified beforehand as being relevant. Each of the modellers filled a questionnaire on these same aspects to ensure that all pre-defined evaluation criteria have been covered.

10 Evaluating Integrated Assessment Tools for Policy Support

245

The evaluations of the SEAMLESS-IF procedure were performed with 11 different policy experts (or groups of experts) and eight modellers (or pairs of modellers). Policy experts who were consulted are representative of different types of institutions (government, resource management institution, agriculture advice services, local public institutions) acting in different political fields such as water management, resource management, agriculture and at different levels from local to regional and national. Modellers were SEAMLESS researchers of different disciplines from biophysical (agro-ecology) to social sciences (e.g. economics, geography). They had different levels of knowledge of the framework and its components. This diversity of policy experts and integrative modellers ensured that the procedure (and accordingly the relevance of the framework specifications) was evaluated in a large range of situations so that the outcome of the evaluation exercises were not narrowed to specific issues and working conditions.

Conceptual Evaluation of Quantitative Tools Integrated assessment tools combine quantitative tools in a single framework. In the case of SEAMLESS three groups can be distinguished: databases (and ontology), numerical models and indicators. These components are of course interrelated, within the modelling chains including the database and the procedure to compute indicators for sustainability assessment. The requirements on each component depend on its place in the overall framework. The conceptual evaluation first determined the expected contribution of each component to the overall objective of the integrated framework (i.e. does it deliver the information required by the whole framework). Then each component was assessed individually (i.e. does it provide the information properly). The two case studies were essential in these evaluations by providing concrete requirements for the whole system from which requirements on components could be derived. This illustrates also the need to carefully choose case studies representative of the target range of studies, because by guiding the conceptual evaluation they drive further development of the components. Different types of quantitative tools require different evaluation procedures. Databases are based on the development of ontology that should encompass the different data, procedures and concepts mobilized in SEAMLESS-IF. Using object-oriented databases and specific UML diagrams helped at analysing the links between these different entities and the conceptual functioning between databases and other quantitative tools (models and indicators). Conceptual evaluation of numerical models is easier because these models describe, in their concepts and equations, a part of the reality the users have in mind for the real systems under study. This principle has been used to communicate with users and detect missing information (Dieste et al. 2003). Such conceptual evaluation was mainly based on flowchart and expert appraisal. Conceptual evaluation of indicators was based on a five-points grid: presence of the indicator (“Is it requested by

246

J.-E. Bergez et al.

the end-users?”), pertinence (“Does it express the requested impact or trend?”), robustness (“How do we know that this indicator will be suitable for the different situation from field to EU?”), sensitivity (“Do we think that the indicator will change with the expected policy?”) and how is it computed (“Do we have the knowledge, the data and the skill to compute it?”). This conceptual evaluation was performed in close relation with the general evaluation procedure. Using the test cases as concrete stories led the selection of indicators. APES (Agricultural Production and Externalities Simulator) can illustrate the conceptual evaluation of a quantitative model in the case of SEAMLESS-IF. APES (see also Chapter 4 of this volume) is a field scale crop model describing the soilplant behaviour under a specific climate and technical management. From the test cases we derived that APES should be able to simulate the yield and externalities of main European crops to be further used as input by the farm model to which it is linked. It also needed to be flexible and modular enough to allow for future additions of modules describing other processes (e.g. carbon sequestration) and types of crops (combining several crops in an agroforestry system). In order to generate proper yields and externalities APES has to take into account water and nitrogen stresses as well as soil and weather variability across Europe, have parameters that can be estimated from available data, and the integration of different modules within APES should be scientifically sound (i.e. without scale and precision discrepancy between modules).

Conceptual Evaluation of User Interfaces When extending evaluation to issues like utility, robustness, rapidity, efficiency, extension possibilities, ease of use, credibility, etc. (Mosqueira-Rey and Moret-Bonillo 2000) the evaluation of user interfaces become important. The way in which integrated frameworks can be accessed, i.e. through the graphic user interface, determines to a large extent the type of study for which the integrated framework can be put. The evaluation of user interfaces is closely linked to the target uses and users of the framework. In the case of SEAMLESS-IF which is a computer-based tool a GUI was developed to set up IA within the framework. The GUI design was therefore derived from the SEAMLESS procedure for integrated assessment (see Fig. 10.1). The conceptual evaluation of the GUI was based on mock-ups, which were screenshots of how the GUI would look like while it was not yet operational to run the model chains. These mock-ups were then evaluated to see whether they would allow the implementation of the full SEAMLESS procedure for integrated assessment of each of the test cases. Apart from allowing a conceptual evaluation of the GUI at an early stage of development, mock-ups also facilitated development of a common understanding and representation between participants of different disciplines by providing a graphical illustration of the proposed functionalities of the GUI.

10 Evaluating Integrated Assessment Tools for Policy Support

247

Technical Evaluation The second type of evaluation consists of assessing whether components function in technical terms and are integrated in such a way that the foreseen analyses can be performed. The case studies played a central role in the technical evaluation by providing an operational benchmark for the required properties of the individual components as well as their integration in the framework.

Technical Evaluation of Procedures for Using the Integrated Framework As discussed before, the technical evaluation of the procedure was performed at the same time as the conceptual evaluation but the technical evaluation focussed on the operational feasibility of the procedure. It was assessed using criteria like the distribution of roles among users, degree of interaction among policy experts and modellers and time needed to complete an integrated assessment. These criteria were made operational using a set of questions that were asked both to policy experts and modellers at the end of the evaluation meeting provided these had not yet been discussed spontaneously: –– Does information required for problem formulation from policy experts match their skills? –– Are the guidelines provided for problem formulation easy to use by the modellers? –– Does the translation of policy questions of the experts into experiments compatible with SEAMLESS-IF match with the modellers’ skills? –– Does problem formulation with guidelines of the SEAMLESS-IF procedure produce the necessary information to perform an assessment? The evaluation revealed that a single meeting is not sufficient to define an IA project in a computer-based tool such as SEAMLESS-IF. This was partly due to the framework not being ready yet, hampering the assessment of whether specific questions could be addressed by the framework, and partly due to its complexity. Another key concern was the necessary level of knowledge of the modellers of different disciplines in order to frame the policy question in a manner suitable for assessment within the modelling chains. Another key reason to have more than one meeting is the time needed for sharing a minimum level of mutual understanding between policy experts and modellers.

Technical Evaluation of Quantitative Tools In the case of quantitative tools the technical evaluation focussed first on assessing the functioning of the individual components after which the combination

248

J.-E. Bergez et al.

of components was tested. In the case of quantitative tools the evaluation relies first on the classical evaluation of numerical models (like the well-developed methods for ecological and agronomical models) to ensure that the outputs are valid. The second aspect is to check that each component delivers the right outputs in the right format to the other components of the modelling chain, to ensure that the framework as a whole derives indicators as output from scenarios as input. Evaluation of the individual components is case-specific since it depends on the modelling concepts used by the discipline from which they originate. Continuing with the example of APES, agronomic validation methods were used to assess the validity of the results. In addition the programming of the model was tested (programming bugs, rapidity, RAM and ROM requirement) as well as its compatibility with multiple operating systems. The latter requirement was essential for the objective to integrate APES in the SEAMLESS-IF framework. The testing itself was split into qualitative and quantitative tests. The qualitative tests consisted of expert appraisals of results. Given the known reaction of crops to modification of soil, weather or management (e.g. reduced yield with no N fertilization or less rain) it was analyzed if the model could reproduce this trend. The quantitative tests then proceeded by more in depth study of the precision, robustness and sensitivity of the APES model. These tests took place through workshops in which mini-applications typical of the agricultural activities to be simulated by APES in SEAMLESS-IF were simulated. These workshops involved three types of participants: – – Mini-applications experts with broad knowledge of the soil, crops and agro-management and able to provide data for calibration and evaluation (both data from test cases regions and field experiments were used) –– Component developers of all components combined in the APES configuration required for the mini-application (e.g. the soil water component, the arable crops component, etc.); and –– Software experts able to change equations and parameters in the code version during the workshop This combination of participants allowed several iterations between evaluation and model development. The quantitative evaluation revealed that the availability of suitable detailed data needed for quantitative testing is hard to obtain for the range of situations representative of Europe. This requirement limits the scope for a full evaluation in line with the agronomic standards, but seems inherent to the development of an integrated framework of the size and scale of SEAMLESS-IF.

Technical Evaluation of the User Interface The technical evaluation of the GUI addresses issues like ease of use, rapidity, ergonomics related to the use of the system through the interface. In addition the

10 Evaluating Integrated Assessment Tools for Policy Support

249

technical evaluation needs to assess whether the user interface is able to define and conduct properly the runs of the underlying components. The requirements of a user interface depend strongly on the background of the users and therefore needs to be tested by potential users. In the case of the SEAMLESS-IF GUI role games were the main tool for testing. These role games involved both groups and individuals of SEAMLESS researchers that acted as integrative modellers who interact with policy experts for an impact assessment with SEAMLESS-IF. Varying with the level of development of the integrated framework these tests evolved during the project from defining a policy question with the GUI (pre-modelling), to quantifying scenarios and running models through the GUI (modelling) and to testing the GUI functionalities for displaying and analyzing results (post-modelling). Outcomes from the testing of the SEAMLESS-IF procedure on preferences of potential end-users on display of results were used in this last step. Each round of testing resulted in a set of recommendations to the GUI developers on screens and functionality, which was prioritized in the light of the needs for implementing the test cases.

System Evaluation The third type of evaluation consists of assessing whether the framework as a whole gives acceptable results that are useful for end-users. A key question in this analysis would also be whether the integrated framework as a whole provides different insights from those gained from an analysis based on the individual components. In the case of SEAMLESS-IF the integrated framework was not yet available at the time of writing, preventing evaluations of the system as a whole. Below the outlines are sketched of the way in which such a system evaluation could be done for the three types of components. Comparing the results obtained for the test case with the framework and with the use of its individual components is the ultimate step of this evaluation.

System Evaluation of Procedures for Using the Integrated Framework When performing a system evaluation of the procedures for integrated assessment the focus will be on whether the procedure succeeds in translating a policy question of end-users into assessed scenarios that provide satisfactory and useful information to support policy decisions. This assessment would encompass not only the quality of the definition of the problem and analysis of results in relation to the policy question, but also the time needed for such an assessment in relation to the organization of integrated assessments by end-users.

250

J.-E. Bergez et al.

System Evaluation of Quantitative Tools A system evaluation of quantitative tools in an integrated framework will focus on validation of the outcomes of the linked set of components included in the framework. One of the big issues here is the problem of chain of errors and uncertainty (Gabbert et al. 2009). Another issue whether problems with the compatibility of different components (models or databases) has been solved through the ontology.

System Evaluation of the User Interface The key question of the user interface is whether it provides access to results that are useful and transparent for end-users. Added to this may be requirements on the way results are displayed, which could vary by type of end-user. These requirements can to some extent be determined beforehand (and are also part of the technical evaluation using the test cases). Use of the operational framework by end-users will no doubt lead to additional requirements, just as in the evaluation of the integrated assessment procedure evaluation criteria were defined during testing.

Some More Lessons to Organize the Evaluation of an Integrated Framework The type of testing done for SEAMLESS-IF is summarized in Table 10.1. The experiences gained can be translated in several key points which can be of use for other projects aiming at evaluating integrated assessment tools: use of prototypes,

Table 10.1 Summary of evaluation levels used for the different types of components Conceptual evaluation Technical evaluation System evaluation − Consistency with − Operational Organisational Procedure (e.g. overall objectives feasibility feasibility SEAMLESS-IF procedure for IA) − Relevance for users − Reliability (quality of outcome) − Uncertainty Quantitative tools Consistency with overall − Reliability analysis (outputs quantitative (e.g. APES) objectives and with analysis) major characteristics of the system it − Programming − Input/output simulates (a crop) problems analysis − Operational feasibility User interfaces Consistency with overall Operational feasibility (e.g. GUI) objectives

10 Evaluating Integrated Assessment Tools for Policy Support

251

use of case studies, timing of testing, independent testers, multidisciplinarity, separating end-users and modellers. Where possible we suggest potential solutions to problems encountered during SEAMLESS-IF.

Use of Prototypes As illustrated in Fig. 10.1 the development of SEAMLESS-IF proceeded from a set of individual components to an integrated framework. This process was made operational through the use of prototypes and transition workshops as illustrated in Fig. 10.3. The first prototype both ran parallel with and was the support of the conceptual evaluation, i.e. assuring that the integrated framework would be internally consistent and suited for addressing the purpose for which it is designed. To assure alignment with end-user demands it was oriented at end-users. It consisted of a series of mock-ups or screen shots showing how the integrated framework would look like and what kind of functionalities it would have. These mock-ups were found to be an important tool not only for the conceptual evaluation, but also to generate a common understanding among the various disciplines involved in building the framework. Whereas the first prototype did not have any functionality in terms of running analyses, the subsequent prototypes did have increasing functionalities allowing the evaluation to switch to technical testing of the framework and its components. The switch from one prototype to the next was marked by so-called transition workshops. These workshops brought together developers and testers to translate the assessment of the prototype into (prioritized) requirements for the next prototype.

Fig. 10.3 The role of prototype testing and transition workshops in the development path of SEAMLESS-IF

252

J.-E. Bergez et al.

Evaluations of the prototypes proved to be instrumental in directing the development and functionality of the integrated framework. A drawback of the development of prototypes is the required time and resources which come at the expense of developing the components and conducting research with it. Based on the experiences in SEAMLESS-IF however these investments were considered as essential for building a common understanding of the status and direction of development of the integrated framework as a whole.

Use of Case Studies Two sets of case studies formed the thread along which the evaluations were performed. These cases studies provided contrasting and realistic applications of the integrated framework. Care was taken to make the applications as representative as possible of the target range of questions (top-down versus bottom-up, economic versus environmental policies, region versus EU scale) to avoid to pull the framework development in a single direction. This allowed to test the general applicability of the tool across different disciplines and scales and to keep it generic enough to maintain the flexibility and broader use of the framework. The case studies turned out to provide a crucial basis for assessing the development of the framework and identifying differences between disciplines that affect the integrated assessment. Furthermore, the case studies define operational minimum requirements to the integrated framework which aid priority setting and allocation of resources in the project.

Timing of Testing The production of an integrated framework such as SEAMLESS-IF required development and integration of different components. This simultaneous development and integration proved to be more complex than expected and resulted in considerable delays in prototype delivery to the testing group. These delays in turn hampered testing and made reliance on the conceptual evaluation developed for this purpose (Thérond et al. 2009) an essential contribution to further development of the framework. The delays in arriving at an operational framework also imply that no tests of the system as a whole are possible during the project. The technical tests were done on partially finished components and needed to strike a careful balance between providing constructive comments to the developers already under pressure and providing early indications that a tool may prove unsuitable to the whole system’s features and validity domain. In the case of SEAMLESS-IF, end-users were also involved in the testing to assure the relevance of the integrated framework. Early involvement of users provides ideas on required features and potential fields of application useful for development of the integrated framework. At the same time user-involvement at an

10 Evaluating Integrated Assessment Tools for Policy Support

253

early stage can interfere with proper testing to avoid deterring potential users and leads to pressure on quick delivery of the integrated framework by fuelling users’ expectations. A balance thus needs to be found to ensure that the integrated framework is relevant while not raising users’ expectations beyond feasible levels.

Independent Testers Following Mosqueira-Rey and Moret-Bonillo (2000) and Sojda (2007) the testing of SEAMLESS-IF started with evaluators independent of tool developers. This allowed a more objective evaluation of the tools and their suitability for the integrated framework and it ensured that an interdisciplinary group was identified in the project with the primary objective of regularly testing prototypes in realistic applications to keep the development of the framework on track (Fig. 10.3). During the project, however, it proved difficult to enrol scientists in evaluation work that is hard to publish in scientific journals, while requiring a high level of involvement to understand and assess prototypes with limited features and a high level of integration of computerized tools from different disciplines. In addition to this push effect the evaluators were also pulled into other parts of the project, for example contributing to the development of links between models, due to their extensive knowledge of the framework and its components. These push and pull factors led to a blurring of the distinction between evaluators and developers at the end of the project. Given these factors at play it seems advisable that evaluators start as external reviewers involved in the conceptualisation of the whole framework based on potential applications, but not on tool development and become involved in integrated tool development once individual components become available for linking.

Multidisciplinarity Evaluating a complex multidisciplinary tool like SEAMLESS-IF is a daunting task. As noticed by Langvad and Noe (2006) multidisciplinary involvement is a precondition for developing relevant decision support systems. At the same time a different disciplinary background compounded by a wide diversity of cultures and practices in an international project like SEAMLESS poses a major challenge to both development and evaluation of the framework. As an illustration, in the evaluation of the integrated assessment procedure the aim was to allow a diversity of ways to frame problems to account for the diversity in backgrounds of both the policy experts and modellers involved. Accounting for this diversity implied testing in different countries. Because of language it was impossible to nominate a single person to record observations (spontaneous comments from policy experts about presentations and the way they interact with integrative modellers for problem

254

J.-E. Bergez et al.

formulation) to assure a uniform recording of observations. Instead a standardized way of recording had to be developed that interfered with the aim of having diversity in problem representations.

Separating End-Users and Modellers The aim of an integrated framework to use models in managing complex situations rather than in sharply defined areas of research, results in people with little modelling or quantitative background relying on models while not being in a position to judge their quality or appropriateness. Linking models from different disciplines in a single framework implies that this danger not only applies to use of the framework by non-modellers as identified by Jakeman et al. (2006), but also to modellers relying on inputs from models outside their discipline. Being unaware of limitations, uncertainties, omissions and subjective choices may lead to invalid conclusions by wrongly interpreting model results and/or applying models for purposes for which they are not designed. The only way to mitigate these risks is to generate wider awareness of the modelling process, choices made, good practice for testing and applying models and the interpretation of model results. Within SEAMLESS-IF these considerations are accounted for in the design of the framework separating pre- and post-modelling from the modelling phase to which only integrative modellers have access. Although this reduces the risks of having users with limited quantitative background running models, it still allows modellers to use tools from other disciplines without sufficient background information. This latter risks can probably only be mitigated by having analyses done by a team of modellers from different disciplines, which may be hard to realize in practice. Acknowledgements The authors wish to thank all the participants of SEAMLESS-IF involved in testing as well as the students that have worked on the testing specific components of SEAMLESS-IF.

References Adelman, L. (1992). Evaluation decision support and expert systems. New York: Wiley. Béguin, P., & Cerf, M. (2004). Formes et enjeux de l’analyse de l’activité pour la conception de systèmes de travail. @ctivités, 1, 54–71. Belhouchette, H., Therond, O., Adenaeuer, M., Kuiper, M., Bigot, G., AlkanOlsson, J., Wery, J., et al. (2007). Documentation of baseline and policy options to be assessed with Prototypes 2 and 3, D6.2.3.3. SEAMLESS Integrated Project, EU 6th Framework Programme, contract no. 010036-2, www.SEAMLESS-IP.org, 112 p. Belhouchette, H., Wery, J., Therond, O., Bergez, J.E., & van Ittersum, M. (2006a). Design of scenarios for integrated impact assessment of interactions between EU policies and agro-ecological technologies using SEAMLESS-IF. Proceedings of the IX ESA Congress (Pt. III), Warsaw (Poland), pp. 11–20.

10 Evaluating Integrated Assessment Tools for Policy Support

255

Belhouchette, J., Wery, J., Therond, O., Duru, M., Bigot, G., et al. (2006b). The major characteristics of environmental policies and agro-ecological technologies to be studied in Test case 2. SEAMLESS Rep. No.13, SEAMLESS Integrated Project, EU 6th Framework Programme, contract no. 010036-2, www.SEAMLESS-IP.org, 87 p. Boehm, B. W. (1981). Software engineering economics. Englewood Cliffs, NJ: Prentice Hall. Boehm, B.W. (1988). A spiral model of software development and enhancement. Computer, May, 61–72. Börjeson, L., Hojer, M., Dreborg, K. H., Ekvall, T., & Finnveden, G. (2006). Scenario types and techniques: Towards a user’s guide. Futures, 38, 723–739. Brunner, N., & Starkl, M. (2004). Decision aid systems for evaluating sustainability: A critical survey. Environmental Impact Assessment Review, 24, 441–469. Darses, F. (2002). A cognitive analysis of collective decision-making in the participatory design process. Participatory Design Conference, Malmö, Sweden. Darses, F., Détienne, F., & Visser, W. (2004). Les activités de conception et leur assistance. In P. Falzon (Ed.), Ergonomie (pp. 545–563). Paris: PUF. Dieste, O., Genero, M., Juristo, N., Mate, J. L., & Moreno, A. M. (2003). A conceptual model completely independent of the implementation paradigm. Journal of Systems and Software, 68, 183–198. Finlay, P. N., & Wilson, J. M. (1991). Validation for decision support systems: Recent developments and findings. Systems Practice, 4, 599–610. Gabbert, S., Van Ittersum, M., Ewert, F., Kroeze, C., Stalpers, S. & Alkan-Olsson, J., (2009). Uncertainty information in Integrated Assessment: The users’ perspective. Regional Environmental Change, in press. Harris, G. (2002). Integrated assessment and modelling: An essential way of doing science. Environmental Modelling and Software, 17, 201–207. Jakeman, A. J., Letcher, R. A., & Norton, J. P. (2006). Ten iterative steps in development and evaluation of environmental models. Environmental Modelling and Software, 21, 602–614. Langvad, A. M., & Noe, E. (2006). (Re-)innovating tools for decision-support in the light of farmers’ various strategies. In H. Langeveld & N. Röling (Eds.), Changing European farming systems for a better future – new visions for rural areas (pp. 335–339). The Netherlands: Wageningen. Lopez, M. (2003). Application of an evaluation framework for analyzing the architecture tradeoff analysis methodSM. Journal of Systems and Software, 68, 233–241. Louhichi, K., Belhouchette, H., Flichman, G., Therond, O., & Wery, J. (2008). Application of FSSIM in two test case regions to assess agro-environmental policies at farm and regional level, PD6.3.2.2. SEAMLESS Integrated Project, EU 6th Framework Programme, contract no. 010036-2, www.SEAMLESS-IP.org, 67 p. Mosqueira-Rey, E., & Moret-Bonillo, V. (2000). Validation of intelligent systems: A critical study and a tool. Expert Systems with Applications, 18, 1–16. Oreskes, N., Shrader-Frechette, K., & Belitz, K. (1994). Verification, validation, and confirmation of numerical models in the earth sciences. Science, 263, 641–646. Parker, P., Letcher, R. A., Jakeman, A. J., Beck, M. B., Harris, G., Argent, R. M., et al. (2002). Progress in integrated assessment and modelling. Environmental Modelling and Software, 17(3), 209–217. Pastre, P. (2005). Apprendre par la simulation: de l’analyse du travail aux apprentissages professionnels. Octares, Collection Formation, 363 p. Rotmans, J. (1998). Methods for IA: The challenges and opportunities ahead. Environmental Modeling and Assessment, 3, 155–179. Rykiel, E. J., Jr. (1996). Testing ecological models: The meaning of validation. Ecological Modelling, 90, 229–244. SEC (2005). Impact Assessment Guidelines (791), from http://ec.europa.eu/governance/impact/ docs/SEC2005_791_IA_guidelines_main.pdf Sharma, M., & Norton, B. G. (2005). A policy decision tool for integrated environmental assessment. Environmental Science and Policy, 8, 356–366. Sinclair, T. R., & Seligman, N. G. (2000). Criteria for publishing papers on crop modeling. Field Crops Research, 68, 165–172.

256

J.-E. Bergez et al.

Sojda, R. S. (2007). Empirical evaluation of decision support systems: Needs, definitions, potential methods, and an example pertaining to waterfowl management. Environmental Modelling and Software, 22, 269–277. Thérond, O., Belhouchette, O., Janssen, S., Louhichi, K., Ewert, F., Bergez, J.E., Wery, J., Heckelei, T., Alkan Olsson, J., Leenhardt, D., & van Ittersum, M. (2009). Methodology to translate policy assessment problems into scenarios: The example of the SEAMLESS Integrated Framework. Environmental Science and Policy, 12(5), 619–630. Tol, R. S. J., & Vellinga, P. (1998). The European Forum on Integrated Environmental Assessment. Environmental Modelling and Assessment, 3, 181–191. Toth, F. L. (2003). State of the art and future challenges for integrated environmental assessment. Integrated Assessment, 4, 250–264. van Ittersum, M., Ewert, F., Heckelei, T., Wery, J., Alkan Olsson, J., Andersen, E., et al. (2008). Integrated assessment of agricultural systems – a component-based framework for the European Union (SEAMLESS). Agricultural Systems, 96, 150–165. van Notten, P. W. F., Rotmans, J., van Asselt, M. B. A., & Rothman, D. S. (2003). An updated scenario typology. Futures, 35, 423–443.

Chapter 11

A Comparison of CAPRI and SEAMLESS-IF as Integrated Modelling Systems Wolfgang Britz, Ignacio Pérez Domínguez, and Thomas Heckelei

Introduction Integrated modelling systems address a wide range of topics by combining different modelling tools. The main motivation for this integration in the context of policy impact analysis is the fact that ‘no model can serve all purposes’ (van Tongeren et al. 2001). Moreover, there exist many stand-alone bio-physical or economic agricultural models with a longstanding record of application. Efforts on development and validation – for large models, the latter is often only possible over a series of repeated applications and their appreciation by academics and policy maker’s (Funtowicz and Ravetz 1994) – might be saved by integrating these existing models instead of expanding each single one in several directions. This chapter tries to compare and assess the main characteristics of CAPRI and SEAMLESS-IF as integrated modelling systems. With this purpose, their conceptual and technical differences and similarities are identified. It is important to note that a modified simulation engine of CAPRI is embedded in SEAMLESS-IF, and given the name SEAMCAP. Within SEAMCAP, important modifications to the CAPRI full stand-alone version are included. First of all, from a technical perspective, the model is adapted to allow for communication with the family of models and databases included in SEAMLESS-IF. Secondly, from a methodological perspective, econometrically or exogenously defined response parameters in the regional models of SEAMCAP are replaced by parameters derived from simulations with a modelling chain (van Ittersum et al. 2008) comprising the Agricultural Production

W. Britz (*) and T. Heckelei Institute for Food and Resource Economics, Chair for Economic and Agricultural Policy, University of Bonn, Nussallee 21, D-53115, Bonn, Germany e-mail: [email protected]; [email protected] I. Pérez Domínguez Public Issues Division, Agricultural Economics Research Institute (LEI) Alexanderveld 5, 2585 DB, The Hague e-mail: [email protected] F.M. Brouwer and M. van Ittersum (eds.), Environmental and Agricultural Modelling: Integrated Approaches for Policy Impact Assessment, DOI 10.1007/978-90-481-3619-3_11, © Springer Science+Business Media B.V. 2010

257

258

W. Britz et al.

Externality System (APES, see Chapter 4), a Farm System SIMulator (FSSIM, see Chapter 5) and the extrapolation tool EXPAMOD (Pérez Domínguez et al. 2009). And thirdly, some model components of CAPRI are not taken up by SEAMCAP, like the farm type layer or the spatial disaggregation module. Therefore, this chapter compares the full stand-alone CAPRI modelling system with SEAMLESS-IF, which includes the SEAMCAP model component. In the following section we start with a short description of CAPRI and its relationship to SEAMLESS-IF. The subsequent three sections compare objectives and structure, concepts for database and coverage as well as types of model links used in the two modelling systems. The succeeding four sections deal with different aspects of modelling agricultural production and agri-environmental interaction at farm and regional level looking into model calibration, technologies, policies and environmental indicators. This is followed by a description of the baseline generation for forward looking impact assessment and a comparison of the software platforms of both systems. Finally, the chapter ends with a brief look into challenges for the future and conclusions.

Short Description of the CAPRI Modelling System and Its Integration in SEAMLESS-IF The core of the CAPRI (Common Agricultural Policy Regional Impact; Britz et al. 2007) model is an economic agricultural sector model aiming to analyse impacts of changes in the European Union (EU) (or international) agricultural policies on European agriculture and global agricultural commodity markets, typically in a forward looking analysis (8–10 years ahead). Technically, the economic core is a static, partial equilibrium model consisting of four interconnected modules: • Regional non-linear programming models of agricultural supply for EU-27, Norway and Western Balkans; • An international multi-commodity model for major primary and secondary agricultural products including bi-lateral trade flows; • An EU market model for young animals and, finally; and • A module generating premium schemes and other policy instruments of the Common Agricultural Policy (CAP) (see also Fig. 11.1). The supply module consists of non-linear programming (NLP) models comprising about 50 crop and animal activities for each of about 250 level-two-regions of the Nomenclature des Unités Territoriales Statistiques (the so-called NUTS 2 level). Each regional model maximises agricultural income at given prices and subsidies subject to resource constraints on land, feed and nutrient requirements as well as policy restrictions. Regional agricultural income is defined as the gross value added (GVA) at producer prices plus direct subsidies (premiums). Costs neither included in the GVA nor covered by the restrictions are captured by a quadratic cost function, whose parameters are either estimated from time series analysis (Jansson 2007)

11 A Comparison of CAPRI and SEAMLESS-IF as Integrated Modelling Systems

Exploitation

Market Model

• Environmental indicators • Welfare analysis

Supply Feed Demand

Supply Models • Nuts 2 / farm type optimisation models

Premiums

259

(Problem: maximisation of agricultural income subject to restrictions)

• Multi-commodity spatial market model • 28 regional world aggregates & EU27

Prices

Young Animal Market Module • Linked optimisation models (Member State level)

Policy Module • Calculation of premiums depending on Common Market Organisations

Herds Acreages

Aggregation

Fig. 11.1 The CAPRI model chain (CAPRI Model Documentation [Britz et al. 2007])

or calibrated to exogenous elasticities. At the same time, the parameters are chosen such that the model recovers the base year price quantity framework. Coupled and decoupled premiums from the first pillar of the CAP are calculated for each activity in the policy model. The global market module (Britz et al. 2007: 96ff.) either differentiates individual countries, for example single EU Member States, US, Canada, China, or regional aggregates, for example the African, Caribbean and Pacific (ACP) countries with preferential access to EU markets. It includes all primary agricultural products of the European Accounts of Agriculture also covered by the supply module plus major secondary products such as dairy commodities, vegetables cakes and oils. The market model is designed as to allow for linkage to the regional programming models, but also observing data availability and technical feasibility. Data on market balances, to name an example, are only available at Member State level, and therefore, market clearing and price formation mechanisms were only implemented at country level for the EU. The market module is formulated as a spatial multi-commodity model allowing for bilateral trade flows based on a “product differentiation by origin” assumption (Armington 1969). It explicitly represents major agricultural trade policy instruments such as (bilateral) tariffs, tariff quotas, and subsidised exports. It follows a template approach, with structurally identical equations for all markets, and applies flexible functional forms calibrated to parameters collected from literature or other modelling systems such as the FAO’s World Food Model (see, e.g. Frohberg and Britz 1994). For projection, trend forecasts on prices, yields and other economic variables are based on outlooks from the European Commission (2004) and the Food and Agriculture Organization of the United Nations (FAO 2003).

260

W. Britz et al.

The market module for young animals links aggregates of the regional programming models at EU Member State level to determine market clearing prices for the different young animal categories such as calves and piglets. Additional constraints ensure consistency between young animal inputs and outputs of the differentiated animal production activities. The policy module calculates activity specific subsidies − originating from pillar one of the CAP relating to market support, − subject to regional and national ceilings in values or physical units, thereby assuring that regional programming models do not violate these policy characteristics. At the same time, the module defines additional domestic market support measures and derives a CAP budget. The overall economic core modelling system is solved by iterating between the different modules. A typical iteration starts with the regional supply models. The results obtained are passed on to the market module by re-calibrating the supply and feed demand functions for the countries covered by regional supply modules. Solving the global market module returns market clearing prices along with supply, demand and trade flow quantities. If ceilings on subsidies are violated, premiums are adjusted. The resulting prices and premiums are passed back and drive the next iteration of the regional supply models. The iterative solution process converges rapidly towards a comparative-static equilibrium characterised by cleared markets and conformity of prices and quantities in both the supply and the market modules. Major outputs of the global market model include trade flows, market balance positions and producer/consumer prices by product and country/country aggregates, whereas the supply module delivers crop acreages and herd sizes with their associated revenues, costs and income, as well as information about feeding and nutrient management practices for each NUTS 2 region. CAPRI is thus unique in its ability to show the regional impact of global changes in agricultural markets and policies on EU agriculture consistently representing mutual feedback between the two. Developed since 1996 and operational since 1999, CAPRI has been gradually expanded and enhanced in many projects and applications, and is regularly applied for policy impact assessment until today. The combination of regionalised supply models and a global multi-commodity model allowed for the impact assessment of various reforms of the first pillar of the CAP, such as the Agenda 2000 policy reform (Britz and Heckelei 2000b), the mid-term review policy reform (Britz et al. 2006) and the sugar market reform (Adenaeuer 2005). In these applications, policy reforms were analysed which changed both market and farm policy instruments such as the introduction of coupled payments while reducing administrative prices. At the same time, pure agricultural trade liberalization proposals could be analysed as well (Junker et al. 2005; Britz and Jacquet 2006). The solution taken up by SEAMLESS-IF was to include a technically modified core CAPRI simulation engine (SEAMCAP) covering the supply-market model interaction as a ‘single’ model component in its model chain with input data coming from EXPAMOD (Pérez Domínguez et al. 2009) and model outputs delivered to the SEAMLESS-IF indicator calculators and in the future also to multi-sectoral models

11 A Comparison of CAPRI and SEAMLESS-IF as Integrated Modelling Systems

CAPRI-Grid

DNDC (Soil N flows)

261

APES

FSSIM-AM

Database Model

FSSIM-MP

Supply Module (Nuts 2 and Farm Types)

EXPAMOD

Market Module

Market Model (SEAMCAP)

Young Animal Trade / Policy Module

GTAP Factor Markets (Labour)

Bio-Energy Model

Dairy Processing Model

CAPRI MODEL CHAIN Fig. 11.2 The CAPRI and SEAMLESS-IF model chains

in the chain (GTAP-Model, see Hertel 1997). Therefore, price and market outcomes are endogenous in SEAMLESS-IF as well using a CAPRI version whose supply response is based on extrapolation of the FSSIM models (see also Chapter 5 of this volume; see Fig. 11.2). Thereby, the trade policy instruments are also included in SEAMCAP, but only a subset is currently explicitly represented in the scenario editor of SEAMLESS-IF limiting the accessibility to implement changes in scenario definitions.

Comparing Objectives and Structure Perhaps the most striking similarity of SEAMLESS-IF and CAPRI is the fact that both modelling systems have been designed as instruments for policy impact assessment on European agriculture at pan-European level. SEAMLESS-IF can be considered a full vertical model chain, with a large range of scales ranging from calculating environmental externalities at field level to multi-sectoral analysis of policies in an open economy (see van Ittersum et al. 2008). CAPRI is an integrated model chain with an ‘economic core model’ linked to satellite models in different

262

W. Britz et al.

directions: vertical (e.g. spatial downscaling) and horizontal (link to biophysical models or processing activities). The vertical approach of SEAMLESS-IF allows the assessment of agrienvironmental policies (from a bottom-up perspective) or multi-sectoral trade agreements (from a top-down perspective) compared to CAPRI. SEAMLESS-IF aims at better exploiting existing agricultural economic, bio-economic and bio-physical simulation models through flexible and automated linking. This shall be achieved by a generic software platform facilitating the generation of model chains flexibly tailored to the objectives of a specific analysis. The issue of modularity receives special attention in SEAMLESS-IF, where all models are linked within an open software infrastructure, which aims at making models substitutable. This requires a clear and uniform way to define generic interfaces of model components. SEAMLESS-IF employs an ‘ontology’ of all system components (see Chapter 9) managed by a knowledge database to serve this purpose. CAPRI is a more standard product in the sense that it supports only one model chain, in which components may be enabled or disabled. Since an iterative linking between some modules is deemed necessary, the separation of model components in the code is less stringent compared to SEAMLESS-IF. The basic structure of CAPRI has been kept constant through the years, but the different modules and their interrelationships have evolved. Also, some new model components, for example introducing spatial downscaling (Leip et al. 2008) or the calculation of an energy indicator (Kempen and Kränzlein 2008), have added to the complexity of this highly integrated modelling system and little attention has been paid to modularity. The issues of data consistency and full EU coverage of agricultural production are key drivers of CAPRI research activities and inherited by SEAMLESS-IF through the SEAMCAP model component.

Differences in Concepts for Database and Coverage CAPRI has strong roots in mathematical programming approaches, dating back to farm type and regional models developed in the 1970s and 1980s at Bonn University for the German agricultural sector models DIES (Bauer 1978) and RAUMIS (Henrichsmeyer et al. 1996). It equally carries genes inherited from economy-wide accounting identities as found in Input-Output (IO) analysis or Computable General Equilibrium (CGE) models. This materialises in a database concept where all economic activities of the farming sector are described both in monetary and physical units and consistent to the definition of the Economic Accounts for Agriculture (EAA). To cover the whole agricultural sector, CAPRI comprises the complete utilizable agricultural area of the EU at regional scale (NUTS 2), at the same time incorporating all crop and animal production activities covered by the EAA. Such an approach was only deemed feasible by using structurally identical template models sourced by a harmonized regional database at European level.

11 A Comparison of CAPRI and SEAMLESS-IF as Integrated Modelling Systems

263

Accordingly, the resolution of CAPRI in terms of inputs, outputs, production activities and regions is driven by data availability, with the EUROSTAT agricultural domain as the main data source. Regional data aggregate to Member State and higher level based on clearly defined and fully consistent aggregation rules. To achieve this database consistency making best use of the information available, CAPRI employs Bayesian techniques (Britz et al. 2007: 15ff.). SEAMLESS-IF does not follow these consistency and full-coverage requirements for all components. For example, the bio-economic farm models (FSSIM, see Chapter 5) are incorporated into the system from a bottom-up perspective, sourced by a bio-physical model (APES, see Chapter 4) for the definition of agricultural processes and by regional surveys on management practices. These models are designed to simulate responses of the bio-physical system regarding crop growth, nutrient fate, the water cycle or soil erosion to a site-specific combination of farming practices and physical characteristics. The bio-physical models may be seen as an instrument to construct the production possibility set on a specific site, by systematically varying different parameters describing the management practices such as rates and timing of fertilizer application, seeding, and irrigation. Combined with a behavioural assumption of maximising expected income penalised by variability to reflect risk aversion and data on costs and revenues associated with each production possibility set, this information allows building farm optimisation models able to allocate land and other resources to current (i.e. observed) and alternative activities. The calibration to crop and livestock production quantities observed in the base year is performed by using a variant of Positive Mathematical Programming (PMP). The ultimate link to aggregate market models is made by extrapolating supply response behaviour in terms of rates of quantity changes per rate of price change (elasticities) from a sample of FSSIM farm type models. This is achieved with the EXPAMOD model (Bezlepkina et al. 2007; Pérez Domínguez et al. 2009) using data from price-sensitivity experiments with FSSIM as observations for the estimation of supply response surfaces for agricultural products depending on variables defining the classification of the SEAMLESS-IF agri-environmental (e.g. rainfall, solar radiation, soil type, carbon content) and farm resources (e.g. availability of land, labour and machinery), both types of variables available at Pan-European level. The regional supply models of CAPRI are calibrated to the extrapolated price elasticities.

Concepts of Linking Models/Components From a methodological perspective, we can identify two types of model linking possibilities here: (i) by means of econometrically estimated response functions, and (ii) by means of (iteratively) modifying the output of model components such as to fit the combined output of other models. The first option has been used in SEAMLESS-IF for linking farm optimisation and aggregated market models and was already described above. The stand-alone CAPRI model also applies this

264

W. Britz et al.

option by including estimated cost functions for dairy products from the European Dairy Industry Model (EDIM) to parameterize dairy processing. Furthermore, such a statistical approach is also applied to represent the relevant part of the simulation behaviour of the bio-physical model DNDC (DeNitrificationDeComposition [DNDC] model, Li et al. 1994) into the spatially explicit CAPRI results (see further below). The second option has been used to link economic models working at different spatial scales, such as the breakdown of results at NUTS 2 regional or farming system level in CAPRI, by using regional or farm accounting statistics (for CAPRI farm type layer, see Britz et al. 2007: 107–114); at different time scales, such as the calibration of agricultural production to econometric projections or expert knowledge at a future point in time (e.g. baseline calibration to DG-AGRI medium-term market outlooks, see European Commission 2007b: 38–45), and different scope, such as the link between the CAPRI aggregated agricultural sector and GTAP as attempted in SEAMLESS-IF (see Fig. 11.2). By doing so, the aggregation bias of single model components is reduced, more flexibility in methodology is achieved and different data sets can be better exploited without forcing the development of a ‘one-fits-all model’. Nevertheless, within this type of framework, it is important to be consistent from an ontology perspective (Chapter 9), i.e. with a clear data structure, so that similar scenarios, model variables and parameters are correctly exchanged. In Fig. 11.2, the main interactions between model components, in both the CAPRI and SEAMLESS-IF model chains, are presented. SEAMLESS-IF consists of a vertical concatenation of models whereas the CAPRI model chain presents a more ‘spider-like’ structure, with vertical (e.g. DNDC) and horizontal links (e.g. factor markets and bioenergy processing linked to the supply module) to models and databases. The part of CAPRI included in SEAMLESS-IF, i.e. the SEAMCAP model component, is comprised in a green box. All model components in the CAPRI model chain outside this box are not included in SEAMLESS-IF. The arrows represent flows of information between modules (one or both directions). The dotted lines connecting model components between both model chains indicate a significant similarity of functionality within the respective modelling chain (for example, DNDC and APES, both dealing with environmental externalities at lower level).

The Farm Type Models in Both Systems NUTS 2 level regions are often characterized by a mix of rather diverse farming systems, so that simulating the average regional impacts of policies, e.g. on the environment, may hide important differences between farms of different specialization. As a response to a request of DG Environment, the CAPRI network developed a farm type layer in 2003 where the database and the supply model were consistently broken down up to eight different farm types, based on FADN (Farm Accounting Data Network) farm typology and data. These single farm type models are based

11 A Comparison of CAPRI and SEAMLESS-IF as Integrated Modelling Systems

265

on the same model template as the regional supply models, allowing their parameterization and exploitation within the same model structure. For each NUTS 2 region, the most important farm types according to acreage/livestock units were identified and explicitly modelled, and the rest of the sector was aggregated into a residual group. An update in 2006 expanded the coverage to some new Member States and introduced a two-dimensional grouping by farm size and specialization. An expansion to all EU Member States where farm group compositions are based on the Farm Structure Survey (EUROSTAT) statistics is currently under way. Differently from CAPRI, the most representative farm types in SEAMLESS-IF are selected from the FADN sample following three criteria: specialization, size and intensity. Here, only a sub-sample of farm types in a wide range of regions across the EU-27 is modelled explicitly, without considering any residual farm type category. In these sample regions, specific surveys have been carried out to obtain the necessary data on management techniques and biophysical conditions for running the APES and FSSIM models. The main advantages of the farm type layers in CAPRI and SEAMLESS-IF compared to the NUTS 2 regional layer are • The ability to capture the policy impacts from a specialized farm perspective; and • An improved supply response representation as the decision space of farmers is more appropriately depicted. The main disadvantage is the lack of time series data on farm type production changes at Pan-European level. Both the CAPRI and SEAMLESS-IF farm type layers still leave room for further improvement. For instance, the introduction of regional markets for agricultural intermediates such as nutrients, manure or fodder, the representation of regional land markets or a better integration of information on production costs from FADN can be mentioned.

Calibration of Farm Type and Regional Programming Models A major challenge in using representative or aggregate programming models is the calibration to observed behaviour. Calibration to observed behaviour must be understood as the ability to recover a given situation, but also to react to changes of determinants in a similar way as observed in the past. The limited number of constraints which can be motivated from economic theory, agronomic knowledge and available data, combined with the requirement to cover a larger list of crop and animal production activities, excluded a classical linear programming (LP) approach for CAPRI. The methodological development of PMP was, therefore, a cornerstone in CAPRI as it allowed for perfect calibration in combination with a smooth simulation response based on observed behaviour. Indeed, a rather visible methodological contribution of the CAPRI team was the step-by-step evolution of the original PMP

266

W. Britz et al.

approach to an econometrically estimated primal-dual approach (Britz and Heckelei 2000a; Heckelei and Wolff 2003; Heckelei and Britz 2005; Jansson 2007). In SEAMLESS-IF, the FSSIM farm type models are comparable to the regional supply models in CAPRI. They originally had been designed as pure mixed integer linear programming models. However, PMP calibration was introduced at a later stage to allow for more realistic simulation behaviour. The standard version of the models used for upscaling are based on a variation of the standard PMP approach (Kanellopoulos et al. 2007) which allows recovery of observed average returns to land, but the FSSIM models are currently not calibrated to empirically estimated supply responses. Furthermore, the survey data used for FSSIM do not necessarily perfectly match regional statistics, so that only relative responsiveness to prices and policies is extrapolated by EXPAMOD, and the SEAMCAP supply model continues to be calibrated to a given price-quantity framework in the base year or baseline (ex-post or ex-ante).

Technology and Externalities In CAPRI, an average technology for each agricultural production activity in a specific region is defined based on regional data on crop areas, herd sizes and yields, along with engineering data on fertilization and feeding practices. Moreover, other data such as Standard Gross Margins or statistical estimates from the FADN are taken into account. The technology is expressed as a vector of Input-Output (IO) coefficients per activity. To perform simulations to a future point in time, the output coefficients of each activity are trend projected and input coefficients adjusted in a proportional fashion adjusted by assumptions on input saving technical progress. Environmental or bio-physical indicators in CAPRI are currently calculated based on two approaches. The more standard one is to borrow emission factors from the literature, such as those delivered by the Intergovernmental Panel on Climate Change (IPCC), and combine them with production activity coefficients. tion processes is linked to cattle production intensity and specific engineering parameters such as digestibility or energy consumption. The second approach consists of a formal link to the biophysical model DNDC (Li et al. 1994). It allows a far more detailed description of water and nutrient flows compared to linear activity approach and it is further detailed below. In SEAMLESS-IF agricultural technologies and externalities are jointly modelled by the APES-FSSIM modelling chain. This approach potentially provides much more detailed information and the ability to represent complex dynamic processes at very low scale and allows for flexible aggregation possibilities. However, the approach is also very demanding regarding data and engineering knowledge. The data requirements and work load to parameterize and calibrate the bio-physical and bio-economic models were the major reasons for the system developers to implement a representative instead of a fully covering database and APES-FSSIM modelling chain.

11 A Comparison of CAPRI and SEAMLESS-IF as Integrated Modelling Systems

267

Modelling of Policies at Farm Level The policy coverage in CAPRI is clearly linked to the layout of the different components. The regional supply models are able to implement to a large extent the different premium schemes under the first pillar of the CAP as well as production quotas and set-aside regimes. Specific consideration was given to the integration of the different implementation options for the so-called ‘Single Farm Payment’ introduced in 2003. Premium ceilings, in values or quantities, were introduced by the CAP for budgetary reasons. The policy module of CAPRI tries to mimic this legislation and carries out the necessary adjustment when premium ceilings are exceeded. At the current stage, payments from the so-called second pillar of the CAP, comprising rural development measures and including the agri-environmental schemes, are not implemented in CAPRI, both for data availability reasons and for missing interfaces in the regional programming models. In SEAMLESS-IF, agri-environmental payments and other farm level schemes under the second pillar of the CAP are explicitly considered through the FSSIM farm-type layer. They can be flexibly introduced by the experiment designer in a farm-type specific manner. However, the impact assessment of these schemes is currently restricted to regional case study analysis. An extrapolation to other regions to allow market level analysis is not possible due to the limited data availability but also due to the regional diversity of these measures in terms of design and implementation.

Agri-environmental Indicators CAPRI, although focusing on the regional analysis of the agricultural sector at the NUTS 2 resolution, has been developed from the beginning also as a tool to analyse agri-environmental interactions (Britz et al. 2003). Pan-European and full sectoral coverage were deemed highly relevant features for environmental analysis. This led to a rather well developed description of nutrient flows in agriculture based on emission factors, and later, to the replication of the IPCC greenhouse gas accounting guidelines for the agricultural sector (Pérez Domínguez 2006). Moreover, CAPRI was also used to investigate abatement options for ammonia emissions in a combined application with the MITERRA and RAINS models (Oenema et al. 2007). An obvious drawback of CAPRI is the fact that it only allows limited endogenous adjustments at the level of specific intensity of single activities, which somewhat restricts its abilities to model effects of emission taxes or management standards. Here, SEAMLESS-IF is able to explicitly represent crop growth, nutrient flows and water-related environmental externalities through the combined use of the APES and FSSIM models (see Chapters 4 and 5). This leads to the possibility of generating regional specific alternative activities not currently used in farm management. These activities may extend the choice set for the analysis of changed incentives for management practices at farm level.

268

W. Britz et al.

The vertical link from CAPRI to bio-physical simulation models and dis-aggregation below the administrative regional level received more attention in the CAPRIDynaspat project (FP6 Project, 2004–2007). These components of CAPRI are not part of SEAMCAP since APES and FSSIM provide a bottom-up approach to assess environmental externalities. To a large extent, this link applies a top-down approach, where only limited bio-physical interactions are described in the regional supply models but rather post-model analysis is used to capture impacts on the environment. Given the non-linearities of bio-physical models, especially with respect to soil parameters, and the need to analyse environmental impacts in a spatial context, a methodology for spatial down-scaling from NUTS 2 level to about 150,000 clusters of 1 × 1 km grid cells covering the EU-27 was developed (Leip et al. 2008). This spatial layer covers all crops in CAPRI, animal stocking densities, crop yields and organic and mineral fertilizer application rates. As full simulations of all crop-site-farm management combination with bio-physical models are too time consuming, a statistical response surface from the bio-physical model DNDC was estimated. It comprised, as explanatory variables soil and climate parameters, mineral and organic fertilizer application and the so-called potential yield, a crop parameter typically used for calibration purposes of DNDC (Britz and Leip 2008). By inverting the regression function so that it allows defining the potential yield to be equal to the observed one, the statistical model is able to reproduce the results of CAPRI and provide biophysical results consistent with the quantity framework provided by the economic model. Following the general CAPRI principle, also this downscaling approach covers all agricultural activities and the whole agricultural area of the EU.

Baseline Generation for Forward Looking Impact Assessment A further feature which CAPRI shares with other economic simulation models is the need for forward looking impact assessment. Agricultural policies require time to be developed and implemented, and their impact clearly depends on the state of the system at the implementation stage. As technical progress, population growth, changes in consumer behaviour and policies inside and outside the EU lead to strong changes in the agricultural and market systems, analysing impacts based on available and often outdated statistical data can provoke a serious bias. Accordingly, there is a need for tools to project a most-likely future state of the system or to offer a plausible range of future states. The projection capabilities of CAPRI, especially at regional scale, led to specific applications e.g. for the European Environmental Agency (Witzke et al. 2004) or for the prospects of agricultural markets and income of the Directorate General of Agriculture and Rural Development (DG-AGRI) 2007 (European Commission 2007a) or DG-AGRI’s Rural Development Outlook 2007 (European Commission 2007b). For the projection, CAPRI integrates a-priori information derived from trend analysis and knowledge from market experts at

11 A Comparison of CAPRI and SEAMLESS-IF as Integrated Modelling Systems

269

DG-AGRI and other studies in an estimation framework which mutually ensures compatibility of the different forecasted elements (such as crop areas and herd sizes, yields and other output coefficients, productions and different elements of product demand, see Britz et al. 2007).

Graphical User Interface The technical realization of CAPRI changed over the last decade, and is, after quite some changes, by now based on a Graphical User Interface (GUI) realised in JAVA and GAMS, without a separate DataBase Management System (DBMS). Interestingly, a policy editor was realized in a very early version, but later abandoned as it was hard to maintain in a period of rapidly evolving policy instruments. Instead, a much more complex policy description was directly introduced in GAMS in order to capture, as close as possible, the changing agricultural policy. This solution clearly allows for a greater degree of flexibility, but in-depth knowledge about the representation of the various policy instruments in the code is required to successfully implement scenarios. The CAPRI GUI is mainly developed as a tool to support trained users in result exploitation and model application, including building up the database and model calibration. It could therefore be called an expert GUI. To exploit results, data and model results including those from indicator calculators are stored in one possible very large, sparse multi-dimensional cube. Result tables (such as prices, market balances, and environmental indicators), maps or graphs are generated as reports (or “views”) from these cubes. Each view applies specific filters and shows the resulting extraction in a specific presentation (table, graph or map). Views are linked to each other, allowing for a kind of drill-down approach by going from more general or aggregate results to more detailed ones. General tables of aggregated indicators give an overview of the main scenario impacts, and can be directly used for reports. Highly detailed technical tables show relevant information for a more refined analysis of the results or for modellers to use during the model debugging process (e.g. convergence statistics or dual values of the optimisation models for each activity and region). The SEAMLESS-IF GUI is more oriented towards policy assessment and supports the user in all stages of the policy assessment (from the definition of a storyline to the analysis of model results by means of exploitation tools). Moreover, SEAMLESS-IF manages a wider range of components compared to CAPRI and supports the integrative modeller in scenario development and selection of the model chain (pre-modelling-phase). The policy expert is given less functionality and is restricted mainly to view results (post-modelling phase). The exploitation tools are based on client-server solutions providing a limited number of visualisation possibilities for preselected indicators and model variables (see Chapter 9).

270

W. Britz et al.

Looking into the Future The different operational stage of the two modelling systems, and the fact that SEAMLESS-IF comprises SEAMCAP as a specific version of CAPRI, renders a comprehensive evaluation difficult. Over the last decade, CAPRI was able to provide analytical support to different EU policy reform steps based on quantitative impact assessment, while at the same time it was expanded, revised and improved. However, its limited modularity has led to a rather large and complex system, so that only a few experienced modellers are able to exploit the full potential of the system or to successfully change components. A key success factor for CAPRI, as for many other model networks, has been human capital in the sense that several key persons continued working with the system over years. Important for building up the network has been a sequence of 3-year EU framework projects (CAPRI, CAP-STRAT, CAPRI-Dynaspat) centred on CAPRI development and application, co-ordinated by the core development team. The network evolved almost naturally during the gradual build up and expansion of the system over a longer period as the projects around CAPRI allowed training graduate students in using, modifying and expanding the system. The increasing demand for agricultural policy impact assessment allowed several of the graduate students trained in the EU framework projects to continue working as post-docs with CAPRI at different institutions, being the basis for the successful application of CAPRI in policy impact assessment studies. Such studies with their typically tight deadlines require more senior staff, well experienced in applying the system, and are thus less suited to provide training for graduate students. New teams and individuals have been attracted to CAPRI by yearly training sessions. The sessions discussed the general concept of CAPRI, communicated methodological and technical changes and conducted hands-on training on application. Building up the network was certainly supported by the growing market for this type of analysis, which kept competition in the network low. At the same time, the increased demand for integrative impact assessment and application of model families made it attractive for research institutions to take CAPRI into their portfolio of tools. From the technical side, the CAPRI model chain uses GAMS (General Algebraic Modeling System) as the only modelling software for all major components, including those dealing with database updates and model parameterization, which certainly reduces maintenance and development costs. However, this has led to a less than perfect implementation in technical terms, for example regarding model structure or documentation of the code. Moreover, GAMS does not support well the development of re-usable components, since basically any object has a global scope and the passing of arguments, as usual for other programming languages, is not supported, hindering encapsulation of components. From the human resource perspective, the discussion on how to come to a fair distribution of maintenance costs of the system is far from finalized. These costs are discouraging when applying for typical research projects which put a focus on further methodological improvements. Moreover, short-time policy impact studies

11 A Comparison of CAPRI and SEAMLESS-IF as Integrated Modelling Systems

271

are not a suitable frame for maintenance work and database updates. This problem might require in the medium to long term a different system of organisation, where access to the model is coupled to the contribution of maintenance or financial funds. With this objective in mind, a CAPRI consortium agreement is currently under discussion amongst research institutions having participated in previous Framework Programme projects and studies involving the CAPRI model. Such an agreement might strengthen the club good character of CAPRI, i.e. one that is characterized by non-rival use, but with the possibility of exclusion allowing to share the costs among users based on private contracts. It remains to be seen if this organisation together with the engagement of the European Commission in the co-development and maintenance of the system will be able to secure the survival of CAPRI in the years to come. Currently, the Institute for Prospective Technological Studies (Joint Research Centre, European Commission) is launching an integrated modelling platform hosting the CAPRI model amongst other well-established agricultural models (see Pérez Domínguez et al. 2008). SEAMLESS faces, compared to CAPRI, somewhat different challenges. At the current stage, little experience can be gained with respect to its future applications, maintenance or adaptation. SEAMLESS-IF faces two important challenges. The first one relates to finding suitable policy questions where clients favour its application over other existing modelling frameworks. CAPRI succeeded in entering the impact assessment market due to its ability to deliver regional impacts of the CAP. Especially, in the field of agri-environmental interactions, the rich layer of bio-physical simulation models represented by the APES-FSSIM model interaction may provide a comparative advantage for SEAMLESS-IF. The second important challenge relates to ensuring maintenance of the system, which requires some continuity in key personnel beyond the current project phase. The maintenance question for SEAMLESS-IF is somewhat different from CAPRI. CAPRI has a rather monolithic structure and does not aim at further modularity. Moreover, little but highly specialized staff is needed for its survival. In the case of SEAMLESS-IF, the key for success will be to find the right balance between modularity (i.e. which components can find an attractive market niche for their further development), complexity (i.e. degree of separation between models, data and IT infrastructure) and attractiveness (i.e. involving research questions interesting for future young researchers). These elements are considered within the current SEAMLESS business plan, which includes a plan for establishing a SEAMLESS association responsible for maintenance of the system. Compared to CAPRI, SEAMLESS-IF offers a far richer set of model components which allow the analysis of policy impacts at the farm type level, their compatibility with the current institutional framework (see Chapter 3) and their effects on the landscape, all within a consistent overall scenario. Most components within SEAMLESS-IF have been developed as stand-alone components, putting emphasis on a clear documentation of required inputs and provided outputs. SEAMLESS-IF may hence address policy questions which cannot be answered by CAPRI and allows for a more tailored analysis. Nevertheless, the simulation behaviour of the APES-FSSIM-EXPAMOD-SEAMCAP model chain remains to be tested;

272

W. Britz et al.

such testing will also reveal how it compares to the full CAPRI stand-alone version, with its econometrically estimated or exogenously defined parameters. Equally, it will be of interest scientifically to compare the bottom-up approach from APES-FSSIM with the top-down down-scaling approach in the CAPRI-Dynaspat to assess environmental externalities.

Conclusions Both SEAMLESS-IF and CAPRI acknowledge that model-chains, linkages across scales and between economic and bio-physical components, are state-of-the-art and necessary attributes of tools for policy impact analysis of the new millennium to respond to societal information needs for policy design. In this chapter, we highlight some important differences and similarities between the two approaches. SEAMLESS-IF has placed a strong focus on developing a flexible framework for hosting and linking different components. It remains to be seen to what extent this flexibility will be exploited in the years to come. Equally, SEAMLESS-IF concentrated on the development and improvement of farm level components, and of establishing interfaces to components at higher scales. The focus on farm level is on bio-economic interrelations and technological adjustments. Exploitation in SEAMLESS-IF is seen from a strong client perspective: client server implementation, a limited but well defined set of indicators, and easy-to-use interfaces. CAPRI’s development is strongly driven by the focus on simulation changes in the CAP and market conditions, so that improving simulation behaviour received much attention. Equally, a decision on the most appropriate components in a model chain for these kinds of scenarios was taken very early, and the components of the chain then developed in an integrated way such that they are almost inseparable by now – which is also one reason that CAPRI is treated as one component in SEAMLESS-IF. The client-user model in CAPRI assumes that model results are mostly exploited by modellers and forwarded to the client in an appropriate format, along with explanatory notes and potential policy recommendations through a combined effort of modellers and policy consultants. Therefore, CAPRI places an emphasis on flexibility in exploitation and access to almost all model results on different scales. Consequently, the whole modelling system including exploitation tools is technically completely client based. Access to data and code (changes) is managed via a server-based versioning system.

References Adenaeuer, M. (2005). Modelling the European sugar sector - Incentives to supply sugar beets and analysis of reform. Options. Dissertation, Bonn University, Germany. Armington, P. S. (1969). A theory of demand for products distinguished by place of production. IMF Staff Papers, 16, 159–178.

11 A Comparison of CAPRI and SEAMLESS-IF as Integrated Modelling Systems

273

Bauer, S. (1978). Quantitative Sektoranalyse als Entscheidungshilfe für die Agrarpolitik - Ein dynamisches Analyse- und Prognosesystem für die Landwirtschaft in der Bundesrepublik Deutschland. Dissertation, University of Bonn, Germany. Bezlepkina, I., Pérez Domínguez, I., Heckelei, T., Romstad, E., & Oude Lansink, A. (2007). EXPAMOD: Component to statistically extrapolate from FSSIM models to other farm types and regions including aggregation to NUTS2: Motivation, description and prototype. PD3.6.11.2 (SEAMLESS Integrated Project, EU 6th Framework Programme, contract no. 010036-2). Britz, W., & Heckelei, T. (2000a). Positive mathematical programming with multiple data points: A cross-sectional estimation procedure. Cahiers d’Economie et Sociologie Rurales, 57(4), 28–50. Britz, W., & Heckelei, T. (2000b, October). Effekte von Agenda 2000 auf den deutschen und französischen Agrarsektor - Anwendung des regionalisierten Modellsystems CAPRI. Paper presented at Gemeinsame Tagung von SFER und Gewisola, Strasbourg. Britz, W., Heckelei, T., & Kempen, M. (Eds.) (2007). Description of the CAPRI modelling system. Final report of the CAPRI-Dynaspat Project. Bonn, Germany: Institute for Food and Resource Economics, University of Bonn. Britz, W., Heckelei, T., & Pérez Domínguez, I. (2006). Land use effects of decoupling: An EU wide, regionally differentiated analysis. Agrarwirtschaft, 55, 215–226. Britz, W., Helming, J., & Pérez Domínguez, I. (2003). Development of models and tools for assessing the environmental impact of agricultural policies. Final report for DG-ENV. Britz, W., & Jacquet, F. (2006). EU-MED Agpol: Impacts of Agricultural Trade Liberalisation between the EU and Mediterranean countries. Proceedings of the Workshop on Euro-Med Association Agreements Agricultural Trade – Regional Impacts in the EU, Jointly organised by DG RTD and DG JRC with the collaboration of DG AGRI, Brussels on 14 February 2006. Britz, W., & Leip A. (2008, February). A statistical meta model of DNDC to estimate nitrogen fate and irrigation water needs at 1 x 1 km grid at Pan-European scale. Paper presented NEU-annual meeting PSG2, Ghent. European Commission. (2004). Prospects for agricultural markets 2003-2010. Retrieved January 9, 2004 from http://europa.eu.int/comm/agriculture/publi/archive/index_en.htm European Commission. (2007a). Prospects for agricultural markets and income in the European Union 2006–2013. Directorate-General Agriculture and Rural Development, Brussels. Retrieved February, 2007 from http://ec.europa.eu/agriculture/publi/caprep/prospects2006b/ index_en.htm European Commission. (2007b). Rural Development in the European Union - Statistical and Economic Information – Report 2007. Retrieved February 2007, from http://ec.europa.eu/ agriculture/agrista/rurdev2007/index_en.htm FAO. (2003). World agriculture: Towards 2015/2030. Summary report. Rome: Food and Agriculture Organization of the United Nations. Frohberg, K., & Britz, W. (1994). The World Food Model and an assessment of the impact of the GATT agreement on agriculture. Research Report. Bonn, Germany: Institute for Agricultural Policy. Funtowicz, S. O., & Ravetz, J. R. (1994). The worth of a songbird: Ecological economics as a post-normal science. Ecological Economics, 10, 197–207. Heckelei, T., & Britz, W. (2005, February). Models based on positive mathematical programming: State of the art and further extensions. Invited paper presented at the 89th EAAE Symposium, Parma, Italy. Heckelei, T., & Wolff, H. (2003). Estimation of constrained optimisation models for agricultural supply analysis based on generalised maximum entropy. European Review of Agricultural Economics, 30, 27–50. Henrichsmeyer, W., Cypris, C., Löhe, W., Meudt, M., Sander, R., von Sothen, F., Isermeyer, F., Schefski, A., Schleef, K.-H., Neander, E., Fasterding, F., Helmcke, B., Neumann, M., Nieberg, H., Manegold, D., & Meier, T. (1996). Entwicklung eines gesamtdeutschen Agrarsektormodells RAUMIS96. Endbericht zum Kooperationsprojekt. Forschungsbericht für das BML (94 HS 021). Bonn/Braunschweig, Germany.

274

W. Britz et al.

Hertel, T. W. (ed). (1997). Global trade analysis: Modeling and applications. New York: Cambridge University Press. Jansson, T. (2007). Econometric specification of constrained optimization models. Dissertation, University of Bonn, Germany. Junker, F., Britz, W., Heckelei, T., Pérez Domínguez, I., & Wieck, C. (2005, June). How sustainable is the latest CAP Reform under the possible trade liberalisation outcomes of the Doha Round? Paper presented at the IATRC Summer Meeting 2005, Sevilla, June 16/17: ‘Pressures for Agricultural Reform: WTO Panels and the Doha Round Negotiations’. Kanellopoulos, A., Berentsen, P., van Ittersum, M.K., & Oude Lansink, A. (2007). Assessing the forecasting capacity of a bio-economic farm model calibrated with different PMP variants. Farming Systems Design 2007: An International Symposium on Methodologies for Integrated Analysis of Farm Production Systems Farm-Regional Scale Design and Improvement (pp. 55–56). Gorgonzola (MI), Italy: Global Print. Kempen, M., & Kränzlein, T. (2008, January). Energy use in agriculture: A modelling approach to evaluate energy reduction policies. Paper prepared for presentation at the 107th EAAE Seminar, Sevilla. Retrieved from http://ageconsearch.umn.edu/bitstream/6652/2/cp08ke15.pdf Leip, A., Marchi, G., Koeble, R., Kempen, M., Li, C., & Britz, W. (2008). Linking an economic model for European agriculture with a mechanistic model to estimate nitrogen losses from cropland soil in Europe. Biogeosciences, 5, 73–94. Li, C., Frolking, S., & Harriss, R. (1994). Modelling carbon biogeochemistry in agricultural soils. Global Biogeochemical Cycles, 8(3), 237–254. Oenema, O., Oudendag, D., Witzke, P., Monteny, G.J., Velthof, G.L., Pietrzak, S., Pinto, M., Britz, W., Schwaiger, E., Erisman, J.W., de Vries, W., van Grinsven, J.J.M., & Sutton, M. (2007). Service contract: Integrated measures in agriculture to reduce ammonia emissions. Final report. (070501/2005/422822/MAR/C1). Wageningen, The Netherlands: Alterra. Pérez Domínguez, I. (2006). Greenhouse gases: Inventories, abatement costs and markets for emission permits in European agriculture – A modelling approach. Dissertation, University of Bonn, Germany. Pérez Domínguez, I., Bezlepkina, I., Heckelei, T., Oude Lansink, A., Romstad, E., & Kannellopoulos, A. (2009). Capturing market impacts of farm level policies: a statistical extrapolation approach using biophysical characteristics and farm resources. Environmental Science and Policy, 12(5) 2009, 588–600. Pérez Domínguez, I., Gay, S. H., & M’Barek, R. (2008). An integrated model platform for the economic assessment of agricultural policies in the European Union. Agrarwirtschaft, 57(8), 379–385. Van Ittersum, M. K., Ewert, F., Heckelei, T., Wery, J., Alkan Olsson, J., Andersen, E., et al. (2008). Integrated assessment of agricultural systems – A component-based framework for the European Union (SEAMLESS). Agricultural Systems, 96, 150–165. Van Tongeren, F., van Meijl, H., & Surry, Y. (2001). Global models applied to agricultural and trade policies: a review and assessment. Agricultural Economics, 26(2), 149–172. Witzke, H.P., Britz, W., & Kuhn, A. (2004). Outlook on selected agricultural variables for the 2005 State of the Environment and the Outlook Report of the European Environmental Agency. Eurocare Final Report to EEA.

Chapter 12

Science–Policy Interfaces in Impact Assessment Procedures Ann-Katrin Bäcklund, Jean Paul Bousset, Sara Brogaard, Catherine Macombe, Marie Taverne, and Martin van Ittersum

Introduction Science and politics serve different purposes. Policy commonly refers to an institutional choice of values (Lee 1993, 2006). Policy options are ways (actions) to favour some values and disfavour others. A policy development process is driven by forces that can be characterised using three dimensions – salience, credibility and legitimacy (Cash and Buizer 2005). Salience relates to the relevance of the policy goals (issue at stake). Credibility addresses the technical quality of the knowledge and information used (validity and accuracy). Legitimacy concerns the society’s interest in the policy. Science on the other hand can be seen as the sum of knowledge produced by the application of systematic methods of inquiry – including experience or learning from actors and institutions, each guided by different epistemologies (what can be known) and different ontologies (what is knowable) (Feynmann 1998). A.-K. Bäcklund (*) Department of Social and Economic Geography, Lund University, Sölvegatan, 12, S-223 61, Lund, Sweden e-mail: [email protected] J.P. Bousset and M. Taverne Dynamics and Functions of Rural Areas, Cemagref, 24, avenue des Landais, BP 50085, 63172, Aubière Cedex, France e-mail: [email protected]; [email protected] S. Brogaard Centre for Sustainability Studies, Lund University, Box 170, S-221 00, Lund, Sweden e-mail: [email protected] C. Macombe Cemagref, 24, avenue des Landais, BP 50085, 63172, Aubière Cedex, France e-mail: [email protected] M. van Ittersum Plant Production Systems Group, Wageningen University, P.O. Box 430, 6700 AK, Wageningen, The Netherlands e-mail: [email protected]

F.M. Brouwer and M. van Ittersum (eds.), Environmental and Agricultural Modelling: Integrated Approaches for Policy Impact Assessment, DOI 10.1007/978-90-481-3619-3_12, © Springer Science+Business Media B.V. 2010

275

276

A.-K. Bäcklund et al.

Knowledge can be characterised by the three dimensions – acquisition, application and ethics. Acquisition relates to what has been found out using enquiry methods. Application relates to how the findings can be put into practice (i.e. technology). Ethics relates to the right and wrong ways to find out new things or to use technologies. But doing science is influenced by the societal norms, which determines its legitimacy. And making a tool to support impact assessment is more than putting science and politics together. Hence, a modelling system like the SEAMLESS Integrated Framework (SEAMLESS-IF) should not be regarded as a strictly scientific/technical application, but as a tool for communication between science and policy. Such interaction is a discernible feature of impact assessment working procedures. It is therefore logical that interaction starts already with potential users, which can contribute to the development of the tool. SEAMLESS is an integrated framework that enables simulation of effects of agricultural and rural development policies and innovations, with a strong emphasis on communication between scientists and stakeholders (Van Ittersum et al. 2008a). This means that impact assessments made by SEAMLESS-IF can be understood as social processes, including but not restricted to their formal end products. A social process perspective directs attention beyond the content of assessment reports and encompasses questions regarding participation, presentation, evaluation, and how the boundaries between the scientific and political dimensions of the assessment are negotiated and legitimized (Hoppe 2005). A deliberative process is one in which values and facts are constructed within a process where ideas are reciprocally confronted with one another as a means for learning. In order for a deliberative process to be scientifically credible, it must consider all forms of knowledge – science, expertise, local and indigenous. It must also ensure the ‘accountability’ of each form of knowledge before relying upon it as a foundation for political choices. To make the process work there must be full and open access to knowledge by all participants. There must also be full transparency in how knowledge is created and used and how values are discovered and applied. In doing so, deliberative processes may address the twin problems of uncertainty and ambiguity. This is done by institutionalizing two basic principles of organizational learning – adaptation through experimentation and iterativeness through continuous monitoring, evaluation, and change (Casey 2005). Following Habermas, we could state that SEAMLESS-IF can contribute to a deliberative impact assessment process, provided that the following conditions are met: (1) The values of the organisation requesting an assessment meet those of the SEAMLESS-IF. (2) The problem definition can be debated between involved parties. (3) The outcome of the assessments allows for a real debate with affected stakeholders. The third condition will normally not be in the hands of the “model owners” but controlled by the policy agencies that are using the SEAMLESS framework. To match the first condition, we compare if the policy experts’ assumptions about science-policy nexus, are compatible with the hypotheses set in SEAMLESS-IF. To meet with the second one, we illustrate how the problem definition created a real debate between policy experts and integrative modellers in the regional test cases.

12 Science–Policy Interfaces in Impact Assessment Procedures

277

And hence in which ways it will be possible for SEAMLESS-IF to contribute to an institutionalisation of a deliberative impact assessment process? This chapter presents the features of the impact assessment system now being implemented in Europe. We will then present SEAMLESS-IF as a science-policy interface, and report on the ways scientific knowledge and policy values were confronted and deliberated in discussions with potential users and in test applications of the SEAMLESS-IF.

Motive and Aim of the European Impact Assessment System To perform impact assessments of policy proposals is gradually becoming an important instrument in European policy making. Since 2003 a formal impact assessment (IA) is required for all regulatory proposals in the EC (EC 2002). The initiative can be traced back to the Lisbon Strategy of 2000, where the European Union set the goal of becoming the most competitive and dynamic knowledge-based economy in the world. In its endeavour to achieve this goal a core priority is to implement a better law making process in the union and the member states. One way to achieve a better knowledge base for new regulations is to submit policy proposals for impact assessment (EC 2006).

Impact Assessment as Part of the Dynamics of Policy The IA procedure at the European Commission stems from a governance concept which assumes that policy programs are the product of complex interaction between government and non-government organizations (researchers included), each seeking to influence the collectively binding decisions that have consequences for their interest. The idea of impact assessment is based on the assumption of the ‘co-production of knowledge’ (Callon 1999). Different kinds of knowledge are negotiating their hybridization, which is a necessity to get forward in the management of the risk-complexity-uncertainty of the implementation of policies. An assessment process that meets the ideal requirement is one that involves stakeholders throughout the work in a major political process, so that the suggestions put forward in the final proposal are anchored on all levels in the European community (Bäcklund 2009). Stakeholders can be consulted about different elements of the IA; the nature of the problem, policy options, impacts, etc. (EC 2005: 9). Assessments are likely to be more thorough and also include sustainability aspects when they concern complex issues without any simple policy solutions. Such situations ‘open up windows and pressures for significant policy change, as well as a demand for new sources of knowledge’ (Turnpenny 2008).1 For an example see the assessment of the Thematic Strategy for Soil Protection (Van-Camp et al. 2004).

1

278

A.-K. Bäcklund et al.

The possibility is also given to wait with consultation until there is a full draft proposal, only applying EU’s minimum standards for consultation. Although the latter policy procedure has been the most frequently used option, at some stage the results will always be subject to public examination. This examination will by nature be politically as opposed to scientifically motivated. Introduction of the IA instrument in European policy making is a helpful step to make economic development more knowledge based and rational but economic development is nonetheless always a struggle between conflicting interests. We are considering here the ‘adversarial model’ by Hoppe (2005), where politics is a non violent power struggle between political parties and/or organized interest groups, that lead to temporary compromises of the public interest (Lindblom 1968). The EC and its administrations are under heavy pressure from influential spheres of power. Also scientists and the use of scientific modelling tools will be subordinate to the political interplay, in which also scientific or other groups contributing to IAs stand a risk of being targeted by a political critique (Bäcklund 2009). The design of the assessment as well as the outcome can be severely criticised on both factual and political grounds (see, e.g. EEAC 2006: 17). Hence, there is a need for scientists to be aware of the political process in which they participate when consulted. When the Commission proposes a new regulation the DG’s argumentation for the policy recommendation made in the proposal has to come across convincingly to decision makers and the public. As it has to be explained how the assessment results support the policy option in the proposal, officers want simplicity and transparency in the modelling process. The models and the assumptions made have to be understood by the policy experts. If it is not comprehensible and transparent it is not politically useful for the administration (Bäcklund 2009). The need for simplification might be a source of conflict between the ambitions of a scientific modeller and a policy expert. The core of this question concerns a ‘cultural’ difference between researchers and practitioners’ way of approaching a problem – especially practitioners in highly politicised administrations. The interactions between these cultures are therefore a main stake when developing and applying a system like SEAMLESS-IF.

IA as a Learning Process The aim of introducing the IA procedure is wider than to merely provide a knowledge base for decision making. IA is introduced as a rational and better way to political decision making but also as a tool to improve internal communication in the Union and to restore an anticipated lack of confidence in European governance. The great aspirations placed on the IA system as a way to achieve increased communication and unity in European politics, to promote mutual learning between EU institutions and Member States and help in “restoration of confidence in government” are expressed in the Mandelkern report preceding the final IA proposal (Mandelkern Group on Better Regulation, 2001).

12 Science–Policy Interfaces in Impact Assessment Procedures

279

The political system in the EU is not based on a voting democracy but on a bargaining system in combination with deliberative democratic principles. Deliberative can be described as ‘collective decision making with the participation of all who will be affected by the decision or their representatives’ in combination with a ‘decision making by means of arguments offered by and to participants who are committed to the values of rationality and impartiality’ (Elster 1998: 8). A prerequisite for a true deliberative process to occur is that it emerges under the conditions of ‘free public reasoning among equals who are governed by the decisions’ (Cohen 1998: 186). The underlying science-policy nexus model is a model of ‘pure learning’. It assumes that scientists and policy-makers cooperate through shared concepts and strategies of innovation. It treats the policy process as a sort of research process, where the policy programme is viewed as a set of hypotheses about the causal links between acts and a desirable future state of affairs. And further putting policy into practice can be seen as a case of social experimentation (Hoppe 2005: 211). To ensure that the assessment work performed also serves its bargaining and deliberative purpose, consultations are made an integral part of the EU’s IA procedures so that it at the same time makes decision-makers and the public aware of likely policy impacts and serves as a tool for communication between them (EC 2002: 3).

Conceptualizing Science–Policy Interactions in SEAMLESS-IF Application SEAMLESS-IF is designed with a set of procedures and computer tools supplied by scientists, which can provide ‘information’ to policy makers, in order to facilitate the assessment of likely short term impacts of a suggested policy, and the exploration of alternative strategies for achieving longer-term goals through influencing land management decisions. The SEAMLESS-IF impact assessment procedures can be viewed as applications of a specific technology – encapsulation of specific enquiry methods (simulation models) and data – for the acquisition of new knowledge. When searching for the domain of intersection between science and policy, we shall first look at the SEAMLESS-IF impact assessment procedure (technology) and then the outputs of the procedures (new knowledge).

The SEAMLESS-IF Impact Assessment Procedure The SEAMLESS-IF impact assessment procedure includes three main phases: pre-modelling, modelling and post-modelling (Therond et al. 2009). During the pre-modelling phase the issue at stake and the associated spatial and temporal scales are defined. The future driving forces at a given time horizon are anticipated

280

A.-K. Bäcklund et al.

and indicators are selected. The choices and selections made during this phase will express the interests, values and motivations of the organisation that has initiated the assessment of a policy or a technological change. The following example shows how policy options can steer an assessment in different directions. The objective of the policy to be tested could be to increase oil-producing crops. In order to achieve this goal a specific group wants to support biofuel. To promote this, the group needs arguments in favour of biofuel. So, the group suggests that new cropping systems should be tested and assessed on an indicator of biofuel energy balance. The group could also be interested in assessing the global performances of cropping systems that include oil-producing crops in comparison to cropping systems without such crops. The group might further like to get ideas about how oil crops could be introduced in new regions, for example at which price level these crops would be accepted among producers. More thought would be needed to suggest changes in policies or practices to assess this question. The main concern for the group is the location of oil-producing crops and the political and economic elements that would favour or disadvantage their introduction as these two factors would have implications on the whole supply chain. During the next step, the modelling phase, the appropriate model chain is used with the relevant data. A consistent micro-macro analysis involves a bio-economic farm model (FSSIM) and an agricultural sector model (CAPRI). The FSSIM farm models are run for the major farm types in a sample of regions within the EU. These farm models are run for a range of prices of the commodities modelled with FSSIM and CAPRI. This results in modelled supplies of commodities for these farm types and regions at a certain price level. The price-supply relationships are then extrapolated to the farm types in regions for which no FSSIM model was run, using the econometric model EXPAMOD. Then the supply models in CAPRI are calibrated to these supply responses allowing derivation of commodity market prices consistent with the farm level behaviour in CAPRI. The prices simulated with CAPRI will then be used for a final run of FSSIM, to simulate the farm behaviour (Van Ittersum et al. 2008a). During the post-modelling phase the impacts of tested policy options and technological changes are analysed and explored, including their institutional compatibility. The post-modelling phase may also include the transformation of the indicators into indices for the comparison between various policy options. Finally, the various results, newly constructed indicators and models are stored for future retrieval and reuse, in a knowledge base, together with annotations, and other process related documents. Both the pre- and post-modelling phases require comprehensive interaction between the two types of actors: policy experts and integrative modellers. Policy experts express the problem to be assessed according to their interest and the political role of their organisation, and thus influence the assessment (definition of problem, policy options, indicators) in a way that supports their stake. Integrative modellers run the SEAMLESS framework. As providers of expert knowledge, data and analysis, they feed scientific knowledge into the policy process, based on their values. This makes that the SEAMLESS-IF procedures can be regarded as a plan for institutionalization of ‘deliberation’.

12 Science–Policy Interfaces in Impact Assessment Procedures

281

The term ‘deliberation’ means that values and facts are constructed within a participatory practice, where they are moulded in a process of mutual confrontation. In deliberative processes it is expected that people through interrogation will discover values – what is important to them, to others, and to society – and learn about their and society’s interests. It is this fundamental commitment to learning that characterizes the use of ‘confrontation’ and ‘interrogation’ among ideas and value. The institutionalization of ‘deliberation’ – i.e. the simultaneous interrogation of knowledge and values – can transform political decision-making (Smith 2000). This process is also described by Habermas as deliberative democracy (1996). Habermas distinguishes three models for the relation between science and policy, among which “the pragmatic model” recognises that “… there is interdependence between values and facts … and the strict separation between the functions of the expert and the politician is replaced by a critical inter-relation” (Van den Hove 2007). Drawing on Habermas’ classification of science-policy relations the SEAMLESS-IF procedure can be viewed as a pragmatic, process-oriented science-policy interface. The main reason for this stems from the fact that SEAMLESS-IF procedures recognise interdependence between values expressed by organisations and techniques to satisfy value-oriented needs. In SEAMLESS-IF procedures, the strict separation between the functions of scientists and policy makers is replaced by a critical inter-relation. In the assessment procedure, the choice of which scientific knowledge is required – i.e. which problem situations will be considered and which are left aside – involves dynamics that pertain to both the scientific and policy domains. The identification of the issue at stake, the choice of relevant disciplines, methodologies, scales, variables, and boundaries, and the strategies to articulate them are elements of the scientific process that are in no way isolated from the socio-political context. In other words, the framing of the question to be considered in the assessment involves value judgements and decisions about what will be considered and what not – from policy experts, and about who will be involved and how – from integrative modellers. Hence the first domain of intersection between science and policy in the SEAMLESS-IF procedure is a result of the fact that processes of selecting, framing and addressing a problem as well as the design of potential solutions pertain to both the scientific and the political spheres. The second domain of intersection between science and policy relates to the selection of models and indicators used to identify impacts of the policy to be tested. Value judgements are also embedded in decisions about which results will be used and how they are interpreted. The essential point here is that the SEAMLESS-IF selection process of models and indicators belongs both to the scientific process and to the policy process. Policy experts may influence the scientific validation process in the direction of interest and values of their organisations. Integrated modellers might feed scientific knowledge into the policy process in accordance with values of the scientific networks by which ‘scientists build up their cognitive authority’. Scientific networks may be used as ‘a powerful vehicle for channelling scientific input to policy-making’ (Van den Hove 2007: 813).

282

A.-K. Bäcklund et al.

The SEAMLESS-IF Outputs as Science–Policy Interface A SEAMLESS-IF integrated assessment can be used as input in an Impact Assessment which provides policy-experts with information about the likely impacts of a specific policy option. This information is provided by indicators and indices, which are tested against a baseline scenario and a selected policy scenario. The baseline scenario can e.g. be constituted by the likely impacts of the CAP reform as it is anticipated to be implemented in 2013 (or any other defined time horizon), including already decided future legislative changes. The scenario is the point of reference for the interpretation of policy effects, or various shocks coming from the ‘market level’ or the ‘farm level’ This makes that the resulting SEAMLESS-IF foresights can be understood as interfaces between ‘system science’ (procedures and computer tools for supporting impact assessment) and ‘social significance’ (the tested policy options and the selected indicators). The analysis made by SEAMLESS-IF can be regarded as ‘social foresights’ (economic, social and ecological needs of society, and means to satisfy these needs – amongst which are technological means and institutional considerations) and consequently the foresights can be understood as science-policy interfaces for the construction of ‘strategic knowledge’. Drawing on Grunwald’s (2004) definition there are three reasons to regard the SEAMLESS-IF foresights as strategic knowledge as they are: –– Context-sensitive combinations of explanatory knowledge about the tested policies/innovative technologies; –– Provisional and incomplete in their descriptive aspects due to the degree of uncertainty in given models and data; –– Non-verifiable in nature since they do not give a representation of an empirical reality. The foresights are an explorative type of knowledge. Their aim is not to forecast the future – which is assumed to be plural – but rather on bounding the uncertainty. By offering a basis for discussing policy impacts they can play as communication devices. Dealing with strategic knowledge for societal application requires a thorough reflection on the premises that are inherent in the integrated assessment models and on the sensitivity of the results to the hidden values and other assumptions of the used models (Morgan and Dowlatabadi 1996; Brugnach et al. 2007). It has been suggested (Van Ittersum et al. 2008b) that uncertainty shall be addressed in SEAMLESS-IF in four steps: identifying the sources of uncertainty in the selected model chain; communication of that uncertainty to the user; establish insight into the effects and/or relative importance of sources of uncertainty relevant to the users; communication of these effects to the users. Problems and challenges for the SEAMLESS-IF science-policy interfaces can be summarized as follows (Table 12.1). During the pre-modelling phase, two challenges must be addressed. As a result of making different selections during the process of problem definition some

12 Science–Policy Interfaces in Impact Assessment Procedures

283

Table 12.1 Problems/challenges for science–policy interfaces in SEAMLESS-IF Science–policy interface Problem/challenge Pre-modelling phase − Problem definition Selection of problem situations looked at/left aside − Resolution Technology (model-data) Post-modelling phase − Discussion Scientific networks as vehicles to channel scientific input to policy; policy issues’ influence on education of scientists − Balance between economic, environmental and social impacts − Strategic analysis − Completion of the database/complexity of indicators − Model uncertainty/sensitivity analysis of the outputs based on the premises of the models

issues will be tackled and others let aside. It is a main challenge to make this outcome evident for the policy makers. The “resolution” challenge relates to the outcome of an informed discussion between the policy maker and the researcher resulting in selection of the suitable data, choice of the right algorithms, and the relevant indicators. During the post-modelling phase, researchers face discussions where results can be contested. In fact, scientific researchers never act alone. They belong to institutions and to networks. Without always being fully aware of it, the researcher brings the ideas from his/her networks into the work. Through the models, hypotheses are introduced in the formulation of the question to be assessed. In turn, the policy makers influence the scientific education. During the strategic analysis other issues have to be addressed, such as how to balance economic, environmental and social impacts; how to select which outputs of the modelling, among all the available ones, shall be considered as ‘the results’; and how the potential causes of uncertainty shall be presented. In conclusion, the EC’s Impact Assessment Guidelines suggest science-policy relations where scientists and experts share their knowledge and capabilities. The design of the procedure and the outputs given by SEAMLESS-IF (as far as we may deem it) come close to that model. The following section reports on the ways scientific knowledge and policy values were deliberated during development of the framework and in test applications.

Case Studies: Establishing a Debate The two cases presented below report about interactions with potential users during development of the SEAMLESS-IF. The first case describes interaction with officers at the EC to explore their views on how the framework should be developed in order to suit assessment work in the DGs. The second case describes the interaction with representatives at regional level in France when setting up assessments in test situations.

284

A.-K. Bäcklund et al.

Case: Interacting with the EC Allocation of research funding is a way of manifesting political prioritisation and expanding knowledge in a certain direction. Given the importance attributed to impact assessment in European policy making it is logical that several large projects with the task to develop frameworks for modelling tools are presently funded by EU’s Framework Programme for Research and Technological Development (e.g. SEAMLESS, SENSOR, EFORWOOD).2 Development of such systems is clearly in line with the Commission’s aim to implement the impact assessment procedure in the EC and in the Member States. But also the way work in the funded projects is organised, is part of a political progression to spread the idea of impact assessment as a political tool. In the research proposals and in the intermittent evaluations of the projects the importance of user interaction with potential users at different political levels during the development of the system has been forcefully underlined. Dissemination and stakeholder interaction are regular requirements in research funded by the EC, but in the recommendations to the SEAMLESS project it has been advocated in quite a pronounced way. It was acknowledged already from the outset of the project that the success of the framework and its impact on EU policies depended on an efficient implementation and integration into user organisations. In the project proposal it was thus stipulated that the program should establish a permanent group of ‘advisors’ among potential ‘users’ that could help to develop the framework in – a process of ‘co-evolution’. Users were defined as organisations/individuals that potentially could use SEAMLESS-IF to support impact assessment processes. For mainly practical reasons the users were divided into prime users (being DGs like Agriculture and Environment) and other users like national policy making or implementation agencies and organisation advocating e.g. agricultural or environmental interests. Forming a Platform for Interaction As the aim was to develop a system that could serve the impact assessments work performed in the Commission, user interaction could very well have been arranged in such a way that the project from the start had been provided with a reference group of experienced assessment leaders in the EC. But this was not the case, instead the researchers in the project had to work their way into the DGs, to establish contact with officers that we identified as possible potential users, and evoke their interest for a modelling tool like SEAMLESS-IF. We wanted to establish a User Forum (UF), with permanent members that were willing to follow the project during its development of the framework. In search of the relevant contacts a large number of officers at different DGs of the Commission, SENSOR: Sustainability Impact Assessment Tools for Environmental Social and Economic Effects of Multifunctional Land Use in European Regions. EFORWOOD: Sustainability Impact Assessment of the Forestry-Wood Chain.

2

12 Science–Policy Interfaces in Impact Assessment Procedures

285

and in the European parliament, as well as at national and regional administrations were approached. We were in contact with more than 30 people by way of email correspondence, personal interviews and organised information meetings in both prime user and other organisations. As a result of this search process, information about the aim of the project and its background in the Commission’s Directive of Impact Assessment has been effectively spread by the researchers. The foregoing can be seen as a purely administrative process but it can also be described as an interactive learning process. While investigating practitioners’ interest in and potential need for modelling tools, the researchers have at the same time advocated the Commissions policy in its own administration, as well as in the European Parliament and in national administrations and interest organisations. The actors involved, the researchers and the officers in the administrations, have participated in a process that may be described as interactive learning. This search process could also be viewed as if researchers acted as lobbyists while anchoring the idea of ex-ante impact assessment in general and assessment by help of modelling tools in particular, among employees in the administrations. When finally the right group of suitable and interested EC officers was targeted it resulted in setting up a User Forum with people that were willing to work with us. There has been some shift of individuals during the years but several people have followed the forums from the start. The UF meetings have been regularly attended by a core group of five to ten representatives from DG Agriculture, DG Environment and DG Economy and Finances, EEA (European Environment Agency) and JRC (Joint Research Centre). It has not been possible to include a representative from the Secretariat General in the UF, which was unfortunate as they have the overall responsibility for the IA development in the EC. One of the learning experiences we made during the initialisation phase was that according to the ‘work culture’ in the services, the DGs preferred not to include representatives from organisations external to the Commission in this type of forum. Originally we intended to create a UF with mixed participation with representatives from the Commission services as well as other organizations such as farmers unions, environmental organisations and the OECD. The seven meetings that have been held have resulted in a knowledge exchange which has been beneficial for the development of the framework. A systematic review of the minutes from the meetings displays that the discussions have mainly revolved around three types of themes: (i) positive and negative feedback on the presented ideas concerning structure and content of the system; (ii) demands on technical performance or knowledge content of the tool; and (iii) strategic advice from the UF concerning best ways to approach and attract the interest of the DG’s and other users. Positive and Negative Feedback from a User Perspective The early phase of interaction focused on explaining the idea of the framework and its components in order to get feedback on its usefulness for the participants’ organizations. It became clear that an integrated framework for modelling policy

286

A.-K. Bäcklund et al.

effects clearly was of great interest to the organizations and that it was timely introduced as the need for knowledge based tools increased along with the growing importance of the IA procedure in the EC. Implementation of a system like SEAMLESS would imply a major change in procedure and quality compared to the present methods, in which answers to a greater extent are estimations based on expert judgments or on fragmented model analysis. Features that particularly appealed to the participants were the open construction of the system where linkages to other models can be made, and that the outcome can be presented with different visualization techniques. The importance of an open design of the framework is related to the anticipation that there will be more general and open questions asked in the future, like: What will happen if food security becomes more important? What happens in the EU if India and China develop into major food importers? Other reasons to keep the framework open would be to investigate for example issue related to water resources which requires integration of hydrological models or to incorporate models that are already operational in their organization. Parallel to the participants’ positive interest in the promising potential that a tool like SEAMLESS-IF would bring about, they demonstrated their role as practitioners concerning another question of openness, but this time openness was not seen as a benefit. From the very start the participants were concerned about not “wasting their time” on development of a tool if the continuation of “the product” and its administration was not secured. This is an unavoidable difference in prime focus between researcher and practitioners. For researchers the development of knowledge is meaningful as such. For practitioners it is meaningful only if knowledge can be administrated and applied. Theoretically both parties understand each others viewpoint, but it would not have become an issue in the discussion between them if the encounter between the two had not been arranged, i.e. the formation of a UF. The issue raised was, “What will happen with SEAMLESS-IF after 2008, when the research project has ended. Who will maintain the models and the data?” The research team was of course well aware of the necessity to provide an answer to this question but for obvious reasons it was impossible to give such guaranties before the development process was begun. The result of this issue being raised in the UF was that the process of finding forms for administration of the system, which was already envisioned in the project description, got an early start. This was also strongly endorsed by the Scientific Advisory Board of the project and independent peer reviewers. As a result the Board of the project developed a business plan, of which the establishment of a SEAMLESS Association was a cornerstone. The establishment of this association is now secured. The major aim of the Association is to maintain, extend and disseminate the major product of the project, i.e. SEAMLESS-IF and its models and data. New research and application projects feedback improvements and extensions to the Association and the Association makes latest versions of SEAMLESS-IF available.

12 Science–Policy Interfaces in Impact Assessment Procedures

287

Which Type of User Interface for Whom? Another important issue in the early stage of the project was the question of how the user interface should be designed. Before this could be decided it was necessary to establish more precisely what kind of user, with what kind of knowledge, should be able to navigate the system for which purpose. In the early discussions it was suggested by UF participants that they would like to have a simple tool on their computer to quickly answer burning questions. Such a tool would of course be very convenient to have in their daily work. It was however difficult to envisage that the system could ever be so user-friendly that any IA-leader in a DG would be able to set up and run a question in the modelling system and yet to provide a flexible framework for a diverse range of applications. The possibility to navigate through the system would be restricted to people that had knowledge in modelling. There is obviously a tendency to do more in-house work on the IAs and also an increasing number of people that have knowledge about modelling in the DGs, but the number of staff that would be capable to perform modelling will continue to be limited. A deeper discussion about this issue in the UF also revealed that even people with knowledge in modelling would hardly have the possibility to spend the time it would take to set up modelling runs. An assessment leader has to devote her or his time to initiate and lead different parts of the impact assessment process and will therefore not be able to personally perform in depth modelling and analysis of complicated issues. In case modelling is used it is usually outsourced. The UF discussion, complemented with information from other DG officers that were currently leading impact assessment projects, led to the decision that we should aim for a system designed for use at different levels. To set up and run new questions, there will be a need for an integrative modeller, i.e. a person that knows the system and its components. The policy experts in the administrations should be provided with an interface where they can alter parameters in already modeled policy issues to test alternative answers by help of minor changes in variables and to view and analyse results. Strategic Motives Behind Requests for Technical Performance and Knowledge Content We met a number of requests concerning technical performance of the tool. These requests were usually concrete and seemingly minor. It was for instance said that it would be beneficial if the tool could be connected to the EC Guidelines for Impact Assessment. Another specific request of this kind was to show the outcome of a modelling run in relation to a policy target, such as a red flag indicating numerical targets that were not achieved. Such requests are rather easy to meet. The former request about adoption to EC guidelines had great influence on the design of the end users interface which is built around the three stages in the integrated assessment process, pre modelling – modelling – post modelling.

288

A.-K. Bäcklund et al.

The latter request can be met through the visualization techniques are elaborated in the presentation part of the interface. The reason for the two requests in the examples above can be traced directly back to the everyday practical assessment work of the services. They are also easily understood and translated by the researchers. Other requests could have their background in more profound political dimensions of the IA work and were not so easily transformed into the tool, like the questions of transparency and uncertainty of results. The participants repeatedly expressed their concern for lack of transparency of modelling systems in general. The demand for transparency stems from a basic need for policy making institutions that their work gains legitimacy. Legitimacy is achieved when the production of knowledge, that a policy is based on, has been conducted in an unbiased way and has treated opposing interests in a fair manner. The modelling part of an assessment must not be suspected of taking over any political agenda. In the practical work situation the issue of transparency means that assessment leaders and other policy makers have to be able to defend and explain modelling results to a critical group of stakeholders. If the modelling system is a black box, where assumptions and other critical parameters are not clear and understandable the outcome might not be perceived as useful (salient) for the officers in charge. The demand for transparency is met with a comprehensive documentation of each of the components and the entire framework. Also the development of a joint ontology for concepts that are shared across components serves the aim of transparency and consistency (Janssen et al. 2009). Further, the need for transparency has triggered a large investment to develop graphical user interfaces for two user types, to define and document the specifications of impact assessment experiments with SEAMLESS-IF. We also learned that the issue of uncertainty of results was important. The closer the process gets to policy implementation the more important the aspect of uncertainty becomes. Information about uncertainty must be based on what users find important and the project therefore needed to get a deeper understanding of this issue. This experience encouraged the project to adopt a user-oriented approach towards uncertainty analysis (Gabbert et al. 2009). As part of this approach a questionnaire with the aim to capture the most important aspects of uncertainty was distributed to participants of the User Forum and other contacts at various DGs, JRC and EEA. The questionnaire revealed which items of uncertainty potential users find most relevant. Although answers from respondents differ, they do give some guidance towards concrete uncertainty analyses in comprehensive model chains such as SEAMLESS-IF. A full blown traditional and forward approach to uncertainty analysis is neither feasible nor effective; instead a focus on relevant sources and aspects of uncertainty as identified by users and scientists is opted for (ibid.). The political character of an IA process was also transformed into requests like the need for a very flexible selection of indicators. A situation where policy makers would be completely free to suggest their own indicators would be the optimal situation. The reason for this was that during the course of an IA, negotiations between several DGs and stakeholders are needed. If the selection of indicators is

12 Science–Policy Interfaces in Impact Assessment Procedures

289

too fixed this could hamper a flexible adaptation to the IA process. A totally flexible selection of indicators is not possible to achieve. Our suggestion for an indicator framework was seen as a useful compromise (Alkan Olsson et al. 2009). Concluding Remark on the UF Experience The interactions with the potential users in the Commission served the development process of SEAMLESS-IF with specifications of requirements and the development of a strategy for maintenance beyond the project’s lifetime. It led to the identification of the need for different user interfaces and the possibility to adopt these according to procedural requirements in the EC’s assessment work. Clearly, the degrees of freedom in a large research project with many partners with a funding agency that requests a clear and detailed project proposal before starting the project are somehow restricted once the projected has started. The interactions however confirmed from the outset that what was proposed and developed by the SEAMLESS project was matching their overall needs. Within the overall aims and methodology the interactions assisted in priority setting within the project and greatly triggered the timely thinking about continuity, maintenance, extension and dissemination. In this way, the two main results of the interactions, i.e. evaluation and specification of requirements and continuity, also serve each other. Requirements which cannot be satisfied (entirely) during the lifetime of the project are an important starting point for the work plan of the SEAMLESS Association.

Case: Interacting with Regional Organisations The application cases are derived from the testing stage of SEAMLESS-IF prototype 2, involving modellers from the project with knowledge about database structures, data formats, quality measures and model components and policy experts representing political or administrative bodies in France. The testing of the pre-modelling phase included selecting and framing of the issue at stake for the organisation involved as well as defining the policy measures to be tested and the experiments to assess these measures. Framing the issue included specifying the ultimate and intermediate goals of the policy to be tested in terms of impact indicators. Defining potential measures to be tested involved specification of changes in the environment of the target actors of the policy (economic instruments, regulation instruments, voluntary instruments) and/or changes in the behaviour of given actors. Designing experiments involved defining the spatial scale of the issue at stake (EU, nation, region, farm-type, etc.), the temporal scale of the implementation of the measures, and the driving forces of the future not handled by the tested policy and that might influence the outcome (alternative economic and/or environmental scenarios). Two different methods for interaction were applied: open discussions around key topics and discussions structured around templates to be filled in during the meeting.

290

A.-K. Bäcklund et al.

Templates included pre-defined typologies of problems, spatial scales, policy options (economic, regulatory, voluntary instruments), scenarios (economic and environmental) and impact indicators (domain, scale). Based on 11 regional test cases involving 17 policy experts and nine integrative modellers (from six different research institutes) the following learning experiences can be reported. General Feedback on the Pre-modelling Process Most of the policy experts appreciated the work steps in the pre-modelling phase and expressed that they helped them to clarify and make their problems explicit. The templates were perceived as good tools to support problem framing and to identify suitable policy options to be tested. The policy experts appreciated that they were asked about their opinion about possible impacts of a policy. This gave them an opportunity to express their experience and knowledge, and gave the modellers a possibility to merge this information with scientific knowledge to specify impact indicators. This process of joint work can circumvent that results will be rejected later on. The modellers experienced the pre-modelling procedures to be an effective way to learn about social values of policy experts and include new knowledge. The test enabled them to better understand the policy experts’ opinions and thereby find ways to support them in formulating their aims and interests and to translate these into a problem possible to assess. If the above reported experiences are generally valid it means that the SEAMLESS-IF pre-modelling procedure can serve as a learning interface for policy experts. Although the general feed-back was promising, some stages of the pre-modelling process came out as critical. Those key situations will be discussed below together with suggestions for improvements that can facilitate the process. The Framing Process Problem framing is the process during which involved actors state in more exact terms what they want to do. Problem framing requires an interpretation of the main aspects of the issue, and how they are connected. The factors and causal relationships will suggest possible policy options and external driving forces. The causal links will sort out which biophysical and agro-management contextual elements that have to be considered, and also demonstrate which alternatives that do not have to be considered (Bardach 2000). The combination of policy options, outlooks and contexts, which may be seen as an ‘experimentation plan’, stems from the problem framing (Janssen et al. 2009). If the problem deemphasises or neglects important dimensions, this can ‘limit understanding and narrow analysts’ vision’ creating ‘blind spots in which people will not see potentially valuable alternatives’ (Stern 1986: 200).

12 Science–Policy Interfaces in Impact Assessment Procedures

291

During the problem definition phase the policy experts mainly expressed their problem in terms of organizational interests, values and motivations, while the modellers needed to know the specifications of policy goals, impact indicators, and the constraints in order to set up an experiment. As a result it appeared that problem definition and specifications for the experiments needed several meetings between policy experts and modellers to become formulated. In some cases, no problem and experiment were defined as policy experts needed more time to specify the issues at stake. In other cases, the modellers needed to check the ability of the models to solve the problem or to find ways to make the problem addressable with SEAMLESS-IF. As a result of discussion with policy makers it was decided that SEAMLESS-IF for comparative purposes should offer interactive mapping tools to co-design the different experiments run. In addition it was suggested that example problems and specifications for experiments should be provided as guidelines. This would make it easier to transform interest and values into impact indicators and experiments to be run. An open discourse about research questions and policy questions at the beginning of the assessment process can help to ensure that a broad framing is achieved. In the test cases the framing procedure was essential to attract the interest of the actors, although the participants sometimes contested the frames. Even the idea that a problem needs to be assessed is a frame and could be contested. Specification of the Policy Option The regional policy experts were interested in assessing both EU policies (e.g. reform of the Common Agricultural Policy) and regional policies, which are designed by their own organization (e.g. policies supporting the introduction of specific crops in new regions). But to specify the policy options to be tested proved to be another difficult step in the methodology, as there is still a lack of support tools to make this process effective. The most difficult step for science to work within policy making is the problem definition. Policy makers express frustration with narrowly defined scientific problems. And, scientists struggle to separate the many strands and pieces of complex policy problems into some assessable options. The specification of policy options is of key importance for a successful assessment and also crucial for the understanding between policy experts and modellers. A policy option is a set of measures (changes of economic and/or regulatory environment) planned to be applied in a given space to solve a problem (difference between actual situation and desired situation). Concluding Remark on the Regional Test Experiences To conclude, the test experiences suggest that SEAMLESS-IF can contribute to make integrated assessment a deliberative process between the scientific and political spheres. Although the general experience from setting up regional test cases is promising, critical stages in the assessment procedure which have to be further developed became visible. Three particular design steps deserve attention: the framing

292

A.-K. Bäcklund et al.

process, the specification of resolution space and specification of the policy to be tested. The tests also gave insights about the importance of pedagogical tools which can favour mutual understanding and debate among the actors to improve the process. An additional recommendation is to sufficiently account for the political, economic, social and scientific context in which the assessment is embedded. For this it is essential to find out the needs of potential users, to take notice of past experience and institutional history, and to ensure appropriate participation in the assessment process. If not, there is a risk of not achieving stakeholders’ involvement (Hilborn 1979; Mintzberg 1980).

Conclusions To perform impact assessments (IAs) of policy proposals is gradually becoming an important instrument in European policy making. The IA procedure in the European Commission stems from a governance concept which assumes that policy programs should be the product of complex interaction between government and non-government organizations (researchers included), each seeking to influence the collectively binding decisions that have consequences for their interest. The idea of IA is based on the assumption of ‘co-production of knowledge’. In order to make a system like SEAMLESS-IF applicable in a European decision-making process, interaction with potential users of the system is needed during its course of development. With the objective to make use of potential users’ opinions, different forms of interaction to enable user involvement in the development of the framework have been performed. The learning that came out of discussions with officers in the EU administration participating in the SEAMLESS User Forum and representatives of regional administrations setting up assessments in test situations in France lead us to the following conclusions. The discussions at the EU level confirmed that the proposed framework developed by the SEAMLESS project is matching the potential users’ overall needs. The feedback on the prototypes served the development of SEAMLESS-IF with specifications of requirements on interface and visualisations needed. It also led to the identification of the need for different user interfaces and the adaptation of these according to procedural requirements in the EC’s assessment work. The interactions further assisted in priority setting within the project and greatly triggered the timely thinking about continuity, maintenance, extension and dissemination. Testing of the assessment procedures at the regional level revealed that scientists, specializing in modelling or systems analysis, can play an important role to facilitate the integration of different types of knowledge into the definition of a policy problem. Although the general experience from setting up test cases is promising, critical stages in the assessment procedure which have to be further developed became visible. Three particular steps in the process deserve attention: the framing process, the specification of resolution space and the specification of policy to be tested.

12 Science–Policy Interfaces in Impact Assessment Procedures

293

The most difficult step for science to work within policy making is the problem definition. Policy makers may express frustration with narrowly defined scientific problems and scientists struggle to separate the many strands and pieces of complex policy problems. An overall conclusion from the interactive encounters between policy makers and scientists is that scientists can have an important role in promoting the process of learning in policy. However, in order to define integrated science-policy problems both science and policy must work in new ways. The scientists need to be able to recognize, understand, and value different forms of knowledge. Indeed, one reason for placing science within the policy process is for scientists to learn about social values, appreciate other ways of knowing, and recognize the limits of science in policy. A role for the scientist as an ‘institutionalized critic’ occurs naturally in a policy analysis process. It is the critical contribution of science to demonstrate when beliefs are supported by evidence and when conventional wisdom is simply wrong. Policy actors, on the other hand, must be prepared to engage in the rather time consuming process which is necessary to arrive at a credible outcome of the problem assessed.

References Alkan Olsson, J., Bockstaller, C., Stapleton, L.M., Ewert, F., Knapen, R., Therond, O., Geniaux, G., Bellon, S., Pinto Correira, T., Turpin, N., & Bezlepkina, I. (2009). A goal oriented indicator framework to support integrated assessment of new policies for agri-environmental systems. Environmental Science and Policy, 12(5), 562–572. Bäcklund, A.-K. (2009). Impact assessment in the European commission – a system with multiple objectives. Environmental Science and Policy, 12(8), 1077–1087. Bardach, E. (2000). A practical guide to policy analysis: The eightfold path to more effective problem solving. New York: Chatham House. Brugnach, M., Tagg, A., Keil, F., & de Lange, W. J. (2007). Uncertainty matters: Computer models at the science–policy interface. Water Resources Management, 21, 1075–1090. Callon, M. (1999). The role of lay people in the production and dissemination of scientific knowledge. Science, Technology and Society, 4(1), 81–94. Casey, A. (2005). Enhancing individual and organizational learning: A sociological model. Management Learning, 36(2), 131–147. Cash, D., & Buizer, J. (2005). Knowledge-action systems for seasonal to inter-annual climate forecasting: Summary of a workshop. Washington, DC: The National Academies Press. Cohen, J. (1998). Democracy and liberty. In J. Elster (Ed.), Deliberative democracy (pp. 185–231). Cambridge: Cambridge University Press. EC. (2002). Communication on impact assessment, COM 2002(276). Retrieved Feb 15, 2007, from http://ec.europa.eu/governance/impact/docs/key_docs/com_2002_0276_en.pdf EC. (2005). Impact assessment guidelines, SEC 2005(791). Retrieved Feb 15, 2007, from http:// ec.europa.eu/governance/impact/docs/SEC2005_791_IA%20guidelines_annexes.pf EC. (2006). A strategic review of better regulation in the European Union. COM 2006 (689). Retrieved Feb 15, 2008, from http://www.ec.europa.eu/enterprise/regulation/better_regulation/ docs/en_689.pdf EEAC. (2006). Impact assessment of European commission policies: Achievements and prospects. Statement of the EEAC – Working Group on Governance. Retrieved Feb 15, 2007, from http:// www.ecologic-events.de/eu-impact assessment/en/documents/EEAC_WG_Gov_IAstatement_ background.pdf

294

A.-K. Bäcklund et al.

Elster, J. (ed). (1998). Deliberative democracy (pp. 1–18). Cambridge: Cambridge University Press. Feynmann, R. (1998). The meaning of it all. Reading, MA: Perseus. Gabbert, S., Van Ittersum, M., Ewert, F., Kroeze, C., Stalpers, S. & Alkan-Olsson, J., (2009). Uncertainty information in Integrated Assessment: The users’ perspective. Regional Environmental Change, in press. Grunwald, A. (2004). Strategic knowledge for sustainable development: the need for reflexivity and learning at the interface between science and society. International Journal of Foresight and Innovation Policy, 1(1–2), 150–167. Habermas, J. (1996). Between facts and norms. Cambridge: MIT Press. Hilborn, R. (1979). Some failure and successes in applying systems analysis to ecological systems. Journal of Applied Systems Analysis, 6, 25–31. Hoppe, R. (2005). Rethinking the science-policy nexus: from knowledge utilisation and science technology studies to types of boundary arrangements. Poiesis & Praxis, 3, 199–215. Janssen, S., Andersen, E., Athanasiadis, I.N., & van Ittersum, M.K. (2009). A database for integrated assessment of European agricultural systems. Environmental Science and Policy, 12(5), 573–587. Lee, K. (1993). Compass and gyroscope. Washington, DC: Island Press. Lee, N. (2006). Bridging the gap between theory and practice in integrated assessment. Environmental Impact Assessment Review, 26(1), 57–78. Lindblom, Ch E. (1968). The policy-making process. Englewood Cliffs, NJ: Prentice Hall. Mandelkern Group on Better Regulation (2001). Final Report. Retrieved Feb 10, 2007, from http://ec.europa.eu/governance/better_regulation/documents/mandelkern_report.pdf Mintzberg, H. (1980). Beyond implementation: An analysis of the resistance to policy analysis. INFOR, 18(2), 100–138. Morgan, M. G., & Dowlatabadi, H. (1996). Learning from integrated assessment of climate change. Climate Change, 34(4), 337–368. Smith, G. (2000). Toward deliberative institutions. In M. Saward (Ed.), Democratic innovation: Deliberation, representation and association (pp. 29–39). London: Routledge. Stern, P. C. (1986). Blind spots in policy analysis: What economics doesn’t say about energy use. Journal of Policy Analysis and Management, 5(2), 200–227. Therond, O., Belhouchette, H., Janssen, S., Louhichi, K., Ewert, F., Bergez, J.-E., Wery, J., Heckelei, T., Alkan Olsson, J., Leenhardt, D., & van Ittersum, M. (2009). Methodology to translate policy assessment problems into scenarios: the example of the SEAMLESS Integrated Framework. Environmental Science and Policy, 12(5), 619–630. Turnpenny, J. (2008). Are we nearly there yet? Lessons for integrated sustainability assessment from EU environmental policy-making. International Journal of Innovation and Sustainable Development, 3(1/2), 33–47. Van den Hove, S. (2007). A rationale for science-policy interfaces. Futures, 39(7), 807–826. Van Ittersum, M., Ewert, F., Heckelei, T., Wery, J., Alkan Olsson, J., Andersen, E., et al. (2008). Integrated assessment of agricultural systems – A component-based framework for the European Union (SEAMLESS). Agricultural Systems, 96, 150–165. Van Ittersum, M., Gabbert, S., & Ewert, F. (2008b). Uncertainty analysis in model chains for integrated assessment. SEAMLESS deliverable PD1.3.11.2, SEAMLESS Integrated Project, EU 6th Framework Programme, contract no. 010036-2, www.SEAMLESS-IP.org, 60p. Van-Camp, L., Bujarrabal, B., Gentile, A.-R., Jones, R.J.A., Montanarella, L., Olazabal, C., & Selvaradjou, S.-K. (2004). Reports of the Technical Working Groups established under the thematic strategy for soil protection. EUR 21319 EN/1, Official Publication of the European Communities, Luxembourg, from http://ec.europa.eu/environment/soil/pdf/vol1.pdf

Chapter 13

Economic Principles of Monetary Valuation in Evaluation Studies C. Martijn van der Heide, Neil A. Powe, and Ståle Navrud

Introduction According to Daily (1997) and Mooney and Ehrlich (1997), the idea that human depends on natural systems dates back as far as Plato, who acknowledged that deforestation led to soil erosion and the drying of springs. In his widely read and influential book Guns, Germs, and Steel (1997), the prominent evolutionary biologist Diamond suggests close links between the ability of natural systems in some parts of the world to provide key services to human societies (such as mitigation of droughts and floods, dispersal of seeds and the preservation of soils and renewal of their fertility) and the successful and military powerful civilizations of those societies. ‘Biodiversity’ – short for biological diversity – includes goods which are an important and familiar part of the economy, and the sustained production of these goods is a service provided to society at low or no cost by natural systems (Levin 1999; Daily 2000). However, it is noteworthy to mention that biodiversity does not only support natural systems, but also contributes to the productivity of agricultural systems. For instance, Kremen et al. (2004) show that nearly one-third of the US food supply by volume depends on animal pollinators, of which bee species are the

C.M. van der Heide (*) LEI Wageningen UR, P.O. Box 29703, 2502 LS, The Hague, The Netherlands e-mail: [email protected] N.A. Powe School of Architecture Planning and Landscape, Claremont Tower, Newcastle University, Newcastle-upon-Tyne, NE1 7RU, UK e-mail: [email protected] S. Navrud Department of Economics and Resource Management, Norwegian University of Life Sciences (UMB), P.O. Box 5033, NO-1432, Aas, Norway e-mail: [email protected] F.M. Brouwer and M. van Ittersum (eds.), Environmental and Agricultural Modelling: Integrated Approaches for Policy Impact Assessment, DOI 10.1007/978-90-481-3619-3_13, © Springer Science+Business Media B.V. 2010

295

296

C.M. van der Heide et al.

most important. The importance of biodiversity in agriculture is well captured by Tilman et al. (1999: 4) in the following words: Humans do not produce food. Other animal and plant species produce it for us. The essence of agriculture is the harnessing of numerous species of plants and animals for human benefit. Many of the advances in agriculture have come from the selection and development of new crops and from genetic refinements in these crops.

Agricultural systems benefit from the existence of a pool of genetic diversity that serves as the primary source of new varieties. In other words, genetic diversity is essential if a crop is to be viable over the long term. Moreover, agriculture depends on the various plant and animal species that are grown as major sources of human food and pharmaceuticals. The relationship between agricultural systems and biodiversity is reciprocal, not one-way. Agriculture, as well as other land-use systems, directly affect the diversity of life. For example, agricultural practices have preserved some species which would otherwise have been driven to extinction by hunting or gathering (Tisdell 1991). In its policy brief, the OECD (2005) suggests that the Mediterranean basis is considered a biodiversity ‘hotspot’, where ‘hot’ refers not literally to high temperatures or warm weather, but to its high level of human-induced agricultural biodiversity. However, agriculture has also been responsible for a direct negative impact on biodiversity. Commercial agriculture has caused extensive habitat loss and degradation – which are among the greatest current and future threats to biodiversity – and it can be considered as one of the most important causes of pollution through its excessive use of chemical inputs, irrigation and mechanised tillage (McNeely et al. 1995; Dirzo and Raven 2003). Moreover, subsidy policies usually distort the response of agricultural producers to market signals and prevent the efficient use and allocation of resources. Price supports and production subsidies can contribute to increasingly intensive and polluting agricultural practices. The effects that agriculture has on biodiversity are ‘externalities’; that is, they are largely external to the market and thus not reflected in market prices. The fundamental reason for this is that many biological assets, such as wild species and natural systems, are characterized by the absence of fully defined property rights. Many of these assets are considered to be public goods, or possess some features associated with such goods. Pure public goods have the characteristics of non-rivalry and non-exclusion (for example, Cornes and Sandler 1996; Ison et al. 2002, and many other textbooks on environmental and ecological economics). Non-rivalry implies that, once the good is provided to a consumer, it can be made available to other consumers at no extra cost, or – in more prosaic, economic terms – the marginal social cost of supplying the asset to an additional individual is zero. For example, a nature area that is protected by or for one agent will benefit everyone else who can access the area. Non-exclusion means that one user cannot prevent consumption by others. Due to the non-exclusion attribute – that is, due to the fact that it is impossible or at least very costly to deny access to a biological asset – markets fail to allocate resources with public good characteristics efficiently.

13 Economic Principles of Monetary Valuation in Evaluation Studies

297

This may be understood by noting that prices do then not signal the true scarcity of the asset (Hanley et al. 1997). So, in summary, conventional markets do not work very well for biological assets. The market provides inadequate incentives for their provision because, due to their non-excludability, individuals can benefit from these assets without paying for them – the free-rider problem. However, the fact that a biological asset is not fully captured in a market, and hence has no price does not mean that it is of no value. Economic analysis can shed light on the value of biodiversity. To put it even stronger, it is our view that economists can provide insights into the potentially high benefits of biodiversity, and can reveal the economic and social pressures that threaten it. Restated in terms of SEAMLESS: through monetary valuation and a proper application of economic principles, economists can support debates on how to integrate biodiversity conservation into agricultural activities. Moreover, and perhaps more important, they can enhance knowledge of the value of biodiversity within agricultural systems. Needless to say that policy makers in general and SEAMLESS end-users in particular can make better choices if the valuation issue is made as explicit as possible. So, the rationale of this chapter is to explore ways to employ valuation methods to assess the economic impacts of agricultural and environmental policies on biodiversity. Based on this rationale, this chapter aims to bring an economics perspective on the benefits of biodiversity in agriculture. It highlights key issues involved in the notion and application of monetary valuation methods for the assessment of biodiversity values in the agricultural landscape. This agricultural biodiversity, or agro-biodiversity, is the variation in all living organisms that are related (and important) to agriculture. It includes not only crop varieties and livestock breeds, but also non-domesticated resources within production systems, such as fields and forests. Our treatment is guided by the following research questions: –– Which approaches and methods of monetary valuation are distinguished? –– What insight do existing valuation studies provide toward monetary values of changes in agro-biodiversity? –– What role can monetary valuation play in agricultural policy impact assessment? –– What is the potential use of monetary valuation methods in SEAMLESS? These four questions are addressed individually in the succeeding four sections of the chapter. A final section presents Conclusions. One last comment before ending this section: there is a huge amount we can say on each of the four research questions. The issue of valuation is well founded in the current (very extensive) literature, and consequently we can dwell considerably longer on this issue than space permits. However, the disadvantage of having lack of space is outweighed by the advantage that the above raised questions are more deserving of work, and probably this work would be challenging. To quote Piet Hein, a Danish mathematician and poet: ‘Problems worthy of attack prove their worth by hitting back’.

298

C.M. van der Heide et al.

Methods for Valuing Biological Assets Looking at the number of papers and monographs published on monetary valuation of biological assets, it could probably be said that it is now a well established practice. It might also be felt that valuation of nature is still controversial because of the combination of theoretical and empirical problems, and the potential importance the resulting values may have in influencing public opinion and policy decisions (Loomis et al. 2000). For example, biologists such as Ehrlich and Ehrlich (1992), argue that ecosystems are complex, indivisible entities that operate on time scales outside the range of human perception, and that they have values that are difficult or impossible to measure (see also Gowdy 1997). Not only biologists, but also scholars from other disciplines (and even economists) may find monetary valuation a ‘hopeless’ exercise. Philosophers, such as Sagoff (2000) and economists, such as Bromley (Vatn and Bromley 1994), dismiss monetary valuation as ethically insupportable and impracticable. Nunes and van den Bergh (2001), on the other hand, claim that monetary valuation of biodiversity can make sense. However, they point out that the various valuation methods should not be considered as universally applicable to all levels of biological diversity or to all types of biodiversity values. Research on the monetary value of biodiversity can be motivated by the desire to better understand the importance of biodiversity (for a much more elaborate treatment of the need for valuation, see, for example, Simpson 2007). Intuitively, the importance of biodiversity to society is best represented by monetary values. However, reality is more complicated than the common intuition suggests. Economics is more concerned with prices than with values or importance. 1 An important difference between prices and values, which is particular relevant for this study, is that prices that arise from market transactions offer, in some sense, objective information, whereas many concepts of value are subjective (Heal 2000). As a result, valuation cannot be dissociated from choice. Economic analysis provides several very ingenious measurement techniques to assign a (subjective) value to the benefits of, or avoided damage due to, changes in agro-biodiversity, and next we review some of these.

Non-demand Approaches Following Bateman and Turner (1993), two basic approaches are distinguished: those which value an asset via a demand curve (see the following subsection), and those which do not. In general, non-demand approaches are based on the principle assumption that if people incur costs to avoid damages caused by lost nature’s services (such as the pollination of agricultural crops by bee species) or to replace nature’s services, then the value of those services are worth at least the amount of Reasoning along Oscar Wilde’s line of thought – who once remarked that a cynic is someone who knows the price of everything and the value of nothing – economists are utterly cynical folks.

1

13 Economic Principles of Monetary Valuation in Evaluation Studies

299

money that people have paid to replace them. There are a number of non-demand approaches, such as the opportunity cost approach, the dose response approach, the preventative expenditure approach, the averting behaviour approach, and the replacement cost approach (see, for example, Turner et al. 1994; Garrod and Willis 1999a). Some examples related to measures of agro-biodiversity value where non-demand approaches might be applied include the following: –– Valuing water quality by measuring the cost of filtering or the chemical treatment of water; –– Valuing an environmental externality by observing the costs people are prepared to incur in order to avoid any negative effects; –– Valuing land degradation by estimating the restoration cost. The examples show that, in general, non-demand approaches are limited to environmental inputs, such as water, air quality and soil. The results of these non-demand approaches can thus be used to demonstrate the importance of these inputs to agriculture, but also to appraise pollution control options.

Demand Approaches: Revealed Preference and Stated Preference Methods Demand curve approaches are broadly divided into revealed preference and stated preference methods. The first type of approach seeks to elicit preferences for biological assets by examining the purchases of related goods in the private market place. In other words, revealed preference methods use market information (such as travel costs or the price of housing) as a proxy to estimate the benefits from biological assets. Stated preference methods consider a change in environmental quality or in biodiversity and, by using survey techniques, they seek to directly measure the value of these changes (Perrings 1995; Garrod and Willis 1999a; van der Heide et al. 2003). As such, these methods derive a value by simulating, or constructing a hypothetical market for the biological asset in question. Although economists are generally much more comfortable with valuations based on actual transactions and observed behaviour in markets than those given in response to survey questionnaires (Heal 2000), for some public goods there are simply no accurate means of inferring preferences from observations. In such circumstances, there are no viable alternatives to asking individuals how much they would be willing to pay for a biological asset or accept in compensation for its loss in a hypothetical situation of payment. Travel Cost Method The travel cost method (TCM), a revealed preference method, is one of the oldest valuation methods employed by environmental economists (Clawson and Knetsch 1966). It is especially useful for assessing the value of outdoor recreation in natural

300

C.M. van der Heide et al.

parks, and to this end, it has been widely used in the USA and to a lesser extent the UK (Perrings 1995). The underlying assumption of the TCM is that the incurred costs of visiting a national park, nature reserve, open space or any other recreational site are directly related to the benefits individuals derive from the amenities within the area, such as hiking, camping, fishing, bird watching and, swimming. The method involves using the value of time spent in travelling, the cost of travel (e.g., petrol costs) and entrance and other site fees as a proxy for computing the demand price of the environmental resource. Because TCM is primarily concerned with recreation and tourism values, the application of the method for quantifying agro-biodiversity values is rather limited. That is, the method is well suited to assessing values related to leisure and recreation in agricultural landscapes, but it is incapable of assessing other agro-biodiversity values, such as water and air quality or the role of bee species in crop pollination. Hedonic Pricing Method The hedonic pricing method (HPM) derives the value of environmental amenities, such as pollution and noise level, from actual market prices of some private goods. Just like the travel cost method, the HPM is based on observed behaviour. By far the most common application of HPM is to the real estate market. House prices are affected by many factors, not only by house characteristics like the number of rooms and the size of the garden, but also by the environmental quality of the surroundings, including proximity to natural areas and the quality and uniqueness of such areas. If the non-environmental factors can be controlled for, then the remaining differences in real estate prices are expected to be the result of environmental differences (Turner et al. 1994). HPM is suitable for the estimation of changes in water and air quality, but is especially appropriate for assessing noise and air pollution (from activities such as large agricultural machinery, landfilling or transport of waste) (Spash and Carter 2002). Contingent Valuation Method TCM and HPM are both revealed preference methods. A common problem with these methods is that in the absence of appropriate data or interdependent market goods, assessing the value of agro-biodiversity is either not possible or will lead to spurious results. Stated preference methods bypass the need to refer to market prices by asking people directly what their willingness to pay (WTP) for a change in environmental quality or biodiversity is. This requires the presentation of a change scenario, for example the loss of bee colonies, a decline in farmland bird species, or the conversion of agricultural land into urban land. The contingent valuation method (CVM) is the most used stated preference method: there are now thousands of papers and studies dealing with the topic (Carson 2000; for an overview of 50 years of CVM, see Smith 2004). CVM has been used extensively

13 Economic Principles of Monetary Valuation in Evaluation Studies

301

in the valuation of biological resources, including rare and endangered species, habitats and landscapes, although it should be recognised that this method will fail for those value categories that the general public has no experience with (Nunes and van den Bergh 2001). It invokes a framework of a contingent (or hypothetical) market, used to indicate what individuals are willing to pay for a beneficial change or what they are willing to accept by way of compensation to tolerate an undesirable change (Garrod and Willis 1999a; Carson 2000; Boardman et al. 2006). The main advantage of CVM is that, in theory and to varying degrees, it is capable of capturing most of the value categories related to agro-biodiversity. The hypothetical nature of CVM offers flexibility in application, and as a result, the method is quite versatile. Nevertheless, results from CVM studies are heavily dependent on the choice of the particular format used to elicit information about the respondent’s preferences. The success CVM has experienced relates directly to the energy, time and money spend on the design state of the contingent valuation survey as well as to the question wording, the question sequencing and the individual interviewers (Diamond and Hausman 1993). Respondents are assumed to be rational and knowledgeable and the best judge of their own interests. Moreover, CVM has many acknowledged problems. The hypothetical character induces the occurrences of an impressive list of potential biases that result from using CVM. They include strategic bias, embedding bias, part-whole bias, starting point bias, and payment vehicle bias (see, for example, OECD 2002; de Blaeij 2003; Boardman et al. 2006). So although CVM has become one of the most widely used non-market valuation technique, debate persists over the reliability of CVM. That is, opinion among scientists is divided. In fact, the debate between advocates and critics of CVM is, partly, a battle of giants, dominated by heavyweight and highly skilled intellectuals on both sides, whose powerful weapons are their words. Many economists have expressed ‘discomfort’ with using the estimates from contingent valuation to measure consumers’ WTP for changes in non-market goods, such as biodiversity, landscape scenery and environmental quality. More than 10 years ago, Diamond and Hausman offered the most dogmatic rejection of the method. They (1994: 62) tell us that when using a CVM ‘… one does not estimate what its proponents claim to be estimating.’ In other words, according to these two CVM sceptics, people simply do not have preferences over what Diamond and Hausman term ‘non-use values’ (see also Vatn 2004).2 Also other well-known scholars, such as Kahneman et al. (1990) and Sagoff (2000) express their doubts with respect to the suitability of CVM for various reasons. Nonetheless, the opinions of these (resolute) detractors has not slowed research in this area – on the contrary. Although proponents of CVM acknowledge that CV studies range from very good to very bad and that the technique suffers from various design problems that require effort and skill to resolve (Smith 1992; Hanemann 1994; Carson 2000), they

Non-use values are values that individuals hold for nature’s goods and services independent of actual use. For example, individual may attach a value to blue whales – because they find it important that these whales will be properly protected – regardless of any personal use, now or in the future. 2

302

C.M. van der Heide et al.

believe that extensive methodological research and quality improvements have already increased the reliability and feasibility of the contingent valuation approach. An important stimulus to the use of CVM was the recommendations of the National Oceanic and Atmospheric Administration’s, or NOAA, panel on contingent valuation (see NOAA 1993). This panel of eminent social scientists, co-chaired by two Nobel Prize-winning economists, Kenneth Arrow and Robert Solow, specified an extensive set of guidelines for CVM survey development, administration, and analysis. It also found that CVM, if appropriately conducted, could convey useful information. Choice Modelling In the field of monetary valuation, choice modelling (CM) is being increasingly applied as an alternative to CVM (Adamowicz et al. 1994; 1998). Like CVM, CM is a stated preference method and is capable of measuring consumer acceptance of multi-attribute commodities. CM has benefited from the work carried out in the field of psychology, economics, and consumer market research, where marketing research scholars employed the technique to evaluate potential new products and new markets for existing products (Garrod and Willis 1999a; for thorough and critical descriptions of CM, see Roe et al. 1996; Farber and Griner 2000, and Louviere et al. 2000). Hanley et al. (2001: 436) define CM as ‘… a family of survey-based methodologies for modelling preferences for goods, where goods are described in terms of their attributes and of the levels that these take.’ As such, the term choice modelling is somewhat broader in coverage than the method of choice experiments; or, to put it more precisely, choice experiments are a derivative of CM.3 Generally, in a CM, respondents are confronted with a number of commodity descriptions, or situations, that differ according to the attributes described (as in hedonic pricing). Respondents are then asked to rank or rate the bundles of attributes, or select the most-preferred one from the set. The basic premise underlying this method is that commodities have a value because of their attributes. In order to decide which commodity they want, people make trade-offs between the attributes (de Blaeij 2003). CM is designed to determine the structure of preferences that underlie the judgement of multi-attribute commodities and to that end, it measures the rates at which people are willing to make such trade-offs (Shechter 2000). The inclusion of at least one monetary attribute, such as the cost of provision or the WTP for conserving the countryside, allows for the derivation of implicit prices for each of the other attributes. Unlike CVM, CM does not directly ask for a WTP, but offers the opportunity to rank all of the alternatives from highest (best) to lowest (worst). Defenders of CM therefore argue that the technique outperforms CVM on the point of strategic behaviour. The indirect way of questioning in the CM allows respondents to explicitly determine trade-offs between different attributes of the commodity under 3 The term choice modelling is often used interchangeably with the term conjoint analysis (e.g. Hanley et al. 2001). Of these two terms, choice modelling seems to be the most commonly used nowadays.

13 Economic Principles of Monetary Valuation in Evaluation Studies

303

valuation. CM emphasises thus not only monetary valuation. Because CM is designed to rank multi-attribute alternatives, it seems more suitable to value agricultural landscapes and the countryside as a whole, which provide a multitude of joint goods and services, than the typical one-dimensional CVM (Farber and Griner 2000). It allows obtaining values for specific characteristics of (landscape) scenarios rather than one specific scenario. On the other hand, CM can be more complicated to implement, and the selection and representation of the several attributes adds to the design problem already associated with hypothetical surveys. Benefit Transfer While the debate over the existing techniques for monetary valuation continues, the search for new approaches remains. The method of benefit transfer is fairly new and still a developing subject (see, for example, the special issue of Ecological Economics on the state-of-the-art and science of benefit transfer, edited by Wilson and Hoehn 2006). Benefit or value transfer involves ‘borrowing’ monetary environmental values estimated at one site (the study site) and applying them to another (the policy site). The attraction of benefit transfer is that it avoids the cost of conducting ‘primary’ studies whereby the benefits of natural or environmental assets or the damages associated with degradation are measured with one or more of the techniques described above (OECD 2002). Navrud (2001; see also Rosenberger and Stanley 2006, and Spash and Vatn 2006) distinguishes two main approaches to benefit transfer: namely, unit value transfer, and function transfer. Unit value transfer is the easiest approach and refers to simply transferring the results of an environmental valuation study to the policy site. The valuation estimates may be left unadjusted, or may be adjusted in some way. Although transferring unadjusted estimates is undesirable, since the values experienced in the study site may not be the same as in the policy site, it is widely practised (OECD 2002). A more sophisticated approach is to transfer the entire benefit function. Thus, instead of transferring only the benefit estimates, a researcher transfers a benefit function, substituting data on independent variables at the new site. Both unit values and value functions can be based on multiple studies. Meta-analysis can be used to take the results from a number of studies, synthesise research findings, and analyse them in such a way that the variations in unit values or value functions found in those studies can be explained (van den Bergh et al. 1997; Navrud 2001; Dalhuisen 2002). The advantage of benefit transfer is that it is both pragmatic and cost-effective and therefore, the method is a very attractive alternative to expensive and time consuming original research to quickly inform decision making (Garrod and Willis 1999a). To date, however, benefit transfer is a controversial issue in various policies contexts. The major reason for this is that the accuracy of transferred estimates is debated because validity tests show that the uncertainty in transfers could be quite large, both spatially and intertemporally. This is especially the case for complex goods, such as biodiversity in agriculture and landscape values.

304

C.M. van der Heide et al.

The benefit transfer approach is recognised as a valid technique for valuation of agro-biodiversity only if the circumstances at the study site are very close to those of the policy site, which is highly unlikely. This does not necessarily imply that the application of benefit transfer should be restricted to transfers within one and the same region. Ready and Navrud (2006) explore the issues that must be addressed when conducting benefit transfers between different regions (namely between countries). They also provide information about the benefit transfer of landscape values (Navrud and Ready 2007).

Monetary Values of Changes in Agro-biodiversity: Some Examples The existing valuation studies of agro-biodiversity and agricultural landscape use TCM, CVM, and CM. TCM is applied only to visitors to the site and as such, it captures only the ‘use value’ arising from recreation and tourism. There are also hedonic price studies valuing open space, but here it is more difficult to isolate the economic value of agricultural landscape. Thus, contingent valuation and choice modelling studies are most relevant for capturing the multiple values of agricultural landscape. OECD (2001, Annex Table 2) provides an overview of results from selected stated preference studies, mostly contingent valuation studies, of agricultural landscape and biodiversity conservation (recall that stated preference methods use ‘hypothetical markets’, described by means of a survey to elicit directly from individuals the monetary value they assign to non-marketed goods and services). It appears that the stated preference methods of CVM and CM have been successfully applied to value rural amenities, see e.g. Dubgaard et al. (1994), Pruckner (1995), Bergland (1998), Santos (1998) and Dachary-Bernard (2004). Below we describe three of these studies in more detail. Pruckner (1995) estimated the economic benefits from agricultural landscapecultivating activities in Austria in 1991. Multifunctionality of agriculture was taken into account, in terms of e.g. protection from avalanches and landslides, erosion, and contribution to the maintenance of rural culture. Tourists were first asked to select from a list of 26 items those that they rated to be important for them at the holiday resort. Then they were asked whether the farmers or some other specialists should provide landscape-related services. Finally, an open-ended contingent valuation question was applied to elicit their willingness to pay for landscape-cultivating services, i.e. mowing grassland, thinning of forests, and taking care of the rural network of walking paths. The mean and median WTP for these services were 9.20 ATS and 3.50 ATS per person per day, respectively.4 Individual WTP varied with the different nationalities of tourists. One ATS (Austrian Shilling) was equivalent to 0.07 ECU (European Currency United) in 1995. This means that, in terms of euros, the mean and median WTP for landscape-cultivating services in Austria were €0.67 and €0.25 per person per day, respectively.

4

13 Economic Principles of Monetary Valuation in Evaluation Studies

305

Dachary-Bernard (2004) applied CM to value the Monts d’Aree landscape in the Finistere region in Brittany in France. The aim of the study was to identify the main attributes of landscape, i.e. moorland, hedged farmland and farmland buildings, and the levels of each attribute. This analysis was used to elicit the WTP of residents, both main and second home owners, and tourists, in terms of the additional cost they would have to pay in order to finance the proposed landscape policies. The payment vehicles used for these alternative policies were local income tax and tourists tax, respectively. Different policy scenarios were created, including the status quo, and different choice sets were constructed. The respondents’ choices were recorded, and utility function estimated for both populations, in order to calculate a monetary measure of welfare changes arising from landscape transformations. The results show that tourists have lower WTP than residents, which seems to be due to the fact that WTP depends on the amount of time respondents spend in the landscape and on the two different payment vehicles used (with tourist tax usually being much lower than typical local income taxes; and respondents anchoring on the expected levels of these payment vehicles). One of the very few valuation studies of agricultural landscape that also tests the validity of benefit transfer was performed by Santos (1998) in the Pennine Dales Environmentally Sensitive Area (ESA) in the UK. Many landscape features are threatened by current changes in farming practice. The Pennine Dales ESA scheme is a voluntary scheme that offers farmers management agreements aimed at conserving the landscape attributes of the Dales. A contingent valuation survey of 422 visitors was undertaken in 1995, in order to estimate visitors’ WTP for the ESA scheme. The value of the most complete policy mix for the average individual is estimated at £112.19 per household per year. These results were then compared with transferred estimates from a list of 19 landscape valuation studies, applying different benefit transfer techniques to test the validity of landscape value transfers. The most accurate transfer techniques in this case was unit transfer and function transfers from a meta-analysis of all studies. Transfer error were in these cases as low as 15–22%. The three studies above studies were all conducted in EU-15 countries. A review of valuation studies in the Czech Republic, Poland and Hungary by Melichar and Ščasný (2004) reveals that only two contingent valuation surveys of agricultural landscape have been performed in these new member states of the European Union. Both of these were in the Czech Republic. Kubickova (2004) applied a in-person contingent valuation survey to a sample of 480 individuals asking for their WTP to maintain agricultural activities contributing to landscape preservation to ensure the conservation of the currently cultivated landscape of the White Carpathians Protected Landscape Area (PLA). WTP was stated as a contribution to a special fund for this PLA, and the mean WTP per person per year was 295–340 CZK and 664 CZK in the open-ended and dichotomous choice elicitation formats, respectively.5 Krumalova et al. (2000) performed a contingent valuation survey of more than 1,000 individuals by asking them their WTP in terms of an increase in their annual tax payment to One CZK = €0.06 in terms of Purchasing Power Parities, which are the corrected exchange rates from GDP per capita in 2004 (for further information about Purchasing Power Parities, see http:// www.oecd.org/dataoecd/61/54/18598754.pdf).

5

306

C.M. van der Heide et al.

improve the quality of agricultural landscape amenities and provide harmonic landscape. The mean WTP was 620 CZK per person per year in the open-ended elicitation format. In theory, the method of benefit transfer can be used to value the agricultural landscape in the Czech Republic, or in the other new member states. Unfortunately, however, the transferability of estimates from both Czech studies can be questioned. In the first contingent valuation study (the White Carpatian PLA) the values relate to a very specific protected landscape, while in the other study the change in the quality of landscape seems diffuse and poorly defined (e.g. a scenario was presented and described as ‘harmonic landscape’). This section has particularly focused on studies that (implicitly) equate agricultural landscape values with the monetary value of agro-biodiversity. Other studies that perform a value assessment of agro-biodiversity are, however, conceivable. For example, one can think of valuing the effect of eutrophication reversal and water restoration processes on agricultural ecosystems. And indeed, a number of studies have applied monetary valuation methods to investigate the public willingness to pay for water quality improvement (see Ewel 1997; Garrod and Willis 1999a). However, most of these studies do not distinguish between the sources of pollution (urban, industrial, agricultural), and therefore it is difficult to isolate the impact that the improvement in water quality has on agro-biodiversity.

Monetary Valuation and Agricultural Policy After some examples of valuation studies in the previous section, we now continue with the potential for valuation studies to be used within policy. Monetary valuation can be used to provide information when deciding whether to go ahead with specific agricultural policies (ex-ante) or the evaluation of policies (ex-post). In either case, the key methodology used has been cost-benefit analysis, which provides a comparison of social costs and benefits and is based within a utilitarian perspective focusing on economic efficiency. Through a consideration of the validity of benefits estimated and the method of the cost-benefit method of aggregation, this section considers the applicability of monetary valuation within agricultural policy.

Estimating Social Costs and Benefits Earlier in this chapter, we described a range of methods for the valuation of non-market benefits occurring due to intervention in the delivery of agricultural policy. However, the validity of the non-market values which have been estimated have been the subject of much debate. In terms of the benefits from agricultural policy, much debate has revolved around the complexities of scenarios and the relationship between the policy and the data available. Debate has particularly focused on the challenges of understanding the policy-on and policy-off scenarios. Any study valuation or otherwise, considering policy

13 Economic Principles of Monetary Valuation in Evaluation Studies

307

outcomes of agricultural policies will be faced with a number of difficulties in understanding the outcomes with and without policy. For example, there may be uncertainty in terms of science, where Zechmeister et al. (2003) questions the link between biodiversity outcomes and the amount of subsidy provided. Where studies are ex-ante, it is difficult to determine in advance how successful they will be. In addition, even where the evaluation is ex-post, policies do not operate in laboratory conditions. Instead, there are often multiple policies affecting the environmental outcomes of agricultural practices, where only the culmination of these policies is observable. If this was not enough, it is also difficult to predict the land use practices of land managers in the policy-off situation (see Hodge and McNally (1998) for a more detailed discussion of these challenges). Such difficulties would be common within all methodologies; however, they pose particular challenges for environmental valuation. For revealed preference methods (TCM, HPM), there may be questions as to how relevant the shadow prices estimated from market transactions are to the policy issues concerned. Such methods are inflexible to the specifics of the scenarios considered. For example, if a biodiversity value is required, this is very problematic using hedonic pricing or travel cost methods, where the benefits accrued from proximity or visits to areas of high biodiversity will be ‘polluted’ with a whole range of other values. Although the very flexibility built within stated preference methods invites application to complex problems intractable to other techniques, such as specific benefit estimation, this in turn has highlighted a host of practical and theoretical problems reflected in a voluminous research literature. In terms of stated preference methods a key issue relating to scenario complexity is the cognitive ability of the respondents. Due to the unfamiliarity of the scenarios valued and the novelty of considering environmental goods and services in terms of WTP, the cognitive challenge for the respondents is high; emphasising the needs to keep the scenario presentation simple but sufficiently detailed to be believable. This is particularly the case when it is realised that no actual money changes hands and that the expressed WTP is hypothetical. Although it is sometimes possible to design scenarios where there is an implied link between response to a WTP question and policy response. Using national payment vehicles and on site recreational survey locations to consider a single scheme, the valuation of agri-environmental schemes do not represent an ideal situation for stated preference. The response-policy link would be likely to be stronger for a broader good and a national survey. Where this is not the case there is a danger that respondents will confuse the specific scheme for a broader good and a which would lead to larger WTP amounts being stated. Given the extent of the challenge, introducing uncertainty of outcomes would add significantly to the cognitive load. This was illustrated by the debate between Hodge and McNally (1998) and Garrod and Willis (1999b) regarding the feasibility of valuing Environmentally Sensitive Areas. Garrod et al. (1994) for example carefully developed a questionnaire with clear policy-on and -off scenarios as demonstrated using painted representations of landscape outcomes. These were then valued using contingent valuation. Hodge and McNally (1998) criticised this study, suggesting it ‘failed to establish a clear and realistic description of the likely outcome of the policy’ (p. 357). It was suggested instead that ‘rather than

308

C.M. van der Heide et al.

simply seeking one valuation for a single choice, it would seem appropriate to explore with respondents their reactions to a range of alternatives so as to work towards some sense of the incremental values related to the extent and characteristics of conservation goods to be produced’ (Hodge and McNally 1998: 366). Although not possible using contingent valuation, the use of choice experiments for example, can provide added flexibility and have since been used extensively by Garrod and Willis. Using experimental groups the potential for using choice experiments to compare environmental outcomes has been explored by Powe et al. (2005). Within these surveys, some participants stated that they could choose between environmental attributes and a statistically significant difference in preference was observed. However, participants generally found such choices difficult. In fact, some thought they had inadequate knowledge to make such decisions. Further to the complexity of the scenarios considered, there is also complexity in values held by the general public. For example, results of experimental economics research have questioned the ability of microeconomic theory to explain behaviour concerning collective goods such as environmental issues. Ostrom’s (2000) review of experimental economics suggests a plurality of values to be held, where some actors follow self-interest and behave in a manner consistent with economic theory, whereas others also follow norms of trust and reciprocity. Although within stated preference surveys it is up to the individuals responding which issues influence their choices made (Hanemann 1994), the following of norms may lead to the use of inconsistent response strategies, making valuations highly sensitive to the particular norm followed and making it more likely that respondents will be influenced by the social situation of an interview. The use of such norms has been commonly observed within stated preference surveys (Powe 2007). In summary, the inflexibility of revealed preference questions often do not provide the opportunity to consider the specific policy or goods and services needing to be valued. Although stated preference approaches are more flexible, dealing with problems of complexity of scenarios, hypothetical nature of the valuations and plurality of values, have led to a whole range of difficulties being observed, questioning the validity and applicability of the valuations. Despite these challenges, the stated preference studies that have considered agri-environmental schemes have provided a wealth of understanding in terms of the attitudes and preferences that the general public holds. Although the valuations themselves would seem to be positively biased, the understanding their elicitation has provided should prove very useful in the design of future agri-environmental schemes; particularly if choice experiments are used.

Aggregation of Costs and Benefits The use of cost-benefit analysis is based on the criteria of Potential Pareto Improvement (PPI), where if the magnitude of the gains from moving from a policy-off to policy-on situation is greater than the magnitude of the losses, then

13 Economic Principles of Monetary Valuation in Evaluation Studies

309

social welfare is increased by making the move even if no actual compensation is made. As such through adding up the social costs and benefits it is possible to say whether a particular policy has/is likely to lead to a PPI and an improvement in social welfare. Although the potential for compensation does exist, the lack of consideration of equity within cost-benefit analysis has been the subject of much criticism (Jacobs 1997; Wilson and Howarth 2002). This objection is widely held, particularly the implicit assumption of cost-benefit analysis that the existing income distribution is optimal (Pearce et al. 2006). This assumes equality in terms of marginal utility of income, such that €1 creates the same additional utility regardless of whether people are rich or poor. Through the use of weighting, cost-benefit analysis could be adjusted, however, this would need to be undertaken based on some form of political consensus and is controversial. This is particularly the case as, within stated preference surveys, individuals are free to consider what they wish, and some respondents may have considered equity as part of this decision process. Wilson and Howarth (2002) argue this is particularly the case for environmental goods and services which are seen to be rich in equity issues. Based on the principles set out by Habermas (1984) and Dryzek (1990) in the context of environment valuation, Wilson and Howarth (2002) suggest that valuation of environmental goods should not result from the aggregation of separately measured individual preferences, but from a process of free and open debate within which participants can go beyond private self-interest to consider wider ranging issues such as equity. A fair process may well lead to a fair outcome, but it is questionable to what extent a fair process can be developed. The issue of equity illustrates the more general question of whether the cost-benefit analysis is just too narrow? This was indeed questioned above for stated preference responses, where individuals struggle to respond in a way consistent with the requirements of cost-benefit analysis or as Gowdy (2004) suggests ‘forcing individual preferences through the narrow funnel of rational choice theory’ (p. 246). To consider this issue further there is a need to consider the results of cost-benefit analysis in practice. Hanley et al. (1999) present the cost-benefit results from a series of studies of agri-environmental schemes, where valuations were elicited using stated preference methods (CVM, CA). The results suggest that in terms of economic efficiency, the schemes have provided good value for money, where the cost-benefit or PPI tests being passed in most cases by some margin. At face value this is a very positive endorsement of agri-environmental schemes, leaving others to consider any compensation adjustments that might be required. However, interpreting the meaning of this finding in the context of what has been said above would be putting too much faith in the valuations generated, for two major reasons. First, it has been assumed within the valuation studies that the agri-environmental scheme will deliver in terms of biodiversity and landscape improvements. As such CBA says nothing about the effectiveness of the policy. This is left to other studies of the agri-environmental scheme take up rates by farmers, and the results of environmental monitoring (Hanley et al. 1999). As noted above, if the scenarios described within the valuation studies (policy-on and policy-off) do not reflect what

310

C.M. van der Heide et al.

is or will actually occur then the valuations will at best need to be adjusted. Second, although the positive bias within the stated preference responses has resulted in most policies passing the PPI tests, a comparison of tests does at least provide a ranking of which studies are likely to have provided the greatest benefits. Care must be taken within such a ranking as the methodology applied varies considerably between studies and the resultant response biases are also likely to vary.

The Role of Monetary Valuation Methods in SEAMLESS The previous sections have shown that monetary valuation of agro-biodiversity can play a supportive role as it may offers – in a tangible way – a means to communicate the attitudes toward and preferences for specific agricultural policies that the general public hold. With all this information behind us, let us try to answer the question what is the potential use of monetary valuation methods in SEAMLESS. Its integrated framework (SEAMLESS-IF) is developed to enable an ex-ante, integrated assessment of agro-environmental policies and agro-technological innovation in the European Union (EU). The ‘founding parents’ of SEAMLESS put it as follows (van Ittersum et al. 2008: 152): ‘SEAMLESS-IF aims at providing analytical capacity to assess sustainability of agricultural systems in the European Union and contributions to the EU’s agricultural systems to sustainable development at large, including some effects on the entire production chain (transport, processing and packing) and other land uses (e.g., nature).’ Although one may wonder what exactly is meant by the term ‘some effects’, this passage invite us to believe that, prima facie, monetary valuation can be a powerful tool in SEAMLESS. Of course, SEAMLESS end-users can opt out of monetary values to losses of agro-biodiversity. But monetary valuation is intended to make it easier to be informed. Any agricultural or agro-environmental decision will involve an implicit valuation that the costs of a decision are worth (or are not worth) the benefits. The role of monetary appraisal within SEAMLESS is to provide a rigorous and explicit assessment of the impacts of the decision on agro-biodiversity so as to aid such agricultural or agro-environmental policy-making. The definition of the impacts (or effects as abovementioned) is crucial here. Monetary valuation differs from other appraisal methodologies, such as Environmental Impact Assessment, because economists mean by impacts the impacts on the welfare of affected parties – as reflected in their preferences (or WTP) concerning these impacts. The goal of monetary valuation is to monetise the impacts of policy measures, rather than describing the nature, scale and significance of these impacts in physical or qualitative terms. As such, monetary valuation is inherently and purely anthropocentric in nature, based on utilitarian principles. Notice that, in general, policy-makers also adopt an anthropocentric orientation, in that they ultimately favour human concerns. Now we have answered the question what the potential use of monetary valuation in SEAMLESS is, we have to consider the logical follow-up question: how can monetary valuation methods be integrated meaningfully within the integrated framework?

13 Economic Principles of Monetary Valuation in Evaluation Studies

311

In SEAMLESS, the impact of policy scenarios is assessed through a set of indicators that capture the key economic, environmental, social and institutional issues of the questions at stake. Monetary valuation typically captures environmental impacts and attaches an economic value to them. By using monetary terms as a common metric, valuation allows for an unambiguous measurement of environmental impacts on human utility or welfare. In SEAMLESS, this metric needs to be euros. The different valuation methods measure the preference of individuals for specific agro-environmental policy measures. Decision-makers, however, are interested in the preferences expressed by the general public. Therefore, it is imperative to bring together or aggregate the preferences of individuals to indicate the overall preferences of society. Here a difficulty arises, because how to devise methods that can handle the aggregation of individual WTP-amounts? For example, specific agro-environmental policy measures for green olives will presumably only have an impact on the biodiversity in olive orchards. Therefore, we can reasonably assume that individuals in, say, Denmark or Belgium are rather indifferent about these measures, in contrast to the people in the Mediterranean countries, who may be directly affected by these specific policy measures.6 A valuation study can be conducted to monetise the impacts of the ‘olive measures’ on the agro-biodiversity. It is justifiable to limit the geographical scope of this study to the preferences of local people living in (or near the olive regions in) Portugal, Spain, Italy, and Greece. The outcome of the study would be an average individual WTP, and the aggregation of this amount – to indicate the overall preferences – should also be restricted to those regions in which individuals are influenced by the specific ‘olive measure’. Some agro-environmental policy measures, however, have a larger geographical impact and may influence the preferences of people in perhaps all European countries. For example, the abolition of the EU’s milk quota system, or trade liberalisation of agricultural products, will have effects far beyond national boundaries. In other words, the impact on agro-biodiversity of such far-reaching policy measures is making itself felt in most European regions, if not in all. This, of course, should be reflected in a wide geographical scope of a valuation study, as well as in the level of aggregation of individual preferences. Let us pause here and reflect, in SEAMLESS-terms, on the story so far. Preferences of individuals for changes in agro-biodiversity patterns (that are due to policy-derived changes in agricultural systems) can be monetised by valuation studies. These studies result in an economic indicator, namely average individual WTP, which can be defined by the amount of benefits associated with the agro-biodiversity resource under consideration. Monetary valuation takes place at the micro level – as it reflects individual preferences – but higher ‘levels of organisation’ can be obtained through aggregating individuals’ preferences to larger spatial scales. Which level of organisation (EU, nation, NUTS-2) is appropriate depends on the size of the population that is affected by the specific agricultural or agro-environmental policy measure that is considered. We are aware of the fact that the public in Northern Europe may attach non-use values to the olive agro-ecosystems in Southern Europe, and therefore, that not all the people in Northern Europe are fully indifferent.

6

312

C.M. van der Heide et al.

In SEAMLESS, indicators aid the translation of specific users’ policy questions ‘into something that SEAMLESS-IF can assess.’ (van Ittersum et al. 2008: 155). For that purpose, so-called indicator frameworks are developed to structure a broad range of indicators, to facilitate their interactive selection by the users, and to provide a balanced selection of indicators across the three dimensions of sustainable development (economic, social and environmental). Economic indicators are assessed through output variable from quantitative model components, in particular SEAMCAP (or CAPRI) and FSSIM (a detailed description of these two models is given in Chapter 11 and Chapter 5 of this volume). For applying monetary valuation techniques, it is important to know whether their methods (as described earlier in this chapter) can be smoothly and seamlessly integrated into the backbone model chain of SEAMLESS-IF that is formed by SEAMCAP, FSSIM and other quantitative models. This question poses a challenge for economists and SEAMLESS-experts alike to consider monetary valuation methods as a model component that possibly can be included in future versions of SEAMLESS-IF. In broad terms, SEAMLESS aims to provide a pan-European integrated database including comprehensive datasets on biophysical, farm, socio-economic and policy variables. Here, there is fair potential for benefit transfer as it applies a dataset developed for one particular use to an alternative situation. However, due to the fact that relatively few valuation studies have been done that focused primarily on agro-biodiversity, especially in the new members states, there is only a small database of values for transfer available. As a result, in the existing valuation surveys, the agro-biodiversity and landscape attributes valued are often very different, making it hard to find studies that have valued the same attributes as those that are considered in a new context or location (to which the estimates can be transferred). To fill the gaps in the database, more valuation studies on agro-biodiversity benefits need to be completed. This can be a Herculean and costly task, because determining the preferences of the general public for agro-biodiversity assets in a consistent and unified manner involves primary data collection. However, by integrating monetary values explicitly into SEAMLESS-IF, the potentially high benefits of agro-biodiversity can be demonstrated, and the pressures on its biological assets that are due to specific policy measures can be revealed more clearly.

Conclusions The relationship between agriculture and biodiversity is extraordinary complex and ambiguous. Agriculture not only depends on biodiversity, but it also has major impacts on the diversity of numerous living organisms and the natural systems in which they live, for example in areas where natural vegetation is completely cleared for crops. In this chapter, we focused from an economic perspective on the benefits of biodiversity in agriculture (agro-biodiversity). Because agro-biodiversity has characteristics of a public good, and the value expressed by its price in conventional markets is usually too low for optimal use and supply, monetary valuation methods

13 Economic Principles of Monetary Valuation in Evaluation Studies

313

can shed light on the benefits of agro-biodiversity, and on the preferences that individuals have for its assets. Generally speaking, the raison d’être of monetary valuation is concern about things that are valuable. After all, due to the public good aspects of agro-biodiversity, the notion that agro-biodiversity is as valuable as its price in the marketplace does not hold. However, the fact that a certain agro-biodiversity asset is not fully captured in a market, and hence has no price does not means that agro-biodiversity values are simply numbers plucked out of the air. This chapter has glanced briefly at a range of monetary approaches and methods for measuring biodiversity values in agricultural landscapes. It has also explored the question whether monetary valuation can be used for agricultural policy impact assessment. In more precise words, can monetary valuation facilitate the assessment of the impact that proposed agricultural or agro-environmental policy actions will have on specific agro-biodiversity assets? The conclusion that can be validly drawn is that monetary valuation gives insight into the preferences for agro-biodiversity that the general public holds. As such, it assesses from a purely anthropocentric viewpoint the impact of specific agricultural or agro-environmental policy measures on agro-biodiversity. Despite some controversies and ethical issues, the application of especially stated preference methods can be very useful in the design of future agricultural and agro-environmental policies; particularly if choice experiments or other forms of choice modelling techniques are used. Also within SEAMLESS monetary valuation can play an informative and supportive role. To that end, we believe that valuation methods, and the results of existing and future valuation surveys, should be linked to the quantitative models in SEAMLESS-IF. However, this integration of valuation model components in later version of SEAMLESS-IF is, at least from our perspective, rather complicated and therefore quite a challenge. Nevertheless, this challenge has the merit of being identifiable and potentially useful. Or, to put it in the hopeful words of the American philosopher Dennett (1995: 38): ‘Problems in science are sometimes made easier by adding complications.’

References Adamowicz, W., Boxall, P., Williams, M., & Louviere, J. (1998). Stated preference approaches for measuring passive use values: Choice experiments and contingent valuation. American Journal of Agricultural Economics, 80(1), 64–75. Adamowicz, W., Louviere, J., & Williams, M. (1994). Combining revealed and stated preference methods for valuing environmental amenities. Journal of Environmental Economics and Management, 26(3), 271–292. Bateman, I. J., & Turner, R. K. (1993). Valuation of the environment, methods and techniques: The contingent valuation method. In R. K. Turner (Ed.), Sustainable environmental economics and management; principles and practice (pp. 120–191). London/New York: Belhaven. Bergland, O. (1998). Can we value the landscape experience? (‘Kan vi sette en pris på landskapsopplevelsen?’) (in Norwegian). In E. Framstad & I.B. Lid (Eds.), The cultural landscape of agriculture; Management of environmental values (‘Jordbrukets kulturlandskap; forvaltning av miljøverdier’) (in Norwegian) (pp. 171–176). Oslo: Scandinavian University Press (‘Universitetsforlaget’).

314

C.M. van der Heide et al.

Boardman, A.E., Greenberg, D.H., Vining, A.R., & Weimer, D.L. (2006). Cost-benefit analysis; concepts and practice (3rd ed.). Upper Saddle River, NJ: Pearson/Prentice Hall. Carson, R. T. (2000). Contingent valuation: A user’s guide. Environmental Science and Technology, 34(8), 1413–1418. Clawson, M., & Knetsch, J. L. (1966). Economics of outdoor recreation. Baltimore: Johns Hopkins University Press. Cornes, R., & Sandler, T. (1996). The theory of externalities, public goods, and club goods (2nd ed.). Cambridge: Cambridge University Press. Dachary-Bernard, J. (2004). Multi-attribute valuation of landscape changes: An application of choice experiments method to the Monts d’Arree rural landscape in Brittany (France). Paper presented at the Agricultural Economics Society – Conference 2–4 April 2004, London, England. Daily, G. C. (1997). Introduction: What are ecosystem services? In G. C. Daily (Ed.), Nature’s services; societal dependence on natural ecosystems (pp. 1–10). Washington, DC: Island Press. Daily, G. C. (2000). Management objectives for the protection of ecosystem services. Environmental Science and Policy, 3, 333–339. Dalhuisen, J.M. (2002). The economics of sustainable water use; comparisons and lessons from urban areas. Dissertation, Tinbergen Institute Research Series No. 290, Vrije Universiteit, Amsterdam. De Blaeij, A.T. (2003). The value of a statistical life in road safety; stated preference methodologies and empirical estimates for the Netherlands. Dissertation, Tinbergen Institute Research Series No. 308, Vrije Universiteit, Amsterdam. Dennett, D. C. (1995). Darwin’s dangerous idea; evolution and the meanings of life. London: Penguin Books. Diamond, J. (1997). Guns, germs, and steel. New York/London: W.W. Norton. Diamond, P. A., & Hausman, J. A. (1993). On contingent valuation measurement of nonuse values. In J. A. Hausman (Ed.), Contingent valuation; a critical assessment (pp. 3–38). Amsterdam/ North-Holland: Elsevier. Diamond, P. A., & Hausman, J. A. (1994). Contingent valuation: Is some number better than no number? Journal of Economic Perspectives, 8(4), 45–64. Dirzo, R., & Raven, P. H. (2003). Global state of biodiversity and loss. Annual Review of Environment and Resources, 28, 137–167. Dryzek, J. S. (1990). Discursive democracy: Politics, policy and political science. New York: Cambridge University Press. Dubgaard, A., Bateman, I., & Merlo, M. (eds). (1994). Identification and valuation of public benefits from farming and countryside stewardship. Brussels: European Commission. Ehrlich, P. R., & Ehrlich, A. H. (1992). The value of biodiversity. Ambio, 21(3), 219–226. Ewel, K.C. (1997). Water quality improvement by wetlands. In G.C. Daily (Ed.), Nature’s services; societal dependence on natural ecosystems (pp. 329–344). Washington, DC/Covelo, CA: Island Press. Farber, S., & Griner, B. (2000). Using conjoint analysis to value ecosystem change. Environmental Science and Technology, 34(8), 1407–1412. Garrod, G. D., & Willis, K. G. (1999a). Economic valuation of the environment; methods and case studies. Cheltenham/Northampton, MA: Edward Elgar. Garrod, G. D., & Willis, K. G. (1999b). Methodological issues in valuing the benefits of environmentally sensitive areas. Journal of Rural Studies, 15(1), 111–117. Garrod, G. D., Willis, K. G., & Saunders, C. M. (1994). The benefits and costs of the Somerset Levels and Moors ESA. Journal of Rural Studies, 10(2), 131–145. Gowdy, J. M. (1997). The value of biodiversity: Markets, society, and ecosystems. Land Economics, 73(1), 25–41. Gowdy, J. M. (2004). The revolution in welfare economics and its implications for environmental valuation and policy. Land Economics, 80(2), 239–257. Habermas, J. (1984). The theory of communicative action. Boston: Beacon Press. Hanemann, W. M. (1994). Valuing the environment through contingent valuation. Journal of Economic Perspectives, 8(4), 19–43.

13 Economic Principles of Monetary Valuation in Evaluation Studies

315

Hanley, N., Mourato, S., & Wright, R. E. (2001). Choice modelling approaches: A superior alternative for environmental valuation? Journal of Economic Surveys, 15(3), 435–462. Hanley, N., Shogren, J. F., & White, B. (1997). Environmental economics in theory and practice. Houndmill/London: MacMillan. Hanley, N., Whitby, M., & Simpson, I. (1999). Assessing the success of agri-environmental policy in the UK. Land Use Policy, 16, 67–80. Heal, G. (2000). Nature and the marketplace; capturing the value of ecosystem services. Washington, DC/Covelo, CA: Island Press. Hodge, I., & McNally, S. (1998). Evaluating the environmentally sensitive areas: The value of rural environments and policy relevance. Journal of Rural Studies, 14(3), 357–376. Ison, S., Peake, S., & Wall, S. (2002). Environmental issues and policies. Harlow: Pearson Education. Jacobs, M. (1997). Environmental valuation, deliberative democracy and public decision making institutions. In J. Foster (Ed.), Valuing nature? Economics, ethics and environment (pp. 211–231). London: Routledge. Kahneman, D., Knetsch, J. L., & Thaler, R. H. (1990). Experimental tests of the endowment effect and the Coase theorem. Journal of Political Economy, 98(6), 1325–1348. Kremen, C., Williams, N. M., Bugg, R. L., Fay, J. P., & Thorp, R. W. (2004). The area requirements of an ecosystem service: crop pollination by native bee communities in California. Ecology Letters, 7(11), 1109–1119. Krumalova, V., Prazan, J., & Drlik, J. (2000). Valuation of selected public goods related to agriculture (‘Ohodnocení Vybraných Veřejných Tatků Pocházejících ze Zemědělství’) (in Czech). Prague: Research Institute for Agricultural Economics, Report for the Ministry of Agriculture of the Czech Republic. Kubickova, S. (2004). Contingent valuation of the landscape amenity function of agriculture in the Protected Landscape Area White Carpathians. In M. Ščasný & J. Melichar (Eds.), Development of the Czech society in the European Union. Part V: Nonmarket valuation methods in environmental area (pp. 259–268). Prague: Matfyzpress, Charles University Environment Center. Levin, S. A. (1999). Fragile dominion; complexity and the commons. Cambridge, MA: Perseus. Loomis, J., Kent, P., Strange, L., Fausch, K., & Covich, A. (2000). Measuring the total economic value of restoring ecosystem services in an impaired river basin: Results from a contingent valuation survey. Ecological Economics, 33(1), 103–117. Louviere, J. J., Hensher, D. A., & Swait, J. D. (2000). Stated choice methods; analysis and applications. Cambridge: Cambridge University Press. McNeely, J. A., Gadgil, M., Leveque, C., Padoch, C., & Redford, K. (1995). Human influences on biodiversity. In V. H. Heywood (Ed.), Global biodiversity assessment (pp. 711–821). Cambridge: Cambridge University Press. Melichar, J., & Ščasný, M. (2004). Introduction to non-market valuation methods and critical review of their application in the Czech Republic. In M. Ščasný & J. Melichar (Eds.), Development of the Czech society in the European Union. Part V: Nonmarket valuation methods in environmental area (pp. 43–94). Prague: Matfyzpress, Charles University Environment Center. Mooney, H. A., & Ehrlich, P. R. (1997). Ecosystem services: A fragmentary history. In G. C. Daily (Ed.), Nature’s services; societal dependence on natural ecosystems (pp. 11–19). Washington, DC: Island Press. Navrud, S. (2001). Comparing valuation exercises in Europe and the United States – challenges for benefit transfer and some policy implications. In OECD. Valuation of biodiversity studies; selected studies (pp. 63–77). Paris: OECD. Navrud, S., & Ready, R. (eds). (2007). Environmental value transfer: Issues and methods. Dordrecht: Springer. NOAA. (1993). Report of the NOAA panel on contingent valuation. Federal Register, 58(10), 4601–4614. Nunes, P. A. L. D., & van den Bergh, J. C. J. M. (2001). Economic valuation of biodiversity: Sense or nonsense? Ecological Economics, 39(2), 203–222.

316

C.M. van der Heide et al.

OECD. (2001). Environmental indicators for agriculture. Methods and results (Vol. 3). Paris: OECD. OECD. (2002). Handbook of biodiversity valuation; a guide for policy makers. Paris: OECD. OECD. (2005). Policy Brief; OECD Observer. Paris: OECD. Ostrom, E. (2000). Collective action and the evolution of social norms. Journal of Economic Perspectives, 14(3), 137–158. Pearce, D., Atkinson, G. & Mourato, S. (2006). Cost-Benefit Analysis and the environment; recent developments. Paris: OECD. Perrings, C. (1995). The economic value of biodiversity. In V. H. Heywood (Ed.), Global biodiversity assessment (pp. 823–914). Cambridge: Cambridge University Press. Powe, N. A. (2007). Redesigning environmental valuation: Mixing methods within stated preference techniques. Cheltenham/Northampton, MA: Edward Elgar. Powe, N. A., Garrod, G. D., & McMahon, P. L. (2005). Mixing methods within stated preference environmental valuation: choice experiments and post-questionnaire qualitative analysis. Ecological Economics, 52(4), 513–526. Pruckner, G. J. (1995). Agricultural landscape cultivation in Austria: An application of the CVM. European Review of Agricultural Economics, 22(2), 173–190. Ready, R., & Navrud, S. (2006). International benefit transfer: Methods and validity tests. Ecological Economics, 60(2), 429–434. Roe, B., Boyle, K. J., & Teisl, M. F. (1996). Using conjoint analysis to derive estimates of compensating variation. Journal of Environmental Economics and Management, 31(2), 145–159. Rosenberger, R. S., & Stanley, T. D. (2006). Measurement, generalization, and publication: Sources of error in benefit transfers and their management. Ecological Economics, 60(2), 372–378. Sagoff, M. (2000). Environmental economics and the conflation of value and benefit. Environmental Science and Technology, 34(8), 1426–1432. Santos, J. M. L. (1998). The economic valuation of landscape change. Theory and policies for land use conservation. Cheltenham/Northampton, MA: Edward Elgar. Shechter, M. (2000). Valuing the environment. In H. Folmer & H. L. Gabel (Eds.), Principles of environmental and resource economics; a guide for students and decision-makers (2nd ed., pp. 72–103). Cheltenham/Northampton, MA: Edward Elgar. Simpson, R. D. (2007). David Pearce and the economic valuation of biodiversity. Environmental and Resource Economics, 37(1), 91–109. Smith, V. K. (1992). Arbitrary values, good causes, and premature verdicts. Journal of Environmental Economics and Management, 22(1), 71–89. Smith, V. K. (2004). Fifty years of contingent valuation. In T. Tietenberg & H. Folmer (Eds.), The International Yearbook of environmental and resource economics 2004/2005 (pp. 1–60). Cheltenham/Northampton, MA: Edward Elgar. Spash, C. L., & Carter, C. (2002). Environmental valuation methods in rural resource management. In F. Brouwer & J. van der Straaten (Eds.), Nature and agriculture in the European Union: New perspectives on policies that shape the European countryside (pp. 88–114). Cheltenham/ Northampton, MA: Edward Elgar. Spash, C. L., & Vatn, A. (2006). Transferring environmental value estimates: Issues and alternatives. Ecological Economics, 60(2), 379–388. Tilman, G.D., Duvick, D.N., Brush, S.B., Cook, R.J., Daily, G.C., Heal, G.M., Naeem, S., & Notter, D. (1999). Benefits of biodiversity (Task Force Report No. 133). Ames, IA: Council for Agricultural Science and Technology. Tisdell, C. A. (1991). Economics of environmental conservation; economics for environmental and ecological management. Amsterdam/London/New York/Tokyo: Elsevier. Turner, R. K., Pearce, D., & Bateman, I. (1994). Environmental economics; an elementary introduction. London: Harvester Wheatsheaf. van den Bergh, J. C. J. M., Button, K. J., Nijkamp, P., & Pepping, G. C. (1997). Meta-analysis in environmental economics. Dordrecht/Boston/London: Kluwer.

13 Economic Principles of Monetary Valuation in Evaluation Studies

317

van der Heide, C.M., van den Bergh, J.C.J.M., & van Ierland, E.C. (2003). Towards an ecologicaleconomic theory of nature policy. In S. Dovers, D.I. Stern, & M.D. Young (Eds.), New dimensions in ecological economics (pp. 121–145). Cheltenham/Northampton, MA: Edward Elgar. van Ittersum, M. K., Ewert, F., Heckelei, T., Wery, J., Alkan Olsson, J., Andersen, E., et al. (2008). Integrated assessment of agricultural systems – a component-based framework for the European Union (SEAMLESS). Agricultural Systems, 96(1–3), 150–165. Vatn, A. (2004). Environmental valuation and rationality. Land Economics, 80(1), 1–18. Vatn, A., & Bromley, D. W. (1994). Choices without prices without apologies. Journal of Environmental Economics and Management, 26, 129–148. Wilson, M. A., & Hoehn, J. P. (2006). Valuing environmental goods and services using benefit transfer: The state-of-the-art and science. Ecological Economics, 60(2), 335–342. Wilson, M. A., & Howarth, R. B. (2002). Discourse-based valuation of ecosystem services: Establishing fair outcomes through group deliberation. Ecological Economics, 41(3), 431–443. Zechmeister, H. G., Schmitzberger, I., Steurer, B., Peterseil, J., & Wrbka, T. (2003). The influence of land-use practices and economics on plant species richness in meadows. Biological Conservation, 114(2), 165–177.

Index

A African, Caribbean and Pacific (ACP) countries, 259 Agenda 2000, 117, 189, 260 Agricultural landscape, 18, 133–154, 297, 300, 303–306, 313 Agricultural management module of FSSIM (FSSIM-AM), 110, 114, 119–123, 125, 126, 212, 229 Agricultural policy, 6, 13, 116, 117, 146, 160, 181, 204, 229, 258, 269, 270, 291, 297, 306, 313 Agricultural production and externalities simulator (APES), 2–4, 6, 66–68, 81, 83–88, 90–103, 111, 119, 120, 122, 123, 125, 126, 181, 212, 229, 246, 248, 249, 261, 263, 265–268, 271, 272 Agri-environmental zonation (AEnZ), 4, 6, 159, 163, 179–181 Agri-environmental zone, 111, 115, 163, 181–183, 211 Agro-biodiversity, 13, 297–301, 304, 306, 310–313 Agro-management, 4, 67, 69–70, 83–86, 93, 96, 97, 125, 126, 129, 248, 290 Agromanagement configuration generator (ACG), 96–97 Application programming interface (API), 93, 218 B Baseline, 2, 117, 118, 127–129, 258, 264, 266, 268, 282 Biodiversity, 13, 15, 18, 28, 29, 40, 47, 49, 50, 123, 145–148, 160, 161, 163, 295–301, 304, 306, 307, 309–313 Bio-economic model, 6, 141, 182, 266

Biogeographical regions map of Europe (BRME), 163 Biomass, 73, 74, 76–80 Biophysical typology, 6, 159–184 Birds Directive, 161 C Calibration, 71, 102, 111, 112, 115, 118, 119, 126, 127, 129, 131, 248, 258, 263–266, 268, 269 Choice modelling (CM), 40, 66, 302, 304, 313 Climate zone, 163, 174, 176, 177, 182, 183 Common agricultural policy (CAP), 13, 116, 117, 160, 181, 229, 258, 291 Common Agricultural Policy Regional Impact (CAPRI) model, 2–4, 6, 114, 118, 127, 181, 212, 229, 257–272, 280, 312 Common Market Organization (CMO), 113 Computable general equilibrium model (CGE model), 3, 262 Computer imagery, 134, 138, 153, 154 Contingent valuation method (CVM), 40, 300–302 CORINE land cover (CLC), 139, 167, 170 Cost-benefit analysis (CBA), 40, 306, 308, 309 Cost of policy inaction (COPI), 39 Create, retrieve, update, delete (CRUD), 218 Crop Model Library (CropMP), 73 Crucial institutional aspect (CIA), 44, 46, 49, 51–56 Cultural heritage, 15, 18, 19 Current activity (CA), 121 D Database management system (DBMS), 269 Data transfer object (DTO), 217

319

320 Decision support system (DSS), 154, 238, 239, 253 Denitrification-decomposition (DNDC) model, 264, 266, 268 Design-by-Contract (DBC), 91 Digital elevation model (DEM), 165, 167 Digital map, 138 Digital surface model (DSM), 138 Digital terrain model (DTM), 138, 166 E Economic accounts for agriculture (EAA), 262 Environmentally sensitive area (ESA), 305, 307 Environmental stratification (EnS), 6, 162, 164, 165, 176 Environmental zone (EnZ), 6, 164–166, 168, 171–176, 183 Erosion, 79–80, 109, 128, 212, 263, 295, 304 European Dairy Industry Model (EDIM), 264 European Livestock Policy Evaluation Network (ELPEN), 189 European size unit (ESU), 190 European soil database (ESDB), 165–167, 169, 170, 175, 183, 184 Ex-ante assessment, 38, 56, 57, 109, 116, 131, 239, 242 Ex-ante policy assessment, 37, 41–43 Ex-post, 19, 23, 24, 37, 39, 64, 109, 266, 306, 307 Extended FAST (E-FAST), 101 Externalities, 17, 30, 63–103, 109, 110, 160, 182, 229, 246, 261, 264, 266–268, 272, 296

Index Global Trade Analysis Project (GTAP), 3, 212, 264 Good agricultural practice (GAP), 161 Graphical user interface (GUI), 65, 83, 97, 121, 122, 125, 215, 216, 220, 238, 242, 269, 288 Graphic data display (GDD), 97–98 H Habitats Directive, 161 Hedonic pricing method (HPM), 300, 307 High nature value, 160, 161, 164, 189 I Indicator, 2, 4, 5, 12, 13, 15–24, 27, 28, 30, 32, 33, 37, 41, 45, 52–55, 57, 66, 100, 109, 114, 127, 137, 152, 153, 166, 189, 191–194, 203, 204, 208, 212, 214, 225, 227–231, 240, 242, 243, 245, 246, 248, 258, 260, 262, 266–269, 272, 280–283, 288–291, 311, 312 Institution, 2, 38, 40–45, 56, 245, 270, 271, 275, 278, 283, 288 Institutional analysis, 5, 38, 41–43, 45, 57 Institutional economics, 37 Institutional policy assessment, 37–57 Integrative modeller, 2, 227, 228, 243, 245, 249, 253, 254, 269, 276, 280, 281, 287, 290 Integrative modeller GUI (IM GUI), 216 K Knowledge manager, 215, 216, 221–222

F Farm Accountancy Data Network (FADN), 4, 17, 24, 25, 27, 30, 31, 33, 109, 114, 121, 122, 129, 180, 181, 193–204, 264–266 Farm structure survey (FSS), 111, 181, 194 Farm system simulator (FSSIM), 6, 110, 229, 258 Farm typology, 110–113, 123, 126, 129, 189, 190, 193, 264 Fourier amplitude sensitivity test (FAST), 101 G G20, 127, 128 General Algebraic Modeling System (GAMS), 5, 118, 124, 269, 270 Geographic information system (GIS), 134, 137, 138, 141, 153 Global agro-ecological zonation (GAEZ), 163

L Landscape, 3, 6, 12–21, 28, 40, 110, 123, 133–154, 160, 161, 163, 182, 189, 271, 297, 300, 301, 303–307, 309, 312, 313 Landscape model, 135, 138 Latin hybercube sampling (LHS), 101 Less favoured area (LFA), 190, 192, 193, 197–198, 200–204 Linear programming (LP), 265, 266 M Market model, 5, 114, 118, 127, 128, 258–261, 263 Mathematical programming module of FSSIM (FSSIM-MP), 30, 110, 113–120, 123, 126, 129, 212, 229

Index Model component explorer (MCE), 93, 96, 100 Model integration, 207–232 3D Modelling, 77 Modelling framework, 6, 38, 56, 65, 68, 83, 86, 88, 93–95, 98, 102, 124–125, 183, 208, 216, 271 Model parameter editor (MPE), 86, 97 Modified universal soil loss equation (MUSLE), 80 Monitoring of agriculture with remote sensing (MARS), 162, 176, 177, 183 Multifunctional agriculture (MFA), 11, 16 Multifunctionality, 5, 11–34, 304 N Nitrate Directive, 5, 38, 44, 47, 49–52, 55–57, 242 Nitrate leaching, 80, 109, 127, 128, 212 Non-linear programming (NLP) model, 258 O Object oriented programming (OOP), 93 Object relational mapping (ORM), 216, 218, 219 Ontology, 4, 88, 89, 96, 100, 103, 124, 125, 208, 211–214, 216, 218, 221, 222, 225, 226, 231, 232, 242, 245, 250, 262, 264, 288 Open modelling interface (Open MI), 5, 6, 125, 216, 223 Organic carbon topsoil (OCTOP), 165, 167–169, 171–175, 182–184 P Parameter estimator (PE), 100–101, 216, 243 Partial general equilibrium model, 212, 226, 258 Pedotransfer function (PTF), 81, 82 Pedo transfer rules database (PTRDB), 166 Percent absolute deviation (PAD), 126, 127 Permanent crop, 161, 200, 201 Pesticide, 28, 43, 48, 63, 69, 71–72, 83, 109, 121, 128, 212 Photosynthetic active radiation (PAR), 76 Plant disease, 74 Policy expert, 216, 227, 228, 269, 278 Policy expert GUI (PE GUI), 216 Policy scenario, 128, 282 Positive mathematical programming (PMP), 112, 118, 126, 127, 263 Post-modelling, 2, 56, 57, 225, 230–231, 243, 249, 254, 269, 279, 280, 283

321 Preference method, 40, 299, 300, 302 Pre-modelling, 2, 56, 225, 227, 243, 244, 249, 269, 279, 282, 283, 289, 290 Principal component analysis (PCA), 28, 30, 166, 168, 169, 182 Procedure for institutional compatibility assessment (PICA), 3, 5, 38, 44–45, 125, 126, 243–246, 250 Production enterprise generator (PEG), 111, 120, 122–123 Production technique generator (PTG), 111, 120, 123 Protected landscape area (PLA), 305 Prototyping, 239 Public good, 12, 296, 312, 313 Q Qualitative evaluation, 15, 16, 20–21, 27, 28, 30–33, 45, 53–55, 192, 248, 310 Quantitative evaluation, 248 R Radiation use efficiency (RUE), 73, 74 Regional typology, 6, 189–205 Representational state transfer (REST), 218 Resource description framework (RDF), 209 Revised universal soil loss equation (RUSLE), 80 Rich internet application (RIA), 219 Rural development, 5, 12, 21, 44, 46, 48, 116, 160, 161, 193, 267, 268, 276 Rural landscape, 13, 14, 20, 133 Rural viability, 5, 18, 27, 29, 161, 189 S Scenario, 2, 6, 39, 56, 57, 70, 117, 118, 126–129, 134, 135, 137, 139, 145–147, 149–153, 211, 214, 238, 240, 242–244, 248, 249, 261, 264, 269, 271, 282, 289, 290, 300, 303, 305–309, 311 SeamGUI, 215, 216, 220, 221, 225, 227–231 SEAMLESS landscape explorer (SLE), 3, 138, 140–142, 145, 146, 152, 153 SEAMLESS version of CAPRI (SEAMCAP), 118, 229, 257, 258, 260–262, 264, 266, 268, 270, 271, 312 SeamPRES, 220, 221, 225, 229–231 SeamServer, 215–218, 220, 229, 231 Seamzone, 163, 173–176, 178–184 Sensitivity analysis (SA), 101, 129, 283 Set-aside, 113, 116, 117, 127, 267

322 Simulation output evaluator (SOE), 96, 98–99 Software architecture, 4–6, 86–87, 208, 215, 217 Soil erosion, 68, 79–80, 109, 128, 212, 263, 295 Soil geographical database of Eurasia (SGDBE), 166 Soil information (SINFO), 81, 82, 164, 170, 179, 183, 184 Soil organic matter (SOM), 78, 79, 127 Soil typological unit (STU), 180 Sustainability, 13, 16, 41, 42, 66, 212, 216, 223, 239, 240, 245, 277, 284, 310 Sustainability Impact Assessment of the Forestry-Wood Chain (EFORWOOD), 223, 284 Sustainability impact assessment tools for environmental, social and economic effects of multifunctional land use in Europe (SENSOR), 137, 223, 284 Sustainable development, 1, 5, 12, 13, 16, 41, 42, 152, 310, 312 T Technical coefficient generator (TCG), 111, 120, 123

Index Temperature in paddy-rice simulation (TRIS), 74 Travel cost method (TCM), 40, 299–300 U User centred design (UCD), 217 User forum (UF), 241, 284 V Virtual reality, 112, 134, 143, 144, 153 Visualisation, 3, 6, 134, 135, 137–139, 141, 145, 152, 154, 269 W Water Erosion Prediction Project (WEPP), 70 Water quality, 20, 28, 49, 299, 306 Web-based client server, 221 Willingness to pay (WTP), 300–302, 304–307, 310, 311 World Trade Organisation (WTO), 127