DEVELOPMENTS IN EARTH SURFACE PROCESSES
1. PALEOKARST: A SYSTEMATIC STUDY AND REGIONAL REVIEW P. BOSAK, D. FORD, J. GLAZEK and I. HORACEK (Editors) [OUT OF PRINT]
2. WEATHERING, SOILS & PALEOSOLS I.P. MARTINI and W. CHESWORTH (Editors)
3. GEOMORPHOLOGICAL RECORD OF THE QUATERNARY OROGENY IN THE HIMALAYA AND THE KARAKORAM JAN KALVODA (Editors) [OUT OF PRINT]
4. ENVIRONMENTAL GEOMORPHOLOGY M. PANIZZA
5. GEOMORPHOLOGICAL HAZARDS OF EUROPE C. EMBLETON and C. EMBLETON-HAMANN (Editors)
6. ROCK COATINGS R.I. DORN
7. CATCHMENT DYNAMICS AND RIVER PROCESSES C. GARCIA and R.J. BATALLA (Editors)
8. CLIMATIC GEOMORPHOLOGY M. GUTIERREZ
9. PEATLANDS: EVOLUTION AND RECORDS OF ENVIRONMENTAL AND CLIMATE CHANGES L.P. MARTINI, A. MARTINEZ CORTIZAS and W. CHESWORTH (Editors)
10. MOUNTAINS WITNESSES OF GLOBAL CHANGES RESEARCH IN THE HIMALAYA AND KARAKORAM: SHARE-ASIA PROJECT RENATO BAUDO, GIANNI TARTARI and ELISA VUILLERMOZ (Editors)
11. GRAVEL-BED RIVERS VI: FROM PROCESS UNDERSTANDING TO RIVER RESTORATION HELMUT HABERSACK, HERVE PIEGAY and MASSIMO RINALDI (Editors)
Developments in Earth Surface Processes, 12
THE CHANGING ALPINE TREELINE: THE EXAMPLE OF GLACIER NATIONAL PARK, MT, USA
Edited by David R. Butler Department of Geography, Texas State University-San Marcos
George P. Malanson Department of Geography, University of Iowa
Stephen J. Walsh Department of Geography, University of North Carolina at Chapel Hill
Daniel B. Fagre US Geological Survey Northern Rocky Mountain Science Center, Glacier National Park, Montana
Amsterdam – Boston – Heidelberg – London – New York – Oxford Paris – San Diego – San Francisco – Singapore – Sydney – Tokyo
Elsevier Radarweg 29, PO Box 211, 1000 AE Amsterdam, The Netherlands Linacre House, Jordan Hill, Oxford OX2 8DP, UK First edition 2009 Copyright Ó 2009 Elsevier B.V. All rights reserved No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means electronic, mechanical, photocopying, recording or otherwise without the prior written permission of the publisher Permissions may be sought directly from Elsevier’s Science & Technology Rights Department in Oxford, UK: phone (+44) (0) 1865 843830; fax (+44) (0) 1865 853333; email:
[email protected]. Alternatively you can submit your request online by visiting the Elsevier web site at http://elsevier.com/locate/permissions, and selecting Obtaining permission to use Elsevier material Notice No responsibility is assumed by the publisher for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. Because of rapid advances in the medical sciences, in particular, independent verification of diagnoses and drug dosages should be made British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library Library of Congress Cataloging-in-Publication Data A catalog record for this book is available from the Library of Congress ISBN: 978-0-444-53364-7 ISSN: 0928-2025 For information on all Elsevier publications visit our website at books.elsevier.com Printed and bound in Hungary 09 10 10 9 8 7 6 5 4 3 2 1
Working together to grow libraries in developing countries www.elsevier.com | www.bookaid.org | www.sabre.org
This book is dedicated with love to our families and to the preservation of Glacier National Park.
CoNTENTs
Editorial Foreword
xiii XV
Preface Acknowledgments
xvii
List of Acronyms
xix
List of Contributors
xxi
1.
Introduction: Understanding the Importance of Alpine Treeline Ecotones in Mountain Ecosystems
1
Daniel B. Fagre 1.
2.
3· 4-
2.
Introduction Importance of High-Elevation Mountain Research Recent Changes to t he Alpine Areas of Glacier National Park Factors Affecting the Alpine Treeline Ecotone
Pattern- Process Relations in the Alpine and Subalpine Environments: A Remote Sensing and GIScience Perspective
1
2
3 6
11
Stephen ). Walsh, Daniel G. Brown, Christine A. Geddes, Dani el ). Weiss, Sean McKnight, Evan S. Hammer, and Julie P. Tuttle 1. 2.
3·
4·
Introduction Study Area 2.1 General setting Geographic areas of research emphasis 2.2 Background and Context Remote sensin g approaches 3-1 Field approaches 3-2 GIS representation 3-3 Data and Methods Medium-grained remote sensing 4-1 High-resolution OEM creation 4-2
11 12 12 14 16 17 18 19 20 20
21
vii
viii
Contents
5·
6.
Selected Applications 5.1 Pixel versus object classification of vegetation 5.2 Linear mixture modeling
24 24 28
S-3 Pattern metrics Conclusions
29 31
3· Ecotone Dynamics: lnvasibility of Alpine Tundra by Tree Species from the Subalpine Forest
35
George P. Malanson, Daniel G. Brown, David R. Butler, David M. Ca irns, Daniel B. Fagre, and Stephen ]. Walsh 1. 2.
3·
4·
s.
6.
Introduction 1.1 Plant's eye view Seeds to Seedlings in Open Tundra Dispersal 2.1 2.2 Protected sites Annual weather 2.3 Seedlin gs Coarse sca le climate 3-1 Endogenous climate modification 3-2 Microclimate 3·3 Soil 3-4 Tree or Krummholz Form Facilitation or Inhibition? Pattern and process 5.1 Conclusion
4· Geomorphic Patterns and Processes at Alpine Treeline
35 36 37 38 40 43 43 44 46 48 50 51 54 55 56
63
David R. Butler, George P. Malanson, Lynn M. Resler. Stephen ]. Walsh, Forrest D. Wilkerson, Ginger L. Schmid, and Carol F. Sawyer 1. 2.
3·
4·
Introduction Coarse-Scale Processes 2.1 Snow avalanches as treeli ne disturbance agents 2.2 Debris flows as treeline disturbance agents Medium-Scale Processes 3.1 Turf-banked terraces Eolian processes at treeline 3.2 Fine-Scale Processes and Landforms 4.1 Turf exfoliation 4.2 Boulders
63 65 65 67 68
69 71 73 73 74
ix
Contents
5· 6.
s.
4· 3 Needle-i ce pans 4-4 Frost heaving and churning Addit ional Comments on t he Possible Role of Animals at Treeline Conclusions
Environmental Controls on Turf-Banked Terraces Daniel J. Weiss, St ephen J. Walsh, Sean A. McKnight, and Evan S. Hammer 1. 2.
3· 4-
5· 6. 7-
Introduction Background Study Site Methods Results Discussion Conclusion s
6. Soils and Pedogenesis at Alpine Treeline
76 79 81 81
ss 85 86 90 92 97 102 105
107
Ginger L. Schmid, David R. Butler, George P. Malanson, and Lynn M. Resler 1. 2.
3-
4·
7-
Int roduction 1.1 Background Field Data Co llection Soil pit data at Lee Ridge and White Calf Mountain 2.1 Soil penetrability and compression on Turf-banked terrace 2.2 treads and risers Results Soil pit data from tree fi ngers and adjacent tundra 3 -1 Effective soil depth and soil compaction data 3-2 Discussion and Conclusions
Canopy Structure in the Krummholz and Patch Forest Zones
107 108 109 109 111 114 114 115 117
119
Evan S. Hammer and Stephen ]. Walsh 1. 2.
3·
Introduct ion Background and Context 2.1 Alpine treeline 2. 2 Leaf area index Study Area Lee Ridge 3-1 Apikuni Cirque 3-2 Cataract Creek 3-3
120 121 121
123 124 125 126 126
Contents
X
Preston Park Baring Basin Scenic Point East Flattop Mountain 3-7 Data and Methods Field data collection 4-1 Digital elevation model and terra in analysis 4-2 Vecto r data sets 4-3 Data processing 4-4 Analysis 3D visualization 5-1 Spatial analysis 5-2 Biophysica l analysis 5-3 Topo-climatic variables 5-4 Results and Discussion Internal canopy structure 6.1 6.2 Between patch structure Conclusions 3-4 3-5 3-6
4·
5·
6.
7·
8 . A Markov Analysis of Tree Islands at Alpine Treeline
127 127 127 127 128 128 128 130 130 132 133 133 139 143 143 144 145 146
151
Lynn M. Resler and Mark A. Fonstad 1.
2.
3·
4·
5·
Introduction Methods 2.1 Markov cha in analysis Embedded Markov chains 2.2 First-order Markov chains 2.3 Results Spatial sequence of coni fer establishment 3-1 Stability at treelin e 3-2 Discussion Establishment characteristics 4-1 Treelin e stability 4-2 Conclusions
151 154 154 157 157 158 158 158 159 160 161 163
9· Modeling Feedback Effects on Linear Patterns of Subalpine Forest Advancement Matthew F. Bekker and George P. Malanson 1.
Introduction 1.1 Treeli ne patterns 1.2 Explanations of treeline pattern Modeling treeline location and dynamics 1.3
167 168 169 170
Contents
10.
xi
2.
Methods FORSKA 2.1 Parameterization 2.2 Modifications 2.3 Site quality 2.4 Simulations 2.5 Results 3· Di scussion 4· Gap models and t reeline environments 4·1 Effec ts of light and mortali ty 4·2 Conclusions 5· Acknowledgments
172 172 172 175 176 178 180 182 182 183 185 187
The Future of Treeline
191
David R. Butler, George P. Malanson, and Stephen j. Walsh
Index
195
EDITORIAL FOREWORD
Ever since David Butler asked me more than three decades ago if he could do his master’s thesis in Glacier National Park, he has been quite devoted to detailed assessments of this scenic alpine terrain. He has even convinced many of his students and numerous colleagues from other institutions to accompany him into the wilder parts of the park, and amazingly enough, none were ever savaged by the famously wild grizzly bears there. Fear of the great griz did not seem to ever slow down their efforts much, in spite of not being able to camp out in the Park, and in fact may have even accelerated their work a time or two in order to avoid certain confrontations with a top predator. Perhaps the potential of being eaten wonderfully focused their minds, because in any case they have produced a useful and important book that allows better understandings of the potential for change in this scenic landscape. Global warming threatens many places in the world in various ways, especially the rise in sea levels worldwide as ice cap and mountain glaciers melt away, or the increasing desertification through declining rainfall and increased drought cycles. In Glacier National Park, especially, the potential loss of the very ice masses that give the Park its name is the most serious aspect of an ongoing trend. In fact over 110 of the 150 glaciers and snow/ice fields of the region have already disappeared over the past century, and to date the remaining 40 glaciers and snowfields continue to decline. But hand in hand with these most obvious changes, the more subtle or hidden aspects of changing vegetation are also occurring. As the treeline, or the so-called timberline, advances upward into the tundra of Glacier National Park in coming decades, some of the great colorful vistas of parts of the park may decline somewhat. This book establishes some important ground rules and methodologies for studying the slowly changing biogeography of the region. The seventeen authors of this volume, under the capable directions of David Butler, George Malanson, and Steve Walsh, and guided by Dan Fagre, have succeeded in elucidating a variety of climatic, geomorphologic, and edaphic controls over the woody vegetation in Glacier National Park. Some of this research may serve as a guideline to coming changes in other montane regions of the world. Furthermore, similar changes would have occurred after the many other global warming times in the Quaternary, such as the immediate post-Pleistocene times in the early Holocene, and again after the Little Ice Age in the late Holocene, so that possible future comparisons by others of this work to those prior events will no doubt be instructive. The authors are to be congratulated for their effort in establishing a good baseline from which future work may spring. John F. Shroder Jr. Editor-in-Chief xiii
PREFACE
Glacier National Park, Montana, was established in 1910 and represents one of the finest examples of formerly glaciated alpine scenery anywhere in the world. The park, along with its Canadian neighbor, Waterton Lakes National Park in Alberta, comprises the Waterton-Glacier International Peace Park, the first such kind of park on Earth. The importance of the park as a scenic and scientific preserve is seen in the designations of Waterton Glacier as an International Biosphere Reserve and a United Nations World Heritage Site. This volume brings together the scientific work of a number of individuals who have been working, individually and collaboratively, on aspects of the alpine treeline ecotone (ATE) in Glacier National Park (GNP) for several decades. The editors of the volume alone collectively number nearly a century of experience working in Glacier Park. The impetus for this current volume was a cooperative research agreement between the US Geological Survey (USGS), through the monumental efforts of Dan Fagre, and the Mountain GeoDynamics Research Group, comprised of David Butler, George Malanson, and Steve Walsh, with contributions from Dan Brown and David Cairns. The research agreement focused on the invasibility of alpine tundra in GNP, with work carried out over the period 1998–2003. A great deal of work and years of research by all parties involved preceded the agreement, and work continues to the present as well. This current volume stands as a testament to these bodies of work, but particularly the work funded by the USGS through the aforementioned cooperative research agreements. The book begins with an introduction to the alpine treeline ecotone by Fagre, who discusses the significance of the ATE as a resource and as a scientific focus of study in an era of increasing global evidence of climate change. He presents the variety of topics and scales of analyses at which the research funded by the USGS has been carried out and sets the stage for the individual chapters that follow. Chapter 2, by Walsh and others, illustrates the array of remote sensing and geographic information science (GISci) techniques we have applied to further our understanding of spatial patterns and processes at the ATE. They describe a variety of sensor platforms used to study the ATE over a range of scales, fieldwork carried out to collect data on soils–vegetation–spectral relationships and statistical approaches utilized to characterize the ATE. The dynamic nature of the ATE is addressed in Chapter 3 by Malanson et al., with the question of, if and where treeline can advance into adjacent tundra, a primary focus. They use a ‘‘plant’s eye’’ approach to examine issues of site suitability for seedling establishment and survival, the form of the twisted wood, krummholz, and the ultimate likelihood of ecotone movement. The powerful legacy of extensive Pleistocene glaciation in GNP created a steep, dynamic geomorphic environment upon which the ATE is emplaced. Butler et al. examine the range and scale of geomorphic landforms and processes, past and xv
xvi
Preface
present, in Chapter 4, that may inhibit or facilitate treeline establishment and upward migration. They show that a wide array of processes operative at coarse, medium, and fine scales create a complex mosaic of geomorphic influences at treeline. One key component of this mosaic, turf-banked terrace treads and risers, is the specific focus of the subsequent Chapter 5 by Weiss et al., who illustrate spatial variations in the morphometry of landforms that provide key habitat for tree seedling establishment in the adjacent alpine tundra. In Chapter 6, Schmid et al. examine the nature of soils at the ATE. Their work focuses on the similarities and differences between soils under krummholz patches and in the surrounding alpine tundra. Echoing conclusions from the geomorphology chapter, they emphasize the significance of microsite variation in soils as related to seedling establishment. They also note that some soils at the ATE may be much older than believed. Hammer and Walsh discuss the structure of the canopy in krummholz and adjacent patch forest zones in Chapter 7. They illustrate the edge effect in describing differences in the canopy structure of krummholz that exist between interior and edge sites. Their results illustrate the importance of slope position, curvature, and wind in shaping the canopy structure of the ATE. Chapter 8 continues to examine krummholz fingers at a fine spatial scale. Resler and Fonstad use Markov analysis to examine the question of the potential role of positive plant interactions in conifer establishment at treeline. They describe geographic variability among study sites in terms of patch occupancy and treeline stability. The role of Pinus albicaulis as a significant initial treeline colonizer is emphasized, as is the fact that establishment is not statistically predictable within a patch. Bekker and Malanson use feedback modeling in Chapter 9 to examine the effects of positive feedback between ecotone pattern and tree establishment at the ATE, using a computer-simulation model validated against a dendrochronological reconstruction. They illustrate that the role of feedback varies over time and discuss how to improve feedback models to better represent ATE conditions and movement in the future. In the final chapter (Chapter 10), the leaders of the Mountain GeoDynamics Research Group consider the ramifications of the results illustrated in this volume. They look ahead with some trepidation to what the ATE in GNP may look like in the future and offer concluding thoughts.
ACKNOWLEDGMENTS
A volume such as this is the culmination of the efforts of many agencies and individuals over the course of many years. We take this opportunity to express our appreciation to those who have believed enough in us to provide us with the funding necessary to support our work and to those who have aided us in many ways. Funding for our work over the years at treeline in Glacier National Park has come from a variety of sources, including the following: • Cooperative Research Agreements between the US Geological Survey and Brown (99CRAG0035), Butler (99CRAG0032), Cairns (99CRAG0026), Malanson (99CRAG0030), and Walsh (99CRAG0034); • a National Science Foundation Small Grant for Exploratory Research (G.P. Malanson, S.J. Walsh, D.R. Butler, and D.B. Fagre, Co-Principal Investigators, Grant Number 0234018); • National Science Foundation grants to Walsh (NSF SES-9111852), Butler (NSF SES-9109837), and Malanson (NSF SES-9111853, SBR-9714347 [with D.G. Brown], BCS-9709810, and BCS 0001738); • a series of Faculty Research Enhancement Grants and a semester of developmental leave to Butler from Texas State University; • a grant from the Biogeography Specialty Group of the Association of American Geographers (AAG) to Cairns; • a doctoral student research grant to Resler from the Geomorphology Specialty Group of the AAG; • a National Park Service Geoscientists in the Park grant to Resler; • the Hodges-Padilla Dissertation Research Scholarship to Resler, and the Ray and Marian Butler Scholarship in Environmental Geography to Resler and Schmid, from Texas State University; • and Quick Response Grants to Butler from the Natural Hazards Research and Applications Information Center of the University of Colorado (1996 and 2004). The National Park Service in Glacier National Park arranged housing and field permits over several field seasons. Kruger Helicopter Service, West Glacier, provided superb flying and aerial views of treeline. Jim Kruger is a legend in Glacier Park for a reason. Dave Wopat, another legend in his own time, performed ‘‘Ring of Fire’’ and ‘‘El Paso’’ for us more times than anyone should have to. Thomas R. Allen, Kelley A. Crews-Meyer, Joseph P. Messina, Fritz Klasner, Dave Selkowitz, Ningchuan Xiao, and Yu Zeng worked with us on a variety of aspects of our treeline research. The data collection efforts of USGS colleagues Karen Holzer, Lisa McKeon, and Blase Reardon have been greatly appreciated. For assistance in the field, we thank Katherine J. (Louise) Alftine, Kathleen M. (Radar Woman) Bergen, Wendy (Wendle) Bigler, Mark A.(Woody) Bowersox, William D. (Farm Boy xvii
xviii
Acknowledgments
Fred) Butler, Dawna L. (Canuckarama) Cerney, Lisa M. (Sunshine) DeChano, Jill (Tokyo) Faine, Dianna Alsup (Mom) Gielstra, Chelsea S. (Spinny) Kelley, Adam (Mad Dog) Krutchinsky, Stephen J. (The Major) MacGregor, Joseph (Jumpin’) Malanson, Chris (LaTrec) Matus, Sean (Sean-O) McKnight, Kate (Thelma) Ramsden, Paul (Dad) Resler, Sean E. (Cato) Savage, Katherine A. (Shorts) Schipke, Dave (Eeyore) Selkowitz, Jackie (Smiley) Smith, Lynn (LYNN!) Smollin, Daniel J. (Cookie) Weiss, Teresa E. (Hat Girl) Welsh, and William F. (Mr. Bill) Welsh. Janet Butler, Ann Fagre, Mary Malanson, and Jeannie Walsh deserve special kudos for putting up with us and our many absences in the field, or in our minds, when we’ve been working on treeline. Their support has been essential and is more appreciated than we can adequately express. The chapters in this book were reviewed by external researchers as well as by a number of internal reviewers who contributed other chapters. Without the selfless contribution of such voluntary reviewers, works of this kind would rarely be possible and we thank the reviewers whole heartedly: Tom Allen, Matthew Bekker, Ryan Danby, John Dixon, Carol Harden, Ullrich Kamp, Lynn Resler, and Fausto Sarmiento.
LIST OF ACRONYMS ANOVA ATE BI CBE DBH DEM DIFN DLLAI DN ESD ETM+ FD GAM GCP GIS GI Sciences GLM GNP GPS LAI LFI LIA ML MSS NDVI NPP NPS OO PAR PCA PDI PDO PLAI RSP RMS SEL SPI TIN TLAI TM TRASP TRI TSI TWI USGS UV VI
Analysis of Variance Alpine Treeline Ecotone Beer’s Index Cumulative Boundary Elements Diameter at Breast Height Digital Elevation Model Diffuse Noninterceptance Drip Line Leaf Area Index Digital Number Effective Soil Depth Enhanced Thematic Mapper (Landsat) Foliage Density Generalized Additive Models Ground Control Point Geographic Information Systems Geographic Information Sciences Generalized Linear Models Glacier National Park Global Positioning System Leaf Area Index Landform Index Little Ice Age Maximum Likelihood Multispectral Scanner (Landsat) Normalized Difference Vegetation Index Net Primary Productivity National Park Service Object Oriented (Classification) Photosynthetically Active Radiation Principal Components Analysis Profile (Soils) Development Index Pacific Decadal Oscillation Projected Leaf Area Index Relative Slope Position Root Mean Square Standard Area of LAI (Leaf Area Index) Snow Potential Index Triangular Irregular Network Total Leaf Area Index Thematic Mapper (Landsat) Relative Radiation Index Topographic Ruggedness Index Terrain Shape Index Topographic Wetness Index United States Geological Survey Ultraviolet Vegetation Index
xix
LIST oF CONTRIBUTORS
Matthew F. Bekker Department of Geography, Brigham Young University, Provo, UT 84602, USA,
[email protected] Daniel G. Brown School of Natural Resources & Environment, University of Michigan, Ann Arbor, MI 48109-1041, USA David R. Butler Department of Geography, Texas State University-San Marcos, San Marcos, TX 78666-4616, USA.
[email protected] David M. Cairns Department of Geography, Texas A&M University, College Station, TX 77845, USA Daniel B. Fagre US Geological Survey, Northern Rocky Mountain Science Center, Glacier National Park, West Glacier, MT 59936-0128, USA,
[email protected] Mark A. Fonstad Texas State University-San Marcos, San Marcos, TX 78666-4616, USA,
[email protected] Christine A. Geddes School of Natural Resources & Environment, University of Michigan, Ann Arbor, MI 48109-1041, USA Evan S. Hammer Department of Geography, University of North Carolina-Chapel Hill, Chapel Hill, NC 27599-3220, USA George P. Malanson Department of Geography, University of Iowa, Iowa City, IA 52242, USA,
[email protected] Sean A. McKnight Department of Geography, University of North Carolina-Chapel Hill, Chapel Hill, NC 27599-3220, USA xxi
xxii
List of Contributors
Lynn M. Resler Department of Geography, Virginia Polytechnic Institute and State University, Blacksburg, VA 24060, USA,
[email protected] Carol F. Sawyer Department of Earth Sciences, University of South Alabama, Mobile, AL 36688-0002, USA,
[email protected] Ginger L. Schmid Department of Geography, Minnesota State University, Mankato, MN 56001, USA,
[email protected] Julie P. Tuttle Department of Geography, University of North Carolina – Chapel Hill, Chapel Hill, NC 27599-3220, USA Stephen J. Walsh Department of Geography, University of North Carolina – Chapel Hill, Chapel Hill, NC 27599-3220, USA,
[email protected] Daniel J. Weiss Department of Geography, University of North Carolina – Chapel Hill, Chapel Hill, NC 27599-3220, USA Forrest D. Wilkerson Department of Geography, Minnesota State University, Mankato, MN 56001, USA,
[email protected]
C H A P T E R
1
Introduction: Understanding the Importance of Alpine Treeline Ecotones in Mountain Ecosystems Daniel B. Fagre
Contents 1. Introduction 2. Importance of High-Elevation Mountain Research 3. Recent Changes to the Alpine Areas of Glacier National Park 4. Factors Affecting the Alpine Treeline Ecotone References
1 1 3 6 8
1. Introduction The alpine treeline ecotone (ATE) is an area of transition high on mountains where closed canopy forests from lower elevations give way to the open alpine tundra and rocky expanses above. The ATE in western US mountains is a highpriority research area because of its ecological importance, interest to managers, and possible role as a signal of climate change. The US Geological Survey’s Global Change Research Program funded a multidisciplinary team of investigators to examine the processes and rates of change in the ATE at Glacier National Park in 1998. This book reports on the outcome of efforts to detect, document, and understand the changes occurring at alpine treeline on Lee Ridge (Figure 1) and at other alpine sites in Glacier National Park. The impetus for this focused research activity was the concern that climate change will ultimately lead to upward movement of trees and the elimination of alpine tundra and associated biodiversity. The task was to determine the underlying mechanisms that are guiding the invasion of the alpine tundra by trees and specify how climate change is potentially driving these mechanisms. The focus was on different scales of pattern and process to understand the drivers of ATE change. This focus assumed that ATE responses to climatic change and variability are strongly influenced by geomorphology and biological feedbacks. More broadly, the ATE research results were intended to Developments in Earth Surface Processes, Volume 12 ISSN 0928-2025, DOI 10.1016/S0928-2025(08)00201-0
Ó 2009 Published by Elsevier B.V.
1
2
Daniel B. Fagre
Figure 1 Ayoung mountain goat stands on a rock outcrop above treeline, GNP. As trees invade open areas above current treeline, the available habitat for mountain goats is diminished.
inform us of the spatial extent and pace of high-elevation changes that are taking place elsewhere in Glacier National Park and the western mountains of the United States.
2. Importance of High-Elevation Mountain Research Mountains cover 20% of Earth’s terrestrial surface, are home to 10% of the global population, and provide 50% of the water used by humans (Messerli and Ives, 1997). Mountains also contain unique species and substantial biodiversity, underpin tourism and recreation in many parts of the world, and provide goods, such as timber, and ecosystem services such as clean air. Clearly, mountains play a large role in society worldwide. Mountains, particularly at the higher elevations, have warmed more rapidly in recent decades than other land areas (Diaz et al., 2003). This is readily apparent in the global retreat of mountain glaciers (Meier et al., 2007). The effects of warmer temperatures also are apparent on alpine vegetation (Pauli et al., 2001) and has led to more scrutiny of alpine areas. The American West is no exception, and since most of the mountains are on federal lands, resource management in mountains is a key concern across many federal agencies including the National Park Service. The processes and dynamics that shape mountain ecosystems start at the top. Here, in the land above the trees, what happens in the harsh environment of the alpine tundra reflects climatic conditions influencing the rest of the mountain
Alpine Treeline Ecotones in Mountain Ecosystems
3
ecosystems but also contributes to the response of entire mountain systems. The degree to which snow is stabilized by high-elevation vegetation, for instance, can influence the amount of water storage and release. This is important because recent changes clearly are evident throughout the high-elevation communities of subalpine meadows, ATEs, and the alpine tundra of North America (cf. Fagre et al., 2003; Inouye, 2008; Rochefort et al., 1994) and elsewhere (Lenoir et al., 2008). To better understand the future condition of our mountains, we need a more thorough understanding of the intricate interactions between climate, geomorphology, and alpine biology. As a potential bellwether for broad-scale changes in mountains, brought on by climate change or increases in global air pollution, we also need to better understand the relationship between the alpine and the mountains as a whole (Seastedt et al., 2004). For many people, there is no better place to start than the ATE, the area on the heights of mountains where trees struggle to live. This is where the limits of tree adaptation to abiotic constraints, such as the harsh climate and rocky substrate, interact with biological processes, such as competition and facilitation of seedling establishment, to produce a complex mosaic of surviving trees. This mosaic contains distinctly different spatial arrangements — patches, clumps, fingers, and ribbons — that hardly resemble the linear characteristics we associate with a ‘‘treeline.’’ This treeline ecotone also marks an area where ‘‘tree’’ is a relative term because of the amazing diversity of individual tree responses to surviving there. These tree-like plants are stunted, flagged, prostrated, layered, abraded, twisted, and tightly clumped behind a sheltering boulder. Clearly, if climatic conditions become less intense, it is here where we should see changes most distinctly and, perhaps, earlier than elsewhere in the mountains.
3. Recent Changes to the Alpine Areas of Glacier National Park The alpine areas of Glacier National Park are cherished and, in fact, define what this park means for Americans. This is evident in the myriad photographs and paintings of the park that focus on summits, glaciers, subalpine meadows, and alpine lakes. It is evident in the descriptions George Bird Grinnell used to convince Congress to create Glacier National Park in 1910. It is evident in the marketing slogans used by the Great Northern Railroad to entice early 20th century visitors to the park that described it as the ‘‘Alps of America.’’ It is the reason that the famed Going-to-the-Sun Road was built precariously along steep mountain slopes: to bring people to the alpine environment. The pilgrimage to Logan Pass, the highest point on the Going-to-the-Sun Road over the Continental Divide, is so important to park visitors that most will not visit the park until it is open. In the spring, before the snow is cleared from the road and people have access to Logan Pass, the local economy loses revenues because the park is the economic engine for the region (Swanson et al., 2003). Many park visitors do not consider the park to be ‘‘open’’ unless the Going-to-the-Sun Road over the mountains is open.
4
Daniel B. Fagre
Why is this access to Logan Pass so important? It is one of the few places where one can drive through spectacular alpine scenery and walk through lush meadows at treeline. Here, park visitors can experience the ‘‘Hanging Gardens,’’ picturesque snow-fed streams flowing through expanses of alpine flowers. Here, visitors see the mountain goat (Oreamnos americanus), which is the mascot and logo for the park as well as numerous local organizations and businesses for the past century (Figure 1). Here, they can experience the small groves of miniature subalpine fir (Abies lasiocarpa), including the krummholz (from ‘‘crooked wood’’ in German) forms that hug the ground in tight clumps. Here, as in other high-elevation treeless areas, is where people may see a grizzly bear (Ursus arctos horribilis). In short, the various alpine areas and the experiences park visitors have there are what people tend to strongly associate with Glacier Park. Given this association and the 1916 National Park Service mandate to conserve the scenery and the natural and historic objects and the wildlife by such means as will leave them unimpaired for the enjoyment of future generations, there is strong reason for management to be concerned about ongoing and potential changes in alpine areas of Glacier Park. At Logan Pass, for example, subalpine fir trees have become taller and have expanded outward from the scattered groves already there (Klasner and Fagre, 2002). As this process continues, the viewscape changes for park visitors and the area of flower-rich meadows diminishes. In turn, the charismatic animals that utilize this area and that visitors enjoy seeing, such as the White-tailed Ptarmigan (Lagopus leucurus), will be impacted. In addition to potentially diminished visitor experiences, there is the possible change in the public’s perceptions of the overall park value. Glacier Park and national parks in general are clearly part of our national heritage. In the view of some Americans, not taking care of our national parks tarnishes our global standing. Park management is sensitive to the public view on their performance and the alpine areas are specifically mentioned as a public focus in much of the general literature on Glacier Park (Rockwell, 2002). The area above treeline covers nearly one third of Glacier National Park (Lesica, 2002), and despite the sparse vegetation compared to lower plant communities, it contains ecologically important resources and important functions. It has significant biological diversity (27% of the park’s plant species are found in the alpine despite the low biomass; Lesica, 2002) and is critical habitat for bighorn sheep (Ovis canadensis), hoary marmots (Marmota caligata), and other species that contribute to the biological richness of Glacier Park. The alpine areas of Glacier National Park potentially affect key resources and ecosystem services. Much of the winter snowpack accumulation and the retention of snow into summer occur above treeline. The retention of snow is critical because it provides a reservoir of slowly released water for the mountain ecosystem during the drier summer months. Changes to the ATE, such as expanding subalpine forests, could alter snowpack accumulation in numerous ways and directly impact lower elevation biota and change mountain hydrology (Figure 2). Because 85% of the water used by people in the western US comes directly from mountains (USGCRP, 2000), these changes to the ATE could have far-reaching effects. The starting zone for snow avalanches is often above treeline (Figure 3). In Glacier National Park, high alpine basins show evidence of frequent avalanche
Alpine Treeline Ecotones in Mountain Ecosystems
5
Figure 2 Snow-covered trees near treeline in GNP. The timing and amount of snow accumulation plays a major role in treeline dynamics.
activity (Butler, 1979). Snow avalanches have geomorphological significance and also are important ecologically by transferring soil, nutrients, and vegetation downslope (Butler et al., 1992). Snow avalanche paths act as firebreaks during forest fires (Malanson and Butler, 1984) and stimulate growth of key vegetation for numerous
Figure 3 The topography of GNP clearly reflects past glaciation and is a major influence on alpine treeline ecotones today. Note the multiple ‘‘treelines’’ at various elevations throughout this scene.
6
Daniel B. Fagre
wildlife species. Clearly, if the starting zones for avalanches are invaded by trees, which often act as anchors to reduce snow instability, there are many cascading changes that could have both direct and subtle effects on mountain resources. Finally, alpine areas, and specifically the ATE, may be potential indicators of long-term climatic shifts. Climatic change logically should be most easily observed in the ATE where tree seedlings are established upslope under more favorable conditions, potentially moving the ATE upslope, or causing a recession downslope when die-backs occur among the older trees under less favorable conditions. The reference to a ‘‘treeline’’ is a misnomer because the ATE is considered to be the upper limit of tree growth that rarely looks linear because of the variation in the mountain environment. ATEs likely were depressed (moved downslope) by temperature in Glacier National Park during the colder Little Ice Age. This is shown by the distribution of very old, dead trees that predated the Little Ice Age and were established on slopes above the current ATE (Carrara, 1989). Recent ATE advances in Glacier Park have been locally documented (Butler and DeChano, 2001; Butler et al., 1994; Cairns, 2001; Cairns and Malanson, 1997, 1998; Klasner and Fagre, 2002) and may be related to warming temperatures.
4. Factors Affecting the Alpine Treeline Ecotone It is important to distinguish the relative roles of climate and geomorphological disturbances and the interactions between them in creating ATEs. Landslides, rockfall, debris flows, and other geomorphological processes can create ‘‘treelines’’ that are not determined by climatic limitations. It also is important to separate possible park management changes from those driven by climatic shifts. For instance, Butler and DeChano (2001) documented some areas in Glacier Park that have had upward forest migrations of 100–250 m since 1935, but probably as a result of both climatic shifts and changes in fire suppression policy. Given that most ATEs are a visible indicator of relatively abrupt ecological transition, changes in ATEs, therefore, should reflect broad-scale climate change in mountains. However, the complex mechanisms underlying treeline changes, and how these interact with climatic drivers and the changing geomorphology of highelevation areas, have led some mountain scientists (Malanson, 2001) to question the utility of the ATE as a climate change indicator. Clearly, resolving why and how ATEs are changing is critical to understanding the future of Glacier Park and to interpreting treeline change as an ecosystem response to global warming. Topographical features and geomorphological processes play an important role in all high-elevation mountain environments (Butler et al., 2007). At Glacier Park, however, these may be more dominant than in other mountains of the western United States. For instance, generally steep topography and the effects of past glaciations have created a landscape where topographic discontinuities (e.g., cirque headwalls and glacier horns) predominate (Figure 3). This results in a treeline that is much more spatially variable than that occurring in the southern Rocky Mountains (Cairns and Malanson, 1997). Becwar and Burke (1982) used topographical maps to
Alpine Treeline Ecotones in Mountain Ecosystems
7
estimate that 80% of the ATE spans an elevation range of 550 m in Glacier Park, compared to 150 m in the southern Rocky Mountains. A more extreme mountain landscape not only limits areas where trees can establish, it also negatively impacts trees that have established where climatic conditions are favorable by subjecting them to more rockfall and other geomorphic disturbance. Geomorphology is also important at smaller scales, especially for soil development in the ecotone (Butler et al., 2007; Malanson et al., 2007). Geomorphology and soils combine to create the landscape on which plants establish and survive in the harsh windy environment at and above treeline (Figure 4). In some cases, geomorphic processes and soils development assist in creating fine-scale environments amenable to tree seedling establishment and survival, whereas in other they cases act to completely preclude establishment success. This influence of geomorphology at treeline, as evidenced by the effects of solifluction treads and risers on seedling establishment, has in concert with surface lithology and stratigraphy major impacts on the ability of trees to pioneer and invade the alpine tundra. Aerial photograph analysis in Glacier National Park has revealed numerous locations where tree establishment is clearly inhibited by geomorphic, lithologic, or stratigraphic and structural control (Butler et al., 2003). On Flattop Mountain, in the center of Glacier Park, this phenomenon is clearly evident where bedrock ridges serve as invasion points for trees, whereas adjacent topographically lower meadows remain snow-covered well into the summer season. The net result illustrates the influence that topography, stratigraphy, and structure have on tree establishment there. Clearly, the role of geomorphology in facilitating or inhibiting tree establishment will vary across the landscape, but it absolutely must be quantified and understood before climatic and biotic causes for ATE change are invoked. To better understand the complex interactions that shape
Figure 4 A climate monitoring station at Lee Ridge, GNP. Winds over 260 kph have been recorded on ridges such as this one during the winter.Wind interacts with geomorphology to determine where trees are able to establish and survive.
8
Daniel B. Fagre
ATEs, a long-term project was initially focused on Lee Ridge in Glacier National Park and then expanded to broader landscapes, including other alpine areas of the western U.S. This book describes an array of studies that were coordinated and intended to provide a holistic understanding of ATEs in a changing world.
REFERENCES Becwar, M.R., Burke, M.J., 1982. Winter hardiness limitations and physiography of woody timberline flora. In: Li, P.H., Sakai, A. (Eds.), Plant Cold Hardiness and Freezing Stress: Mechanisms and Crop Implications, Vol. 2. Academic Press, New York, pp. 307–323. Butler, D.R., 1979. Snow avalanche path terrain and vegetation, Glacier National Park, Montana. Arctic and Alpine Research 11, 17–32. Butler, D.R., DeChano, L.M., 2001. Environmental change in Glacier National Park, Montana: An assessment through repeat photography from fire lookouts. Physical Geography 22, 291–304. Butler, D.R., Malanson, G.P., Bekker, M.F., Resler, L.M., 2003. Lithologic, structural, and geomorphic controls on ribbon forest patterns in a glaciated mountain environment. Geomorphology 55, 203–217. Butler, D.R., Malanson, G.P., Cairns, D.M., 1994. Stability of alpinetreeline in Glacier National Park, Montana, USA. Phytocoenologia 22, 485–500. Butler, D.R., Malanson, G.P., Walsh, S.J., 1992. Snow-avalanche paths: Conduits from the periglacial-alpine to the subalpine depositional zone. In: Dixon, J.C., Abrahams, A.D. (Eds.), Periglacial Geomorphology. John Wiley, London, pp. 185–202. Butler, D.R., Malanson, G.P., Walsh, S.J., Fagre, D.B., 2007. Influences of geomorphology and geology on alpine treeline in the American West – more important than climatic influences? Physical Geography 28, 434–450. Cairns, D.M., 2001. Patterns of winter desiccation in krummholz forms of Abies lasiocarpa at treeline sites in Glacier National Park, Montana, U.S.A. Geografiska Annaler 83A, 157–168. Cairns, D.M., Malanson, G.P., 1997. Examination of the carbon balance hypothesis of alpine treeline location in Glacier National Park, Montana. Physical Geography 18, 125–145. Cairns, D.M., Malanson, G.P., 1998. Environmental variables influencing the carbon balance at the alpine treeline: A modeling approach. Journal of Vegetation Science 9, 679–692. Carrara, P.E., 1989. Late Quaternary Glacial and Vegetative History of the Glacier National Park Region, Montana, USA. Geological Survey Bulletin 1902, Denver, CO. Diaz, H.F., Eischeid, J.K., Duncan, C., Bradley, R.S., 2003. Variability of freezing levels, melting season indicators, and snow cover for high-elevation and continental regions in the last 50 years. Climatic Change 59, 33–52. Fagre, D.B., Peterson, D.L., Hessl, A.E., 2003. Taking the pulse of mountains: Ecosystem responses to climatic variability. Climatic Change 59, 263–282. Inouye, D.W., 2008. Effects of climate change on phenology, frost damage, and floral abundance of montane wildflowers. Ecology 89, 353–362. Klasner, F.L., Fagre, D.B., 2002. A half century of change in alpine treeline patterns at Glacier National Park, Montana, USA. Arctic, Antarctic, and Alpine Research 34, 49–56. Lenoir, J., Gegout, J.C., Marquet, P.A., de Ruffray, P., Brisse, H., 2008. A significant upward shift in plant species optimum elevation during the 20th century. Science 320, 1768–1771. Lesica, P., 2002. Flora of Glacier National Park. Oregon State Univ. Press, Corvallis. Malanson, G.P., 2001. Complex responses to global change at alpine treeline. Physical Geography 22, 333–342. Malanson, G.P., Butler, D.R., 1984. Avalanche paths as fuel breaks: Implications for fire management. Journal of Environmental Management 19, 229–238. Malanson, G.P., Butler, D.R., Fagre, D.B., Walsh, S.J., Tomback, D.F., Daniels, L.D., et al., 2007. Alpine treeline of western North America: Linking organism-to-landscape dynamics. Physical Geography 28, 378–396.
Alpine Treeline Ecotones in Mountain Ecosystems
9
Meier, M.F., Dyurgerov, M.B., Rick, U.K., O’Neel, S., Pfeffer, W.T., Anderson, R.S., et al., 2007. Glaciers dominate eustatic sea-level rise in the 21st century. Science 317, 1064–1067. Messerli, B., Ives, J.D. (Eds.), 1997. Mountains of the World: A Global Priority. Parthenon Publishing, London. Pauli, H., Gottfried, M., Reiter, K., Grabherr, G., 2001. High mountain summits as sensitive indicators of climate change effects on vegetation patterns: The ‘‘Multi Summit-Approach’’ of GLORIA (Global Observation Research Initiative in Alpine Environments). In: Visconti, G.Beniston, M.Iannorelli, E.D., Barba, D. (Eds.), Global Change and Protected Areas. Kluwer, Dordrecht, pp. 45–51. Rochefort, R.A., Little, R.L., Woodward, A., Peterson, D.L., 1994. Changes in subalpine tree distribution in western North America: A review of climate and other factors. The Holocene 4, 89–100. Rockwell, D., 2002. Exploring Glacier National Park. Globe Pequot Press, Guilford, CT. Seastedt, T.R., Bowman, W.D., Caine, T.N., McKnight, D., Townsend, A., Williams, M.W., 2004. The landscape continuum: A model for high-elevation ecosystems. BioScience 54, 111–121. Swanson, L.D., Nickerson, N., Lathrop, J., Archie, M.L., Terry, H.D., 2003. Gateway to Glacier: The Emerging Economy of Flathead County. National Parks and Conservation Association, Washington, DC. USGCRP, 2000. Climate Change and America: Overview Document. A Report of the National Assessment Synthesis Team. U.S. Global Change Research Program, Washington, DC.
C H A P T E R
2
Pattern—Process Relations in the Alpine and Subalpine Environments: A Remote Sensing and GIScience Perspective Stephen J. Walsh, Daniel G. Brown, Christine A. Geddes, Daniel J. Weiss, Sean McKnight, Evan S. Hammer, and Julie P. Tuttle
Contents 1. Introduction 2. Study Area 2.1. General setting 2.2. Geographic areas of research emphasis 3. Background and Context 3.1. Remote sensing approaches 3.2. Field approaches 3.3. GIS representation 4. Data and Methods 4.1. Medium-grained remote sensing 4.2. High-resolution DEM creation 5. Selected Applications 5.1. Pixel versus object classification of vegetation 5.2. Linear mixture modeling 5.3. Pattern metrics 6. Conclusions References
11 12 12 14 16 17 18 19 20 20 21 24 24 28 29 31 32
1. Introduction Understanding the patterns and dynamics of spatially extended resources, such as subalpine forest, tundra, and snow cover, requires a range of spatially explicit observational and analytical approaches. Additionally, in specific, observations were required in the course of our collective research enterprise to understand alpine and subalpine landscape systems, such that we needed to pursue innovative approaches to surface observation, characterization, and analysis. In this chapter, we describe a Developments in Earth Surface Processes, Volume 12 ISSN 0928-2025, DOI 10.1016/S0928-2025(08)00202-2
Ó 2009 Elsevier B.V. All rights reserved.
11
12
Stephen J. Walsh et al.
set of activities that draws on tools and techniques in remote sensing, geographic information science, and environmental modeling to support specific analytical objectives in the overall project. This chapter summarizes work that has been ongoing for a period of more than 15 years. Although this means that a number of efforts were aimed at specific objectives, we seek to synthesize these efforts and indicate both the innovative methodologies we have developed and/or applied within the context of Glacier National Park (GNP), Montana, and the scientific questions that these efforts have supported. Central to our approach to observing and understanding the alpine and subalpine environments is an explicit interest in how patterns and processes interact with one another. We are interested in characterizing spatial patterns in ways that facilitate connections and interactions with characterizations of spatial and ecological processes, through models that are described in other chapters in this volume. We take a particular interest in how the scales of observation and characterization affect our understanding of pattern–process relationships, as spatial patterns are scale-dependent, as are the ways in which those spatial patterns are affected by and feed back to affect processes. This chapter is organized into the following sections after this brief introduction. First, we provide a brief description of the study areas within GNP where the work we are describing has been conducted. Second, we provide some background on the kinds of tools and technologies we are working with and summarize literature relevant to the work we are doing. Third, we describe the various remotely sensed data sets we have acquired for this work and how we have processed these data to make them useful for our biophysical investigations. Fourth, we describe several applications and analyses conducted using these data and some of the methodological innovations involved in these studies. Finally, we close with a discussion of how this works fits within the broader research effort of the Mountain GeoDynamics Research Group and a brief set of conclusions.
2. Study Area 2.1. General setting GNP is located in the northwestern corner of Montana, USA (Figure 1). The Park extends south from the border between the United States and Canada approximately 80 km and is bounded by the Great Plains to the east and the Flathead River Valley to the west. The dominant physical features of the Park are the Livingston and Lewis mountain ranges, which stretch from the northwest to southeast through the Park’s center. As part of the Northern Rocky Mountains of the United States, GNP is characterized by a rugged landscape. Although the Park was named for the remnant Little Ice Age glaciers present when it was established in 1910, glaciation during the Pleistocene was responsible for sculpting the numerous cirques, U-shaped valleys, and other classic alpine glacial features found in the Park. The dramatic effects of glaciation were enhanced by the relatively soft limestone and argillite that comprise the Park’s mountains (Carrara, 1990).
Pattern–Process Relations in the Alpine and Subalpine Environments
Figure 1
13
Location map of Glacier National Park, Montana.
The vegetation of the Park is a mixture of species common to the Rocky Mountains and the Pacific Northwest. Although species from these ecosystems are found together, they tend to be spatially segregated as a result of orographic precipitation, with species from the Pacific Northwest more commonly found on moist slopes west of the Continental Divide and species from the Rocky Mountains occurring more commonly on the drier eastern slopes. In addition to east–west differences, vegetation in the Park responds to varying climatic conditions along an elevation gradient. The closed forests found at lower elevations gradually give way to a subalpine zone containing the alpine treeline ecotone (ATE): a mosaic of stunted upright trees, subalpine meadows, krummholz tree islands, and alpine tundra. Above the subalpine is the alpine zone, which is characterized by tundra, bare and lichen-covered rock, snow, and ice. Like all environments, GNP is subject to periodic disturbances that vary both in size and frequency. The most significant disturbance agent is forest fire at lower elevations, which periodically burns thousands of hectares of timber. Snow avalanches are another important disturbance agent, and while not large in size individually, collectively they occupy a significant portion of the landscape (Butler and Walsh, 1990). Unlike forest fires, snow avalanches tend to reoccur in the same paths, which are often coincident with stream channels, topographic depressions, and faults and fractures (Butler and Walsh, 1990). The physical beauty of the glacially sculpted mountains, the remaining glaciers themselves, the mosaic of vegetation, lakes both large and small, and abundant wildlife attract over one million visitors to the Park each year. Despite the large number of visitors, the Park’s status as a national park has greatly reduced anthropogenic impacts on the landscape, leaving a relatively pristine environment ideal for scientific study.
14
Stephen J. Walsh et al.
2.2. Geographic areas of research emphasis Research utilizing remote sensing techniques has been conducted at a number of spatial scales in our various studies in GNP. For instance, Allen (1998) looked at the entire Park, while others (Allen, 1995; Allen and Walsh, 1996; Brown, 1994; Butler and Walsh, 1990) have focused on the east-central portion of Park. More recently, as the spatial resolution of remotely sensed imagery has improved, our research has expanded to include detailed pattern and process relations (Walsh et al., 2003a) at the hillslope scale and finer. For most projects that incorporate remotely sensed imagery, measurements taken in the field are required to associate site characteristics with data collected remotely. Field data acquired for this purpose in GNP have been collected from a number of sites distributed throughout the eastern one-half of the Park. Several of the sites have been used repeatedly for various research endeavors, thereby allowing both site expertise and a deep data archive to be developed. Primary field sites in GNP include Baring Basin, Cataract Creek, Divide Mountain, East Flattop Mountain, Lee Ridge, Piegan Pass, Preston Park, Scenic Point, and Siyeh Pass (Figure 2). In addition to their accessibility by trails, these field sites were selected because they contain components of the ATE and interesting geomorphic features, with varying topographic settings. The Baring Basin field site is located in the valley between Going-to-the-Sun Mountain and Goat Mountain. Sexton Glacier is perched above the valley, and meltwater from it flows down Baring Creek from the northwest to the southeast. The lower portion of Baring Basin contains closed canopy forest, which becomes a mosaic of krummholz patches and meadows with increasing elevation. The Cataract Creek field site is in the upper portion of the U-shaped valley just below Piegan Pass and Cataract Mountain. This site has a relatively low slope angle and a north to northwest aspect. The vegetation is composed primarily of large krummholz tree islands surrounded by alpine tundra. The Divide Mountain field site is located on the east slope of the mountain with the same name. With east to southeast-facing slopes, the site overlooks the Great Plains of western Montana. The vegetation at the Divide Mountain field site is characterized by krummholz trees and tree islands, often located behind microtopographic features including boulders and solifluction terraces. East Flattop Mountain is a north-south trending mountain located on the eastern edge of the Park. This field site offers one of the largest expanses of alpine tundra and ATE in GNP. The great size of this site contributes to a high diversity of ATE types, ranging from wide ecotones with abundant krummholz islands to abrupt transitions from upright trees to alpine tundra. Lee Ridge slopes from the flank of Gable Mountain toward the Canadian border to the north. The ridge is a relatively narrow strip of ATE surrounded by closed canopy forests to the east, north, and west. The vegetation on Lee Ridge consists of alpine tundra with krummholz tree islands and linear, finger-like sections of trees extending from the forests below, parallel to the dominant wind direction. Like Divide Mountain, Lee Ridge has prominent periglacial patterned ground features that influence the establishment and growth of trees within and above the ATE. The Piegan Pass site is located between Cataract Mountain and Pollock Mountain, and above the headwall at the south end of Cataract Creek valley. The field site is situated within a small, sheltered cirque. The vegetation at this field site
Pattern–Process Relations in the Alpine and Subalpine Environments
Figure 2
15
Study sites in GNP displayed on an air photo mosaic.
is limited to alpine tundra, and prominent periglacial features are found here. Preston Park is located just below Siyeh Pass, in a valley that slopes gently from east to west. A cirque forms the head of the valley, which is currently occupied by a glacial tarn. Lower portions of this study site are covered by meadows with some patches of stunted trees. Higher in the study site, the trees take on more of a krummholz growth form and are surrounded by patches of tundra and bare rock. The Scenic Point site is found in the gently sloped and generally east-facing basin between Scenic Point and Bison
16
Stephen J. Walsh et al.
Mountain. Scattered krummholz islands can be found within the alpine tundra vegetation that dominates this field site, and several forms of periglacial patterned ground can be found here as well. Siyeh Pass is located just above the headwall at the upper end of Baring Basin. Siyeh Pass lies between Matahpi Peak and Mount Siyeh and separates Baring Basin and Preston Park. This site has no alpine treeline, but abundant periglacial patterned ground features are present in the alpine tundra.
3. Background and Context Remote sensing and GIScience research conducted in the Park by the Mountain GeoDynamics Research Group has tended to focus on (a) resource assessments of snow avalanche paths and debris flows (Butler and Walsh, 1990; Walsh and Butler, 1997; Walsh et al., 1994a), assessment of lake turbidity and basin morphometry (Brown and Walsh, 1991, 1992), identification of deltaic wetlands (Butler et al., 1991); (b) examination of scale dependence of plant biomass and topography (Bian and Walsh, 1993; Walsh et al., 1997), characterization of the ATE (Allen and Walsh, 1993, 1996; McGregor, 1998), and the development of cartographic, empirical, and mechanistic models that assess the variability of the ATE over time and space scales (Brown et al., 1994; Walsh and Kelly, 1990; Walsh et al., 1998). In addition, spatially explicit digital data are being used to examine the effects of geomorphic and biogeographic processes and patterns influencing the spatial organization and variability of alpine treeline patterns and typology (Allen, 1995; Allen and Walsh, 1996; Brown, 1994), vegetation characteristics and spectral responses associated with the Leaf Area Index (LAI) (Brown, 2000; Hammer, 2004), and modeling of treeline position (Brown, 1992; Walsh et al., 1994b). The central issues of the research involve the sensitivity of tundra to invasion of trees, that is, the ecotone as an indicator of climate change (Malanson, 2001). Remote sensing of snow avalanche paths is another theme of considerable interest (Butler and Walsh, 1990; Walsh et al., 1990, 1994a, 2004), as is remote sensing of geomorphic features including periglacial patterned ground (Walsh et al., 2003a, 2003b). Finally, analysis of glaciers and perennial snow fields using remotely sensed data has been examined by Allen and Walsh (1993) and Allen (1998). Inaccessibility and remoteness make remote sensing an attractive method for conducting research in GNP, but these same factors create a set of challenges that must be addressed for a successful measurement and evaluation campaign. From a fieldwork perspective, working in the Park is difficult, because (a) the relatively short alpine summers leave few snow-free days when field sites are accessible; (b) long, steep, and time-consuming hikes, which are typically required to reach the field sites, limit the amount of work that can be done in any single day and the amount and type of equipment that can be carried by backpack; and (c) trails in the Park are periodically closed and backcountry camping for research is not allowed, due to safety issues related to grizzly bears. The rugged landscape of the Park also presents a set of challenges for processing remotely sensed data. One challenge is the accurate georectification of imagery for an area with relatively few stable ground control points (GCPs) and few obvious features identifiable on air photos that can
Pattern–Process Relations in the Alpine and Subalpine Environments
17
be used instead. Another image processing challenge is the effect of topography on spectral responses as a consequence of illumination differences caused by terrain settings and daily and seasonal changes in solar geometry.
3.1. Remote sensing approaches Remote sensing research in the park is constrained by a set of challenges common to the research community, namely balancing the issues of imagery cost, availability, and resolution. Landsat imagery has been used with great success in many research applications (Allen and Walsh, 1996; Brown, 1994) due to its medium-grained (30-m pixel size) spatial resolution, a reasonable spectral resolution (six bands at 30-m spatial resolution), and a coverage area large enough to capture all of GNP in a single scene. Unfortunately, many of the features present in the ATE are smaller than 30 m, a fact that has limited the scale at which detailed pattern–process analyses could be done with Landsat imagery. In recent years, higher spatial-resolution multispectral data, such as ADAR-5500 digital aircraft data (Figure 3) and IKONOS satellite imagery (Figure 4), have been added to our data archive. These data have been applied to studies of snow avalanche paths (Walsh et al., 2004) and periglacial patterned ground (Walsh et al., 2003b). A number of image processing and analytical techniques have been applied to remotely sensed data to address our research questions. Common image processing methods include supervised and unsupervised classification (Allen, 1995), linear mixture modeling (Brown, 2000), principal components analysis (PCA) (Walsh et al., 2003a), and the application of tasseled-cap transformations (Walsh et al., 2004). Analytical methods have included pattern metrics and clustering techniques
Figure 3
ADAR-5500 digital aircraft data of Lee Ridge draped over a 10 -m USGS DEM.
18
Stephen J. Walsh et al.
Figure 4 Ikonos satellite image of scenic point and the surrounding two medicine area draped over a 10 -m USGS DEM.
(Allen, 1995), and a host of statistical analyses ranging from descriptive statistics, linear regression analysis, and analysis of variance (ANOVA) (McKnight, 2004) to generalized linear models (GLM) and generalized additive models (GAM) (Brown, 1994).
3.2. Field approaches An important step in most remote sensing applications is the association of information extracted from remotely sensed data with data gathered in the field. These associations are made possible by collecting GPS locations for all field data samples that are differentially corrected to increase the accuracy of the located points to generally +5 m. Many types of field data can be gathered depending on the research objectives of the project. The simplest case involves gathering ground control data for accuracy assessments of image classifications. More involved methods include collecting detailed information on vegetation characteristics at the scale of the plant (e.g., size, age, species, growth form, and diameter at breast height), the patch (e.g., size, shape, and height), or the stand (e.g., composition and successional stage). Other than locations, the field data most often collected for our studies that use remotely sensed data are measurements of LAI and spectral reflectance curves for reference land cover types. LAI measurements have been collected using Li-Cor LAI-2000 Plant Canopy Analyzers (Brown, 2001; Hammer, 2004). These devices estimate LAI using the canopy gap method, in which under-canopy and overcanopy measurements are taken with a hemispherical optical sensor, and LAI values are calculated as a function of the amount of light intercepted by the plants. Spectral response measurements are taken using a Li-Cor 1800 spectroradiometer, which measures radiation at wavelengths between 300 and 1,100 nm, divided into user-defined band widths. Spectral reflectance curves (Figure 5) are then generated
19
Pattern–Process Relations in the Alpine and Subalpine Environments
5.00E – 01 4.50E – 01 4.00E – 01
% Reflectance
3.50E – 01
Bare Dead Snow Tree Tundra Wet tundra
3.00E – 01 2.50E – 01 2.00E – 01 1.50E – 01 1.00E – 01 5.00E – 02 0.00E + 00 330
430
530
630
730
830
930
1030
Wavelength (nm)
Figure 5 Spectral reflectance curves generated for common land cover types in GNP using data acquired in situ with a Li-Cor 1800 spectroradiometer.
by plotting the ratio of the amount of light reflected from the surface of an object to the amount of incoming light at each wavelength.
3.3. GIS representation A great benefit of working with remotely sensed data is that, unlike discrete data points, the image data are typically continuous across the coverage area, thereby allowing a large number of samples to be taken for the variables of interest. To make the most of this characteristic, it is useful to have other spatial data sets, typically in the form of GIS coverages, which can act as additional variables in statistical analyses. Research in the Park has often been linked to the landscape through GIS data layers. For example, digital elevation models (DEMs) and measures derived from them, such as slope angle, slope aspect, soil moisture potential, curvature, surface roughness, incident solar radiation, and snow potential index, have been used as independent variables and as descriptors of study sites as part of statistical analyses (Allen, 1995; Brown, 1994; Hammer, 2004) and scientific data visualizations (Walsh et al., 2003a). For these variables, US Geological Survey DEM (30-m
20
Stephen J. Walsh et al.
resolution) has served as the primary data source, and the derivative surfaces have been generated using GIS software packages. More recently, the 30-m DEM of the Park has been augmented to derive a 10-m DEM and a 5-m DEM constructed using stereo photogrammetry. Other data sets that have been useful for research include surficial geology (Carrara, 1990) and bedrock geology (Whipple, 1992), and soils data (Dutton and Marrett, 1997; Land and Water Consulting Incorporated, 1999).
4. Data and Methods 4.1. Medium-grained remote sensing Of the medium-resolution data sets we have been working with, those on the Landsat satellite platforms have seen the most application. We have used the Multispectral Scanner (MSS) data, with four spectral bands and a spatial resolution of 79 m, to take advantage of a longer time series and for analysis of broad-scale changes over time. More of our work has taken advantage of the enhanced spectral and spatial resolutions of the Thematic Mapper (TM) and Enhanced Thematic Mapper (ETMþ) sensors, with a 30-m spatial resolution and seven and nine spectral channels, respectively. The spatial resolution of the TM sensor has provided a nearly ideal balance between spatial detail and our ability to address questions over broad landscapes. The spectral characteristics include measuring reflected light in the visible (0.4–0.7 mm), near-infrared (0.7–1.4 mm), and middle infrared (1.5–2.5 mm) wavelengths and emitted radiation in the thermal infrared wavelengths (10–12.5 mm). Most of our applications have involved converting raw digital numbers (DNs) in the TM image to radiance by (a) applying gain and offset values for sensor calibration provided in the image header file and (b) adjusting the visible channel values to correct for path radiance by subtracting the DN values obtained in clear lakes. Next, radiance values were converted to exoatmospheric reflectance using the method of Leprieur et al. (1988). Because of the rugged nature of the terrain, this conversion has necessarily included a correction for the effects of terrain on solar incidence and illumination using the DEMs (Brown, 1994). Solar incidence angle was calculated using available 30-m resolution DEMs and the solar angles published in the image header files according to the equation cos I = cos Z cos S þ sin Z sin S cosðAS AZ Þ
ð1Þ
where S = slope of ground surface, As = aspect of ground surface slope, Az = solar azimuth from north, and Z = solar zenith angle. We made an initial attempt to use RADAR data to map vegetation structural characteristics at the alpine treeline, by acquiring a single image from the RADARSAT sensor. We placed corner reflectors at three locations within the Lee Ridge study site to help identify GCPs for georeferencing. Unfortunately, the rugged nature of the terrain, together with the highly reflective barren and tundra surfaces,
Pattern–Process Relations in the Alpine and Subalpine Environments
21
made identification of the corner reflectors impossible. In addition, discrimination of land covers was not possible in this context. Further analysis of the RADARSAT image was therefore not pursued.
4.2. High-resolution DEM creation We used soft copy photogrammetry to extract a high-resolution DEM from a pair of stereo images, that is, images with overlapping coverage and a series of GCPs (Ackermann, 1996). A 5-m-resolution DEM was created for a portion of Lee Ridge to support our analysis of the environmental determinants of treelinevegetation patterns. The product was evaluated with respect to field data and the corresponding US Geological Survey 7.5-min DEM, a 30-m-resolution DEM created by manual profiling of a stereo model along parallel transects. A pair of high-resolution 1:15,000 scale aerial photographs was acquired by the Department of the Interior in August of 1999 and scanned at a resolution of 3 m. GCPs were acquired with Trimble GeoExplorer GPS receivers in the summers of 1999 and 2001 (nominal accuracy of 1–5 m in the horizontal dimension and 1–25 m in the vertical dimension). The 37 GCPs were manually located on the images and their coordinates entered. To improve the spatial distribution of the GCPs and to include points in the northern portion of the study area, an additional 40 GCPs were established by identifying common points in the DEM and the scanned images and extracting the DEM values at those locations. To establish the relationship between the pair of photographs, 39 tie points, that is, points that could be located in each image of the pair, were entered in the same ‘‘point and click’’ fashion. The automatic generation of a DEM from a stereo model involves finding conjugate points on both images, calculating the displacement between matching pixels and interpolating and densifying the image, that is, surface fitting (Ackermann, 1996). A DEM with a 5-m resolution was extracted and then filtered with a 3 3 neighborhood averaging filter, a common step used to remove high-frequency spatial artifacts of the DEM extraction process. To facilitate comparison for accuracy assessment, the USGS DEM was resampled to a 5-m resolution with a nearest neighbor algorithm. The database of GPS elevations was used as a source of elevation data against which the photogrammetric and USGS DEMs could be compared. The intent of this study was to interpret the quality of a photogrammetrically derived DEM with respect to a 7.5-min USGS DEM. Accordingly, in the following discussion, ‘‘error’’ refers to the disagreement between points of known elevation, that is, GCPs measured in the field and corresponding values in each of the DEMs. Twenty-six of the 37 GCPs acquired with the GPS receivers fell within the bounds of the study area and were used in error analysis. Relationships between absolute error and the elevation and slope of pixels were examined. Difference maps were evaluated to assess the difference in geomorphic detail of the DEMs. Two difference maps were calculated and analyzed as the signed and unsigned value of the difference between the photogrammetric DEM and the USGS DEM. A small portion of the ridge was more closely examined using a set of control points collected in situ with a Topcon GTS-226 total station theodolite (accuracy
22
Stephen J. Walsh et al.
of 15 cm) in the summer of 2002. A lattice of 278 GCPs was acquired traversing the top of the ridge in the northernmost portion of the study area. The 278 GCPs from the total station served as an objective control set to intensively evaluate terrain form. Total station elevations were georeferenced, as closely as possible, to the photogrammetric and USGS DEMs, allowing for independent examination of landscape form over the area sampled. Both the USGS and photogrammetric DEMs captured the major terrain features, with the photogrammetric DEM exhibiting considerably more detail in terms of surface morphology (Figure 6). The channels to the east and west of the ridge were much better defined in the photogrammetric DEM, as was microtopography over the entire study area. The distributions of elevations were statistically similar. The mean absolute error was 5 m in the USGS DEM and 6 m in the photogrammetric DEM. Errors in the photogrammetric DEM ranged from –26 to þ9 m, with nearly two-thirds of pixels sampled in the photogrammetric DEM falling within 7 m of the GCPs. The USGS DEM showed a narrower range of error (–14 to þ4 m), but a lower proportion of pixels was included within 7 m of the GCP elevations (approximately 63%). Differences between GCP and DEM elevations were highest in the valleys to either side of the ridge in both DEMs. Possible reasons for the spatial pattern of error relate to the geomorphic character of the landscape, scant ground control in the valleys, and canopy bias. Areas of steep slope, in particular, are more susceptible to error, because elevation changes rapidly as a function of distance (Bolstad and Stowe, 1994). The highest errors were noted in areas of steep slopes ( 20°) in both DEMs. The region of control in the photogrammetric DEM was constrained by the limitations of data collection, as well as the ability to identify common points in the stereo images and USGS DEM. GCPs were much more concentrated on the top of the ridge, leaving the areas beyond the region of control to rely on the estimation of elevation modeled from the distant GCPs. More extensive collection (a)
(b)
Figure 6 Five-meter DEMs for Lee Ridge created from (a) stereo photogrammetry and (b) resampling 30 -m USGS DEM.
Pattern–Process Relations in the Alpine and Subalpine Environments
23
of GCPs in the valleys could potentially improve valley elevation predictions in the photogrammetric DEM. Difference maps elucidated the spatial pattern of elevation disagreement between photogrammetric and USGS DEMs. The patchwork pattern of differences results from the incongruence of resolution in the DEMs. The photogrammetric DEM reports 36 elevation values for each individual value in the USGS DEM, allowing for the depiction of more variability. Differences were relatively small over most of the study area. Identical values in both DEMs were reported for approximately 8% of the pixels. Nearly three-quarters of all pixel values in the photogrammetric DEM were within 7 m of the USGS DEM and 93% were within 14 m. However, there were areas of concentrated differences in elevation that gave rise to the majority of the disagreement. Values in a small saddle in the ridge toward its northern end were generally predicted to be several meters higher in the photogrammetric DEM. Conversely, elevations along the length of the valley to the west of the ridge were consistently predicted to be several meters lower in the photogrammetric DEM. The total station control set was collected in an area on the ridge of fairly high difference between the derived and USGS DEMS. Absolute differences varied slightly; the unsigned average difference was 2.8 m with the photogrammetric DEM and 4.8 m with the USGS DEM. Corresponding photogrammetric DEM values versus total station values showed a strong relationship (R2 = 0.75), while the relationship was far weaker for corresponding USGS DEM and total station values (R2 = 0.19), suggesting that the photogrammetric DEM better represented actual landscape variability. Results were also examined in a spatial context. Patterns of discrepancy were generally the same in both the photogrammetric and USGS DEMs, progressing from a slight overestimation of elevations in the DEMs at the western edge of the control area to an underestimation of elevations along the eastern edge. The main difference between the patterns of discrepancy in the DEMs was on the eastern slope of the ridge where the photogrammetric DEM predicted a more gradually sloping surface toward the valleys, more closely matching the total station control set. This is in part due to the grain of the photogrammetric DEM, which allowed for a finer resolution of terrain variation. The concentrated discrepancy along the eastern periphery of the subregion was coincident with dramatic elevation change, as the ridge abruptly slopes to the valley to the east. These findings are consistent with those of Bolstad and Stowe (1994), who reported that estimated elevations were less accurate on steep slopes in the southern Appalachian Mountains. They postulated that error on steep slopes was associated with terrain variability and ruggedness. This area was also coincident with the area of greatest disparity between photogrammetric and USGS DEMs, suggesting the superiority of the photogrammetric DEM in the study subregion, as disparity can be attributed to error in the USGS DEM, encompassing part of the eastern lip of the ridge. In summary, soft copy photogrammetry was found to create a product superior to the USGS DEM in certain respects critical to the larger goals of this collection of research. The most obvious benefit relates to resolution; the photogrammetric DEM had a 5-m resolution, while the USGS 7.5-min DEM had a spatial resolution
24
Stephen J. Walsh et al.
of 30 m. In other words, for each elevation value in the USGS DEM, the photogrammetric DEM reported 36 elevation values, capturing far greater detail in terms of surface morphology. Additionally, the total station control points showed a strong relationship with photogrammetric DEM values (R2 = 0.75), while the relationship was far weaker for corresponding USGS DEM and total station values (R2 = 0.19).
5. Selected Applications In this section, we describe three different applications in which we put the remotely sensed images to use in service of the broader scientific goals of our research. We describe methods we have used to generate ecological and geomorpholoical information products from the multiple images.
5.1. Pixel versus object classification of vegetation We compared the accuracy of classifications derived using object-oriented (OO) classification and traditional per-pixel methods. The goal was to classify a 1-m aerial image of a portion of GNP into vegetation structural classes (i.e., forest, meadow, tundra/barren, and snow). The OO approach relies on knowledge-based membership functions that explicitly define rules to classify regions (i.e., contiguous groups of pixels) rather than applying a single decision rule on a per-pixel basis. Accuracy of the classifications resulting from the two methods was assessed with the principles of fuzzy classification, allowing for the scrutiny of mixed land cover types within regions as well as within individual pixels. For the analysis, we used the four bands from the 1-m ADAR-5500 multispectral image, plus a fifth band to represent the Normalized Difference Vegetation Index (NDVI), an indirect measure of vegetation health and vigor (Lillesand and Kiefer, 2000) over the same area. The OO classification (Figure 7a) involved segmenting the data set into regions, using eCognition software, and then training a classification tree to classify patches to one of four classes (i.e., forest, meadow, tundra/barren, and snow). The ADAR5500 image was segmented into approximately 26,600 polygon objects based on spectral qualities of pixels in the five bands and the contiguity of like pixels. Training polygons were selected through a stratified sample across the range of conditions within each vegetation structural class. The meadow class showed the greatest amount of intraclass variability, thus requiring the largest number of training polygons (30 total were selected). Twenty polygons were selected for each of the other classes. The representativeness of each polygon was assessed by viewing its boundary overlaid on falsecolor infrared and grayscale NDVI image displays, as well as examining the mean values of each band for the polygon. The decision rules that made up the trained classification tree were derived from a set of spectral and form variables describing the regions. The variables were imported to S-Plus (Statistical Sciences, 1999) where the tree model was fit by successively splitting data into homogeneous subsets using the likelihood ratio, a deviance
25
Pattern–Process Relations in the Alpine and Subalpine Environments
(a) Object-Oriented Classification
(b) Maximum Likelihood Classification
Vegetation Structural Class Forest Meadow Tundra/Barren Snow
Figure 7 Results from object- and pixel-based classifications of land cover at Lee Ridge using ADAR imagery. (a) Object-oriented classification and (b) maximum likelihood classification.
measure, to maximize dissimilarity between and similarity within the resultant groups. The classification tree quantified break points in the dependent variables that could be used to segregate polygons into nested classes resulting in a binary tree structure. Decision rules derived from the classification tree were then input in eCognition, and the segmented polygons were classified in the series of steps defined by the classification tree. For comparison, the more standard maximum likelihood (ML) decision rule was applied to the data in a pixel-by-pixel manner rather than to polygons in a raster-based GIS (Figure 7b). A total of 90 polygons (signatures of 20 representative forest, tundra, and snow polygons and 30 representative meadow polygons) were ‘‘grown’’ from seed pixels for the purposes of creating a training set. Each polygon was evaluated in the context of falsecolor infrared and grayscale NDVI displays. The ML decision rule was then applied to the data set to classify each pixel to one of the four classes. The accuracies of both the object-based and ML classifications were assessed by visually interpreting and assigning classes for 100 randomly selected polygons and 100 randomly selected pixels, hereafter referred to as the polygon reference data and the pixel reference data. Interpreted classes were compared with classes assigned by the two classification algorithms to the respective polygons and pixels. Thus, the accuracy of each thematic map was evaluated twice, once with the reference data based on polygons and a second time by reference data based on individual pixels. For the OO classification, 100 polygons were selected at random from the collection of segmented regions. The segmented image was intersected with the classified ML image to produce a series of polygons, each with a single class assignment in both classifications, from which 100 could be selected at random.
26
Stephen J. Walsh et al.
The accuracy assessment methodology followed a fuzzy classification protocol suggested by Gopal and Woodcock (1994). Accuracy values for each patch or pixel ranged from 1 (absolutely wrong) to 5 (absolutely right). Each polygon/pixel of the evaluation sets received four scores indicating the degree of membership in each of the four vegetation structural classes. A series of fuzzy set operations was then used to evaluate the scores in terms of frequency, magnitude, source, and nature of errors. Fuzzy set operations resulted in a total of eight tables for each classification strategy (four for each assessment method). The frequency of matches and mismatches between the classified and interpreted classes was measured as the number of sites within each assigned class in the map that had the best classification according to visual interpretation, identified by the MAXIMUM operation as the class with the highest score. The number of sites within each assigned class that had an acceptable class label was identified with scores 3 by the RIGHT operation. Together, these operations revealed how many samples from the assessment set described the vegetation class best and how much of the sample described the vegetation class to an acceptable level, as defined by the analyst. Both classification strategies captured much of the variability in vegetation structural patterns on Lee Ridge. Forest, meadow, and tundra/barren vegetation classes covered 23–25, 40–45, and 31–32% of the study area, respectively, with snow covering less than 3%. The meadow class exhibited the greatest difference in areal coverage between the two classifications. Visually, the classifications produced similar spatial distributions of the vegetation classes. The main discrepancy between the two products was the degree of landscape patchiness, with the ML classification producing a greater number of small patches and greater fragmentation at class boundaries. A Boolean difference map revealed that the majority of the disparity in the classifications was concentrated at patch boundaries. Another area of disagreement was in the northernmost part of the study site. Here, the ML classifier favored meadow in the east while the classification tree predicted a greater proportion of tundra and forest relative to the meadow class. Similarly, forest was classified as nearly contiguous by the OO classifier in the northwest while the ML classifier predicted far more patches of meadow. The frequencies of matches and mismatches between classes derived from the classification approaches and reference classes revealed that the OO classifier performed better, with 95% of the evaluation polygons having the best class assignment versus 59% with the ML classifier and 94 versus 86% according to the pixel-based evaluation. The majorities of the polygons and pixels in each class were assigned to the best class by both classification methods, as was evidenced by the consistently low improvement in match values between the MAXIMUM and the RIGHT operations. The greatest improvement was seen in the meadow class of the ML classification with an improvement of 22% by the polygon-based assessment and 10% by the pixel-based assessment. The ML classifier performed better by both measures (i.e., MAXIMUM and RIGHT operations) when evaluated using a pixel-based evaluation; an improvement of 27% over the polygon-based assessment was reported for best class assignments and an 18% improvement in acceptable class assignments. The OO classifier performed well regardless of the evaluation procedure.
Pattern–Process Relations in the Alpine and Subalpine Environments
27
The most interesting results, as they relate to within-class variability, were expressed as the source and nature of errors. Multiple class membership was substantial in both polygons and pixels of the reference data sets. Multiple memberships were witnessed in up to 68% of polygons and 44% of pixels within a given class. Thus, the presence of multiple classes was not limited to the segmented polygons, revealing the inability to resolve completely land cover classes of patchy land cover types in spite of the availability of fine-grained data. Summarizing data over image objects in the OO classification likely removed some of the within-class variability, especially extreme values. However, the presence of multiple cover types within pixels, thus objects, suggests that within-class variability was inherent in the data and remained substantial in image objects. Polygons and pixels with a single membership were classified properly in the OO classification in nearly every case, while such polygons and pixels were frequently misclassified by the ML classifier. In the OO classification, forest and meadow polygons were ultimately segregated on the basis of brightness, the mean value in the blue band, and the ratio of the red band to brightness. The structure of the classification tree suggested the greatest complexities in discerning forest and meadow classes in the OO classification. Of the four classes, the forest and meadow classes were most similar spectrally, especially in the NDVI layer, as both are characterized by having abundant vegetation in contrast to tundra/barren and snow. Due to noise, environmental variability, and variation in vegetation structure, pixels of the same class do not have identical signatures, with structurally complex or diverse vegetation producing more varied signals. It was expected that the greatest spectral variability would exist in both of these classes because of the difference in tree species composition on either side of the ridge, the general progression of tree species from standing upright at lower elevations to lateral growth form dominance at the highest elevations (Treitz et al., 1992) and varying hydrologic conditions in the meadow class. For example, meadow was present on the upper, convex part of the ridge where conditions are relatively dry as well as in the lower, concave valleys where snowmelt is plentiful. A tenet of landscape ecology states that spatial heterogeneity influences ecological process. Both classification strategies have potential for improvement by including elevation as a surrogate for many other variables influencing process as a data band. Elevation and its derivatives share relationships with numerous other variables, including temperature, atmospheric pressure, atmospheric moisture, precipitation, solar radiation, wind exposure, soil moisture, and snow accumulation and melt through variations in slope, aspect, and curvature. Using MSS data, Walsh (1987) showed that cover type, slope angle, and slope aspect accounted for much of the spectral variability in mean band response (bands 4–7) at 182 sites in Crater Lake National Park, Oregon, with canopy characteristics being secondary to the terrain conditions in this respect. Further, he suggested the use of digital terrain models in analyses that include topography of significant roughness and relief. A high-resolution DEM of the area was known to exist. However, it was not used in this research because it was part of a larger study on landscape patterns that necessitated independence between topography and vegetation.
28
Stephen J. Walsh et al.
5.2. Linear mixture modeling To provide continuous surfaces of the intensity of tree cover across the ATE, for the purposes of comparison with models described elsewhere in this volume, we sought to map the LAI of trees with Landsat TM data. A three-stage approach was tested for addressing the problem of mixed pixels in biophysical value estimation (Brown, 2000). First, we used spectral unmixing to obtain estimates of the percentage of each pixel that was composed of tree, tundra, bare rock, and shadow (Adams et al., 1995; Gong et al., 1994). Spectral signatures obtained through on-screen image interpretation were used to define the four end-members used for mixture modeling. Although there was generally good agreement between the orthophoto- and the field-based estimates of percent trees, the fraction in trees tended to be overestimated by unmixing. The root mean square (RMS) errors from the unmixing ranged from 0 to 9.41 across the entire Park, in units of reflectance, with the highest values occurring in lakes and snow patches. Because water and snow were not included as end-members, the error in fraction estimates were quite high in areas dominated by these cover types. At snow-free treeline sites, however, the RMS errors were generally well below 1.0. The second step involved adjusting the pixel Vegetation Index (VI) values so that they represented the VI of the tree-only portion of the pixel, assuming an average VI for background components. We calculated both the simple band ratio (SR) between near-infrared reflectance (TM band 4) and red reflectance (TM band 3) and NDVI, which normalizes based on the sum of reflectances in these two bands. In essence, a second mixture model was created, this time for the VI signal. The approach assumes that the VI value for a pixel results from the linear combination of VI values from each of the end-member components, following VIpixel = f tree VItree þ f tundra VItundra þ f bare VIbare þ f shadow VIshadow þ " ð2Þ where VIpixel is the index value observed for the pixel, " is measurement error, and the inputs are the f’s and VIs for each end-member. By assuming an average VI for each of the nontree components and rearranging the equation, an estimate of the tree LAI is obtained using Equation (3): VItree = VIpixel f tundra VItundra f bare VIbare f shadow VIshadow "
ð3Þ
Brown (2000) demonstrated that the mean was generally a good estimate of the VI values in bare areas and shadow, but that substantial variability in tundra areas, which included both wet and dry tundra, may limit the accuracy of the estimate in Equation (3). Finally, the adjusted VI values were used to estimate LAI, using equations from White et al. (1997), which estimated the total LAI based on two different vegetation indices derived from field data collected from several different ecosystems within the Park. The estimates were compared with tree-only LAI (LAItree) values
Pattern–Process Relations in the Alpine and Subalpine Environments
29
measured in the field using a Li-Cor LAI-2000 and spatially referenced through differential GPS. The results (Brown, 2000) indicated that the fit of VI–LAItree relationships was improved using the unmixing-based adjustments of VI. Although the results for the SR index were weak, the NDVI–LAI relationship developed and published by White et al. (1997) provided reasonably good estimates of LAItree (+0.60 LAI) when applied to adjusted NDVI values (NDVItree). Maps of the LAI specific to conifer trees were then generated by adjusting VI values to represent only the tree component of pixels. The results indicate that adjusted NDVI can be used to predict the LAI of trees within mixed pixels much better than does unadjusted NDVI, which overestimates the LAI, because it includes nonarboreal vegetation. A number of issues affect the accuracy of LAI estimates in practice: the accuracy of estimates of end-member proportions obtained through unmixing, the adequacy of the end-member average NDVI for removing the effects of nonarboreal NDVI contributions, the nonsynchronous nature of the satellite flight and field work, and the accuracy of field estimates of LAI made using the LAI-2000.
5.3. Pattern metrics The ATE is important, because of possible sensitivity to climate change. Complex topography, disturbance processes, and geology in GNP likely represent additional local sources of variation in treeline dynamics and pattern, complicating treeline monitoring efforts. Landscape spatial metrics, however, can be used to quantify landscape structure and identify different forms of alpine treeline, as a first step toward unraveling the interplay between regional and local influences on ATE dynamics. Allen and Walsh (1996) integrated remotely sensed digital image processing, terrain modeling, and landscape structural analysis in a GIS to identify ATE pattern types in eastern GNP. Preprocessing of multidate Landsat TM data for the study area consisted of atmospheric and geometric corrections, as well as topographic normalization to improve classification accuracy. A combination of unsupervised and ML supervised classification was performed, yielding 16 cover types with an overall accuracy of greater than 90% and a kappa statistic of 0.89. Ancillary environmental data on geology and disturbances were available from a previously developed digital spatial database (Allen and Walsh, 1993; Brown, 1992; Walsh et al., 1989). Additional data layers for other important environmental gradients in the study area were derived from DEMs, including slope angle, slope aspect, exposure, solar insolation, topographic wetness, and slope curvature. Watershed modeling of DEMs was used to divide the study area into topographically homogeneous sub-basin slopes spanning the ATE. Subsequent morphometric analysis of these slopes based on spherical variance of topography enabled selection of a topographically stratified, random sample of slopes, ensuring adequate representation of ATE types. Pattern metrics, that is, algorithms used to assess the composition and spatial structure of land cover types or landscape conditions, were calculated and standardized for each slope in the sample and chosen to characterize ecotone composition, patchiness, and abruptness by quantifying patch shape and
30
Stephen J. Walsh et al.
density, density and complexity of edges, interspersion of patch types, and diversity of patch types. Patches were distinguished by stature, cover, canopy structure, and species. Pattern metrics were grouped by correlation analysis into classes representing diversity, texture, patch density, edge complexity, and abruptness. Seven metrics were chosen from these classes to minimize between-class correlation. Preliminary use of Ward’s clustering revealed that four to six clusters would group the slope samples, and k-means clustering solutions were generated for four, five, and six clusters. Split-sample replication and discriminant analysis enabled selection of the six-cluster solution as the best number to distinguish ATE pattern types for the study area. ANOVA of four theoretically similar but independent pattern metrics confirmed the robustness of the six-cluster solution. Statistical and GIS interpretation of the six clusters revealed meaningful differentiation of ATE patterns into a gradient from zonal (i.e., well-sorted patches) to heterogeneous (i.e., interspersed patches) types. Analysis of environmental gradients by cluster revealed significant differences between clusters in mean elevation, as well as in coefficients of variation of slope and solar radiation potential. These differences illustrate the combined regional control of climate and hypothesized local controls of slope disturbance and geomorphic processes on treeline pattern. Results demonstrate the value of combining remote sensing, pattern metrics, and GIS to quantify differences in alpine treeline pattern in GNP and elsewhere. Relating treeline pattern to underlying processes requires the use of landscape structural statistics that measure and are sensitive to changes in the pattern characteristic of interest (Bowersox and Brown, 2001). For example, the transition from closed-canopy subalpine forest to alpine tundra may be abrupt or gradual, depending on the spatial rate of change of cover types. The level of abruptness is thought to reflect environmental gradients and ecological processes such as competition and feedback (Malanson, 1997; Malanson and Butler, 1994). Patchiness, consisting of decreasing levels of tree cover and interspersion of cover types, however, may affect interpretation of any measure of abruptness, necessitating a landscape statistic that maximizes sensitivity to abruptness and minimizes sensitivity to patchiness (Bowersox and Brown, 2001). Bowersox and Brown (2001) simulated ecotone pattern for 25 combinations of abruptness and patchiness (with 50 replications each) and compared 11 patch and boundary statistics across the simulated patterns. A Landsat TM image encompassing ATE patterns in the Park was used as guidance for the cell size (30 m 30 m) and extent (630 m 630 m) of the simulations. Simulations were visually compared to the Landsat image to ensure approximation of real ecotone patterns. The average range, mean, and standard deviation of surface values for each group of replications were minimized to ensure that differences among groups would reflect primarily differences in spatial pattern. Both patch and boundary statistics can be derived from remotely sensed images. However, while patch statistics rely on classification of pixels into discrete cover types, which entails a loss of information, boundary statistics may be calculated from ecologically relevant, continuous, derived surfaces such as percent vegetative cover or the NDVI. Boundaries are identified on the surface as ‘‘spatially contiguous
Pattern–Process Relations in the Alpine and Subalpine Environments
31
locations with high rates of change’’ (Bowersox and Brown, 2001); location, width, shape, and distribution of boundaries are used to formulate the statistics. Each statistic was subjected to a two-way factorial ANOVA using both abruptness and patchiness levels (i.e., main effects) and a single-factor ANOVA using only abruptness levels (i.e., simple effects). The fractional contribution of variance from each factor was calculated to represent main-effect sensitivities to abruptness and patchiness and simple-effect sensitivity to abruptness. Pairwise multiple comparisons (Bonferroni method) were used to determine whether each statistic responded as expected externally (i.e., across abruptness levels) and consistently internally (i.e., across patchiness levels). Values for main-effect sensitivities to abruptness and patchiness, simple-effect sensitivity to abruptness, external consistency, and internal consistency were standardized to range from 0 to 1. Each statistic’s overall sensitivity to abruptness and insensitivity to patchiness could then be represented in a ‘‘suitability ranking’’ by summing its values for the five sensitivity and consistency properties. Although patch-based statistics exhibited reasonably high suitability rankings, two boundary statistics developed specifically for this study, dispersion (DISP) and cumulative boundary elements (CBE), exhibited the highest suitability rankings for detecting abruptness with relative insensitivity to patchiness. This study illustrates the promise of using landscape statistics to represent ecologically relevant ecotone characteristics such as abruptness.
6. Conclusions Much of the work presented in this chapter was conducted to provide data inputs to an integrated project aimed at understanding ecological and environmental processes at alpine treeline over different scales. The mapping efforts were specifically designed to address questions related to the spatial pattern of vegetation, continuous vegetation properties, and environmental parameters that affect vegetation across the ATE. Some of our mapping efforts paralleled efforts to model processes that generate vegetation patterns. Because the models focused on tree establishment, growth, and death, we focused several of our efforts on mapping the tree canopy. This led us to develop an innovative approach to mapping tree cover. At the fine scale, using aerial digital imagery at a spatial resolution of 1 m, we found that object-based classification was much more suited to identification of tree patches within the treeline vegetation mosaic. At the medium scale (i.e., spatial resolution 30 m), and because of the patchy nature of the treeline ecotone, we developed an approach to mapping continuous variables that describe tree distributions and characteristics, including percentage tree cover and the LAI of trees. The discrete approach to representing tree cover at fine resolutions and the continuous approach to representing tree cover at medium resolutions was consistent both within our mapping efforts and our process models. In addition to mapping vegetation, it was necessary for us to develop innovative approaches to mapping environmental conditions. In the rugged landscape of GNP, a key environmental variable of interest was terrain. We worked with several
32
Stephen J. Walsh et al.
existing DEMs, and developed our own high-resolution (5 m) DEM from stereo aerial photographs. These elevation data were then used to derive a range of other variables related to temperature, snow pack, soil characteristics, and more. These variables were then available for use in our process models and for statistical analyses of the patterns of vegetation (Brown, 1994). A key aspect of many of our analyses has been the ability to map vegetation and environmental characteristics in a continuous fashion across large, and sometimes inaccessible, areas of the Park. Remote sensing provides the means to collect this information. Continuous coverage permits analysis of patterns in ways that field-based samples do not. For example, analysis of patch patterns (Allen and Walsh, 1996) permits a holistic landscape analysis perspective that integrates biophysical controls with ecological processes. The frequent trade-off with wide-area coverage of this sort is the loss in thematic information associated with remote data collection. For example, much of our work focuses on landscape structural features (e.g., trees) as opposed to floristic diversity. However, through integration of field observations and sophisticated remote sensing image processing, we have been able to extract information at a finer level of detail than that available through traditional image classification approaches. Analyses of observed patterns have led us to a clearer understanding of how vegetation patterns relate to environmental factors and disturbances (Brown, 1994). They have helped us classify treeline landscapes and identify places that deviate from the general patterns (Allen and Walsh, 1996). The have helped us identify the scale dependence of vegetation patterns and the scales at which these patterns are organized (Bian and Walsh, 1993). Through such analysis, we have identified the appropriate scales for mapping and modeling patterns. Our understanding of the ATE is incomplete when we focus solely on mapping patterns. This project has explicitly focused on the pattern–process interactions by integrating observed patterns with models of processes. By coupling and comparing spatial outputs from models with the observed patterns, we have sought to gain a clear understanding of both the patterns of vegetation at the alpine treeline and why those patterns exist. These efforts complement work in other places (e.g., Bader and Ruijten, 2008 in the Andes, Schneevoigt et al., 2008 in the Alps, Wundram and Loffler in Norway), but Glacier National Park provides an exceptional perspective on the importance of the local site because of its diversity of alpine treeline pattern and process connections.
REFERENCES Ackermann, F., 1996. Techniques and strategies for DEM generation. In: Greve, C. (Ed.), Digital Photogrammetry, An Addendum to the Manual of Photogrammetry. American Society for Photogrammetry and Remote Sensing, Bethesda, MD, pp. 135–141. Adams, J.B., Sabol, D.E., Kapos, V., Filho, R.A., Roberts, D.A., Smith, M.O., et al., 1995. Classification of multispectral images based on fractions of endmembers: Application to landcover change in the Brazilian Amazon. Remote Sensing of Environment 52, 137–154. Allen, T.R., 1995. Relationship between Spatial Pattern and Environment at the Alpine Treeline Ecotone, Glacier National Park, Montana. Doctoral Dissertation, University of North Carolina, Chapel Hill, NC.
Pattern–Process Relations in the Alpine and Subalpine Environments
33
Allen, T.R., 1998. Topographic context of glaciers and perennial snowfields, Glacier National Park, Montana. Geomorphology 21, 207–216. Allen, T.R., Walsh, S.J., 1993. Characterizing multitemporal alpine snowmelt patterns for ecological inferences. Photogrammetric Engineering and Remote Sensing 59, 1521–1529. Allen, T.R, Walsh, S.J., 1996. Spatial and compositional pattern of alpine treeline, Glacier National Park, Montana. Photogrammetric Engineering and Remote Sensing 62, 1261–1268. Bader, M.Y., Ruijten, J.J.A., 2008. A topography-based model of forest cover at the alpine tree line in the tropical Andes. Journal of Biogeography 35, 711–723. Bian, L., Walsh, S.J., 1993. Scale dependencies of vegetation and topography in a mountainous environment of Montana. Professional Geographer 45, 1–11. Bolstad, P.V., Stowe, T., 1994. An evaluation of DEM accuracy: Elevation, slope, and aspect. Photogrammetric Engineering and Remote Sensing 60, 1327–1332. Bowersox, M.A., Brown, D.G., 2001. Measuring the abruptness of patchy ecotones: A simulationbased comparison of landscape pattern statistics. Plant Ecology 156, 89–103. Brown, D.G., 1992. Topographical and Biophysical Modeling of Vegetation Patterns at Alpine Treeline. Doctoral Dissertation, University of North Carolina, Chapel Hill, NC. Brown, D.G., 1994. Predicting vegetation types at treeline using topography and biophysical disturbance variables. Journal of Vegetation Science 5, 641–656. Brown, D.G., 2000. A spectral unmixing approach to leaf area index (LAI) estimation at the alpine treeline ecotone. In: Millington, C.Walsh, S.J., Osborne, P.E. (Eds.), GIS and Remote Sensing Applications in Biogeography and Ecology. Kluwer, Dordrecht, pp. 7–21. Brown, D.G., Cairns, D.M., Malanson, G.P., Walsh, S.J., Butler, D.R., 1994. Remote sensing and GIS te4chniques for spatial and biophysical analyses of alpine treeline through process and empirical models. In: Michener, W.K.Brunt, J.W., Stafford, S.G. (Eds.), Environmental Information Management and Analysis. Taylor and Francis, London, pp. 453–481. Brown, D.G., Walsh, S.J., 1991. Compatibility of non-synchronous in-situ water quality data and remotely sensed spectral information for assessing lake turbidity levels in complex and inaccessible terrain. GeoCarto International 6, 5–11. Brown, D.G., Walsh, S.J., 1992. Relationship between the morphometry of alpine and sub-alpine basins and remotely sensed estimates of lake turbidity, Glacier National Park, Montana, USA. Physical Geography 13, 250–272. Butler, D.R., Malanson, G.P., Walsh, S.J., 1991. Identification of a deltaic environment in an alpine finger lake. Environmental Professional 13, 352–362. Butler, D.R., Walsh, S.J., 1990. Lithologic, structural, and topographic influences on snow avalanche path location, East Glacier National Park, Montana. Annals of the Association of American Geographers 80, 362–378. Carrara, P.E., 1990. Surficial Geologic Map of Glacier National Park, Montana. USGS Map 1-1508-D, Miscellaneous Investigations Series 24, 387–395. Dutton, B.L., Marrett, D.J., 1997. Soils of Glacier National Park, East of the Continental Divide. Land and Water Consulting Inc., Missoula, MT. Gong, P., Miller, J.R., Spanner, M., 1994. Forest canopy closure from classification and spectral unmixing of scene components: Multispectral evaluation of an open canopy. IEEE Transactions in Geoscience and Remote Sensing 32, 1067–1080. Gopal, S., Woodcock, C., 1994. Theory and methods for accuracy assessment of thematic maps using fuzzy sets. Photogrammetric Engineering and Remote Sensing 60, 181–188. Hammer, E.S., 2004. Canopy Structure in the Krummholz and Patch Forest Zones, Glacier National Park, Montana, USA.. MA Thesis, University of North Carolina, Chapel Hill, NC. Land and Water Consulting Incorporated, 1999. Soils of Glacier National Park. Missoula, MT. Leprieur, C.E., Durand, J.M., Peyron, J.L., 1988. Influence of topography on forest reflectance using Landsat Thematic Mapper and digital terrain data. Photogrammetric Engineering and Remote Sensing 54, 491–496. Lillesand, T.M., Kiefer, R.W., 2000. Remote Sensing and Image Interpretation. John Wiley, New York. Malanson, G.P., 1997. Effects of feedbacks and seed rain on ecotone patterns. Landscape Ecology 12, 27–38.
34
Stephen J. Walsh et al.
Malanson, G.P., 2001. Complex responses to global change at alpine treeline. Physical Geography 22, 333–342. Malanson, G.P., Butler, D.R., 1994. Tree-tundra competitive hierarchies, soil fertility gradients, and the elevation of treeline in Glacier National Park, Montana. Physical Geography 15, 166–180. McGregor, S.J., 1998. An integrated geographic information system approach for modeling the suitability of conifer habitat in an alpine environment. Geomorphology 21, 265–280. McKnight, S.A., 2004. Composition and Spatial Structure of Turf-Banked Terraces in Glacier National Park, Montana. MA Thesis, University of North Carolina, Chapel Hill, NC. Schneevoigt, N.J., van der Linden, S., Thamm, H-P., Schrott, L., 2008. Detecting Alpine landforms from remotely sensed imagery. A pilot study in the Bavarian Alps. Geomorphology 93, 104–119. Statistical Sciences, 1999. S-PLUS, Version 2000 for Windows. Mathsoft Inc., Seattle. Treitz, P.H., Howarth, P.J., Suffling, R.C., 1992. Application of detailed ground information to vegetation mapping with high spatial-resolution digital imagery. Remote Sensing and Environment 42, 5–82. Walsh, S.J., 1987. Variability of Landsat MSS spectral responses of forests in relation to stand and site characteristics. International Journal of Remote Sensing 1, 105–120. Walsh, S.J., Bian, L., McKnight, S., Brown, D.G., Hammer, E.S., 2003a. Solifluction steps and risers, Lee Ridge, Glacier National Park, Montana, USA: A scale and pattern analysis. Geomorphology 55, 381–398. Walsh, S.J., Butler, D.R., 1997. Morphometric and spectral analyses of debris flows: Components of a natural hazards methodology. GeoCarto International 12, 59–70. Walsh, S.J., Butler, D.R., Allen, T.R., Malanson, G.P., 1994a. Influence of snow patterns and snow avalanches on the alpine treeline ecotone. Journal of Vegetation Science 5, 657–672. Walsh, S.J., Butler, D.R., Brown, D.G., Bian, L., 1990. Cartographic modeling of snow avalanche path location within Glacier National Park, Montana. Photogrammetric Engineering and Remote Sensing 56, 615–621. Walsh, S.J., Butler, D.R., Brown, D.G., Bian, L., 1994b. Form and pattern in the alpine environment: An integrative approach to spatial analysis and modeling in Glacier National Park, USA. In: Haywood, I.D., Price, M.F. (Eds.), Mountain Environments and GIS. Taylor and Francis, London, pp. 189–672. Walsh, S.J., Butler, D.R., Malanson, G.P., 1998. An overview of scale, pattern, and process relationships in geomorphology: A remote sensing and GIS perspective. Geomorphology 21, 183–205. Walsh, S.J., Butler, D.R., Malanson, G.P., Crews-Meyer, K.A., Messina, J.P., Xiao, N., 2003b. Mapping, modeling, and visualization of the influences of geomorphic processes on the alpine treeline ecotone, Glacier National Park, Montana, USA. Geomorphology 53, 129–145. Walsh, S.J., Cooper, J.W., Von Essen, I.E., Gallagher, K.R., 1989. Image enhancement of Landsat Thematic Mapper data and GIS data integration for evaluation of resource characteristics. Photogrammetric Engineering and Remote Sensing 56, 1135–1141. Walsh, S.J., Kelly, N.M., 1990. Treeline migration and terrain variability: Integration of remote sensing digital enhancements and digital elevation models. Proceedings, Applied Geography Conference 13, 24–32. Walsh, S.J., Moody, A., Allen, T.R., Brown, D.G., 1997. Scale dependence of NDVI and its relationship to mountainous terrain. In: Quattrochi, D.A., Goodchild, M.F. (Eds.), Scaling in Remote Sensing and GIS. Lewis Publishers, Boca Raton, FL, pp. 27–55. Walsh, S.J., Weiss, D.J., Butler, D.R., Malanson, G.P., 2004. An assessment of snow avalanche paths and forest dynamics using Ikonos Satellite data. Geocarto International 19, 85–93. Whipple, J.W., 1992. Geologic Map of Glacier National Park, Montana. US Geological Survey Miscellaneous Investigations Series Map I-1508-F, Washington, DC. White, J.D., Running, S.W., Nemani, R., Keene, R.E., Ryan, K.C., 1997. Measurement and remote sensing of LAI in Rocky Mountain montane ecosystems. Canadian Journal of Forest Research 27, 1714–1727. Wundram, D., Loffler, J., 2008. High resolution spatial analysis of mountain landscapes using a low-altitude remote sensing approach. International Journal of Remote Sensing 29, 961–974.
C H A P T E R
3
Ecotone Dynamics: Invasibility of Alpine Tundra by Tree Species from the Subalpine Forest George P. Malanson, Daniel G. Brown, David R. Butler, David M. Cairns, Daniel B. Fagre, and Stephen J. Walsh
Contents 1. Introduction 1.1. Plant’s eye view 2. Seeds to Seedlings in Open Tundra 2.1. Dispersal 2.2. Protected sites 2.3. Annual weather 3. Seedlings 3.1. Coarse scale climate 3.2. Endogenous climate modification 3.3. Microclimate 3.4. Soil 4. Tree or Krummholz Form 5. Facilitation or Inhibition? 5.1. Pattern and process 6. Conclusion References
35 36 37 38 40 43 43 44 46 48 50 51 54 55 56 57
1. Introduction Alpine tundra is an island biome and its ecotone with forest is subject to change. Like oceanic islands, alpine tundra is subject to invasion. Here we use invasion in the sense of the replacement of species of one biome by those of another, not referring to exotic species of other regions; however, although the isolation of mountaintops might create some of the same conditions that allow invasion as on islands, the supposed relative species richness of alpine tundra (we call into question the scale of comparisons) may differentiate them. The importance of Developments in Earth Surface Processes, Volume 12 ISSN 0928-2025, DOI 10.1016/S0928-2025(08)00203-4
Ó 2009 Elsevier B.V. All rights reserved.
35
36
George P. Malanson et al.
Earth system processes at global to local scales may also differ (cf. Malanson and Butler, 2002). The invasion of tundra by tree species is an expected outcome of climatic change. The most notable biological response to climatic change is the shift in the ranges of species (Malanson, 1993), and the movement of tree species to higher elevations is a well-documented response to climate warming (Rochefort et al., 1994). Advances (and retreats) of trees have been documented for periods during the last 15,000 years, and recent advances of tree species into tundra have been recorded (Alftine and Malanson 2004; Bekker, 2005; Kullman, 1988, 2002; Lloyd and Graumlich, 1997; Resler et al., 2005). Thus, the movement of the ecotone between subalpine forest and alpine tundra has been suggested as an indicator of climatic warming, and this point has been debated (Kupfer and Cairns, 1996). Holtmeier (2003) concluded that the coupling is too weak, with too many intervening processes, for use as an indicator. The invasion of tundra by trees will have consequences for the tundra biome as invasion does for other island flora and fauna (Malanson et al., 2007a, 2007b). Currently, disjunct distributions and consequent genetic diversity may be lost if tundra species are replaced by trees. Relict species (also disjunct), surviving in disequilibrium with climate and potential competitors (tree species), are most likely to suffer. As trees invade tundra, habitat for tundra animals will be lost and these island populations will also be reduced. The reverse effects of feedbacks to the climatic system – changes in albedo, roughness, and evapotranspiration – would be limited by alpine tundra area. Whereas Ko¨rner (1998) (Hoch and Korner, 2003, 2005) has advocated an explanation of treeline that rejects the carbon balance theme that we develop, we must say at the outset that we, like several others (Cuevas, 2000; Smith et al., 2003), believe that the focus should be on seedling establishment, which differs from Korner. The question of trees at the 3 m upright stem point is of secondary importance for understanding ecotone dynamics. Holtmeier (2003) summarized much of what is known about mountain treelines globally in his valuable book. Much of the book is devoted to a factor by factor discussion of the environmental conditions that determine treelines as static phenomena. He also includes biotic effects of regeneration and the influence of the plants on the site factors. He then briefly covers the dynamics of past and present changes of treelines. Here, we will examine what we see as the most important steps in the dynamics and bring in the factors identified by Holtmeier (2003) with, of course, emphasis on what we see in Glacier National Park (GNP), Montana. Much of our detailed work is on Lee Ridge near the northeast corner of the park, but we have study sites across nearly its entire north-south extent. To link the dynamics to the factors we adopt a ‘‘plant’s eye view.’’
1.1. Plant’s eye view To examine the invasibility of tundra, we take a plant’s eye view. In his seminal work, Harper (1977) advocated a plant’s eye view for plant ecology. By applying an environmental sieve (sensu van der Valk, 1981), that is, an analysis of factors that would eliminate potential species from a community at multiple time steps, at every stage of a plant’s life cycle one can best see its relations with other biota and the
Ecotone Dynamics
37
abiotic environment. The plant’s eye view is a trans-temporal and multifactorial expansion of Liebig’s Law of the Minimum. We apply a plant’s eye view to the potential invasion of alpine tundra by subalpine trees. The environmental sieve, from the plant’s eye view, must include its requirements, its tolerances, and its ability to avoid what it cannot tolerate. We consider as necessities the requirements of photosynthesis, growth, and reproduction: carbon, water, nutrients, and energy. We consider as factors to tolerate as factors that would decrease photosynthesis or growth and factors to avoid are those that would destroy existing growth (i.e., stress or disturbance, sensu Grime, 1979). The two categories can overlap, so that disturbance can be tolerated or stress avoided. We categorize tolerances/avoidances as shortages in the necessities plus other factors that reduce photosynthesis and disturbance, in the sense of loss of modules, such as predation, the physical damage of abrasion, and losses due to desiccation or freezing. One approach that we have used in the past to capture the plant’s eye view of resources has been to simulate the carbon dynamics of tree species in the alpine treeline ecotone. This explicit carbon balance approach does not always accurately predict treeline location or pattern, but where it does not it can be accounted for by nonequilibrium conditions rather than mis-specification, and it does allow us to examine krummholz as well as tree forms (Cairns, 2005; Cairns and Malanson, 1997, 1998). Given a plant’s eye view, the local conditions become extremely important. At the end of this chapter, we discuss how local scale processes propagate across scales into landscape patterns, but the local process focus sets the tone for this and the other chapters in this book. We concentrate on aspects of microtopography (and microgeomorphology) and microclimate because these are the factors that matter: from the plant’s eye view! We consider three of the major life stages of a potential woody invader of alpine tundra: seed stage, which includes initial location and germination, seedling stage, when the invader is small enough to have rather direct contact with tundra species, and sapling or krummholz stage, when the plant grows upright or prostrate.
2. Seeds to Seedlings in Open Tundra We summarize the seed to seedling stage as a sieve located in time between the processes of dispersal and establishment (Figure 1). From a plant’s eye view, not all places are equal at the seed stage, and seeds differ widely in their requirements and tolerances. A seed carries its own limited energy source, but for successful establishment it needs to reach soil for water and nutrients. The seed needs to be able to anchor itself in soil and to reach sunlight. To do so it must reach what we call a protected site. The importance of protected or safe sites for tree species establishment in the alpine zone has been suggested for several places (Li et al., 2003, Tyrol; Vostral et al., 2002, Vermont; Erschbamer et al., 2001, Alps; see Eastham and Jull, 1999 for a relevant silvicultural study). The importance of protection should become clear below.
38
George P. Malanson et al.
Vegetation Solifluction
Boulders
Animals Next year?
Protected site?
No Next year?
Yes
Seed predation?
No
Annual weather OK?
Yes
Establishment
Dispersal
No
Yes Next year?
Winter desiccation Spring frost Summer drought Fall frost Sudden < –40
Figure 1 The environmental sieve for a tree seed to become a seedling in alpine tundra.
Holtmeier (2003) pays considerable attention to the production of seeds. Because seed production at treeline is so limited (Oosting and Reed, 1952) and the species involved can disperse by wind and birds, we are less concerned with local seed production but more concerned with where seeds disperse to (cf. Marr, 1977).
2.1. Dispersal Seed rain can be an important component of tree advance into tundra (Malanson, 1997, 2001), but the geography of dispersal, at least two scales, determines the effects. At a coarse scale, seed dispersal depends on long distance mechanisms, such as wind and birds, and seed sources away from treeline: in the subalpine forest. The trees near treeline do not produce seeds in abundance (Tranquillini, 1979). At this scale, the primary concern is the loss of sources of Pinus albicaulis (whitebark pine). Due to blister rust, the number of seed producing whitebark pine in the subalpine forest is greatly reduced. In GNP, Kendall et al. (1996) reported 44% mortality, and Resler and Tomback (2008) observed blister rust at treeline. A seed arrives at a tundra site primarily by one of two means – carried by wind or a bird. These two routes are based on major differences among the seeds themselves and have consequences for where the seeds get to in tundra. There are similarities, however, in the tundra sites that are starting points at this stage. Here we will consider the primary genuses that occur at treeline sites in North America: Abies, Picea, and Pinus. Abies and Picea seeds have large wings that are large relative to seed size and are dispersed by wind (Young and Young, 1992). As a genus, Pinus is quite diverse in seed morphology and dispersal mechanism (Young and Young, 1992). Some have
Ecotone Dynamics
39
quite large wings, but those most commonly found at treeline sites (Pinus albicaulis, Pinus flexilis, and Pinus aristata) are wingless and have large nuts attractive for birds and rodents to cache. The most effective dispersal agent for these pines is Clark’s nutcracker (Nucifraga Columbiana; Tomback, 2001 and references therein). At a fine scale, it is the sites where seeds stay that matter. While wind would distribute seeds widely, the high wind across alpine tundra makes it an unlikely resting place. Seeds can stay in only a few places: where the wind will eddy in an aerodynamic shelter. For bird-dispersed seeds, the choice is made by the birds matter. Clark’s nutcracker will bury seeds where they will not be covered by deep snow, such as on exposed ridges (Tomback, 2001), but they also need soil in which they can dig; this is also found in aerodynamic shelters or where animal burrowing or turf exfoliation have exposed it (the latter may also be somewhat sheltered, depending on wind direction; it usually is on Lee Ridge). The coincidence of sites where seeds will stay and fine soils sets the stage for further development. Although arriving by wind or by bird, both types of seeds need to find an amenable seedbed. They need to arrive in a place where its radicle can reach soil and its cotyledons can reach sunlight. Most sites in tundra do not meet both criteria. Many tundra locations do not have easily accessible soil. A common condition is to find a rocky surface with little growth. Due to internal processes such as solifluction, which concentrates stones on the surface, or deflation by persistent wind, some sites provide no soil at all (Butler and Malanson, 1989; Butler et al., 2004). Seeds blown here will land on stones and not have access to soil. These sites are probably less attractive to birds, since they would be quite difficult to dig onto. In places where these processes, solifluction in particular, have developed fine sediments at the surface, these sites are occupied by tundra plants. In dry exposed sites, tundra may occupy stripes or the risers (the more vertical parts) of stairsteps created by solifluction. This vegetation is often tightly packed and may prevent a wind-dispersed seed from falling close enough to the surface for its radicle to reach soil. Birds could dig into this mat but may not find it easy. In moister sites, tundra grows larger but soil may be more accessible for seeds carried by either birds or wind. At this stage, seeds may be in a good position for germination but face difficulties at a later stage, discussed below. The potential good germination may be why Clark’s nutcrackers have been observed to prefer windswept ridges – in addition to finding a site where the seed will not be retrievable in early spring due to late-lying snow (Tomback, 2001). Thus, barriers to success in reaching the soil include landing on rock or an impenetrable surface (such as frozen ground) or snow. Saturated soil, as found just below a snow patch, is also a barrier, as metabolism becomes aerobic on germination. A protected site is one at which soil with a useable range of moisture is exposed. A number of studies have identified soil surface conditions as a limiting factor, but it varies, with spruce being more sensitive than fir or pine (Alexander and Shepperd, 1990; Alexander et al., 1990; Arno and Hoff, 1990). Seed dispersal by wind will not be random. The topographic sites with fine soils are also places more likely to final seed deposition, but these sites also have the denser tundra, which can trap and hold seeds that fall into it. Thus, it is only a limited range of topographic protection that will be best for seeds.
40
George P. Malanson et al.
2.2. Protected sites We have identified four major factors that can create a protected site: solifluction, boulders, vegetation, and animal activity. 2.2.1. Solifluction One place in which the soil resource is available is where solifluction processes open this resource (Figure 2). Solifluction is a mass movement of the surface layers of soil that occur during the summer when saturation is maintained by deeper permafrost. Many tundra sites are underlain by active or relict solifluction (Bowman and Damm, 2002; Butler and Malanson, 1989; Malanson et al., 2007b). Solifluction produces a surface that is a mix of stony areas – inhospitable to seeds – and vegetated soil surfaces. Characteristic patterns of stripes or steps of alternating stony treads and vegetated risers, together comprising turf-banked terraces, are common. Several remote sensing analyses have been able to quantify these patterns (Walsh et al., 2003a, 2003b). The vegetated soil presents a different challenge for the seed in that dense cover may still prevent it from reaching and penetrating the soil. In some places, however, active and even relict solifluction can, through continuing geomorphic activity, create sites that provide seeds with the necessities. Butler et al. (2004) have described the process of turf exfoliation. The lower part of turf-banked terraces can erode, producing a small deposit of fine soil that might be ideal for seeds. They noted that seedlings have been establishing at such microsites, and Resler et al. (2005) reported that 28% of establishment occurred in such places. 2.2.2. Boulders Other areas where fine soil can accumulate would be in the lee of microtopographic features such as boulders where eolian deposition of fine soil is possible (Figure 3). These are sites where tundra is likely to be well established, however.
Figure 2
Seeds can find a protected site with fine soil among solifluction treads and risers.
Ecotone Dynamics
Figure 3
41
Another protected site for seeds is in the lee of a boulder.
On the other hand, areas with fine soils are attractive to burrowing and seedburying animals, which opens up this resource. Resler et al. (2005) reported that 60% of establishment occurred in the lee of boulders. 2.2.3. Vegetation The other feature that alters eolian deposition is the vegetation itself. Seastedt et al. (2004) have characterized the alpine environment as one of deflation and loss of material and the treeline, by which they mean full upright trees, as one of deposition and accumulation of material. Maher et al. (2005) and Maher and Germino (2006) reported additive positive effects of surrounding herb and tree species vegetation on seedling establishment of conifers in the treeline ecotone. We still need to look at differences within the alpine that are differentiated by a range of kinds and sizes of vegetation. Patches of krummholz may be able to accumulate fine sediment within and immediately downwind of their canopy, but the relatively high wind and streamlined shape of krummholz patches may minimize this input. In dwarf tree patches, where the trees have grown upright from seedlings or developed in to upright forms from what began as krummholz, the aerodynamics can lead to deposition more easily, and in this type of patch we do observe lenses of fine sediment within the upper soil profile. The vegetation itself can, on the other hand, lead to a surface that is not accessible for a seed to reach the soil resource. Dense mats of tundra can keep a seed up away from the surface. Mats of Dryas would seem dense enough to do this although seeds might eventually fall through – but a filter can reduce numbers. Within krummholz and dwarf tree canopies, we find mats of indurated duff (Figure 4). These mats are made up of needles that are cemented together; they have a hard surface and are 3–4 cm thick. We discuss their hydrologic role below, but the surface could prevent a radicle from reaching soil resources. In addition to mats, dense upright tundra, characterized by Lesica (2002) as turf, could also limit the ability of a seed to reach the soil surface.
42
George P. Malanson et al.
Figure 4 Mats of indurated duff under krummholz and dwarf tree canopies may alter the exchange (infiltration and evaporation) of water between the soil and the atmosphere.
2.2.4. Animals Even more directed will be dispersal by animals. For trees invading tundra, the role of birds in moving and burying pine nuts may be crucial. Tomback (2001) summarized the importance of the connection between whitebark pine seeds and Clark’s nutcracker. They two are codependent. From our perspective, the dependence of the pine on the bird is a limiting factor, but the interdependence is also important. While we have emphasized protected sites above seed production, here is a point where seed production becomes most important in our region. Whitebark pine have been decimated in recent years by white pine blister rust (Cronartium ribicola), an exotic disease fungus. This loss of trees means a weaker seed source and potentially fewer nutcrackers, together meaning less establishment in tundra. In addition to directly carrying seeds and choosing and creating a seed bed, animals can create soil surface conditions that make a microsite amenable to seedling establishment just by digging. Small mammals create burrow mounds and their burrows occasionally erode. Among larger mammals, grizzly bears tear out large areas of dense tundra. Each brings fine sediment to the surface. While we have not as yet observed seedling establishment on these sites, they do have some of the conditions needed for successful germination. We may not have seen seedlings at such sites because only seeds that avoid predation can be successful. The seeds do not have avoidance strategies, but species can compensate by more plentiful or mast production. In addition to the Clark’s nutcrackers that use seeds they buried that did not become seedlings, grizzly bears eat these seeds, but usually at lower elevations where they are cached by red squirrels. It is hard to imagine that a seed that falls on a burrow digging would not be preyed on by the local ground squirrel or marmot.
Ecotone Dynamics
43
At very early stages following germination, spruce seedlings are susceptible to predation by Juncos (Noble and Shepperd, 1973). Rodents can also have an impact but are probably more important as seed predators than seedling predators. Large animals can browse and trample seedlings; this effect varies greatly across western North America, because the numbers of large animals (mountain sheep and goats would be primary at treeline) have been greatly lowered by humans during the past century.
2.3. Annual weather In addition to finding a protected site and not being eaten, a seed must survive long enough on its own resources to establish in the face of climatic extremes. While we treat microclimate below in more detail, and we explicitly recognize that the general microclimate necessary for growth is critical in the months after seed reserves are used, from a seed’s perspective we simply look at extremes that may affect it here. Extremes of radiation, desiccation, and cold are important. Summer desiccation is the major factor in first-year seedling mortality (Alexander and Shepperd, 1990). Drought at any time can kill seedlings, but the more commonly prolonged dry soils of summer are the main limit. Full sunlight can inhibit germination in some species, and the high-intensity radiation at some high elevation sites (spruce more so than fir), possibly combined with high stem temperatures, can also lead to seedling mortality (Alexander and Shepperd, 1990). Germino et al. (2002) have reported better survival rates for seedlings in moderately dense tundra where the negative effects of competition are outweighed by the positive effects of reduced incident radiation. Fall frost can also kill first-year seedlings if they are very young (<10 weeks). In subsequent years, spring (early growing season) frosts are more likely to cause mortality. Lastly, frost heaving has been identified as a source of mortality for seedlings which can remain a threat for several years (Alexander and Shepperd, 1990). This process is part of the geomorphic process of solifluction, and so is a reversal of the potential for the production of protected sites. A seedling that establishes in a spot of turf exfoliation can be later ‘‘turfed out’’ (Kajimoto, 2002, in Japan). This process may be worse on other sites within the tundra, however, especially on the stony treads where frost sorting of stones is ongoing (cf. Butler and Malanson, 1998).
3. Seedlings At the seedling stage, we examine two related steps: survival and growth (Figure 5). At this stage, we consider them to be closely related because we judge both based on carbon gain and allocation. Holtmeier (2003) concluded that carbon allocation was more important than carbon gain and he treated the carbon balance as one of many factors, but here he was focusing on tree forms, and he based his argument primarily on net primary productivity whereas we balance the input with a full accounting of losses, including those of disturbance. Because we are focused on seedlings, we believe that carbon balance is the critical factor and that carbon
44
George P. Malanson et al.
Soil
Nutrients
Water Krummholz Yes1
No Site condition OK? >minC
Microclimate
Figure 5
Yes
Tree or krummholz
Yes2
Tree
Establishment
Mortality
Wind effects
The environmental sieve for a tree seedling to become a tree or krummholz.
allocation comes second. We recognize that an individual seedling can have periods of negative carbon balance (Germino and Smith, 1999), but to survive and establish, a seedling must maintain a positive carbon balance over the course of a year (larger plants may be able to withstand longer negative periods). Seedling growth then depends on how positive the balance is. The calculation of carbon balance depends primarily on microclimate (which includes water in all ways), but plant physiology is the operating system, which differs among species (and individuals), and secondarily on nutrients. From the plant’s eye view, the sieve is the sum of the effects on carbon balance of the microclimate and soil effects rather than any single factor (although one can alone be negative enough to be the sole causal factor). We unpack microclimate is several steps (Figure 6). Long-term trends in climate, including global climate change, are constraints on what can occur at local sites. Modifying the global pattern are regional trends, such as the Pacific Decadal Oscillation (PDO), seasonal shifts, and synoptics.
3.1. Coarse scale climate The Pacific Decadal Oscillation (PDO) is likely to be an important influence on treeline dynamics, at least in the PNW and northern Rockies. The PDO is a mode of North Pacific sea surface temperature variability on a multidecadal time scale (Mantua et al., 1997). Analyses of PDO focus on the contrast between its positive and negative phases. The PDO index from 1900 to present shows one major negative phase (from the late 1940s to the late 1970s, the end reported as 1977), but considerable variation exists within the multidecadal trends. The PDO has climatic correlations that are strongest in the Pacific northwest and northern Rocky Mountains (Zhang et al., 1997), where, importantly from our perspective, more snow is found when the index is negative (Kovanen, 2003; Moore et al., 2002; Selkowitz et al., 2002). Thus, the PDO affects
45
Humidity
Air temperature
Leaf temperature
Soil temperature
Microclimate Soil moisture
Infiltration
Shade
Roughness
Rain
Snow
UV
PAR
Albedo
Endogenous mod Interception
Radiation
Wind Strength
Direction
Precipitation
Annual weather
Synoptics PDO, etc.
Climate
Seasonality
Ecotone Dynamics
Snow bank
Figure 6
Multiple scales mediate the translation of macroclimate on plant microclimate.
forest growth in the Pacific Northwest (Kadonaga et al., 1999; Peterson and Peterson, 2001; Peterson et al., 2002). The effect of PDO on treeline in GNP is through snow. Selkowitz et al. (2002) established that GNP had greater spring snow pack during the negative phase of the PDO, and Alftine et al. (2003) determined that the temporal pattern of seedling establishment on Lee Ridge indicated that it was associated with the negative phase of the PDO during the 1950s–1970s, during which period it accelerated (probably due to positive feedbacks), but it stopped after the PDO switched to positive phase in 1977. Synoptic conditions translate regional weather to site weather. The patterns of circulation and thus the boundaries between air masses and the direction of atmospheric movements such as storm fronts varies synoptically and are in turn controlled by larger patterns such as sea surface temperatures and the interaction of global circulation with topography. The most significant synoptic factors in the northern Rocky Mountains would be modification of the mean summer and winter positions of the polar jet stream. Different synoptic conditions for GNP would mean altered wind strength and direction and different amounts of precipitation due to upstream effects of topography. We can see major differences in the effects of individual storms during different synoptic conditions where the amounts of precipitation on different aspects of a mountain are reversed. Different synoptic conditions might also have different lapse rates. At the next stage, the annual weather experienced at a site can be divided into components of wind, radiation, and precipitation (we consider temperature as an outcome of modification of these components). Wind velocity is high due to elevation, and its velocity and direction are modified locally by topography. Radiation, high in both photosynthetically active radiation (PAR) and UV, depends on elevation and latitude. Precipitation can be high due to orography, but relationship is not 100%; rain is often intense; snow is in large proportion but much of it blows away. These factors vary through the year as constrained by the level above, but the components that are of most interest, from the plant’s eye view,
46
George P. Malanson et al.
are what affects the needles, stems, and roots, but these effects are best considered after they have been modified endogenously.
3.2. Endogenous climate modification How do the plants themselves modify the climate they experience? They alter the basic conditions of environmental physics. This type of microclimate modification has been studied in excruciating detail, and here we pick out the observations and lines of reasoning that apply at alpine treeline. 3.2.1. Wind Wind direction and velocity is modified by surface roughness, which is in large part due to the plants. A small lone seedling has little effect on wind, especially in comparison to the surrounding topography and tundra plants. Slightly larger seedlings, however, begin to affect flow, and patches of small trees or krummholz definitely alter the wind environment. Wind affects trees directly via pressure and abrasion (Holtmeier, 2003; Milne, 1995) and indirectly by modifying the other factors of microclimate and soils. The slowing of wind by plants thus has two major directions for effects on plants themselves. Winds contribute to the breaking of branches, removal of leaves/needles, and the reduction of cuticle on leaves. These factors together are losses of carbon and nutrients that must be accounted for in the carbon balance. The extent to which the vegetation lowers wind speeds also lessens these losses. Slower wind speed affects the way in which winds alter the other components of microclimate. Wind evens out temperatures locally, reducing the differences caused by differences in albedo, and so lessening the higher canopy temperatures of a midsummer day. Trees and krummholz friction lessens this modification, especially in the interior of canopies. High winds also mean high transpiration losses of water and lower water use efficiency, but again friction lessens this problem. Lower wind speeds allow a thicker leaf boundary layer through which water loss must diffuse before reaching the turbulent atmosphere, thus increasing water use efficiency. Perhaps most importantly, wind redistributes snow, and vegetation changes the pattern. At high elevations in GNP, much of the snow is removed by wind. Similar conditions have been observed elsewhere in the Rockies (Hiemstra et al., 2002; Seastedt et al., 2004). The vegetation causes snow to be deposited in eddies, causing accumulations much higher than average precipitation. Krummholz canopies can be completely full of snow up to their tops. 3.2.2. Radiation The radiation is affected by reflectivity and absorption. Dwarf trees and krummholz differ in albedo from tundra, with lower reflectivity. Overall, the increased energy absorption leads to differences in temperatures, discussed immediately below. The direct differences in radiation are due to shade. A developing canopy creates shade. In general, shade will limit photosynthesis, but it can also limit damaging UV radiation. It is possible, but untested, that the upper needles in the very dense canopies found in krummholz may serve a primary role in shading the needles
Ecotone Dynamics
47
lower in the canopy and not actually add much to net primary productivity (NPP) themselves. Our findings for krummholz in GNP is that Leaf Area Index (LAI) is tine the neighborhood of 12, and this is all compressed into a canopy 20–30 cm in depth, below which there is an empty volume except for stems. Light penetration to the surface is 10 % of ambience. Although tundra is a high-light environment, shading and competition for light can be important. Germino and Smith (1999) documented increased photosynthesis for new seedlings in shade in Abies lasiocarpa and Picea engelmannii at Wyoming treeline. The other radiation-related endogenous modification is by albedo. The darker green of conifer needles absorbs more radiation than tundra plants or bare rock surfaces. This absorption raises leaf (needle) temperature. This higher temperature could raise NPP, except that it also leads to a higher respiration rate and to higher potential evapotranspiration and probably lower water use efficiency. Another radiation component is outgoing radiation at night. Greater exposure to the sky, the inverse of shade, means cooler nighttime temperatures. Germino and Smith (1999), in conjunction with their findings regarding shade from solar radiation, also found increased photosynthesis when sky exposure was reduced at night. They even showed that the architecture of needles on the ends of branches might be adaptations to these stresses. Germino et al. (2002) noted that survival of emergent seedlings was lower on the southern edges of tree islands and on slopes with a southern aspect. They concluded that high sunlight leads to higher seedling mortality because it exacerbates moisture stress. Additionally, high sunlight immediately following colder nighttime temperatures also led to higher seedling mortality on the eastern side of tree islands as opposed to the western side. Germino and Smith (1999) concluded that exposure, both day and night, was the most important factor of seedling survival. Smith et al. (2003) put more emphasis on sky exposure than any previous review. Exposure to the sky at night lowers seedling temperatures, while exposure during the day can increase temperatures but may be more complicated in terms of the spatial relations among sites and plants than Smith et al. (2003) suggested. These ideas have been supported by additional site-specific studies (Johnson et al., 2004; Maher and Germino, 2006; Maher et al., 2005). 3.2.3. Precipitation Interception and infiltration together have a great effect on soil moisture. Alpine tundra has a full range of soil moistures, ranging from saturated (when not frozen) sites near snow banks or glaciers that provide constant inflow to dry conditions in which precipitation as rain is intense and has relatively high runoff to precipitations as snow which blows off. Where precipitation and infiltration can be increased by endogenous modification, the soil moisture of otherwise dry sites can be greatly increased. Infiltration is affected by the root structure of the vegetation, which is greater for krummholz and trees than for alpine tundra species, but at alpine treeline an additional factor is the role of imbricated litter and duff. Under krummholz, we have found a layer composed primarily of intact but desiccated needles that is about 4 cm thick. The layer appears to be glued together, possibly by resin from the
48
George P. Malanson et al.
needles, as it can be cut and lifted as plates. It is to some degree water repellant but does allow infiltration; we have no experimental results, just observation of pouring 0.5 l of water onto a plate for over 5 min, after which infiltration had begun. While it thus may be repellant to a summer rain, it likely might be saturated and less repellant during spring snowmelt. It may also inhibit water loss by evaporation from the soil. Interception is normally thought of as a loss of water from the system because much of it evaporates (what is not drip or stemflow). But this rain-centered view does not capture the dynamics of snow. Snow that is intercepted can sublimate away, but more important is that intercepted snow can be trapped by the canopy and not blown elsewhere; moreover, snow blowing from elsewhere can be trapped. Snow accumulates not just on the leaves and branches of vegetation, but in stands of dwarf trees, and krummholz can form snow drifts and even fill the entire space up to the top of a krummholz canopy with a dense wind pack. Even upright trees can be encased in snow (or rime), which insulates them against some damage, but this can result in breakage. The snow bank has both positive and negative effects on the environment of seedlings. Most notably, the snow trapped by the vegetation provides a moisture source longer into the summer, when the site might otherwise be too dry. Conversely, if the snow lasts too long, the growing season is too short. The snow bank also affects soil temperature, similar to shading. For the problem of tree seedlings moving into tundra, one of the most important factors is that the snow provides protection from the negative effects of wind, but they must first establish with other protection before this feedback takes effect. Another cause of mortality to branches and thus a carbon loss is snow mold, Herpotrichia coulteri. Although this problem has been studied mostly in Europe, we see evidence of it in GNP.
3.3. Microclimate The outcome of endogenous modification, from the plant’s eye view, is the microclimate experienced by leaf, stem, and root tissue. The two components of microclimate are temperature and moisture. 3.3.1. Temperature The overall association of treeline with elevation and altitude indicate that temperature plays a central role. Shade and albedo together modify temperatures, primarily during summer days. Higher temperatures would seem to promote increased NPP in an environment that at a coarse scale is cold limited, but at specific places, the temperature of individual needles and higher water loss combined with higher respiration rates will lower NPP. The effect of trees species on the local temperature is to increase canopy temperature during the growing season. The darker canopies of conifers have a lower albedo and absorb more energy than tundra plants or exposed rock. The higher canopy temperatures, while leading to higher rates of photosynthesis all else being equal, also lead to higher rates of respiration and to higher evapotranspirative losses. Also, Ko¨rner (1998) identified the role of shade as an endogenous
Ecotone Dynamics
49
modification leading to lower soil temperatures as an important negative feedback regulating tree advance into tundra. Soil temperatures are definitely colder under krummholz on a summer afternoon than they are under tundra. This effect cannot limit initial seedling establishment and development, however, except under and immediately adjacent to trees and krummholz. This effect may still be important if this location is otherwise the majority of protected sites. He is concerned with the allocation of carbon, finding that the products of photosynthesis cannot be effectively allocated to growth if the soil temperature is low, but he is focused on the tree stage. Temperature through the growing season affects metabolism. The microclimate temperature for a seedling will be determined by the regional macroclimate as modified by local albedo. In tundra environments, conifers usually have a lower albedo than forbs or semiwoody prostrate forms (in GNP, Salix spp.). The length of the growing season is also an energy factor that varies from a plant’s eye view. The length of the frost-free season is a primary control on the amount of carbon that a seedling can gain. In addition to the macroclimatic temperature regime, the important factor in determining the growing season is snow. 3.3.2. Moisture The water available for photosynthesis is that which is available in the soil for plants minus that which they lose through transpiration. Most of the treeline sites that we see in GNP are quire dry. Because snow is blown off by wind, the actual annual precipitation is much less than would be expected (Figure 7). Although there is somewhat deep soil that can hold moisture during the summer (Schmid, 2004), the generally coarse soils and steep slopes result in relative aridity. This aridity is highly variable in space, being modified by the vegetation itself, especially in snow banks that can persist into the summer. For initial seedling establishment, the dry surface soils are a strong barrier, but once tree species vegetation is established, the feedbacks
Figure 7
Much of the alpine zone has snowfall blown away and is thus dry.
50
George P. Malanson et al.
in fine soils and organic matter plus a potential snowbank change the moisture of the site drastically. This change is one of the greatest differences between the growth faced by a new seedling and established krummholz or dwarf trees. A moisture challenge faced primarily by the established krummholz or dwarf trees, however, is winter desiccation. When soils are frozen, losses of moisture from leaf tissue, as may occur on relatively warm, sunny days, cannot be replaced and the needles can die. The incidence of this damage is difficult to predict, but it is probably related to wind exposure in GNP, given that it is most common on higher slopes with a southwest exposure and, within these, in the parts of patches with little protection (Cairns, 2001). For seedlings, the plant’s eye view sieve reduces to a narrow opening wherein seedlings can simultaneously avoid cold temperatures and frosts, desiccation, UV, and abrasion in the protection provided by larger plants and to some degree, at least initially, by microtopographic features.
3.4. Soil The characteristics of soil that allow positive growth/carbon balance for a tree species seedling in alpine tundra are the presence of fines, which hold water and nutrients. Within the standard (climate, organisms, relief, parent material, and time) representation of soil genesis (first formulated by Dokuchaev, 1883), we will take time as given, subsume relief in geomorphology, focus on the microclimate, and differentiate organisms to examine the roles of litter and duff, animals, and micro-organisms. Parent material: It provides the minerals that, depending on weathering (and thus on the other factors), determine the proportion of fines and the status of mineral nutrients. On Lee Ridge, the underlying Cretaceous shale is overlain by alluvium rich in limestone from the Helena formation. • Microclimate: Weathering rates and the transport of fines are affected by soil temperature and infiltrating water. • Litter/duff: These contribute organic matter. Organic content under tundra is minimal, and the development of trees and krummholz can alter near-surface organic matter greatly by adding intact and decomposed matter. Organic matter increases the structure of the soil, water-holding capacity, and nutrients, for example, organic matter and nitrogen are often highly correlated (cf. Malanson and Butler, 1994 for this relation at treeline in GNP). • Microorganisms: Bacteria especially alter the status of nutrients, especially nitrogen. Their role relative to vegetation change in tundra has been documented in the Arctic, where increasing shrub cover traps snow that maintains higher winter sol temperatures, which allows more microbial activity and processing of nitrogen. While no evidence is available, the same may be true in alpine tundra. • Geomorphology: Geomorphic processes that alter the surface and near-surface distributions of soil texture and structure. Solifluction and creep especially separate soil per se from larger clasts. • Animals: The digging and eating, living and dying of small mammals alters soil water and nutrients. Animal activity increases infiltration and adds and concentrates organic matter.
Ecotone Dynamics
Figure 8
51
Tree seedlings face competition from tundra species when both are the same size.
In general, we do not know enough about soils. From a plant’s eye view at the seedling stage, getting resources relative to other plants is crucial. Competition thus plays a role. The tough competition for tree seedlings invading tundra is the tundra itself (Figure 8). Germino and Smith (1999) found that while some grassy tundra helped seedling establishment by providing protection from incoming UV and lowering nighttime cooling, dense tundra inhibited seedlings. Malanson and Butler (1994) hypothesized that on slopes with higher nutrients, tundra would grow more densely and trees would have difficulty invading, and throughout GNP we do find a positive association of treeline elevation with substrate clast size. Cairns (1999) found further evidence that local differences in nutrients between krummholz and tundra sites within the ecotone indicated some positive feedback. However, we have also observed an association of advancing tree species in fingers and patches with bordering denser tundra, which we termed a ‘‘green wave’’ on Lee Ridge (Walsh et al., 2003b), that perhaps indicates a heretofore unrecognized negative feedback in which trees, perhaps by trapping snow, facilitate tundra which then resists new trees.
4. Tree or Krummholz Form If the site conditions are sufficient that a positive carbon balance can be maintained, the next question at treeline is whether the form will be krummholz or a tree, possible a dwarf tree (Figure 9). This is an important distinction because the feedbacks to modifiers of microclimate and to a lesser extent soil depend on this form. Trees and krummholz will have different aerodynamic and radiation effects, and thus a series of differences in microclimate; they will have different litter conditions and thus potentially different soils; they will provide different habitats for animals.
52
(a)
George P. Malanson et al.
(b)
Figure 9 Growth forms at treeline are usually krummholz (a), but can be dwarf trees (b).
Krummholz can change form to become trees, and studies have focused on that conditions under which a krummholz patch can begin to extend vertical leaders that become trees. We have observed a pattern in which the krummholz develops initially and then, as it enlarges in a linear finger aligned with the prevailing wind (a hedge, sensu Holtmeier, 1982), tree forms develop at the downwind end later in the process (Figure 3). We consider the tree form to be the default and krummholz in need of explanation. That the development of krummholz is a response to high wind seems to be a given, but the development of vertical leaders from krummholz, and thus a step to establish a tree form secondarily, has been linked to other factors of climatic amelioration (Millar et al., 2004). Niklas (1998, 1999) provided a mechanical rationale for linking the development of krummholz directly to wind. He argued that for plants to survive in wind, they must reduce drag and torque on their stems. He noted that tall trees reduce drag and torque by having horizontally broad canopies; where massive trunks are not an option the horizontal form of krummholz may also reduce torque. One aspect of tree response to torque is twisting, which leads to the definition of krummholz. Studies on the relationship of wind to canopies in commercial studies only indicate that progress could be made in determining these forces. It may be necessary to consider the mechanical and carbon depleting effects of wind in combination. As above, the effect of wind on carbon balance is through its effects on needles and their local environment: decreased boundary layer thickness, decreased summer temperature, abrasion of the surface, and mechanical damage. All of these lead to increased evapotranspiration and to winter kill via needle desiccation. The parts of the plant that are most exposed – those growing upward – are more likely destroyed, leaving growth to proceed along the lower growing, prostrate branches. One characteristic that we have observed that has not been reported elsewhere in the literature is a form that we call upright krummholz. This form has distinct individual stems (Figure 10) (but below ground they may result from some layering) that begin with horizontal growth but turn upward and form a vertical stem of about 1 m in length. At the top of this stem, there is a dense canopy of needles 20–30 cm in depth similar to that found on prostrate krummholz, and the overall
Ecotone Dynamics
53
Figure 10 Upright krummholz has vertical stems but they end in a compressed dense canopy.
canopy has an aerodynamic form similar to that of the prostrate form. It appears that many of the trees on the lower part of Lee Ridge began >100 years ago with this horizontal turning vertical J shape and later became full-size trees ( >10 cm DBH, >5 m height). Endogenous processes turn macroclimate into microclimate, which affects NPP, carbon balance, and growth. The key endogenous processes, created by the development of vegetation and, from our perspective central to allowing the advance of tree species into alpine tundra, are roughness, infiltration, interception (including snowbanks), shade, and albedo. Roughness is the factor that represents the effect of vegetation on wind. Vegetation creates friction and reduces wind momentum and velocity. As a result, eddies form in the vicinity of vegetation, and a layer of nominally still air at the surface of each leaf or needle, through which gases must diffuse, is thicker. The thicker this boundary layer, the lower is water loss by evapotranspiration and the higher will be leaf temperature on summer days. Less wind allows the local effects of the endogenous modifications of moisture and energy inputs to be relatively stronger. Shade and albedo together modify temperatures, primarily during summer days. Higher temperatures would seem to promote increased NPP in an environment that at a coarse scale is cold limited, but at specific places, the temperature of individual needles and higher water loss combined with higher respiration rates will lower NPP. Seasonal variation is important. An increased canopy of tree species leads in winter to greater snow storage and potentially higher soil temperatures, certainly more insulated sols. This could contribute to winter desiccation, a notable problem in the treeline ecotone. In summer, the increasing canopy will create conditions that favor photosynthesis by limiting moisture loss but at some points shade will limit carbon storage but limit UV damage. In spring and fall, canopies maintain higher temperatures that can increase NPP while moisture is abundant, helping
54
George P. Malanson et al.
lengthen the growing season, and it is possible that the effect of shading on soil temperature may be reversed with canopy lowering radiation loss, at least for a short period.
5. Facilitation or Inhibition? The importance of the role of vegetation itself in determining the advance of tree species into tundra will depend on where and when one looks and at what one looks (Table 1). Most notably, Ko¨rner (1998) (Hoch and Korner, 2003, 2005) has argued that the primary limitation on treeline is the presence of trees. They contend that trees cast shade, leading to cooler soil temperatures, which inhibits the allocation of carbon for tree growth. This logic is one of a negative feedback in which the position of treeline will be maintained. We and others (Smith et al., 2003), on the other hand, have argued in favor of a positive feedback (Bekker et al., 2001; Malanson, 1997). We concede Korner’s mechanistic argument for where a line of full size (3 m height) trees is found, but we are looking at something else. The advance of tree species into tundra is the important problem, and it cannot be limited by shading. A small seedling does not cast enough shade to limit itself. If
Table 1 Feedbacks observed and hypothesized in the alpine treeline ecotone
Positive feedbacks Microclimate Albedo Temperature Wind Snow accumulation Potential evapotranspiration Shade UV radiation Sky exposure (night temperature) Soil Input of organic matter Trapping of fines (via wind effect) Negative feedbacks Microclimate Shade PAR Soil temperature Wind Snow accumulation Soil Imbricated duff
Ecotone Dynamics
55
Korner’s hypothesis was the limiting factor for treeline, we might expect to see a carpet of small seedlings out in tundra that, once they grow enough to cast shade, they would die. Nowhere in the world is this carpet of seedlings observed. Instead, there are two lines of argument for positive feedback or facilitation: observation and simulation. The stronger is based on observation. There is a positive association of new seedlings with extant trees or krummholz in most treeline locations (Callaway, 1998; Germino et al., 2002; Hattenschwiler and Smith, 1999; Maher et al., 2005). Also, Marr (1977) reported expansion of krummholz islands by layering that showed positive feedback. Simulations show that the spatially aggregated fractal patterns observed at treeline would be difficult to produce without facilitation. How do these factors combine to affect the possible advance of tree species into tundra? We have several simulation studies that help elucidate this issue in GNP by examining spatial pattern.
5.1. Pattern and process Observed spatial pattern is primarily via remote sensing. In addition to the coarse scale patterns that locate and define treeline where processes will be of interest (Walsh et al., 1998; see Brown et al., 1994 for a rationale for a multiscale approach to modeling), higher resolution analysis links pattern to process. Allen and Walsh (1996) established that the spatial pattern of alpine treeline in GNP is often fractal. Allen et al. (2004) examined possible controls of alpine treeline using geostatistics; they reported that application of semivariograms followed by computing fractal dimension indicated scales of spatial autocorrelation. Fractal patterns are to be expected in nature, but the spatial aggregation inherent is important for our argument. In addition, Cairns and Waldron (2003) examined the ecotone in GNP for evidence of a sigmoid wave pattern. They found this pattern, but not at all sites due to several complicating factors and scale issues. Malanson et al. (2001) modeled the resource averaging hypothesis. This hypothesis states that treelines are caused by diminishing, and patchy resources and trees need to gain soil resources from a larger area, so that when the average over that larger area drops below a minimum trees cannot be sustained while smaller plants can. Malanson et al. (2001) found that the hypothesis would produce a pattern of trees that should closely match the observed pattern of soil resources regardless of averaging. Using indicators of soil resource pattern described by Malanson et al. (2002), they were able to conclude that the observed patterns of trees on Lee Ridge do not match the pattern of soil resources in tundra, and so the resource averaging hypothesis is not explanatory at this level of pattern; the more complex patterns suggested that positive feedback was operating. Walsh et al. (2003b) used this same simulation but forced the effect of linear solifluction features into it. The results were more similar to the observed patterns, but observed linear forms at Lee Ridge run across (at about a 20 angle) the solifluction forms, not along them, and the simulation is unable to reproduce this effect. At alpine sites, the high wind is a dominant factor. At Lee Ridge, we know that the winds are consistently from the southwest when at high velocity. Alftine and
56
George P. Malanson et al.
Malanson (2004) included wind direction in a cellular automaton simulation of the invasion of tundra by trees. Only when they included positive feedback effects of trees on the probability of further invasion n their immediate neighborhood did the simulation produce spatial patterns similar to those observed on Lee Ridge. The simulations matched observations best when the positive feedbacks were skewed directionally to account for wind, that is, when the sites just downwind of existing trees had their probability of seedling establishment increased, the simulated pattern was most like the observed pattern. Zeng and Malanson (2006) created a very abstract simulation of alpine treeline in order to investigate the consequences of feedbacks on system dynamics (cf. Zhang et al., 2008 at a coarser scale). They assumed that positive feedback would vary with distance from other trees in a bell shape, because the negative effects should cancel out the positive ones immediately adjacent to trees, or in a spot of tundra crowded by surrounding trees, positive effects would reach their maximum at some distance when negative effects were negligible and would thereafter decline. This assumption rests on the observations that the affects of shade on reduced PAR and soil temperature, the effects of imbricated duff, and the possibility of too-deep snow all occur immediately under or adjacent to trees while the effects of reduced wind on PET, the increased summer moisture of a larger snowpack, and the effects of increased nutrients might have broader spatial spread. The most important result of their simulation was that the rate of advance of trees up a slope varied considerably due to differences that developed in the spatial pattern of the trees themselves, independent of exogenous drivers such as rate of climate change or change in the underlying site conditions. Advance sometimes accelerated and later decelerated. While at first glance it appeared that there was a cycle in this rate, the advance was in fact fractal. Zeng and Malanson (2006) concluded that a multiscale explanation was necessary: At the local scale, individual tree establishment and mortality drive the dynamics. At a middle scale, patch dynamics of groups of trees affect their surroundings differently depending on density and pattern. At the landscape scale, pattern sometimes begins with many small patches preceding an acceleration and at other times the patches coalesce across the landscape, lowering the positive feedback effects. From our perspective here, what is important is that (a) facilitation dominates the dynamics of an advancing treeline and (b) the advance is thus effectively decoupled from exogenous drivers such as climate change.
6. Conclusion The invasion of alpine tundra by tree species must be a two-step process. First, tree seedlings must pass an environmental sieve that is strongly associated with geomorphic patterns and processes. Second, positive feedback, the facilitation of seedling establishment and growth by other conifers, either krummholz or dwarf trees, is critical to developing enough cover to make a difference in tundra. Both steps are primarily in terms of how a plant gets the resources it needs – energy, water, nutrients – or avoids the aspects of the environment that are lethal. Individual plants have to pass the first sieve to escape from the seedling to the sapling
Ecotone Dynamics
57
stage. They then can potentially develop into krummholz or, if passing another sieve, into a tree form. In both instances, they need the right combination of multiple factors for a considerable but finite amount of time. Thus, the invasion of tundra by tree species must initially be spatially and temporally restricted, and then the vegetation itself must overcome some of these restrictions by its own growth in order to pass the next sieve or expand via growth or reproduction. Climate change can alter these sieves or relations in multiple ways. First, it can alter the resources as in change in the precipitation, the evapotranspiration, or the length of the growing season, and it can alter the access to resources as in modifying disturbances such as solifluction activity. Second, the negative aspects of resource use could be mitigated, for example, with fewer killing frosts or droughts; however, amelioration of one factor could exacerbate another as in frosts and winter desiccation. Third, climate change can alter the relative importance of feedback. The importance of spatial pattern and the relevance of self-organization to constraining and/or accelerating the rate of invasion would then change. In GNP, we have seen mixed messages that indicate the complexity, in the complexity science sense (Malanson, 1999) of the pattern–process relationship at alpine treeline (cf. Rietkerk and Van de Koppe, 2008 for a discussion of vegetation in general). Where we initially saw little if any advance of trees into tundra based on repeat photography (Butler et al., 1994; Klasner and Fagre, 2002), we now see that it depends where and when one looks. In some environments, trees advanced upslope rather quickly as the Little Ice Age ended. This advance probably led to the early development of ribbon forest on some sites which now have a pattern of meadow and mature forest. Possibly related was the advance that documented on the lower slopes of Lee Ridge. Even since the time of the early photographs, however, we see advances in places that were not photographed, with Lee Ridge being the best example. We conclude that GNP exhibits sensitivity to the initial spatial conditions – in terms of both where trees were and what the upslope resources were – that prevailed at the end of the Little Ice Age and that this difference has been affected somewhat by variation in the climate within the park, especially in wind strength and direction. GNP provides an excellent location for pushing the study of treeline into the arena of complexity science, but comparison with other environments in western North America and globally will be necessary in order to appreciate the full importance of the spatial and temporal legacies on how process and pattern can change our present alpine treeline ecotone.
REFERENCES Alexander, R.R., Shearer, R.C., Shepperd, W.D., 1990. Abies lasiocarpa (Hook.) Nutt. Subalpine fir. In: Burns, R.M., Honkala, B.H. (Eds.), Silvics of North America Volume 1, Conifers. USDA Forest Service Agricultural Handbook 654, Washington, DC, pp. 60–70. Alexander, R.R., Shepperd, W.D., 1990. Picea engelmannii Parry ex Engelm. Engelmann spruce. In: Burns, R.M., Honkala, B.H. (Eds.), Silvics of North America Volume 1, Conifers. USDA Forst Service Agricultural Handbook 654, Washington, DC, pp. 187–203. Alftine, K.J., Malanson, G.P., 2004. Directional positive feedback and pattern at an alpine tree line. Journal of Vegetation Science 15, 3–12.
58
George P. Malanson et al.
Alftine, K.J., Malanson, G.P., Fagre, D.B., 2003. Feedback-driven response to multidecadal climatic variability at an alpine treeline. Physical Geography 24, 520–535. Allen, T.R., Walsh, S.J., 1996. Spatial and compositional pattern of alpine treeline, Glacier National Park, Montana. Photogrammetric Engineering and Remote Sensing 62, 1261–1268. Allen, T.R., Walsh, S.J., Cairns, D.M., Messina, J., Butler, D.R., Malanson, G.P., 2004. Geostatistics and spatial analysis: Characterizing form and pattern at the alpine treeline. In: Bishop, M., Shroder, J. (Eds.), Geographic Information Science and Mountain Geomorphology. Springer Verlag-Praxis Scientific, London, pp. 189–214. Arno, S.F., Hoff, R.J., 1990. Pinus albicaulis Engelm. Whitebark pine. In: Burns, R.M., Honkala, B.H. (Eds.), Silvics of North America Volume 1, Conifers. USDA Forst Service Agricultural Handbook 654, Washington, DC, pp. 268–279. Bekker, M.F., 2005. Positive feedback between tree establishment and patterns of subalpine forest advancement, Glacier National Park, Montana, USA. Arctic, Antarctic, and Alpine Research 37, 97–107. Bekker, M.F., Malanson, G.P., Alftine, K.J., Cairns, D.M., 2001. Feedback and pattern in computer simulations of the alpine treeline ecotone. In: Millington, A.C., Walsh, S.J., Osborne, P.E. (Eds.), Applications of GIS and Remote Sensing in Biogeography and Ecology. Kluwer, Dordrecht, pp. 123–138. Bowman, W.D., Damm, M., 2002. Causes and consequences of alpine vascular plant diversity in the Rocky Mountains. In: Korner, C., Spehn, E.M. (Eds.), Mountain Biodiversity. Parthenon Publishing, Boca Raton, FL, pp. 35–47. Brown, D.G., Cairns, D.M., Malanson, G.P., Walsh, S.J., Butler, D.R., 1994. Remote sensing and GIS techniques for spatial and biophysical analyses of alpine treeline through process and empirical models. In: Michener, W.K.Brunt, J.W., Stafford, S.G. (Eds.), Environmental Information Management and Analysis. Taylor and Francis, London, pp. 453–481. Butler, D.R., Malanson, G.P., 1989. Periglacial patterned ground, Waterton-Glacier International Peace Park, Canada and USA. Zeitschrift fur Geomorphologie 33, 43–57. Butler, D.R., Malanson, G.P., 1998. Site locations and characteristics of miniature patterned ground, eastern Glacier National Park, Montana, USA. Landform Analysis 2, 45–50. Butler, D.R., Malanson, G.P., Cairns, D.M., 1994. Stability of alpine treeline in northern Montana, USA. Phytocoenologia 22, 485–500. Butler, D.R., Malanson, G.P., Resler, L.M., 2004. Turf-banked terrace treads and risers, turf exfoliation and possible relationships with advancing treeline. Catena 58, 259–274. Cairns, D.M., 1999. Multi-scale analysis of soil nutrients at alpine treeline in Glacier National Park, Montana. Physical Geography 20, 256–271. Cairns, D.M., 2001. Patterns of winter desiccationin krummholz forms of Abies lasiocarpa at treeline sites in Glacier National Park, Montana, USA. Geografiska Annaler 83A, 157–168. Cairns, D.M., 2005. Simulating carbon balance at treeline for krummholz and dwarf tree growth tree forms. Ecological Modelling 187, 314–328. Cairns, D.M., Malanson, G.P., 1997. Examination of the carbon balance hypothesis of alpine treeline location, Glacier National Park, Montana. Physical Geography 18, 125–145. Cairns, D.M., Malanson, G.P., 1998. Environmental variables influencing carbon balance at the alpine treeline ecotone: A modeling approach. Journal of Vegetation Science 9, 679–692. Cairns, D.M., Waldron, J.D., 2003. Sigmoid wave transitions at alpine treeline. Geografiska Annaler 85A, 115–126. Callaway, R.M., 1998. Competition and facilitation on elevation gradients in sub-alpine forests of the northern Rocky Mountains, USA. Oikos 82, 561–573. Cuevas, J.G., 2000. Tree recruitment at the Nothofagus pumilio alpine timberline in Tierra del Fuego, Chile. Journal of Ecology 88, 840–855. Dokuchaev, V.V., 1883. Russian Chernozem. Selected Works of V.V. Dokuchaev, vol. 1, pp. 14–419. Moscow, 1948 and Israel Program for Scientific Translations, 1967, Jerusalem. Eastham, A.M., Jull, M.J., 1999. Factors affecting natural regeneration of Abies lasiocarpa and Picea engelmannii in a subalpine silvicultural systems trial. Canadian Journal of Forest Research 29, 1847–1855. Erschbamer, B., Kneringer, E., Schlag, R.N., 2001. Seed rain, soil seed bank, seedling recruitment, and survival of seedlings on a glacier foreland in the Central Alps. Flora 196, 304–312.
Ecotone Dynamics
59
Germino, M.J., Smith, W.K., 1999. Sky exposure, crown architecture, and low-temperature photoinhibition in conifer seedlings at alpine treeline. Plant Cell and Environment 22, 407–415. Germino, M.J., Smith, W.K., Resor, A.C., 2002. Conifer seedling distribution and survival in an alpine-treeline ecotone. Plant Ecology 162, 157–168. Grime, J.P., 1979. Plant Strategies and Vegetation Processes. Wiley, New York. Harper, J.L., 1977. Population Biology of Plants. Academic Press, London. Hattenschwiler, S., Smith, W.K., 1999. Seedling occurrence in alpine treeline conifers: A case study from the central Rocky Mountains, USA. Acta Oecologica 20, 219–224. Hiemstra, C.A., Liston, G.E., Reiners, W.A., 2002. Snow redistribution by wind and interactions with vegetation at upper treeline in the Medicine Bow Mountains, Wyoming, USA. Arctic, Antarctic, and Alpine Research 34, 262–273. Hoch, G., Korner, C., 2003. The carbon charging of pines at the climatic treeline: a global comparison. Oecologia 135, 10–21. Hoch, G., Korner, C., 2005. Growth, demography and carbon relations of Polylepis trees at the world’s highest treeline. Functional Ecology 19, 941–951. Holtmeier, F.K., 1982. ‘‘Ribbon-forest’’ and ‘‘hedges.’’ Strip-like distribution patterns of trees at the upper timberline in the Rocky Mountains. Erdkunde 36,142–153. Holtmeier, F.-K., 2003. Mountain Timberlines: Ecology, Patchiness, and Dynamics. Kluwer, Dordrecht. Johnson, D.M., Germino, M.J., Smith, W.K., 2004. Abiotic factors limiting photosynthesis in Abies lasiocarpa and Picea engelmannii seedlings below and above the alpine timberline. Tree Physiology 24, 377–386. Kadonaga, L.K., Podlaha, O., Whiticar, M.J., 1999. Time series analyses of tree ring chronologies from Pacific North America: Evidence for sub-century climate oscillations. Chemical Geology 161, 339–363. Kajimoto, T., 2002. Factors affecting seedling recruitment and survivorship of the Japanese subalpine stone pine, Pinus pumila, after seed dispersal by nutcrackers. Ecological Research 17, 481–491. Kendall, K.C., Schirokauer, D., Shanahan, E., Watt, R., Reinhart, D., Renkin, R., et al., 1996. Whitebark pine health in northern Rockies national park ecosystems: A preliminary report. Nutcracker Notes 7, 16. Klasner, F.L., Fagre, D.B., 2002. A half century of change in alpine treeline patterns at Glacier National Park, Montana, USA. Arctic, Antarctic, and Alpine Research 34, 49–56. Ko¨rner, C., 1998. A re-assessment of high elevation treeline positions and their explanation. Oecologia 115, 445–459. Kovanen, D.J., 2003. Decadal variability in climate and glacier fluctuations on Mt Baker, Washington, USA. Geografiska Annaler 85A, 43–55. Kullman, L., 1988. Holocene history of the forest-alpine tundra ecotone in the Scandes Mountains (Central Sweden). New Phytologist 108, 101–110. Kullman, L., 2002. Rapid recent range-margin rise of tree and shrub species in the Swedish Scandes. Journal of Ecology 90, 68–77. Kupfer, J.A., Cairns, D.M., 1996. The suitability of montane ecotones as indicators of global climatic change. Progress in Physical Geography 20, 253–272. Lesica, P., 2002. A Flora of Glacier National Park. Oregon State University Press, Corvallis. Li, M., Yang, J., Krauchi, N., 2003. Growth responses of Picea abies and Larix decidua to elevation in subalpine areas of Tyrol, Austria. Canadian Journal of Forest Research 33, 653–662. Lloyd, A.H., Graumlich, L.J., 1997. Holocene dynamics of treeline forests in the Sierra Nevada. Ecology 78, 1199–1210. Maher, E.L., Germino, M.J., 2006. Microsite differentiation among conifer species during seedling establishment at alpine treeline. Ecoscience 13, 334–341. Maher, E.L., Germino, M.J., Hasselquist, N.J., 2005. Interactive effects of tree and herb cover on survivorship, physiology, and microclimate of conifer seedlings at the alpine tree-line ecotone. Canadian Journal of Forest Research 35, 567–574. Malanson, G.P., 1993. Comment on modeling ecological response to climatic change. Climatic Change 23, 95–109. Malanson, G.P., 1997. Effects of feedbacks and seed rain on ecotone patterns. Landscape Ecology 12, 27–38.
60
George P. Malanson et al.
Malanson, G.P., 1999. Considering complexity. Annals, Association of American Geographers 89, 746–753. Malanson, G.P., 2001. Complex responses to global change at alpine treeline. Physical Geography 22, 333–342. Malanson, G.P., Butler, D.R., 1994. Tree – tundra competitive hierarchies, soil fertility gradients, and the elevation of treeline in Glacier National Park, Montana. Physical Geography 15, 166–180. Malanson, G.P., Butler, D.R., 2002. The Western Cordillera. In: Orme, A. (Ed.), Physical Geography of North America. Oxford University Press, Oxford, pp. 363–379. Malanson, G.P., Butler, D.R., Fagre, D.B., 2007a. Alpine ecosystem dynamics and change: A view from the heights. In: Prato, T., Fagre, D. (Eds.), Sustaining Rocky Mountain Landscapes: Science, Policy and Management of the Crown of the Continent Ecosystem. RFF Press, Washington, DC, pp. 85–101. Malanson, G.P., Butler, D.R., Fagre, D.B., Walsh, S.J., Tomback, D.F., Daniels, L.D., et al., 2007b. Alpine treeline of western North America: Linking organism-to-landscape dynamics. Physical Geography 28, 378–396. Malanson, G.P., Butler, D.R., Cairns, D.M., Welsh, T.E., Resler, L.M., 2002. Variability in an edaphic indicator in alpine tundra. Catena 49, 203–215. Malanson, G.P., Xiao, N., Alftine, K.J., 2001. A simulation test of the resource averaging hypothesis of ecotone formation. Journal of Vegetation Science 12, 743–748. Mantua, N.J., Hare, S.R., Zhang, Y., Wallace, J.M., Francis, R.C., 1997. A Pacific interdecadal climate oscillation with impacts on salmon production. Bulletin of the American Meteorological Society 78, 1069–1079. Marr, J.W., 1977. The development and movement of tree islands near the upper limit of tree growth in the southern Rocky Mountains. Ecology 58, 1159–1164. Millar, C.I., Westfall, R.D., Delany, D.L., King, J.C., Graumlich, L.J., 2004. Response of subalpine conifers in the Sierra Nevada, California, USA, to 20th-century warming and decadal climate variability. Arctic, Antarctic, and Alpine Research 36, 181–200. Milne, R., 1995. Modelling mechanical stresses in living Sitka spruce stems. In: Coutts, M.P., Grace, J. (Eds.), Wind and Trees. Cambridge University Press, Cambridge, pp. 165–181. Moore, G.W.K., Alverson, K., Holdsworth, G., 2002. Variability in the climate of the Pacific Ocean and North America as expressed in the Mount Logan ice core. Annals of Glaciology 35, 423–429. Niklas, K.J., 1998. The influence of gravity and wind on land plant evolution. Review of Palaeobotany and Palynology 102, 1–14. Niklas, K.J., 1999. A mechanical perspective on foliage leaf form and function. New Phytologist 143, 19–31. Noble, D.L., Shepperd, W.D., 1973. Gray-headed juncos important to first season mortality of spruce. Journal of Forestry 71, 763–765. Oosting, H.J., Reed, J.F., 1952. Virgin spruce-fir of the Medicine Bow Mountains, Wyoming. Ecological Monographs 22, 69–91. Peterson, D.W., Peterson, D.L., 2001. Mountain hemlock growth responds to climatic variability at annual and decadal time scales. Ecology 82, 3330–3345. Peterson, D.W., Peterson, D.L., Ettl, G.J., 2002. Growth responses of subalpine fir to climatic variability in the Pacific Northwest. Canadian Journal of Forest Research 32, 1503–1517. Resler, L.M., Butler, D.R., Malanson, G.P., 2005. Topographic shelter and conifer establishment and mortality in an alpine environment, Glacier National Park, Montana. Physical Geography 26, 112–125. Resler, L.M., Tomback, D.F., 2008. Blister rust prevalence in krummholz whitebark pine: Implications for treeline dynamics, Northern Rocky Mountains, Montana, USA. Arctic, Antarctic, and Alpine Research 40, 161–170. Rietkerk, M., van de Koppe, J., 2008. Regular pattern formation in real ecosystems. Trends in Ecology and Evolution 23, 169–175. Rochefort, R.M., Little, R.L., Woodward, A., Peterson, D.L., 1994. Changes in the distribution of subalpine conifers in western North America: A review of climate and other factors. The Holocene 4, 89–100.
Ecotone Dynamics
61
Schmid, G.L., 2004. The Role of Soils in Recording Environmental Change at Alpine Treeline in Glacier National Park, Montana. Unpublished doctoral dissertation, Department of Geography, Texas State University-San Marcos. Seastedt, T.R., Bowman, W.D., Caine, T.N., McKnight, D., Townsend, A., Williams, M.W., 2004. The landscape continuum: A model for high-elevation ecosystems. BioScience 54, 111–121. Selkowitz, D.J., Fagre, D.B., Reardon, B.A., 2002. Interannual variation in snowpack in the crown of the continent ecosystem. Hydrological Processes 16, 3651–3665. Smith, W.K., Germino, M.J., Hancock, T.E., Johnson, D.M., 2003. Another perspective on altitudinal limits of alpine timberlines. Tree Physiology 23, 1101–1112. Tomback, D.F., 2001. Clark’s nutcracker: Agent of regeneration. In: Tomback, D.F., Arno, S.F., Keane, R.E. (Eds.), Whitebark Pine Communities. Island Press, Washington, DC, pp. 89–104. Tranquillini, W., 1979. Physiological Ecology of the Alpine Timberline. Springer-Verlag, New York. van der Valk, A.G., 1981. Succession in wetlands – a Gleasonian approach. Ecology 62, 688–696. Vostral, C.B., Boyce, R.L., Friedland, A.J., 2002. Winter water relations of New England conifers and factors influencing their upper elevational limits. I. Measurements. Tree Physiology 22, 793–800. Walsh, S.J., Bian, L., McKnight, S., Brown, D.G., Hammer, E.S., 2003a. Solifluction steps and risers, Lee Ridge, Glacier National Park, Montana, USA: A scale and pattern analysis. Geomorphology 55, 381–398. Walsh, S.J., Butler, D.R., Malanson, G.P., 1998. An overview of scale, pattern, and process relationships in geomorphology: A remote sensing and GIS perspective. Geomorphology 21, 183–205. Walsh, S.J., Butler, D.R., Malanson, G.P., Crews-Meyer, K.A., Messina, J.P., Xiao, N., 2003b. Mapping, modeling, and visualization of the influences of geomorphic processes on the alpine treeline ecotone, Glacier National Park, Montana, USA. Geomorphology 53, 129–145. Young, J.A., Young, C.G., 1992. Seeds of Woody Plants in North America. Dioscorides Press, Portland, OR. Zeng, Y., Malanson, G.P., 2006. Endogenous fractal dynamics at alpine treeline ecotones. Geographical Analysis 38, 271–287. Zhang, Y.A., Peterman, M.R., Aun, D.L., Zhang, Y.M., 2008. Cellular automata: Simulating alpine tundra vegetation dynamics in response to global warming. Arctic, Antarctic, and Alpine Research 40, 256–263. Zhang, Y., Wallace, J.M., Battisti, D.S., 1997. ENSO-like interdecadal variability: 1900–1993. Journal of Climate 10, 1004–1020.
C H A P T E R
4
Geomorphic Patterns and Processes at Alpine Treeline David R. Butler, George P. Malanson, Lynn M. Resler, Stephen J. Walsh, Forrest D. Wilkerson, Ginger L. Schmid, and Carol F. Sawyer
Contents 1. Introduction 2. Coarse-Scale Processes 2.1. Snow avalanches as treeline disturbance agents 2.2. Debris flows as treeline disturbance agents 3. Medium-Scale Processes 3.1. Turf-banked terraces 3.2. Eolian processes at treeline 4. Fine-Scale Processes and Landforms 4.1. Turf exfoliation 4.2. Boulders 4.3. Needle-ice pans 4.4. Frost heaving and churning 5. Additional Comments on the Possible Role of Animals at Treeline 6. Conclusions References
63 65 65 67 68 69 71 73 73 74 76 79 81 81 82
1. Introduction Geomorphology encompasses the landforms of a landscape and the past and present surficial processes that have sculpted them. In the alpine treeline ecotone of Glacier National Park (GNP), geomorphic processes operative at a variety of temporal and spatial scales have created the spectacular landscape we see today. These processes in turn exert an overarching influence on the nature of soil development in the alpine environment (Butler et al., 2007). Geomorphology and soils combine to create the landscape on which plants endeavor to become established and survive in the severe climatic environment at and above treeline. In some cases, geomorphic processes and soil development assist in creating Developments in Earth Surface Processes, Volume 12 ISSN 0928-2025, DOI 10.1016/S0928-2025(08)00204-6
Ó 2009 Elsevier B.V. All rights reserved.
63
64
David R. Butler et al.
environments amenable to tree seedling establishment and survival, whereas in other cases act to completely preclude establishment success. The geomorphic processes of the alpine environment do not carve landforms in a sterile environment separate from atmospheric and tectonic forces but should instead be considered as part of a continuum of processes operative in the alpine environment (Seastedt et al., 2004). We suggest that this environment also needs to be examined at a continuum of spatial and temporal scales, sensu Malanson and Butler (2002) (cf. Malanson et al., 2007; Post et al., 2007; Walsh et al., 1997). Temporal scales are significant because disequilibrium exists in the GNP alpine environment. The impact of the Laramide Orogeny, responsible for the Lewis Overthrust and its attendant mountain-building processes in GNP, simply cannot be overstated. GNP also still bears sharp testament to the widespread, intense effects of Pleistocene glaciation and is still adjusting valley sideslopes and vegetation zones to the changing landforms, slopes, and climatic conditions that have characterized the post-Pleistocene period (approximately the past 14,000 years in GNP). The Little Ice Age (LIA) (approximately 1500–1850 AD in GNP) also impressed its own set of geomorphic and climatic characteristics on the GNP alpine environment and adjustments to that period of glaciation and cold climate continue. Many local treelines were elevationally depressed by the harsher climate and geomorphic processes of the LIA. Glacial and rock glacier advances during the LIA created harsh, often barren surfaces that could support trees under modern climatic conditions. Such sites, still being revealed as glaciers recede in the post-LIA period (since approximately 1850 in the American West), are only slowly undergoing treeline expansion today, primarily due to limited soil resources. Landslides on U-shaped glacial valley sideslopes, and their attendant effects in depressing treeline below its climatic optimum (Figure 1), are yet another example of landscape adjustment to climatic and geomorphic effects produced at different intensities and temporal scales.
Figure 1 Rockslides and talus deposits preclude upward advancement of trees along the eastern edge of the Lewis Overthrust Fault, on Singleshot Mountain.
Geomorphic Patterns and Processes at Alpine Treeline
65
The general elevation and shape of alpine treeline have traditionally been defined on the basis of air temperature (e.g., the 10°C isotherm for the warmest month), soil temperature, and/or net carbon balance of plants (Holtmeier, 2003). While these factors clearly influence the location of the ecotone, ongoing and recent geomorphic processes and landforms also influence and, in some instances, control the elevation, pattern, and, especially, dynamics of alpine treeline. Becwar and Burke (1982) cited a mean timberline elevation in GNP of 1,870 + 220 m above sea level (asl), and Brown (1994) described an ‘‘open forest’’ category analogous to the alpine treeline ecotone that ranges between 1,700 and 2,100 m asl. As Becwar and Burke’s (1982) work illustrated, the elevation of treeline in GNP is highly variable when compared to treeline elevation in the Colorado Rockies of Rocky Mountain National Park. Much of this variability in GNP is a result of a more active geomorphic environment that exists in the Colorado Rockies. In this chapter, we examine the geomorphic processes and landforms that influence the position and pattern of treeline at a variety of spatial scales (coarse, medium, and fine) in GNP.
2. Coarse-Scale Processes At coarse scales in GNP (e.g., across valleys), snow avalanches, debris flows, landslides, and rockfall/talus deposits typically originate above, but many descend well below and impact, the treeline (Butler and Malanson, 1996; Butler and Walsh, 1994; Walsh et al., 1994; Wilkerson and Schmid, 2003). Snow avalanches and debris flows are especially widespread in the treeline ecotone in GNP (Butler, 1979; Butler and Walsh, 1990; Butler et al., 1992; Walsh et al., 1990, 1994, 2004) and distinctly depress treeline below the regional, climatically influenced, treeline elevation along valley flanks. The following sections provide a basic summary of our previous work on these processes that has revealed the strong influence of avalanches and debris flows on treeline elevation in GNP.
2.1. Snow avalanches as treeline disturbance agents Snow avalanche paths in GNP number well in excess of 1,000, and in some valleys as much as 50% of the area may be covered by avalanche paths (Butler, 1979; Butler and Walsh, 1990; Walsh et al., 2004) (Figure 2). Throughout much of GNP, because of their widespread distribution on both sides of the Continental Divide, snow avalanches are the primary control on the position of treeline. In general, the higher the starting zone, the more likely it is that avalanches will descend into the forest below. As the avalanches pass into the forest, trees are uprooted, decapitated, or severely tilted, coarse and fine clasts are excavated and deposited downslope, and a disturbance treeline well below climatic limits is the net result (Butler, 2001; Butler and Malanson, 1985, 1990, 1992; Butler et al., 1992; Walsh et al., 1994, 2004). Funneling topography enhances the likelihood of avalanche descent well below, and associated subsequent disturbance depression of, climatic treeline (Butler and Malanson, 1992).
66
David R. Butler et al.
Figure 2 Snow avalanche paths where the disturbance-induced treeline is well below climatic treeline, Pinchot Creek drainage.
The geographic pattern of avalanche paths, and by extension the shape of the disturbance treeline, in GNP is strongly influenced by spatial patterns of lithology, structure, and topography (Butler and Malanson, 1990, 1992; Butler and Walsh, 1990; Butler et al., 1992; Walsh et al., 1990). Avalanche path positions are largely determined by the spatial pattern of resistant lithologies such as metamorphic outcrops, structural lineaments, and topographic couloirs (Butler and Walsh, 1990; Walsh et al., 1990) (Figure 3). In east-central GNP, Butler and Walsh (1990) mapped 121 avalanche paths and identified the primary controls on their spatial position – the position of >50% of the paths was controlled by geologic structure, approximately 30% were controlled by topography, and >10% by lithology. All three groups descended to terminal elevations well below the climatic treeline range of approximately 1,700–1,800 m asl (Becwar and Burke, 1982; Brown, 1994). Topographically and lithologically controlled avalanche paths have the greater impacts on treeline, with both groups descending to average elevations of approximately 1,535 m asl (Walsh et al., 1994). These paths also have larger runout zones, and thus greater impact on the subalpine forest; topographically controlled paths average over 32,000 m2 in area, and lithologically controlled paths average over 41,000 m2. Although more numerous in the east-central GNP area, structurally controlled avalanche paths have less overall impact on treeline, and with an average of only approximately 25,000 m2 much smaller runout zones while descending only to an average elevation of 1,700 m asl. (Walsh et al., 1994). It remains to be seen whether snow avalanche frequency and magnitude will be affected by projected climate changes of the 21st Century. Reduced snowpacks could reduce the spatial extent of snow avalanches and allow the treeline to ascend into the lower reaches and lateral flanks of current avalanche paths, but avalanche meteorology is notoriously difficult to predict under the best of circumstances.
Geomorphic Patterns and Processes at Alpine Treeline
67
Figure 3 Snow avalanches create avalanche boulder tongues that extend well into the alpine treeline ecotone, Preston Park. Path positions are controlled by the presence of a diorite sill (dark band running across wall on either side of right-hand bifurcated snowpack). Debris flows also occur at these sites (Figure 4).
2.2. Debris flows as treeline disturbance agents The net effect of debris flows in the alpine treeline ecotone is similar to that of snow avalanches, that is, to maintain a nonclimatic, geomorphically controlled treeline considerably below the elevation of climatic treeline. Butler and Walsh (1994) and Walsh and Butler (1997) illustrated the widespread nature of debris flows at alpine treeline throughout east-central GNP and illustrated their role in depressing the altitudinal limit of treeline in geomorphically active basins. They recognized four primary types of debris flows: debris flows/torrents occurring within the confines of snow-avalanche paths; debris flows generated by seasonal melting of semipermanent snowpatches; debris flows unassociated with either avalanche paths or snowfields, that is, debris flows primarily triggered by heavy rain onto talus and colluvial debris (Figure 4) (DeChano and Butler, 2001); and debris flows produced by meltwater from glacial moraines. The first three categories are widespread throughout eastern GNP and therefore have major impacts as disturbances impacting alpine treeline, particularly because more than half of their sample of 157 debris flows terminated well below the general alpine treeline ecotone. Data in Butler and Walsh (1994) also illustrated that the importance of debris-flow type varies spatially. In the Cataract Creek valley, for example, where climatic conditions are amenable for an upward treeline advance, avalanche-path debris flows preclude upward movements. Snow-patch debris flows, on the other hand, are especially important along the east-facing wall of Baring Creek valley in depressing treeline well below its climatic optimum. Moraine debris flows are limited in number but may also
68
David R. Butler et al.
Figure 4 Debris flows directly impacting adjacent krummholz patches, Preston Park.
impact treeline position in specific locations such as at the head of Canyon Creek above Cracker Lake. Wilkerson and Schmid (2003) showed that debris flows occur with great frequency in several eastern basins in GNP and used absolute- and relative-dating techniques to determine both absolute and relative ages of debris-flow deposition in areas of alpine treeline (Wilkerson et al., 2002). Wilkerson mapped more than 2,400 debris flows in the Park east of the Continental Divide (Wilkerson, 2004). The greatest density of debris flows occurs in cirques near, but not immediately on, the Continental Divide, with numbers declining toward the eastern margin of the park. This distribution has obvious ramifications for the positioning of treeline near the Continental Divide. During the period 1994–2003, debris flows at treeline sites frequently impacted and destroyed trees attempting to colonize higher altitude sites (Wilkerson, 2004), illustrating the significant geomorphic roles of debris flows in depressing treeline below its climatic optimum throughout basins in eastern GNP.
3. Medium-Scale Processes At a medium, landform-specific scale, some surfaces are inhospitable to seedling establishment and survival. Examples include boulder deposits such as talus deposits, avalanche boulder tongues, and solifluction treads. Avalanche boulder tongues (Figure 3) are periodically swept by snow avalanches, precluding tree seedling establishment. Input from rockfall, as well as a dearth of soil, prevents seedling establishment on active talus deposits, although if rockfall rates and
69
Geomorphic Patterns and Processes at Alpine Treeline
magnitude change under global change scenarios, some talus deposits could stabilize and become amenable for seedlings.
3.1. Turf-banked terraces Past geomorphic processes also dramatically influence current processes and the position of alpine treeline at the medium, landform scale. Across widespread areas of the eastern uplands of Glacier Park, relict turf-banked terrace tread-and-riser topography (Butler and Malanson, 1989, 1999; Butler et al., 2004; Carrara, 1990; Malanson et al., 2002; Walsh et al., 2003a, 2003b) provides sufficiently gentle slopes for soil development in the vicinity of current alpine treeline. The terraces support well-developed soils (Schmid, 2004) and lichen coverage on surface clasts that suggest a late-Pleistocene age for their last major period of activity (Butler and Malanson, 1989; Carrara, 1990). The morphometry of such treads, which we have measured at a variety of sites across the eastern treeline ecotone uplands (Table 1), has been described in some detail in Butler and Malanson (1989), Walsh et al. (2003a), and Butler et al. (2004). From the perspective of tree seedling establishment at alpine treeline, turfbanked terrace risers are difficult sites for a tree seed to penetrate because of the dense covering of Dryas octapetala (Figure 5). Treads are equally difficult sites for seedling establishment because of their exposure to wind and drying. Malanson et al. (2002) revealed that no differences exist in effective soil depth between the terrace treads and the risers (each averaging approximately 15 cm in depth; Figure 6). Butler et al. (2004) showed that drier sites, such as the steeper, welldrained White Calf site, had treads and risers more impenetrable than those at the gentler sloping Divide Mountain site, but penetrability at each individual site did not vary between treads and risers (Table 2). The overall effect on the potential for trees or krummholz patches to affect their neighborhood depends on the scale of the features (Zeng et al., 2007).
Table 1 Average widths of solifluction terrace treads and risers, Lee Ridge, White Calf and Scenic Point sites
Transect name
Treads
Risers
Lee Ridge 1 Lee Ridge 2 Lee Ridge 3 Lee Ridge 4 Lee Ridge 5 White Calf 1 White Calf 2 Scenic Point
93 (44) 94 (34) 95 (46) 69 (35) 66 (31) 59 (11) 82 (45) 77 (40)
99 (49) 114 (42) 99 (39) 99 (54) 84 (35) 85 (82) 119 (93) 60 (31)
Mean and standard deviation (in parentheses) in centimeters. Data from Butler et al. (2004).
70
Figure 5
David R. Butler et al.
Typical widths of relict turf-banked terrace treads and risers, Divide Mountain site.
Figure 6 Soil depth on turf-banked terrace treads and risers, Lee Ridge site. Malanson points to the lower soil horizon transition into regolith below.
71
Geomorphic Patterns and Processes at Alpine Treeline
Table 2 Penetrometer measurements along transects across solifluction treads and risers, Lee Ridge and White Calf sites
Group Lee Ridge site Treads Risers White Calf site Treads Risers
N
Mean
SD
138 138
2.84 2.87
0.668 0.816
45 42
4.11 3.92
0.563 0.830
Means of treads and risers at each site are not statistically different (t-test). Triplicate penetrometer measurements were collected on 138 treads and 138 risers at Lee Ridge and on 45 treads and 42 risers at White Calf. Table reproduced from Butler et al. (2004).
3.2. Eolian processes at treeline In order to quantify the amount of eolian influx under current conditions at and above treeline, sediment traps were installed in 1999 in five krummholz fingers and adjacent tundra sites at Lee Ridge (Figure 7); adjacent to the two United States Geological Survey (USGS)-installed weather stations on Lee Ridge; and within, and adjacent to, two krummholz islands above treeline at Scenic Point. Five additional traps were installed in Preston Park and Baring Basin in 2000. Sediment traps were harvested in the summer of 2003. The positioning of the traps was designed to assess whether sedimentation rates vary at windward versus leeward
Lee Ridge Instrumentation
A
A
B B
Sediment Trap Weather Station Sensor Mast
Figure 7
Sediment trap sampling distribution, Lee Ridge.
72
Figure 8
David R. Butler et al.
Tundra sediment trap downwind of krummholz finger, Scenic Point site.
sites in krummholz fingers and whether krummholz sites collect more sediment than adjacent open tundra sites (Figure 8). Sediment analysis is on-going; in general, however, very few sand- or silt-sized particles were deposited in the traps; duff and litter comprised the primary component of each trap’s collected material, suggesting little eolian sediment influx into treeline soils under current conditions. Two traps at Scenic Point did collect 1-cm pebbles, illustrating the scouring and transport ability of wind in the treeline ecotone in GNP. Additional observations on East Flattop Mountain illustrated that eolian processes are sufficiently robust to move 2þ cm-diameter rocks across the surface of the tundra, where they collected in a rock dune abutting the krummholz. The rock dune (Figure 9) appears to have created a great deal of stress on the krummholz patch. If such eolian rock dunes occur elsewhere in GNP, they would be another addition to the group of mediumscale geomorphic processes that preclude treeline establishment or survival.
Geomorphic Patterns and Processes at Alpine Treeline
73
Figure 9 Wind-blown rocks up to 4^5 cm in diameter bury a tree finger in the treeline ecotone on the top of East Flattop Mountain. Note the dying krummholz. Burial occurred between first visit in 1992 and follow-up in 2001.
4. Fine-Scale Processes and Landforms Fine-scale processes such as frost heaving and churning are inimical to seedling survival, whereas other processes such as turf exfoliation (Rasenabscha¨lung) may facilitate seedling establishment by creating microsites that are less compacted, more penetrable by seeds, and sources of moisture retention. Placement of individual boulders across the local surface is also important, as boulders offer leeward shelter for seedling establishment in sites where the microclimate might otherwise be too harsh for their survival.
4.1. Turf exfoliation Turf exfoliation (Rasenabscha¨lung) is a denudation process active in periglacial areas which destroys a continuous ground vegetation cover by removing the soil exposed along small terrace fronts (Pe´rez, 1992). Processes associated with turf exfoliation include needle ice, desiccation, deflation, collapse of overhanging terrace edges, surface runoff, soil piping, and throughflow. The net effect is creation of microsites differing in soil compression and penetrability compared to unexfoliated risers and adjacent treads. The exfoliated sites are significantly easier for seeds to penetrate than either the adjacent treads or the vegetated risers (see Chapter 6 in this volume on soils of treeline and Butler et al., 2004 for a complete discussion of the methodology and data supporting this statement). We have observed needle ice, collapse of overhanging terrace edges, and evidence of throughflow via soil piping at our sites adjacent to treeline in GNP (e.g., Figure 10 at Divide Mountain). These sites are dynamic, exposing
74
David R. Butler et al.
(a)
(b)
(c)
(d)
Figure 10 Evidence of throughflow and soil piping at turf exfoliation sites, Divide Mountain site. (a) Water entering the tunnel at lower left below the lens cap along this exfoliated turf front exits by the rock to the right of the lens cap. A ‘‘natural bridge’’ of turf and Dryas separates the entry point from the egress point. A second natural bridge, above and to the right of the lenscap, separates the egress tunnel from another needle-ice pan upslope. (b) Close-up of soil piping exit points by the exit rock shown in (a). The natural bridge connecting these two sites was first observed in 2003 and still in existence in September 2005 (c); by 2006, it had collapsed (d).
fine-grained sediment at the base of exfoliating risers where seedlings become established. We have observed, over the course of several seasons and years, how exfoliation exposes fresh sediment and offers microclimatic protection from the wind that sweeps across the tundra. These microsites are one of two primary sites in the tundra where we observe seedling establishment and survival; the other is in the lee of boulders.
4.2. Boulders Seedlings establishing in tundra at three study sites (Canyon Creek, Lee Ridge, and Divide Mountain) in GNP were almost exclusively concentrated either at the base of exfoliated turf risers or in the lee of boulders extending upward above the general tundra surface (see Resler, 2004 and Resler et al., 2005 for the complete data set of 211 conifer patches examined in the field) (Figure 11); in both cases, protection is offered from severe tundra winds and ice-crystal blasting, and fine-grained material
Geomorphic Patterns and Processes at Alpine Treeline
75
Figure 11 Krummholz established in the lee of a pair of large Altyn limestone boulders at the base of Divide Mountain. Prevailing wind direction is from lower right to upper left.
(at the base of exfoliated risers or in the lee eddy behind tundra boulders where loess can be deposited) is available for seedlings to take root. Geographic variability exists among study sites with regards to conifer shelter in eastern GNP (Resler et al., 2005). At Lee Ridge, a small majority (51%) of sheltered establishment sites are associated with exfoliated terrace risers. This finding highlights the importance and availability of terrace risers as a shelter source at this site. Large boulders scatter the landscape among the terrace risers at Lee Ridge and are associated with 41% of establishment sites, with the remaining classified as a combination of boulder/turf exfoliated riser interaction. The large majority of conifer establishment sites (75%) in Cataract Creek Basin are associated with terrace risers, and boulders rank second (16%) for serving as sites for conifer establishment. The majority of conifer establishment sites at Divide Mountain are also associated with exfoliated terrace risers (52%); however, boulders also provide an additional important shelter source (26%). The majority of boulders are found close to their source at the base of Divide Mountain. Although exfoliated terrace risers form exceed boulders as initial establishment microsites, their protection as seedlings grow is less effective. Conifer mortality, defined as establishment sites where 100% of foliar material is dead, was significantly higher (df = 2, 2 = 11.52, p < 0.001) at vegetated exfoliated riser sites than at boulders: 84% of conifer mortality at Lee Ridge, Cataract Creek, and Divide Mountain was located in association with terrace risers, as opposed to 16% near boulders and 0% near combination shelters (Resler et al., 2005). A conceptual model developed by Resler (2006) illustrates the role of microscale shelter, within the constraints of landscape (medium to coarse)-scale processes, in modifying
76
David R. Butler et al.
Microscale pattern and process Landscape-scale pattern and process
Landscape-scale pattern and process Shelter availability:
Macroscale topography
Boulders Terrace risers Combination shelter
Treeline advance New establishment of trees Growth of existing trees
Regional climate Wind Precipitation Average growing season temperature
Local site improvements: Wind protection Shade Soil moisture retention Increased soil depth; penetrability
Solifluction Freeze–thaw Rockfalls
Treeline retreat Age-related mortality Stress-related mortality
Geomorphic slope processes Conifer establishment Conifer mortality Conifer growth
Figure 12 Conceptual diagram illustrating how microscale shelters improve local site conditions that assist in conifer seedling establishment and growth.
microsite conditions and producing sites amenable to tree seedling establishment and survival (Figure 12).
4.3. Needle-ice pans Pe´rez (1992) and Grab (2002) have described needle-ice pans, ovoid to circular depressions 1–6 m in diameter and with a steep scarp 20–40 cm high affected by exfoliation. Both Pe´rez and Grab suggested that the pans may have originated from animal trampling and hoof disruption of the turf surface. We have observed pan depressions at our Divide Mountain and Lee Ridge study sites on low slopes of <10°. Pan-like depressions are particularly well developed at Divide Mountain (Figure 13). We measured the long, cross-slope axis and the shorter, downslope axis for 30 pan depressions at Divide Mountain, including those illustrated in Figure 13. Slope angles ranged from nearly flat (1–3°) at the largest pans on the south-facing, south side of the study site, to gentle (4–8°) at the most well-developed, smaller pans on the north-facing, north portion of the study site. Pans were elongated ovoids, typically ranging from about 60 to 140 cm on the shorter, downslope axis and about 130 to 360 cm on the longer, across-slope axis. The complete pan data set is presented in Table 3. The surface of pans frequently illustrated well-developed frost sorting and development of fine-scale patterned ground (sensu Butler and Malanson, 1989, 1999) (Figure 6). We did observe one animal hoofprint within a pan depression
77
Geomorphic Patterns and Processes at Alpine Treeline
(a)
(b)
Figure 13
Table 3
Large frost pans at the Divide Mountain site, in July (a) and November (b) 2003.
Cross-slope and downslope dimensions of Pan depressions at Divide Mountain
Cross-slope axis (cm) 230 490 320 220 140 352 125 237 315 173 280 272 250 235 508 290 185 235 365 220
Downslope axis (cm) 125 110 195 130 88 140 78 115 272 100 210 195 125 125 375 180 87 120 132 115 (Continued)
78
Table 3
David R. Butler et al.
(Continued )
Mean Standard deviation
Cross-slope axis (cm)
Downslope axis (cm)
95 220 165 175 110 130 90 130 225 205
50 65 90 95 50 55 40 80 90 75
232.9 102.8
123.6 70.8
(sensu Grab, 2002; Pe´rez, 1992) at Divide Mountain. We also observed several forms of animal activity that could be associated with the initiation of turf disruption and subsequent exfoliation: 1. large ungulate hoofprints, sunk 4–8 cm deep below the surrounding surface, were seen along the edge of a tread at Lee Ridge, attesting to the possibility of animals serving as the initial disrupting agent exposing a vegetated riser to subsequent turf exfoliation processes; 2. burrows of ground squirrels were seen in association with turf overhangs, along exfoliated risers and a few pans (Figure 14);
(a)
(b)
Figure 14 Gopher casts (a) and ground squirrel burrow openings (b), Lee Ridge study site. Note the small pine seedling associated with the gopher casts (a) and left of the largest limestone boulder (a). The ground squirrel burrow opening (b) is emplaced within an exfoliated riser front on the upward side of a frost pan. A 49-mm lens cap provides scale in both photos.
Geomorphic Patterns and Processes at Alpine Treeline
79
3. burrowing and surface sediment displacement associated with gophers (Figure 14). Fewer than 1% of the treeline sites studied in detail in the field, however, experienced meaningful animal-induced disruptions. This conclusion calls into question the role of animals as initiators of turf exfoliation. We have, on the other hand, observed widespread frost churning and subsurface needle ice associated with numerous pans at the Divide Mountain site. In general, then, needle-ice pans effectively inhibit tree establishment, particularly where surfaces are strongly disturbed by frost churning (below).
4.4. Frost heaving and churning On the treads of relict turf-banked terraces in eastern GNP, active frost processes churn the upper several centimeters of soil and produce miniature patterned polygons. This frost churning and heaving is especially notable in needle-ice pans on low-angle, turf-bank terrace surfaces at Divide Mountain. Active soil churning is, of course, inimical to the success of tree seedling establishment and survival. Because no previous data documenting the amount and rate of frost heaving and churning at sites adjacent to treeline in GNP exist, we have embarked on a multiyear program designed to document the level of frost activity in needle-ice pans at our Divide Mountain site. In summer 2003, we placed 15 US one-cent pieces (‘‘pennies’’) in the center of miniature patterned ground polygons (sensu Butler and Malanson, 1989, 1999; Butler et al., 2004; Wilkerson, 1995) at three sites at Divide Mountain and two sites at Siyeh Pass (Figure 15), and observations at these sites are ongoing. We
Figure 15 2003.
Positioning of pennies in fine-grained patterned ground centers in the summer of
80
David R. Butler et al.
photographed the position of the pennies in each pan and aligned the pennies so that each coin was placed with the ‘‘tails’’ side facing up. All coins were oriented on the same compass direction. We rephotographed these three sites again in summer, 2004 (Figure 16), and added two additional sites. The results of this initial season of pennies monitoring are shown in Tables 4 and 5. Numerous pennies were flipped and/or rotated by frost heaving, and many pennies are undergoing at least partial burial (Figure 16). The loss of approximately 15% of the pennies in one season suggests that they were completely buried by frost churning. Clast movement within the pans is also clearly visible from year-to-year photographic comparison.
(a)
(b)
Figure 16 Frost churning results, summer of 2004, at same site as shown in Figure 13 (a and b). Pennies have been heaved, rotated, overturned, and partially buried.
Table 4 Pennies recovered from pan depressions, Divide Mountain, July 22, 2003–July 30, 2004
N Pennies
N Recovered
% Recovered
11 12 15 38
73.3 80.0 100 84.4
Pan 1, 15 Pan 2, 15 Pan 3, 15 Total pop 45
Table 5 Summary detail of all 45 pennies
• 38 of 45 pennies recovered • 17 pennies tipped up from horizontal, ranging from 20 to 90° of tilt (45% of 38 total recovered pennies) • 11 pennies overturned with Lincoln head up (29% of 38 total recovered pennies) • At least 13 pennies rotated, ranging from 45 to 180° (34% of 38 total recovered pennies) • 15 pennies partially buried (39% of 38 total recovered pennies) • 36 of recovered pennies illustrated changes (95% of 38 total recovered pennies)
Geomorphic Patterns and Processes at Alpine Treeline
81
Our results, although preliminary, provide a picture of an extremely active frost environment on needle-ice pans in GNP. As long as the processes of frost heaving, churning, and burial continue, it is extremely unlikely that seedling establishment at such sites would occur.
5. Additional Comments on the Possible Role of Animals at Treeline Charismatic megafauna, for example, mountain goats (Oreamnos americanus), bighorn sheep (Ovis canadensis), and grizzly bears (Ursus arctos horribilis) have been widely studied in GNP. Although not at the extensive level seen in European cases where reindeer herds interact with arctic treeline environments (Cairns and Moen, 2004), herbivory from large ungulates is widespread in the GNP alpine environment. The trampling and wallowing effects of mountain goats and bighorn sheep have been known for some time and are well documented, with Chadwick’s work (1983) in GNP ecosystem especially noteworthy. Trampling by goats and sheep is of sufficient environmental impact to produce well-marked animal trails on the landscape (Butler, 1995, 2002; Eaton, 1917) and may yet prove to be significant at treeline in assisting or precluding successful seedling establishment. The excavational activities of grizzly bears, described by Butler (1992) and Tardiff and Stanford (1998), is concentrated in the alpine treeline ecotone and may also prove important in affecting seedbed viability. The role of ground squirrels as a major geomorphic agent at treeline was illustrated in nearby southern Alberta (Gardner et al., 1983; Smith and Gardner, 1985) and may yet also prove to be locally significant as in such cases as illustrated in Figure 14.
6. Conclusions Much remains to be determined concerning the role of geomorphology at the fine scale in the alpine treeline ecotone of GNP, but it is clearly a significant player in determining the success or failure of tree seedling establishment, survival, and growth at the fine scale (Butler et al., 2004; Malanson et al., 2002; Resler, 2006; Resler et al., 2005). At the medium and coarse scale, geomorphic processes clearly inhibit the upward establishment and migration of treeline at elevations where climatic conditions alone would allow tree establishment and survival. The steep, heavily glaciated environment of GNP is a difficult and dynamic environment that proves highly challenging for tree establishment and survival. This situation may change under scenarios of global climatic change (see concluding chapter), but at present geomorphology provides only minor assistance to tree success and gives a large dose of failure. It will be interesting to see if and when this relationship changes and in what directions.
82
David R. Butler et al.
REFERENCES Becwar, M.R., Burke, M.J., 1982. Winter hardiness limitations and physiography of woody timberline flora. In: Li, P.H., Sakai, A. (Eds.), Plant Cold Hardiness and Freezing Stress: Mechanisms and Crop Implications, , Vol. 2. Academic Press, New York, pp. 307–323. Brown, D.G., 1994. Comparison of vegetation-topography relationships at the alpine treeline ecotone. Physical Geography 15, 125–145. Butler, D.R., 1979. Snow avalanche path terrain and vegetation, Glacier National Park, Montana. Arctic and Alpine Research 11, 17–32. Butler, D.R., 1992. The grizzly bear as an erosional agent in mountainous terrain. Zeitschrift fu¨r Geomorphologie 36, 179–189. Butler, D.R., 1995. Zoogeomorphology – Animals as Geomorphic Agents. Cambridge University Press, Cambridge. Butler, D.R., 2001. Geomorphic process-disturbance corridors: A variation on a principle of landscape ecology. Progress in Physical Geography 25, 237–248. Butler, D.R., 2002. Visualizing animal impacts on the landscape: Remote sensing in the physical geography classroom. Geocarto International 17, 67–74. Butler, D.R., Malanson, G.P., 1985. A reconstruction of snow avalanche characteristics in Montana, USA using vegetative indicators. Journal of Glaciology 31, 185–187. Butler, D.R., Malanson, G.P., 1989. Periglacial patterned ground, Waterton-Glacier International Peace Park, Canada and USA. Zeitschrift fu¨r Geomorphologie 33, 43–57. Butler, D.R., Malanson, G.P., 1990. Non-equilibrium geomorphic processes and patterns on avalanche paths in the northern Rocky Mountains, USA. Zeitschrift fu¨r Geomorphologie 34, 257–270. Butler, D.R., Malanson, G.P., 1992. Effects of terrain on excessive travel distance by snow avalanches. Northwest Science 66, 77–85 Butler, DR, Malanson, GP, 1996. A major sediment pulse in a subalpine river caused by debris flows in Montana, USA. Zeitschrift fu¨r Geomorphologie 40, 525–535. Butler, D.R., Malanson, G.P., 1999. Site locations and characteristics of miniature patterned ground, eastern Glacier National Park, Montana, USA. Landform Analysis 2, 45–49. Butler, D.R., Malanson, G.P., Resler, L.M., 2004. Turf-banked terrace treads and risers, turf exfoliation, and possible relationships with advancing treeline. Catena 58, 259–274. Butler, D.R., Malanson, G.P., Walsh, S.J., 1992. Snow-avalanche paths: Conduits from the periglacial-alpine to the subalpine-depositional zone. In: Abrahams, A.D., Dixon, J.C. (Eds.), Periglacial Geomorphology. Wiley, London, pp. 185–202. Butler, D.R., Malanson, G.P., Walsh, S.J., Fagre, D.B., 2007. Influences of geomorphology and geology on alpine treeline in the American West – more important than climatic influences? Physical Geography 28, 434–450. Butler, D.R., Walsh, S.J., 1990. Lithologic, structural, and topographic influences on snow-avalanche path location, eastern Glacier National Park, Montana. Annals of the Association of American Geographers 80, 362–378. Butler, D.R., Walsh, S.J., 1994. Site characteristics of debris flows and their relationship to alpine treeline. Physical Geography 15, 181–199. Cairns, D.M., Moen, J., 2004. Herbivory influences tree lines. Journal of Ecology 92, 1019–1024. Carrara, P.C., 1990. Surficial Geologic Map of Glacier National Park, Montana. US Geological Survey Miscellaneous Investigations Series, Map I-1508-D, Washington, DC. Chadwick, D.H., 1983. A Beast the Color of Winter – The Mountain Goat Observed. Sierra Club Books, San Francisco. DeChano, L.M., Butler, D.R., 2001. Analysis of public perception of debris flow hazard. Disaster Prevention and Management 10, 261–269. Eaton, W.P., 1917. A Rocky Mountain game trail. Harper’s Magazine 136, 111–123. Gardner, J.S., Smith, D.J., Desloges, J.R., 1983. The Dynamic Geomorphology of the Mt. Rae Area: A High Mountain Region in Southwestern Alberta. University of Waterloo Department of Geography Publication Series No. 19. Waterloo, ON.
Geomorphic Patterns and Processes at Alpine Treeline
83
Grab, S.W., 2002. Turf exfoliation in the high Drakensberg, Southern Africa. Geografiska Annaler 84A, 39–50. Holtmeier, F.-K., 2003. Mountain Timberlines: Ecology, Patchiness, and Dynamics. Kluwer, Dordrecht. Malanson, G.P., Butler, D.R., 2002. The Western Cordillera. In: Orme, A. (Ed.), The Physical Geography of North America. Oxford University Press, Oxford, pp. 363–379. Malanson, G.P., Butler, D.R., Cairns, D.M., Welsh, T.E., Resler, L.M., 2002. Variability in an edaphic indicator in alpine tundra. Catena 49, 203–215. Malanson, G.P., Butler, D.R., Fagre, D.B., Walsh, S.J., Tomback, D.F., Daniels, L.D., et al., 2007. Alpine treeline of Western North America: Linking organism-to-landscape dynamics. Physical Geography 28, 378–396. Pe´rez, F.L., 1992. Processes of turf exfoliation (Rasenabscha¨lung) in the high Venezuelan Andes. Zeitschrift fu¨r Geomorphologie 36, 81–106. Post, D.M., Doyle, M.W., Sabo, J.L., Finlay, J.C., 2007. The problem of boundaries in defining ecosystems: A potential landmine for uniting geomorphology and ecology. Geomorphology 89, 111–126 Resler, L.M., 2004. Conifer Establishment Sites on a Periglacial Landscape, Glacier National Park, Montana. Unpublished Doctoral Dissertation, Department of Geography, Texas State UniversitySan Marcos. Resler, L.M., 2006. Geomorphic controls of spatial pattern and process at alpine treeline. Professional Geographer 58, 124–138. Resler, L.M., Butler, D.R., Malanson, G.P., 2005. Topographic shelter and conifer establishment and mortality in an alpine environment, Glacier National Park, Montana. Physical Geography 26, 112–125. Schmid, G.L., 2004. The Role of Soils in Recording Environmental Change at Alpine Treeline. Doctoral Dissertation, Department of Geography, Texas State University-San Marcos. Seastedt, T.R., Bowman, W.D., Caine, T.N., McKnight, D., Townsend, A., Williams, M.W., 2004. The landscape continuum: A model for high-elevation ecosystems. BioScience 54, 111–121. Smith, D.J., Gardner, J.S., 1985. Geomorphic effects of ground squirrels in the Mount Rae area, Canadian Rocky Mountains. Arctic and Alpine Research 17, 205–210. Tardiff, S.E., Stanford, J.A., 1998. Grizzly bear digging: Effects on subalpine meadow plants in relation to mineral nitrogen availability. Ecology 79, 2219–2228. Walsh, S.J., Bian, L., McKnight, S., Brown, D.G., Hammer, E.S., 2003a. Solifluction steps and risers, Lee Ridge, Glacier National Park, Montana, USA: A scale and pattern analysis. Geomorphology 55, 381–398. Walsh, S.J., Butler, D.R., 1997. Morphometric and spatial analyses of debris flows: Components of a natural hazards methodology. Geocarto International 12, 59–70. Walsh, S.J., Butler, D.R., Allen, T.R., Malanson, G.P., 1994. Influence of snow patterns and snow avalanches on the alpine treeline ecotone. Journal of Vegetation Science 5, 657–672. Walsh, S.J., Butler, D.R., Brown, D.G., Bian, L., 1990. Cartographic modeling of snow avalanche path location within Glacier National Park, Montana. Photogrammetric Engineering and Remote Sensing 56, 615–621. Walsh, S.J., Butler, D.R., Malanson, G.P., 1997. An overview of scale, pattern, process relationships in geomorphology: A remote sensing and GIS perspective. Geomorphology 21, 183–205. Walsh, S.J., Butler, D.R., Malanson, G.P., Crews-Meyer, K.A., Messina, J.P., Xiao, N., 2003b. Mapping, modeling, and visualization of the influences of geomorphic processes on the alpine treeline ecotone, Glacier National Park, Montana, USA. Geomorphology 53, 129–145. Walsh, S.J., Weiss, D.J., Butler, D.R., Malanson, G.P., 2004. An assessment of snow avalanche paths and forest dynamics using IKONOS satellite data. Geocarto International 19, 85–93. Wilkerson, F.D., 1995. Rates of heave and surface rotation of periglacial frost boils in the White Mountains, California. Physical Geography 16, 487–502. Wilkerson, F.D., 2004. The Spatial and Temporal Hazard of Debris Flows in Glacier National Park, Montana. Doctoral Dissertation, Department of Geography, Texas State University-San Marcos. Wilkerson, F.D., Schmid, G.L., 2003. Debris flows in Glacier National Park, Montana: Geomorphology and hazards. Geomorphology 55, 317–328.
84
David R. Butler et al.
Wilkerson, F.D., Schmid, G.L., Butler, D.R., 2002. Lichenometric dating of hazardous debris flows, Glacier National Park, Montana. Papers and Proceedings of the Applied Geography Conferences 25, 76–84. Zeng, Y., Malanson, G.P., Butler, D.R., 2007. Geomorphic limits to self organization in alpine forest-tundra ecotone vegetation. Geomorphology 91, 378–392.
C H A P T E R
5
Environmental Controls on Turf-Banked Terraces Daniel J. Weiss, Stephen J. Walsh, Sean A. McKnight, and Evan S. Hammer
Contents 1. Introduction 2. Background 3. Study Site 4. Methods 5. Results 6. Discussion 7. Conclusions References
85 86 90 92 97 102 105 105
1. Introduction Turf-banked terraces are a type of periglacial patterned ground found in alpine environments. These features result from a combination of (a) cryoturbation, the process by which materials are mechanically mixed, sorted, and transported by repeated freezing and thawing (Kessler and Werner, 2003); and (b) solifluction, the slow, gravitational, downslope movement of water-saturated, seasonally thawed materials (Thomas and Goudie, 2000). They may reflect past conditions but affect current hydrology (Christiansen et al., 2008; Raab et al., 2007). In Glacier National Park (GNP), Montana, turf-banked terraces are characterized by alternating stripes of flat, stone-covered treads or steps and angled risers covered by alpine tundra vegetation, primarily Dryas octopetala (Butler and Malanson, 1989). Other forms of turf-banked terraces include lobe-shaped turf-banked terraces that have vegetation growing on both the treads and the risers (Hansen-Bristow and Price, 1985; Hugenholtz and Lewkowicz, 2002). Comparatively, massive solifluction lobes with boulder-covered risers can be found in mountains from South Africa to the Front Range of Colorado (Boelhouwers, 1999). Butler et al. (2007) and Zeng et al. (2007) suggested that they may play an important role in treeline ecotone dynamics. Developments in Earth Surface Processes, Volume 12 ISSN 0928-2025, DOI 10.1016/S0928-2025(08)00205-8
Ó 2009 Elsevier B.V. All rights reserved.
85
86
Daniel J. Weiss et al.
Turf-banked terraces in GNP showed little sign of downslope movement during a 3-year study at Siyeh Pass by Bamberg and Major (1968) in which the motion of the vegetation mats was measured using steel rods driven into the terraces to a depth of approximately 75 cm. This conclusion was affirmed by Butler and Malanson (1989), who described turf-banked terraces at Scenic Point and Piegan Pass as ‘‘inactive under current climatic conditions’’ based on an analysis of lichens growing on large clasts found on the stone treads. These assessments of terrace stability are very reasonable considering that periglacial activity can be limited even in areas much closer to alpine glaciers (Haugland, 2004) than sites currently containing turf-banked terraces in GNP. The processes by which the stone-covered treads remain vegetation-free are uncertain, but needle ice and frost heave may impede vegetation establishment through mechanical disturbance and surface churning that exposes fine sediments to erosion by wind and water (Butler et al., 2004; Douglas and Harrison, 1996). Evidence of surface churning in GNP was reported by Bamberg and Major (1968). Previous research conducted in the Park by members of the Mountain GeoDynamics Research Group has identified a number of environmental conditions coincident with the presence of turf-banked terraces, including elevation, slope angle, slope aspect, and soil moisture potential (Butler et al., 2004; McKnight, 2004; Walsh et al., 2003a, 2003b). These observations led us to hypothesize that characteristics of the turf-banked terraces might also reflect underlying environmental conditions and, therefore, be correlated with geographic information systems (GIS)-based measures of topo-climatic settings and geomorphic conditions. Specifically, we hypothesize that turf-banked terrace size is positively correlated with topo-climatic settings, including steepness, moisture availability, coldness, and exposure to wind, as these conditions generally favor the downslope movement, saturation, and surface churning necessary to produce the terraces. To test these possible relationships, we use statistical analysis to explore the relationships between the spatial characteristics of turf-banked terraces and the topographic settings in which they exist. The composition and spatial structure of solifluction steps and risers were sampled through in situ methods involving the use of 50-m field transects and a run-length measurement approach in which the widths of the alternating stone-covered treads and vegetation-covered risers for 55 transects distributed throughout the eastern side of the Park were measured. Digital elevation models (DEMs) were used to characterize the environmental settings of the field transects. The fundamental intent was to describe the environmental controls on the measured composition and spatial organization of the solifluction steps and risers in GNP, Montana.
2. Background Five previous studies have looked specifically at the environmental settings of turf-banked terraces in GNP, Montana. The first study, conducted by Bamberg and Major (1968), found that variations in soil moisture conditions were the most
Environmental Controls on Turf-Banked Terraces
87
important environmental factor affecting the observed pattern. They suggested that the turf-banked terraces were formed under previously existing conditions as only slight downslope movement of the terraces was occurring at the time of their work, thereby indicating that the sites were no longer active. Further, they noted that turf-banked terraces were most common on east-facing aspects and that exposure to wind (i.e., presence on exposed ridges) was most associated with their geographic distribution. In GNP, exposure to wind is important for the redistribution of snow, which insulates the ground and may diminish the effects of periglacial processes such as solifluction. However, the relationship between exposure to wind and sorted stripes may be more complicated as wind appears to be a control on sorted stripes even in maritime climates dominated by diurnal freeze–thaw cycles (Holness, 2001). Butler and Malanson (1989) identified three major types of patterned ground on the east side of the Park including polygons, turf-banked terraces, and larger terraces dominated by frost shattering. They found that slope and elevation generally distinguished pattern ground types and that turf-banked terraces in GNP are found at middle elevations, approximately 2,200–2,475 m, with slopes sufficiently shallow to facilitate soil moisture retention for solifluction processes to occur. Unlike Bamberg and Major (1968), Butler and Malanson (1989) found turfbanked terraces to be more prevalent on slopes with northeast and northwest aspects. Further, they identified a combination of frost-creep and solifluction processes as responsible for producing the turf-banked terraces within the Park. They also inferred that the turf-banked terraces found in Scenic Point and Siyeh Pass were last active during the Little Ice Age. In contrast, periglacial features found just east of the Park date from the Wisconsin glacial period or earlier (Karlstrom, 1990). The most recent studies on turf-banked terraces in the Park were conducted by Walsh et al. (2003a, 2003b) and McKnight (2004). Walsh et al. (2003b) used spatially explicit data organized within a GIS environment to map, model, and visualize selected direct and indirect geomorphic processes that were hypothesized to influence the composition and spatial organization of the alpine treeline ecotone in GNP. GIScience techniques were used to consider biotic and abiotic controls of alpine treeline with particular emphasis on relict solifluction terraces, evidence of nonlinearity in the development of a catena, and the location of isolated boulders dispersed within alpine tundra as possible nurse rocks for seedling establishment at treeline (Resler et al., 2005). They found that characterizing the alpine treeline ecotone within a threedimensional context showed promise for improving spatial simulation models of alpine treeline and defining landscape attributes within space–time contexts. Solifluction terraces, catena development, and boulders within the alpine tundra were selected for study because they constrain (or advance) subsequent geomorphic and biogeomorphic processes, thus leading to disequilibrium of the alpine treeline ecotone. Walsh et al. (2003a) used semivariograms to examine the spatial organization and the scale-dependent relations of solifluction steps and risers on Lee Ridge, a site located in the northeastern portion of Glacier National Park.
88
Daniel J. Weiss et al.
Datasets for this research included 50-m field transects, GPS coordinates of the transect starting points, and digitally enhanced ADAR-5500 aircraft imagery that was spatially coincident with the field transects. Wavelet transformations of the ADAR-5500 high-spatial resolution digital aircraft data were used to describe the spatial organization of solifluction steps and risers, whereas run-length coding was used to describe the composition and spatial structure of the steps and risers for the field transects. Semivariance was computed by plotting (a) the lagged distance of steps and risers using a 0.1-m spatial resolution to summarize the actual run-lengths derived for the field transects and (b) the lagged distance of the steps and risers using a 1.0-m spatial resolution to summarize the spatial structure of the steps and risers assessed by the ADAR-5500 image data for a sampling range extending from 1 to 20 m (or pixels). Finally, distance from Gable Mountain, the dominant peak upslope of Lee Ridge, was used as a surrogate for elevation in describing the pattern of solifluction steps and risers characterized by the field transects and the remote-sensing imagery. Findings include that the plots of the semivariograms for the field transects at lower elevations had smaller amplitudes and shorter wavelengths than those at higher elevations. Field observations confirm that the terrace surfaces on Lee Ridge are compacted, exhibit some induration, and are supportive of large-diameter Xanthoria elegans lichens, all conditions indicative of the relict nature of the solifluction processes on the ridge. Localized patches of biopedoturbation by ground squirrels and grizzly bears disrupt the surface of some solifluction terraces and expose fine-grained sediments to erosion from runoff. These exposed patches of fine-grained sediment may also serve as potential seedbeds for tree expansion on Lee Ridge. McKnight (2004) examined the composition and spatial structure of 55 turf-banked terrace transects distributed throughout the east side of GNP. These transects also comprise the data set used for the analysis presented in this chapter, and all authors of this chapter participated in the acquisition of these data. Figure 1 graphically summarizes the pattern of steps and risers and the compositional make-up of the field transects sampled in Divide Mountain (n = 19), Lee Ridge (n = 16), Piegan Pass (n = 3), Scenic Point (n = 8), and Siyeh Pass (n = 9). McKnight (2004) used these field transects to generate (a) measures of semivariance (i.e., sill, range, and nugget variance) that were computed to examine scale-dependent relationships and pattern for the onedimensional field transects; (b) ecological pattern metrics that were used to assess the spatial organization and composition of the alternating sequences of steps and risers for the field transects; and (c) cluster analysis that was used to examine typologies of solifluction patterns captured in the 55 field transects. Findings indicate that the study sites were significantly different with respect to their mean stripe widths and variances and that measures of tread and riser characteristics tended to cluster according to study-site elevation. McKnight (2004) suggested the exploration of relationships between turf-banked terrace characteristics and environmental conditions other than elevation as an avenue for future research.
Environmental Controls on Turf-Banked Terraces
89
Figure 1 Spatial organization of solifluction steps and risers on fifty-five 50 -m field transects distributed over five sub-basins within GNP.The transects are displayed from upslope (left) to downslope (right), the black segments represent vegetated risers, the white segments represented nonvegetated treads, and the names indicate the sub-basin and number of the field transect.
90
Daniel J. Weiss et al.
3. Study Site GNP is located in the northwest corner of Montana on the border between the United States and Canada. Established in 1910, GNP is approximately 4,100 km2 in size. The Park is bounded on the west by the Flathead River Valley and on east by the Blackfoot Indian Reservation. The Park is part of the northern Rocky Mountains and sits astride the western Continental Divide. It is dominated by two mountain ranges that trend northwest to southeast through the Park. The Livingston Range is on the western side of the Park, stretching from the Canadian border to Lake McDonald, and the Lewis Range runs through the center of the Park, stretching from the Canadian border to the southern tip of the Park at Marias Pass. GNP varies in relief from 940 m valley bottoms to summits just over 3,000 m in elevation. The landscape was uplifted by the Rocky Mountain orogeny 150 million years ago but owes much of its current character to periods of alpine glaciation during the Pleistocene. The dominant rock types in GNP are relatively soft limestones, shales, and argillites. These brittle sedimentary rocks have experienced intense frost shattering as a result of the harsh climatic conditions, leading to a surface dominated by rock fragments with few boulders compared to mountains with more resistant substrate. The five study sites used for this research are Divide Mountain, Lee Ridge, Piegan Pass, Scenic Point, and Siyeh Pass (Figure 2). McKnight (2004) selected these sites with the intent to sample a variety of topographic aspects, slopes, and elevations. Most topographic aspects are represented in the data set, with the exception of southwest-facing slopes due to a lack of accessible sites with this slope aspect. Slope values in the data set range from 1 to 17°, and the elevation values range from 2,037 to 2,391 m. All five sites are located east of the Continental Divide to minimize climatic differences (e.g., precipitation differences between the east and the west sides of the Park due to orographic precipitation). The research reported here was conducted using the same data set as McKnight (2004) although all the authors participated in data collection. These study sites used for this research correspond to several sites examined by Bamberg and Major (1968) and by Butler and Malanson (1989). The Divide Mountain study site is located just outside the eastern Park boundary in the Blackfoot Indian Reservation. The site is generally east facing, with some southeast-facing slopes. The approximate elevation is 2,140 m. A total of 19 transects were collected at this site, 7 in 2002 and 12 in 2003. The Lee Ridge study site is located near the northern Park boundary. The site is a long, gently sloping ridge that slopes from the south (2,150 m) downhill to the north (1,825 m). The aspect of the site ranges from northwest to northeast. Patterned ground is much more pronounced on the southern (higher) portion of the ridge, with turf-banked terraces occurring mostly perpendicular to the slope (Figures 3 and 4). Sixteen transects were collected at this site, 12 in 2001 and 4 in 2002. The Piegan Pass study site is located near the center of the Park in a well-sheltered glacial cirque. The site is west facing and at an elevation of approximately 2,110 m. Three transects were collected at this site in 2002. The Scenic Point study site is located in the southeastern portion of the Park. It is situated in an exposed, broad, semicircular basin approximately 2,270 m in elevation. The
Environmental Controls on Turf-Banked Terraces
91
Figure 2 Study site locations in GNP. The image background is a Landsat Thematic Mapper image acquired on August 20, 1999.
Figure 3
Turf-banked terraces on Lee Ridge ^ note the GPS receiver for scale.
92
Daniel J. Weiss et al.
Figure 4 Turf-banked terraces on Lee Ridge.
aspect of this study site ranges from southeast to northeast. Scenic point was one of the study sites examined by Butler and Malanson (1989), who described the pattern ground found here as having large areas of nonsorted stripes with some miniature sorted polygons on the treads of turf-banked terraces. A total of eight transects were collected at this site during 2002. The Siyeh Pass site is situated in a high saddle at an elevation of approximately 2,400 m. It is a broad south to southeast-facing area found at the head of a hanging valley, just above a small glacial cirque. The site is home to large expanses of tundra and nonsorted patterned ground. This site was examined by Bamberg and Major (1968) and Butler and Malanson (1989). Six transects were collected here in 2002, while three transects were collected in 2003.
4. Methods Two statistical approaches were followed to examine the relationships between the characteristics of turf-banked terraces and the GIS-based measures of topo-climatic settings and geomorphic conditions. The first approach involved the use of simple linear regression to model relationships between single environmental variables and solifluction characteristics. The second approach used multiple linear regression to model the full set of environmental variables, following their transformation using principal components analysis (PCA). For both approaches, the set of turf-banked terrace characteristics were derived from fifty-five 50-m field transects collected from the five study basins in GNP. These measures of turfbanked terrace characteristics were similar to those used by McKnight (2004). The transects were measured using a run-length approach in which a 50-m tape was stretched perpendicular to the tread-and-riser morphology and the widths of the alternating rock and vegetation stripes were recorded to a precision of 0.1 m. Sample transect locations were selected opportunistically in areas with clearly
93
Environmental Controls on Turf-Banked Terraces
defined turf-banked terrace morphology. Sample transects were collected in sets of 2 or 3 parallel transects with 10–20 m between each transect. The sets of parallel transects were distributed on slopes with no vertical overlap (i.e., transect start points were always placed beyond the end points of the previous transect set). A GPS unit was used to collect UTM coordinates at the start point of each transect, which corresponded to the highest elevation of the transect. A transect bearing was also taken at the start point of each transect using a compass. The GPS points and the transect bearings allowed the transects to be organized as linear features within a GIS environment. To assure the consistency of measurements, we followed a standard protocol for determining the edges of the alternating stone and vegetation stripes, which was reviewed each year prior to collecting field data. The core-dependent variables used in the analyses included the mean (Figure 5) and standard deviation (Figure 6) values for treads, risers, and all stripes from the field transects, calculated on a per-transect basis. Five additional dependent variables were also tested, including the ratio of tread and riser widths and a set of four onedimensional pattern metrics calculated by McKnight (2004). These variables are useful for quantifying the texture and spatial structure of the turf-banked terraces by including width information for both treads and risers into single measures. The pattern metrics selected include Aggregation Index (AI), Contagion (CONTAG), Landscape Division Index (DIVISION), and Splitting Index (SPLIT) (McGarigal and Marks, 1995). The set of pattern metrics was processed using FRAGSTATS (Version 3.3) software (McGarigal and Marks, 1995). The input data for the onedimensional pattern metrics calculation consisted of the transect samples segmented 16
Frequency
14 12
Tread Riser
10
All
8 6 4 2 0 –
–
43 2. – 28 .28 2. 2 – 13 .13 2. 2 – 98 .98 1. 1 – 83 .83 1. 1 – 68 68 1. 1. – 53 .53 1. 1 – 38 .38 1. 1 – 23 .23 1. 1 – 08 08 1. 1. – 93 .93 0. 0 – 78 .78 0. 0 – 63 .63 0. 0
48
0. 48
0.
33
0.
33
–
0.
0
Width (m)
Figure 5
Histogram of the frequency distribution of stripe width of the field transects.
94
Daniel J. Weiss et al.
20 18 16
Frequency
14 Tread Riser All
12 10 8 6 4 2 0
47 2. – 6 36 .3 2. – 2 5 25 .2 2. – 2 4 14 .1 2. – 2 3 03 .0 2. – 2 2 92 .9 1. – 1 1 81 .8 1. – 1 0 70 .7 1. – 1 9 59 .5 1. – 1 8 48 .4 1. – 1 7 37 .3 1. – 1 6 26 .2 1. – 1 5 15 .1 1. – 1 4 04 .0 1. – 1 3 93 .9 0. – 0 2 82 .8 0. – 0 1 71 .7 0. – 0 0 60 .6 0. – 0 9 49 .4 0. – 0 8 38 .3 0. – 0 7 27 .2 0. – 0 16 6 1 0. –
0
0.
Standard Deviation (m)
Figure 6 Histogram of the frequency distribution of stripe-width standard deviation of the field transects.
into 0.1-m intervals. AI and CONTAG range from 0 to 100%. These indices are equal to zero when patches on the landscape are disaggregated and equal to 100 when the landscape consists of a single patch. CONTAG and AI will increase as dispersion decreases, but they differ in that CONTAG also incorporates the inequitable distribution of patch types (McGarigal and Marks, 1995). Effectively this means that if the width increases more for treads than risers or vice-versa, CONTAG will increase more rapidly than AI. DIVISION and SPLIT relate to the subdivision of the landscape into separate patches. The SPLIT ranges from 1 to the number of cells (line segments in the case of this research) on the landscape squared, with a 1 indicating that the landscape consists of a single patch. Higher values of the SPLIT indicate that the landscape is increasingly subdivided into smaller patches. The DIVISION ranges from 0 to 1 with 1 indicating that the landscape consists of a single patch. The SPLIT and DIVISION differ in that SPLIT is more sensitive to a broader range of transect widths. As is often the case with analyses utilizing pattern metrics, the differences between the selected metrics are subtle and the metrics are correlated with each other. Varying relations among such metrics mean that interpretation should be cautious (Wang and Malanson, 2007). All the selected pattern metrics quantify landscape patterns using the patch size and patch-size frequency distribution of landcover classes. For this research, we anticipate that the pattern of turf-banked terrace morphology is related to the hypothesized environmental controls. Specifically, we expect that steeper slopes, slopes with north to northeast aspects, and slopes with high soil moisture potentials experienced more intense periglacial activity and have larger terraces and more regular turf-banked
Environmental Controls on Turf-Banked Terraces
95
terrace morphology as a result. Evidence to support these hypotheses includes wider treads and risers and less subdivided (clumpier) patterns of treads and risers. We expected a more complicated relationship between slope shape (i.e., convexity and concavity) and turf-banked terrace morphology, because concave areas favor the capture and retention of blowing snow in winter. This effect might encourage more intense solifluction through the addition of moisture, but snow cover also reduces diurnal temperature differences that can lead to needle ice activity and the related surface churning that keeps the stone-covered treads vegetation-free. The independent variables used in the analysis were selected to approximate the observations by Bamberg and Major (1968), Butler and Malanson (1989), and Walsh et al. (2003a, 2003b). These variables included slope angle, cosine aspect, sine aspect, transformed aspect, elevation, Topographic Convergence Index (TCI), total curvature, profile curvature, and plan curvature. Aspect was split into three variables to capture the north-south gradient (cosine aspect), the west-east gradient (sine aspect), and the southwest-northeast gradient (transformed aspect). Transformed aspect = –1 cosine (aspect-45) is a simple method of capturing relative temperature differences between the warmest and the coolest slope aspects that result from unequal radiation loading (Beers et al., 1966 as cited in Urban et al., 2000). Transformed aspect is a useful proxy for temperature because southwest-facing slopes in the northern hemisphere are in direct sunlight during the afternoon and typically have higher daytime high temperatures than neighboring slopes with different aspects. Elevation will also capture temperature differences between the sample transects that result from adiabatic cooling, but for samples with similar elevations that are in close proximity to each other, localized temperature will be affected more by the amount and timing of received solar radiation than by elevation. TCI uses a DEM to approximate soil moisture based on the notion that water flowing downhill is transferred from each cell to that cell’s most direct downslope neighbor. Total curvature is the two-dimensional curvature value, while plan and profile curvature relate to one-dimensional horizontal and vertical curvature, respectively. For the three curvature measures, a value of zero indicates a flat surface, positive values indicate a convex surface, negative values indicate a concave surface, and the absolute value of each cell is the relative magnitude of curvature. All independent variables were created from a USGS, 10-m DEM using ArcInfo. Continuous raster surfaces covering the entire Park were generated for each variable, and the values associated with each transect were extracted using the GPS points and the bearings of the field transects. Three transects were deleted from the analysis as they were judged to be statistical outliers during preliminary data exploration. These outlier transects contained treads that experienced vegetation infilling as a result of their position near treeline. In the first stage of the statistical analysis, linear regression models were computed to test the relationships between each turf-banked terrace characteristic (dependent variable) and each of the terrain (independent) variables. The second stage of the analysis tested the relationship between single turf-banked terrace characteristics and the full set of environmental variables. Traditional multiple regression models were not used for this analysis because, as is often the case with environmental data, there was collinearity among the independent variables (Table 1). To overcome this challenge, PCA was used to transform the independent
96
Table 1 Correlation among independent variables associated with the field samples of turf-banked terrace morphology
Slope Slope Cosine aspect Sine aspect Transformed aspect Elevation TCI Total curvature Profile curvature Plan curvature p-value < 0.05.
Sine aspect
Transformed aspect
Elevation
TCI
Total curvature
Profile curvature
Plan curvature
1.000 0.352 0.106 0.323
0.352 1.000 0.262 0.757
0.106 0.262 1.000 0.201
0.323 0.757 0.201 1.000
0.078 0.642 0.175 0.401
0.208 0.061 0.022 0.200
0.107 0.216 0.252 0.176
0.054 0.132 0.210 0.206
0.129 0.212 0.131 0.034
0.078 0.208 0.107
0.642 0.061 0.216
0.175 0.022 0.252
0.401 0.200 0.176
1.000 0.045 0.267
0.045 1.000 0.261
0.267 0.261 1.000
0.202 0.241 0.898
0.182 0.088 0.384
0.054
0.132
0.210
0.206
0.202
0.241
0.898
1.000
0.062
0.129
0.212
0.131
0.034
0.182
0.088
0.384
0.062
1.000 Daniel J. Weiss et al.
Cosine aspect
Environmental Controls on Turf-Banked Terraces
97
variables into a set of principal components that were orthogonal to each other. Once calculated, the principal components were used as the independent variables in the multiple linear regression analysis. The downside of using PCA is that the interpretation of the regression analysis is more challenging because the principal components typically represent combinations of several independent variables. Therefore, correlations between the principal components and the original independent variables were examined to infer links between PCA outputs and terrain characterizations.
5. Results In the first stage of analysis, several of the independent variables were significantly correlated with characteristics of the tread and riser morphology (Table 2), but the relationships were relatively weak (Table 3 and Figure 7). Of the dependent variables that were significantly correlated with topographic variables, mean tread width and the ratio of riser and tread widths showed the strongest relationships. In contrast, mean riser width, stripe-width standard deviation, and DIVISION were not significantly correlated with any of the independent variables. The independent variables that were most frequently and most significantly correlated with characteristics of the turf-banked terraces were elevation, slope angle, topographic convergence index, total curvature, and profile curvature. Mean tread width was and positively correlated with slope angle, elevation, and topographic convergence index and negatively correlated with total curvature and profile curvature. The same set of independent variables was correlated with the ratio of tread and riser widths, but the signs of the correlations were reversed. This relationship indicates that mean tread width and the ratio of riser and tread widths were probably reflecting the same general set of effects. In the second phase of the analysis, the full set of independent variables was reduced to eight principal components that contained all the variability of the original data set. The linear relationships between each turf-banked terrace characteristic and each principal component are shown in Table 4. Of the eight principal components, only three were significantly correlated with any of the turf-banked terrace characteristics. Using the correlation and loading matrices generated as part of the PCA, the principal components are described with respect to the original set of environmental variables. In this case, the first principal component (PC1) is almost completely associated with elevation, the second principal component (PC2) is associated with TCI and slope angle, and the fifth principal component (PC5) is associated with the three aspect measures. The results of the multiple regression models are shown in Table 5. Several of the dependent variables are significantly correlated with the principal components, but most of the variability in any of the dependent variables could not be explained by all the variability in the set of independent variables. Among the models that produced significant p-values, the ratio of riser and tread width shows the strongest result with an R2 of 0.448.
98
Table 2 Correlation coefficients between turf-banked terrace characteristics and environmental variables
Slope
p-value < 0.05. p-value < 0.01.
Cosine aspect
Transformed aspect
Elevation
TCI
Total curvature
Profile curvature
Plan curvature
0.329
0.194
0.187
0.239
0.386
0.397
0.346
0.299
0.157
0.241
0.266
0.13
0.176
0.358
0.166
0.360
0.25
0.291
0.215
0.095
0.105
0.196
0.204
0.042
0.236
0.177
0.164
0.275
0.069
0.24
0.136
0.249
0.262
0.267
0.268
0.043
0.025
0.184
0.002
0.298
0.011
0.162
0.039
0.008
0.072
0.269
0.163
0.188
0.2
0.163
0.257
0.215
0.224
0.017
0.340
0.005
0.086
0.063
0.362
0.299
0.436
0.354
0.247
0.024 0.246 0.246 0.037 0.155
0.26 0.165 0.201 0.26 0.262
0.083 0.041 0.122 0.076 0.111
0.348 0.241 0.251 0.346 0.286
0.129 0.156 0.123 0.114 0.028
0.189 0.11 0.179 0.172 0.096
0.028 0.326 0.202 0.05 0.193
0.01 0.265 0.209 0.03 0.225
0.042 0.183 0.02 0.051 0.033
Daniel J. Weiss et al.
Mean tread width Tread width standard deviation Mean riser width Riser width standard deviation Mean strip width Strip width standard deviation Riser/tread ratio AI CONTAG DIVISION PLADJ SPLIT
Sine aspect
R2-values for significant relationships between turf-banked terrace characteristics and environmental variables
Slope Mean tread width Tread width standard Deviation Mean riser width Riser width standard deviation Mean strip width Strip width standard deviation Riser/tread ratio AI CONTAG DIVISION PLADJ SPLIT
Sine aspect
Cosine aspect
Transformed aspect
0.108
Elevation
TCI
Total curvature
Profile curvature
0.149
0.157
0.119
0.089
0.128
0.129
Plan curvature
0.085
0.075
Environmental Controls on Turf-Banked Terraces
Table 3
0.089 0.115
0.131
0.090
0.190
0.125
0.121 0.106 0.120 0.082
p-value < 0.05. p-value < 0.01.
99
100
1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 0
3
6
9
12
15
18
Mean tread width (m)
Mean tread width (m)
Tread width vs. Slope R 2 = 0.1085
Slope (degree)
Tread width vs. Tropographic convergence index R 2 = 0.1575 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 0 5 10 15 20 TCI Tread width vs. Curvature R 2 = 0.1195
1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 2,000 2,050 2,100 2,150 2,200 2,250 2,300 2,350 2,400 2,450 Elevation (m)
Mean tread width (m)
Mean tread width (m)
Tread width vs. Elevation R 2 = 0.1493 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 –2
–1
0
1
2
Curvature
Daniel J. Weiss et al.
Figure 7 Linear trends between selected turf-banked terrace characteristics and individual environmental variables.
Mean tread width Tread width standard deviation Mean riser width Riser width standard deviation Mean strip width Strip width standard deviation Riser/tread ratio AI CONTAG DIVISION PLADJ SPLIT
PC1
PC2
0.149 0.129
0.183
0.090
0.131
0.089 0.119
PC3
PC4
PC5
0.098 0.132 0.086 0.153
PC6
PC7
PC8
Environmental Controls on Turf-Banked Terraces
Table 4 R2-values for significant relationships between turf-banked terrace characteristics and principal components
0.119 0.154 0.178
p-value < 0.05. p-value < 0.01.
101
102
Daniel J. Weiss et al.
Table 5 p-values and R2-values for the multiple regression models using the full set of principal components as the independent variables
Mean tread width Tread width standard deviation Mean riser width Riser width standard deviation Mean strip width Strip width standard deviation Riser/tread ratio AI CONTAG DIVISION PLADJ SPLIT
p-value
R2-value
0.001 0.019 0.148 0.009 0.521 0.019 0.000 0.381 0.037 0.056 0.383 0.087
0.425 0.302 0.331 0.303 0.448 0.275
p-value < 0.05. p-value < 0.01.
6. Discussion We found that tread width is significantly correlated with slope angle, elevation, TCI, total curvature, and profile curvature. These relationships are also reflected in the ratio of riser to tread width. Although causal relationships cannot be inferred with only regression analysis, this result does support the hypothesis that steep, moist slopes found at high elevations may have experienced more intense solifluction and, therefore, possess larger turf-banked terrace treads than those found in lower, drier, more gently sloping areas. Tread width is negatively correlated with total curvature and profile curvature, indicating that treads are wider in areas with locally concave topography. This finding did not support our hypothesis that exposure to wind, and the consequential snow removal, would increase tread width through more intense, freeze–thaw-related surface churning related to reduced snow insulation. A possible explanation for this finding is that the additional snow deposited in concave areas was less important as insulation that hindered surface churning than the moisture from snow melt was for its contribution to the surface saturation required for solifluction. Our hypothesis that colder slopes would have experienced more intense solifluction, and therefore retain larger terraces, was not supported by significant correlations with the measures of slope aspect. The significant, negative relationship between elevation and tread width, however, suggests that temperature may still be an important control on turfbanked terrace characteristics. In contrast to tread width, riser width was not significantly correlated with any of the hypothesized environmental controls. Possible explanations for this finding
Environmental Controls on Turf-Banked Terraces
103
include (a) that other environmental variables not included in this analysis may be controlling riser width; (b) that vegetation infilling or covering of treads may be a more common occurrence than we anticipated; (c) that natural stochastic variability of the periglacial processes that produced the turf-banked terraces is more evident in riser widths than tread widths, effectively masking the environmental controls of the riser widths; or (d) the attribute of turf-banked risers that responds most strongly to environmental controls is riser angle rather than riser width. With respect to the variability of turf-banked terraces, elevation and curvature are significantly correlated with the standard deviation of the tread widths, indicating that treads become less organized in higher areas and in concave areas. This relationship might reflect the importance of lower temperatures and decreased growing season length due to higher elevation and greater snow accumulation. These factors may impede the establishment of tundra vegetation found on the risers and lead to narrower, irregular, or even bare risers. Riser width standard deviation has a significant, albeit weak, negative relationship with slope angle. This finding may reflect greater irregularity in turf-banked terrace morphology at low slope angles where solifluction was less consistent and less pronounced. Pattern metric results showed few significant correlations with the environmental variables. One exception was transformed aspect, which has a significant, positive correlation with aggregation (AI) and a significant, negative correlation with SPLIT. These results indicate that slopes with aspects closer to southwest tend to have turf-banked terrace treads and risers that are generally wider and less fragmented. This relationship was also suggested by the positive correlation between transformed aspect and mean strip width (i.e., the mean width of both treads and risers). Overall, these results do not support our hypothesis that turfbanked terraces will be larger (i.e., have more aggregated, less subdivided treads and risers) on colder, northeast-facing slopes. It is difficult to draw conclusions from these results because the widths of treads and risers do not respond individually to any measures of slope aspect. The weak relationships between the pattern metrics and the full set of environmental variables in the multiple regression analysis also suggest that the one-dimensional patterns of turf-banked terraces cannot be explained by the set of environmental variables derived for this research, given the limitations of the available data sets. The most significant of these limitations is that the environmental variables, as produced from a 10 m DEM, are at too coarse of a scale to capture microtopographic features that may influence turf-banked terrace pattern. While many of the regression models proved to be statistically significant, only about 45% of the variance in any turf-banked terrace characteristic was explained by the full set of hypothesized environmental controls. Several explanations for this finding include (a) the small sample size of only 55 field transects that may not have provided sufficient variability of environmental conditions and statistical power to extract important relationships and to test for multilevel controls on solifluction patterns, (b) the relatively coarse-grained DEM used to produce the environmental variables that had obvious limitations for representing key morphological processes that affect these landforms, (c) the limited set of environmental variables that were used to describe geomorphic processes related to tread and riser morphology that
104
Daniel J. Weiss et al.
may not have included the variables most influential on turf-banked terrace formation, and (d) important measures for providing insight into controls on turf-banked terraces such as riser angle and the extent of vegetation infilling were not part of the field data collection. Environmental variables likely to be important for turf-banked terrace formation that were not part of our data set include soil and substrate measurements such as those used by Hugenholtz and Lewkowicz (2002), but soils data related to turf-banked terraces in GNP are currently limited to a set of soil penetrability measures, for which no nonrandom patterns were found (Malanson et al., 2002). Another challenge inherent to working with periglacial features is that they are polygenetic in nature (e.g., the same form of the patterned ground can arise from different processes, and similar processes can produce different forms of patterned ground) (French 1996; Washburn, 1980). This view is supported for our data set by McKnight (2004) who found that turf-banked terraces at the five study sites were significantly different with respect to tread, riser, and stripe mean width. We next compared our findings with those generated by comparable analyses conducted on similar features. For instance, Douglas and Harrison (1996) conducted the most closely related study when they examined environmental constraints on turf-banked terraces in southeast Iceland. They found that mean tread width decreased as slope increased; tread and riser morphology broke down on slopes above 19°; the effect of slope aspect was unclear, because most of their study sites were on south- or east-facing slopes and exposed to wind; and low slope locations did not allow sufficient downslope movement to disrupt the vegetation mat. Unlike Douglas and Harrison (1996), our research did not explore the slope limits of turf-banked terrace formation, but none of our sample turf-banked terraces are found on slopes steeper than 19°, and previous research has suggested that a lessorganized form of periglacial patterned ground described as ‘‘spotted tundra’’ occurs on steeper slopes in the Park (Butler and Malanson, 1989). Unlike Douglas and Harrison (1996), however, we did find and sample turf-banked terraces on very low slope areas (eight transects with slopes between 1 and 5°). With respect to slope aspect, we had a more complete sample in our data set but did not find strong statistical relationships between slope aspect and turf-banked terrace characteristics. Unlike Douglas and Harrison (1996), we found that mean tread width increased slightly as slope increased, indicating treads become more frequent rather than wider to account for the increased horizontal distance they must span at lower slope angles. Several possibilities exist for this difference. First, the processes responsible for producing and maintaining turf-banked terraces might differ between southeastern Iceland and northern Montana. Evidence for this was found in the mean tread widths reported by both studies. Douglas and Harrison (1996) found mean tread widths ranging from just under 1 to over 5 m wide. In contrast, the mean tread widths in our data ranged from 0.32 to 1.4 m. A second possibility is that the studies gathered fundamentally different field measurements. Our data originated as ground control information for a remote-sensing application dealing with the form and function of the alpine treeline ecotone. In this capacity, the data had to accurately reflect the widths of the stripes and, therefore, considered vegetation that had encroached onto the flat treads as part of the riser, particularly at lower elevation sites with presumably milder climatic conditions and less needle ice
Environmental Controls on Turf-Banked Terraces
105
activity to prevent vegetation establishment on the stone-covered treads. In contrast, Douglas and Harrison (1996) were explicitly measuring the geomorphic characteristics of the turf-banked terraces. If we assume that our findings and the findings by Douglas and Harrison (1996) are comparable, it seems that the turf-banked terraces can respond differently to slope angle and aspect conditions. In fact, our findings with respect to tread width are more similar to those from research on steep, nonvegetated slopes in the Bolivian Andes conducted by Francou et al. (2001). They found that (a) sorted stripes were nonsignificantly correlated with ‘‘other factors, such as slope aspect and elevation,’’ (b) stone-covered stripes (flat) were relatively uniform in width, and (c) stripes covered by fine materials (angled) varied in width and were positively correlated with slope angle. Conceptually, this would mean that at low-slope angles, more stone-covered stripes, rather than wider ones, account for the increased horizontal distance the terraces span relative to the change in elevation. Although the environment in which Francou et al. (2001) worked is dramatically different than the environment of GNP, they identify many of the same periglacial processes as responsible for producing the striped pattern ground that we have described.
7. Conclusions In GNP, we found several statistically significant relationships between characteristics of periglacial patterned ground and environmental variables. Our results indicated that the stone-covered treads widen as slope, elevation, topographic moisture index, and the degree of slope concavity increase. These results support our hypothesis that characteristics of turf-banked terraces may be correlated with environmental conditions likely to influence solifluction. In contrast, we did not find strong relationships between measures of the spatial pattern of turf-banked terraces and hypothesized environmental controls that may indicate that turf-banked terrace patterns result from unaccounted-for environmental variables and/or influences from microtopographic features that were below our scale of analysis. Future research on the environmental controls of turf-banked terraces would benefit from using a larger sample size, incorporating additional environmental variables that may influence turfbanked terrace characteristics (i.e., soil and substrate conditions), including more field measurements such as riser angle and the extent of vegetation infilling on treads, and through the use of higher-spatial-resolution DEMs (e.g., LIDAR) that would allow environmental controls to be tested at finer scales of analysis.
REFERENCES Bamberg, S.A., Major, J., 1968. Ecology of the vegetation and soils associated with calcareous parent materials in three alpine regions in Montana. Ecological Monographs 38, 127–167. Beers, T.W, Press, P.E., Wensel, L.C., 1966. Aspect transformation in site productivity research. Journal of Forestry 64, 691–692.
106
Daniel J. Weiss et al.
Boelhouwers, J.C., 1999. Relict periglacial slope deposits in the Hex River Mountains, South Africa: Observations and paleoenvironmental implications. Geomorphology 30, 245–258. Butler, D.R., Malanson, G.P., 1989. Periglacial patterned ground, Waterton-Glacier International Peace Park, Canada and USA. Zeitschrift fur Geomorphologie 33, 43–57. Butler, D.R., Malanson, G.P., Resler, L.M., 2004. Turf-banked terrace treads and risers, turf exfoliation, and possible relationships with advancing treeline. Catena 58, 259–274. Butler, D.R., Malanson, G.P., Walsh, S.J., Fagre, D.B., 2007. Influences of geomorphology and geology on alpine treeline in the American West – more important than climatic influences? Physical Geography 28, 434–450. Christiansen, H.H., Sigsaard, C., Humlum, O., Rasch, M., Hansen, B.U., 2008. Permafrost and periglacial geomorphology at Zackenberg. Advances in Ecological Research 40, 151–174. ¨ raefi, southwest Iceland: Morphometry, Douglas, T.D., Harrison, S., 1996. Turn-banked terraces in O rates of movement, and environmental controls. Arctic and Alpine Research 28, 228–236. Francou, B., Le Mahaute, N., Jomelli, V., 2001. Factors controlling spacing distances of sorted stripes in a low-latitude alpine environment. Permafrost and Periglacial Processes 12, 367–377. French, H.M., 1996. The Periglacial Environment. Longman, Essex. Hansen-Bristow, K.J., Price, L.W., 1985. Turf-banked terraces in the Olympic Mountains, Washington, USA. Arctic and Alpine Research 17, 261–270. Haugland, J.E., 2004. Formation of patterned ground and fine-scale soil development within two late Holocene glacial chronosequences: Jotunheimen, Norway. Geomorphology 61, 287–301. Holness, S.D., 2001. The orientation of sorted stripes in the maritime Subantarctic, Marion Island. Earth Surface Processes and Landforms 26, 77–89. Hugenholtz, C.H., Lewkowicz, A.G., 2002. Morphometry and environmental characteristics of Turf-Banked Solifluction Lobes, Kluane Range, Yukon Territory, Canada. Permafrost and Periglacial Processes 13, 301–313. Karlstrom, E.T., 1990. Relict periglacial features east of Waterton-Glacier Parks, Alberta and Montana and their paleoclimatic significance. Permafrost and Periglacial Processes 1, 221–234. Kessler, M.A., Werner, B.T., 2003. Self-organization of sorted patterned ground. Science 299, 380–383. Malanson, G.P., Butler, D.R., Cairns, D.M., Welsh, T.E., Resler, L.M., 2002. Variability in an edaphic indicator in alpine tundra. Catena 49, 203–215. McGarigal, K., Marks, B.J., 1995. FRAGSTATS: Spatial analysis pattern program for quantifying landscape structure. USDA Forest Service General Technical Report PNW-351. McKnight, S.A., 2004. Composition and Spatial Structure of Turf-Banked Terraces in Glacier National Park, Montana, USA. M.A. Thesis, University of North Carolina at Chapel Hill. Raab, T., Leopold, M., Volkel, J., 2007. Character, age, and ecological significance of Pleistocene slope deposits in Germany. Physical Geography 28, 451–473. Resler, L.M., Butler, D.R., Malanson, G.P., 2005. Topographic shelter and conifer establishment and mortality in an alpine environment, Glacier National Park, Montana. Physical Geography 25, 112–125. Thomas, D.S.G., Goudie, A., 2000. The Dictionary of Physical Geography. Blackwell, Oxford, UK. Urban, D.L., Miller, C., Halpin, P.N., Stephenson, N.L., 2000. Forest gradient response in Sierran landscapes, the physical template. Landscape Ecology 15, 603–620. Walsh, S.J., Bian, L., McKnight, S., Brown, D.G., Hammer, E.S., 2003a. Solifluction steps and risers, Lee Ridge, Glacier National Park, Montana, USA: A scale and pattern analysis. Geomorphology 55, 381–398. Walsh, S.J., Butler, D.R., Malanson, G.P., Crews-Meyer, K.A., Messina, J.P., Xiao, N., 2003b. Mapping, modeling, and visualization of the influence of geomorphic processes on the alpine treeline ecotone, Glacier National Park, MT, USA. Geomorphology 53, 129–145. Wang, Q., Malanson, G.P., 2007. Patterns of correlation among landscape metrics. Physical Geography 28, 170–182. Washburn, A.L., 1980. Geocryology: A Survey of Periglacial Processes Environments. John Wiley, New York. Zeng, Y., Malanson, G.P., Butler, D.R., 2007. Geomorphic limits to self organization in alpine forest-tundra ecotone vegetation. Geomorphology 91, 378–392.
C H A P T E R
6
Soils and Pedogenesis at Alpine Treeline Ginger L. Schmid, David R. Butler, George P. Malanson, and Lynn M. Resler
Contents 1. Introduction 1.1. Background 2. Field data collection 2.1. Soil pit data at Lee Ridge and White Calf Mountain 2.2. Soil penetrability and compression on Turf-banked terrace treads and risers 3. Results 3.1. Soil pit data from tree fingers and adjacent tundra 3.2. Effective soil depth and soil compaction data 4. Discussion and Conclusions References
107 107 109 109 111 114 114 115 117 117
1. Introduction Alpine treeline soils in Glacier National Park (GNP), Montana, have traditionally been considered to be shallow, Lithic Cryochrepts in the Ptarmigan Series, with minimal horizon development (Dutton and Marrett, 1997). These soils are thin but with well-developed 02 horizons and stony to very stony sandy loam to sand A, B, and C horizons (Butler and Malanson, 1989; Nimlos and McConnell, 1962). We posited the following hypotheses in examining soils in the alpine foresttundra ecotone (AFTE) of GNP: 1. Do tundra soils in the AFTE differ from soils formed underneath tree fingers? 2. In widespread areas of relict solifluction treads and risers in the AFTE, do tread soils differ substantially from vegetated riser soils? 3. Do AFTE soils illustrate characteristics of long-term soil development when developed on geomorphically stable, isolated slopes?
Developments in Earth Surface Processes, Volume 12 ISSN 0928-2025, DOI 10.1016/S0928-2025(08)00206-X
Ó 2009 Elsevier B.V. All rights reserved.
107
108
Ginger L. Schmid et al.
The overarching research question behind these hypotheses is why conifer seedlings are able to become established in certain locations above current alpine treeline, but not in others, that is, the ‘‘sieve effect’’ mentioned in the chapter on Ecotone Dynamics. Our observations and data were collected primarily at the Lee Ridge, White Calf Mountain, Divide Mountain, and Scenic Point study sites (Figure 1), with additional observations from the head of Cataract Creek. Data reported and summarized here include previously unpublished observations from Schmid (2004), as well as from published reports (Butler and Malanson, 1989; Butler et al., 2004; Malanson et al., 2002; Resler, 2006; Resler et al., 2005).
1.1. Background Burns and Tonkin (1982) provided a mode of alpine soil development within which the setting of soils within eastern GNP near treeline can be examined. They described the typical soils characterizing seven types of alpine microenvironments, ranging from dry windblown sites to wet meadows and semipermanent snowpacks. Topographic site conditions are the most important pedogenic
114° 35′W 49° 00′N
Montana
1
Glacier St. Mary
National 2
Park
3
4
West Glacier
East Glacier Park
N
0 48° 00′N
10 20 Kilometers 113° 00′W
Figure 1 The four primary study sites for AFTE soils in eastern GNP. 1, Lee Ridge site; 2, Divide Mountain site; 3,White Calf Mountain site; and 4, Scenic Point site. Cataract Creek site mentioned in text is approximately located 10 km east of the ‘‘a’’ in the word ‘‘National’’ in center of park.
109
Soils and Pedogenesis at Alpine Treeline
factor because they control both snow cover and loess distribution, which in turn influence the rate of soil development and its tendency toward a ‘‘climax state.’’ Malanson and Butler (1994) suggested that tundra soils near treeline in eastern GNP may actually be sufficiently more fertile than the adjacent forested soil such that alpine tundra plants outcompete trees along the ecotone, extending tundra vegetation to lower-than-expected elevations and precluding the upward advance of treeline. However, in a more spatially detailed study, Cairns (1998) reported a positive effect of krummholz on soil nutrients. These studies expand on work done on Niwot Ridge at treeline (Parker and Sanford, 1999; Seastedt and Adams, 2001; Withington and Sanford, 2007; see also Weiss et al., 2005) and in alpine tundra (Bowman et al., 1993; Fisk et al., 1998; Holtmeier and Broll, 1992; Litaor et al., 2002).
2. Field Data Collection 2.1. Soil pit data at Lee Ridge and White Calf Mountain At both Lee Ridge and White Calf Mountain, Schmid (2004) excavated 50-cm deep soil pits in four tree fingers that approximately parallel the topographic contours across slope and in the adjacent alpine tundra. Characteristics of the two sites are illustrated in Table 1. Seven soil pits per finger, plus two control pits, were excavated at the study sites for a total of 60 soil pits. At each finger, the seven soil pits were excavated in the pattern shown in Figure 2, with three pits in the finger in windward, middle, and leeward locations, and four pits in the surrounding tundra in upwind, downwind, upslope, and downslope positions. Each pit was sampled to depths of 50 cm. In each soil pit, soil profiles were sampled and for each horizon present, the following variables were collected: horizon depth, boundary characteristics, dry and moist Munsell colors, texture, percent coarse fragments by volume, structure, consistence, pH, HCl reaction, roots, pores, and presence of clay films.
Table 1 Site characteristics
Elevation Slope angle Slope aspect Prevailing wind Parent material Dominant lithology
Lee Ridge
White Calf
2,000–2,150 m 7° (1–15) W–N (273–358) WSW Till/colluvium Altyn formation (limestones and dolomites)
2,000–2,200 m 22° (14–32) ESE (90–145) SSW Till/colluvium Altyn formation (limestones and dolomites)
110
Ginger L. Schmid et al.
Upslope X Leeward X
X
X
Tundra Tree finger
X
Windward X
X Downslope
Soil pit locations
Figure 2
Soil pit sampling scheme for an individual tree finger.
All soil horizons were sampled and described according to standard soil survey methods (Soil Survey Division Staff, 1993; Soil Survey Staff, 1975; USDA SCS, 1974). A Profile Development Index (PDI) (Birkeland, 1999; Harden and Taylor, 1983) was subsequently calculated for each soil pit, using the most relevant fieldmeasured variables (Table 2). The actual number of soil properties used in calculating a PDI can be tailored to the environment under study, and thus our variables focus strictly on the characteristics present or most relevant at our study sites. Details of the PDI development may be found in Schmid (2004).
Table 2 Soil pit variables used for calculating PDI values at Lee Ridge and White Calf sites
PDI property (Birkeland, 1999; Harden and Taylor, 1983) Rubification
Color paling
Reason property not used for Lee Ridge and White Calf
Reason property used for Lee Ridge and White Calf
• Few obvious Bt horizons present • May be well developed, so below 50 cm excavation • May be masked due to periglacial mixing • May not have formed yet • Parent material chroma not readily established • Not recommended nor practical with light-colored parent material (Continued)
111
Soils and Pedogenesis at Alpine Treeline
Table 2
(Continued )
PDI property (Birkeland, 1999; Harden and Taylor, 1983)
Reason property not used for Lee Ridge and White Calf
• Color darkening due to organic enrichment common in all pits
Melanization
Color lightening
Total texture
Dry consistence Moist consistence
• Not recommended nor practical with light-colored parent material • Texture of original parent material difficult to predict, especially wet consistence • Value for parent material difficult to predict • Value for parent material difficult to predict • Good distribution present in most pits
Clay films Structure
• Minimal structure observed in majority of pits • Potentially limited by periglacial activity • Readily observable changes from pH 8 of parent material
pH
Carbonate morphology
Mottling Clast weathering
Reason property used for Lee Ridge and White Calf
• Morphology calculation requires use of color paling and color lightening (see above) • Not present • Highly variable within pits
2.2. Soil penetrability and compression on Turf-banked terrace treads and risers We examined turf-banked terrace treads and risers at Scenic Point, Divide Mountain, Cataract Creek, and Lee Ridge. Morphometry of these terraces, as well as an additional data set collected on Lee Ridge by Walsh et al. (2003), has been described in Chapter 5. Treads and risers are especially well developed at Lee Ridge (Figures 3 and 4) and Divide Mountain.
112
Ginger L. Schmid et al.
Figure 3 Tree fingers extend out into the turf-banked terrace tread and riser topography of the Lee Ridge study site.
(a)
(b)
Figure 4 Summer (a) and winter (b) views of the Lee Ridge study site.Winter view courtesy of the US Geological Survey’s Glacier Field Station.
Soils and Pedogenesis at Alpine Treeline
113
At Lee Ridge and Scenic Point, we measured effective soil depth (ESD) using a steel rod pushed into tundra soil until resistance from underlying clasts was encountered. The details of the methodology have been described in Malanson et al. (2002); here we note that several hundred penetration-depth measurements were collected on both turf-banked terrace treads and on risers. At Lee Ridge and White Calf Mountain, we also measured soil compaction using a hand-held ELE International steel pocket penetrometer. We collected triplicate measurements on 138 treads and risers at Lee Ridge and 45 tread–riser sets at White Calf (see Butler et al., 2004, for a more detailed description of the methodology). We also collected compaction data at 30 turf-exfoliated sites (as described in the previous chapter) and adjacent terrace treads at Lee Ridge, and at 15 sites at Divide Mountain. We examined soil characteristics underlying turf-banked terrace treads and risers along an incised bank at the eastern edge of the Lee Ridge site (Figure 5). For all data on (a)
(b)
Figure 5 Incised edge of Lee Ridge site (a) reveals little apparent difference in soil characteristics beneath turf-banked terrace treads (unvegetated) and risers (vegetated) (b).
114
Ginger L. Schmid et al.
ESD and compaction, means were calculated and compared for statistical differences using t-tests.
3. Results 3.1. Soil pit data from tree fingers and adjacent tundra Soil pit data from Lee Ridge and White Calf revealed no differences in the soils developed in tundra and the soils developed in the adjacent tree fingers although research by Malanson and Butler (1994) documented that vegetation in tree fingers (25 AFTE samples, all from farther south in GNP) has been present long enough to produce differences in soil nutrients (0–10 cm) compared to the tundra soils. No difference existed in soils developed at different slope locations, that is, no difference existed between leeward, windward, upslope, or downslope locations within the tundra, nor within tree fingers, nor in comparing soils of tree fingers with soils of adjacent tundra. All 60 soil pits (7 pits 4 fingers 2 sites, plus two control pits at each site) illustrated well-developed soil profiles and PDI values, with no significant differences in depth of A horizons (Table 3) or thickness of O horizons (Table 4). No significant erosion or excessive colluvial accumulations were identified, indicative of relatively stable, long-term horizon development. No C horizons were encountered within the 50-cm excavation depth in any pit. Organic enrichment (dark, near-mollic soil colors) was common throughout the majority of profiles sampled. Upper horizon pH (A and B1/Bt horizons) were more acidic than the pH 8 of parent material (Table 5). Carbonate accumulations were only present in lower horizons at the White Calf site and were absent from the majority of profiles on Lee Ridge. Table 3 Mean depth of A horizon (cm) for Lee Ridge (LR) and White Calf (WC) sites in tundra (t-) and finger (f-) pits as shown in Figure 2
Site
Location t-wind
t-up
t-lee
t-down
f-wind
f-center
f-lee
LR-1 LR-2 LR-3 LR-4 All
15 15 13 13 14
13 12 10 10 11.3
6 10 12.5 6 8.6
14 15 10 10 12.3
12 8 20 6.5 11.6
8 15 12.5 11 11.6
8 10 10 11 9.8
WC-1 WC-2 WC-3 WC-4 All
20 16 8 6 12.5
7 8 10 10 8.8
14 14 6 11 11.3
14 8 7 12 10.3
8 10 10 7.5 8.9
12 5 6 10 8.3
9 8 9 12.5 9.6
115
Soils and Pedogenesis at Alpine Treeline
Table 4 Mean depth of O horizon (cm) for Lee Ridge (LR) and White Calf (WC) sites in tundra (t-) and finger (f-) pits as shown in Figure 2
Site
Location t-wind
t-up
t-lee
t-down
f-wind
f-center
f-lee
LR-1 LR-2 LR-3 LR-4 Average
7 7 7 8 7.3
7.25 7 7.5 7.5 7.3
7.25 6.5 7 7 6.9
7 7 7.5 7.5 7.3
7 7 7.5 6.5 7
6.5 7.5 7 7 7
7 7 8 7.5 7.4
WC-1 WC-2 WC-3 WC-4 Average
7.5 7 7 7.5 7.3
7.5 7 7 7.5 7.3
7.5 8 7 7.5 7.5
8 7.5 6.5 7.75 7.4
6.5 7.5 6.5 6.5 6.8
7 7.75 6 7.5 7.1
7 7.75 5.5 7.5 6.9
Table 5 2001
pH of A horizon for Lee Ridge (LR) and White Calf (WC) sites, Glacier National Park,
pH A-Horizon Tundra pits (1,2,3,4) Wind
Up
Lee
Down
(Finger)
7 7 7 8 7.3 7.5 7 7 7.5 7.3
7.25 7 7.5 7.5 7.3 7.5 7 7 7.5 7.3
7.25 6.5 7 7 6.9 7.5 8 7 7.5 7.5
7 7 7.5 7.5 7.3 8 7.5 6.5 7.75 7.4
LR-1 LR-2 LR-3 LR-4 Average WC-1 WC-2 WC-3 WC-4 Average
Finger pits (5,6,7) Wind Center 7 7 7.5 6.5 7.0 6.5 7.5 6.5 6.5 6.8
6.5 7.5 7 7 7.0 7 7.75 6 7.5 7.1
Lee 7 7 8 7.5 7.4 7 7.75 5.5 7.5 6.9
3.2. Effective soil depth and soil compaction data Schmid’s (2004) data illustrated no discernible differences between soils in tundra and soils under tree fingers in the AFTE. Soil penetrability and soil compaction data similarly revealed no distinct differences between soils developed on tundra treads and tundra risers. A t-test revealed no significant
116
Table 6
Ginger L. Schmid et al.
Effective soil depth (cm)
40-m grids
Site 1 Site 2 Site 3 Site 4 Site 5 Site 6 Site 7 Site 8 Site 9 Site 10 Site 11 Site 12 All data
4-m grids
Mean
SD
CV
Mean
SD
CV
15.54 16.72 10.64 39.70 10.84 12.36 11.18 21.61 18.16 6.52 26.64 18.32 17.36
6.69 12.30 7.88 27.91 5.01 7.58 9.25 7.61 10.02 4.45 12.00 11.49 14.41
0.43 0.74 0.74 0.70 0.46 0.61 0.83 0.35 0.55 0.68 0.45 0.63 0.83
22.44 11.16 21.72 11.20 8.08 9.62 22.52 10.84 5.54 22.44 45.36 17.36
10.95 4.97 15.80 6.05 5.24 7.18 13.35 8.21 5.64 11.12 25.33 15.92
0.49 0.45 0.73 0.54 0.65 0.75 0.59 0.76 1.00 0.50 0.56 0.92
Statistics on effective soil depth (ESD) (N = 25 per grid)
differences between ESD means (in cm) for treads and risers (Table 6) (15.91 and 15.27, respectively; Malanson et al., 2002). Soil compaction data (Table 7, from Butler et al., 2004) for turf-banked terrace treads and risers also revealed no significant differences (t-test) in mean compaction at either the Lee Ridge nor the White Calf sites, and no differences between sites. The only statistically significant (t-test) differences in compaction data occur between tundra treads and adjacent turf-exfoliated sites (Table 8) such as those described in Chapter 4, with exfoliated sites significantly less compacted and thus, more easily penetrable than the adjacent tread surfaces.
Table 7 Penetrometer measurement along transcects across solifluction treads and risers
Group Lee Ridge Treads Risers White Calf Treads Risers
N
Mean
SD
138 138
2.84 2.87
0.668 0.816
45 42
4.11 3.92
0.563 0.830
117
Soils and Pedogenesis at Alpine Treeline
Table 8 Penetrometer measurements on turf exfoliation sediment and adjacent treads
Group Lee Ridge Exfoliation Tread Divide Mountain Exfoliation Tread
N
Mean
SD
30 30
1.0 2.3
0.5 0.9
15 15
1.03 1.82
0.68 0.75
4. Discussion and Conclusions No differences in long-term soil development in tundra versus tree finger soils indicate that both primary soil pit sample sites (Lee Ridge and White Calf) have been geomorphically stable for extended periods of time, largely because these gently sloping uplands were ice-free during the later advances of Pleistocene glaciation. The results at these AFTE sites suggest that long-term soil development is not significantly altered by climate change associated with glacial pulses in adjacent drainages. The spatial pattern of advance of trees and krummholz into tundra at the AFTE is not controlled by ESD, differences in soil compaction between tundra treads and risers, nor differences in tundra and tree finger soils. As was shown in other chapters in this volume, topographic shelter and turf exfoliation provide sites where seedlings become established in alpine tundra. Soil compaction at turf-exfoliated sites was the only soil-related variable illustrating any significant differences in the AFTE as related to site suitability for seedling establishment. Soils at the AFTE may be significantly older and possess better-developed profiles than previously believed, at least in site locations such as at Lee Ridge and White Calf Mountain (and probably at Scenic Point as well), where late-Pleistocene glaciation did not directly impact upon broad, gently sloping surfaces at the AFTE.
REFERENCES Birkeland, P.W., 1999. Soils and Geomorphology. Oxford University Press, New York. Bowman, W.D., Theodose, T.A., Schardt, J.C., Conant, R.T., 1993. Constraints of nutrient availability on primary production in two alpine tundra communities. Ecology 74, 2085–2097. Burns, S.F., Tonkin, P.J., 1982. Soil-geomorphic models and the spatial distribution and development of alpine soils. In: Thorn, C.E. (Ed.), Space and Time in Geomorphology. Allen and Unwin, London, pp. 25–43. Butler, D.R., Malanson, G.P., 1989. Periglacial patterned ground, Waterton-Glacier International Peace Park, Canada and USA. Zeitschrift f u¨r Geomorphologie 33, 43–57. Butler, D.R., Malanson, G.P., Resler, L.M., 2004. Turf-banked terrace treads and risers, turf exfoliation and possible relationships with advancing treeline. Catena 58, 259–274.
118
Ginger L. Schmid et al.
Cairns, D.M., 1998. Modeling controls on pattern at alpine treeline. Geographical and Environmental Modelling 2, 43–64. Dutton, B.L., Marrett, D.J., 1997. Soils of Glacier National Park East of the Continental Divide. Land and Water Consulting, Missoula, MT. Fisk, M.C., Schmidt, S.K., Seastedt, T.R., 1998. Topographic patterns of above- and belowground production and nitrogen cycling in alpine tundra. Ecology 79, 2253–2266. Harden, J.W., Taylor, E.M., 1983. A quantitative comparison of soil development in four climatic regions. Quaternary Research 20, 342–359. Holtmeier, F-K., Broll, G., 1992. The influence of tree islands and microtopography on pedoecological conditions in the forest alpine tundra ecotone on Niwot Ridge, Colorado Front Range, USA. Arctic and Alpine Research 24, 216–228. Litaor, M.I., Seastedt, T.R., Walker, D.A., 2002. Spatial analysis of selected soil attributes across an alpine topographic/snow gradient. Landscape Ecology 17, 71–85. Malanson, G.P., Butler, D.R., 1994. Tree-tundra competitive hierarchies, soil fertility gradients, and treeline elevation in Glacier National Park, Montana. Physical Geography 15, 166–180. Malanson, G.P., Butler, D.R., Cairns, D.M., Welsh, T.E., Resler, L.M., 2002. Variability in an edaphic indicator in alpine tundra. Catena 49, 203–215. Nimlos, T.J., McConnell, R.C., 1962. The morphology of alpine soils in Montana. Northwest Science 36, 99–112. Parker, E.R., Sandford, R.L., 1999. The effects of mobile tree islands on soil phosphorus concentrations and distribution in an alpine tundra ecosystem an Niwot Ridge, Colorado Front Range, USA. Arctic, Antarctic, and Alpine Research 31, 16–20. Resler, L.M., 2006. Geomorphic controls of spatial pattern and process at alpine treeline. Professional Geographer 58, 124–138. Resler, L.M., Butler, D.R., Malanson, G.P., 2005. Topographic shelter and conifer establishment and mortality in an alpine environment, Glacier National Park, Montana. Physical Geography 26, 112–125. Schmid, G.L., 2004. The Role of Soils in Recording Environmental Change at Alpine Treeline in Glacier National Park, Montana. Doctoral dissertation, Department of Geography, Texas State University-San Marcos. Seastedt, T.R., Adams, G.A., 2001. Effects of mobile tree islands on alpine tundra soils. Ecology 82, 8–17. Soil Survey Division Staff, 1993. Soil Survey Manual. Handbook No. 19, US Department of Agriculture, Washington, DC. Soil Survey Staff, 1975. Soil Taxonomy: A Basic System of Soil Classification for Making and Interpreting Soil Surveys. Handbook 436, USDA Soil Conservation Service, Washington, DC. USDA Soil Conservation Service, 1974. Definitions and Abbreviations for Soil Descriptions. USDS SCS West Technical Service Center, Portland, OR. Walsh, S.J., Bian, L., McKnight, S., Brown, D.G., Hammer, E.S., 2003. Solifluction steps and risers, Lee Ridge, Glacier National Park, Montana, USA: A scale and pattern analysis. Geomorphology 55, 381–398. Weiss, L., Shiels, A.B., Walker, L.R., 2005. Soil impacts of bristlecone pine (Pinus longaeva) tree islands on alpine tundra, Charleston Peak, Nevada. Western North American Naturalist 65, 536–540. Withington, C.L., Sanford, R.L., 2007. Decomposition rates of buried substrates increase with altitude in the forest-alpine tundra ecotone. Soil Biology and Biochemistry 39, 68–75.
C H A P T E R
7
Canopy Structure in the Krummholz and Patch Forest Zones Evan S. Hammer and Stephen J. Walsh
Contents 1. Introduction 2. Background and Context 2.1. Alpine treeline 2.2. Leaf area index 3. Study Area 3.1. Lee Ridge 3.2. Apikuni Cirque 3.3. Cataract Creek 3.4. Preston Park 3.5. Baring Basin 3.6. Scenic Point 3.7. East Flattop Mountain 4. Data and Methods 4.1. Field data collection 4.2. Digital elevation model and terrain analysis 4.3. Vector data sets 4.4. Data processing 5. Analysis 5.1. 3D visualization 5.2. Spatial analysis 5.3. Biophysical analysis 5.4. Topo-climatic variables 6. Results and Discussion 6.1. Internal canopy structure 6.2. Between patch structure 7. Conclusions References
Developments in Earth Surface Processes, Volume 12 ISSN 0928-2025, DOI 10.1016/S0928-2025(08)00207-1
120 121 121 123 124 125 126 126 127 127 127 127 128 128 128 130 130 132 133 133 139 143 143 144 145 146 147
Ó 2009 Elsevier B.V. All rights reserved.
119
120
Evan S. Hammer and Stephen J. Walsh
1. Introduction Alpine treeline is an ecotone or transition zone between two ecosystems (Holland et al., 1991) — the lower elevation subalpine forests and the higher elevation treeless tundra communities. The alpine treeline ecotone (ATE) begins where the subalpine forest transitions from closed to open canopy forest and extends upslope to timberline, the maximum elevation of trees on the slope. The vegetation in this ecotone consists of open forests at lower elevations. As elevation increases and the habitat turns more harsh, the trees become shorter in stature, more distinctly wind shaped, and the continuity of the open forest gives way to tree clumps, patches, or islands. These gnarled patches of trees, deformed by a combination of severe environmental constraints such as ice and wind, are known as krummholz. Continued movement upslope or into more harsh conditions is accompanied by a general decrease in the density of the krummholz patches, a general reduction in stature, and an increase in the dominance of tundra vegetation, as well as nonvegetated rock, snow, and ice surfaces. While several studies have observed different types of krummholz patches on the basis of vegetation height and organization (Baker and Weisberg, 1997; Benedict, 1984; Hadley and Smith, 1987; Marr, 1977; Wardle, 1968), few have attempted to characterize the structure of different krummholz patch types through direct measurement. Occurrences of flagged and mat krummholz forms are well documented (Cox, 1933; Griggs, 1938; Hadley and Smith, 1986, 1987; Holtmeier, 1981; Johnson and Billings, 1962; Pontecorvo and Bokhari, 1975; Wardle, 1968). Some studies suggest a continuous gradation between these distinct forms, while others imply that most patch structures can be described by these two extremes and one or two intermediate growth forms. Vegetation that persists in the alpine environment is influenced by endogenous (e.g., terrain settings) and exogenous (e.g., climate) factors that combine in complex ways to determine the position, extent, composition, pattern, development, form, and spatial expression of the ATE. The contextual effects of site and situation further mediate pattern–process relations at the ATE through biotic (e.g., species) and abiotic (e.g., distances from the Continental Divide) factors and disturbances that are scale dependent. Many studies in Glacier National Park (GNP), Montana, USA, conducted by the Mountain GeoDynamics Research Group, explore pattern and process relationships affecting the ATE. In general, these studies have (a) modeled the spatial patterns of treeline vegetation with respect to topo-climatic and disturbance processes (Alftine and Malanson, 2004; Brown, 1992, 1994; Walsh et al., 1990, 1994; Zeng and Malanson, 2006; Zeng et al., 2007); (b) examined the influence of scale on the spatial organization of treeline communities (Bian and Walsh, 1993; Walsh et al., 1997); (c) explored the spatial patterns of treeline vegetation at the scale of the hillslope unit (Allen, 1995; Allen and Walsh, 1996); (d) examined the influence of the spatial patterns of snow accumulation and snow melt on treeline vegetation (Allen, 1995; Allen and Walsh, 1993; Brown, 1992, 1994; Walsh et al., 1994); and (e) assessed pattern–process relations at the alpine treeline ecotone (Bowersox and
Canopy Structure in the Krummholz and Patch Forest Zones
121
Brown, 2001; Brown, 2001; Cairns, 2001; Cairns and Waldron, 2003; Malanson et al., 2001; McGregor, 1998; Walsh et al., 2003a, 2003b, 2003c). Using these prior themes and initiatives as context, the basic intent of this research was to (a) explore the spatial structure of the ATE in GNP, (b) associate pattern and process relationships in the study of ATE, particularly in krummholz island vegetation communities, through a set of terrain and pattern descriptors, and (c) determine whether these pattern–process relationships are generalizable across several ATE sites that are located on the east side of the Continental Divide of GNP or whether topographic complexity precludes explanation of pattern using descriptors of geographic site and situation. To accomplish these research objectives, field data were collected to examine the internal structure of krummholz patches. Data were collected at regular intervals along transects that extended across the major and minor axes of each selected patch. For each sample point, patch width, patch height, species, and the Leaf Area Index (LAI) were collected using a LI-COR LAI-2000 plant canopy analyzer. The LAI-2000 uses a fisheye optical device to measure light interception that is used to calculate light transmission through the canopy as a measure of canopy gap fraction used to derive LAI (Gower and Norman, 1991). All data collected in the field were located using GPS coordinates so that terrain descriptors derived within a GIS environment could be associated through spatial and statistical approaches. Additional structural variables for the selected krummholz patches were generated by deriving 3D representations of the patches: diffuse noninterceptance (DIFN), foliage density (FD), and drip-line LAI (DLLAI) were calculated. The goal of this project was to (a) examine the internal structure of krummholz patches within GNP, (b) assess the spatial pattern of key physiological descriptors of krummholz patches for a number of topographically varied basins positioned on the east side of the Continental Divide, and (c) use field- and lab-based approaches for characterizing krummholz patches and graphically and statistically analyzing them through data visualization methods and quantitative approaches that integrated topographic and positional descriptions of the sampled krummholz patches.
2. Background and Context 2.1. Alpine treeline Arno and Hammerly (1984) describe the internal structure of the ATE through the various vegetation limits that are found across this landscape: the upper limit of the closed canopy forest or the ‘‘forest line,’’ the upper limit of upright trees, or the ‘‘treeline,’’ and the upper limit of krummholz, and the ‘‘krummholz limit’’ of stunted and distorted, island vegetation. Baker et al. (1995) use these defined limits to divide the ATE into four zones: closed forest zone, patch forest zone, krummholz zone, and the tundra zone. It is important to note that all the limits described by Arno and Hammerly (1984) may not be distinct entities on the landscape. In some locations, usually where ATE is stretched into a gradual transition, these limits may be readily distinguished. At other sites, with more abrupt treelines, these limits
122
Evan S. Hammer and Stephen J. Walsh
may occur in such close proximity that they coalesce into a single forest edge (Arno and Hammerly, 1984; Baker et al., 1995). In such cases, the zones described by Baker and Weisberg (1997) often are not distinguishable. The study sites chosen for this project are all gradual transitions and, in most cases, each of these ‘‘limits’’ and ‘‘zones’’ are distinguishable. The majority of the krummholz patches examined in the study occur in the krummholz zone, while a few are located in the patch forest zone. As trees reach their altitudinal limit at the ATE, the form of the trees becomes deformed due to the environmental stresses encountered in this extreme setting. Krummholz takes a number of forms: symmetrical tree islands, flagged krummholz, and krummholz mats (Hansen-Bristow and Ives, 1985; Wardle, 1968). These are somewhat idealized groupings, because the gradation between upright trees and krummholz mats is often so gradual that it is difficult to classify many of the krummholz forms within an area into a single class. Additionally, while these forms are ordered, moving upslope from forest to tundra, this is not a strict altitudinal sequence. Small variations in terrain as well as differing growth forms between tree species may result in different krummholz forms in very close proximity, along the same contour, or at the same distance from the closed canopy forest. In addition to the general krummholz forms, one specific type of krummholz encountered in this study is the tree island (Figure 1). While this term is sometimes used to reference any krummholz mat form (Daubenmire, 1943), the more common use describes a krummholz patch in an extremely windy treeline site, which may also be referred to as a mobile tree island (Benedict, 1984; Brooks et al., 1996; Holtmeier and Broll, 1992; Ko¨rner, 1999; Marr, 1977; Parker and Sanford, 1999; Pauker and Seastedt, 1996). The high winds and sparse shelter often result in long, narrow patches of krummholz whose major axes extend from windward to leeward orientations. In an active mobile tree island, the windward end of the patch, or the
Figure 1 A typical krummholz growth form found in GNP, Montana,USA.
Canopy Structure in the Krummholz and Patch Forest Zones
123
patch origin, is battered by wind and slowly dies away. At the same time, the origin of the patch provides some shelter to the area downwind that may be sufficient to permit the growth of upright trees in the latter reaches of the patch, with implications for snow accumulation. It is in this area that the plant is able to extend itself. The process of extension and die-off may result in the slow migration of tree islands across an otherwise tundra-covered site (Marr, 1977). It is important to note that while the speed of the movement may be tied to climate conditions at the site, as long as the wind remains the dominant control on the islands, they will continue to move parallel to the wind. The prevailing winds may or may not move in the same direction as the forest-alpine treeline transition. A number of studies have recognized and examined tree islands specifically, and many others have noted the presence of these unique features more generally. An early study by Daubenmire (1943) noted the presence of ‘‘island-like clumps’’ of krummholz. Also noted was the importance of the process of layering, which is a form of asexual reproduction that is critical to the maintenance of the tree island (Wardle, 1968). While these initial studies serve as a general foundation, providing the basic tree island structure, processes of movement, and typical species content, a new wave of research has moved beyond simple description. Recent examinations of tree islands have shifted the focus to the impact that forest patches have on their surroundings by altering site conditions and through generating positive feedback mechanisms for seedling establishment and growth (Alftine and Malanson, 2004; Alftine et al., 2003; Bekker, 2005). Recent tree island literature has focused on how the islands influence soil development. Holtmeier (1992) emphasizes the impact of tree islands on microtopography and pedogenesis.
2.2. Leaf area index For this research, a LAI-2000 plant canopy analyzer, manufactured by LI-COR, was used to make rapid calculations of LAI across krummholz patches. LAI is a measurement of the density of the leaves or needles in a plant canopy derived by calculating the total leaf surface area per unit ground area (Cairns and Malanson, 1998; Franklin et al., 1997; Mackay and Band, 1997; White et al., 1998). The LAI2000 uses a hemispherical sensor to measure diffuse short-wave radiation. The sensor’s field of view is divided into five angular bands: 0–13°, 16–28°, 32–43°, 47–58°, and 61–74°. A complete measurement using the device involves taking a reference measurement above the canopy of interest, followed by one or more below the canopy. The below canopy readings are then normalized by the above canopy reading to derive canopy gap fraction that is converted to LAI (Welles and Norman, 1991). Because LAI is calculated from the average gap fraction across the five zenith angles (as well as being averaged across the number of below canopy measurements), an error term, standard error of LAI (SEL), is also calculated for each LAI measurement. Through the use of the LAI-2000 support software, additional information may be extracted for each sample. LAI is typically calculated in forests that have canopies of relatively continuous height. This is considered in the calculation of LAI from canopy gap fraction by assuming that light reaching the outer viewing rings travel
124
Evan S. Hammer and Stephen J. Walsh
further than light reaching the center rings. In the case of krummholz patches, however, path length to each of the view rings on the optical sensor is highly sensitive to canopy geometry. For this reason, LAI for the sampled patches, defined as DLLAI, was calculated by providing detailed canopy geometry information to the LAI-2000 postprocessing. A recalculation of the LAI equations with path lengths from the sensor rings to the canopy edge produces the FD (the foliage area per unit canopy volume) at a sample point. DLLAI is calculated: canopy volume DLLAI = FD ð1Þ drip line area
3. Study Area The general study site for this research is the east side of GNP (Figure 2). GNP lies on the eastern edge of the northern Rocky Mountain Province – the eastern boundary of GNP separates the gently rolling hills of the Great Plains from the dramatic topography of the Rocky Mountains. Geologically, the Lewis Overthrust fault divides GNP and contributes to the topographic structures found in the Park, as well as the diversity of rock types. Topographically, GNP straddles the Continental Divide and represents the eastern-most establishment of ecological
Lee Ridge Apikuni Cirque Cataract Creek East Flattop Preston Park Baring Basin Scenic Point
Glacier National Park Park boundary Study site boundaries
Continental Divide Streams N
Lakes 0
Figure 2
5
10
20
30
40
50 Km
Study sites in GNP, Montana, USA.
Canopy Structure in the Krummholz and Patch Forest Zones
125
communities found in the Pacific northwest and the western-most extent of ecological communities of the Great Plains. Specific field sites were selected east of the Continental Divide that represent a variety of ATE settings. Two mountain ranges exist in the central portion of GNP, the Lewis Range in the east and the Livingstone Range in the west. Both ranges run from the northwest to southeast direction with the Livingstone Range terminating about midway into the Park and the Lewis Range running the full length of the east side of GNP. The Continental Divide follows the Lewis Range until it reaches the northernmost portion of GNP where it shifts to the Livingstone Range. These ranges were uplifted during the Laramide orogeny, in a process which is also responsible for the massive displacement of rocks that form the Lewis Overthrust fault (Carrara, 1989). Snow avalanches and debris flows are among the most conspicuous mass movements within the Park that serve to shape the ATE through disturbance regimes (Butler, 1986; Butler and Walsh, 1990; Walsh and Butler, 1989, 1997; Walsh et al., 1990; Wilkerson and Schmid, 2003). Field data for this research were collected from seven field sites. One site, Lee Ridge, was an intensive study site where a detailed sampling routine was undertaken as part of a research program funded by the US Geological Survey to the Mountain GeoDynamics Research Group. The remaining sites, Apikuni Cirque, Baring Basin, Cataract Creek, East Flattop Mountain, Preston Park, and Scenic Point, employed a less intensive sampling scheme (Figure 2) that emphasized diversity of form and setting.
3.1. Lee Ridge Lee Ridge is asymmetrical extending northeast from Gable Mountain with a steep cliff dropping into the Lee Creek drainage on the east side and a gentle slope down to the Belly River on the west side (Figure 3). As a result of this orientation, the
Figure 3
Lee Ridge, GNP, Montana, USA.
126
Evan S. Hammer and Stephen J. Walsh
majority of the surface area of the ridge lies in the Belly River drainage to the west. The remaining portion of the ridge drains into Lee Creek to the east. The lower portions of the study area lie in the upper subalpine forest that gives way to alpine tundra. The north end of Lee Ridge, where the elevation is lowest, consists of a subalpine forest of primarily Lodgepole pine (Pinus contorta) interspersed with Subalpine Fir (Picea abies), 5-needle pine (Pinus flexilis and Pinus albicaulis), and Engelmann spruce (Picea engelmannii). In contrast, across most of the higher areas of the ridge, as it approaches Gable Mountain, the trees are arranged in a finger or island pattern. At the lower end, where the forest edge and the islands have grown together, the fingers are found reaching from the subalpine forest into tundra. At higher elevations, the islands are more clearly separated from the forest. Between these vegetation islands are various types of tundra communities including Dryas terraces and moist meadow communities. Solifluction lobes and terraces and the sequential steps and risers are a conspicuous feature (McKnight, 2004; Walsh et al., 2003a, 2003b, 2003c).
3.2. Apikuni Cirque Apikuni Cirque is also in the northeast portion of GNP. It is located just north of Lake Sherburne, southwest of the Lee Ridge site. This study site is located just below the upper cirque in a valley surrounded by Mount Henkel, Mount Altyn, and Apikuni Mountain. Apikuni Falls is the general terminus of the study site to the southeast, while the western terminus is a steep headwall. Wind appears to be a significant factor shaping treeline at this site. In general, the entire area is protected on three sides by steep valley walls. The only opening is above the falls to the southeast, and the area around this opening is exposed to very heavy winds. Wind patterns throughout the remainder of the valley are not as obvious. Vegetation is sparse throughout the study site with the dense vegetation clustered near the headwall to the west. The vegetation, primarily krummholz, becomes sparser downstream along the creek. Away from the creek, vegetation becomes sparser yet with increasing elevation.
3.3. Cataract Creek Southwest of Apikuni Cirque, the Cataract Creek study site is located at the headwall of Cataract Creek valley. This study site is surrounded on three sides by steep valley walls with Mount Siyeh to the southeast, Cataract Mountain to the south, and the Bishop’s Cap to the west. Unlike Apikuni, however, which is bounded by a narrow outlet, downstream of this site is a wide U-shaped valley. Wind is a major factor at the Cataract Creek study site. The prevailing winds are down-valley, originating around Mount Siyeh to the east-southeast and running north and northwest. Vegetation at this site is predominantly in the form of krummholz patches interspersed with tundra communities. Like Lee Ridge, solifluction steps and risers arranged generally normal to the slope on lobes and terraces are predominant in numerous settings. The krummholz patches at this site have the distinction of being the largest patches of all the sites examined.
Canopy Structure in the Krummholz and Patch Forest Zones
127
3.4. Preston Park Preston Park is immediately southeast of Cataract Creek, between Mount Siyeh to the north and Matahpi Peak to the south. The eastern boundary of the site is a steep headwall where Siyeh Creek originates. While much of Preston Park is exposed to strong winds that flow through the valley from the pass to the east, this study site exhibits some degree of protection by virtue of its location against the headwall. The eastern portion of the study site includes a narrow corridor that funnels winds. At this location, there is evidence that the vegetation has been wind trained. Down-valley, to the east of this study site, is a more open forest and meadow community. Closer to the headwall there are sparse krummholz patches and tundra communities.
3.5. Baring Basin Baring Basin is in the east-central portion of GNP, immediately south-southeast of Preston Park, separated by Siyeh Pass. The headwall leading up to Siyeh Pass forms the northern border of the Baring Basin study site. To the west of the Baring Basin site is Going-to-the-Sun Mountain, and to the east is Goat Mountain. Baring Basin opens into the St. Mary Valley to the south. Winds are extreme in this north-south trending basin. Krummholz flags suggest that the prevailing winds are down-slope. The vegetation along the river is dense, but upslope from the river the vegetation quickly turns to krummholz patches interspersed with tundra. The tundra at this location is primarily alpine meadow, and the dominance of grasses in these meadows distinguishes Baring Basin from the other sites.
3.6. Scenic Point Scenic Point is located in the extreme southeast corner of GNP, in a large topographic bowl-shaped formation. Bison Mountain forms the southern boundary of the study site, Scenic Point is to the north, and an unnamed peak rises in the west. This basin represents the headwaters of Fortymile Creek, and water flows east out of the study site. In spite of the peaks surrounding this site, it is very exposed. It is likely that both wind and solar insolation have a strong influence on the vegetation found at this site. Scrubby forests can be seen in the lower portions of the site, extending upslope on the south side of the creeks.
3.7. East Flattop Mountain East Flattop Mountain is located on the east-side of GNP. The middle portion (north to south) of the eastern boundary of the Park lies on the eastern portion of East Flattop Mountain. The study site is in a bowl-shaped feature on the east flank of East Flattop Mountain that overlooks the town of St. Mary. The east slopes of East Flattop Mountain are characterized by a diverse combination of treeline communities. In some locations, treeline is extremely abrupt with closed canopy forest giving way to tundra with little or no transition. Other locations have forests
128
Evan S. Hammer and Stephen J. Walsh
composed of trees that appear as overgrown flag-mat krummholz with upright symmetrical crowns growing from a layered base. Downslope is a spruce-fir forest that gives way to a patch forest, then to the krummholz zone that gives way to open tundra.
4. Data and Methods The data used in this project came from two primary sources. First, field data collection was used to locate krummholz patches and to gather detailed patch composition, geometry, and structural data using a plant canopy analyzer, and GPS units to acquire locational coordinates. Second, a number of geospatial, digital data products were acquired and variables derived to gather information about the position, topography, and biophysical setting of the sampled krummholz patches. A more detailed explanation of the data collection process follows.
4.1. Field data collection At the Lee Ridge site, the most intensively sampled site, 28 krummholz patches were identified for the study. A GPS was used to determine the location of the patches so they would align with other data layers organized in a GIS database. The major and minor axes of each sampled patch were identified and the length was recorded. A sampling interval was determined based on the patch length. The goal was to have at least five sample points along the major axis. Most patches were sampled at every 3 m along the major axis, but some of the smaller patches were sampled at every meter, and a few large patches were sampled at every 5 m. At each sample point, the species, canopy height, canopy width, and LAI were recorded. LAI samples were collected using the LICOR LAI-2000 plant canopy analyzer. A view cap was placed on the device to restrict the field of view to 30° to prevent the inclusion of the operator in the spectral response pattern and to decrease the inclusion of direct sunlight as it has the potential to distort measurements (LI-COR, 1990). The LAI-2000 was positioned so that it could view the 30° segment of the sample cone that faced upslope.
4.2. Digital elevation model and terrain analysis Elevation data were extracted from the 10-m digital elevation model (DEM) product produced by the USGS. In addition to the standard terrain descriptors such as elevation, slope angle, and slope aspect, several variables were derived through the topographic partitioning of watersheds from the DEM by first generating a flow direction surface, derived by passing a focal filter or kernel over each cell of the DEM. The direction of steepest descent is defined, and water flow direction is assigned (Moore et al., 1991). Once flow direction has been generated, flow accumulation may be calculated. Flow accumulation simply measures the number of cells whose flow reaches the
Canopy Structure in the Krummholz and Patch Forest Zones
129
target cell or the upslope contributing area of each cell. By assigning a threshold value, a drainage network was derived from the flow accumulation surface that was similar to the stream network derived by the US Geological Survey (the 1:24,000 scale topographic map series) but with greater uniformity due to a less subjective process of determining features. Finally, watershed boundaries were derived that identified the ridgelines separating drainages (Moore et al., 1991). The potential soil moisture at each 10-m cell of the base DEM for the study area was determined using the Topographic Wetness Index (TWI) (Beven and Kirkby, 1979) that considers the size of the upslope contributing area and the slope of neighboring cells. TWI is calculated as TWI = ln ð2Þ tan where is the contributing area and is the slope. Several indices have been developed and used in this analysis to measure the amount of exposure at a site based on different assumptions. The Landform Index (LFI), Terrain Shape Index (TSI), and Topographic Ruggedness Index (TRI) are all general measures of exposure that consider different scales of importance. LFI is intended to measure the topographic setting of a site in a regional context. Low values of LFI indicate that a site is on an exposed ridge, while high values of LFI indicate that the site is in a protected valley (McNab, 1993). TSI is a similar measure, but it calculates the local topographic setting of a site. Low values of TSI indicate a location on a ridge with steep side slopes, while high TSI values signify a site in a ravine (McNab, 1989). TRI measures the variability of the landscape on a cell by cell basis. Low values of TRI indicate a level surface (suggesting a more exposed site), while high values indicate that there are large changes in elevation between neighboring cells (suggesting a more protected site) (Riley et al., 1999). More specific exposure indices focus on the importance of topographic position in determining direct beam solar radiation flux, snow accumulation, and exposure to wind. Brown (1992) developed a program called SOLARPOT to model potential direct solar insolation based on elevation, slope angle, and slope aspect grids for a site. SOLARPOT Index (SPI) calculates potential direct insolation (Sd) using the following formula from Dubayah et al. (1990): Sd = SCexp I cos Z
ð3Þ
where SC is the solar constant (2,880 cal/cm2day), is the optical depth, I is the solar incidence angle, and Z is the solar zenith angle. The output of this program is the predicted potential direct solar insolation at each pixel for a single-user defined day. Another approach was used to measure radiation potential at a site as a relative radiation index (TRASP) (Roberts and Cooper, 1989). TRASP assigns relative radiation values on the basis of aspect with output values ranging
130
Evan S. Hammer and Stephen J. Walsh
from 0 (for north-northeast slopes) to 1 (south-southwest slopes). The equation used to calculate TRASP is h i cos ðaspect 30Þ 180 TRASP = 1 ð4Þ 2 A similar approach has been used for modeling wind exposure. The Beer’s Index (BI) is a measure of the relative orientation of a site to its dominant wind direction. Southwest dominant winds were represented as follows: BI = cosðaspect 45Þ þ 1 ð5Þ where aspect offset of 45° from the aspect is used to orient the index; at other locations the offset is changed to agree with dominant winds, if they are not from the southwest; the output of the BI measurement of relative wind exposure ranges from 0 (a cell facing the wind) to 2 (a cell facing opposite the wind). The potential for snow accumulation is strongly influenced by wind direction (Allen, 1998; Allen and Walsh, 1993). It is not surprising that Cairns (2001) used a similar approach to modeling wind in his Snow Potential Index (SPI). SPI models potential snow accumulation in a cell as a function of wind, surface curvature, and elevation: SPI =
½cos ð225 aspectÞcurvatureðElev Elevmin Þ Elevrange
ð6Þ
4.3. Vector data sets Point locations of each krummholz patch sampled were collected using GPS units and were differentially corrected using postprocessing software provided with the hardware. The location data were exported into a GIS data layer for integration with the remaining vector and raster data sets. Patch geometry data collected in the field were integrated with the point location layer to derive the ‘‘polylines’’ representing major and minor axes of each patch, points representing sample location, and polygons representing patch boundaries.
4.4. Data processing 4.4.1. Canopy structure In this study, canopy structure is examined at two scales. First, individual sample points were examined to understand the organization of individual krummholz patches. Second, the sample points were aggregated to the level of the krummholz patch for the purpose of examining how krummholz canopy structure varies across the ATE. Variables measured in the field to represent canopy structure included LAI, canopy height, and canopy width. Using the LAI-2000, FD, DIFN, and DLLAI were generated. Unlike the LAI calculation, FD considers variation in the canopy height, while DLLAI considers sample area and sample volume, and canopy height.
Canopy Structure in the Krummholz and Patch Forest Zones
131
Figure 4 Three-dimensional representation of a krummholz canopy.
To calculate these secondary variables of canopy structure, measures of sample area, sample volume, and path length from each sample point to the top of the canopy were derived. Using the field data, 3D models of each krummholz canopy were developed (Figure 4). A Visual Basic program was written to automate these measurements. The program requires two data tables as input, one containing patch length and the other containing sample point location, height, and width measurements. The program iterates through the data for each krummholz patch. First, the sample height and width data are used to generate a triangulated irregular network (TIN) representing the top of the patch canopy. Creation of a TIN surface was chosen over the interpolation of a raster grid surface, because the TIN was better able to preserve the height values along the major axis and better maintained the character of the patches. Using the sample point location and the TIN, the distance from the sample point to the canopy (path length) was extracted. A second TIN model was developed that estimated the sample area of the LAI2000. This surface was used to extract the volume and area of the sample that cannot be extracted directly from the TIN surfaces. Each TIN is converted to a high-resolution grid surface, and a cut/fill operation is used to extract the area and volume of the sample. The output is a single table containing a record for each sample point and four fields: sample ID, path length, sample area, and sample volume. A second program uses the output of the first program to calculate the values of FD, DIFN, and DLLAI using the formulas previously presented. 4.4.2. Sample position within patch We hypothesized that an ‘‘edge effect’’ may influence the growth of vegetation within each patch. Weisberg and Baker (1995a) found a significant difference in
132
Evan S. Hammer and Stephen J. Walsh
leader growth at the leeward, windward, and interior of krummholz patches in Rocky Mountain National Park, Colorado, USA. Therefore, positional measurements derived for each sample point in this study included distance to midpoint, distance to centroid, distance to edge, distance to windward edge, 3D distance to edge, normalized distance to edge, and normalized distance to windward edge. Additionally, based on Weisberg and Baker (1995a) who compared leader growth from the patch edge and patch interior by defining the edge as the outer 20% of the patch and the interior as the remaining portion of the patch, samples have been categorized as interior or edge based on this 20% decision rule. Krummholz trees have more foliage at the top of their canopy and dwarf trees have the opposite (Cairns, 2005). 4.4.3. Patch position on landscape Determining the position of krummholz patches in the ATE is more difficult than deriving position within a krummholz patch, because the boundaries of the ATE are not as well defined. Several variables were generated as surrogates for the position of a patch within the ATE. Distance to nearest stream and distance to nearest ridgeline are both local measures that show the position of a patch on a hillslope. Distance to nearest drainage line is almost identical to distance to nearest streams. Drainage lines are representations of streams derived through automated terrain analysis and are a more standardized drainage network than the traditional stream network data on the topographic base maps, because local fine-grained terrain conditions were characterized through the 10-m DEM. Distance to nearest peaks is similar to distance to ridgelines. Both values are measures of the distance from the krummholz patch to the top of the ATE, but they are a more general measure not sensitive to the local variation in ridgelines than distance to nearest ridgeline considered. Distance to Continental Divide is more of a regional measurement intended to place each patch in its broader context.
5. Analysis The analysis of the krummholz patch canopy structure is divided into four steps. First, field data are transferred to a 3D GIS environment where digital models of each krummholz patch and sample regions are developed for exploratory data analysis and hypothesis development. Additional dependent and independent variables are also generated during this step. Second, a spatial analysis is performed at two scales. At the scale of the individual patch, canopy structure is examined with respect to the position of the sample point and relative to the patch center and edge. At the scale of the study site (a typical alpine treeline ecotone landscape), structural canopy data are aggregated to the patch, and spatial analysis is used to consider the influence by the position of the patch within the ecotone. Third, a biophysical analysis examines the relationship between the patch canopy and variables representing the biological and geomorphic setting of the patch. These variables include a set of terrain variables (i.e., elevation, slope, curvature, and TWI), a set of
Canopy Structure in the Krummholz and Patch Forest Zones
133
exposure variables (i.e., TSI and TRI), and a set of topo-climatic variables (i.e., SA, SPI, TRASP, and BI).
5.1. 3D visualization During the field data collection of LAI, several descriptive data sets were collected at the individual sample point. The species located at the sample point were recorded as well as canopy height and canopy width (relative to the major axis). These data, as well as patch length, are used to digitally reconstruct each krummholz patch in a 3D GIS environment. Additionally, 3D models of each sample region are also generated. The driving factor was to generate DLLAI values for each sample location. DLLAI, like LAI, is a measure of leaf area per unit of ground area. Traditionally, LAI was developed for use in a forest environment with relatively consistent canopy height. The LAI-2000 follows this assumption when generating LAI values following a set of above and below canopy measurements. As a consequence of the form of krummholz patches, it is difficult to compare LAI values with extreme variability in the size and shape of the actual region of the canopy sampled. DLLAI uses the same raw light transmittance data used to generate LAI but also considers the distance from the sample point to the top of the canopy, as well as the area and volume of the sampled portion of the canopy. By developing a 3D digital model, key variables are extracted. FD was also calculated using the data generated during this analysis. FD considers only the distance from the sample point to the canopy, excluding area and volume. In addition to providing the data for these two additional variables (DLLAI and FD), the 3D models allowed for the measurement of volume and area for each of the krummholz patches. These variables, as well as the surface area to volume ratio (a 3D shape index), are part of the spatial analysis. The importance of patch centroid over major axis midpoint for characterizing krummholz patches was recognized as a result of this visualization.
5.2. Spatial analysis The spatial analysis examines the factors controlling the spatial distribution of the internal canopy structure of krummholz patches in the study site, followed by an examination of the factors that influence spatial patterns of krummholz patch growth across the alpine treeline ecotone. 5.2.1. Within krummholz patches The objective of the spatial analysis was to identify variations in patch structural characteristics that indicate distinct patch edge and interior growth patterns. Correlation coefficients calculated for canopy structural variables and sample position are provided in Table 1. A significant positive correlation between sample height and distance from patch edge indicates that the canopy is higher in the middle and lower at the edges. Similarly, a significant positive correlation between LAI and
134
Table 1
Correlation matrix of canopy structural data and sample position data
n = 183 a = 0.05 To wind
To midpoint
To centroid
To edge
To edge 3D
To edge norm
To wind norm
0.0211 0.2144 0.0561 0.0632
0.0411 0.2342 0.0415 0.0876
0.0294 0.2511 0.0412 0.1111
0.3616 0.3090 0.0824 0.3905
0.2114 0.4770 0.3541 0.8668
0.2281 0.4044 0.0180 0.3103
0.1177 0.2555 0.0558 0.0030
Evan S. Hammer and Stephen J. Walsh
LAI Foliage Density DLLAI Height
Position within patch
Canopy Structure in the Krummholz and Patch Forest Zones
135
distance to patch edge indicates that in the center of the patch, there are more needles per unit of ground area than at the patch edge. Significant correlations also exist between FD and all positional variables. The directions of these correlations indicate that the canopy density is higher at the edge of patches and lower in the interior. This set of correlations points to the importance of considering canopy height when measuring needle density in this type of environment. LAI is a measure of leaf area per unit ground area: the ground area of the samples diminishes rapidly toward the ends of the major axis and remains relatively consistent through the remainder of the patch. This decrease in the ground area of the sample is primarily a result of the narrowing of the patch at the windward and leeward ends. Patch height also exerts an influence, but it is always secondary to patch width that acts as an absolute limit to the distance that the sample area can spread from the major axis. FD, by contrast, is a measure of leaf area per unit canopy volume measured. While canopy width is important in determining the volume of the sampled region of the patch, canopy height is more important, because two parts of the patch with equal width can produce sample areas of dramatically different volume depending on the height of the patch. Table 1 shows that canopy height is significantly correlated with distance to patch edge. The correlation between FD and distance to patch centroid and patch midpoint is significant, suggesting that the importance of patch height is not restricted to the patch edge. Next, variation between the canopy structure of the patch edge and the patch interior was analyzed. Two approaches were taken to categorize samples as interior, windward edge, edge, or other. For the 20% interior/edge division, samples within 20% of the windward edge of the patch were categorized as ‘‘windward’’ samples, samples within 20% of other edges are categorized as ‘‘edge’’ samples, and samples within 20% of the midpoint of the major axis were categorized as ‘‘interior.’’ The second approach categorizes samples within 25% of the windward edge as windward and samples within 25% of the other edges as ‘‘edge.’’ Samples were taken within 25% of the patch center and were ‘‘interior’’ samples, and remaining samples were categorized as ‘‘other.’’ ANOVA was used to determine whether the windward edge, edge, interior, and other samples represented different populations with respect to canopy structural values (Table 2). The ANOVA suggests distinct differences between canopy edge and interior. By separating out the windward edge from the remaining edge, the results of the ANOVA indicate that the mean LAI is highest in the interior, slightly lower at the windward edge, and lower still at the remaining edge locations. 5.2.2. Across krummholz patches To explore the spatial organization of canopy structure across the ATE, correlations were calculated between canopy structural variables and variables representing distances to key local and regional reference features. Distance to ridgeline, distance to streams, and distance to drainage are surrogates for the local topographic position of the patch, as they are potential measures of the position of a patch on the slope. Distance to ridgeline was calculated by taking the shortest distance from each
136
Table 2
Comparison of windward, edge, and interior structural samples; ANOVA results for patch location of 20 (a) and 25% (b) edge interior
(a)
(b) n
93 31 41 21
Pr(F):
LAI DLLAI log(FD)
LAI
DLLAI
log(FD)
4.283 4.573 3.579 5.229
2.866 3.248 2.498 3.022
2.449 2.811 2.510 1.795
0.00098096 0.09549413 0.00004519
All Windward Edge Interior Other Pr(F):
183 33 55 29 66 LAI DLLAI log(FD)
Means LAI
DLLAI
log(FD)
4.363 4.621 3.453 4.974 4.723 0.00009099 0.05520808 0.00000013
2.742 3.240 2.342 2.816 2.794
2.245 2.790 2.405 1.797 2.035 Evan S. Hammer and Stephen J. Walsh
All Windward Edge Interior
n
Means
137
Canopy Structure in the Krummholz and Patch Forest Zones
sample patch to the upslope boundary of the basin containing the patch. Distance to streams measures the shortest distance from a sample patch to a stream. Streams used for this measurement were derived from 1:24,000 scale USGS topographic maps. Distance to drainage was calculated by using a DEM to develop a topographic model of surface water flow. Drainage lines were created from this surface by indicating a minimum flow accumulation. This approach provided a stream network that closely conformed to observed drainage patterns but represented those patterns more uniformly than the networks printed on topographic maps. Distance to drainage was a measure of the shortest distance from a sample patch to a point along this uniform drainage network. Distance to peaks is a more general measure of the topographic position that places less emphasis on local topographic variability. Distance to the (Continental) Divide is a more regional measure of the same relative positioning appropriate for comparing samples from several study sites. 5.2.3. Lee Ridge At the Lee Ridge study site, our most intensively sampled site, canopy structural data were aggregated by averaging the samples on a patch by patch basis. Correlations were used to compare these aggregate measures of canopy structure to the patch positional information (Table 3). According to Table 3, there was no significant correlation between patch average DLLAI and any of the distance variables. Patch average LAI is correlated with the distance to the Divide as well as distance to peaks. FD and canopy height are correlated with these variables, as well as to the distance to streams. Canopy height is correlated with distance to drainage as well. As previously mentioned, the correlation with distance to Divide, though statistically significant, is likely not important to canopy structure at this scale. The direction to the Divide from Lee Ridge is roughly the same as the direction to the nearest peak. So, it is likely that this is a better explanation for the high correlation than an actual relationship between distance to divide and any of the canopy structural variables measured in the krummholz at Lee Ridge. Similarly, at Lee Ridge there is a stream that lies between many of the patches and Gable Mountain. Though this is the nearest stream, it is not the stream
Table 3
Correlation between patch average structural variables and landscape position
n = 28 a = 0.05
To streams
To peaks
To ridge
To drainage
To divide
LAI Foliage density DLLAI Height
0.3569 0.4695
0.5845 0.7048
0.2480 0.1964
0.3479 0.3385
0.4636 0.6010
0.1207 0.4268
0.2188 0.7064
0.1272 0.0220
0.0602 0.3295
0.1376 0.5639
138
Evan S. Hammer and Stephen J. Walsh
immediately down-slope of these patches. It is likely that the high correlation between patch structure and distance to streams is not indicative of the importance of local slope position with respect to stream location, but rather, it is due to the importance of the distance to the nearest peak. Thus, it seems likely that the meaningful relationships discovered in this case are between distance to peaks and LAI, FD, and canopy height. It is difficult to know whether the lack of correlation with the other variables is indicative of the low influence of local slope position on canopy structure or the inability of these variables to represent local topographic position. The effects of position in the ATE on the variance in canopy structure were also examined, but no significant correlations were revealed. Because of the intricate internal canopy structure explained in the earlier section, a method for considering this was also examined. Interior to edge ratios were calculated for each structural variable in an attempt to ascertain whether ATE position was related to the interior edge patterns. No significant correlations were found between interior to edge ratios of structural variables and ATE position at Lee Ridge. 5.2.4. All sites Similar analyses were performed using the 28 krummholz patches at Lee Ridge along with 29 krummholz patches from several other sites on the east side of the Continental Divide of GNP. Because sample height data were only collected at Lee Ridge, they were not available to calculate the additional structural variables: DLLAI, FD, and DIFN. When samples from the additional locations were considered, several different significant correlations are revealed. Table 4 shows that distance to peaks, which had been the primary significant correlation at Lee Ridge, is only significantly correlated with the variance in LAI across all sites. All of the other positional variables were found to be significantly correlated with the minimum and average LAI values for all patches. The correlations between average LAI and distance to streams, drainage, and ridgeline all appear to be in agreement and all seem to be capturing the effect of topographic position. Distance to streams and distance to drainage are generally the same. Both distance to streams and distance to drainage are intended to represent the location of a patch relative to the valley bottom or stream course. The inverse
Table 4 Correlation between patch aggregate LAI and landscape position at all
n = 57 a = 0.05
To streams
To peaks
To ridgeline
To drain point
To cont. divide
Min LAI Max LAI Ave LAI SD LAI Var LAI
0.5777 0.2540 0.5835 0.1774 0.1407
0.2467 0.1107 0.2468 0.2677 0.2773
0.5146 0.1187 0.4493 0.2400 0.2532
0.4574 0.0168 0.4008 0.3200 0.3103
0.4755 -0.0954 0.4064 0.3618 0.3644
139
Canopy Structure in the Krummholz and Patch Forest Zones
relationship between average LAI and each of these variables suggests that patch average LAI decreases with increasing distance from the valley bottom or stream course. A significant direct correlation between average LAI and distance to ridgeline corroborates this finding. As distance to ridgeline increases, a patch is located further downs-slope, and the average LAI values increase. This corresponds to the decreasing distance to streams and drainage. Beyond patch average LAI, it is notable that most of the distance variables are significantly correlated with patch minimum LAI but none are significantly correlated with patch maximum LAI. Though patch average LAI is decreasing with movement up-slope, and more low LAI values are present, some locations within the patch are able to thrive in a manner that is not constrained by local topographic position. This seems consistent with the significantly increasing standard deviation and variance of LAI as patch distance to drainage increases. These findings were further corroborated by generating an index of relative slope position (RSP). RSP considers both the upslope distance to ridgeline and the downslope distance to the valley bottom (Hatfield, 1999). Output values range from 0 to 100, with 0 indicating valley bottom, and 100 representing a ridge top. Table 5 indicates significant negative correlations between RSP and both minimum and average LAI indicating that LAI values decrease as patches move further upslope. Additionally, the significant positive correlation between RSP and both variance and standard deviation of LAI suggests that LAI may be increasingly difficult to predict as krummholz approaches its upper topographic limits.
5.3. Biophysical analysis The final analysis step involved exploring the relationship between canopy structure and biophysical processes occurring across the ATE. Three sets of variables were derived for use in this analysis. A set of terrain variables that were derived to represent geomorphic processes included elevation, slope angle, slope aspect, planform curvature, and profile curvature. General exposure is modeled based on the TSI and the Terrain Ruggedness Index (TRI). Topo-climatic processes are represented by SA (inverse wetness), SPI (snow), TRASP (radiation), and BI (wind).
Table 5
Correlation between relative slope position and LAI
n = 57 a = 0.05
RSP
Min LAI Max LAI Ave LAI SD LAI Var LAI
0.49449 0.07948 0.43621 0.31636 0.30296
140
Evan S. Hammer and Stephen J. Walsh
5.3.1. Lee Ridge A direct correlation between curvature and FD on Lee Ridge (Table 6) indicates that FD increases as the surface moves from upwardly concave to upwardly convex. This seems opposite of what might be expected, because a concave surface would seem to provide more shelter to a patch and therefore, be more amenable to dense canopy growth. At the same time, the analysis of the internal patch canopy indicates that the most exposed portion of the patch, the windward end, produced higher values for all of the canopy structural variables suggesting that exposure may actually result in a denser canopy. This would be consistent with the finding that more exposed convex slopes produce a denser canopy. Elevation showed significant correlation with canopy height, LAI, and FD at Lee Ridge. These correlations indicate decreasing canopy height and increasing canopy density with increasing elevations. This seems to contradict the earlier findings that canopy density decreased with movement upslope. Patch size may be playing an undetermined role in the relationship between terrain and canopy structure. LAI does not consider canopy height, while FD does. Neither variable considers the size of the region of the patch that the LAI sample represents. DLLAI, which does consider the size of the sampled region, does not show a significant correlation with elevation. This may indicate that the correlations are a result of decreasing patch size with increasing elevation. The variance of canopy structural variables showed some correlation with surface curvature, and more specifically, planform curvature (Table 7). The correlation between variation in DLLAI and planform curvature corresponds with the Table 6
Correlation between patch average structural variables and terrain
n = 28 Elevation Curvature Planform a = 0.05 curvature LAI Foliage density DLLAI Height
Profile Slope Slope curvature (degrees) (percent)
0.3937 0.5625
0.1731 0.3946
0.1363 0.3206
0.1579 0.3478
0.1083 0.1689
0.1112 0.1725
0.1918 0.5504
0.2598 0.3492
0.3291 0.3643
0.0837 0.2086
0.1353 0.2766
0.1282 0.2770
Table 7 Correlation between variance in structural variables and terrain
n = 28 Elevation Curvature Planform a = 0.05 curvature LAI Foliage density DLLAI
Profile Slope Slope curvature (degrees) (percent)
0.0943 0.0911
0.4363 0.0882
0.4003 0.1642
0.3281 0.0359
0.2426 0.0054
0.2405 0.0064
0.2302
0.3032
0.3869
0.0944
0.2335
0.2268
Canopy Structure in the Krummholz and Patch Forest Zones
141
earlier finding of correlation between DLLAI and minimum flow accumulation. Planform curvature is a measure of the convergence of slope which influences flow accumulation. The earlier finding was that DLLAI increased with average minimum flow accumulation. This correlation suggests that DLLAI is more stable (variance decreases) with increasing slope convergence (increased planform slope value). 5.3.2. All sites The biophysical analysis continued with an examination of the terrain variables across all of the sample patches (Table 8). Curvature and planform curvature in particular show a significant correlation with maximum LAI and the standard deviation of LAI values. This correlation suggests that higher levels of slope convergence result in lower patch maximum LAI values. Perhaps by limiting the range of LAI values in a patch, this decreases the variability in patch LAI measures as indicated by the negative correlation between planform curvature and standard deviation of LAI. Slope aspect measurements were grouped into four categories representing the four cardinal directions. While using eight directions may have been more telling, four were chosen in hopes of creating more balanced categories. This was somewhat successful, as 7 patches are on north-facing slopes, 17 on east-facing slopes, 2 on south-facing slopes, and 31 on west-facing slopes. Only southern aspects had significantly different patch average LAI values than any of the other directions, but only two patches were sampled. Patches on north- and south-facing slopes also show lower variances, but this is also probably best explained by the low sample sizes on these two slope aspects. Of the remaining analysis, most indicate no significant influence from slope aspect. The minimum LAI, however, indicates that patches on east-facing slopes have significantly higher LAI values than patches on either north-facing or westfacing slopes. At least on Lee Ridge where winds move from the west-southwest, this seems like a reasonable finding. East-facing slopes should provide protection from harsh winds, as well as some protection from incoming solar radiation that should be the most direct on southwest-facing slopes. Why this only seems to affect minimum LAI is unclear, but perhaps these krummholz patches are typically able to maintain a certain minimum level of productivity in most conditions. When these patches are stressed and exposed, they seem to continue to maintain the capacity to produce at a high average rate and even reach similar maximum production levels as in sheltered sites. At the same time, however, some parts of the patch sustain greater damage and cannot be maintained as well as in a sheltered location, thus the lower minimum LAI values. It is also reasonable to suspect that a greater range of aspects may be more useful in determining why LAI responds to some aspects differently than others. Rather than try to test several aspects at this point to try to determine the specific reason for the correlation between LAI and aspect, the analysis will move on to a set of topo-climatic variables. Many consider the effect of slope aspect in their calculations.
142
Table 8 Correlation between patch aggregate LAI and terrain at all sites
Elevation
Curvature
Planform curvature
Min LAI Max LAI Ave LAI SD LAI Var LAI
0.1062 0.0731 0.0393 0.1362 0.0629
0.0937 0.2253 0.1645 0.3017 0.2552
0.0043 0.2953 0.2137 0.2819 0.2427
Profile curvature
Slope (degrees)
Slope (percent)
TRI
0.1573 0.0666 0.0508 0.2133 0.1759
0.2563 0.0883 0.1899 0.1297 0.1490
0.2481 0.0885 0.1813 0.1261 0.1470
0.2673 0.1119 0.1772 0.1680 0.1793
Evan S. Hammer and Stephen J. Walsh
n = 57 a = 0.05
143
Canopy Structure in the Krummholz and Patch Forest Zones
Table 9
Correlation between patch LAI and topo-climatic variables
n = 57 a = 0.05
SA
SPI
Beers
TRASP
Min LAI Max LAI Ave LAI SDLAI Var LAI
0.0682 0.3035 0.2203 0.2591 0.2329
0.1005 0.2259 0.0917 0.2585 0.2127
0.4653 0.2648 0.5286 0.0102 0.0515
0.0205 0.0694 0.0387 -0.0919 0.1595
5.4. Topo-climatic variables A subset of the topo-climatic variables previously described were used to examine the influence of soil moisture, snow accumulation, wind, and solar radiation on canopy structure. While methods for modeling solar radiation were discussed earlier, the TRASP model, which is a simple measure of whether or not a cell faces the southwest, was chosen over the SOLARPOT approach primarily based on ease of use. The SOLARPOT model, however, does appear significantly more robust, as it considers several factors such as shadowing by landscape features and the changing position of the sun in the sky. The other topo-climatic variables compared with LAI included SA, SPI, and BI. Table 9 provides the results of the correlation between these two sets of variables. Interestingly, SA is only significantly correlated with maximum LAI values. The particular wetness index used is actually an inverse wetness measure, so a negative correlation actually indicates the correspondence between high LAI and moist areas. Neither SPI nor TRASP show any significant correlation with LAI which is interesting, because both snow and radiation are important to the growth processes of the vegetation in this area. The possible weaknesses of TRASP have already been addressed, and perhaps SOLARPOT would have been more revealing. SPI is a little more surprising, especially given the high correlation between BI and both minimum and average LAI. Given that the first part of the SPI model is essentially the same as BI, the lack of correlation is difficult to explain. Perhaps the use of a relative term for elevation caused some problems in this study, which examines data from so many locations that often exhibit extremely irregular relief.
6. Results and Discussion The purpose of this research was to explore patterns of krummholz canopy growth at alpine treeline. Two levels of organization were explored, canopy structure within individual krummholz patches and the canopy structure across the ATE. At each level of analysis, the importance of location was examined. In the analysis of structure across the ATE, the influence of several biophysical variables was also studied.
144
Evan S. Hammer and Stephen J. Walsh
6.1. Internal canopy structure The study of within canopy structure began with an examination of the influence of species type on the measurements of each of the canopy structure variables. Categorizing samples on the basis of the species type recorded at the sample point revealed that LAI values recorded for pines (Pinus contorta and Pinus flexilis and Pinus albicaulis) were lower than LAI values recorded for subalpine fir (Abies lasiocarpa). Spruce (Picea engelmannii) also recorded higher LAI values than the pines. Neither DLLAI nor FD showed similar distinctions between species. While all tree canopies show severe deformation from the impact of wind, pines respond to this influence differently than spruce and fir. Spruce and fir are able to adapt to these harsh conditions through the process of vegetative layering. When layering takes place, canopies tend to take the form of low stature krummholz mats. Pines, however, lack the capacity to reproduce through layering (Holtmeier, 1981). Given these considerations, the differences in LAI and FD that were observed for different species may be a result of the height differences in these samples or patches rather than an actual difference in the density of leaf area per unit ground area. A comparison of canopy height measurements categorized by species hints at this, because though not significantly different statistically, the canopy height and volume values recorded for pines were higher than those for spruce and fir. These findings further suggest that the differences in LAI and FD may be related to canopy height and volume. If this is the case, the fact that DLLAI measurements did not differ across species suggests that efforts to derive this variable was worthwhile, because it was buffered from the influence of changing canopy height and volume in its calculation of canopy density. An additional test categorized samples based on the number of species found in the patch. In this case, both LAI and FD values were significantly higher in single species patches than those recorded in patches containing four species. There was also a general trend of decreasing LAI and FD values from single species patches to patches with two, three, and four species. Further, though significant differences were not noted in DLLAI values, all measurements, including DLLAI, showed lower variance for single species patches than for those containing multiple species. Patches containing a single tree species appear to be able to produce a more consistent canopy that may produce a greater number of needles per ground area, and the range of recorded values should certainly be smaller as well. The analysis of the influence of patch position on canopy structure, one of the primary objectives of this research, also produced some notable results. Correlations were calculated between LAI, FD, DLLAI, and several positional variables based on the distance to the edge, interior, or windward end of the patch. LAI and FD were both positively correlated with variables measuring distance to patch edge. The influence of height is again likely to be playing a role, as DLLAI did not significantly correlate with any measures other than 3D distance to edge. Canopy height was also found to have significant positive correlation with distance to patch edge indicating that the canopy becomes lower as it approaches the patch edge. Measurements of distance to patch edge proved to be more important than measurements of the distance to patch center. One explanation for this is that over a
Canopy Structure in the Krummholz and Patch Forest Zones
145
set of patches ranging from just a few meters in length to 55 m long, the distance to the patch center across multiple patches varies greatly, while distance to the edge and the meaning of that value remains consistent. A normalized measure of distance to the patch center may have been a more effective tool. By categorizing samples as windward, edge, and interior and then using ANOVAs to compare these classes, LAI, FD, and DLLAI were all found to have significantly different values based on patch position. Interior samples recorded the highest LAI and FD values, while DLLAI measures were highest at the windward edge. While the divisions used in this study are slightly different, there is a clear distinction between edge and interior values of these measurements of needle density within the canopy. Coupled with the findings of Weisberg and Baker (1995b) which show that leader growth in krummholz differed between the edge and interior of the patch, there is strong evidence of meaningful structure within patches of krummholz vegetation which must be examined further to be fully understood.
6.2. Between patch structure Across the ATE, canopy structure was examined with respect to position in the ecotone and the relationship between canopy structure and biophysical variables representing potential moisture, snow accumulation, wind, and solar radiation. Unlike the position of a sample in a patch, which has an easily measurable edge and center, position of a patch in the ATE is more difficult to define. While the ATE generally moves from the forest edge (down-slope) to the lower limit of tundra (up-slope), actual ATE landscapes tend to be more complex than this. At Lee Ridge, for example, forests are found down-slope in two and sometimes three directions as a result of the site being located on a ridge-top. While the dominant ridge begins in the north and rises to the south, there is significant wind influence at this site, and the winds blow generally west to east. Distances to peaks, ridgelines, streams, and the Continental Divide were all used as surrogates for the position of the patch within the ATE. In general, these are all measurements of the position of the patch on a slope with the major distinction being a matter of scale. Distance to local features reflected local variations in slope, while the distance to more regional features was less influenced by local variability. Of the variables used, the local measures of distance to ridgelines, distance to streams, and distance to drainage all had significant correlations with LAI. A further analysis examined the correlation between RSP and LAI. In each case, the position of a patch on a slope is significantly correlated, with LAI values higher at lower slope positions closer to the valley floor and decreasing upslope toward the ridgeline. Further, variance and standard deviation of LAI increases with movement upslope. While measures of patch average LAI and patch minimum LAI were significantly correlated with the position of the patch on the slope, this correlation did not extend to patch maximum LAI values. This suggests that while some part of the patch is able to maintain high plant density at higher slope positions, the density of the surrounding patch is decreasing, or perhaps becoming responsible for a larger proportion of the patch as slope position increases.
146
Evan S. Hammer and Stephen J. Walsh
The analysis of the influence of biophysical variables was divided between a set of terrain variables and a set of topo-climatic variables. Most of the terrain variables showed no significant correlation with LAI measurements. Curvature, however, had a strong negative correlation with the standard deviation of LAI, and planform curvature was negatively correlated with both patch maximum LAI and standard deviation of LAI. This indicates that as slopes become more convex, the maximum attainable LAI values on these slopes decrease and the LAI measures become more consistent (perhaps because of a lower possible range of values). This is probably a result of the decreasing amount of shelter provided to a krummholz patch as the slope moves from concave to convex. The wetness and snow potential variables examined in the topo-climatic analysis both consider curvature in their calculation for this reason. The notable result of the topo-climatic analysis is the significance of wind. Increases in the values of the wind index, which represent increasing wind sheltering, were correlated with increasing values of average and minimum LAI. Similarly, the wetness index, in which increasing values represent drier sites, showed strong negative correlation with maximum LAI. Moist sites seem to encourage patches to have higher maximum LAI values, while higher levels of wind exposure limit the average and minimum LAI values.
7. Conclusions Several findings suggest that the distribution of LAI across individual patches, as well as across the ecotone, is not a random occurrence. Edge effects were hypothesized to control these distributions. Within krummholz patches, the canopy structure showed distinct differences between edge and interior. Though all of the boundaries of the krummholz patch were influenced by external factors, the way that these influences express themselves seem highly dependent on which edge is being studied. In this particular analysis, wind was a significant influence. The process of wind interaction with krummholz edges differs depending upon whether the edge is at the windward or leeward end of the patch. At the level of the ATE, slope position, curvature, and wind are all correlated with canopy structure. While slope position was intended to represent a surrogate for patch position within the ATE, it is questionable whether some other process was actually being represented. Perhaps, a set of variables relating to the distance from the subalpine forest edge and/or the distance to tundra would be more revealing regarding the role of patch position in the ATE. The influences of wind and planform curvature seem to be in agreement with one another. Many patches were located in stream channels and gulleys that have planform concavity. These sites are amenable to krummholz growth as a result of the shelter they offer from prevailing winds. In sum, the ATE in GNP is highly variable in its pattern–process relations (Allen and Walsh, 1993). Considerable topographic complexity exists, as well as notable disturbance regimes (e.g., snow avalanches and debris flows) that further
Canopy Structure in the Krummholz and Patch Forest Zones
147
shape the ATE. In this research, field sampling of krummholz patches through the calculation of LAI and derivatives was a useful way to discriminate among forest patches, further mediated by geographic site and situation, notably terrain settings and geomorphic conditions affecting biotic and abiotic factors (cf. Malanson et al., 2007). The research indicates the relevance of LAI to distinguish krummholz patches, the importance of deriving 3-dimensional models of forest patch canopies, and the terrain settings that describe patch form, state, and conditions.
REFERENCES Alftine, K.J., Malanson, G.P., 2004. Directional positive feedback and pattern at an alpine tree line. Journal of Vegetation Science 15, 3–12. Alftine, K.J., Malanson, G.P., Fagre, D.B., 2003. Feedback driven response to multidecadal climatic variability at an alpine treeline. Physical Geography 24, 520–535. Allen, T.R., 1995. Relationship Between Spatial Pattern and Environment at the Alpine Treeline Ecotone, Glacier National Park, Montana. Doctoral Dissertation, University of North Carolina, Chapel Hill. Allen, T.R., 1998. Topographic context of glaciers and perennial snowfields, Glacier National Park, Montana. Geomorphology 21, 207–216. Allen, T.R., Walsh, S.J., 1993. Characterizing multitemporal alpine snowmelt patterns for ecological inferences. Photogrammetric Engineering and Remote Sensing 59, 1521–1529. Allen, T.R., Walsh, S.J., 1996. Spatial and compositional pattern of alpine treeline, Glacier National Park, Montana. Photogrammetric Engineering and Remote Sensing 62, 1261–1268. Arno, S.F., Hammerly, R.P., 1984. Timberline: Mountain and Arctic Forest Frontiers. The Mountaineers, Seattle. Baker, W.L., Honaker, J.J., Weisberg, P.J., 1995. Using aerial photography and GIS to map the forest-tundra ecotone in Rocky Mountain National Park, Colorado for global change research. Photogrammetric Engineering and Remote Sensing 61, 313–320. Baker, W.L., Weisberg, P.J., 1997. Using GIS to model tree population parameters in the Rocky Mountain National Park forest-tundra ecotone. Journal of Biogeography 24, 513–526. Bekker, M.F., 2005. Positive feedback between tree establishment and patterns of subalpine forest advancement, Glacier National Park, Montana, USA. Arctic, Antarctic, and Alpine Research 37, 97–107. Benedict, J.B., 1984. Rates of tree-island migration, Colorado Rocky Mountains, USA. Ecology 65, 820–823. Beven, K.J., Kirkby, M.J., 1979. A physically based, variable contributing area model of basin hydrology. Hydrological Sciences 24, 43–69. Bian, L., Walsh, S.J., 1993. Scale dependencies of vegetation and topography in a mountainous environment of Montana. Professional Geographer 45, 1–11. Bowersox, M.A., Brown, D.G., 2001. Measuring the abruptness of patchy ecotones – A simulationbased comparison of landscape pattern statistics. Plant Ecology 156, 89–103. Brooks, P.D., Williams, M.W., Schmidt, S.K., 1996. Microbial activity under alpine snowpacks, Niwot Ridge, Colorado. Biogeochemistry 32, 93–113. Brown, D.G., 1992. Topographical and Biophysical Modeling of Vegetation Patterns at Alpine Treeline. Doctoral Dissertation, University of North Carolina, Chapel Hill. Brown, D.G., 1994. Predicting vegetation types at treeline using topography and biophysical disturbance variables. Journal of Vegetation Science 5, 641–656. Brown, D.G., 2001. A spectral unmixing approach to leaf area index (LAI) estimation at the alpine treeline ecotone. In: Millington, A.C., Walsh, S.J., Osborne, P.E. (Eds.), Gis and Remote Sensing Applications in Biogeography and Ecology. Kluwer, Boston, pp. 7–21.
148
Evan S. Hammer and Stephen J. Walsh
Butler, D.R., 1986. Snow-avalanche hazards in Glacier National Park, Montana: Meteorologic and climatologic aspects. Physical Geography 7, 72–87. Butler, D.R., Walsh, S.J., 1990. Lithologic, structural, and topographic influences on snow avalanche path location, eastern Glacier National Park, Montana. Annals of the Association of American Geographers 80, 362–378. Cairns, D.M., 2001. A comparison of methods for predicting vegetation type. Plant Ecology 156, 3–18. Cairns, D.M., 2005. Simulating carbon balance at treeline for krummholz and dwarf tree growth forms. Ecological Modelling 187, 312–328. Cairns, D.M., Malanson, G.P., 1998. Environmental variables influencing carbon balance at the alpine treeline ecotone: A modeling approach. Journal of Vegetation Science 9, 679–692. Cairns, D.M., Waldron, J.D., 2003. Sigmoid wave transitions at alpine treeline. Geografiska Annaler 85A, 115–126. Carrara, P.E., 1989. Late Quaternary Glacial and Vegetative History of the Glacier National Park Region, Montana. US Geological Survey Bulletin 1902, Washington. Cox, C.F., 1933. Alpine plant succession on James Peak, Colorado. Ecological Monographs 3, 299–372. Daubenmire, R.F., 1943. Vegetational zonation in the Rocky Mountains. The Botanical Review 9, 326–393. Dubayah, R., Dozier, J., Davis, F.W., 1990. Topographic distribution of clear-sky radiation over the Konza Prairie, Kansas. Water Resources Research 26, 679–690. Franklin, S.E., Lavigne, M.B., Deuling, M.J., Wulder, M.A., Hunt, E.R., 1997. Estimation of forest leaf area index using remote sensing and GIS for modelling net primary production. International Journal of Remote Sensing 18, 3459–3471. Gower, S.T., Norman, J.M., 1991. Rapid estimation of leaf area index in conifer and broad-leaf plantations. Ecology 72, 1896–1900. Griggs, R.F., 1938. Timberlines in the northern Rocky Mountains. Ecology 19, 548–564. Hadley, J.L., Smith, W.K., 1986. Wind effects on needles of timberline conifers: Seasonal influence on mortality. Ecology 67, 12–19. Hadley, J.L., Smith, W.K., 1987. Influence of krummholz mat microclimate on needle physiology and survival. Oecologia 73, 82–90. Hansen-Bristow, K.J., Ives, J.D., 1985. Composition, form, and distribution of the forest-alpine tundra ecotone, Indian Peaks, Colorado, USA. Erdkunde 39, 286–295. Hatfield, D.C., 1999. TopoTools – A collection of topographic modeling tools for ARC/INFO. In: Proceedings of the 20th Annual ESRI International Users Conference. ESRI, San Diego, CA. Holland, M., Risser, P.G., Naiman, R.J., 1991. Ecotones : The Role of Landscape Boundaries in the Management and Restoration of Changing Environments. Chapman and Hall, New York. Holtmeier, F.-K., 1981. What does the term ‘‘krummholz’’ really mean? Observations with special reference to the Alps and the Colorado Front Range. Mountain Research and Development 1, 253–260. Holtmeier, F.-K., Broll, G., 1992. The influence of tree islands and microtopography on pedoecological conditions in the forest alpine tundra ecotone on Niwot Ridge, Colorado Front Range, USA. Arctic and Alpine Research 24, 216–228. Johnson, P.L., Billings, W.D., 1962. The alpine vegetation of the Beartooth Plateau in relation to cryopedogenic processes and patterns. Ecological Monographs 32, 105–135. Ko¨rner, C., 1999. Alpine Plant Life : Functional Plant Ecology of High Mountain Ecosystems. Springer, Berlin. LI-COR, 1990. LAI-2000 Plant Canopy Analyzer. Instruction Manual. LI-COR, Lincoln, NE. Mackay, D.S., Band, L.E., 1997. Forest ecosystem processes at the watershed scale: Dynamic coupling of distributed hydrology and canopy growth. Hydrological Processes 11, 1197–1217. Malanson, G.P., Xiao, N., Alftine, K.J., 2001. A simulation test of the resource-averaging of ecotone formation. Journal of Vegetation Science 12, 743–748. Malanson, G.P., Butler, D.R., Fagre, D.B. 2007. Alpine ecosystem dynamics and change: a view from the heights. In: Prato, T., and Fagre, D. (Eds.), Sustaining Rocky Mountain Landscapes: Science, Policy and Management of the Crown of the Continent Ecosystem. RFF Press, Washington D.C., pp. 85–101.
Canopy Structure in the Krummholz and Patch Forest Zones
149
Marr, J.W., 1977. The development and movement of tree islands near the upper limit of tree growth in the southern Rocky Mountains. Ecology 58, 1159–1164. McGregor, S.J., 1998. An integrated geographic information system approach for modeling the suitability of conifer habitat in an alpine environment. Geomorphology 21, 265–280. McKnight, S.A., 2004. Composition and spatial structure of turf-banked terraces in Glacier National Park, Montana, USA. M.A. Thesis, University of North Carolina, Chapel Hill. McNab, W.H., 1989. Terrain shape index: Quantifying effect of minor landforms on tree height. Forest Science 35, 91–104. McNab, W.H., 1993. A topographic index to quantify the effect of mesoscale landform on site productivity. Canadian Journal of Forest Research 23, 1100–1107. Moore, I.D., Grayson, R.B., Ladson, A.R., 1991. Digital terrain modelling: A review of hydrological, geomorphological, and biological applications. Hydrological Processes 5, 3–30. Parker, E.R., Sandford, R.L., Jr., 1999. The effects of mobile tree islands on soil phosphorus concentrations and distribution in an alpine tundra ecosystem on Niwot Ridge, Colorado Front Range, USA. Arctic, Antarctic, and Alpine Research 31, 16–20. Pauker, S.J., Seastedt, T.R., 1996. Effects of mobile tree islands on soil carbon storage in tundra ecosystems. Ecology 77, 2563–2567. Pontecorvo, G., Bokhari, M., 1975. Hedge-like habit of Juniperus excelsa at high altitude on the southern Zagros mountains in Iran. Proceedings of the Royal Society of London, B 188, 507–508. Riley, S.J., DeGloria, S.D., Elliot, R., 1999. A terrain ruggedness index that quantifies topographic heterogeneity. Intermountain Journal of Sciences 5, 23–27. Roberts, D.W., Cooper, S.V., 1989. Concepts and techniques of vegetation mapping. In: Johnson, F.D. (Ed.), Proceedings, Land Classifications Based on Vegetation: Applications for Resource Management. USDA Forest Sevice General Technical Report INT-257, pp. 90–96. Walsh, S.J., Bian, L., McKnight, S., Brown, D.G., Hammer, E.S., 2003a. Solifluction steps and risers, Lee Ridge, Glacier National Park, Montana, USA: A scale and pattern analysis. Geomorphology 55, 381–398. Walsh, S.J., Bian, L., McKnight, S., Hammer, E., Erlien, C.M., 2003b. Characterizing the spatial pattern of solifluction steps and risers in an alpine environment: Application of digital image enhancements and geostatistics. ASPRS 2003 Annual Conference Proceedings, Anchorage, AK. Walsh, S.J., Butler, D.R., 1989. Spatial pattern of snow avalanche path location and morphometry, Glacier National Park, Montana. GIS/LIS 0 89 Proceedings, pp. 286–294. Walsh, S.J., Butler, D.R., 1997. Morphometric and multispectral image anaysis of debris flows for natural hazard assessment. Geocarto International 12, 59–70. Walsh, S.J., Butler, D.R., Allen, T.R., Malanson, G.P., 1994. Influence of snow patterns and snow avalanches on the alpine treeline ecotone. Journal of Vegetation Science 5, 657–672. Walsh, S.J., Butler, D.R., Brown, D.G., Bian, L., 1990. Cartographic modeling of snow avalanche path location within Glacier National Park, Montana. Photogrammetric Engineering and Remote Sensing 56, 615–621. Walsh, S.J., Butler, D.R., Malanson, G.P., Crews-Meyer, K.A., Messina, J.P., Xiao, N., 2003c. Mapping, modeling, and visualization of the influences of geomorphic processes on the alpine treeline ecotone, Glacier National Park, Montana, USA. Geomorphology 53, 129–145. Walsh, S.J., Moody, A., Allen, T.R., Brown, D.G., 1997. Scale dependance of NDVI and its relationship to mountainous terrain. In: Quattrochi, D.A., Goodchild, M.F. (Eds.), Scale in Remote Sensing and GIS. CRC Press, Boca Raton, FL, pp. 27–55. Wardle, P., 1968. Engelmann spruce (Picea engelmannii Engel.) at its upper limits on the Front Range, Colorado. Ecology 49, 483–495. Weisberg, P.J., Baker, W.L., 1995a. Spatial variation in tree seedling krummholz growth in the foresttundra ecotone of Rocky Mountain National Park, Colorado, USA. Arctic and Alpine Research 27, 116–129. Weisberg, P.J., Baker, W.L., 1995b. Spatial variation in tree regeneration in the forest-tundra ecotone Rocky Mountain National Park Colorado. Canadian Journal of Forest Research 25, 1326–1339. Welles, J.M., Norman, J.M., 1991. Instrument for indirect measurement of canopy architecture. Agronomy Journal 83, 818–825.
150
Evan S. Hammer and Stephen J. Walsh
White, J.D., Running, S.W., Thornton, P.E., Keane, R.E., Ryan, K.C., Fagre, D.B., Key, C.H., 1998. Assessing simulated ecosystem processes for climate variability research at Glacier National Park, USA. Ecological Applications 8, 805–823. Wilkerson, F.D., Schmid, G.L., 2003. Debris flows in Glacier National Park, Montana: Geomorphology and hazards. Geomorphology 55, 317–328. Zeng, Y., Malanson, G.P., 2006. Endogenous fractal dynamics at alpine treeline ecotones. Geographical Analysis 38, 271–287. Zeng, Y., Malanson, G.P., Butler, D.R., 2007. Geomorphic limits to self organization in alpine forest-tundra ecotone vegetation. Geomorphology 91, 378–392.
C H A P T E R
8
A Markov Analysis of Tree Islands at Alpine Treeline Lynn M. Resler and Mark A. Fonstad
Contents 1. Introduction 2. Methods 2.1. Markov chain analysis 2.2. Embedded Markov chains 2.3. First-order Markov chains 3. Results 3.1. Spatial sequence of conifer establishment 3.2. Stability at treeline 4. Discussion 4.1. Establishment characteristics 4.2. Treeline stability 5. Conclusions References
151 154 154 157 157 158 158 158 159 160 161 163 163
1. Introduction Tree islands are an important and dynamic component of the alpine treeline landscape in Glacier National Park (GNP), Montana. Composed of an individual tree or several conifers of varying species, tree islands are landscape elements distinct from the upper subalpine continuous forest cover in terms of composition, pattern, and function. Pattern across the alpine treeline ecotone in GNP is often manifest as a transition from upright, closed canopy trees to dwarfed flagged trees or stunted trees, to tundra interspersed with sparse krummholz trees, and eventual replacement by krummholz (Allen and Walsh, 1996; Bekker, 2005). Tree islands of varying sizes may be found throughout the alpine treeline ecotone, and conifers in tree islands may assume any of these growth forms. Though alpine treeline dynamics are often researched from a landscape scale perspective with emphasis on the relationship between adult tree growth and Developments in Earth Surface Processes, Volume 12 ISSN 0928-2025, DOI 10.1016/S0928-2025(08)00208-3
Ó 2009 Elsevier B.V. All rights reserved.
151
152
Lynn M. Resler and Mark A. Fonstad
abiotic drivers such as mean temperature (Daubenmire, 1954; Tranquillini, 1979), topography (Brown, 1994), and disturbance factors (Butler and Walsh, 1994), analysis of fine-scale components of alpine treeline, such as tree islands, may reveal information on treeline stability from the perspective of establishment, survival, and mortality of conifers. Specifically, the establishment and survival of seedlings is important to the geographic position and spatial pattern of treeline on a landscape scale (Callaway, 1995; DeAngelis et al., 1986), because such fine-scale processes ultimately determine the upslope migration of trees (Germino et al., 2002; Hattenschwiler and Smith, 1999). Tree islands at alpine treeline in GNP provide shelter and ameliorate local climate conditions, thereby aiding in subsequent conifer establishment (Bekker, 2005; Resler et al., 2005). The sheltering effects of tree islands have also been shown at other western treelines (Germino et al., 2002). In this regard, the establishment of a conifer at alpine treeline might be dependent, at least to some degree, upon the presence of a preexisting conifer. In an abiotically stressful environment, such as alpine treeline, trees experience greater climatic stresses than those at lower elevations (Ko¨rner, 1999). Therefore, the moderating effects of existing conifers and surface microtopography are particularly important for seedling establishment and survival (Figure 1) because they expand local resource availability through shading and offering structural protection from wind (Benedict, 1984; Marr, 1977), increase the water-holding capacity of soils (Lamb and Chapman, 1943; Pe´rez, 1987, 1991), and serve as seed or sediment traps. In highly exposed tundra locations, conifer establishment is patchy and new seedlings are found growing in sheltered locations, such as in the lee of microtopographic features such as boulders (Figure 2), periglacial patterned ground, and topographic
Figure 1 A seedling growing in the lee of a larger tree island at Lee Ridge; a common phenomena at alpine treeline in GNP. Predominant wind direction is from the west (top of the photo).
A Markov Analysis of Tree Islands at Alpine Treeline
153
Figure 2 Pine establishing in the lee of a boulder at upper limits of the alpine treeline ecotone, Lee Ridge study site.
hollows that serve as shelter (Resler et al., 2005). Research by Callaway et al. (2002) has indicated that the relative importance of positive plant interactions such as facilitation (as opposed to resource limiting interactions such as competition) increases in increasingly stressful abiotic environments. Many tree islands in GNP are composed of conifer species that, under less stressful conditions, would not likely coexist. For example, Pinus albicaulis, Pinus contorta, and Larix lyallii have been found coexisting in a single patch, within a relatively small area (Resler, 2004). The purpose of this study is to characterize fine-scale attributes of alpine treeline, such as conifer establishment characteristics, species associations, and conifer
154
Lynn M. Resler and Mark A. Fonstad
growth in tree islands at three alpine treeline study locations in GNP, Montana. Specific objectives of this research are (a) to characterize the species-specific patterns of conifer establishment in tree islands throughout the alpine treeline ecotone using Markov chains, a statistical technique useful for analyzing sequences of data and (b) to determine whether there are predictable patterns to the spatial establishment of conifers in exposed patches. We are interested in the characteristics and dynamics of conifer establishment in tree islands for what they may reveal about landscape-scale ecotone dynamics. Because each conifer species at alpine treeline has different environmental tolerances, reproductive strategies, and susceptibility to environmental impacts, each has different potential to effect community structure, diversity, and ultimately, the stability of alpine treeline.
2. Methods We chose three representative sites (Lee Ridge, Divide Mountain, and Cataract Creek Basin) in GNP, Montana, immediately above alpine treeline, on the eastern slopes of the Continental Divide. Slopes east of the Divide are typically dominated by a dry, continental climate regime (Finklin, 1986). All sites are windswept periglacial environments and are found at elevations ranging from 2,050 to 2,150 m. Tree islands, including individual trees and islands with multiple trees, are typically found in conjunction with boulders, other vegetation (such as Salix spp. or trees), or microtopography created by the presence of periglacial features that form a step-like pattern on slopes with gradients from 3 to 7° (Resler et al., 2005). Tree islands may range in species composition from one species to several species. Elevation and relative topography at the sites precluded Pleistocene glaciations, and seedlings have grown in the last 50 years (Alftine and Malanson, 2004). While each site is similar with regard to microtopography, macrotopography differs. Variation in macrotopography among sites influences variation in predominant wind directions. These variations in wind directions result in patches of different shapes at each site and likely influence seed dispersal patterns. Sampling boundaries at each study site were delineated by placing a 30 m 30 m quadrat in each of five elevation zones (beginning near the edge of closed forest, moving upslope). Species intercepts along transects oriented lengthwise along each tree island captured species richness and abundance data that detailed the establishment characteristics of each tree island (Mueller-Dombois and Ellenberg, 1974). Measurements along transects began with the edge of the shelter source, where the measurement was considered zero.
2.1. Markov chain analysis Markov chains, when used in conjunction with probability statistics, are useful for testing data sequences for nonrandom arrangements in ordered successions of mutually exclusive states (Davis, 1973; Stevens, 1990). Markov chain analysis determines whether a series of data have the Markov property, the property that
155
A Markov Analysis of Tree Islands at Alpine Treeline
each successive data point is interdependent with the preceding data point. A series may be a sequence of data measured at uniform spatial or temporal intervals (firstorder Markov chains) or they may be a sequence of categorical transitions (embedded Markov chains). In this particular study, we are interested in analyzing the sequence of conifer species establishment in tree islands. Embedded and first-order Markov chains assessed spatial patterns of plant establishment (Figure 3). Embedded chains determined whether fine-scale patterns to conifer establishment are predictable. In many conifer establishment sites, especially those located on Lee Ridge, a sequence to plant establishment within a linear microsite would suggest that the establishment of a species is dependent upon the previous establishment of another species to improve its local environment. First-order Markov chains are used to characterize the geographic variability of conifer growth and ecotone stability among each study site. In order to use Markov chains in a geographic analysis, three types of data are needed: (a) a classification method that separates definable, individual states into time or space, (b) data to determine the transfer probabilities from one state to another, and (c) initial conditions. These conditions are met in the following ways. Mutually exclusive states are composed of individual conifer species, and the linear transects recorded such species information. Combined ordered species data
Embedded Markov chain Abies lasiocarpa
Transition 3
First-order Markov chain 70 65 60 55 50
Pinus contorta
45 Transition 2
40 35
Picea engelmannii
30 Transition 1
25 20
Wind direction
Pinus albicaulis
Boulder
15 10 5 cm Boulder
Figure 3 Conceptual difference between sampling for embedded and first-order Markov chains.
156
Lynn M. Resler and Mark A. Fonstad
determined the probabilities that one species follows another in space. Initial conditions are those conditions before the establishment of any conifer species in lee of a shelter. Because we had several sequences of observations from different locations within and among study sites, we calculated several transition chains for each location. Chi-square analyses ascertained whether the total combination of sequences composed of the linear vegetation data differed significantly from expected transitions from one state to another if the states were independent of each other (Davis, 1973). In the field, we recorded information on 14 species and/or states found in tree islands at alpine treeline. Therefore, the species data along lengthwise transects were classified into 14 mutually exclusive states, as listed in Table 1. A 14 14 transition frequency matrix generated from these data showed the number of times a species is succeeded by another species. A database contained the series of species measurements, with each species coded as a unique numerical category. Each data series began with an initial state, defined as whatever species occurs adjacent to the shelter. For example, if P. albicaulis was found growing in the immediate lee of a boulder, and this plant was followed by Abies lasiocarpa, the coding would be recorded as 3, 4. This type of coding reflects the change of one mutually exclusive state to another. Classified series included all transects with at least one transition. The total frequencies for all chains for each of the three study sites were added to form a total transition matrix representing the combination of species transitions of each chain, for each site. The counts for the total matrix became the input for the expected frequency matrix. Chi-square analysis allowed testing of the two matrices for the Markov property.
Table 1 Mutually exclusive states, including species and topographic characteristics, and associated coding for Markov analysis
States Small herbaceous Salix spp. Pinus albicaulis Abies lasiocarpa Picea engelmannii Pinus contorta Dead Pinus albicaulis Dead unknown Juniperus communis Psuedotsuga menzesii Juniperus horizontalis Larix lyallii Bare Rock
Code 1 2 3 4 5 6 7 8 9 10 11 12 13 14
A Markov Analysis of Tree Islands at Alpine Treeline
157
2.2. Embedded Markov chains Embedded Markov chains are those in which observations are taken only when there is a change in state from one species to another (Figure 3). In embedded Markov chains, transitions cannot reflect change from one state to itself. Therefore, this type of model is ideal for testing the significance of conifer establishment sequences within each patch. For embedded models, coding required chains that did not reflect length of the patches but only transitions between states. Therefore, the species transects were recorded as a succession of different species in space. Computation of embedded Markov chain models is different than that of firstorder Markov chains to compensate for the restriction of no self-transition (Stevens, 1990). Davis (1973) outlines the computation of embedded Markov chains. In this research, a 14 14 matrix is generated from observed species transitions based on the number and nature of transitions observed in the coded Excel Markov chains. The total number of possible transitions determined the 14 14 size of the matrix. From this matrix and its associated transition probabilities, an expected transition matrix was generated, based on the assumption that, for example, if species 6 is independent of species 2, then P(6|2) = P(6)P(2). The total of the column divided by the total number of transitions computed the occurrence probabilities of each transition. Once the two matrices are derived, a 2 test in the form of Equation (1) n X ðOi Ei Þ 2 2 ¼ ð1Þ Ei i¼1 that has (number of rows – 1)(number of columns – 1) degrees of freedom (132 in this study) will test the chain to see whether a state at one part is partially dependent, in a probabilistic sense, on the preceding state, and thus check for the Markov property. This test compared the observed transition matrix to the matrix that would be expected if successive states were independent. If there is a significant difference between the observed and the expected matrices, then we can assume that the chain has the Markov property (Davis, 1973) and therefore can conclude that the establishment of one species lacks independence (in a probabilistic sense) with the preceding state.
2.3. First-order Markov chains First-order Markov chains are those observations which occur at equal intervals in space or time (Figure 3). Therefore, first-order chains can reflect self-transitions. We coded first-order Markov chains in the same manner as embedded Markov chains and recorded observations at a regular interval rather than only at state transitions. Rather than testing the sequential establishment of species in a patch, first-order Markov chains served as an indicator of ecotone stability at the three study sites. A patch with several small individuals of different species may suggest the tendency of a patch to support multiple species and hence a level of dynamism that may not be present in a patch that has only one large patch of the same species. Absorbing states, those that show the tendency to remain in the same state, suggest a level of stability within a patch.
158
Lynn M. Resler and Mark A. Fonstad
A critical part of Markov chain analysis is correctly choosing the appropriate sampling interval (in this case, distance along a lengthwise transect). Selecting an interval too large may miss important within-patch variability, and selecting an interval too small may suspiciously enhance the Markov property. In this study, we chose a 5-cm interval at which to take observations. This interval allowed measurement of the smallest individual at a site.
3. Results 3.1. Spatial sequence of conifer establishment Results of the embedded Markov chain tests did not reveal a significant (p > 0.05) spatial sequence to conifer establishment in exposed conifer patches at any of the three study sites. However, some important species-specific patterns of conifer establishment within patches did emerge. Species order indicated that initial conifer establishment, or position of occupancy found immediately adjacent to the shelter source, was occupied in most patches by P. albicaulis (Table 2). Cataract Creek Basin was the only site where P. albicaulis did not reflect the highest percentage of initial occupancy (Table 3). Here, A. lasiocarpa was the dominant initial colonizer. The total percent occupancy is shown in Table 3.
3.2. Stability at treeline First-order Markov chains assessed the within-patch dynamics of conifer species at treeline. State stability may be thought of as the tendency for a species to transition into the same state. Patches that exhibit a high degree of self-similarity (fewer species present, or large amounts of growth) are more stable or have more
Table 2 Percent occupancy of initial position (position immediately adjacent to shelter source) in tree island transect sequences
Species Abies lasiocarpa Juniperus communis Juniperus horizontalis Picea engelmannii Pinus albicaulis Pinus contorta Pseudotsuga menziesii Dead
Percent (including dead conifers) n = 128
Percent (excluding dead conifers) n = 128
19 5 1 19 30 4 2
23 6 1 24 39 5 2
20
NA
Sample size reflects combined samples from each study site.
159
A Markov Analysis of Tree Islands at Alpine Treeline
Table 3 Percent initial position (position immediately adjacent to shelter source) occupancy for each study site, excluding dead conifers
Species
Percent, Lee Ridge n = 37
Percent, Cataract Creek Basin n = 53
Percent, Divide Peak n = 31
3 8 0 3 70 16 0
41 9 0 39 6 0 5
10 0 3 16 71 0 0
Abies lasiocarpa Juniperus communis Juniperus horizontalis Picea engelmannii Pinus albicaulis Pinus contorta Pseudotsuga menziesii
absorbing states. Those that show a high richness and transition numbers indicate dynamism within the establishment site through a tendency toward change. Geographic variability of sites may depend on a number of site conditions including wind, seed sources, and exposure. Results of the 2 test for the Markov property for first-order Markov chains for Lee Ridge (2 = 39,540.53), Cataract Creek Basin (2 = 1,024.42), and Divide Peak (2 = 11,428.29) were found to be significant at the p < 0.0001 level for 169 degrees of freedom. The results of this test indicate that when factoring in length of the transitions as a measure of growth of a species and allowing for self-transitions, transects at the three study sites do exhibit the Markov property. Self-transitions reflect the amount a species can grow before it is replaced by another species in a sequence. A high number of self-transitions may indicate that the establishment site has reached an absorbing state, and growth of one species dominates a patch’s dynamics. Therefore, a high number of transitions between states of species may indicate that the site is dynamic and is suitable for establishment by numerous species. Interspecies transitions are therefore an indicator of species richness. Among the three sites, Lee Ridge reports the highest number of transitions (104) from one state to another, indicating that its species composition is the highest and the site is the most dynamic of the three. Divide Peak showed 51 total interstate transitions, and Cataract Creek Basin showed only 29, indicating that a large portion of the patches have a very low species richness, and many large patches are occupied by one species only, in this case, A. lasiocarpa.
4. Discussion This study examines fine-scale attributes of alpine treeline, such as conifer establishment characteristics, species associations, and within-patch conifer growth, which may reflect coarse-scale ecotone dynamics in GNP. Based upon the results,
160
Lynn M. Resler and Mark A. Fonstad
three major findings of this study are as follows: (a) establishment order of conifers within a tree island is not statistically predictable, (b) P. albicaulis is an important initial treeline colonizer, and (c) there is geographic variability among study sites in terms of patch occupancy and treeline stability.
4.1. Establishment characteristics First, this study demonstrates that the establishment of conifers within a patch does not follow a predictable spatial sequence. For example, P. albicaulis, which has extremely hardy seedlings (Mellmann-Brown, 2005; Tomback et al., 2001) is not always followed in spatial sequence by the less hardy P. contorta. This finding suggests that the spatial order of conifer establishment in tree islands is based on stochastic events, such as seed dispersal in favorable microsites. This is particularly true for P. albicaulis seedlings, which are dispersed primarily by Clark’s nutcrackers. Nutcrackers often cache seeds in microsites near objects, such as at the base of rocks, trees, or logs (Tomback, 1978, 1998), which provide shelter against wind or shade (Mellmann-Brown, 2005). The finding that there is no predictable sequence to conifer establishment within a tree island also suggests that the operating processes within a conifer patch above treeline are not successional in nature to the extent that competitive exclusion is the predominant operative process. Rather, what is evident on the landscape in GNP is a number of conifers that exist in a small location that would normally not exist together under alternative conditions more amenable to growth. Dominant controlling processes may be more abiotic than biotic in areas of high stress (Callaway, 1995, 1998; Callaway and Walker, 1997; Callaway et al., 2002; Choler et al., 2001) and facilitation is important to overall treeline patterns, even if this facilitation is structural in nature (Smith et al., 2003). The physical presence of a plant or rock may expand local resources necessary for conifer growth (Grabherr et al., 1995), acting as an ecosystem engineer (Jones et al., 1994). Additionally, the facilitating influence of mycorrhiza to conifers at alpine treeline may complement microclimatic amelioration at the local scale (Hasselquist et al., 2005). 4.1.1. Initial establishment Conifer establishment in high-elevation and highly exposed sites at alpine treeline in GNP is often in the lee of periglacial shelters such as surface boulders or periglacial patterned ground (Butler et al., 2004; Resler et al., 2005). We found that the initial position next to a shelter source, that is, the position immediately adjacent to the shelter, is most frequently occupied by P. albicaulis. Again, this is a pattern explained by the selection of suitable caching sites of the Clark’s nutcrackers (Tomback, 1978, 1998). In addition, the microsites for initial tree island establishment must also be conducive to long-term survival of the tree (Resler et al., 2005). The hardiness of whitebark pine seedlings (Mellmann-Brown, 2005; Tomback et al., 2001) suggests that a whitebark pine may have a better ability to establish and survive in these microsites at treeline than other treeline conifer species.
A Markov Analysis of Tree Islands at Alpine Treeline
161
The initial position next to the shelter represents a very important position in terms of environment, niche, and positive feedbacks. The first conifer species to establish next to a shelter initiates a positive feedback cycle whereby the local environment is moderated to allow for subsequent establishment and survival. Positive feedback mechanisms have been shown to generate pattern in ecotones in general (Malanson, 1999; Wilson and Agnew, 1992) and specifically, vegetation patterns at the alpine (Alftine and Malanson, 2004; Bekker, 2005; Zeng and Malanson, 2006) and arctic (Svienbjo¨rnsson et al., 2002) treeline ecotone. Furthermore, conditions had to be appropriate for initial establishment to occur. The plant occupying the niche is indicative of environmental conditions at this particular site at the time it was established, and therefore, initial establishment of a conifer at treeline is species-specific due to varying environmental tolerances among species. The establishment of a hardy conifer such as P. albicaulis, for example, may enable the subsequent establishment of less resilient conifer species, resulting in increased conifer diversity at treeline. Knowledge about the species that occupies the first spatial position next to the shelter may be indicative of local conditions at the time of establishment. Initial occupancy by P. albicaulis and subsequent establishment by other species implies that (a) local microclimatic conditions improve following establishment by P. albicaulis and (b) that post-Little Ice Age (LIA) conditions were less favorable for conifer establishment than they are currently. The former observation indicates the importance of positive feedback mechanisms in the amelioration of local site quality. The latter is especially apparent at Lee Ridge, where establishment sites lower in elevation are typically richer in diversity and larger in size than highelevation sites where establishment sites often consist of a single individual (typically a spruce or a pine). Butler (1986) reported similar results in the Lemhi Mountains, Idaho. There, Pinus flexilis, an initial post-LIA colonizer in subalpine meadows, was replaced by other conifer colonizers as conditions became milder. Establishment of P. albicaulis at treeline has significant implications for the treeline environment of GNP. Despite current regeneration at highly exposed treeline locations, the long-term potential for P. albicaulis to contribute to the treeline and lower elevation ecosystems is uncertain given the mortality of whitebark pine at treeline due to blister rust, an invasive and introduced fungal pathogen (Resler and Tomback, 2008; Tomback and Resler, 2007). Given the potential importance of whitebark pine as a tree island initiator (Resler, 2004; Resler and Tomback, 2008; Tomback and Resler, 2007), the pattern of tree islands on the alpine treeline landscape in GNP could be altered, as well as conifer diversity and composition at alpine treeline.
4.2. Treeline stability The results of the first-order test, allowing for transitions from one state to itself, indicate that transects at the three study sites do exhibit the Markov property. Although we cannot assume anything about the sequential nature of establishment from these results, we may make inferences about conifer growth and patch diversity. Conifer diversity at treeline may be indicative of ecotone stability.
162
Lynn M. Resler and Mark A. Fonstad
Hacker and Gaines (1997) suggest that positive plant interactions can increase landscape biodiversity, resulting in increased ecotone stability. Growth of conifer species within each conifer patch number is indicated in the Markov chains as the number of self-transitions. Diagonal dominance on a transition matrix suggests relative stability among the states (Dale et al., 2002). It implies that the local conditions are right for its self-perpetuation and thus occupancy and dominance of the niche by one or few conifer species. Diagonal dominance also suggests competitive advantages; a conifer has a high amount of resources and may compete for the most space in a patch. A large patch with 100% foliar cover by one species alone suggests that while it may not be in equilibrium in terms of growth, it may be stable and less dynamic than one with five different species. Dynamism may tell us that there are numerous species that can thrive in a particular patch, implying a potential for increased treeline biodiversity. If biodiversity is an important indicator of ecosystem health (Tilman, 1999), then site stability as recorded by interspecies transitions suggests that there is important geographic variability among the three study sites in terms of ecosystem health. Where species richness and patch dynamism are high (such as at Lee Ridge), there may be greater potential for treeline advance than at sites with lower species richness (such as Cataract Creek Basin and Divide Peak). Low richness sites may be more susceptible to stochastic disturbance events that would disrupt conifer establishment and survival. Indicators of species richness, namely species richness counts and the number of interspecific transitions as shown in the Markov chains, determine that the lowest level of dynamics occur at Cataract Creek Basin. Species richness is 5, and transitions from one species to another are only 29 for the entire site. The transition transfer number indicates that the majority of patches found at Cataract Creek Basin are composed only of one species. Additionally, Cataract Creek Basin is the only site of the three where Pseudotsuga menziesii was established. This is likely due to the site’s location adjacent to the Continental Divide and the moister site conditions at Cataract Creek. The western portion of the park has a higher occurrence of P. menziesii than does the east, and therefore the seed source is likely from trees residing on the western ranges. Indicators of species richness, namely species richness counts and the number of interspecies transitions as shown in the Markov chains, demonstrate the dynamics intermediate between those at Cataract Creek Basin and Lee Ridge. Species richness is 5, and transitions from one species to another are 51 for the entire site. Divide Peak does not contain any unique species. Lee Ridge is the most dynamic treeline site of the three chosen for this study. Indicators of species richness include species richness counts (simply the number of different conifer species per site) and the number of transitions from one species to another, as shown in the results of the first-order Markov chains. Lee Ridge shows the highest conifer richness (with a total of seven conifer species). Additionally, Lee Ridge exhibits more than twice as much change from one species to another than does Cataract Creek Basin or Divide Peak. Lee Ridge is the only study site where Larix lyallii was recorded. Additionally, it is the only site among the three where P. contorta is found. A likely source for L. lyallii is Canada, as Lee Ridge lies only
A Markov Analysis of Tree Islands at Alpine Treeline
163
4 miles from the Canadian border. An even-age stand of lodgepole pine dominate the lower portions of the ridge below the subalpine zone. This study indicates geographic variability in patch diversity among the three study sites, all located on the eastern slopes of GNP. This finding suggests that varying fine-scale factors may result in different landscape scale treeline patterns in different places, depending upon how the geography of a site influences establishment, growth, and survival.
5. Conclusions Fine-scale factors of alpine treeline, specifically species-specific characteristics of tree islands such as growth, and conifer establishment patterns were examined using Markov chains. Results of the Markov analyses revealed that establishment order of conifers within tree islands is not statistically predictable, indicating that stochastic processes and modification of local climate in sheltered sites are important for establishment of conifers. P. albicaulis was found to be the conifer species most frequently located immediately adjacent to a shelter source – a function of both nutcracker caching and hardy seedlings. Fine-scale characteristics of conifer establishment in tree islands may be important in making links between local and landscape level dynamics, given the unique environmental tolerances of each conifer. It will be interesting to see how tree island composition in GNP responds as our environment continues to change.
REFERENCES Alftine, K.J., Malanson, G.P., 2004. Directional positive feedback and pattern at an alpine tree line. Journal of Vegetation Science 15, 3–12. Allen, T.R, Walsh, S.J., 1996. Spatial and compositional pattern of alpine treeline, Glacier National Park, Montana. Photogrammetric Engineering and Remote Sensing 62, 1261–1268. Bekker, M.F., 2005. Positive feedback between tree establishment and patterns of subalpine forest advancement, Glacier National Park, Montana, U.S.A. Arctic, Antarctic, and Alpine Research 37, 97–107. Benedict, J.B., 1984. Rates of tree-island migration, Colorado Rocky Mountains, USA. Ecology 65, 820–823. Brown, D.C., 1994. Comparison of vegetation-topography relationships at the alpine treeline. Physical Geography 15, 125–145. Butler, D.R., 1986. Conifer invasion of subalpine meadows, central Lemhi Mountains, Idaho. Northwest Science 60, 166–173. Butler, D.R., Walsh, S.J., 1994. Site characteristics of debris flows and their relationship to alpine treeline. Physical Geography 15, 181–199. Butler, D.R., Malanson, G.P., Resler, L.M., 2004. Turf-banked terrace treads and risers, turf exfoliation, and possible relationships with advancing treeline. Catena 58, 259–274. Callaway, R.M., 1995. Positive interactions among plants. Botanical Review 61, 306–349. Callaway, R.M., 1998. Competition and facilitation on elevation gradients in subalpine forests of the northern Rocky Mountains, USA. Oikos 82, 561–573. Callaway, R.M., Brooker, R.W., Choler, P., Kikvidze, Z., Lortie, C.J., Michalet, R., et al., 2002. Positive interactions among alpine plants increase with stress. Nature 417, 844–848.
164
Lynn M. Resler and Mark A. Fonstad
Callaway, R.M., Walker, L.R., 1997. Competition and facilitation: Asynthetic approach to interactions in plant communities. Ecology 78, 1958–1965. Choler, P., Michalet, R., Callaway, R.M., 2001. Facilitation and competition on gradients in alpine plant communities. Ecology 82, 3295–3308. Dale, M., Dale, P., Edgoose, T., 2002. Using Markov models to incorporate serial dependence in studies of vegetation change. Acta Oecologia 23, 261–269. Daubenmire, R.F., 1954. Alpine timberlines in the Americas and their interpretation. Butler University Botanical Studies 11, 119–136. Davis, J.C., 1973. Statistics and Data Analysis in Geology. John Wiley, New York. DeAngelis, D.L., Post, W.M., Travis, C.C., 1986. Positive Feedback in Natural Systems. Springer, New York. Finklin, A.I., 1986. A Climatic Handbook for Glacier National Park with Data for Waterton Lakes National Park, Intermountain Research Station, U.S.D.A. Forest Service, Ogden, Utah. Germino, M.J., Smith, W.K., Resor, A.C., 2002. Conifer seedling distribution and survival in an alpine-treeline ecotone. Plant Ecology 162, 157–168. Grabherr, G., Gottfried, M., Gruber, A., Pauli, H., 1995. Patterns and current changes in alpine plant diversity. In: Chapin, F.S., Ko¨rner, C. (Eds.), Arctic and Alpine Biodiversity. Springer Verlag, Berlin, pp. 161–181. Hacker, S.D., Gaines, S., 1997. Some implications of direct positive interactions for community species diversity. Ecology 78, 1990–2003. Hasselquist, N., Germino, M.J., McGonigle, T., Smith, W.K., 2005. Variability of Cenococcum colonization and its ecophysiological significance for young conifers at alpine-treeline. New Phytologist 165, 867–873. Hattenschwiler, S., Smith, W.K., 1999. Seedling occurrence in alpine treeline conifers: A case study from the central Rocky Mountains, USA. Acta Oecologica 20, 219–224. Jones, C.G., Lawton, J.H., Shachak, M., 1994. Organisms as ecosystem engineers. Oikos 69, 373–386. Ko¨rner, C., 1999. Alpine Plant Life: Functional Plant Ecology of High Mountain Ecosystems. Springer, Berlin. Lamb, J., Jr., Chapman, J.E., 1943. Effect of surface stones on erosion, evaporation, soil temperature, and soil moisture. Journal of the American Society of Agronomy 35, 567–578. Malanson, G.P., 1999. Considering complexity. Annals, Association of American Geographers 89, 746–753. Marr, J.W., 1977. The development and movement of tree islands near the upper limit of tree growth in the southern Rocky Mountains. Ecology 58, 1159–1164. Mellmann-Brown, S., 2005. Regeneration of whitebark pine in the timberline ecotone of the Beartooth Plateau, U.S.A.: Spatial distribution and responsible agents. In: Broll, G., Keplin, B. (Eds.), Mountain Ecosystems: Studies in Treeline Ecology. Springer, Berlin, 97–115. Mueller-Dombois, D., Eldenberg, H., 1974. Aims and Methods of Vegetation Ecology. Wiley, New York. Pe´rez, F.L., 1987. Soil moisture and the upper altitudinal limit of giant paramo rosettes. Journal of Biogeography 14, 173–186. Pe´rez, F.L., 1991. Soil moisture and the distribution of giant Andean rosettes on talus slopes of a desert paramo. Climate Research 1, 217–231. Resler, L.M., 2004. Conifer Establishment Sites at the Alpine Treeline Ecotone, Glacier National Park, Montana. Doctoral Dissertation, Department of Geography, Texas State University-San Marcos. Resler, L.M., Butler, D.R., Malanson, G.P., 2005. Topographic shelter and conifer establishment and mortality in an alpine environment, Glacier National Park, Montana. Physical Geography 26, 112–125. Resler, L.M., Tomback, D.F., 2008. Blister rust prevalence in krummholz whitebark pine: Implications for treeline dynamics, northern Rocky Mountains, Montana, USA. Arctic, Antarctic, and Alpine Research 40, 161–170. Smith, W.K., Germino, M.J., Hancock, T.E., Johnson, D.M., 2003. Another perspective on altitudinal limits of alpine timberlines. Tree Physiology 23, 1101–1112.
A Markov Analysis of Tree Islands at Alpine Treeline
165
Stevens, R.L., 1990. Markov-chain analysis as a pedagogic tool. Journal of Geological Education 38, 288–293. Svienbjo¨rnsson, B., Hofgaard, A., Lloyd, A., 2002. Natural causes of the tundra-taiga boundary. Ambio 30, 23–29. Tilman, D., 1999. The ecological consequences of changes in biodiversity: A search for general principles. Ecology 80, 1455–1474. Tomback, D.F., 1978. Foraging strategies of Clark’s nutcracker. Living Bird 16, 123–161. Tomback, D.F., 1998. Clark’s nutcracker (Nucifraga columbiana). In: Poole, A., Gill, F. (Eds.), The Birds of North America. The Birds of North America, Inc., Philadelphia, No. 331. Tomback, D.F., Anderies, A.J., Carsey, K.S., Powell, M.L., Mellmann-Brown, S., 2001. Delayed seed germination in whitebark pine and regeneration patterns following the Yellowstone fires. Ecology 82, 2587–2600. Tomback, D.F., Resler, L.M., 2007. Invasive pathogens at alpine treeline: Consequences for treeline dynamics. Physical Geography 28, 397–418. Tranquillini, W., 1979. Physiological Ecology of the Alpine Timberline. Springer-Verlag, New York. Wilson, J.B., Agnew, A.D.Q., 1992. Positive-feedback switches in plant communities. Advances in Ecological Research 23, 263–336. Zeng, Y., Malanson, G.P., 2006. Endogenous fractal dynamics at alpine treeline ecotones. Geographical Analysis 38, 271–287.
C H A P T E R
9
Modeling Feedback Effects on Linear Patterns of Subalpine Forest Advancement Matthew F. Bekker and George P. Malanson
Contents 1. Introduction 1.1. Treeline patterns 1.2. Explanations of treeline pattern 1.3. Modeling treeline location and dynamics 2. Methods 2.1. FORSKA 2.2. Parameterization 2.3. Modifications 2.4. Site quality 2.5. Simulations 3. Results 4. Discussion 4.1. Gap models and treeline environments 4.2. Effects of light and mortality 5. Conclusions Acknowledgments References
167 168 169 170 172 172 172 175 176 178 180 182 182 183 185 187 187
1. Introduction The development and maintenance of several types of visually striking vegetation patterns are controlled by positive feedback between ecological pattern and process (Rietkerk and van de Koppel, 2008). For example, interactions between vegetation patterns and climatic, hydrologic, and geomorphic processes produce ‘‘fir waves’’ (Sato and Iwasa, 1993; Sprugel, 1976) and ‘‘ribbon forest’’ in subalpine environments (Billings, 1969; Butler et al., 2003; Holtmeier, 1982) and ‘‘tiger bush’’ in arid and semiarid environments (Tongway et al., 2001; White, 1969). These interactions are particularly visible at ecotones, where the influence of positive Developments in Earth Surface Processes, Volume 12 ISSN 0928-2025, DOI 10.1016/S0928-2025(08)00209-5
2009 Elsevier B.V. All rights reserved.
167
168
Matthew F. Bekker and George P. Malanson
feedback may affect the position, pattern, and dynamics of the boundary between the adjacent communities. Wilson and Agnew (1992) discussed the potential for positive biotic feedback, where plants modify the environment they experience, to produce sharp boundaries on gradual abiotic gradients, to maintain vegetation mosaics, or to amplify or reduce the rate of vegetation change at ecotones. The potential for feedback to accelerate vegetation change has important implications for the sensitivity of ecotones to climate change (Bader et al., 2007; Malanson et al., 2007). Because ecotones often represent the distributional limits of the communities they divide, it has been suggested that they might be particularly responsive to climate change (Neilson, 1993; Risser, 1995). Yet, others have suggested that the rate of movement of the alpine treeline would be too slow to be useful as a climate change indicator (Cairns and Malanson, 1997; Holtmeier, 2003; Kupfer and Cairns, 1996; Noble, 1993). If positive feedbacks that would accelerate vegetation change are present, however, the ecotone would respond more rapidly. The goal of this research was to examine the effects of positive feedbacks between ecotone pattern and tree establishment using a computer simulation model validated against a dendrochronological reconstruction (Bekker, 2005) of subalpine forest advancement on Lee Ridge in Glacier National Park (GNP), Montana. We used a hybrid simulation that combines a physiologically mechanistic model with a model of forest-stand dynamics in order to represent (a) multispecies interactions between spatial pattern and process; (b) mechanistic responses of tree growth to climate; and (c) changing environmental and vegetative conditions. We hypothesized that the simulation could produce the same pattern and dynamics of post-Little Ice Age forest advancement observed on Lee Ridge if positive feedbacks were included, that positive feedback can accelerate vegetation change at the alpine treeline ecotone, and that feedback in ecotone dynamics on Lee Ridge has remained important under changing climatic conditions.
1.1. Treeline patterns Most alpine treeline ecotones in the Rocky Mountains represent a gradual upslope transition from closed subalpine forest to patches of dwarfed trees, to extremely deformed, low-growing patches of krummholz, and finally tundra vegetation (Allen and Walsh, 1996; Baker et al., 1995; Hansen-Bristow and Ives, 1985). Such gradual transitions are common for ecotones generally as well (Gosz, 1993). In some extremely windswept treeline locations, the patches or ‘‘islands’’ of krummholz extend into linear strips, oriented parallel to the prevailing wind (Marr, 1977). Krummholz islands usually consist of subalpine fir (Abies lasiocarpa) and Engelmann spruce (Picea Engelmannii), which expand leeward by layering, a process in which branches are buried and generate new roots. Thus, an entire hedge or island may consist of only one individual. Slightly larger linear strips composed of both krummholz and dwarfed trees, termed ‘‘hedges,’’ have also been described (Holtmeier, 1982). An interesting characteristic of both krummholz and dwarfed trees is that they have been shown to respond to climate change by changing their growth form to taller, more upright forms (Hessl and Baker, 1997; Kullman, 1986;
Modeling Feedback Effects on Linear Patterns of Subalpine Forest Advancement
169
Payette et al., 1985; Weisberg and Baker, 1995) when climatic conditions improve. Such changes in growth form can also produce a positive feedback, as the taller branches reduce wind speed and improve conditions for the patch and leeward areas (Earle, 1993; Hessl and Baker, 1997). The ecotone on Lee Ridge contains linear patterns of forest that differ from krummholz islands and hedges in three respects. First, while krummholz and hedges are usually found several meters upslope from the closed forest, on Lee Ridge the linear strips extend like ‘‘fingers’’ from closed forest into the tundra (Figure 1). Second, these fingers consist primarily of lodgepole pine (Pinus contorta var. latifolia) and contain primarily tall, upright, separate individuals that have been established from seed. Some of the larger pines have a pronounced upslope basal sweep, growing along the ground in the direction of the prevailing wind for as much as a few meters before turning upright. Some also exhibit a ‘‘krummholz base’’ (cf. Earle, 1993), characterized by an array of dead branches near the base of the tree. These observations suggest that forests at this site formerly contained a mix of krummholz and dwarfed trees, which have been able to change their growth form as the forest has advanced into the tundra.
1.2. Explanations of treeline pattern Stevens and Fox (1991) suggested that treeline was largely a phenomenon expressing the carbon balance of trees. This hypothesis suggests that the position of the treeline corresponds to the point at which a multitude of environmental factors
Study site
Figure 1 Lee Ridge and example of fingers (inset) near the study site. Photo D. Cairns, inset G. Malanson.
170
Matthew F. Bekker and George P. Malanson
cause a tree to produce less photosynthate than is lost through respiration. Stevens and Fox (1991) proposed two additional theories to explain treeline position and patterns. The resource-averaging hypothesis assumes that resources are heterogeneously distributed as they gradually decline along a gradient. In this case, the resources represent a fine-grained pattern for a large tree, but a coarse-grained pattern for a smaller plant. Thus, larger plants such as trees will always receive an average amount of resources, whereas smaller plants may receive high amounts. Moreover, as resources continue to decline along the gradient, trees will not be able to obtain the resources they need to survive, giving smaller plants a competitive advantage. The apical dominance hypothesis addresses the tendency for trees to grow taller rather than spreading laterally to dominate a site quickly. At many cold treeline sites, trees begin to reproduce through layering due to harsher conditions with increasing proximity to the treeline. This hypothesis suggests that apical dominance in trees ultimately sets a limit to their ability to tolerate the loss of their terminal buds, which corresponds to the limit of upright trees. Positive feedbacks may compound or diminish the effects of other biotic or abiotic processes affecting the location and pattern of treeline ecotones. The consistent windward to leeward alignment of the fingers on Lee Ridge suggests the importance of directional, wind-mediated switches in producing the pattern. The deposition of wind-blown snow has been shown to be a particularly important control on treeline patterns and dynamics (Minnich, 1984) in a feedback loop (Hiemstra et al., 2002, 2006). Walsh et al. (1994) found that an intermediate amount of snow is optimal for tree growth in GNP. Intermediate snow cover protects plants from damaging winds and airborne particles (Frey, 1983; Hadley and Smith, 1986), provides moisture after peak stream flow in the spring and summer (Berg, 1987), and increases input and availability of soil nutrients (Bowman, 1992; Williams et al., 1998). Ko¨rner (1998) suggested that negative feedbacks are also present at treeline. He argued that trees are at a disadvantage relative to shrubs and forbs due to their tendency to raise their canopies into cold air and to lower root zone temperatures during the growing season through shading. This hypothesis is also in contrast to the carbon balance hypothesis, contending that cold temperatures inhibit carbon sinks, or growth per se, rather than reducing carbon production.
1.3. Modeling treeline location and dynamics Computer simulation models that have been used to investigate pattern–process interactions at treeline may be separated into three general classes. The first class includes the JABOWA-FORET (Botkin et al., 1972; Shugart and West, 1977) family of stand or ‘‘gap’’ models, originally designed to simulate succession after the creation of forest-canopy gaps. These models simulate the establishment, growth, and mortality of trees phenomenologically; the processes are related directly to environmental parameters or state variables by empirical equations, without incorporating explicitly the underlying tree physiology. Physiological models, in contrast, simulate plant-level processes in response to climate and hydrology more mechanistically, in that they only use equations from below the level of interest. For example, carbon balance is simulated mechanistically as interactions of
Modeling Feedback Effects on Linear Patterns of Subalpine Forest Advancement
171
photosynthesis, respiration, transpiration, and carbon allocation, but each of these lower level processes is modeled phenomenologically. The third class, cellular automata, tends to examine general spatial relations based on the effects of neighbors (Alftine and Malanson, 2004; Zeng and Malanson, 2006; Zhang et al., 2008). Cairns and Malanson (1997) used the physiological model ATE-BGC to test the carbon balance hypothesis at the alpine treeline in GNP. They found that carbon balance is in dynamic equilibrium with climate, but the treeline is not. The predicted locations of the treeline based on the carbon balance hypothesis varied over the simulation, when in fact the treeline in the park has been stable for the past century (Butler et al., 1994). Therefore, Cairns and Malanson (1997) concluded that the carbon balance hypothesis is useful for predicting the potential, but not actual, location of the treeline. Using a spatially explicit version of the JABOWA-FORET model, Malanson (1997) simulated the dynamics of an ecotone as might occur on a mountain treeline, incorporating a positive feedback represented by an increase in the site quality of a given cell if neighboring cells were occupied. He found waves of mortality running through the forest toward the treeline. Because the mortality of individual trees was simulated as a function of their age, size, and current growth, mature trees in any one row tended to die at the same time. This row of dead trees caused a reduction in site quality in the adjacent row, causing many of those trees to die, and so on for subsequent rows. When the positive feedback was removed from the simulations, the moving wave of mortality did not occur. In a simulation test of the resource-averaging hypothesis, Malanson et al. (2001) found that treeline patterns produced by the model closely corresponded to variations in underlying resources but did not match vegetation patterns observed at treeline ecotones. They suggested that biotic feedbacks, perhaps with a directional component, might be necessary to produce observed patterns. Alftine and Malanson (2004) found that spatial patterns of the upper portion of Lee Ridge could be approximated in a cellular automaton if directional positive feedback was included. Recently, several researchers have argued for the integration of different types of models to provide a better understanding of vegetation dynamics (Bossel, 1991; Friend et al., 1993; Wild and Winkler, 2008). For example, Brown et al. (1994) used a combination of regional- to landscape-scale empirical models and plant- to stand-scale physiological models to assess the sensitivity of alpine treeline to climate change. They argued that a more complete hierarchy of models, including a stand model between the empirical and physiological models, would provide a better understanding of the multiscale influences on vegetation patterns at treeline ecotones. Bekker et al. (2001) combined a physiological model with a stand model to investigate the levels of feedback needed to maintain krummholz vegetation. Very high levels of feedback, which would produce sharp boundaries on smooth environmental gradients (Malanson, 1997), were needed to match observed leaf area patterns. They concluded that feedback was necessary to maintain krummholz but was not sufficient alone to produce the complex patterns observed at treeline. These patterns were a result of either directional feedbacks or a combination of pattern in the abiotic environment and feedback.
172
Matthew F. Bekker and George P. Malanson
2. Methods Lee Ridge is located in the northeast corner of GNP. It is a north-south trending spur that slopes gradually downward to the west and north from the base of Gable Mountain, thus creating a bidirectional elevation gradient (Figure 1). Data from a weather station established by the US Geological Survey (USGS) – Biological Resources Division on Lee Ridge indicate that the ridge is dominated by strong westerly winds year round and is located in one of the driest areas in the park.
2.1. FORSKA Because our research question involves multispecies dynamics, we used a gap model as the principal vehicle for the simulations. Since the introduction of the first gap model, many variants have been produced to expand the range of forest types that could be modeled or to simulate transient responses of forests to climate change. These variants maintain the core of the model, simulating the basic processes in the same way as the original. The gap model FORSKA (Leemans, 1989), however, was developed with improved, more mechanistic growth functions and was altered to represent the crowns of conifers instead of deciduous trees. The details of how growth is simulated can be found in Leemans and Prentice (1989). Here we briefly discuss aspects of the model that are necessary to describe how we incorporated the effects of positive feedback. Tree establishment in the current version of FORSKA (Prentice et al., 1993) is determined by the sapling establishment rate parameter (E0 ), de-rated by two factors: (a) a light multiplier, which represents the proportion of light reaching the forest floor relative to maximum light intensity and (b) a suite of climatic multipliers. Both the light and climatic multipliers range from 0 to 1. The actual number of saplings that appear in a given timestep is determined from a Poisson distribution, representing natural variance in establishment from year to year. The primary growth equation is d 1 Wtot H 2 ð1Þ ðD HÞ ¼ B SL ð Pz zÞdz f ðenvironmentÞ Wmax dt where D is DBH, H is tree height, W is biomass, B is bole height, SL is the vertical density of leaf area, Pz is the proportion of maximum possible annual assimilation achieved by leaves at depth z in the canopy, is the species-specific growth-scaling factor, represents the cost of maintaining sapwood with increasing size, and f (environment) represents the same climatic multipliers as in the establishment routine.
2.2. Parameterization Patch size is one of the most critical parameters affecting the behavior of gap models. If patch size approximates the maximum crown area of the largest trees, then a single tree may dominate the patch, and gap-phase successional dynamics are
Modeling Feedback Effects on Linear Patterns of Subalpine Forest Advancement
173
initiated when the dominant tree dies (Shugart and West, 1979). We set the patch size to 0.003 ha (30 m2) to facilitate easy comparison with field data, which were collected in 0.003-ha segments along the fingers. Moreover, this size is small enough that a single tree could affect the microenvironmental conditions of its own and neighboring patches, but not completely dominate a single patch. Most estimates of parameters for site (Table 1) and species (Table 2) were obtained from published silvicultural, allometric, or modeling studies. Pinus flexilis (limber pine) was not included in the simulations because it represented only 1% of trees at the site. Some species parameters were not available in the literature, but a sensitivity analysis (Leemans, 1991) found that none of these had more than a slight effect on the model. Therefore, values of these parameters were set to those used in the original version of FORSKA to simulate spruce-pine forests in Sweden (Prentice and Leemans, 1990). Of these, only the sapwood maintenance cost factor is species-specific; the values used for Scots pine (Pinus sylvestris) were assigned to lodgepole pine, and those for Norway spruce (Picea abies) were assigned to Engelmann spruce and subalpine fir. All other parameters were obtained from field data. However, data collected within the study site were not used to estimate parameters in order to provide a more robust validation of the model results. The initial slope of diameter : height was calculated using regression analyses of extensive data collected in GNP (Jensen et al., 1993). Average light intensity was obtained from a meteorological field station established by the USGS-BRD on Lee Ridge. Finally, the sapling establishment rate was determined from field data collected approximately 1–2 km upslope from the study area, above the current limit of closed forest (Alftine et al., 2003). This is consistent with the hypothesis that our study site was once above treeline but has been since filled in with trees.
Table 1 Site parameters in FORSKA
Parameter name
Value
Source
Average light intensity (mmol/m2/s) Light extinction coefficient
500
Field data
0.5
Biomass conversion factor (kg/cm2/m) Maximum biomass (mg/ha) Canopy vertical integration step (m) Mortality curve steepness parameter Initial DBH for saplings (cm)
0.035
Friend et al. (1993); Keane et al. (1996) Brown (1978)
600 0.5
Leemans (1991) Leemans (1991)
999.0
Leemans (1991)
1.0
Leemans (1991)
174
Table 2 Species parameters in FORSKA
Pico
Pcen
Abla
Source
Maximum height (m)
32.0
49.0
40.0
Initial diameter: height slope Light compensation point (mmol/ m2/s) Half-saturation point (mmol/m2/s) Growth scaling const. (cm2/m/ year) Sapling establishment rate (ha1/ year) Initial leaf area: D2 ratio (m2/cm) Threshold for index of vigor Intrinsic mortality rate (year1) Suppressed mortality rate (year1) Sapwood turnover rate (year1) Sapwood maint. cost (cm2/m2/ year)
0.93 29.2
0.66 21.2
0.59 21.2
Nikolov and Helmisaari (1992); Burns and Honkala (1992) Field data (Jensen et al., 1993) Keane et al. (1996)
530.9 89.0
378.4 100
435.4 150
Keane et al. (1996) Nikolov and Helmisaari, (1992); Urban et al. (1993)
21.0
5.0
1.0
Field data
0.137 0.025 0.0046 0.46 0.004 0.2
0.379 0.025 0.0046 0.46 0.004 0.05
0.343 0.025 0.0046 0.46 0.004 0.05
Kaufmann et al. (1982) Leemans (1991) Leemans (1991) Leemans (1991) Leemans (1991) Leemans (1991)
Pico, lodgepole pine; Pcen, Engelmann spruce; Abla, subalpine fir.
Matthew F. Bekker and George P. Malanson
Parameter name
Modeling Feedback Effects on Linear Patterns of Subalpine Forest Advancement
175
2.3. Modifications Two principal modifications to FORSKA were required for this research. First, we altered the model to run on a grid of cells (each representing a patch in the original FORSKA model) that can interact with neighboring cells through positive feedback. In each timestep, the model examines the biomass of the eight cells immediately surrounding the cell of interest and increases or decreases site quality according to the following equation (Malanson, 1997): dSB Q¼ ð2Þ Bmax where S is feedback strength, d is a parameter to alter feedback in different directions, B is biomass in kg/ha, and Bmax is maximum biomass on a cell in kg/ha. Values for the parameter d were based on the assumption that existing trees increase snow deposition on their leeward side due to reduced wind speeds, thus providing protection from wind, increased moisture, and increased input and availability of soil nutrients to downwind trees. Observations and snow surveys on Lee Ridge show that snow depth is positively correlated with tree cover and that snow is concentrated around trees, with plume-shaped drifts extending several feet leeward of tree groups (Geddes et al., 2005). The surveys, however, contained insufficient quantitative detail to determine feedback directly. The actual values for d were based on Scott et al. (1993), who provided detailed measurements of wind speed and snow depth around ‘‘tree islands’’ at the arctic treeline in Canada. They found that wind speeds were reduced to 40% of ambient levels immediately downwind of the islands. Snow depth was approximately four times greater within the islands compared to treeless areas, and a snowdrift extended several meters leeward, with a peak depth 2.5 times greater than within the islands. In addition to these areas of reduced wind speeds and higher snow deposition, they also measured higher wind speeds upwind and around the sides of the trees near the ground, which created a horseshoe-shaped trough in the snow in these areas. These measurements and observations are consistent with the snow surveys and observations on Lee Ridge, as well as descriptions from other arctic (Payette and Filion, 1985) and alpine (Camarero et al., 2000) environments. We used these observations to set the relative proportions of the values of d (with the sum of all values = 1), according to the location of occupied cells with respect to a focal cell (Figure 2). Thus, if the focal
0.0
0.0
0.0
–0.2
0.4
–0.2
0.0
1.0
0.0
Figure 2 Values for the parameter d in the feedback equation. Each value represents the relative influence of that cell being occupied on feedback to the focal (center) cell.
176
Matthew F. Bekker and George P. Malanson
cell were occupied, its site quality would be increased (with the specific amount dependent on biomass and feedback strength), but the increase would be 2.5 times greater if the windward cell were occupied. Conversely, if the horizontally adjacent cells were occupied, the focal cell would have its site quality reduced. The second alteration to FORSKA involved the f (environment) portion of the growth equation. This term relates climate parameters to growth phenomenologically, without explicitly considering underlying physiology. Thus, leaf and cell processes are represented collectively as a black box. Because the results of interactions between these processes in response to climate are not predictable, it was necessary to use a physiologically mechanistic model to simulate these interactions and produce a more general measure of site quality at a level of detail compatible with the gap-modeling framework.
2.4. Site quality ATE-BGC is a version of FOREST-BGC (Running and Coughlan, 1988) that was modified to incorporate alpine treeline-specific conditions and processes including light attenuation through a krummholz canopy and the physiological effects of low temperatures and desiccation (Cairns, 1994, 2005). The model predicts carbon balance based on photosynthetic inputs and carbon outflows due to respiration and tissue loss and produces several measures of tree growth that could be used to represent site quality. For this research, a measure of potential growing conditions was needed, which would set a base level for the processes of establishment and growth within FORSKA. Thus, we chose gross primary productivity (GPP) as the measure of site quality. FOREST-BGC and ATE-BGC have successfully simulated growth in upright forests in subalpine environments (Running and Coughlan, 1988) and krummholz patches in treeline environments (Cairns, 1998; Cairns and Malanson, 1997, 1998), respectively. However, our research question required a model that could simulate a transition in site quality from a treeline environment with krummholz to a subalpine environment with upright trees. To represent this transition, we ran the fully modified ATE-BGC model, and then removed the effects of the alpine treeline modifications in four even steps, so that the final step represented the original FOREST-BGC model (Table 3). The effects of the adjusted parameters and constants are as follows: (a) The reduction of the TLAI-PLAI (Total Leaf Area Index – Projected Leaf Area Index) conversion factor from 3.57 to its original value of 2.2 represents the gradual change in light attenuation through the canopy as vegetation changes from krummholz to upright trees; (b) the reduction of the winter injury multiplier to 0 represents the reduction and eventual removal of carbon loss due to desiccation under improving climatic conditions; (c) the increase and convergence of the f (mat) values toward 1 represents gradual increases in mesophyll conductance values as a result of the reduced impact of hard frosts and lethal freezes in a less harsh environment; and (d) the reduction of SPmax to 0 effectively increases the growing season, as plants will reach photosynthetic maturation earlier in the spring.
Altered parameters and constants in ATE-BGC and FOREST-BGC
Model parameters and constants TLAI-PLAI conversion factor Winter injury multiplier Photosynthetic maturation (f(mat)) After 1 hard frost After 2 hard frosts After 3 hard frosts After lethal freeze Complete maturation level (SPmax) Site parameters Elevation (m) Soil depth (cm) Orange LAI (%) Site type
ATE-BGC
Step1
Step2
Step3
Step 4 (FOREST-BGC)
3.57 1.0
3.23 0.75
2.89 0.5
2.55 0.25
2.20 0.0
0.35 0.25 0.20 0 1000
.51 .44 .40 0.25 750
.68 .63 .60 0.5 500
.84 .81 .80 0.75 250
1 1 1 1 0
2153 11.4 7.4 1
2110 12.4 5.6 2
2067 13.4 3.7 2
2024 14.4 1.9 3
1981 15.4 0.0 3
Modeling Feedback Effects on Linear Patterns of Subalpine Forest Advancement
Table 3
See text for explanation.
177
178
Matthew F. Bekker and George P. Malanson
In addition to these gradual changes in model parameters and constants, we also used different site parameters to represent the transition from an environment supporting krummholz to one containing upright forest (Table 3). We used parameters from a location above the current treeline on Lee Ridge to represent krummholz conditions and gradually changed the values to those representing our study site for the upright forest conditions. These changes thus involve a space-for-time substitution representing changes in climate and vegetation. The values for the site parameters elevation and soil depth were measured directly at the high-elevation site and the study site and extrapolated evenly to represent the transition between krummholz and upright forests. The orange LAI parameter indicates the amount of desiccated foliage, represented as a percentage of total LAI, measured at 7.4% for high-elevation sites on Lee Ridge and assumed to be 0 for upright forests. The site type parameter denotes tree form (type 1 = krummholz, type 2 = dwarfed, type 3 = upright), with associated diameter, height, and density values.
2.5. Simulations We ran each of the five resulting variants of the model with different LAI input levels to produce a grid of GPP values (Table 4), representing a transition of potential growing conditions from a treeline-krummholz site to a subalpineupright forest site. This grid was then used by FORSKA as a look-up table to determine site quality for each timestep. During a FORSKA run, feedback increases the site quality of any cell in which biomass has increased in that cell or its ‘‘windward’’ neighbor and decreases the site quality of those with neighbors perpendicular to the direction of the wind (Figure 2). At the end of each timestep (1 year), FORSKA calculates LAI and average tree height on each cell, using those values to determine site quality for the next timestep by locating the corresponding GPP value on the look-up table. The decision to use tree height to determine location on the y-axis of the table was based on the observation that trees gradually decrease in height with increasing elevation at many alpine treeline ecotones (Baker et al., 1995; Hansen-Bristow and Ives, 1985). The gradual changes in model parameters and LAI values used to create the look-up table do not imply an assumption of smooth changes in climate and vegetation over time. The objective was to provide a smooth gradient of potential conditions on which FORSKA could determine (through feedback) how abrupt or gradual the changes would be. In addition, using four steps to alter the BGC parameters produced differences in GPP values from cell to cell (by row) that were similar (0.0–0.02) to those produced by the different LAI inputs (by column). Thus, more steps would probably not significantly reduce the differences between the GPP values, while fewer steps would likely produce larger differences, making the gradient less smooth. For all simulation runs, we set the initial site quality to 0.0, representing a harsh site above treeline (LAI = 1 and M = 4 in Table 3). However, because site quality at this level would not allow establishment to occur in FORSKA, we initialized the model with 10 randomly placed 10-year old lodgepole pine
M4 M3 M2 M1 M0
0.00 0.00 0.00 0.03 0.04 1
0.04 0.05 0.06 0.06 0.08 2
0.06 0.07 0.09 0.09 0.11 3
0.07 0.09 0.10 0.11 0.13 4
0.08 0.09 0.11 0.12 0.14 5
0.09 0.10 0.12 0.13 0.15 6 LAI
0.09 0.10 0.12 0.14 0.16 7
0.09 0.11 0.13 0.14 0.16 8
0.10 0.11 0.13 0.14 0.17 9
0.100 0.111 0.133 0.145 0.177 10
0.10 0.11 0.13 0.15 0.18 11
0.10 0.11 0.13 0.15 0.18 12
The values are simulated from varying Leaf Area Index (LAI) input values and five levels of modification to the FOREST-BGC model representing a transition from treeline conditions to upright forest conditions (M0 = original FOREST-BGC model, upright forest conditions, and M4 = ATE-BGC, treeline conditions). Mean tree height and LAI are calculated by FORSKA and used to determine the corresponding site quality on the look-up table for the next timestep.
Modeling Feedback Effects on Linear Patterns of Subalpine Forest Advancement
Table 4 Gross primary productivity and mean tree height values used to determine site quality for a given cell in FORSKA
179
180
Matthew F. Bekker and George P. Malanson
saplings, representing trees that established in ‘‘safe sites.’’ The decision to use lodgepole pine saplings to initialize the model was based on its characterization as a pioneer successional species. The model was run for 250 years, based on field observations of the oldest trees. Because the dendrochronological data collected at our study site were not used to parameterize FORSKA, they provide an independent set of observations to allow validation of the simulations. However, the soil depth parameter was used in the BGC simulations, and the sapling establishment rate and light intensity parameters were determined from nearby sites on Lee Ridge, so the results may still only be applicable to forests at this latitude, elevation, and slope aspect.
3. Results The pattern of advancement in 50-year increments for an average run is shown in Figure 3. Two strips of occupied cells gradually developed over the 250-year simulation period, with advancement taking place only immediately upslope of occupied cells. When all trees died on a given cell, trees in the adjacent upslope cell usually died soon thereafter. These two processes created a general pattern of older-to-younger trees from the bottom to the top of the grid, with the pattern repeating in places along each strip (particularly evident in column 4), as a result of moving waves of establishment and mortality caused by the presence or absence of feedback from downslope cells. By the end of the simulation run, strips in both columns reached the top of the grid, representing advancement of 19–24 cm/year. It is also evident that the strips of occupied cells gradually migrated upslope, that is, all trees in the furthest downslope cells died and did not re-establish while upslope cells became occupied. The overall migration rate of the strips was about 10 cm/year. Most of these results are consistent with field observations (Bekker, 2005). In particular, the repeating pattern of older-to-younger trees was observed, and a patch of dead trees is present at the base of most of the fingers, suggesting that they are gradually migrating upslope. The simulated advancement rates are also comparable, but slightly faster than the observed values of 8–15 cm/year. Seedling establishment in the field was restricted to areas within 5 m leeward of existing trees from 1720 to about 1850 as predicted by the model, but thereafter seedlings established several meters away from existing trees. The comparison between simulated and observed basal area and density for each species and for the study area is shown in Figure 4. For the study area, basal area increased up to feedback = 100, then leveled off immediately, very close to the observed value of 37 m2/ha, with a slight increase at feedback = 140. This pattern was similar for lodgepole pine, the dominant species in the study area, although the predicted value did not come as close to the observed at any feedback level. The model under-predicted basal area for subalpine fir and over-predicted the value for
Modeling Feedback Effects on Linear Patterns of Subalpine Forest Advancement
(a)
(b)
(d)
(e)
181
(c)
Establishment (yr) 1980–1999 1960–1979 1940–1959 1920–1939 1900–1919 1880–1899 1860–1879 1840–1859 1820–1839 1800–1819 1780–1799 1760–1779 1740–1759 Initialized–dead Initialized–living
Figure 3 Advancement graphics for a standard simulation run in 50 -year increments: (a) 1800, (b) 1850, (c) 1900, (d) 1950, and (e) 2000. Increasing elevation and wind direction is from the bottom to the top of the grid.
Engelmann spruce. Moreover, there was no clear relationship between basal area and feedback for either species. The model under-predicted density for all species, regardless of feedback level. Subalpine fir had the largest discrepancy between predicted and observed values and Engelmann spruce the smallest. Although some slight increase between feedback = 80 and feedback = 100 occurred, density was generally insensitive to feedback level.
182
ALL
Matthew F. Bekker and George P. Malanson
ALL
3,000.0
Basal area (m2/ha)
Density (/ha)
2,500.0 2,000.0 1,500.0 1,000.0 500.0
40.0 35.0 30.0 25.0 20.0 15.0 10.0 5.0
0.0
0.0 80
100
120
140
Density (/ha)
2,000.0 1,500.0 1,000.0 500.0
120
140
80
100
120
140
80
100
120
140
80
100
120
140
25.0 20.0 15.0 10.0 5.0 0.0
80
100
120
140 ABLA
ABLA 600.0
3.5
500.0
3.0
Basal area (m2/ha)
Density (/ha)
100
30.0
0.0
400.0 300.0 200.0 100.0 0.0 100
120
2.0 1.5 1.0 0.5
140
PCEN
PCEN 9.0 8.0 7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0 2 Basal area (m /ha)
160.0 140.0 120.0 100.0 80.0 60.0 40.0 20.0 0.0 Model Field
2.5
0.0 80
Density (/ha)
80 PICO 35.0 Basal area (m2/ha)
PICO 2,500.0
80
100
120
140
Feedback level
Feedback level
Figure 4 Modeled and observed density and basal area for the entire stand (all species) and individual species in response to increasing feedback levels in FORSKA, from a standard simulation run. See Table 2 for explanation of species codes.
4. Discussion 4.1. Gap models and treeline environments Many simulation models have required some modification when used to represent forests other than those for which they were originally developed. The alterations that were made to FORSKA in order to accurately simulate conifers instead of
Modeling Feedback Effects on Linear Patterns of Subalpine Forest Advancement
183
deciduous broadleaf trees and the modification of FOREST-BGC for use at treeline are examples (Cairns, 1998; Leemans, 1989). The inability of FORSKA to produce stand structure raises questions regarding its appropriateness for direct use in treeline studies. Growth functions and parameter estimates that are accurate for lower elevation, closed-canopy forests may be unrealistic for treeline environments.
4.2. Effects of light and mortality Light is a key component of gap models, and FORSKA includes a more realistic representation of canopy structure and thus of competition for light, compared to previous gap models. However, the fundamental ideas behind competition for light in gap models may be inappropriate for this research. A ‘‘patch’’ in a gap model is intended to represent an opening in the canopy resulting from a fallen mature tree, through which light can enter to reinitiate the process of succession. All areas surrounding the patch are essentially assumed to already contain mature trees, because no light enters from outside the patch. Thus, once the patch, or gap, becomes filled with the canopy of a new tree, very little light can reach the forest floor. However, this is just the opposite situation with the forest fingers on Lee Ridge; a patch represents a group of trees surrounded on two sides by treeless areas (coves). Thus, even when mature trees occupy a patch, light can reach the forest floor via the coves adjacent to the fingers. Treeline environments also differ from lower elevations with regards to causes and rates of mortality. Trees at high altitudes are more exposed to abiotic stresses, such as high winds, low temperatures, and damaging frosts. In addition, they are exposed to greater light intensities as discussed above, which may reduce the effects of competition. In a review of the representation of mortality in gap models, Keane et al. (2001) illustrated that much less emphasis has been placed on the accurate parameterization and validation of mortality than on processes related to establishment and growth. They identified several potential problems with the current representation of mortality in the models, three of which are directly relevant here. First, many gap models exhibit ‘‘temporal inflexibility’’ (Hawkes, 2000), which is the assumption that the relationships between climate, growth, and mortality remain unchanged as climate changes. Second, the ‘‘intrinsic’’ mortality rate is often simply considered to be a ‘‘fudge factor’’ to keep trees from living indefinitely, with no ecological basis. Finally, growth-dependent or ‘‘suppressed’’ mortality rates, based on slow growth rates primarily as a result of competition for light, may be inaccurate when modeling species near the edge of their fundamental niche. This last problem is probably most relevant to this research. Keane et al. (2001) suggest that fast-growing trees may be more susceptible to damage or death by drought, winter desiccation, or frost than slow-growing trees. Moreover, many studies have shown that once adult trees are able to establish in harsh treeline environments, they exhibit a high degree of inertia, in that they are unlikely to die, even though their growth rates may be very slow (Lavoie and Payette, 1994; Lloyd and Graumlich, 1997). The high number of cells in which all trees died during the standard simulation runs, coupled with the low-density values, suggest that the representation of mortality may indeed be inaccurate for this environment. Moreover, we identified 194 dead trees in the field, representing 17% of all trees. In contrast, the number of simulated dead trees was more than five times the number of living trees.
184
Matthew F. Bekker and George P. Malanson
To investigate the combined effects of light, mortality, and feedback on forest structure, we ran the model with combinations of increased light (10 and 50%) and decreased suppressed mortality rates (0.046 and 0.0046), with varying feedback levels. The altered light values are similar to those used to simulate increases in light due to cutbank erosion along a riparian forest edge (Kupfer and Malanson, 1993). These changes, particularly the suppressed mortality rates, allowed some trees to survive through the entire simulation period with much lower feedback levels (Figure 5). In ALL 3,000.0
Basal area (m /ha)
ALL 60
2
Density (/ha)
2,500.0 2,000.0 1,500.0 1,000.0 500.0 1
25
60
20 10 1
25
60
120
1
25
60
120
1
25
60
120
1
25 60 Feedback level
120
PICO 35 Basal area (m /ha)
2,000.0
2
Density (/ha)
30
120
PICO 2,500.0
1,500.0 1,000.0 500
30 25 20 15 10 5 0
0 1
25
60
120
ABLA 600
ABLA 5 4
2
Basal area (m /ha)
500 Density (/ha)
40
0
0.0
400 300 200 100
3 2 1 0
0 1
25
60
120
PCEN
PCEN 20 2
Basal area (m /ha)
300 Density (/ha)
50
200 100 0 1
25
60
120
Feedback level
15 10 5 0
Field Suppressed mortality = .046; shading reduction = 10% Suppressed mortality = .046; shading reduction = 50% Suppressed mortality = .0046; shading reduction = 10% Suppressed mortality = .0046; shading reduction = 50%
Figure 5 Modeled and observed density and basal area for the entire stand (all species) and individual species in response to increasing feedback levels in FORSKA, after reductions in the suppressed mortality rate and shading effects. See Table 2 for explanation of species codes.
Modeling Feedback Effects on Linear Patterns of Subalpine Forest Advancement
185
almost all cases, basal area was highest for the stand and for each species at the highest light levels and lowest mortality rates at all feedback levels. In addition, values for the stand and for each species either matched or exceeded observations, and in most cases, this value was reached with feedback at relatively low levels. The combination of increased light and reduced mortality also increased density values for the stand and for all species, but even at the highest feedback levels, the values still fell short of most observations. Overall stand and lodgepole pine densities each approached 50% of the observed values, and subalpine fir density was greatly underestimated. Values for Engelmann spruce, however, exceeded observations. The alterations to mortality and light also affected patterns of advancement. Linear strips of trees were still produced by the model, but migration of the strips was either greatly slowed or stopped, and the repeating pattern of older-to-younger trees was replaced by a simple gradient of decreasing age with increasing elevation. Advancement rates were not affected by the changes, but throughout the simulation trees were able to establish and survive in cells that were not immediately upslope from occupied cells.
5. Conclusions The ability of the model to reproduce general patterns and rates of forest advancement, particularly the repeating pattern of old-to-young trees, suggests that biotic feedback is an important control on treeline patterns on Lee Ridge. The role of feedback in producing and maintaining linear krummholz islands and hedges composed of subalpine fir and Engelmann spruce has been shown previously (Benedict, 1984; Holtmeier, 1982; Marr, 1977). The fingers represent feedback-controlled vegetation patterns that are similar to krummholz islands and hedges, but they consist primarily of lodgepole pine, a species that does not reproduce vegetatively through layering. Lodgepole pine is apparently able to change from a prostrate to upright growth form in response to climate change, just as subalpine fir and Engelmann spruce establish vertical leaders above krummholz canopies. These changes amplify the positive feedback effect by causing greater deposition of wind-blown snow. The simulations suggest that the fingers also migrate upslope, just as krummholz islands do; the migration rate of 10 cm/year produced by the model is comparable to reconstructed rates of krummholz migration (Benedict, 1984; Marr, 1977). The inability of the model to accurately represent the change in the relative importance of feedbacks over time (i.e., seedling establishment only immediately leeward of existing trees before 1850 followed by more extensive establishment in less-protected areas thereafter) is probably due to the fact that site quality was only increased by endogenous feedback, whereas in reality all areas experienced improving exogenous conditions associated with the end of the Little Ice Age. The model had greater difficulty in accurately representing forest structure. The relationship between basal area and density is perhaps the most significant aspect to consider in explaining this discrepancy. The predictions of basal area in the model were much closer to observations than were predictions of density. This result,
186
Matthew F. Bekker and George P. Malanson
combined with the fact that the number of dead trees in the simulations greatly exceeded the number observed in the field, indicates that too few trees were able to reach maturity in the model. The result is that a few trees, which are larger in diameter than those in the field, dominate the patches, instead of many smaller trees. Increasing light and reducing mortality separately only slightly improved predictions of stand structure, and the results were inconsistent. However, when these alterations were combined, the effects were both of higher magnitude and more consistent, illustrating the importance of interactions between these processes. These alterations allowed trees to survive through the simulation period even with feedback levels close to 0. This prediction is consistent with the observations of vegetational inertia in treeline environments mentioned above. FORSKA plants new trees as ‘‘saplings’’ (1.4 m height), so they represent individuals that have survived the very high-mortality seedling stage. Pacala et al. (1996) reported the results of extensive work on model development and testing with field data to analyze forest structure and dynamics at several scales. One of their most important findings concerned the relationship between response to light and mortality in gap models. In particular, they found that including species-specific mortality functions, which are based on shade tolerance, is crucial in accurately simulating forest response to climate change or other novel conditions. For example, they compare the results of a simulation study analyzing species-specific responses to increased atmospheric CO2 using their SORTIE model with results from previous JABOWA models. The SORTIE simulation predicted that shade-intolerant species would increase in dominance under elevated CO2 conditions, while the JABOWA models predicted dominance by shadetolerant species. They suggest that this discrepancy can be explained by differences in assumptions of growth and mortality relationships in the two models. In SORTIE, the enhanced CO2 conditions allow shade-intolerant species to survive in the understory. Thus, because the growth-dependent mortality functions of the shadetolerant species allow them to survive under the canopy even without increased CO2, only shade-intolerant species are benefited by the increase, giving them a competitive advantage. These results are consistent with the recommendations of Keane et al. (2001) regarding the need for better parameterization, representation, and validation of mortality in gap models. These models are one of the bases for global simulations of ecological response to climate change, with feedbacks to the climate system (Bonan, 2008; Purves and Pacala, 2008); their performance at the real points of change – ecotones – is significant. This research has important implications for testing hypotheses about treeline location and pattern. First, carbon balance is a viable starting point from which the role of feedback in producing pattern can be investigated. Second, the observation of prostrate lodgepole pines changing to an upright growth form under improved conditions lends support to the apical-dominance hypothesis. These trees have a stronger tendency than spruce or fir to maintain a single leader as the dominant, apical bud, and therefore protection of this bud is more crucial. Positive feedback from existing trees can improve conditions for leeward individuals, protecting their apical buds and allowing further growth-form change and, consequently, continued feedback. Finally, the prediction from the model of a few trees
Modeling Feedback Effects on Linear Patterns of Subalpine Forest Advancement
187
surviving to maturity, and then inhibiting the further growth of new trees through shading, is consistent with Ko¨rner’s (1998) negative-feedback, growth-limitation hypothesis. However, it is this aspect of the simulation that does not match observations. The dendrochronological reconstruction of forest advancement used to validate this simulation (Bekker, 2005) indicates that the relative importance of feedback has varied over time. Feedback was dominant, amplifying the rate of forest advancement, during climatic conditions that were intermediate in harshness; with improving climatic conditions, feedback became unnecessary for the establishment and survival of trees. This and other models need to be improved in order to accurately represent this trade-off between feedback and climate in treeline environments. Three modifications are needed: (a) a representation of exogenous climate change that will affect all potential establishment sites needs to be incorporated; (b) the representation of mortality needs to be more ecologically based and to include species-specific rates; and (c) the representation of competition for light when simulating forests that do not exhibit true gap-phase dynamics needs to be reconsidered. Such a model could be used to identify treeline patterns and conditions, potentially applied to particular sites, in which feedback is most likely to produce an acceleration of vegetation change and thus would be most beneficial to monitor as potential indicators of climate change.
Acknowledgments We thank Dave Cairns for assistance and suggestions regarding the BGC models.
REFERENCES Alftine, K.J., Malanson, G.P., 2004. Directional positive feedback and pattern at an alpine treeline. Journal of Vegetation Science 15, 3–12. Alftine, K.J., Malanson, G.P., Fagre, D.B., 2003. Feedback-driven response to multidecadal climatic variability at an alpine treeline. Physical Geography 24, 520–535. Allen, T.R., Walsh, S.J., 1996. Spatial and compositional pattern of alpine treeline, Glacier National Park, Montana. Photogrammetric Engineering and Remote Sensing 62, 1261–1268. Bader, M.Y., van Geloof, I., Rietkerk, M., 2007. High solar radiation hinders tree regeneration above the alpine treeline in northern Ecuador. Plant Ecology 191, 33–45. Baker, W.L., Honaker, J.J., Weisberg, P.J., 1995. Using aerial photography and GIS to map the forest-tundra ecotone in Rocky Mountain National Park, Colorado, for global change research. Photogrammetric Engineering and Remote Sensing 61, 313–320. Bekker, M.F., 2005. Positive feedback between tree establishment and patterns of subalpine forest advancement, Glacier National Park, Montana, USA. Arctic, Antarctic, and Alpine Research 37, 97–107. Bekker, M.F., Malanson, G.P., Alftine, K.J., Cairns, D.M., 2001. Feedback and pattern in computer simulations of the alpine treeline ecotone. In: Millington, A.C., Walsh, S.J., Osborne, P.E. (Eds.), GIS and Remote Sensing Applications in Biogeography and Ecology. Kluwer, Dordrecht, pp. 123–138.
188
Matthew F. Bekker and George P. Malanson
Benedict, J.B., 1984. Rates of tree-island migration, Colorado Rocky Mountains, USA. Ecology 65, 820–823. Berg, N.H., 1987. Krummholz snowdrifts: Hydrologic implications at a Colorado treeline site. In: Troendle, C.A., Kaufmann, M.R., Hamre, R.H., Winokur, R.P. (Technical Coordinators). Management of Subalpine Forests: Building on 50 years of Research. USDA Forest Service General Technical Report GTR-RM-149, pp. 141–144. Billings, W.D., 1969. Vegetational pattern near timberline as affected by fire-snowdrift interactions. Vegetatio 19, 192–207. Bonan, G.B., 2008. Forests and climate change: Forcings, feedbacks, and the climate benefits of forests. Science 320, 1444–1449. Bossel, H., 1991. Modelling forest dynamics: Moving from description to explanation. Forest Ecology and Management 42, 129–142. Botkin, D.B., Janak, J.F., Wallis, J.R., 1972. Rationale, limitations, and assumptions of a northeastern forest growth simulator. IBM Journal of Research and Development 16, 101–116. Bowman, W.D., 1992. Inputs and storage of nitrogen in winter snowpack in an alpine ecosystem. Arctic and Alpine Research 24, 211–215. Brown, J.K., 1978. Weight and density of crowns of Rocky Mountain conifers. USDA Forest Service Research Paper INT-RP-197. Brown, D.G., Cairns, D.M., Malanson, G.P., Walsh, S.J., Butler, D.R., 1994. Remote sensing and GIS techniques for spatial and biophysical analyses of alpine treeline through process and empirical models. In: Michener, W.K., Stafford, S., Brunt, J. (Eds.), Environmental Information Management and Analysis: Ecosystem to Global Scales. Taylor and Francis, Philadelphia, pp. 453–481. Burns, R.M., Honkala, B.H. (Technical Coordinators), 1990. Silvics of North America: 1. Conifers. Agriculture Handbook 654. USDA For. Ser., Washington, DC. Butler, D.R., Malanson, G.P., Bekker, M.F., Resler, L.M., 2003. Lithologic, structural, and geomorphic controls on ribbon forest patterns in a glaciated mountain environment. Geomorphology 55, 203–217. Butler, D.R., Malanson, G.P., Cairns, D.M., 1994. Stability of alpine treeline in Glacier National Park, Montana, USA. Phytocoenologia 22, 485–500. Cairns, D.M., 1994. Development of a physiologically mechanistic model for use at the alpine treeline ecotone. Physical Geography 15, 104–124. Cairns, D.M., 1998. Modeling controls on pattern at alpine treeline. Geographic and Environmental Modeling 2, 43–63. Cairns, D.M., 2005. Simulating carbon balance at treeline for krummholz and dwarf tree growth forms. Ecological Modelling 187, 312–328. Cairns, D.M., Malanson, G.P., 1997. Examination of the carbon balance hypothesis of alpine treeline location in Glacier National Park, Montana. Physical Geography 18, 125–145. Cairns, D.M., Malanson, G.P., 1998. Environmental variables influencing the carbon balance at the alpine treeline: A modeling approach. Journal of Vegetation Science 9, 679–692. Camarero, J.J., Gutierrez, E., Fortin, M.-J., 2000. Spatial pattern of subalpine forest-alpine grassland ecotones in the Spanish Central Pyrenees. Forest Ecology and Management 134, 1–16. Earle, C.J., 1993. Forest Dynamics in a Forest-Tundra Ecotone, Medicine Bow Mountains, Wyoming. Ph.D. Dissertation, University of Washington. Frey, W., 1983. The influence of snow on growth and survival of planted trees. Arctic and Alpine Research 15, 241–251. Friend, A.D., Shugart, H.H., Coughlan, J.C., 1993. A physiology-based gap model of forest dynamics. Ecology 74, 792–797. Geddes, C.I.A., Brown, D.G., Fagre, D.B., 2005. Topography and vegetation as predictors of snow water equivalent across the alpine treeline ecotone at Lee Ridge, Glacier National Park, Montana, USA. Arctic, Antarctic, and Alpine Research 37, 197–205. Gosz, J.R., 1993. Ecotone hierarchies. Ecological Applications 3, 369–376. Hadley, J.L., Smith, W.K., 1986. Wind effects on needles of timberline conifers: Seasonal influence on mortality. Ecology 67, 12–19. Hansen-Bristow, K.J., Ives, J.D., 1985. Composition, form, and distribution of the forest-alpine tundra ecotone, Indian Peaks, Colorado, USA. Erdkunde 39, 286–295.
Modeling Feedback Effects on Linear Patterns of Subalpine Forest Advancement
189
Hawkes, C., 2000. Woody plant mortality algorithms: Description, problems, and progress. Ecological Modelling 126, 225–248. Hessl, A.E., Baker, W.L., 1997. Spruce-fir growth form changes in the forest-tundra ecotone of Rocky Mountain National Park, Colorado, USA. Ecography 20, 356–367. Hiemstra, C.A., Liston, G.E., Reiners, W.A., 2002. Snow redistribution by wind and interactions with vegetation at upper treeline in the Medicine Bow Mountains, Wyoming, USA. Arctic, Antarctic, and Alpine Research 34, 262–273. Hiemstra, C.A., Liston, G.E., Reiners, W.A., 2006. Observing, modelling, and validating snow redistribution by wind in a Wyoming upper treeline landscape. Ecological Modelling 197, 35–51. Holtmeier, F.-K., 1982. ‘‘Ribbon-forest’’ and ‘‘hedges.’’ Strip-like distribution patterns of trees at the upper timberline in the Rocky Mountains. Erdkunde 36, 142–153. Holtmeier, F.-K., 2003. Mountain Timberlines: Ecology, Patchiness, and Dynamics. Kluwer, Dordrecht. Jensen, M.E., Hann, W., Keane, R.E., Caratti, J., Bourgeron, P.S., 1993. ECODATA-A multiresource database and analysis system for ecosystem description and evaluation. In: Jensen, M.E., Bourgeron, P.S. (Eds.), Eastside Forest Ecosystem Health Assessment, Volume II, Ecosystem Management: Principles and Applications. USDA Forest Service General Technical Report PNW-GTR-318, pp. 203–217. Kaufmann, M.R., Edminster, C.B., Troendle, C.A., 1982. Leaf area determinations for subalpine tree species in the central Rocky Mountains. USDA Forest Service General Technical Report RM-GTR-238. Keane, R.E., Austin, M., Field, C., Huth, A., Lexer, M.J., Peters, D., et al., 2001. Tree mortality in gap models: Application to climate change. Climatic Change 51, 509–540. Keane, R.E., Morgan, P., Running, S.W., 1996. FIRE-BGC – A mechanistic process model for simulating fire succession on coniferous forest landscapes of the northern Rocky Mountains. USDA Forest Service Research Paper INT-RP-484. Ko¨rner, C., 1998. A re-assessment of high elevation treeline positions and their explanation. Oecologia 115, 445–459. Kullman, L., 1986. Recent tree-limit history of Picea abies in the southern Swedish Scandes. Canadian Journal of Forest Research 16, 761–771. Kupfer, J.A., Cairns, D.M., 1996. The suitability of montane ecotones as indicators of global climatic change. Progress in Physical Geography 20, 253–272. Kupfer, J.A., Malanson, G.P., 1993. Observed and modeled directional change in riparian forest composition at a cutbank edge. Landscape Ecology 8, 185–199. Lavoie, C., Payette, S., 1994. Recent fluctuations of the lichen-spruce forest limit in subarctic Quebec. Journal of Ecology 82, 725–734. Leemans, R., 1989. Description and simulation of stand structure and dynamics in some Swedish forests. Acta Universitatis Upsaliensis 221. Uppsala, Sweden. Leemans, R., 1991. Sensitivity analysis of a forest succession model. Ecological Modelling 53, 247–262. Leemans, R., Prentice, I.C., 1989. FORSKA, a general forest succession model. General Report 89/ 2, Institute of Ecological Botany, Uppsala, Sweden. Lloyd, A., Graumlich, L., 1997. Holocene dynamics of treeline forests in the Sierra Nevada. Ecology 78, 1199–1210. Malanson, G.P., 1997. Effects of feedbacks and seed rain on ecotone patterns. Landscape Ecology 12, 27–38. Malanson, G.P., Butler, D.R., Fagre, D.B., Walsh, S.J., Tomback, D.F., Daniels, L.D., et al., 2007. Alpine treeline of western North America: Linking organism-to-landscape dynamics. Physical Geography 28, 378–396. Malanson, G.P., Xiao, N., Alftine, K.J., 2001. A simulation test of the resource-averaging hypothesis of ecotone formation. Journal of Vegetation Science 12, 743–748. Marr, J.W., 1977. The development and movement of tree islands near the upper limit of tree growth in the southern Rocky Mountains. Ecology 58, 1159–1164. Minnich, R.A., 1984. Snow drifting and timberline dynamics on Mount San Gorgonio, California, USA. Arctic and Alpine Research 16, 395–412. Neilson, R.P., 1993. Transient ecotone response to climatic change: Some conceptual and modelling approaches. Ecological Applications 3, 385–395. Nikolov, N., Helmisaari, H., 1992. Silvics of the circumpolar boreal forest tree species. In: Shugart, H.H., Leemans, R., Bonan, G.B. (Eds.), A Systems Analysis of the Global Boreal Forest. Cambridge University Press, New York, pp. 13–84.
190
Matthew F. Bekker and George P. Malanson
Noble, I.R., 1993. A model of the responses of ecotones to climate change. Ecological Applications 3, 396–403. Pacala, S.W., Canham, C.D., Saponara, J., Silander Jr., J.A., Kobe, R.K., Ribbens, E., 1996. Forest models defined by field measurements: Estimation, error analysis and dynamics. Ecological Monographs 66, 1–43. Payette, S., Filion, L., 1985. White spruce expansion at the tree line and recent climate change. Canadian Journal of Forest Research 15, 241–251. Payette, S., Filion, L., Gauthier, L., Boutin, Y., 1985. Secular climate change in old-growth tree-line vegetation of northern Quebec. Nature 315, 135–138. Prentice, I.C., Leemans, R., 1990. Pattern and process and the dynamics of forest structure: A simulation approach. Journal of Ecology 78, 340–355. Prentice, I.C., Sykes, M.T., Cramer, W., 1993. A simulation model for the transient effects of climate change on forest landscapes. Ecological Modelling 65, 51–70. Purves, D., Pacala, S., 2008. Predictive models of forest dynamics. Science 320, 1452–1453. Rietkerk, M., van de Koppel, J., 2008. Regular pattern formation in real ecosystems. Trends in Ecology and Evolution 23, 169–175. Risser, P.G., 1995. The status of the science examining ecotones. BioScience 45, 318–325. Running, S.W., Coughlan, J.C., 1988. A general model of forest ecosystem processes for regional applications. I. Hydrologic balance, canopy gas exchange and primary production processes. Ecological Modelling 42, 125–154. Sato, K., Iwasa, Y., 1993. Modeling of wave regeneration in subalpine Abies forests: Population dynamics with spatial structure. Ecology 74, 1538–1550. Scott, P.A., Hansell, R.I.C., Erickson, W.R., 1993. Influences of wind and snow on northern treeline environments at Churchill, Manitoba, Canada. Arctic 46, 316–323. Shugart, H.H., West, D.C., 1977. Development of an Appalachian deciduous forest succession model and its application to assessment of the impact of the chestnut blight. Journal of Environmental Management 5, 161–179. Shugart, H.H., West, D.C., 1979. Size and pattern of simulated forest stands. Forest Science 25, 120–122. Sprugel, D.G., 1976. Dynamic structure of wave-regenerated Abies balsamea forests in the northeastern United States. Journal of Ecology 64, 889–911. Stevens, G.C., Fox, J.F., 1991. The causes of treeline. Annual Review of Ecology and Systematics 22, 177–191. Tongway, D.J., Valentin, C., Seghieri, J. (Eds.), 2001. Banded Vegetation Patterning in Arid and Semiarid Environments. Springer, New York. Urban, D.L., Harmon, M.E., Halpern, C.B., 1993. Potential response of pacific northwestern forests to climatic change, effects of stand age and initial composition. Climatic Change 23, 247–266. Walsh, S.J., Butler, D.R., Allen, T.R., Malanson, G.P., 1994. Influence of snow patterns and snow avalanches on the alpine treeline ecotone. Journal of Vegetation Science 5, 657–672. Weisberg, P.J., Baker, W.L., 1995. Spatial variation in tree seedling and krummholz growth in the forest-tundra ecotone of Rocky Mountain National Park, Colorado, USA. Arctic and Alpine Research 27, 116–129. White, L.P., 1969. Vegetation arcs in Jordan. Journal of Ecology 57, 5461–5464. Wild, J., Winkler, E., 2008. Krummholz and grassland coexistence above the forest-line in the Krkonose Mountains: Grid-based model of shrub dynamics. Ecological Modelling 213, 293–307. Williams, M.W., Brooks, P.D., Seastedt, T., 1998. Nitrogen and carbon soil dynamics in response to climate change in a high-elevation ecosystem in the Rocky Mountains, USA. Arctic and Alpine Research 30, 26–30. Wilson, J.B., Agnew, A.D.Q., 1992. Positive-feedback switches in plant communities. Advances in Ecological Research 23, 263–336. Zeng, Y., Malanson, G.P., 2006. Endogenous fractal dynamics at alpine treeline ecotones. Geographical Analysis 38, 271–287. Zhang, Y.A., Peterman, M.R., Aun, D.L., Zhang, Y.M., 2008. Cellular automata: Simulating alpine tundra vegetation dynamics in response to global warming. Arctic, Antarctic, and Alpine Research 40, 256–263.
C H A P T E R
1 0
The Future of Treeline David R. Butler, George P. Malanson, and Stephen J. Walsh References
194
The chapters in this volume have attempted to provide a coherent picture of the characteristics of the alpine treeline ecotone in Glacier National Park (GNP). These studies contribute to theories of treeline location and formation by its focus on the site and landscape-scale dynamics that may reflect the initial stages of treeline advance. By considering treeline at a mesoscale, the growth, death, and establishment of seedlings and saplings in the open tundra matrix are observable and measurable. It is the survival of these young conifers above current locations of treeline that will ultimately determine whether treeline will advance up a mountain slope. While their growth into trees is also of importance, a change from tundra to an extensive krummholz zone is a more significant step in terms of ecosystem processes and biodiversity. Treeline response to climate change in GNP will depend on what? Treelines in GNP varied widely in their 20th century dynamics. In many areas, no change can be seen between the present and patterns shown in photographs from 90 years ago (Butler et al., 1994); in others trees have become denser or, in a few locations, have advanced upslope tens of meters (Butler and DeChano, 2001; Klasner and Fagre, 2002). This variability in spatial response to sparsely documented changes the park experienced during the past century indicates that we should expect variability in the way treeline responds to future climate changes. Here and elsewhere, the local context is so important that broad predictions are not reliable. The focus area of our studies has been on the eastern front of the Park, where the climate is dominated by continental conditions with more variable temperatures and drier hydrology (Finklin, 1986). There, the response to a warmer climate will depend primarily on whether it becomes wetter or not. If wetter, the potential exists for tree species to advance into tundra; if not, trees could even retreat downslope (Cairns and Malanson, 1997). But even this potential may go unfulfilled. The studies throughout this book have emphasized the importance of specific, fine-scale, geomorphic and soils processes and conditions that create hospitable sites for tree seedlings, as well as coarser-scale processes such as snow avalanches and debris flows that act to preclude treeline advance into otherwise climatically favorable elevations. The directionality and intensity of treeline Developments in Earth Surface Processes, Volume 12 ISSN 0928-2025, DOI 10.1016/S0928-2025(08)00210-1
Ó 2009 Elsevier B.V. All rights reserved.
191
192
David R. Butler et al.
responses to these changes could increase or decrease with climate change. Wetter conditions could produce more turf exfoliation if the temperature variability at daily scales, that is, freeze thaw cycles, is increased. Greater annual temperature variability will not help tree seedling establishment, however, because it may mean that the number of days with freeze–thaw cycles would be reduced during the steeper annual changes, reducing the effectiveness of turf exfoliation in creating amenable nurse sites for seedling establishment. Wetter conditions might also increase the frequency and/or magnitude of avalanches and debris flows, negating the fine-scale effects of fostering processes such as turf exfoliation. Finally, feedback mechanisms must play a role, once the feedback loop is initiated. All of these factors will be dependent upon location-specific conditions as well as regional climate change. The slope geomorphology at fine and coarser scales will interact with the synoptic climate to determine how any regional climate factors will be translated to local microclimates that treelines experience. The dependencies are such that specific expectations are unrealistic. Over the last century, studies have attempted to discern the composition and spatial structure of alpine treeline ecotones and the corresponding pattern–process relations that function across space and time scales. Early efforts focused on finding a single explanation for the occurrence of alpine treeline and Arctic timberline (cf. Arno and Hammerly, 1984). The emphasis shifted, however, as it became clear that several different environmental factors influence the composition and spatial pattern of the alpine treeline ecotone. As summarized in the preceding chapters, our work in GNP has illustrated this over and over again. The collective works on alpine treeline by the members of the Mountain GeoDynamics Research Group have not been conducted in a vacuum in GNP. Complementary studies examining a variety of other forms of environmental change in GNP have also been underway for many years. Snowpack variability is a concern from the perspective of water availability as well as from that of safety issues associated with the annual opening of Going-to-the-Sun Road each spring (Selkowitz et al., 2002a, 2002b). The frequency and magnitude of snow avalanches, and their effect upon the disturbance-induced treeline ecotone, have been a focus of work both by the USGS (Reardon and Lundy, 2004; Reardon et al., 2004) and members of our group (Sawyer and Butler, 2006; Walsh et al., 2004). Other ecotones in the Park that are patterned through the interaction of climate, climate change, and geomorphology also illustrate recent changes where trees are becoming established in formerly nontreed sites (Figure 1) (Butler et al., 2003a, 2003b; Cerney and Butler, 2004). The USGS is actively involved in documenting glacial recession throughout the park, through on-site measurements as well as through the use of repeat photography (Hall and Fagre, 2003; Key et al., 2002). The common thread running through all these studies is an emphasis on change. As Fagre mentioned in this volume’s introduction, visitors to GNP come to see the Park’s outstanding natural scenery. This scenery is changing before their eyes, and they will have to come to grips with this reality. Treeline in the park viewed by our grandchildren will not be the treeline we have studied over these many years. The world is changing because of climate change, and treeline is a part of that world of change. We hope our work has helped increase understanding of why that world is
The Future of Treeline
193
Figure 1 Nontreeline ecotone changes in GNP ^ at lower treeline, where trees are invading grassland (top); in subalpine meadows, again where trees are invading formerly unforested sites (middle); and in‘‘snow glades’’separating ribbon forests (bottom).
194
David R. Butler et al.
changing and will continue to do so for the foreseeable future. The future of treeline is continuing change, and it will be fascinating to observe the nature of those changes.
REFERENCES Arno, S.F., Hammerly, R.P., 1984. Timberline: Mountain and Arctic Forest Frontiers. The Mountaineers, Seattle. Butler, D.R., DeChano, L.M., 2001. Environmental change in Glacier National Park, Montana: An assessment through repeat photography from fire lookouts. Physical Geography 22, 291–304. Butler, D.R., Malanson, G.P., Bekker, M.P., Resler, L.M., 2003a. Lithologic, structural, and geomorphic controls on ribbon forest patterns. Geomorphology 55, 203–217. Butler, D.R., Malanson, G.P., Cairns, D.M., 1994. Stability of alpine treeline in Glacier National Park, Montana, USA. Phytocoenologia 22, 485–500. Butler, D.R., Resler, L.M., Gielstra, D.A., Cerney, D.L., 2003b. Ecotones in mountain environments: Illustrating sensitive biogeographical boundaries with remotely sensed imagery in the geography classroom. Geocarto International 18, 63–72. Cairns, D.M., Malanson, G.P., 1997. Examination of the carbon balance hypothesis of alpine treeline location, Glacier National Park, Montana. Physical Geography 18, 125–145. Cerney, D.L., Butler, D.R., 2004. Examining montane ecotone change with repeat photography. Papers of the Applied Geography Conference 27, St. Louis, MO, 111–121. Finklin, A.I., 1986. A Climatic Handbook for Glacier National Park, with Data for Waterton Lakes National Park. USDA Forest Service INT-GTR-204. Hall, M.P., Fagre, D.B., 2003. Modeled climate-induced glacier change in Glacier National Park, 1850–2100. Bioscience 53, 131–140. Key, C.H., Fagre, D.B., Menicke, R.K., 2002. Glacier retreat in Glacier National Park, Montana. In: Williams, R.S., Ferrigno, J.G. (Eds.), Satellite Image Atlas of Glaciers of the World, Chapter J, Glaciers of North America. US Geological Survey Professional Paper 13686-J, pp. J365–J381. Klasner, F.L., Fagre, D.B., 2002. A half century of change in alpine treeline patterns at Glacier National Park, Montana, USA. Arctic and Alpine Research 34, 49–56. Reardon, B.A., Fagre, D.B., Steiner, R.W., 2004. Natural avalanches and transportation: A case study from Glacier National Park, Montana, USA. In: Proceedings of the International Snow Science Workshop, Jackson, WY. September 2004. Reardon, B.A., Lundy, C., 2004. Forecasting for natural avalanches during spring opening of the Going-to-the-Sun Road, Glacier National Park, USA. In: Proceedings of the International Snow Science Workshop, Jackson, WY. September 2004. Sawyer, C.F., Butler, D.R., 2006. A chronology of high-magnitude snow avalanches reconstructed from archived newspapers. Disaster Prevention and Management 15, 313–324. Selkowitz, D.J., Fagre, D.B., Reardon, B.A., 2002a. Interannual variations in snowpack in the Crown of the Continent Ecosystem. Hydrological Processes 16, 3651–3665. Selkowitz, D.J., Fagre, D.B., Reardon, B.A., 2002b. Spatial and temporal snowpack variation in the Crown of the Continent Ecosystem. In: Proceedings of 70th Annual Meeting of the Western Snow Conference. Granby, CO, pp. 98–109. Walsh, S.J., Weiss, D.J., Butler, D.R., Malanson, G.P., 2004. An assessment of snow avalanche paths and forest dynamics using Ikonos satellite data. Geocarto International 19, 85–93.
INDEX
Abies 38 Abies lasiocarpa, subalpine fir 4, 47, 126, 144, 156, 158–159, 168, 173, 185 Abruptness 30–31 absolute 68 activity 78 ADAR-5500 17, 24, 88 Albedo 47, 49–50, 53 Altyn formation/limestone 75, 109 Analysis of variance, ANOVA 135–136, 144 Andes 105 Animals 41–43, 79, 81 Apical dominance 170, 186 Apikuni Cirque 125, 126 Aridity 49 Aspect 180 ATE-BGC 171, 176 Autocorrelation 55 Avalanche, see snow avalanche boulder tongue 67–68 Baring Basin/Creek 14, 67, 71, 125, 127 Beer’s Index/BI 130, 133, 143 Bighorn sheep, see Ovis canadensis Biodiversity 162, 191 Bison Mountain 15 Blister rust, Cronartium ribicola 38, 161 Bolivia 105 Boulder(s) 40–41, 73–75, 152, 156, 160 Bounder 87 Burrow(s) 78 Canopy 14, 18, 22, 27, 30–31, 41, 46–48, 52–54, 121, 123–124, 128, 130–135, 137–140, 143–144, 146, 151, 170, 172, 176, 183, 186 Canyon Creek 68, 74
Carbon allocation/balance/dynamics 36, 43–44, 49–51, 53, 65, 169–171 Cataract Creek 14–15, 67, 75, 108, 111, 125–127, 154, 158–159, 162 Catena 87 Cellular automata 56, 171 Clark’s nutcracker, see Nucifraga columbiana Classification 17, 24–27 Clast 51, 86 Climate/climatic change(s) 1, 3, 6, 16, 29, 36, 56–57, 66, 81, 117, 191–192 Climatic optimum 68 Climax 109 Colluvium 109 Colorado 65, 85 Competition 3, 51, 153, 187 Complexity science 57 Concavity 95, 105 Continental Divide 13, 65, 68, 90, 120–121, 124–125, 132, 137–138, 145, 154, 162 Convexity 95 Cotyledon 39 Couloir 66 Cracker Lake 68 Cryoturbation 85 Dating 68 Debris flow(s) 65, 68, 125, 146, 191–192 Deflation 73 Digital elevation model, DEM 21–24, 32, 86, 95, 103, 105, 128–129, 132, 137 Desiccation 43, 50, 53, 176, 183 Diffuse noninterceptance, DIFN 121, 130, 138 Disequilibrium 64 Dispersal 38–40 Disturbance 6, 7, 13, 36, 86, 120, 162 disturbance treeline 65–66, 192 195
196
Diversity 30 Divide Mountain (Peak) 14, 69–70, 73–80, 87, 108, 111, 113, 117, 154, 159, 162 Drainage 132 Dryas, Dryas octopetala 69, 126, 185 Duff 21, 47, 56 Dwarf tree 41, 46, 56, 168 East Flattop Mountain 14, 72–73, 125, 127 Ecosystem engineer 160 Ecosystem services 2 Edge effect 131, 146 Elevation 87, 92, 95, 105, 180 Engelmann spruce, see Picea engelmannii Eolian 72 Facilitation 3, 51, 54–56, 153, 160 Feedback 36, 54–56, 180, 184–185, 192 negative 170, 187 positive 123, 161, 167–169, 186 Fire 13 Firebreak 5 Foliage density, FD 121, 130, 133–135, 137–138, 140, 144–145 Forest 24, 26 subalpine 11, 30 spruce-pine 173 FOREST-BGC 176, 183 FORET 170–171 FORSKA 172–180, 182 Fractal 55–56 FRAGSTATS 93 Freeze-thaw 87, 102, 191 Front Range 85 Frost 43, 79, 87, 183 heaving, churning 73, 79–81 pan(s) 77–78, 80 sorting 76 Gable Mountain 14, 87 Gap model(s) 170, 172, 183 Geographic information science/system, GIS 12, 16, 19–20, 86, 92–93, 128, 130, 132–133
Index
Geomorphic, geomorphology 6, 7, 50, 56, 63, 86, 191–192 Geostatistics 55 Germination 39 Glacial recession 192 Glaciation 64, 117, 154 Goat Mountain 14 Going-to-the-Sun Mountain 14 Going-to-the-Sun Road 3 Gopher(s) 78–79 GPS 87, 93, 121, 128, 130 Grizzly bear, see Ursus arctos Gross primary productivity, GPP 176, 178 Ground squirrel(s) 42, 78, 81 Growing season 48 Growth form(s) 122 Hedge 168–169, 185 Herbivory 81 Herpotrichia coulteri, snow mold 48 IKONOS 17 Indicator 6, 36, 168 Infiltration 47–48, 53 Insolation 129 Interception 47–48, 53 Invasion/invasibility 35–36, 56–57 Island 35 JABOWA 170–171, 186 Junco 43 Krummholz 4, 13–15, 41, 46–47, 49–54, 56, 69, 71–72, 75, 117, 120–124, 126–128, 130–133, 135, 137–139, 141, 143–144, 146–147, 151, 168–169, 171, 176, 178, 185, 191 Lagopus leucurus, white-tailed ptarmigan 4 Lake McDonald 90 Landform Index, LFI 129 Landform(s) 63–65 Landscape ecology 27 Landscape 32, 64, 191 Landslide(s) 64–65
197
Index
Laramide Orogeny 64, 125 Larix lyallii, subalpine larch 153, 156, 162 Latitude 180 Layering 144, 170 Leaf area 171 Leaf Area Index (LAI) 16, 18, 28–29, 47, 121, 123, 128, 130–131, 133–147, 176, 178 Lee Ridge 1, 14, 20, 26, 36, 56–57, 69–71, 74–76, 78, 87, 90, 108–117, 125–126, 128, 137, 140–141, 145, 152–155, 159, 161–162, 168, 172–173, 175, 180 Lemhi Mountains 161 Levation 86 Lewis Overthrust (fault) 64, 124–125 Lewis Range 12, 90, 125 Lichen 69, 86–87 LIDAR 105 Liebig’s Law of the Minimum 37 Light 184 Limberpine, see Pinus flexilis Lineaments 66 Lithology 66 Litter 47, 51 Little Ice Age, LIA 6, 12, 57, 64, 161, 168, 185 Livingston Range 12, 90, 125 Lodgepole pine, see Pinus contorta Logan Pass 3 Macrotopography 154 Mammals 50 Mapping 31 Markov, Markov chain(s) 154–155, 157–159, 161–163 Marmota caligata, marmot 5, 42 Meadow(s) 3, 4, 24, 26, 108, 127, 161, 193 Metrics, pattern, spatial 29–31, 93–94 Microclimate 44, 48–50, 51, 53, 73, 160, 192 Microorganisms 50 Microtopography, microsite 22, 36, 123, 152, 154, 160 Mixture modeling 17, 28–29 Mortaility 38, 47, 75, 171, 184 Mosaic 3, 13, 168
Mountain sheep, see Ovis canadensis Multispectral Scanner, MSS 20 Mycorrhiza 160 National Park Service 2 Needle ice 73, 79, 95, 104 needle-ice pan(s) 76, 79, 81 Net primary productivity, NPP 47–48, 53 Niche 161–162 Nitrogen 50 Normalized Difference Vegetation Index, NDVI 24, 28–29 Nucifraga Columbiana, Clark’s nutcracker 39, 160, 163 Nurse rocks 87 Nutrient(s) 5, 109, 114 Oreamnos americanus, mountain goat 4, 43, 81 Organic matter 50 Ovis canadensis, bighorn sheep, mountain sheep 5, 43 Pacific Decadal Oscillation, PDO 44–45 patch, patches, patchiness 3, 14–15, 18, 24, 26–32, 41, 46, 50–52, 56, 88, 94, 120– 124, 126–128, 130–133, 135–141, 143–147, 157–163, 168–169, 172–173, 175–176, 180, 183, 186 Pattern metrics see metrics Pattern, spatial 12 Patterned ground 14–16, 76, 79, 85, 152, 160 Penetrometer 70, 113, 116–117 Periglacial 14–16, 110, 154, 160 Potential Evapotranspiration, PET 56 Photosynthesis 48 Photosynthetically active radiation, PAR 45, 56 Physiological models 170–171 Picea 38 Picea abies, Norway spruce 173 Picea engelmannii, Engelmann spruce 47, 126, 144, 168, 173, 181, 185 Piegan Pass 14, 87 Pinus 38 Pinus albicaulis, whitebark pine 38–39, 42, 126, 144, 153, 156, 158, 160–161, 163
198
Pinus aristata 39 Pinus contorta, lodgepole pine 126, 144, 153, 160, 162–163, 169, 180, 186 Pinus flexilis, limber pine 39, 126, 144, 161 Pinus sylvestris, Scots pine 173 Pleistocene 12, 64, 69, 90, 117, 154 Precipitation 45, 47–48 Preston Park 14, 15, 71, 125, 127 Principal components analysis, PCA 17, 92, 95–96 Protection 36, 39–43 Pseudotsuga menziesii, Douglas fir 162 RADAR 20 RADARSAT 20 Radiation 43, 45–47, 129, 143, 145 relative radiation index, TRASP 129–130, 133, 139, 143 short-wave 123 Radicle 21, 39 Rasena¨bschalung 73 Relative slope position, RSP 139, 145 Relict species 36 Remote sensing 12, 14, 16–21, 55, 87 Resin 47 Resource averaging 55, 170 Respiration 47, 170 Ribbon forest 167, 193 Riser, see solifluction Rock glacier 64 Rockfall 68 Rocky Mountains, Rockies 65, 124 Rocky Mountain National Park 65, 132 Rocky Mountain Orogeny 90 Roughness 53 Safe sites 180 Salix spp. 49, 154 Sapwood 173 Scale 12, 31, 35–36, 65, 105 microscale 76 Scenic Point 14–15, 69, 71–72, 87, 108, 111, 113, 117, 125, 127 Sediment trap(s) 71–72 Seed(s) 36–43 dispersal 160
Index
Seed rain 38 Seedbed 39 Seedling 36–51, 56, 74, 78, 108, 152, 154, 191 establishment 7, 36, 49, 64, 68–69, 73, 75–76, 79, 81, 87, 117, 123, 162, 180, 191 survival 64 Self-organization 57 Semivariagram, semivariance 87 Sexton Glacier 14 Shade 48–49, 53 Shelter(s) 75–76 Sieve 36, 57, 108 Sigmoid wave 55 Simulation 55, 168, 170 Siyeh Pass 14–15, 79, 87, 92 Sky exposure 47 Slope 90, 180 angle 27, 86, 97, 102, 105 aspect 27, 86, 133, 139, 141, 143, 105 curvature 95, 102 shape 95 Snow 5, 11, 16, 24, 45–46, 50–51, 53, 120, 145, 170, 175 Snow avalanche 5, 6, 13, 16 path(s) 65–67 runout zone(s) 66 Snow mold, see Herpotrichia coulteri Snow Potential Index, SPI 130, 139, 143 Snowpack 108, 192 Soil(s) 7, 44, 50–51, 104, 107–108, 152, 191 color(s) 109 compaction 113–117 depth 70, 113–114 development 63, 108–109, 123 effective soil depth 69, 115–117 horizon(s) 70, 107, 109, 114–115 moisture 49–50, 86, 129 nutrient(s) 109, 114 penetrability 69, 115 pH 109, 111, 114–115 piping 73–74 profile(s) 109, 114 PDI 110–111, 114 throughflow 73–74
199
Index
SOLARPOT Index, SPI 129, 133, 139, 143 Solifluction (also see patterned ground) 7, 39–40, 55, 68–69, 71, 85–86, 97, 102–104, 107, 126 also Riser(s) 70, 73–75, 78, 111, 113, 115–117, also Tread(s) 79, 92, 185 SORTIE 186 South Africa 85 Spatial heterogeneity 27 organization 16, 86–88, 120 pattern, process 12, 22–23, 30–31, 55–57, 168, 171 scale 14, 65 structure 29, 86, 88, 93, 105, 121 Species richness 162 Spectral reflectance 18 response 16 signature 16, 28 Stress 36 Structure, geologic 66 Subalpine fir, see Abies lasiocarpa Sweden 173 Synoptic 45
Torque 52 Trampling 81 Transpiration 46, 49 Tread, see solifluction Tree finger(s) 14, 51–52, 107, 110, 114, 117, 169–170, 173, 180, 183, 185 Tree island(s) 47, 55, 122–123, 151–156, 160–161, 168–169, 175, 185 Triangular irregular network, TIN 131 Tundra 11, 13–15, 24, 26, 30, 35–36, 38, 47, 53, 57, 71–72, 74, 104, 107, 109–110, 114–115, 117, 120, 122, 126–127, 145–146, 152, 169, 191 Turf exfoliation 73–76, 78–79, 116–117, 191 Turf-banked terrace(s) 40, 69–70, 79, 85–105, 111, 113, 116
Talus 65, 67–69 Temperature 48–49, 95 Terrain Shape Index, TSI 129, 133, 139 Texture 30 Thematic Mapper, TM 20, 28 Till, glacial 109 Topographic complexity 121 Topographic moisture 105 Topographic Ruggedness Index, TRI 129, 133, 139 Topographic Wetness Index, TWI 129, 132 Topography, see macrotopography, microtopography Topoclimate, topoclimatic 120, 141, 143
Wallow(s), (ing) 81 Water use efficiency 47 Wave(s) (also see sigmoid wave) 167, 171, 180 green 51 Wavelet 87 White Calf 69, 71, 108–111, 113–117 Whitebark pine, see Pinus albicaulis White-tailed ptarmigan, see Lagopus leucurus Wildlife 6 Wind 45–46, 52, 56, 87, 130, 169, 178, 183
Ungulate(s) 78, 81 Ursus arctos, grizzly bear 4, 42, 81 US Geological Survey, USGS 1 Ultraviolet, UV 46, 51, 53 Vegetation index, VI 28 Viewscape 4 Visualization 121
Xanthoria elegans 87
