No title

Advances in

ECOLOGICAL RESEARCH VOLUME 9

This Page Intentionally Left Blank

Advances in

ECOLOGICAL RESEARCH Edited by

A. MACFADYEN School of Biological and Environmental Studies, New University of Ulster, Coleraine, County Londonderry, Northern Ireland

VOLUME 9

1975

ACADEMIC PRESS London New York San Francisco A Subsidiary of Harcourt Brace Jovanovich, Publishers

ACADEMIC PRESS INC. (LONDON) LTD. 24/28 Oval Road London NW1 Un&d States Edition published by ACADEMIC PRESS INC. 11 1 Fifth Avenue New York, New York 10003

CopyripHc 1975 by ACADEMIC PRESS INC. (LONDON) LTD.

All Rights Reserved No part of this book may be reproduced in any form by photostat, microfilm or any other means, without written permission from the publishers

Library of Congress Catalog Card Number: 62-21479 ISBN: 9-12-0.13909-x

PRINTED I N GREAT BRITAIN B Y T. AND A. CONSTABLE LTD., EDINBURGH

Contributors to Volume 9 G . E. BLAU,Computation Research, The Dow Chemical Company, Midland, Michigan 48640, USA. THOMASM. HINCKLEY,School of Forestry, University of Missouri, Columbia, M O 65201, USA. W. W. MURDOCH, Department of Biological Sciences, University of California, Santa Barbara, California 93106, USA. W. BROCK NEELY,Ag-Organics Product Department, The Dow Chemical Company, Midland, Michigan 48640, U S A . A. OATEN,Department of Biological Sciences, University of California, Santa Barbara, California 93106, USA. JOHNPROCTOR, Biology Department, University of Stirling, Stirling, Scotland. GARY A. RITCHIE,Weyerhaeuser Go., Tacoma,W N 98401, U S A . STANLEY R. J. WOODELL, Botany School, University of Oxford, Oxford, England.

V

This Page Intentionally Left Blank

Preface By chance, this volume could be said to contain two “botanical” and two “zoological” papers, two of which are more theoretical and two “applied”. It is to be hoped that the volume will not be regarded in that light. Each of the papers reflects features of current ecological thinking which transcend systematic and other boundaries and demonstrate that, as ecological science matures, such divisions are of limited relevance. An exciting feature of recent ecological science has been its closer associations with ethology. When two disciplines come together in this way there is a particular need for broad reviews which can be understood by readers from both areas. Murdoch and Oaten’s contribution would seem t o meet that need well. I n Volume 8 the article by Krebs and Myers demonstrated the relevance of intraspecific behaviour to small mammal population dynamics. This volume’s contribution goes on to demonstrate the importance of behavioural characteristics to the stability properties of predator-prey relationships and of whole communities. The paper by Blau and Neely is of general interest for two reasons. First it explains with great clarity an approach to systems modelling and to techniques for discriminating between alternative ecological hypotheses. Secondly these procedures are demonstrated in practice with reference to data on the distribution of an insecticide in an aquatic system containing soil, plants and fish. Changes of distribution and movements of the chemical between these components are modelled over a period of time and a procedure is given for assessing environmental hazards more generally in a multi-compartment system. The progress of ecology has frequently been carried forward by techniques of no great complexity or sophistication which have the advantages of robustness and cheapness and which permit extensive replication under field conditions. Earlier examples are simple census methods, techniques for measuring soil carbon dioxide emission and devices such as the sugar inversion technique for temperature integration. To laboratory scientists such methods often appear crude and inaccurate, but to an ecologist who has considered the relations between the magnitude and the variability of individual field readings, such methods can be superior in terms of cost effectiveness to those which, at greater expense, give unnecessary precision. It is absolutely essential, vii

viii

PREFACE

however, that such aimple devices be employed with a full understanding of their limitations and sources of error. It is in this spirit that Ritchie and Hinckley offer apaper on the Pressure Chamber for measuring Xylem pressure potential in plants, This has now been widely used in comparisons between species and between environmental variables, sometimes uncritically and without due consideration of error sources. The authors consider theory and operation of this basically simple device over a range which should greatly extend the confidence with whiuh it can be used by ecologists. The conceptual gap between a detailed analysis of the physiological effects of abiotic factors on plants and the establishment of whole plant communities is well illustrated by Proctor and Woodell’s review of serpentine soils and their flora. Frequently, in the past, authors have generalized from limited observations and have isolated particular factors as “the cause” of the floral peculiarities of serpentine. As these authors show, however, contrary examples can frequently be cited to many such simplifications and it is only through a much broader ecological approach that one can comprehend the determinants of this highly specialized flora and suggest means of improving fertility in those serpentine areas which have come under human management. The continuing success of “Advances in Ecological Research” is most encouraging. It must derive mainly from the high level of activity in ecology at the present time but it also argues strongly for the quality of the papers which are being offered and the loyalty of subscribers, despite inevitable price rises. I should like to add also that the volumes with which I have been wociated owe much to a publishing team whose standards of professionalism and workmanship are quite outstanding. March, 1975 A. MACFADYEN

Contents CONTRIBUTORS TO VOLUME

9

-

V

vii

PREFACE

Predation and Population Stability W. W. MURDOCHand A. OATEN I. Introduction 11. Mostly Field Observations A. Evidence for Instability B. Evidence for Stabilizing Mechanisms C. Summary111. Stability Analysis A. Density-dependence in the Prey Population B. The Prey has a Refuge C. One Class of Prey is Invulnerable D. Spatial Heterogeneity E. Accelerating Functional Response F. Graphical Analysis G. SummaryIV. One-prey Species A. Functional Response in a Patch of Prey B. Predators’ Responses to Patchiness C. A Model of Predator Behaviour, and its Consequences D. General Criteria for Stability E. Other Studies of Patchiness F. SummaryV. Two-prey Species A. Relative Attack Rates B. Functional Response-Two-prey Species C. SummeryVI. Learning and Functional Response VII. Other Responses by Predators VIII. Concluding Remarks Acknowledgements References Appendix1 Appendix I1 -

ix

2 4 4 5 13 13 17 19 22 22 25 28 32

-

-

-

-

33 35 38 41 51 54 58 59 59 95 104 105 112 117 120 120 126 130

X

CONTENTS

Mathematical Model Building with an Application to Determine the Distribution of DursbanB Insecticide added to a Simulated Ecosystem G. E. BLAUand W. BROCKNEELY I. Introduction 11. Model Building Techniques A. Types of Mathematical Models B. Model Building Procedure C. The Design Problem and the Analysis Problem D. The Likelihood Approach to Model Discrimination E. Example of Model Discrimination by Likelihoods F. Parameter Estimation Procedures G. Tests of Model Adequacy H. Conclusion 111. The Environmental Fate and Distribution of DUMBAN@ Added to an Ecosystem A. Introduction B. Description of the Ecosystem C. Building theModel D. Discussion of Results E. Conclusion References -

133 134 134 135 138 139 143 144 145 148 149 149 149 161 100 162 162

The Pressure Chamber as an Instrument for Ecological Research GARYA. RITCHIEand THOMAS M. HINCKLEY I. Introduction A. Plant Water Status B. A Brief Historicd Perspective C. Objectives 11. Theory and Methodology A. Theoretical Consideratiom and Terminology B. Apparatus C. Procedures D. Calibration E. Precautions F. Memurements of P on Conifer Needles G. Use of the Pressure Chamber to Determine Osmotic and Matric Potentials H. Where to Sample 111. Review of Ecological Studies A. Some Physical Relationships B. Plant Responses to Supply and Demand C. Expression and Interpretation of Data D. P in Relation to Habitat E. P in Relation to some Plant Factors IV. Other Applications of the Pressure Chamber A. Pathology, Entomology, Pollution Effects B. Leaf Folding.in Legumes -

166 166 167 169 169 169 171 173 174 183 192 193 196 200 200 202 206 216 218 229 229 230

xi

CONTENTS

C. Water Relation of Roots D. Frost Hardiness E. Cultural Applications F. Other Applications V. Some Unresolved Questions A. Why Does P Fail to Meet the Gravitational Potential Gradient? B. Why is 0 Bars Never Achieved? C. Is There Substantial Resistance to Flow Between Leaf and Stem? D. Do Plants at Night Act as Tensiometers? VI. Concluding Statement Acknowledgements References -

23 1 232 232 233 234 234 236 236 238 240 243 243

The Ecology of Serpentine Soils JOHN

PROCTOR and STANLEY R. J. WOODELL

I. Introduction 11. Geology and Soils . A. Geology B. Weathering and Pedogenesis C. Clay Minerals -

-

D. A General View of Serpentine Soils

111. The Vegetation of Serpentine

~

25G 250 256 259 261 262 263 27 1 27 1

1V. The Reasons for Serpentine Infertility A. Physical Properties of Serpentine Soils B. Low Levels of Nitrogen, Phosphorus and Potassium in Serpentine Soils C. Nickel, Chromium and Cobalt in Serpentine Soils D. Calcium and Magnesium E. Other Unusual Chemical Features of Possible Importance to Plants V. Animals on Serpentine Soils VI. Fungi and Bacteria in Serpentine Soils VII. Evolution on Serpentine A. Eootypic Differentiation B. EndemicsC. The Exclusion of Serpentine Endemics from other Soils D. Plants Showing Disjunct Distribution on Serpentines E. Morphological Differences shown by Serpentine Plants F. Speciation VIII. Conclusions Acknowledgements References -

336 338 338 340 340 342 343 345 346 347 349 350 350

Author Index Subject Index Cumulative List of Titles

367 377 385

-

~

-

-

-

273 276 314

This Page Intentionally Left Blank

Predation and Population Stability w. w.

MURDOCH

and

A. OATEN

Department of Biological Sciences, University of California, Santa Barbara, California, U.S.A.

-

-

-

I. Introduotion11. MostlyFieldObservations A. Evidence for Instability B. Evidence for Stabilizing Mechanisms 1. Refuges for the Prey 2. Invulnerable Class of Prey 3. Spatial Heterogeneity C.Surnmary III. Stability Analysis A. Density-dependence in the Prey Population B. ThePreyhasaRefuge C. One Claas of Prey is Invulnerable D. Spatial Heterogeneity E. Accelerating Functional Besponse F. Graphical Analysis 0.SummaryIV. One-prey Species A. Functional Response in a Patch of Prey B. Predators’ Responses to Patchiness C. A Model of Predator Behaviour, and its Consequences D. General Criteriafor Stability . E. Other Studies of Patchiness F. SummaryV. Two-prey Species A. Relative Attack Rates 1. Predator Switching and Apostatic Selection 2. SearchImage B. Functional Response-Two-prey Species 1. Experimental Results 2. Models C.Summary VI. Learning and Functional Response VII. Other Responses by Predators- VIII. ConcludingRemarks Acknowledgements References . Appendix1 Appendix11 1

-

-

-

-

-

~

-

~

~

~

~

-

~

-

~

~

~

~

~

~

~

-

~

-

~

~

. .

-

~

.

~

~

~

-

-

-

-

-

-

-

-

.

.

-

-

-

2 4 4 6 6 6 9 13 13 17 19 22 22 26 28 32 33 36 38 41 61 64 68 69 69 69 80 96 96 98 104 106 112 117 120 120 120 130

2

w. w.

MURDOCH

and

A. OATEN

I. INTRODUCTION Although populations of prey and predators fluctuate in nature, at least some, and perhaps most, are in some sense stable: they persist locally for long periods of time. Furthermore, such natural populations fluctuate less violently than do pest species, and thus appear to be relatively stable. Although their densities and stability are affected by many factors in the environment, we can gain some insight by isolating the interaction between prey and predator and enquiring whether it tends to increase or decrease the stability of the system. That question itself is very complex and a complete analysis would include the effects of refuges, spatial heterogeneity, population movements, seasonal events, relative growth rates of prey and predator, their reproductive and death rates, the influence of prey and predator densities and age distributions upon attack rates, and so on. I n this paper we ignore many of these complexities and concentrate on only a few, giving brief mention to others in passing. Sections I1 and 111, which can be considered together as a separate unit, look a t some features that are in a sense “extra” or ancillary to the interaction itself. Our approach in those sections is as follows: given the fact that simple models of the predator-prey interaction are fundamentally unstable, what is the effect of adding some ancillary complications that have been found in real systems? Thus Section I1 describes some potentially stabilizing features that have been found in real systems (refuges, invulnerable classes of prey, spatial heterogeneity) while Section I11 adds these to simple predator-prey models and shows that indeed they are stabilizing. That section also touches. upon two obvious features of reality, namely time lags, which destabilize the interaction, and density dependence (e.g. resource limitation) in the prey, which is stabilizing. In Section 111we also place the functional response in the context of a general model (the Lotka-Voltema equations) and present a criterion for estimating its stabilizing effect. Perhaps the stability we observe in nature can be explained by features such as refuges, spatial heterogeneity and an invulnerable age class, and certainly we believe they contribute. However, our second and most extensive type of analysis (Sections I V and V) ignores these essentially external features, and asks whether the interaction itself is stabilizing, and in pa,rticular whether the short-term response of an individual predator to variations in prey density (the functional response) can be stabilizing (and, if SO, when and how). An attractive reason for asking this question is that the behaviour of an individual predator can change quickly in response to prey density, so m y stabilizing effect will operate with only a short time lag. Furthermore,

PREDATION AND POPULATION STABILITY

3

almost all of the simple experiments done on functional response suggest it is destabilizing, so there is an interesting challenge to see whether we can recognize those complications that are present in real systems that affect the predator’s behaviour and thus might alter the response so that it becomes stabilizing. Here we show that either of two complications can indeed change the response in this way. The first is the fact that prey may occur in discrete patches in space. We show that the need for the predator to move between patches is stabilizing. Secondly, the presence of another one (or more) species of prey combined with a change in the predator’s behaviour (“switching”) can also be stabilizing. I n discussing each of these two ideas we first summarize the existing experimental and observational evidence and then provide a mathematical model. Since the switching idea is similar to the idea of “search image”, familiar to ornithologists, we devote a section to discussing search image and try there to clarify some misunderstandings concerning the relationships among search image, switching, and functional response. Thirdly, in two very brief sections (VI and VII) we discuss the relationship between learning and functional response and indicate the importance of two features (developmental response and numerical response) of the interaction that we have omitted in the rest of the paper. Two thorny problems have not been discussed in detail, but deserve a mention. First, we make no further effort to be more rigorous about what we mean by “stability” in field populations, relying mainly on the reader’s own experience and intuition, but we are quite explicit in defining criteria for stability in the models and experiments. (Roughly speaking, stability in the general models is “return to equilibrium after perturbation”, and in experiments and models dealing with functional reaporwe, stability is equivalent to requiring that the attack rate of the predator increase faster than proportionately as prey density increases.) We wsume, as do most ecologists, that these two concepts of stabilitya rough and ready description of the field situation, and the rigorous mathematical definition-are related, and that features that lend stability to models will also tend to add stability to field populations. Second, we do not approach in any formal way the possibility that natural predator-prey systems may owe much of their stability to a long shared evolutionary history: the prey species we see are those that have evolved sufficient defense mechanisms for some individuals to survive in each generation, while the predator species likewise have oharacteristics that assure some individuals an adequate harvest of prey. The models we use are simply not designed to compare the stability properties of pairs of non-co-evolved species with pairs that are co-

w. w.

4

MURDOCH

and

A. OATEN

evolved; they do not specify enough characteristics of the organisms to represent anything other than an essentially arbitrary pairing of species. I n this way they are perhaps more analogous to laboratory and agricultural predator-prey systems than to “natural” interactions. We do not discuss in any general way the relationship between the number of species in the interaction, and the system’s theoretical stability, for the very good reason that this field has been covered in a series of elegant papers by May. References can be found in May (1973b). Finally, while we intend that the various sections of this paper should relate to each other, we have tried to write each major section so that it can be read alone, independently of the others.

11. MOSTLY FIELDOBSERVATIONS A.

EVIDENCE FOR INSTABILITY

We know from field studies that predators can drive their prey to extinction, a process usually seen either for brief transitional periods or, where it is not transitional, as a recurring phenomenon over a part of the prey’s range. I n this latter case the predator accounts for at least one edge of the prey’s distribution, or, in patchy habitats, for its presence or absence. The best documented examples come from aquatic environments. For example, Brooks and Dodson (1965)noticed that the larger zooplankton in Crystal Lake, although present in 1942 before the alewife fish (Alosa) was introduced, were missing by 1964, presumably owing to the fish’s depredations. Some larger species of zooplankton were totally absent and others were greatly reduced. This change in time was similar to differences between two groups of lakes in Connecticut, those without alewives being dominated by large Cladocera such as Daphnia sp. and calanoid copepods such as Diaptomus sp., while those lakes with alewives were missing these species and instead were dominated by small cladocera (e.g. Bosmim sp.) and copepods (e.g. Cyclops). Perhaps the most dramatic case of predation almost driving a population t o extinction is the effect of lamprey upon the fish populations in the Great Lakes (Baldwin, 1964). I n this caae again, the predator was introduced into the system by man. Some strong experimental evidence has recently been produced in set of artificial ponds at Cornell University (Hall et al., 1970). One set of ponds had fish added and another had no fish, with striking effects. I n one year of the experiment, the large zooplankter Ceriodalphnia contributed 63% of the zooplankton biomaas in the absence of fish, and

PREDATION AND POPULATION STABILITY

5

only 3% when fish were present; other, smaller, species increasing concomitantly. Interestingly, while the species composition changed, the overall biomass of zooplankton did not. (The manipulation of invertebrate predators had little effect on the prey species. But the actual manipulation may have been less successful in this case.) These striking examples of the local extinction or virtual extinction of a species from an area presumably are relatively rare events, and it is perhaps significant that those that we know about have resulted from man’s intervention. Connell (1970) has demonstrated with convincing experiments that the lower limit to the distribution of a barnacle (Balanusglandula)on the rocky intertidal in Washington is determined by predation, mainly by a variety of small predators. On San Juan Island (off the Washington coast) the barnacle settles throughout the shore but survives only on the upper shore. Each year all the mortality on the lower shore can be attributed to a whelk, Thais. The mechanism was demonstrated by excluding the whelk from small cages low on the shore, in which B. glandula then survived. It appears that higher on the shore the snails, which take several hours to drill into a barnacle, do not have time during high tide to complete an attack. Kitching and Ebling (1967) provide other examples of predation’s setting limits to the distribution of sessile animals on the seashore and, extending predation to cover grazing by herbivores, they show that sea-urchins determine the local distribution of algae by driving it extinct in local patches. Whether such extinction and determination of distribution is more common in aquatic habitats, or simply more conspicuous there, is not clear. Predators may also increase the instability of systems that nevertheless persist. For example, Varley and Gradwell’s (1968) study of the winter moth suggests that two parasites operated with a time lag, which if true would cause the moth numbers to oscillate. Instability can also arise, of course, as a result of the prey population’s escaping from the control of the predator. Morris (1963) suggests that the spruce budworm has become a pest in Canada at regular intervals, even though it is a native species,because a combination of suitable conditions allows it to increase so rapidly that the predators cannot keep up. B.

E V I D E N C E FOR S T A B I L I Z I N G M E C H A N I S M S

It is almost a tautology that the predator-prey systems that we see are those that are, in a sense, stable-they have avoided extinction. This must often result from the fact that the two species have evolved together, the prey evolving defense mechanisms that allow a fraction of

6

w. w.

MURDOCH

and

A. OATEN

the population to avoid being eaten, the predator developing attacking mechanisms that allow it to reap a harvest. I n natural communities the evolutionary mechanisms that produce stability through co-adaptation may be subtle and hard to find. For example, they may reside in the size, behaviour and physiology of some prey individuals that become invulnerable, and these characteristics may be only marginally different in the prey that are vulnerable. Perhaps the most obvious phenomenon that needs explanation is the persistence of the prey, since we know that predators can drive prey extinct. We discuss two mechanisms that produce such persistence, the presence of a refuge for the prey and the presence of an invulnerable class of prey. 1. Refuges f o r the prey The simplest mechanism that explains how some prey can exist is the existence of refuges. The barnacle B abnw glandula is a good example (Connell, 1970). The whelk Thais eats all the barnacles that settle low on the shore each year. But higher on the shore the period between low tides is too brief to permit a whelk to complete drilling a barnacle. The barnacles above this height are therefore in a refuge that provides settlement to the entire shore and food to the whelk. Other examples can be found in Connell (1972).

2. Invulnerable class of prey The following two examples, a moose-wolf system and a barnaclepredator system, may illustrate the importance of the existence of an invulnerable class in the prey. The moose-wolf story may also show that predators can stabilize a herbivore population that otherwise would be unstable, though the evidence is by no means overwhelmingly convincing. The moose-wolf interaotion on Isle Royale (Mech, 1966; Jordan et al., 1967) is an example of that archetypal predator-prey system, the large predator and its ungulate prey. Mech records estimated and guessed fluctuations in abundance of the moose from their colonization of the island in the early 1900’s until the present. There is evidence from population declines, emaciated carcasses, and severely eaten browse that on two occasions the population over-exploited its food supply, crashed, and then subsequently increased. Thus, before the wolves reached the island around 1948 and reached a steady density of about 22, the moose population was unstable, i.e. it fluctuated violently and dangerously over-exploited its food supply. Between 1967 and 1966 the food supply seems to have increased, for a fairly high moose population was being supported (possibly 800-1000) but there was no evidence of over-browsing.

PREDATION AND POPULATION STABILITY

7

From the later 1960‘s or 1960 until the most recent estimate, the wolves have been stable, and so have the moose. This does not constitute a very convincing experiment; it has no controls or replication. But the observations of Mech and Jordan at least provide a plausible case that the new stability is caused by the wolves. The wolves take young ( < one year old) or older debilitated ( > six years) individuals. The one t o six year age group seems to be invulnerable to predation, because they can fight off attacks and outrun the wolves. It is clear that the existence of an essentially invulnerable prey class is not the whole story. First, the moose must be partly controlled by the rate at which they can take in food, and in this respect Mech suggests that the food supply probably increased in recent years. Furthermore, there remains the question: why do the wolves not eat all of the newborn each year, thus preventing any moose from becoming invulnerable?Partly the answer must lie in the size of the wolf population, which is determined partly in turn by the rate at which vulnerable prey become available, partly by the abundance of alternative prey in spring, and partly by their own social dominance behaviour (there are dominant males and very little reproduction). Nevertheless, a case can be made that the invulnerable class of moose contributes to the persistence of the moose population. That this story should be the main worked-out example of a predator stabilizing its prey in a natural situation is strong evidence that we are in great need of field studies that demonstrate the phenomenon and that analyze the mechanisms by which it occurs. Connell (1974) has illustrated the significance of an invulnerable class in a barnacle species that contrasts nicely with B. glandula. The larger individuals of the barnacle, Balanus carioswr, coexist on the shore with several species of their predators and are not eaten out every year aa is B. glandula. The prey become invulnerable by becoming too large for the predator; once a B. cariosus individual is two years old the predators cannot kill it. But all younger barnacles are vulnerable. As a consequence B. cariosw escapes through the predation bottleneck only in unusual years-when the predator populations are temporarily reduced by severe physical conditions. Thus, the interaction produces dominant year-classesin the prey. Even though survival is intermittent, the total population of B. carioswr is relatively stable (Fig. 1). It is well known that many predators are quite selective in the sizes of prey they eat, and that the size taken increases with predator size, so we might guess that one invulnerable prey size (either large- or small-size classes) will often explain prey persistence. Connell’s barnacle study is particularly illustrative, since the system was upset in 1967 when a new predator (a large starfish species) appeared on the scene. The large

w. w.

8

MURDOOH

and

A. OATEN

B. curiosus can be eaten by the starfish, which quickly wiped out the older year classes of barnacles (Connell, 1974). A couple of points are worth noting in these two examples. First, the invulnerable class probably is the age group with the highest reproductive value. Second, we have accounted for only one aspect of the stability-why the predator does not eat all the prey. I n the case of the wolves, perhaps they can handle any “excess” moose that are produced, thus also preventing the prey’s increase. Wolf social behaviour is also a significant feature, at least in stabilizing the wolf population. Finally, this kind of invulnerability is rather similar to a refuge, except that afraction of the prey population, rather than a fixed number, is

90

-

0

60 /

c

63

al

30 N

-----__ o‘o-o-o--o

55

‘

O ~ O - O - o - -

O L

4 5 59 60

61

62

63

64

65

66

67

68

69

70

71

Years

FIQ.1. The percentage of the surface covered by the barnacle Balanue c a r i o m at the mid-tide level on the rocky shore of San Juan Island, Washington, U.S.A., from 1969 to 1971. The numbers 50, 55, 63 refer to year classes of barnacles that escaped predation and constituted the invulnerable classes. No data were obtained in 1963 and 1964 (dotted curves). Starfish reached the area in 1967 and subsequently ate most of the barnacles. Reproduced with permission from Connell (1974).

safe. The fraction is likely to vary because the range of individuals that fall into the invulnerable class may change with prey and predator densities, prey food supply and so on. The analogous mechanism in plant-herbivore interactions may be the evolution of chemical and other defenses that make parts of the plant a t least temporarily inedible (Whittaker and Feeny, 1971). Examples of predators adding stability can be found in the biological control literature. Unfortunately, successful caaes of biological control generally are not studied and explained, their having worked apparently being reward enough. Again, the examples are not particularly convincing or very enlightening since the mechanisms are not well understood. Icerya, the cottony cushion scale, persists in California orchards

PREDATION AND POPULATION STABILITY

9

and has done so together with its predator Vedalia and parasite Cryptochaetum for 80 years or so. It does so at very low densities, and almost all of its mortality is caused by these enemies (Quesada, 1969). It is not known whether local populations are stable or appear intermittently, but the system as a whole appears to be more stable than was the prey alone, and more stable than when the predator is accidentally reduced by insecticides. Hypericum (Klamath weed) and its enemy, the chrysomelid beetle Chrysolina, provide a similar situation (Huffaker and Kennett, 1959), again including herbivory under predation. 3. Spatial heterogeneity A third set of mechanisms involve spatial heterogeneity. It seems likely that this feature of natural ecosystems will go a long way to explaining their stability, though there is not much field evidence concerning predator-prey systems. Some fairly good evidence comes from analogous herbivore-plant interactions. For example, in the balsam fir forests of Canada there appear to be three circumstances in which trees are spared devastatingly heavy attack by the spruce budworm (Morris, 1956, 1963). First, when susceptible trees are among other trees of a non-susceptible age; second, when they are among hardwood species; and third, when they are in isolated stands rather than in continuous swathes. Our interpretation of these different kinds of heterogeneity (among individuals, among species and in space) is that they all function in a similar way, as barriers to the rapid dispersal of the “predator”. Perhaps there is more than one way in which such reduced dispersal can stabilize the system. One interpretation is that it can produce a mosaic of sub-systems out of phase with each other, since extermination of prey at any one time is likely to remain localized. Alternatively the heterogeneity may serve as a sort of relative refuge, making susceptible individuals harder to find. Thus two stabilizing mechanisms may be operating in the trees: an invulnerable age class and heterogeneity, i.e. intermingling of the susceptible and the unsusceptible classes. Probably the existence of invulnerable age classes of tree is the main feature that prevents local extinction of tree populations, though the spatial heterogeneity seems t o help. One normally thinks of spatial heterogeneity operating by producing differences among spatially separated sub-systems within the system. These differences should cause fluctuations in different sub-systems to have a different frequency or to be out of phase with each other, and when some movement of predator and prey individuals occurs between sub-systems, the whole system might thereby be stabilized. It would

30

w. w.

MURDOCH

and

A. OATEN

really be useful to have a field example illustrating this hypothesis, which has obvious intuitive appeal. Since any population owes its existence in part to chance, any form of patchiness, anything that partially uncouples the fate of one segment of the population from that of another, should decrease the probability of simultaneous bad (or good) times for all segments. The only example we know of has several factors causing significant prey mortality, though predation may be the most important. Landenberger (1973) has studied mussel clumps on pier pilings on the coast of Southern California over seven years. Landenberger’s pier had dozens of pilings with very variable sizes of mussel clumps, and some had no clump. The mussels are attacked by starfish, which not only eat mussels but also weaken the clump. When storms or even heavy swells come, clumps that are particularly large and/or heavily attacked by stariish fall from the piling to the sandy b o t t o m t h e clump becomes extinct or severely reduced in size. Starfish tend to stay where there are clumps, and there is a time-lag before they leave an area when the food is gone. Thus, patches of mussels are continually becoming extinct, or greatly reduced in size, but at the same time some clumps are growing. Except in so far as settlement is seasonal, clumps are out of phase with each other in their waxing and waning, so that there is a very low probability that all clumps become extinct together. There is also immigration (i.e. planktonic larvae) from clumps elsewhere to the empty spots, so the set of interacting patches is actually much larger than a single set of pier pilings. Huffaker’s (1958) famous orangelmite study probably provides a laboratory example of this sort of spatial heterogeneity, although there are some problems in interpreting it in this way. First, it is really not clear that in his complex systems the different parts were out of phase; changes in density seemed to be more or less synchronous throughout the system. I n addition, the initial distribution of mites, as well as the physical structure, was changed in the more complex treatment. The main mechanism may have been slowing up the predator’s dispersal relative to the prey’s. Finally, a laboratory study by Luckinbill (1974) suggests the existence of a mechanism similar to the spruce budworm situation, but physical structures play no part. (See also an interesting study using Protozoa by Salt (1967).) Luckinbill has managed to get Didinium and Paramecium to persist together in the laboratory, without adding any structural refuges, by reducing the frequency of encounters between prey and predator by slowing up the movements of both species by increasing the viscosity of the medium, using methyl cellulose. This would not allow persistence, however, if the predator population’s

PREDATION AND POPULATION STABILITY

11

capacity for attack could become very high at low prey density. This was prevented by lowering the productivity of the system. Luckinbill did a series of experiments in which he showed the following: 1. Under “normal” circumstances, with Didinium attacking Paramecium which in turn was feeding on bacteria in cerophyll, Paramecium

is driven extinct very quickly (in about 10 hours), as Gause (1964) observed. 2. When velocities were slowed by making the medium slightly viscous (by adding methyl cellulose),the system persisted for one or two weeks, and several oscillations occurred, but the predator always became extinct because it drove the prey to very low densities, then crashed. In this case, where the prey persisted, their minimum density became too low t o maintain the Didinium population. 3. I n the viscous medium, when the food supply for Paramecium was increased, their average size increased, but not their total biomass. As a consequence Didinium was able to increase more rapidly and reached higher densities, driving the prey to extinction. 4. Finally, persistence of predator and prey in the viscous medium was obtained by halving the rate at which bacteria were given to Paramecium. I n this case Paramecium density was not much reduced, but at high density individuals were thinner. Peak numbers of Didinium were not much reduced, but they were less well fed and apparently therefore did not attack the prey so heavily during prey troughs, so that the prey population recovered more rapidly and did not become so sparse as previously. Although the mechanisms are not well studied in the example above, it may serve as a model for field situations in which physical barriers that reduce the frequency of contacts and the rate of dispersal of the predator, when combined with some mechanism that limits the maximum number of predatory attacks (either by limiting predator numbers or quality), can maintain both populations. There are field studies that suggest that the predator-prey interaction can be stabilizing, but they do not demonstrate what the stabilizing mechanisms are. For example, Varley and Gradwell (1968) have shown by key factor analysis that predation upon the pupae of the winter moth, probably by carabid beetles and other invertebrate predators, is density-dependent. Hassell (1969) has shown a similar effect upon Cyzenis, one of the parasites of the winter moth. Harcourt and Leroux (1967) summarized key factor analyses of 12 species of insect pests in Canada. I n six of these studies, predation (including parasitism) was the key factor and in each case the predation was stabilizing. But for

12

w. w.

MURDOCH

and

A. OATEN

none of these six studies do we know whether such things as refuges or heterogeneity are important, or if the stability arises purely from the way the predators respond to variation in prey density, for example via their numerical or functional response. Field studies are also available that show that such responses do occur (Sections IV and V), but their ability to stabilize the predator-prey interaction has not been demonstrated. If we accept the interpretation of the examples presented in this section, then the field observations provide us with a phenomenon to be explained : some prey populations that are otherwise unstable are stabilized when a predator population is added (the moose-wolf example and perhaps some from biological control). What characteristics of the predator-prey interaction (i.e. features lacking in the prey system alone) lead to this stability? This puts the question in its most positive form. Next we develop a rather weaker statement of the phenomenon-to-beexplained. There is good evidence that in natural communities prey populations persist for long periods together with the species that prey upon them. They are stable in this sense. They conform to stricter criteria of stability also; they appear to fluctuate much less than prey populations in agro-ecosystems, and at least some of them show rather small changes in abundance (Richards, 1961 ; Varley and Gradwell, 1963). By contrast, we presented evidence above that some predators are capable of causing extinction of their prey populations. A weaker statement of the question is therefore as follows: (1) How do some prey persist in the presence of their predators-why are they not driven t o extinction? The implication here is that there are one or more mechanisms offering protection of some sort to the prey, which suggests a second question: (2) Why do prey not become excessively abundant so that they over-exploit their food supply or other resources as occurs in artificial systems? (In some simple systems, like the seashore, of course the resource, namely space, is not over-exploited but simply used to full capacity.) There may be different answers to each question, and indeed the second may have more to do with the prey-food interaction than the predator-prey interaction, i.e. question ( 2 ) may really be question (1) at a lower trophic level. For example, possibly natural vegetation is often resistant in some way to defoliation. But we will also consider the possibility that predators are responsible for stabilizing the prey at high densities. I n Section I11 we discuss briefly how the mechanisms described in this section can be incorporated into the general theory of predation. I n both sections the emphasis is on stabilizing mechanisms (refuges, invulnerability, heterogeneity) that are ancillary or additional to the

PREDATION AND POPULATION STABILITY

13

“pure” interaction between predators and their prey. By “pure” we mean interactions that arise from changes in numbers of prey or predators, or from changes in the attack rate of individual predators in a system with no refuges, no spatial heterogeneity, no differences among individuals and so on. This pure interaction is the concern of the remainder of the paper after Section 111.

c. S U M M A R Y There is evidence from the field that predators can cause prey populations to be unstable, and even drive them extinct, though the latter seems either to occur in only part of the prey’s habitat or to be caused by man’s disruptive activities. By contrast, many predators and their prey persist together in nature over long periods of time. Some stabilizing mechanisms that can account for this and that have been described in the field include prey refuges, an invulnerable class of prey, and spatial heterogeneity, including barriers to dispersal. Some laboratory work suggests that simply reducing interactions between predator and prey can be stabilizing. However, we have no explanation for the stability of most systems and in particular do not know whether predators can stabilize their prey in the absence of the mechanisms listed above.

111. STABILITYA N A L Y S I S One way to explore the stabilizing properties of various mechanisms or relationships is to insert them into a simple model of a predator-prey system. The classical Lotka-Volterra model is a useful vehicle for this procedure since, in its simplest form, it describes a system that is balanced on the knife edge of neutral stability: the system oscillates but the oscillations are neither damped nor expanding. Any changes in the system will tip it over into either stability or instability. I n this section we briefly describe that system and the consequences of changing some of the assumptions, in particular those involving the mechanisms we discussed in the previous section. I n addition, this analysis helps to put into perspective our later material on functional response and switching. I n the Lotka-Volterra model the rates of change of the prey and predator populations are written

dH - = aH-bHP at d_P - -CP+dHP at

w. w.

14

MURDOCH

and

A. OATEN

where H is the prey (or host) density and P is the predator density. The instantaneous rate of increase of the prey, in the absence of predation, is a H . Each predator is assumed to take a constant fraction b of the prey population and there is no interaction among predators. Thus the number of prey killed is a linear function of both prey density and predator density. The predators have an “accidental” death rate c P . Prey are transformed t o predators at an efficiency, say e, that is independent of prey and predator densities, so that the number of new predators added is e ( b H P ) = d H P . There are no time lags. We now examine the stability properties of this system. There is an equilibrium with both H and P greater than zero when aH - b H P = 0 and - cP + d H P = 0 . The equilibrium values are P* = a/b and H* = c / d . We then want to know what happens when the system is perturbed, i.e. when the densities are moved away from their equilibrium values. The effect of small perturbations on the rates of change is determined as follows. We let H = H*(1 + h ) and P = P*(1 + p ) so that h and p give the amounts, as fractions of H* and P* respectively, by which H and P deviate from equilibrium. Since H* and P * are constant, dH dP dP - = H* ah - and - = P * -, so that ( 1 ) can be re-written at at at at

H*

dh

=

P * dP -= at

uH*( 1 + h)- bH*P*( 1 + h)(1 + p )

- cP*(1 + p ) + dH*P*( 1 + h)(1 + p )

(2)

Cancelling H* and P * where possible, replacing them by c/d and a / b respectively otherwise, and expanding, we can reduce (2) to

dh _ -- - a p ( l + h ) at dP - = ch(1 + p ) at

(3)

For small deviations from equilibrium, both h and p would be much smaller than 1 , so their product can be ignored. Thus (3) becomes, approximately,

Thus, by examining small perturbations, we obtain two simple linear simultaneous equations with two unknowns which can therefore be solved. A method of solving such simultaneous first-order linear

PREDATION AND POPULATION STABILITY

16

differential equations is outlined in Appendix I. From (A9) of this Appendix, we see (since r = v = 0, 8 = -a and u = c) that (4) gives solutions which, aa functions of t , oscillate with constant amplitude, and period v/2/ac. These well-known results are displayed in Fig. 2 in the usual phase diagram. The Lotkrt-Volterra mods1 is of course a gross oversimplification, though Lotka (1925) and others (MacArthur, 1970; May, 1971) have

H

FIG.2. Each point in the plane represents the density of the prey (H) and the predator (P)populations.The ellipsesare solutionsto the Lotka-Volterra equations and show neutral stability; in each solution the populations cycle at constant amplitudes and frequencies.

noted that its stability properties are representative of a rather wide range of less simple models. Perhaps the most glaring omission is the absence of time lags, and these well illustrate how delicate is the model’s neutral stability. Time lags always tend to make the system unstable. This is illustrated graphically in a simple way in Fig. 3. Time lags can be incorporated into the model simply by writing the equations indifference rather than differential form:

Ht+, = aHt-bHtPt Pt+l = - c P t - d H t P t Figure 3 corresponds to a model of this sort. Bartlett ( 1957) incorporated time lags into the differential equations, in a more specific manner, by assuming that both predator and prey take a fixed time to grow to maturity. Then

16

w. w.

MURDOCH

and

A. OATEN

Notice that the predators attack at a rate appropriate to current prey density and die at a rate determined by current predator density. The new prey added at time t are those born t o prey which were alive at time t - Atl, while the rate of change of the predators at time t is determined by the number of interacting predators and prey at time t - Atp Another realistic source of instability is random variation. However, from several studies, it seems that the stability of the model is not very sensitive to stochastic variation. Bartlett (1957) showed that stochastic variation in the simple Lotka-Volterra model, although it destabilizes

r

P

H

FIG.3. A diagrammatic representation of lag causing instability in the predator (P)and prey (H) populations. The ellipse is a neutrally stable solution without time lags. A lag of At causes the system to have the vector at time t + At that is appropriate at time t , so the system spirals outwards along the dotted lines, the oscillations increasing in amplitude with time. The appropriate vector a t t + At is indicated by the solid line tangent to the ellipse a t that point. 1

the system and drives it extinct, does so to a rather small degree, so that the probability of extinction in any short period is very small. Thus time lags are more important destabilizers than is stochastic variation. This conclusion will be reinforced below when we discuss Leslie and Gower’s (1960) and May’s (1973a)work with time lag and stochastic variation in models with density-dependence. Stochastic variation probably becomes more important when the stable solution is a limit cycle that has values close to the axes in the phase diagram (Gilpin, 1972; Rosenzweig, 1972).

We now turn to examining some ways of changing the LotkaVolterra model so that it becomes stabilized. This is done in a somewhat loose way. Because of the mathematical difficulties, we do not always try to incorporate stabilizing factors into ( 5 ) or any other version containing time lags. Rather, we incorporate these factors into the

PREDATION AND POPULATION STABILITY

17

system (l),which we then linearize by the use of the Taylor expansion and the assumption that, since we are concerned only with relatively small perturbations from equilibrium, we need consider only first-order terms. Then we will claim our introduced factors are “stabilizing” if the linearized systems satisfy the stability criteria of (AS) and (A9) of Appendix I. A.

D E N S I T Y - D E P E N D E N C E I N THE PREY POPULATION

I n the model in Eqn (1) the prey increase exponentially when there are no predators. The model can be stabilized by adding an upper limit

P

H

FIU.4. Diagram of the Lotka-Volterra equation with the prey density-dependent. The system is stabilized. Oscill&tionsdecrease in amplitude with time aa the system approaches the equilibrium.

t o prey density (Fig. 4). This limit is incorporated into the equation by

making the prey growth term logistic:

dH - = aH-mH2-bHP at where m = a / K and K is the usual “saturation density” that appears in the logistic model. An early version of this model was introduced and discussed by Leslie (1948, 1968). Leslie’s work on prey-predator equations seems to have had less than its share of attention, and we hope our discussion here may help bring it the notice it deserves. Leslie stresses the fact that there will be an upper limit to the rates of increase of both prey and predator, upper limits not recognized in the Lotka-Volterra equations. These upper limits are approached under favourable conditions: for the predator, when the number of prey per predator is large; for the prey, when the number of predators (and

18

w. w.

MURDOCH

and

A. OATEN

perhaps the number of prey also) is small. These considerations lead him to the difference equation model:

which Leslie suggests is made more realistic by the inclusion of a logistic term for the prey, so

(We have changed Leslie’s notation to conform with ours and with Leslie and Gower (1960).) These equations are rather different from the usual Lotka-Volterra system, but there are similarities: if?! , and a1 are very small, we can 1 1-r ( 7 ) : Ht+,=X,Ht( 1 - PHt) - A,a,HtPt, which is the LotkeVolterra equation with a logistic term added. Also, Leslie shows that the continuous time version of (7) is

use the relation - = 1 + r + r 2 + . .. e l + r (for r small) to get, from

Again, the first equation is the Lotka-Volterra equation with a logistic term added. The second equation is less familiar, and does not fit our usual notions of the predator converting captured prey into new predators. Nevertheless, the equation is in standard logistic form, with the “carrying capacity” of the predator’s environment being some proportion (r2/a2)of the prey numbers. I n addition, Leslie claims it provides a good fit to some data of Gause (for detailed discussion see Leslie (1957, 1958)). I n any case the system (7) is one of the simplest having maximum growth rates which each population approaches under (different) favourable conditions. The differential equations, when linearized near equilibrium, yield a stable system, according to the criteria of (A8) and (A9). Leslie does not discuss the difference equations, (7), separately, relying on their similarity to the differential equations. However, he does give a numerical example which seems very clearly to return to equilibrium by a series of damped oscillations. Equation (6) and its continuous time analog are also discussed in these contexts, and also give stability, though the return to equilibrium is slower than for (7).

PREDATION AND POPULATION STABILITY

19

Leslie and Gower (1960) considered stochastic versions of these models, simulated on a computer. Survival of the system appears to have depended quite heavily on the choice of parameters, since the variances of Ht and Pt, and their covariance, behave roughly like linear functions of H* and P*, the equilibrium values. Consequently, in a very small environment, where H* and P * are low (around 5 to lO)--e.g. because ,fll and a2,which are roughly the effect of the predator on the prey and of the prey on the predator, are high-extinction can be highly probable and can occur quite quickly. When H * and P * are larger (around 100) the system survives almost indefinitely in the presence of stochastic variation. May (1973a) also shows that the equations with a logistic prey are sensitive to time lags.

B. T H E

PREY HAS A REFUGE

When we come to incorporating the mechanisms observed in the field or laboratory (Section 11)into the LotkeVolterrrt models we have to make simplifications, and this generally involves choosing between possible simplifications. The two obvious candidates with respect to a prey refuge are (a)a constant number of prey are safe and (b) a constant fraction is safe. We examine these in turn. Connell’s barnacles (Section 11)have a safe zone on the shore, and the assumption of a fixed number of refuges is perhaps not too outrageous in this case. St Amant (1970) has shown that such a refuge always stabilizes the Lotka-Volterra model, both for the differential and difference form. For the differential case we write dH - = aH-Pb(H-k) at

dP

= at

-cP+Pd(H-k)

where k is the fixed number of prey in the refuge. For the stability analysis we first find the equilibrium values by solvingaH-Pb(H-k)

= Oand -cP+Pd(H-k)

=

C

OtogetH* = k + -

d

adk a and P * = -+-. Taking H = H * ( l + h ) and P = P * ( l + p ) we get, bc b after eliminating second-order terms (i.e. terms involving the product dh adk dP = - -h + up, and - = - dH*h. Referring to our criteria for hp), C at stability of such systems (Appendix I, assertions (As) and (A9)) we see

20

that “r+v” = T - v

w. w. =

MURDOCE

-

adk

-

and

and A. OATEN 8u =

C

-adH*. Thus, whether the

( i)

system oscillates depends on whether 4ad k+-

>

(?)2;

but if

it does, the oscillations have decreasing amplitude, since “T-V”

=

-adk/c
and if it does not, then (As) (b) holds, and I, and h decrease steadily to zero without oscillation. I n either case, the system is stable. It is rather more difficult to find real systems that approximate to the assumption of a fixed fraction in the refuge. If one class of prey were invulnerable (Section 11), and that class formed a constant fraction, then this would be mathematically the same as having a constant fraction in a refuge. But when, say, an age class is invulnerable, the prey is unlikely to have a stable age distribution and the fraction that is invulnerable will therefore vary. Also, vulnerability is likely to vary with prey and predator density. I n the absence of a good motivating field example we simply proceed with the analysis of fractional refuge. Leslie and Gower also consider the case where a constant fraction, 1 - k, of the prey is safe from attack, and found that their very small systems which had previously gone rapidly to extinction (see discussion above) were now able to survive for long periods for suitable values of k (in their case, k near 0.2 or 0.3 seemed optimal). With the fraction in the refuge at 0.8,extinction time increased 3000-fold over that of the model with no refuge. Bailey et al. (1962) also looked at the effects of a fractional refuge in a difference equation model, in this case for a parasite and its host. Since the difference equation has time lags, the model without refuges is always unstable unless density-dependence or some other stabilizing feature is added. When Ht+l is the number of hosts surviving parasitism and Pt+, is the number of parasites emerging, the basic NicholsonBailey model without a refuge is

Ht+,

=

FHte-aPt

Pt+, = P ( H t - Hte-aPt)

(9)

where a is the “area of discovery” (i.e. the fraction parasitized per parasite) so that, allowing for multiple parasitism and assuming a random distribution of attacks among prey, e - W is the fraction of hosts at time t escaping attack by the Pt parasites. F is the host’s rate of increase in the absence of parasitism. The area of discovery here is equivalent to b, the proportion of the prey killed per predator, in the Lotka-Volterra formulation. Notice that it is assumed that the

PREDATION AND POPULATION STABILITY

21

generation times of the parasite and the host are precisely the same and that at each generation they coincide at exactly the same point in time. Bailey et al. then add to this model a fractional refuge, i.e. there is a fixed fraction r0 of the prey that is inaccessible to the parasites. They then show that stability is possible only for a combination of a narrow range of values for F, the hosts’ reproductive rate, and a narrow range of r o (i.e. 1- k in Leslie’s notation), the inaccessible fraction. Stability is also possible when all hosts are accessible but vary in their degree of accessibility, again for a narrowly defined set of combinations of F and “the percentage of least accessible hosts”. Stability in this situation is more likely when the least accessible class of hosts is much more abundant than other classes. However, the fraction of inaccessible hosts that yields stability is severely bounded both above and below, as is the host’s power of increase, P. The basic model is modifled to incorporate variable accessibility by adding a probability density distribution, f(a), which specifies the distribution of accessibilities among prey: ~

t

=+F

H~ ~ e - e t f ( a ) h 0

Pt+,

= F&-Ht+1

(10)

Hassell and May (1973) build upon a base similar to that of Bailey al. (1962). They also start with the basic Nicholson-Bailey model, but they assume that hosts vary in their degree of accessibility according to whether the area they are in has many other hosts or few other hosts; since parasites are attracted to dense patches (see Section IV), the average host in a sparsely populated area may be less at risk, analogous to Bailey et al.’s lower accessibility. Where Bailey et al. employ a probability distribution, f (a),to designate relative accessibility, Hassell and May simply assign different proportions (a1)of the hosts to different patches, and this distribution of hosts remains fixed through time. Hosts in different patches then have different relative chances of being attacked because the parasites aggregate at denser patches according to the rule /It = cap$, where /It is the proportion of the parasites at the patch with a4 of the hosts. Positive values of p give aggregation. (This point is discussed in more detail in Section I V E.) The actual distribution of prey incorporated into the model is extremely patchy: hosts are abundant in one patch (highly “accessible”) and rare everywhere else. Hassell and May then show: et

1. Increasing parasite aggregation (i.e. increasing p ) increases stability. Increasing aggregation of the parasite is equivalent to making the hosts B

22

w. w.

MURDOCH

and

A. OATEN

in the sparse patches more and more inaccessible, i.e. it moves the system nearer to a fractional refuge. 2. Greater stability ensues if the hosts that are in sparsely populated patches are distributed among more patches. This operates to the same end as the first mechanism. 3. Stability is increased if the easily accessible (dense) fraction of hosts is close to one-half. If a large fraction is in the dense patch the system is unstable when the host rate of increase is low; if the fraction is small, stability is possible only when the host has a low rate of increase and the parasite is strongly aggregated. 4. Stability is possible only for a very narrow range of (low) rates of host increase. Smith (1972) noted that prey will vary in the ease with which they are captured, and that under some circumstances this might be related to prey density. This would operate somewhat like a refuge, with a variable fraction being difficult to find and catch. Where, as Smith suggests, the fraction increases as prey density declines (because the remaining prey are difficult to get), this will have a stabilizing effect on the interaction.

c. O N E

CLASS O F P R E Y I S I N V U L N E R A B L E

The situation where a single class of prey is invulnerable (Section 11) has not been modelled but is mathematically somewhat analogous to the presence of a refuge. If the dynamics were such that the invulnerable class were always a constant fraction of the prey population the analogy would be exact. However, we expect that the invulnerable fraction would not be constant, except in the very unlikely circumstance that the prey are maintained at a stable age distribution (assuming it is an age class that is invulnerable). The invulnerable age class can be added to the model by treating it as a second prey population which is fed from the initial population, and feeds back into the population. Since we have not completed the analysis of such systems (Oaten and Murdoch, 1 9 7 4 ~we ) will not discuss this problem further.

D.

SPATIAL HETEROGENEITY

I n discussing spatial heterogeneity we have in mind a system similar to Landenberger’s (1973) mussel clumps on pilings (Section 11) where each part of the sub-system is fluctuating independently of other parts, except that there is movement of prey and predator among sub-

PREDATION AND POPULATION STABILITY

23

systems. We also assume that, unlike the mussel clump situation, there is no source of immigrants from outside the whole system. St Amant (1970) modelled such a system. Suppose the predator and prey occur in sub-systems which are “coupled” by movement of individuals in which the proportion of the population that emigrates is fixed, i.e. density-independent. Assume each sub-system is a LotkaVolterra system with no density-dependence, so that each sub-system on its own undergoes fluctuations of constant amplitude. For a model with two sub-systems, and using subscripts to designate the subsystem, we have

The fraction of prey leaving area 1 is incorporated directly into a, by subtraction, and similarly for a2,c1 and c2. a, P, A and 6 are the fractions of the population in one sub-system that leave and go to the other sub-system. These equations are not so easy to solve as our previous systems. For a start, we must solve a system of four non-linear simultaneous equations in order to find the equilibrium values. It is not even clear (from the equations) whether positive equilibrium values exist or are unique, though it would be surprising if this were not true. St Amant (1970) assumed H f , H,*, P: and P l existed and studied the behaviour of the system near this hypothetical equilibrium. Again taking

Hi

= H:(l

+ht)

etc., and ignoring terms of second order in h’s and p’s, he shows that near equilibrium the equations can be reduced to a linear system, of the general form given in Appendix I, in which the coefficients agi are all negative. This strongly suggests the equilibrium is stable since it suggests that large positive values of, say, h, will give negative values of ah# so that h, decreases, while large negative values of h, will cause k, to increaae; however, this argument is not sufficient to prove stability. Thus, by using a (somewhat unrealistic) counter-example, in

24

w. w.

MURDOCH

and A. OATEN

the one-prey, one-predator system discussed in Eqn (As) (a)of Appendix I , it is easy to show that the system can be unstable even if the coefficients act are negative. Suppose r = v = -1, but the other coefficients are large, say 8 = u = 2, then the system is unstable. St Amant also ran several computer simulations of a coupled system in which there is a generational time lag (i.e. difference equations). His conclusions are somewhat tentative, but are, in brief, that: (i)if the time lag is short enough, the system is stabilized within an appropriate range of migration rates; if the time lag is too long, there are no migration rates capable of stabilizing the system; (ii) for systems with time lags that are short enough, it is the degree of coupling (i.e. the migration rates) that is crucial: the two systems must oscillate out of phase with each other, so coupling must remain weak. All the systems with lags that St Amant considers, however, would be unstable without the coupling. I n his report on computer simulation of various aspects of spatial heterogeneity, Smith (1972) also concluded that a variety of mechanisms can produce heterogeneity and that it must be a strong stabilizing force in natural ecosystems. I n similar work, Allen (1974) assumes that parasites and their hosts are distributed in k patches; that in each generation fixed proportions, CEtj of the hosts and p f j of the parasites, in patch j move to patch i; and that the matrices {dgj} and {pgj}are transition matrices for an irreducible acyclic Markov chain. Within a patch, he assumes that the relationship between the populations at time t , of prey ( H t ) and predator (Pt),and the populations at time t + 1 are given by the Hassell-Varley model, Ht+, = o31t exp { -QP:-m} and Pt+, = d t- Ht+l, 0 < m < 1, or by the special case, m = 0, the Nicholson-Bailey model. Unfortunately, as Allen himself recognizes, in order to make the problem tractable mathematically he has had to make a number of assumptions which tend to remove the advantages that heterogeneous distributions seem, intuitively, to have: he assumes that the parameters a,m and Q are the same in each patch, that the transition matrices {&} and {ptj} are doubly stochastic (i.e. the rows sum to 1, as well aa their columns), that “the” equilibrium point (which is not necessarily unique) is the one for which all patches have equal numbers of hosts and of parasites, and that there is no random fluctuation. He concludes that, under these conditions, the multi-patch extension will not be stable unless the singlepatch version is, and may not be even then; and he arrives at similar conclusions for a Lotka-Volterra system of difference equations. However, the strength of his assumptions, and those of St Amant, suggest that the whole question of the relationship between spatial heterogeneity and stability is still rather open. The assumption that

PREDATION AND POPULATION STABILITY

25

a, m and Q are the same in all patches is particularly serious since it removes the machinery that keeps different patches oscillating at different frequencies.

E. A C C E L E R A T I N G

FUNCTIONAL RESPONSE

The stabilizing mechanisms discussed to this point correspond to the observations that have been made in real predator-prey systems, discussed in Section 11. The remaining sections of this paper examine other aspects, namely the consequences for stability arising from the response of the individual predator to changes in the density of its prey (or to changes in the densities of two or more prey). I n the L o t k e Volterra model this interaction is summarized in the second term of the prey equation, bHP. The Lotka-Volterra equations assume that each predator takes a constant fraction of the prey population regardless of its density. This is clearly unrealistic; at very high prey densities the predator must either get full or run out of time in which to eat more prey. Thus its functional response must level off. The effect of satiation is to destabilize the system, for the reasons we discuss below. Before that discussion, though, we state the problem more generally: instead of PbH for the second term of the LotkaVolterra model, we have Pf(H) where f(H) describes the relationship between prey density and number of prey eaten per predator. This function, the functional response, might take a variety of forms, but we are especially interested in two types (see Section IV): an increasing, decelerating response and a sigmoid response (see Fig. 8, Section IV); the linear response up to satiation can be thought of as a limiting case of the decelerating function. Now we examine the consequences of these different functions for stability. Our equations are now dH/dt = aH-Pf(H) anddPldt = -cP+dPf(H). The equilibrium values are H* = f - l ( c / d ) and = H*(l + h ) and P = P*(l + p ) , we get

P* = (ud/c)H*. With

H

dh/dt = a(1+ h) -ad( 1+p)f(H*( 1 + h)}/c

+

and dpldt = - c( 1+ p ) d( 1+p)f{H*( 1+ h)}. We expand f in Taylor Series about H*: f{H*+hH*} =f ( H * ) + hH*f ’(H*)+ &(hH*)2f”(H*) + ... Ignoring higher (than first) order terms in h, we substitute f(H*)+hH*f’(H*) for f(H*+hH*) into the equations, ignore terms involving the produot, hp, and after some

26

w. w.

MURDOCH

and

A. OATEN

rearranging obtain dh/dt = rh-ap and dp/dt = dH*f‘(H*)h where = a(1-(d/c)H*f ’ ( H * ) } .Again using the stability criteria of assertions (As) and (A9) of Appendix I we see that

T

(i) the solutions oscillate if r 2< &H*f’(H*); these oscillations are damped (i.e. their amplitudes decrease as t increases) if r < O , are constant if T = 0 and increase if r > 0; (ii) if r2 2 &H*f ’(H*)> 0, the solutions do not oscillate; h and p go to zero (so the populations return to equilibrium) if r < 0 and increase indefinitely otherwise; (iii) the equilibrium is unstable if d H * f ’ ( H * ) 5 0 : our approximating system converges to a new equilibrium if adH*f’(H*) = 0 and r < 0 , but otherwise h and p increase indehitely. This, however, is an unlikely case: it requires f ’ ( H * ) 5 0 , i.e. that prey intake per predator should not increase when the prey population increases. Thus the essential determinant of whether the functional response is stabilizing is whether r < O or not. Since c/d = f ( H * ) this condition is equivalent to 1 - H*f’(H*)lf(H*)< 0 , i.e. f ’ ( H * )> f ( H * ) / H * . This is the condition we use as a criterion of stabilizing functional response in Sections IV and V (see also May (1973a), Eqn (29)). It corresponds roughly, though not exactly, to intuitive ideas of convex or accelerating functional response in which predation intensity increases as the prey population increases. We sketch some stabilizing and destabilizing functional responses in Fig. 5 . Both the condition and the sketches make it clear that, in a stabilizing functional response, near equilibrium, any change in prey population size results in a greater than proportionate change in the predation rate. The main weakness of this criterion for stabilizing functional response is that it requires knowledge of H * , which in turn requires knowledge of c / d . It would be useful to have a criterion by which the tendency of the functional response to stabilize the interaction could be measured without reference to the (possibly unknown) other parameters of the interaction. We cannot, in fact, provide a single criterion, but we can indicate three features of the functional response that need to be considered. We restrict attention to functional responses that are either sigmoid or decelerating. First, for such functions, it is the small values of H for which f ’ ( H )>f ( H ) / H .More precisely, the values of H satisfying this criterion will form an interval of the form (0,Hm}: Hm is the maximum value of H satisfying the criterion. For decelerating functional responses, Hm = 0. The size of Hm is one measure of the tendency of the functional response to stabilize: since H * < Hm is required for stability, a large value of H , means a wider range of values of H* for

PREDATION AND POPULATION STABILITY

21

which stability occurs. The larger H * is, the less likely is extinction to occur through random fluctuations. Secondly, for the same reason, persistence of the population is more likely if P* is large; since P* = a H * / f ( H * ) ,this suggests that a functional response with a large

FIQ.6. The stability criterion for the functional response, f(H). The criterion, f ’ ( H ) > f ( H ) / H ,is a requirement that the slope of the tangent at H (4in ( a ) )be greater than the slope of the solid line joining the point (II,f ( H ) )to the origin ( 0 in (a)). This is equivalent to requiring the tangent line to meet the abscissa at a positive value. The dashed lines are all tangent to the curve. I n (a) the functional response is type 2, and destabilizing at every value of H . I n ( b ) the response is sigmoid; it is stabilizing for any value of H smaller than H , , such aa H,, and destabilizing for any value larger than H , , such as H,. Note that H , is greater than the inflection point.

value of H m / f ( H m ) is more likely to be stabilizing. Thirdly, one might measure the tendency of functional response to stabilize by the range of possible parameters, a, c and d, for which the corresponding LotkaVolterra system is stable. But this system is stable provided f ’ ( H * )> f ( H * ) / H * ,

28

w. w.

YURDOCH

and

A. OATEN

where H* is the point for which f ( H * ) = c / d ; this condition will be satisfied provided H* c Hm and, since f is an increasing function, H* will be less than Hm if f ( H * ) < f ( H , ) , i.e. if c / d < f ( H m ) . Thus the larger f ( H m ) is, the larger is the range of values of the parameters (specifically,of c / d ) for which the corresponding Lotka-Volterra system is stable. Summarizing, let H , be the largest value of H for which

f’(H)> f ( H ) / H . Then we would expect f to be more likely to enhance stability if Hm Hm, -and f ( H m ) are large. .f (H,) Notice that the functional response in these equations is treated as an instantaneous predation rate. This is one of the justifications for a procedure described in the next section, namely that in experiments and models concerning functional response the total prey density is kept fixed even though prey are being eaten.

F.

GRAPHICAL ANALYSIS

The approach described above, for analyzing the consequences for stability of a variety of assumptions, is especially useful because each assumption is a cog in a piece of machinery, namely the Lotka-Volterra equations, that models the whole predator-prey system. A t times, however, the machinery is somewhat restrictive, or we may not like to be encumbered by its over-simplified parts, or it simply may lack intuitive appeal. We present in the next few pages a simple method of assessing the stability characteristics of the functional response. The main function of our presentation is expository, since in order to find out if the predation is stabilizing, we would need to know about more than the functional response. If we were willing to assume that the number of predators did not change, and that the properties of individual predators remained constant, then we could also use the method to examine the effects of the predator population’s total response. The effects of predation (or at least of functional response alone) upon the change in the numbers of prey from one generation to the next can then be illustrated using the “stock-recruitment’’ curves shown in Fig. 6. The abscissa is the prey population size in one generation and the ordinate is its size in the next generation, which gives the abscissa value for that generation. The 45’ line is drawn t o facilitate reflecting values from one axis to the other.

PREDATION AND POPULATION STABILITY

29

Suppose the prey would increase by a factor X every generation in the absence of predation, i.e. there is no density-dependence. This is shown by the line OR, drawn for X = 3. Thus, at each generation t + 1, the population would be thrice the value at t. (There must be some limit to population size, possibly set by resources, at which OR will bend over and then cross the 45" line. For our purposes we assume that this limit is higher than the range of densities illustrated.) Now suppose that between generations the number of prey surviving predation is described by one of the curves labelled (1) to (3), so that the vertical distance of a point on the curve to OR gives the number of prey killed

",+I

0 "t

FIG.6. A graphical method of assessing functional response. O R is the density in generation t + 1 after the population in generation t has tripled. This density is reduced by predation according to one of 3 schemes (seeFig. 7) to give the density at t + 1 illustrated in curves ( l ) , (2) or (3). Equilibrium values exist where the curves cross the 46' line. Equilibria are stable where the curves cross the 46' line from above, as a t H,, provided the slope of the curve is not decreasing too steeply: slope must be > - 1. Hul and H m cross from below and are unstable equilibria.

30

w. w.

MURDOCH

and

A. OATEN

between time t and t + 1, which is f ( H ) . We assume that predation at time t operates upon a population of size H t .

Ht

FIG.7. Three functional responses analyzed in Fig. 6. These correspond to cases ( l ) , (2) and (3) described in the text. H, is explained in Fig. 5.

I n the absence of predation, Ht+l = AHt. With predation,

where f ( H t ) is the loss to predation. We examine three types of predation (Fig. 7). (1) I n the first type a constant fraction of the H t prey are eaten so that

Ht+, = H t ( h - a )

(12)

where a is the fraction killed by predation (density-independent predation). (2) The second type corresponds to a type 2 functional response (Section IV), with Ht+l =

AHt - B(Ht)a,0 < u < 1

(13)

We use the exponential form rather than, say, the disc equation (Holling, 1959b) merely for simplicity. Both yield increasing decelerating curves (inversely density-dependent). (3) Finally, corresponding to a type 3 functional response, the third type at first accelerates and then decelerates to an asymptote, as prey density increases:

PREDATION AND POPULATION STABILITY

31

the second term being one of the simplest equations for an S-shaped response. We now compare these responses. The first situation is very straight. forward and does not need to be illustrated: if the curve lies entirely above the 45’ line in Fig. 6, then the predator never causes enough mortality to reduce the prey’s rate of increase below zero. The prey population escapes and increases in each generation, even when the predator is causing density-dependent mortality as in case (3). The second situation, in which the curves lie all or in part below the 46’ line, is graphed in Fig. 6. I n case (1) the prey is reduced in density at each generation and finally driven extinct by the density-independent predation. In case (2) there is an equilibrium density a t Hul, but the equilibrium is unstable: below this density the population is driven extinct by predation, and if the population ever exceeds Hul it increases every generation thereafter. Perhaps the most interesting model is one in which predation is initially density-dependent, i.e. f ( H t ) is initially accelerating but “satiation” operates at high prey densities so that the predation curve decelerates and then a roughly constant number of the prey is eaten (case (3), Fig. 7). Predation is then sufficiently intense so that the curve intersects with the 45’ line at the equilibrium point, Hs. Since satiation occurs, a second intersection point must occur at a higher density, Huz. These are equilibria such that when the density is either H,, or H,, it will not change, but is a stable equilibrium and H,, is unstable. When Ht < Ha, the population increases towards H8. When

Hs < H t < Hum the population decreases towards Ha. One point deserves stressing here. Between 0 and the point Hm of the sigmoid curve in Figure 7, predation is density-dependent; above the point Hm, the predation rate (i.e. proportion killed) decreases with increasing prey density; i.e. it is inversely density-dependent in this range. But it continues to be stabilizing up to H,, because losses to predation in this range are greater than the number needed to keep the population replacing itself. When Ht > H,, the population increases at each succeeding generation. In real populations h will not be a constant. I n some seasons that are good for the prey h will be unusually large and the prey population will jump beyond the stabilizing range of the predator unless that range is large. Thus, the stabilizing capacity of an S-shaped response clearly depends upon the prey density at the stable equilibrium point and the range over which the curve remains below the 45’ line. For a given predation response, these in turn are determined by the rate of increase of the prey. For a given prey rate of increase, the predation curve will

32

w. w.

MURDOCH

and

A. OATEN

lie below the 45" line over a larger range the steeper the curve is, and the longer it is accelerating. Thus, even though acceleration over a small range of prey density near the origin is potentially stabilizing, in general we should be more interested in the range of prey densities over which the curve is stabilizing, relative to the normal range of variability in prey densities. One could also use this graphical tool to examine the consequence of total response under limited conditions. Evaluation of total response is discussed in Section VII. I n this section we have not tried to be exhaustive in our examination of stability analysis. For example, we have not looked at the effects of interaction between predators (Hassell and May, 1973), at the effects of varying the efficiency a t which predators convert prey to new predators, or at the consequences of adding extra species to the system (May, 1971). Furthermore, no consideration has been given to mechanisms that tend to maintain the predator population when the prey becomes very scarce. For example, carabid beetles (Murdoch, 1966), seashore snails, and probably invertebrate predators in general, can survive for very long periods (weeks or even months) with no food, which certainly must help to stabilize the predator density. We have also limited ourselves to the linear case where perturbations from equilibrium are small. Readers interested in examining stability far from equilibrium may find recent papers on limit cycles of interest (Gilpin, 1972; May, 1972). For a purely graphical analysis see a series of interesting papers by Rosenzweig (e.g. Rosenzweig, 1973). G.

SUMMARY

The Lotka-Volterra differential equations for predator and prey populations have neutral stability, but become unstable when time lags are added, either by writing the model as a pair of difference equations or by incorporating lags directly into the differential form. Stochastic variation is also destabilizing, but we suspect that time lags are a more potent source of instability than is such stochastic variation. Stabilizing mechanisms, including density-dependence in the prey, prey refuges, and spatial heterogeneity when M e r e n t sub-systems have different parameter values, stabilize the non-lag form of the model, but their ability to stabilize the model with lag depends upon the length of the lag time relative to other features of the model. Prey refuges probably are more stabilizing than is spatial heterogeneity. The remaining sections in the paper center around short-term responses of predators to changes in the density of their prey, and in the present section we place that discussion in context by describing

PREDA’ITON AND POPULATION STABILITY

33

the effects upon stability of various kinds of functional responses that might replace the assumption of linearity in the Lotka-Volterra model. Finally, we present a simple graphical way of viewing functional response, as an aid to the intuition. IV. O N E - P R E Y SPECIES I n this section we review briefly what is known about how individual predators respond to changes in the density of a single prey species, changes that occur within an interval that is short relative to the predator’s life-span. Such functional responses to prey density reflect changes in the attack rate of a predator whose characteristics (e.g. age and size) remain essentially constant during the interval. I n accord with the general aim of this paper, we are interested especially in whether or not predation can be stabilizing; that is, we are interested in the absolute predation rate as a function of prey density. I n the situation where the predator chooses from prey of different species, the absolute attack rates will of course be affected by changes in the relative frequency of attacks on Werent species; however, we will deal with that question separately later (Section V). We also examine the effect of patchiness in the prey leading to the predator distributing its time non-randomly among different patches. The term “functional response” was introduced by Solomon (1949), and Holling (1959a)recognized three forms the response might take. All responses level off at high prey densities because the predator becomes satiated and/or runs out of time in which to eat more prey. From Fig. 8 it can be seen that type 1produces density-independent mortality up to satiation; type 2 produces inversely (or negatively) density-dependent mortality over the entire range; and only type 3 produces densitydependent mortality. The effect of satiation (or running out of time) is to produce inverse density-dependence. The curves are expected to rise initially because contacts with prey increase with prey density. The rise would be linear if each contact (as between molecules of a gas) took no time or did not affect the speed at which the predator or prey moved. However, since some time must be spent handling each prey caught, the amount of time available for searching decreases with prey density, causing the curve in response type 2 to decelerate along its length. A second factor might enhance this effect: if a predator slows its hunting rate rn it becomes more satiated, and it is likely to be more satiated more of the time a t high prey densities than at low densities, then its average hunting rate will decline with prey density. Clearly, in type 3 some other factor(s) must operate at low prey densities to reverse these trends. Many explanations

34

w. w.

MURDOCH

and A.

OATEN

are possible. For example, the predator’s efficiency of search and/or capture may increase with the number of meals eaten per unit time; or he may hunt faster as he receives increasing amounts of stimulus from the prey; or the prey may behave differently as their density increases (a factor we would like to remove from most laboratory experiments); or the predator’s response to patchiness in the prey may

Prey density

FIG.8. Three types of functional response designated by Holling (1959a). For each type the number (and percentage) of prey killed per unit time by a single predator is graphed against prey density. Only type 3 yields density-dependent mortality.

cause a sigmoid curve. Where there are several prey species, the explanation may be quite complex, and we deal with this in Section V. The effects of various kinds of functional response upon prey stability is discussed near the end of Section 111. We first examine the functional response in a homogeneous environment, i.e. the predator is inside a large patch containing many prey

PREDATION AND POPULATION STABILITY

35

(Sub-section A). Thereafter we look at overall functional response when the predator is faced with prey distributed among patches (Subsection B).

A.

FUNCTIONAL R E S P O N S E I N A PATCH O F P R E Y

Probably most predators and parasites are not prey- or host-specific, but some are (e.g. some Coccinellid beetles and hymenopterous wasps) and many other predators may in practice be prey-specific so long as their preferred prey is abundant (e.g. owls feeding on wood-mice). Laboratory studies have naturally concentrated mainly on this simplest of situations, and we now have a wide range of laboratory studies showing the form of the functional response to one prey species (Table I). The experimental techniques used have varied, but usually each predator is presented with a fixed number of prey; the prey are replaced as they are eaten or the number presented is large enough so that predation removes only a small fraction (say 10%) of the prey. Usually, different predators are presented with different prey densities; this has the advantage of making each datum statistically independent, and it also excludes any effect of a “memory” from the predator’s having fed at a different density previously. Thus, the analogous field situation would be an instantaneous picture of attack rates of predators that h d themselves in patches containing different prey densities. The experiment is also analogous to a predator faced with a prey whose density varies through time, provided there is no “memory” effect, and since we are concerned with stability, it is this latter relationship with time that interests us. I n some experiments the same predator has been offereddifferent densities in sequence, which has the potential advantage of reducing variability among individuals; however, if there is a memory effect then the outcome will depend upon the order in which densities are presented. Clearly, when prey are eaten, the prey density will change. However, most functional response experiments try to avoid this effect, either by replacing prey or by allowing only a small fraction to be eaten. This is because we really need a description of the instantaneous feeding rate at each density (see Section I11 E). I n population models (as distinct from the functional response), of course, prey numbers are allowed to change. However, functional response refers to very short-term experiments so that the approximation involved in assuming fixed prey density i s appropriate. An alternative way of viewing such experiments is to think of prey density as the rate at which prey become available per unit time. Then we imagine some process (such as recruitment to the

TABLEI Functional reaponae of predators, in the laboratory, given differ& dmaitiee of a single prey species Response Predator

Prey

type

Source

Protozoa Stentor

Stentor Stentor

Tetrahymm Ezcglerm Chhrnydmunma

I

Flagellatea

3 3 3

Rapport (1974) Rapport (1974) Rapport (1974)

insects

Acheta (Gryllidae) Hierod& (Mantidae) C&a (Corixidae) Lethoceros (Bellastomatidae) Notonecta (Notonectidae)

Housefly puparia Adult housetlies Mosquito larvae Tadpoles Mosquito larvae

P o d h (Pentatomidae) Acdliw, (Dytiscidae) Syrphua (Syrphidae)

Webworm larvae Mosquito larvae Psyllid l a ~ a e

2

2 2

Holling (1965) Holling (1965) Holling (1965) Holling (1965) Holling (1965) and Fox and Murdoch (1974) Holling (1965) Holling (1965) Clark (1963) (field data)

% P

Parasitic Insecls Emark Exidtxhthk

Nt?7neritua

Whitefly Almond moth larvae Moth larvae

Dahlbminw Exenterua c a d P l e o l Q p bm3izonua Praon txmoletum

Sawfly cocoons SawRy larvae Sawfly cocoons Aphids

Other Invertebrates Arternia; (Crustacea) Daphniu (Crustacea) TyphlaEromecs (mite) T h k (snail) Amnthino (snail) Pisaster ( s t a h h )

Burnett (1964) Takahashi (1968) Taylor (1972) (Fig. 31 in text) Burnett (1954) Griffiths (1969) Griffiths (1969) Messenger (1968)

Holling (1965) Holling (1965) Holling (1965) Murdoch (1969) Murdoch (1969) Landenberger (1973) (Fig. 30 in text)

43Algae and yeaat

mtes MusSelS MusSelS "urban snails

Fish Cyprin~ (carp) Rectilw, (roach) Alburnua (bleak) S d m (trout)

Bream roe Chironomid larvae Daphnia Amphipods

2 2 2

land2

Holling (1965) Holling (1965) Holling (1965) Ware (1971)

w. w.

38

MURDOCH

and

A. OATEN

vulnerable prey class) as providing a continuous supply of prey at that density. I n 1965 Holling was able to list some examples of functional response, and the number of examples has since been doubled. Table I shows that almost all the results are recognizably one or the other of the 3 types shown in Fig. 8, and that almost all the predators show type 2 responses when given only one prey species. This result applies to such a wide range of organisms that one might suspect it is the basic and most widespread response. There are, however, five examples of type 3 responses, two in parasitic insects and three in the Protozoan, Stentor, that are probably real and examplify a mechanism that yields a type 3 response that would persist in nature (Burnett, 1964; Takahashi, 1968; Rapport, 1974). A possible explanation for the initial acceleration in the response is that the hunting (or feeding) rate increases at lower densities as some stimulus deriving from the presence of the prey increases. For the parasites it may even be possible that there is a threshold odour level below which they do not hunt at all. I n the case of the protozoan Stentor (Rapport, 1974), this would mean that the cilia that cause the feeding currents would beat faster as the concentration of the prey’s metabolites increased. Such a mechanism could operate in the field and produce type 3 responses there. However, no good evidence is available to test these suggestions. One can imagine other situations that would lead to S-shaped curves in the field; for example, the prey might become more vulnerable as their density increases because only a fixed number of refuges is available. We discuss below how patchiness might also cause this. However, in the absence of such complicating factors, type 2 curves seem to be the rule among predators feeding upon one species of prey. Thus the functional response in these circumstances generally will be destabilizing. Unfortunately we have not found any field data to illustrate this situation. One set of data not mentioned in Table I, Reed’s (1969) for bluegill feeding on mosquito larvae, and Landenberger’s (1973) data for starfish feeding on turban snails, are discussed in detail in Section VI. Finally, the results of two studies (Mori and Chant, 1966; Sandness and McMurty, 1970) do not fit easily into Holling’s scheme. B.

PREDATORS’ RESPONSES TO PATCHINESS

We have assumed to this point that the predator is searching within a patch of prey, or that the prey’s distribution in space does not influence the predation rate. However, prey come in different densities

PREDATION AND POPULATION STABILITY

39

in space, often corresponding to patches of a particular habitat, but also even if the habitat is homogeneous; indeed, patchiness greater than random is probably the most general type of distribution. Furthermore, it has been known for a long time that at least some predators behave in such a way that they concentrate their attack upon denser patches of prey. For example, Fleschner (1950) working with mites, and Banks (1957) and Dixon (1959) using ladybirds, showed that the predators tended to remain for some time in the immediate vicinity of their most recent meal. The predator’s turning rate increased immediately after a meal and stayed high for a short period, before random search was resumed. Thus for a dense patch the probability is high that a long sequence of meals will be taken from the patch. Each predator will therefore stay longer in patches with many prey than in patches with few, and if all the predators behave the same way, more predators will occur at the denser patches. Hassell (1968), using Varley’s field data on the predators of the winter moth, showed that the parasitic tachinid fly Cyzenis aggregated at dense prey patches, spending more time at such patches. Landenberger (1968) attracted starfish to a small subtidal area, and kept them there, by adding food. Other workers (see Hassell and May, 1973, for a review) have shown that some predators are actually attracted to areas of high prey density. Hassell (1971)also did laboratory experiments in which the parasite Nemeritus hunted for hosts that were a t different densities. Hassell’s idea was to present the parasite “with the choice of a range of different host densities at the same time”. Consequently, the prey (almond moth larvae, Ephestia) were placed at different densities in petri dishes which were placed in a small box, either 0 - 5 m2 or 0-05 m2 in area. Each box contained either 16 or 15 such dishes, in total representing six different densities. Hassell showed that, in the smaller box, the parasite spent almost all of its time at the highest host density and that this resulted in a slightly accelerating curve of number of contacts versus host density. (Such data are not presented for the larger container. The main point of Hassell’s experiments was to examine the effect of interaction between parasites, but we are not concerned with that problem here and discuss only the experiment that used a single parasite.) It is not clear whether the parasite could distinguish from a distance those petri dishes with high prey densities; presumably it could, or at least needed very little time to “sample” a petri dish, so that the parasite was actually making a choice among simultaneously available host densities. (These experiments therefore differ fundamentally from the model we discuss below. I n the models we assume (a)that the predator has to take time to search a “patch” to determine the reward rate, even if the patch is empty, and (a) that time is spent in transit from one patch to another.

w. w.

40

MTJRDOCH and A. OATEN

In Hassell’s small container, the time lost in transit (when only one parasite was present) must have been negligible. As we will show, these differences in the models are crucial.) These various observations discussed above of predators spending more time where the prey is denser, support the self-evident proposition that predators are adapted to find their prey and, other things being equal, ought to go preferentially where their prey are easiest to find. Fishermen who follow flocks of sea-birds have for centuries worked successfully on the basis of this proposition. In general, it will clearly be of selective value for the predator to spend more time in dense prey patches than in sparse patches (unless too many other predators do the same). The choice is less clear, however, for the prey. If we assume that the predator has a type 2 response, then the average prey in a dense patch has a lower probability of being eaten during every unit of time the predator spends there than does the average prey in a sparse patch during every unit of time the predator spends in that patch. Only if the predator spends enough extra time in the dense patch will the probability of attack for the average prey there exceed that in a sparse patch. From the point of view of the individual prey the matter can be put succinctly as follows. Let the functional response of the average predator, i.e. the number of prey killed per unit time while the predator is in the patch, bef(H), where H is the prey (or host) density in the patch. f ( H ) is increasing but decelerating for a type 2 response. Let g ( H ) be the length of time (or the proportion of the total fixed time) that the average predator spends in a patch of density H . The question is, is the productf(H) g ( H ) an accelerating function? Does the predator cause mortality that is density-dependent in space? This question has been given some attention recently (e.g. Royama, 1970) and we return to it below. But the broader question we are asking here incorporates this relationship and goes beyond it. Our question is, how does the variation in the predation rate in space affect the likelihood that the predators can cause density-dependent mortality on the whole prey population as it varies through time? Does mortality that is density-dependentin space lead to mortality that is density-dependent in time? The variation in predation rate through time, as a function of total prey density, is what we call hereafter “overall functional response’’ I n another paper (Oaten et al., 1974) we explore in some detail the effects of patchiness upon overall functional response, by (1) deriving several models of a predator’s behaviour in a patch, (2) calculating on the computer the overall functional response of a predator behaving according to these models, and facing different degrees of prey patchi-

.

PREDATION AND POPULATION STABILITY

41

ness, and (3) exploring a more general criterion for relating predatory behaviour to functional response. I n the present paper we select for illustrative purposes only one model of predatory behaviour, two general types of prey distribution, and we summarize the conclusions from the more general analysis. I n doing so we discuss the significance of the following factors with respect to the stabilizing properties of the functional response: 1. Time spent in transit between patches. 2. The relation between transit time and the time taken to handle a

prey individual. 3. Degree of prey patchiness. 4. Prey patchiness as a function of prey density.

Although we include in this analysis ( a )the time taken to search an empty patch and (b) the relationship between prey density in a patch and time spent in the patch, we do not discuss the effects of variation in these variables, though such variation no doubt affects the stabilizing properties of the overall functional response.

c. A

MODEL OF PREDATOR BEHAVIOUR, A N D ITS CONSEQUENCES

There are three basic assumptions to be made for a model of the kind we want to investigate here. These concern: the way the predator searches in the patch (how good is he at finding prey?); the way the predator decides when to leave the patch and try elsewhere; and the way the prey are distributed among the patches. We deal with these questions in turn. First, we assume that, within a patch, the predator is searching randomly, at constant speed, for randomly distributed prey, and that his radius of perception is small. If this radius is r and the speed is u, then in a time interval (t, t + 6t) the predator would, if it travelled in a straight line, search an area 2ru6t: a rectangle, u6t units long and 2r wide. In fact the line is not straight, but if 6t is small it is nearly so, and the difference between 2ru6t and the true area searched is o(6t): i.e. goes to zero faster than 6t does (o(st)/st+O).(Technically, this requires the predator’s path to be differentiable at all but a finite number of points.) Thus the probability that the predator does not discover a particular prey during (t, t + 6t) is 1 - (2rulA)St+ o(&), where A is the area of the patch. (We msume all patches have the same area.) If there are k prey in the patch at some time t, the probability he discovers none of them is (1 - A6t + o(8t))k,where h = 2ru/A. (This assumes the prey are randomly and independently distributed about the patch.) This will hold for any t , where t is the time since the last

42

w. w.

MURDOCH

and

A. OATEN

prey wtw disposed of, provided we can ignore the probability that the predator will be able to see the next prey from the place where he caught the last one: it is for this reason we assume r is small. If T is the time it takes to find the next prey, then the probability that T is greater than t + at, given that T is greater than t (i.e. given the predator has been unsuccessful up to t ) , is just the probability he discovers no prey during (t, t + 8t). Thus

P(T>t+6t I T > t ) = (I-h8t+0(8t))k

(16)

P(T > t + St) , so subtracting P(T > t )

1 from each side,

The left side of (16) is

dividing by 6t and letting 8t+O we have

a (log P(T > t ) } =

--

dt

- hk

which solves to give

P(T > t ) = e--Xkt (In fact our first assumption, really, is (17)-we do not question why it is true, for the moment; there may be sets of assumptions, other than those given above, that lead to it.) Second, we assumo that the predator adopts the following strategy: if he has found no prey before time to, he will leave the patch; if his first prey is caught before to, he will “handle” it (eat it, digest it etc.), and then, starting his “clock” at zero again, begin searching for a second prey; if t, time units of this search are unsuccessful, he will leave the patch, but if he is successful before then, he will “handle” his prey, restart his “clock”, and begin a search for the third prey. In general, after handling the ith prey, the predator starts his “clockyy again, and will search for the (i+1)th prey, leaving the patch if unsuccessful by time ti. While it is unrealistic to assume the times, to, t,, ... are exactly adhered to, this assumption seems to fit the behaviour of some predators-Coccinellid beetles for example-fairly well if the t i s are taken to be random variables with small variances. We shall make two further assumptions about the ti’s. The first is that they are all the same, t say; i.e. ti is independent of i, the number of prey he has eaten in the patch so far. The second is that t is constant-in particular, that it is not affected by the average density in other patches encountered by the predator. These last two assumptions essentially require the predator not to have too much behaviour. In particular, the second one says that the predator does not compare the reward rate of the patch he is in with the rates in other patches he has visited.

PREDATION AND POPULATION STABILITY

43

Our third assumption concerns the distribution of patch densities. In fact, our real concern is with the densities the predator is likely to encounter: if he has some way-movement, sound, smell etc.--of distinguishing the more dense patches, the distribution of the densities he encounters may not be the same aa the “true” distribution of prey densities. We will assume, however, that they are the same-again, a predator without much behaviour-and will consider several simple possibilities. I n each, we assume the distribution of the number of prey in a randomly chosen patch is of a particular type or family-we consider only Poisson and Negative Binomial-and that the only parameter required to determine which member of the family applies is the mean. This provides no difficulty for the Poisson distribution, which involves only one parameter, its mean; but the Negative Binomial involves two: usually R and p , where the random variable concerned is the number of failures before the Rth success in a sequence of Bernoulli trials in which each trial has probability p of being a success. The mean of the Negative Binomial is Rqlp and the variance is Rqlpa. Two obvious ways of varying the mean, h, of this distribution are to fix p (and q ) while varying R, and to fix R while varying p . The difference between these methods is that the first keeps the ratio of variance to mean a constant, lip, while the second has the variance ( = patchiness) increase faster than the mean: if h is the mean, the variance is h(h/R+ l),so the variance increases like the square of the mean. We have used both these methods of varying h. Our procedure has been as follows. First, suppose a predator enters a patch containing D prey. The number of prey he eats there, and the length of time he stays, are both random variables. Their distributions can be deduced from our fist two assumptions (random search and the predator’s way of deciding when to leave). These distributions are complicated, however, and we have computed only g(D), the mean or expected number of prey the predator will eat in a patch initially containing D prey, and s(D), the mean time he will search in such a patch. We then suppose the patch is chosen randomly from a collection of patches, containing varying numbers of prey. Thus D becomes a random variable whose distribution we will assume to depend only on h, the mean number of prey in a patch. As we have said, we will take the form of the distribution to be either Poisson, with mean and variance h; Negative Binomial, with mean h and variance h/p for several choices of p between 0 and 1 ; or Negative Binomial with mean h and variance h(h/R+ 1) for several choices of R. Since D is random, the mean functions g ( D )and e(D)are also random. Technically, they are conditional means, given D. Accordingly they will

44

w. w.

MURDOCH

and

A. OATEN

have distributions which depend on h. We do not compute the exact distributions of g ( D ) and s(D),which would be like the distribution of D but would concentrate on the possible values of g(D) and s ( D )rather than those of D. We do, however, want the means of g ( D ) and s ( D) which, for a given form of the distribution on D, depend only on h, so will be designated G(h) and S ( h ) . Thus, on an average visit, the predator will eat G(h)prey and spend time S(h)in searching. However S(h) does not account for all the time the predator devotes to the patch. He must also spend time handling the prey he has caught, and he must spend time getting to the patch from the patch he last searched. We assume that, on the average, the predator takes time T to travel from one patch to another; and that the time it takes to handle one prey individual is a random variable, independent of the number of prey eaten, whose mean is 7.With these assumptions the average time spent in (or getting to) a patch is 7+S(h)+ 7G(h)when the average prey density is h. Accordingly we take the overall functional response, the number of prey eaten per unit time as a function of average prey density, to be

(It is intuitively clear that the average number of prey eaten per unit time is the average number of prey eaten per patch visit divided by the average time-including handling and transit times-spent on a patch visit. This intuition can be theoretically justified: see Cox (1970), especially Chapters 4 and 9.) Note that we have changed notation slightly here. Functional response is usually expressed as a function of total prey density, H . We have it here as a function of the average patch density of the prey. If there are M patches, H = Mh. We do not want to specify M other than to say it is “very large”. I n fact it would need to be infinite if our hypothesized distributions for the number of prey in a randomly chosen patch were exact. We need M to be large for these distributions to be approximately correct, and in order that the predator’s shortterm effect on the total prey population be negligible (we do not, however, ignore his effect on the patch he is in). If f ( H ) is the usual functional response, the relationship between it and our form of it is f ( H ) = F(H/M). We saw earlier (Section I11 E) that if we allow for functional response in a LotkeVolterra system, and write dH/dt = a H - P f ( H ) and dP/dt = - cP + dPf ( H ) , then the equilibrium point H*

= f-l(c/d),

P* = adH*lc,

PREDATION AND POPULATION STABILITY

45

is stable provided

where h* = H * / M , so this condition can be rewritten as

If we do not know c or d, we cannot say whether or not a particular F will stabilize the system it is in. We can, however, ask how large is the range of values of cld for which this condition holds. Or, writing cld as u, we ask for the range of values of u for which, when

F(h) =

U,

F ’ ( h )> F(h)/h.

The larger this range of values is, the more “stabilizing” is the functional response.

D

FIG.9. The solid curve is the expected number of prey eaten g(D),from a patch of initial density D . The dotted curve is the expected length of time s ( D ) spent searching a patch of initial density D . Search time does not include handling time or transit time. The range of D is 0 to 400. Note that g ( D ) and a ( D )do not depend on the average patch density, h.

Turning to the model, we now give formulae for s ( D ) and g ( D ) ,and then for S(h) and Q(h).The mean number eaten is D-1

9(D) =

c

fl

l-I

n=O j = O

(1 -exp { -

W-.W)

(19)

w. w. MURDOCH and

46

A. OATEN

and the expected search time is D-1

1

n (1-exp { - h(D n

-j)tj)

+

D-1

These formulae are derived in the Appendix to Oaten et al. (1974). These expressions for g(D) and s(D) are rather lengthy, though basically simple. We have simplified them by assuming the t2)s are all the same, i.e. tg = t for i = 0, 1, 2, ..., but this does not shorten them. Some approximation is possible if we assume that prey are replaced aa they are eaten, but this seems a very drastic assumption when some patches may contain very small numbers of prey. Rather than do this, we have calculated the exact values on the computer. On the computer we have calculated the following values:

(4

g(D) =

D-1

n

n=O

j=O

c n

(1-exp{-W-j)t))

and

+t

n (1 - exp{

D-1

- h(D - j ) t ) )

j=O

the mean number eaten and the mean search time, respectively, for a predator visit to a patch initially containing D prey. We arbitrarily take t = 1 ; this is just a matter of choice of time scale. Our choice of A, the proportion of the area of a patch searched per unit time by the predator, is then not arbitrary. It should, no doubt, be based on experimental evidence. We have not done this. We tried several values of A, and computed values of g(D) and s ( D ) for D = 1, 2, ... 100. For many of the values of A, g(D)was either close to D or close to 0, even at D = 100; i.e. the predator essentially ate everything in the patch, or nothing. These values seemed less interesting, so we decided on A = 0.05 which suffers neither of these defects. We have plotted s(D) and g(D) in Fig. 9, with D ranging from 0 to 400. It will be seen that both increase throughout the range. It would be more realistic to bound both functions, since the predator’s total time in the patch-searching time and handling time-must be limited. However, such a bound would apply more to the large values of D, while our main interest is in the smaller values. We also note that, though both functions are initially accelerating, g is essentially linear

PREDATION AND POPULATION STABILITY

47

for large values, while s is decelerating. When D gets large, it becomes virtually certain that the predator will catch a first prey in a time very close to l/AD; thus g ( D )- g ( D - 1) gets close to 1 (i.e. g'(D) gets close to 1 so g is essentially a straight line) and s(D ) - s(D - 1) gets close to l/AD (i.e. s'(D) gets close to l/AD so s behaves like l / A log D). (b) We computed S(h) and G(h),the average values of s(D) and g(D), under several different distributions on D . That is, we computed

iw)=

2 s(D)p(D/h)and QW 2 9(D)P(D/h), =

D=O

D-0

where p ( D / h ) is the probability that a randomly chosen patch has density D when the average patch density is A. Of course, we did not really take the sum to 00; but if the sum is taken, say, to N, it is not difficult to calculate a bound on the remainder; we took N large enough for the bound to be smaller than 0.001. The distributions we used were: (i) the Poisson distribution, with mean h, for h = 0.5, 1, 2, 4, 6, ... 96, 104, 110, ... 600 (see Table 11); (ii) the Negative Binomial distribution, with: (A) mean h and variance 2h, for h = 2, 4, 6, ... 100; (B) mean h and variance 4h, for h = 6, 12, 18, 150; (C) mean h and variance 10h, for h = 9, 18, 27, ... 225; (D) mean h and variance h(h+ 1) (i.e. the variance is roughly the square of the mean) for h = 2 , 5 , 8 , 1 1 , ... 98 (see Table I1 (b); (E) mean h and variance h(h/4+ 1) for h = 4, 12, 20, 28, ... 100. It would, perhaps, have been better to use the same values of h for each distribution. We did not do so partly because the distributions themselves impose restrictions on the possible values of the means: e.g. for the Negative Binomial, R must be an integer and, in cme (C), we had to take p = 0.1, so the mean is necessarily 9R. It may seem rather extreme to have the variance increase as the square of the mean (case (ii) (D)),but Dixon (1966) has shown that for some populations of aphids the variance increases this fast.

...

(c) For each computation of S(h) and G(h)we computed values of the overall functional response F(h) = values of

and 7,namely 0.05, 0.1, 0-5 and 2.0. T

T

G(h) We used 30 pairs of qG(h)' = 0.1, 0.5, 1, 5, 10 and 50, and 7 = 0.01, ~~

T +&h)

+

(a) For each computed functional response (i.e. for each F(h)computed above) we estimated rates of change by considering, for each pair of successive values of h, say h, < h,, the value of F(h,) - '(h).This is the h, - h,

48

w. w.

MUBDOCII

and

A. OATEN

value of P‘(h) somewhere between hl and h,; if P is accelerating, P’(h,)>P’(h)>F’(h,), while the inequalities are reversed if F is P,( h ) we decelerating. To test our criterion for stability, P ’ ( h ) > h computed all values of

W , )- P(h1) - -ml) ha - hi

hl

and of m 2 ) - P(hJ

ha - hl

- -W

2 ) .

ha ’

if F is accelerating, the former overestimates F’(hl)- F(hl) and the hl latter underestimates F‘(h,) - P(h2),while if P is decelerating the h2

former underestimates and the latter overestimates. The tendency of a sigmoid F to stabilize the interaction can, as we have noted before (Section I11 E), be measured by three things. If h , is the largest value of h for which P’(h,)> ’(lam), these three things hTn are: h, itself, F(hm)and h,/P(h,). We recall, from Section I11 E, that a large value of h, means a large range of prey equilibrium values for which stability will result; a large value of hm/P(hm) means a large range of predator equilibrium values; and a large value of F(hm)means a wide range of possible choices of the parameters c and d in the Lotka, Volterra equations, for which stability will result. We have tabulated h, and P(h,), for various values of T and r] and for two distributions, in Table I1 (a) and (b). These values are approximate. If P is initially F(h) i.e. at hm. accelerating, then it must be decelerating when P’(h) = -, h

.

-

underestimaie it, and hence be negative. Accordingly we take h i to be the first value of h, for which the first estimate is positive and the second negative. This is illustrated in Fig. 10. Table I1 includes only the extremes of the six distributions we listed above-the most even (Part (a)) and the most patchy (Part (b)). The entire set of results does not show a clear relationshipbetween the degree of patchiness and the range over which stability occurs, although they all show the same pattern with respect to variations in T and 7.

49

PREDATION AND POPULATION STABILITY

TABLEI1 Values of hm and P(hm)for v a h vduso of r (transit t h e between p a t c h ) an& q (hara&lins time) 0.01

0.05

8

0.1

1

22

52 14557

120

102

132

32

26 2.3123

0.1

_

~

0.1884 84

0,6917

0.2297

2.0

0.5

8

23 1.7360

0 0.3604

1.4183

0

-

~-

38 0.5

32 2.4403

77

1

98

0.7494

I

95

0.7057

26

77

92

0.6604

0.1906

04759

1.4322

I

17

47

65 1.6136

0

0.6786

14359

71 1.8775

0

35

56 1.881 1

0

04510

1.6809

59 2.2525

0 0.5270

20

35 1.9966

65 5

1.5712

38 2.5231

14

29 1.8888

44 1

50

_

34

138 1.6496

0.05

0.0493

0.7492

216 2.1678

___-___

I

1.9037

252 3.0142

0.1

76

2.3349

324

0.6835

120

3.0892

50

1.866

0.0243

6

54

2.2486

156

1 0.0243

94

3.0979

10

1 1.2402

-

0

0.0160

42

2.2070

5

0.7484

0 0

0.5

1.2177

64

1

0 0

0.02 18

34 1.7513

2.0

0

0.0441

46

0.5

0.5

0.3560

0.5

0.1

0.2163 53

0.4453

0.1986

I n each equare, the top left entry is the value of hm, estimated aa described in the text. The bottom right entry is the corresponding value of P(hm). In Table (a) the distribution of patch densities is Poisson,with h = 0.6, 1, 2, 4, 96, 104, 110, ..., 600, and variances = h. See case (i) in text. In Table (b) the distribution of patch densities is Negative Binomial with R = 1 and p = 1/3, 1/6, 1/9, ..., 1/99. Thus the means, h, are 2, 6, 8, ..., 98, and the variances are h(h+ 1). See caw (ii) (D) in the text.

...,

50

w. w.

MURDOUH

and

A. OATEN

Some of the features of the table were to be expected. For example we see that, in every case, hm (the largest density a t which overall functional response is stabilizing) increases as either T (time spent in transit between patches) increases or r ) (time taken to handle one prey individual) decreases, and that F(hm)increases as r) decreases. It is not clear, intuitively, what to expect of F(Rm)as T increases: increasing T increases hm, but also tends to decrease F by increaaing the denominator. From the table it seems that the effect of increasing T is at first to

hl

hz

h3

Fia. 10. Our calculations of h , from the computer output. The tangent line at h, must lie between ( a )and ( b ) . We take h , to be the value of h, for which, as in the diagram, line ( a )meets the abscissa a t a negative value and line ( b ) meets it at a positive value. This will happen if - Wl) - F ( h )- F(hz) --P(ha) < and m z ) > o, ha - hz ha hz - hl hZ respectively. This should ensure that the tangent line a t h, should pass through or near the origin, as is required for h,.

mz)

It is much more difficult to compare the results from the different distributions, e.g. to find a relationship between the variability of patch densities and the values of hm and F(hm). There appears to be a tendency for the distributions with large and rapidly increasing variances to give larger values of hm when T is small, but for these values of hm not to increase much w T increases (see Table I1 (b)).Conversely, the more even distributions seem to give very small (often zero) values for hm when T is small, but large values for hm when T is large (see Table I1 (a)). We have speculated on possible reasom for this in the next section, but the results are not sufficiently clear-cut, or consistent across all the distributions we have considered, for us to claim the discovery

PREDATION AND POPULATION STABILITY

61

of any kind of universal law. The question of the effect on stability of variability in patch densities, as a function of average patch density, remains open.

D.

Q E N E R A L C R I T E R I A FOR S T A B I L I T Y

Conclusions drawn from a model such as we have presented must be quite tentative. Not only have we evaluated the various functionssearch time and number eaten-only at a few selected points, but also the model itself involves many assumptions, some of them unlikely to be true. These range from the probably harmless, such as that the predator will search for a fixed, exact time before deciding to leave a patch and look for a better one, to the possibly serious, such as that the predator's behaviour in a patch is determined only by his experience in the patch and not at all by his experiences in other patches: that is, that he does not allow his estimate of the overall density to influence his behaviour. (We hope, in fact, to develop more realistic models which allow for this factor in a later paper.) It would be preferable to have a more general picture of the relationships between patchiness, predator behaviour and stabilizing functional response. We attempt to provide this in some detail elsewhere (Oaten et al., 1974), without succeeding in giving a clear and unambiguous picture. Here we will merely summarize some of the points we make there. I n this section, as in the previous one, we take the expected number of prey eaten by the predator during a visit to a randomly chosen patch to be G(h), where h is the average number of prey per patch. The average time devoted to a patch visit is the sum of T, the average transit time-the time taken to get to the patch; S(h), the average time spent searching the patch; and 7G(h), the average time spent handling prey in the patch. Using again the renewal theory result from Cox (1970), we take the functional response to be

We again use the criterion established in Section I11 and used in the previous section, that F is stabilizing a t h if F'(h)> P(h)/h.Applying this to (21), we find after some rearrangement that F is stabilizing if

where we have omitted the argument, h, from the functions S, a and

52

w. w.

MURDOCH

and

A. OATEN

G’. Alternatively, writing V = G/h (so V is, in a sense, the vulnerability of a single prey individual: if there are N patches, V / N is the proportion of total prey that will be eaten in an average patch visit), F is stabilizing at h if

(T-kS)V’> V(qV+s’) (23) From (23) we notice that a minimum requirement for stability is that either SV’ > VS‘ or TV’ > q V 2 . The first of these is unlikely to hold anywhere: it can be rewritten as (V/S)‘> 0, so is a requirement that the proportionate increase in V be greater than that in S ; but if patches are chosen randomly and prey are distributed randomly, V is the proportion of a single patch area searched in time S, so cannot increase faster, proportionately, than S (unless the predator’s search rate increases, a possibility we ignore here). Indeed V should increase more slowly than S since extra search time will be wasted on previously searched areas. Thus stability seems to require TV‘ > q V 2 . If this is to hold not only at h but also at all values below h, as would be the case for sigmoid functional response, V will need to be at least as large as the solution

Tvv(o)( V ( 0 )is the limit of B(h)/h

of V’ = ~ V % /which T , is V,(h) =

qhV(0)’ aa h+O; it can be taken as the proportion of an empty patch that the predator will search before giving up, so V ( 0 ) = 1-exp{-hS(O)).) Obviously, V is more likely to be larger than V , if V , is small; and V , is small if T / T is small and if V ( 0 ) is small. Thus we expect stability over a wider range of values of h if transit time is large, handling time is small and the time spent searching an empty patch is small. We have so far assumed that S , the average search time, is an increasing function of the average density, h. It is in fact possible for (23) to be satisfied if S is decreasing, so both S and V are negative. One can also think up circumstances in which this might happen: e.g. the predator might, when h is low, conserve his energy by travelling between patches as little as possible. However, a decreasing S seems rather unlikely, and we do not pursue this possibility here. Nor do we consider the possibility that the search rate, A, declines with h, though this is more probable. The variability in patch densities has played little part in the discussion so far. It seems likely that this variability will enter into the functional response in three ways. The first is that G(h) and S(h), the average number eaten and average search time when the average patch density is h, are different from g(h)and s(h),the expected number eaten and search time in a single patch whose density is h. (Technically, if D is the number of prey in a randomly chosen patch, Q(h) = E{g(D)) T-

PREDATION AND POPULATION STABILITY

53

while g(h) = g(E{D}),and similarly for S and 8 . ) The difference is a result of the non-linearity of g and s and, generally, will be greater when the variance of the distribution of D is greater. Jensen’s Inequality says that if g is an accelerating function, G(h)> g(h);the same is true of s and S, but it seems likely that g should accelerate more than 8. We might expect this to apply particularly for smaller values of h; for larger values, g might become more nearly linear and 8 might become decelerating, as happened in our model (see Fig. 9) when h (or D ) became so large that any additional prey was virtually always found (so g ( D + l ) - g ( D ) is near 1) in a time close to l / h ( D + l ) (so s(D+ 1) -s(D) is near l / h ( D + 1)). The effects of these tendencies, on (23), may be that, for highly variable distributions, V’ is initially large enough to satisfy TV’ > q V 2 even for small T and large 9, and to come close enough to SV‘> VS’ for (23) to hold. However, the larger V is the more rapidly it must increase to maintain T V ’ > ~ Vso ~ ,that a V which increases rapidly early is unlikely to maintain this relationship for very long, even for large T and small 7.At first the tendency for this relation to fail may be compensated for by the greater (for high variability) decrease in S‘, but this is a short-lived effect since S quickly becomes essentially linear. A more even distribution, however, might yield a V which increases more slowly initially so that, while it may never satisfy TV’ > q V 2 for small T and large 7,it may satisfy it over a wide range of h-values if T is large enough and 7 small enough. These speculations might suggest that highly variable distributions are more likely to yield stability (i.e. satisfy (23)for smaller values of 7 , the transit time, or larger values of q ,handling time) though probably at low values of h (prey density). This is somewhat borne out by the calculations made from our model, but those results are, in this respect, far too inconclusive for us to base any claims on them. The other ways we might expect variability to effect functional response involve its influencing the predator’s behaviour. There are two obvious possibilities. The first is that, in a highly variable situation, the predator is less likely to be able to judge overall. patch density accurately from his experiences, and that, in any case, the overall patch density will not be a good guide to the density of the patch he is in. This suggests that, when variability is high, the predator is likely to stay in patches that reward him well, and leave those that reward him badly, paying relatively little attention to his experiences in other patches. (Thus the behaviour in our model is more appropriate to highly variable distributions than to even ones.) Such behaviour may cause both S and V to be rapidly increasing functions of h, so that (23) might more easily be satisfied than for an even distribution. Finally, if the predator is not choosing patches randomly but is able 0

54

w. w.

MURDOCH

and

A. OATEN

to choose the more dense patches, variability will be an influential factor in determining stability. What its influence will be depends on how the predator chooses. One possibility is that by movement, smell, sounds, amount of leaf damage etc., the predator can distinguish between patches which are above or below some threshold density. If h is low, a highly variable distribution would yield more patches above this threshold than an even distribution would, so that for low values of h we might expect Q(h) to increase more rapidly when variability is high, so that, again, (23) might be easier to achieve when T is low and q high. The situation may be reversed when h is high, since a highly variable distribution might yield many more patches below the threshold density than an even distribution. Thus, if T and q are large enough to yield stabilizing functional responses, the stability might persist into larger values of h for the even distribution. These remarks, however, are little more than speculations. There is clearly a need for much more work, both modelling and experiment, before we can be confident we understand the role of variability here.

E. O T H E R

STUDIES O F PATCHINESS

The effects of patchiness on functional response or on prey stability have rarely been examined. Two recent papers which have discussed patchiness are those by Royama (1970) and Hassell and May (1973).

Royama actually considers a different question than we do. He looks at the number of prey eaten in a patch per unit of total search time, i.e. he is interested in the variation in the attack rate in space, rather than in time. Furthermore, his concern is less with a single species distributed in several patches than with several species, each living in a different “niche” (“the place where the prey species mainly occurs”). For any one of these species, the number eaten per unit of time spent by the predator in its niche, as a function of its density, may be linear (as in the LotkeVolterra equations), concave (type 2), or sigmoid; but if the density of this species should increase, the predator will spend more of his time in its niche, so the number eaten (of this species) per unit of the predator’s total search time may be convex. I n symbola: Suppose P,(h) is the number of prey taken by the predator per unit of the time the predator spends in the prey’s niche, when the prey’s density is h; and T(h)is the proportion of the predator’s hunting time spent in this niche when the density is h. (T(h)would depend on the density of other prey species too, but we assume these are constant.) Then F(h) = T(h)F,(h)is the number of prey of this species taken by the predator per unit of total hunting time. Royama then asserts that

PREDATION AND POPULATION STABILITY

55

Ah

if F,(h) is Holling’s function, P,(h) = ___ (where r] is handling time 1 +ugh and A is search rate), then F(h) will be sigmoid regardless of the form of T(h) provided only (presumably) that T(h) is increasing. This is clearly not true in general, but would be true if T increased fast enough, e.g. linearly. I n terms of our criterion, that F is stabilizing at h if F’(h)> F(h)/h,Royama’s F(h) is stabilizing if

T’(h)+ Xr]h T’(h)- T(h)) > o . h Hassell and May (1973) consider the functional response of a parasite to a patchily distributed host species. I n their model “D”, they assume there are n patches, 1, 2, ... n, and that in the ith patch (i = 1, 2, ... n ) there is a proportion at of the Ht hosts and a proportion 188 of the Pt parasites. Using the Nicholson-Bailey model, they have that, in the ith patch, the number of hosts surviving to the adult stage will be Hst = aaHt exp { - aglPt};these hosts produce an average of F offspring each, so at time t + 1 we have

(

Pt+, = FHt

2

at(1- exp { - apd‘t})

i-1

= -Ht+l

F ’

From these equations, conditions on P,a, {at}, {St} and the parasite equilibrium value P * can be established for the stability of the equilibrium. The procedure is, as for the Lotka,-Volterra equations, to take Ht = H*( 1 + ht) and Pt = P*( 1 + p t ) , expand non-linear functions by Taylor Series and drop all terms of more than fist order in ht and pt. One can determine the stability of the resulting linear difference equations by methods similar to those of Appendix I. Details may be found in the Appendix to Hassell and May (1973). n

C {a& i-1

1

exp ( - a&P*)}< 1 - F do not clearly reveal the kinds of sets {a(}, or the relation between them and the sets {pi}, that are conducive to stability, Hassell and May simplify the problem by assuming each pr = caf ,where c is a normalizing Because the stability conditions aFP*

n

= 1) and p is a “parasite

constant (to ensure i=l

aggregation index”.

w. w.

56

MURDOCH

and

A. OATEN

The larger the value of p, the more the parasites tend to aggregate in the areas of highest host density. In particular, obviously, 0 < p < 1 gives a rising decelerating curve, p = 1 gives a linear relation between pg and at, and p > 1 implies an accelerating curve. By varying the values of p, Hassell and May are able to investigate the extent to whioh aggregation by the parasite tends to stabilize the system. Unfortunately, even with this simplification, it is necessary to look at a particular example to get a feeling for the effects of changes in {at} and p on stability. The example Hassell and May consider has 1-a

and at = -for i = 2, 3, ..., n. Thus there is one dense patch n- 1 of hosts and (n- 1) equally thinly populated patches. It is then shown that (1) stability increases as p increases (i.e. fewer restrictions are required on other parameters for stability), in fact p generally must be > 1; (2) stability increases as n increases; (3) stability is greater when a is near 1/2 (see Section I11 B). Finally, after a brief discussion of a fractional refuge in terms of model D, an extension, model E, is given to allow for parasite interference. In this case, as in the more general cases of model D, the stability conditions are better given in diagrams, for which the reader is referred to the original paper. Roughly, however, the results are the same as for model D, with the addition that increasing interference increases stability. It is necessary, in work of this kind, t o make assumptions one knows to be unrealistic, in order to simplify the mathematics enough for the drawing of general, easily stated conclusions. Hassell and May's explicit assumptions seem to be ( 1 ) there are n patches and, for every point in time t , {at} and {pz} describe the distribution of hosts and parasites respectively; (2) within a patch, the Nicholson-Bailey assumptions apply; (3) 8.r = cay for some p ; (4) where there is parasite interference, it can be described by a single parameter m which reduces the effective (from the prey's point of view) parasite population in a patch from a1 = a

ptpt $0

(BzPt)'-".

We will not concern ourselves with (2) and (4): Hassell and May discuss both, especially (4), pointing out their defects. Assumption (3) is an admitted device to permit discussion of the effect of aggregation on stability. It is certainly reasonable to try to do this in terms of a single parameter and (the acid test for a critic) we have not been able to think of a better. It also corresponds well to standard cases: p = 0 means the parasite is distributed evenly among the patches; p = 1 means its distribution is identical to that of the host; p > 1 means the parasites tend to concentrate disproportionately in areas where hosts are abundant. Nevertheless, the system is not so simple as it may seem.

PREDATION AND POPULATION STABILITY

For instance the normalizing constant, c, is

Eli ap

-l;

67

thus cay, the

proportion of predators in the ith patch, depends not only on the proportion of hosts in the ith patch, but also on how the remaining hosts are distributed in the remaining patches. This is unavoidable

c f(ag) n

(if

,9g

=f(at)

so

c at n

= 1 whenever

=

1, then !(al) = at)

i= 1

i=l

and perhaps not unreasonable, but deserves notice. I n addition, having fir = caip means that no parasite spends time in an empty patch. I n fact, these parasites do not spend any time comparing different patches of prey or moving between patches: each parasite does all its parasitizing in one patch. A more serious assumption, not discussed by Hassell and May, is (1). This assumption implies that, in every generation, the hosts and parasites rearrange themselves so as to reconstitute the original distribution. After the emergence of the (t + 1)th parasite generation, the proportions in the ith patch are, for the host a'# = at yr/s and for the predator pg = at(1- y t ) / ( l - s ) where ya = exp ( - aP&) and 8

=

f atyg.

i=l

It is clear that these will not, in general, be the same as at and ,9r respectively, and may be far away. It is, however, possible that something like the required rearrangement does take place. If the parasite attacks host larvae, which are relatively immobile, while both the adult parasites and the adult hosts are highly mobile, it seems quite possible that the adult hosts may redistribute themselves according to the distribution of their food supply (which may be relatively stable), and that the adult parasites either follow the adult hosts or redistribute themselves later according to the distribution of the next generation of host larvae. Still, though it is possible, it is not very likely, and this somewhat implicit assumption may be the least realistic and the hardest to improve of those made in the paper. It is not easy to compare the system investigated by Hassell and May with the kind of system we have been discussing and modelling here. They are considering difference equations for a parasite and host with synchronous generations; we are considering differential equations in which generation time plays no role. The rate a t which prey are killed is a linear function of the number of predators in our model, but a decelerating function, exp { - a/?tBtPt), in theirs (this takes account of superparasitism). Our model is unlikely to be stable when transit timethe time taken to travel from patch to patch-is small or zero; theirs doe8 not have parasites travelling from patch to patch except initially,

68

w. w.

MURDOCH

and

A. OATEN

before they have begun to parasitize, so that transit time does not enter into their calculations. On the other hand, their model contains stabilizing features that ours lacks: they ignore handling time as “negligible”, which is often, though not always, the case; their system contains an implicit density-dependent migration of both hosts and parasites between generations, so that the proportions of each remain constant in every patch; and, when = cat$ no parasites are searching empty patches. We do not raise these points as criticisms of Hassell and May, but rather to note some of the differences in assumptions between their model and ours, and to show some of the limitations they have been forced to accept in order to obtain explicit answers in a quite general model. By using the very simple structure of the Nicholson-Bailey population models, Hassell and May have been able to produce interesting broad generalities about the host-parasite interaction, at the cost of omitting many parts of that interaction, some of which probably are unimportant to stability, but others of which may in some cases be crucial. By going to a rather more detailed and realistic model (though one still far from reality), we have been able to include a couple of the more important additional parts of the interaction, but in so doing have lost generality in so far as we cannot tie in functional response either with prey and predator’s rates of increase or with time lags, as can be done in population models.

F.

SUMMARY

When a predator is exposed to different densities of a single prey species in a homogeneous environment the almost universal result is that the number of prey killed per time unit increases at a decelerating rate (type 2, destabilizing functional response). A few predators (and insect parasites) have been found that give a type 3 response (initially accelerating, stabilizing). However, in nature, prey occur in patches that are separated in space and that contain different prey densities, and predators tend to stay longer in areas of high prey density. We present a model of such predatory behaviour that also requires the predator to spend time travelling between patches (transit time), searching empty patches, and handling prey. We conclude that transit time is a stabilizing force, the range of stabilizing conditions increasing with transit time. The range of conditions for which transit time is stabilizing decreases as handling time increases and probably also as the time taken to search an empty patch increases. The relationship between stability and the variance of prey density among patches is not clear, though our general criteria for stability suggest that increasing

PREDATION AND POPULATION STABILITY

69

variance should yield more stability. This relationship would be strengthened if the predator could actively select areas of high prey density or could relate the reward rate in a patch to the average reward rate. V. T W O - P R E YSPECIES

A.

RELATIVE ATTACK RATES

I n this section we are concerned with the short-term response of individual predators to variation in prey abundance where more than one species of prey is available. Since we want to know if predators can stabilize their prey, we are ultimately interested in the relationship between the density of each prey species and the numbers of that prey that are killed, a relationship that is examined in sub-section V B. However, there is good evidence that predators react to the relative frequency of their prey species, and that to predict absolute attack rates on each prey species we need to determine the relationship between the relative frequency of the prey species in the environment and their relative frequency in the diet. I n sub-section V A we examine this relationship and explore several predation mechanisms that determine it. For simplicity we will consider only two prey species.

1. Predator switching and apostatic selection I n Section I V we reviewed the results obtained when a predator is given varying densities of one prey species, and we noted that the almost universal finding is that such mortality is destabilizing (type 2 functional response), unless patchiness or some other complication is added. However, many predators feed on several prey species and they might cause stabilizing mortality upon some or all of these species as they distribute their attacks among the prey species in response to the species’ relative frequency. I n this section we consider the possibility that, in a system with two prey species, the individual predator might concentrate a disproportionate fraction of its attack potential upon the more abundant species, and might correspondingly spare the rare species. As the relative frequency of the two prey species changes, the predator might then alter its diet so as to concentrate on the alternative prey when it becomes the more abundant. This is what Murdoch (1969) termed “switching”. Clearly, since switching refers only to relative numbers of attacks and to relative prey abundances, one cannot determine directly if the mortality on either prey is deneity-dependent, as distinct from frequency-dependent. However, the process clearly is one that might lead to stabilizing mortality in

60

w. w.

MURDOCH

and

A. OATEN

some circumstances. I n this section we provide a criterion for switching and then explore mechanisms that might produce it. We want t o explore here a quite restricted part of predation. How does the relative contribution of a prey species to the diet of a single predator vary with its relative frequency in nature? Also, we want to look at the diet at each particular relative frequency of prey, for a very short time only, so short that we can assume the prey abundances are not changed by predation. That is, essentially we want to look at the instantaneous behaviour under a range of conditions. For this reason the theory (and the experiments where possible) are set up so that prey density is fixed, or at least changes very little, during the experiment. Consider fist the null case. The predator faces two prey species at densities H , and H,, and we assume it encounters them in proportion to their actual abundances. Some fraction of the encountered individuals in each species will be eaten. The ratio of these fractions is taken to be the preference the predator has for one species over the other at any given relative frequency of the two prey. This preference can be thought of as the dumber of times the predator will select one species over the other. If, as the relative abundances of the two prey vary, preference remains constant, the ratio will remain constant (even if the actual fractions eaten are changing, say with hunger). Letting c be the preference, the null case is written:

where N , and N , are the number eaten of prey species 1 and 2, respectively, and H , and H , are the densities of prey species 1 and 2, respectively. It is perhaps immediately obvious that we have a problem in estimating preference, c. This will be a minor, statistical, matter if preference remains constant for all values of H J H , ; but how are we to meaaure preference if the null hypothesis is wrong, i.e. if preference varies as a function of H,/H,? This can be solved in practice by fist measuring preference in the situation where the two prey are equally abundant. Having obtained this estimate of c, we can ask if the same value obtains a t other values of H J H , (in which case the null hypothesis is accepted). We follow this sequence throughout this section, and for convenience we call the preference when the two prey species are equally abundant, the “preference at equality”. In the simplest case of the null model the prey are equally preferred at equality (i.e. there is no preference at 50:60), and also at every other value of H J H , . There the fractions taken from both prey species

PREDATION AND POPULATION STABILITY

61

are equal and their ratio, c, is 1, so that the ratio in the diet is always simply the ratio given. When there is a preference at equality, c > 1 if species 1 is preferred and c < 1 if species 2 is preferred. The null hypothesis is that c is then a constant at all values of H,/H,. The switching hypothesis is that c is in fact not constant, but increases as R , / H , increases. The simplest form that switching might take would

HI

’

H2

FIG. 11. Switching and the null hypothesis. The ratio of species 1 to species 2 in the diet (Nl/Na)is plotted against the ratio available ( H l / H a ) .The straight lines represent the null case (a)where there is no preference between the species and (a) where species 2 is preferred five times as much as species 1. In both caaes the curve shows switching when preference (c) increases aa (Hl/Ha)a.

be for c to increase linearly with the ratio H J H , . Then if cf is the . even be a step preference at equality, N J N , = C ~ ( H , / H , )c~ could function. Figure 11 shows the expected diets under the four circumstances obtained by combining: preference at equality or no preference at equality and switching or no switching. It should be noted that in using “preference”, we are not implying any particular behaviour but simply that the predator does not take prey equally when they are equally available. A stochastic version of this model, allowing for a decrease in prey density as they are eaten and providing a method of estimating

62

w. w.

MURDOCH

and

A. OATEN

preference, can be found in Manly et al. (1972). The method assumes that an experiment is run until half of the original prey are eaten. Ratios are often awkward to deal with (e.g. when N , = 0) and in practice the data are best transformed to proportions. Where the proportion of species 1 in the diet is

P, =

~

N1

N,+ N,’

then the null cam is

written :

P -

c=,

- cH,+H,

and letting the proportion of species 1 in the food available be PI, then

P -

CPl

- l-P,+cF,

Thus, in this form, the null hypothesis of constant preference predicts that P, as a function of P, is a curve (see Fig. 12) except where there is no preference, and a line of slope 1 results. As usual, switching implies that c is an increasing function of P,, and predicts that observations will lie below the expected “constant preference” curve at low values of P, and above the curve at high values of P,. I n particular, if we assume that c increases linearly with HJH,, as we did above, we have

c’Fla - (1 - P1)2+c’P12

P -

where c = c’ when F , = 0.5. The above theory does not place any constraints on absolute prey density ( H , + H , ) or on the absolute numbers eaten. Two important considerations are therefore omitted. ( 1 ) What is the relation between prey density (Ha) and number killed (Nt)? We need to specify this relationship in order to explore the consequences of switching for prey stability. We propose to do that first by finding out in this section, what are the mechanisms that lead to switching. Obviously, different mechanisms will have different consequences for the relationship Nt = f ( H a ) , so we need to specify them first. ( 2 ) Total prey density almost surely has some effect on whether or not switching occurs. Thus, ideally, we would like to do factorial experiments in which a range of prey ratios is examined at each of a range of total prey densities. Again, we will have more insight into this probable effect when we have specific mechanisms to work with. I n the meantime, it is probably better to keep total density fixed to avoid this factor. A completely analogous model, “apostatic selection”, has been developed in the literature of ecological genetics (Clarke, 1962), though

PREDATION AND POPULATION STABILITY

63

the null case seems to have been first stated in an explicit algebraic form by Clarke (1969). I n this case, predation is upon two morphs of the same species occurring at varying relative frequencies. The motivating idea here is to have the predator favour (i.e. spare) the rare

FI FIU.12. Switching and the null hypothesis. The proportion that species 1 forms of the diet (Pl)is plotted against the proportion it forms of the food available (Pl).In case (a)there is no preference in the null case; in case ( b ) species 2 is preferred five times as much as species 1 in the null case. The switching curves are the function (27) in each case (see text).

morph and so maintain a polymorphism. Since the two approaches are formally the same, we treat them together below in searching for tests against data. a. Tests of the model. For the sake of clarity, although no doubt with

some unfortunate corollary loss of drama, we shall here summarize the general procedure and the conclusions of this section. We examine several sets of results from different predator-prey systems, and in each case we first examine predators’ preference when the two prey are equally abundant, “preference at equality”. We then examine the diets that obtain when the predators are presented with prey at unequal abundances, to see if preference is constant or if switching has occurred. The conclusions are as follows: 1. If preference at equality is strong, and consistent among predators, then it remains constant at other values of HJH,, i.e. the null hypothesis is accepted. 2. If preference at equality is weak when averaged over the predator population, but is strong and highly variable among individual

64

w.

W. MURDOCH

and

A. OATEN

predators, then switching occurs when H,/H, is vaned (i.e. preference is a function of H,/H,). 3. If preference at equality is weak, and consistently weak among predators, then it remains constant at other values of Hl/H2, and the null hypothesis is confirmed. Turning now to tests of the models, note that the model assumes that at a given ratio of prey, H , and H , remain fixed through time, so that the prey must be replaced as they are eaten, or only a small fraction of the prey should be eaten, or their relative abundances should remain essentially unchanged as they are eaten. Unless stated otherwise, the data below come from experiments where these assumptions were approximated. The arguments for using models with such constraints were discussed earlier in Section IV. I n most of the experiments discussed below, the total number of prey is also held fixed as prey ratios are varied. This situation will not often occur in nature, but neither will obvious alternative situations such as keeping one species a t a constant density while varying the other. The advantage of fixing total prey density is that only one factor (the ratio) varies in the experiment, whereas if one species is kept fixed and the other varied, then both total density and the ratio are varying together, and a factorial design would be more appropriate. Naturally, in this case it is important that individual prey items of different species be roughly comparable in size if the total density of food is to remain constant in this design. The first set of experiments was done with seashore snails (Murdoch, 1969). I n one case they showed strong preference at equality and, in the second, weak preference a t equality. The results of these experiments suggested the three generalizations listed above (Murdoch, 1969; Murdoch and Marks, 1973). I n the following pages we present results from a wide range of predators to try to determine under what circumstances switching does or does not occur and, where it occurs, what the mechanisms might be.

b. Evidence for switching Seashore snails Thais and Acunthina attack a wide range of sedentary prey on the rocky shores of southern California. Each snail (or group of snails) was provided with one of several ratios of the two prey species, a ratio that was maintained over several weeks either in the laboratory or on the shore (Murdoch, 1969). With this experimental design it is assumed that we can extrapolate from the results of differences among treatments at one point in time to differences in time as prey abundances vary, though

PREDATION AND POPULATION STABILITY

66

we might expect predators to exhibit a lag as they respond to a new prey ratio. I n the first two situations the predators were given two similar species, but showed strong preferences at equality and no switch occurred (see “Evidence for the Null Hypothesis”, p. 79). Given these results it seemed that the snails would be more likely to have labile preferences if they were presented with two prey species between which they had weak preference at equality. Acanthina was, therefore, presented with a mussel (Mytilus edulis), and a barnacle (Balanusglandula). Preference at equality was weak in all the experiments; however, a striking result was that there was great variability among individual snails, some snails feeding mainly upon the barnacles and others mainly upon the mussels (Fig. 13) even though they were all given equal numbers of the two prey. An explanation for this variability is that at first the snails had no preference (they were nai‘ve to these prey), but that the first meal and each meal thereafter had an effect on preference, so that meals were not a series of independent “Bernoulli” trails, but rather each meal affected subsequent meals. This effect was established by showing that a snail could be trained to prefer either species by feeding it a pure diet of that species. (If meals had been an independent series of trials, as in coin-flippingexperiments, then we would expect a binomial distribution, as is approximated by other predators, e.g. Ladybirds in Fig. 13). Analysis of snail diets through time showed that the snails maintained the same preference, after it was established, so long as they were kept at the same prey ratio. Thus, in the 50: 50 prey ratio situation, we may think of the predators’ diets diverging from this given ratio at least initially, thus building up the extreme diets seen in Fig. 13. The existence of initially weak preference is supported by the fact that the average diet was about 5 0 : 5 0 , by other experiments in which half the snails ate a mussel and half ate a barnacle when given a choice between one of each species, and by the fact that these snails were then easily trained to prefer the alternative species. Such a tendency to become conditioned, or trained, to one of the species can lead to switching when ratios are varied. This can be seen from the following argument (Murdoch, 1969, Appendix). Suppose species 1 forms 75% of the food available. On the first meal the snail will eat species 1 with probability 0-75. Then 0.75 of the snails have eaten species 1 for their first meal. If we assume that having the meal of species 1 causes a training effect, ci, then the probability that these snails will eat species 1 on the next meal is 0.75+ a. Assuming further that training is symmetrical, those snails that ate species 2 first have probability 0 - 7 5 - a of eating species 1 next time. Then, for the n+ l t h meal, the probability that the average snail will eat species 1 is

w. w.

66

MURDOCH

and

A. OATEN

6 4

(a 1

2

12 (f

8

1

4

8 6

(91

4 2

0

0

0.2 0.4 0.6 0.8 1.0

0 . 2 0.4 0.6 0.8 1.0 Proportion

of one species in the diet'

FIG.13. Each histogram is the number of predators (frequency)that had a given proportion of one out of two species in its diet when offered equal numbers of the two species. Snails (a),pigeons (b), Stentor attacking Eugkna and ChlanzydO~n~na8 ( c ) , and guppies (d) had weak and variable preferences. Stentor attacking EugZena and T e t r a h y m a (e) had strong consistent preferences; ladybirds (f)and bluegill ( 9 ) had weak consistent preferences.

PREDATION AND POPULATION STABILITY

67

+

0-76 a(2Pn- l), where Pn is the probability that the nth meal wm species 1, and of course no probability can exceed 1. Clearly, since P, must be 0.75 or greater for all n, the probability Pn+1 is always greater than 0.76, which gives switching.

The invariant correspondence between training and switching in this simple model arises from the assumption that the effects of training are symmetrical. It is in fact possible to have asymmetrical training effects that do not lead to switching (Oaten and Murdoch, 1974a, b). I n the case of the snails, switching indeed occurred when the prey ratios were varied (Fig. 14), but only when “patches” of the abundant prey had been provided before the experiment, to ensure that training

FIU.14. Switching in snails. P,is the percentage of barnacles in the food available end P , is the percentage in the diet. The other prey waa a mussel. The line is the null hypothesis with no preference.

to the abundant prey occurred. When no “patches” were provided, no switch occurred, though there was enormous variability among replicates so that the null model provided a poor description of the data. Thus, even where average preference at equality is weak, variable, and the mechanism of conditioning has been demonstrated, the circumstances permitting the mechanisms to operate must be present. On the other hand, these circumstances (prey patchiness) do occur on the seashore and elsewhere, and we might therefore expect this mechanism to operate there (see Murdoch, 1969, for a more extended discussion of this point). One would also predict that when a predator has become trained to one species and then the alternative prey becomes the more abundant,

68

w. w.

MURDOCH and A. OATEN

the predator will turn t o attack disproportionately the newly abundant species. To test this in the snail experiment described above, one set of treatments were fed pure diets of either barnacles or mussels and then were given a preponderance of the alternative species in the food available. Figure 16 shows the change in the diets in one of these treatments over a five-week period, from 81% barnacles to 14% barnacles. However, the diets of the snails in the corresponding treatment (trained on mussels then given mainly barnacles) did not change in a consistent fashion with time.

FIU. 15. Shift in diets of snails through time. These snails were first trained to barnacles then given 83% mussels for five weeks. Each point is the mean of six observations.

The snail results are, therefore, somewhat equivocal. The training treatment illustrates the mechanism but “forces” the switching; and the change in the diet through time was asymmetrical between treatments. There is some evidence concerning the mechanism by which the snails switch. The snails hunt at least partly by smell (Wood, 1968). As they hunt they move over the prey, making contact with many individual prey with their tentacles and foot, and usually seeming to try, unsuccessfully, to attack several before finally settling down to attack a particular prey. Thus, they appear to be rejecting prey, in the sense that many prey are encountered but not attacked. Now, in order to switch they must concentrate upon one prey species to the relative exclusion of the other, and presumably (but not necessarily) the acceptance rate for the abundant prey increases. But diets are so extreme and so few of the rare prey are eaten that the rate of rejecting the rare prey must also rise, so that the snails must actually reject some

PREDATION AND POPULATION STABILITY

69

opportunities for attack that they would otherwise use. This appears to be less efficient behaviour than keeping rejection rates low, but possibly it is more efficient physiologically to concentrate on one prey species at a time-even if the rate of feeding does not increase, the metabolic cost of feeding may decrease. Alternatively, it may be more efficient to use one method of attack rather than two, and different procedures are used for attacking barnacles and mussels. Unfortunately Murdoch’s 1969 results do not provide data for examining feeding rates. This mechanism might be expected t o lead to switching at high total prey densities, but not at low prey densities, since at low densities the intervals between meals are long, and it would probably be more profitable to bear the cost of a mixed diet (if such a cost exists) than the cost of a high rejection rate. These results suggest that weak average preference at equality is a necessary but not sufficient condition for switching. However, switching should occur when different predators have extremely different preferences at equality. I n this case, preference at any time might be strong but should be reversible by training. The other necessary condition is that, as abundance8 vary, there be an opportunity for the predator to become trained (or conditioned) to the more abundant prey.

Birds-pigeons Murton’s data (Murton, 1971) from pigeons feeding on various seeds in a field, show interesting similarities to the snail results. Murton distributed maple peas and tic beans on an area 84 m square, both at equal and unequal densities. The seeds were distributed evenly and together over the same area. Pigeons that came to feed in the area were caught (the bait had a narcotic) and their crop contents examined. The study is unusual in that large numbers of birds were involved (80 or so) and that each bird had many “prey” in its crop (occasionally up to 90). The birds fed in flocks, which may have had some effect upon diets (see below). When the abundance of maple peas and tic beans was very similar, the birds, on the average, showed only a very slight preference for maple peas over tic beans (c = 1-59 for maple peas). But there was enormous heterogeneity among birds, with the majority showing significantly extreme diets, i.e. they were not choosing at random but presumably during the experiment had developed a “preference” for one or other species of prey, with roughly equal numbers preferring each species (Fig. 13). Similar results were obtained with other pairs of seed types. This is precisely the divergent feeding process exhibited by the snails, and would lead one to predict that switching should occur if prey ratios were varied. Murton then provided pigeons with an extreme ratio, 41 tic beans

70

w. w.

MKKRDOCR and A. OATEN

and 9 maple peas, i.e. 82% tic beans. With c for maple peas = 1-59,the expected percentage of beans in the diet, under the assumption of constant preference, is 74%. An observed percentage greater than this would indicate that preference for tic beans was greater when tic beans were at a high relative frequency. The diets of 81 pigeons were measured (Dr Murton kindly provided us with the original data which is not in the paper). Of these, 11 pigeons ate fewer than 10 seeds and we ignored them, though they showed the same sort of results as the other 71. The mean percentage of beans in the diets of the remaining 71 was 90-7y0, providing clear evidence for switching. Interestingly, two pigeons out of the 81 showed an extreme preference for the rare prey maple peas; their crops contained 0% and 4.9% tic beans. This is the kind of result that is expected on the basis of a probabilistic model of switching, but which is hard to demonstrate unless large numbers of predators are examined.

Birds-quail I n experiments done by Manly et al. (1972) pairs of quail were given a choice between pieces of pastry coloured either red or blue. The relative frequency of these two colours was varied among treatments, the total number of “prey” given being 20. Prey were not replaced and the experiment was stopped when 10 had been eaten. Each pair of quail was run 10 times on 10 different days at a given prey frequency. Manly et al. developed a stochastic technique for measuring preference when prey frequencies are changing as prey are eaten. They showed that preference for red is an increasing function of the relative abundance of red, i.e. switching occurred. (In fact they graph In a, which is In l/c in our notation, which was roughly linear with prey frequency.) We have plotted their data in Fig. 16 which shows switching clearly. Since a large proportion of the prey were eaten, the “proportion offered’’ was calculated as the mean of the initial proportion and the final proportion. We are not able to say whether these results do or do not fit the prediction concerning variable diets. Preference at equality was weak (c = 1.5), but not particularly variable. However, each observation was obtained from two birds and was a mean of 10 separate trials for the particular pair of quail. The paper does not say if other food was given between meals, but we guess that it was. For these reasons we would expect the means to be much less variable than individual observations based on individual predators. Both the quail and the pigeons switched when presented with two prey that were mixed together in the same area. The switching mechanism is less clear than in the snails, since we do not know what

PREDATION AND POPULATION STABILITY

71

the relative “discovery” rates were for the two prey. However, presumably the birds ate most of the grain or pastry that they saw, rather than seeing some but rejecting them. If this was so it seems that there was a variable “non-discovery” or “ignoring” rate, and that an individual of the rare prey type had less chance of being seen.

0 Fl

FIQ.16. Switching in quail given red and blue pastry. P , is the proportion of red pastry in those given, P , is its proportion in the diet; each point is the mean of 10 observations on two quail. The curve is the null hypothesis with weak preference for red pastry. Data from Manly et al. (1972).

(Tinbergen’s search image notion may be appropriate here-see

Section

V A2.) Presumably the selective value of this behaviour is that the increased probability of “seeing” any given individual of the abundant species that is within visual range more than compensates for the reduced probability of “seeing” a given individual of the rare species in that range.

Fish-rudd The earliest results that may demonstrate switching are from Popham’s (1941) work with fish attacking several morphs of a corixid bug. He gave three rudd a choice between two corixid morphs occurring at three different frequencies. The data show a very slight but statistically significant deviation from expected under a null model of no frequency-dependent predation (no switching) (Elton and Greenwood, 1970). Unfortunately there seems to have been no replication, and we cannot check the behavioural basis. Preference at equality (c = 3) was the strongest found so far in a, predator that exhibited switching.

72

w. w.

MURDOCH

and

A. OATEN

Profozodtentor Dr David Rapport has kindly allowed us to analyze some unpublished data obtained by providing the ciliate protozoan Stentor with pairs of prey species (Rapport, 1974). The predators were allowed to feed for 20 min on a suspension of the two prey species, during which time they ate about 10% of the prey. They were then placed in formalin and the diets were determined by counting the prey inside the predators, which are transparent. There were three pairs of prey, each giving different

Fl

FIQ. 17. Switching in Stentor given Euglena and Chlamydomonae. F , is the proportion of Euglena in the food available, P , is its proportion in the diet; each point is the mean of eight observationsand the bars represent k 2 S.E. The cullre is the null hypothesis with weak preference for Euglena. Data from Rapport (1974).

results. In each caae the density of one prey was the same in all treatments while the density of the alternative prey was vaned, so that total prey density varied among treatments. The experiments have the advantage of high replication ( 12 per treatment) but unfortunately the number of prey eaten frequently was small ( c10) at the lower densities. Stentor was given a choice between Euglena and another flagellate, Chlamydomonas. There was no equal abundance treatment, so we cannot estimate preference independent of differences in relative prey abundances. However, the data suggest that preference at equality would have been weak. When Euglena was the less abundant species, preference

PREDATION AND POPULATION STABILITY

73

was very weak (c N 1.5). When Euglena became the more abundant species there was a clear increase in preference for Euglena, providing good evidence for switching. This is illustrated in Fig. 17, where in order to draw the expected curve, we chose the value of c midway between its value in the two central treatments. I n this case it seems that the change in preference was discontinuous. When Euglena was rarer, c ranged from 1.5 to 1.6; when Euglena was the more abundant prey, c waa roughly 6 , except for the very high value when Euglena formed 0.73 of the available prey (Fig. 17). Thus there may have been a shift in behaviour from one mode to another when Euglena became abundant. Where preference was weak, and especially in the treatments closest to equal abundances, it was highly variable among replicates (Fig. 13). Furthermore, large numbers of prey were eaten in these treatments (up to IOO), so that the variability is real. Thus, this experiment provides a further confirmation of the prediction relating switching to diverse diets when prey are equally common. We do not know what the mechanism is that leads to switching. Possibly a variable rejection rate is involved since the prey come to Stentor mixed together in a current of water. Fish-guppies I n the experiments discussed so far the two prey species were distributed either together or with a very large overlap. Thus the predator could encounter the two species together by using the same searching behaviour. The predictions discussed so far also concerned such a situation, with the mechanism leading to switching conceived as some sort of “training” to the abundant species. I n this section we present some new experiments designed to test an entirely different mechanism that could operate in a different kind of situation. Consider a predator that attacks two prey species that occur in two different sub-habitats (areas) within the predator’s habitat. There will then be some cost to the predator in leaving one of the areas and searching for prey in the alternative area, but there will also be a potential benefit in that in the other area, the other prey might be more abundant and, therefore, more rewarding (presuming the prey are equally nutritious, catchable and so on). The predator might respond in one of at least two fashions, especially if the prey in each area are in patches so that the predator moves from patch to patch within an area. The predator might remain in one area with a certain probability, provided it is rewarded at or above some threshold rate, leaving the area only after the reward rate falls below the threshold value. Alternatively, the predator at intervals might leave the area it is in and “sample” the alternative area; if the reward rate there is greater, then the predator

74

w. w.

MURDOCH

and

A. OATEN

stays there and “samples” from that area. I n the first case the predator responds to the reward rate as it relates to an absolute standard, in the second it has to be able to compare current and recent reward rates. Either mechanism, but especially the second, could lead to the predator’s spending more time in the area that contains the more abundant prey species, and having that prey represented disproportionately in its diet. No learning or training is involved; the predator simply plays the better of sampling should not be bet more of the time. Presumably the too great or the predator might sample very infrequently. Also the absolute density of the prey may affect the outcome, as we discuss below. Guppies were placed in a 22-gallon hatchery aquarium with one prey species that stayed on the bottom (Tubifkid worms) and one that floated on the top (Drosophilu). The guppies perceived the prey from leas than 6 cm away and the water was 20 cm deep, so the fish clearly had to leave one “sub-habitat” to search for prey in the other. However, the time taken in transit was small, so the cost of going to sample was small. The results reported here are incomplete, and the complete results, together with details of the methods, will be found in Murdoch et al. (1974). The movements of the predator and the times of attacks and of meals were recorded by observers on a continuous recorder. Prey were replaced as they were eaten, care being taken to prevent the fish from seeing the delivery of the new prey. If the predator was hungry and not “upset” it attacked and ate all perceived prey. Sometimes the Drosophila was too large to swallow and would be spat out, though it was generally attacked again and eaten. Thus, unlike the snails and birds described above, there was no rejection rate or “ignoring rate”. Predation at equal prey densities, by fish naive to the two prey, was examined by allowing each of 20 predators to take four meals at a density of one Drosophilu and one Tubificid. The results (Fig. 13) demonstrate the characteristic extreme diets and average lack of preference at equality that lead us to predict switching. The average diet oontained 46% Tubificids, but 15 fish ate only one or the other of the two prey species. Corresponding to the extreme diets, the seven fish that ate only TubXcids spend an average of 96% of their time a t the bottom, while the eight fish that ate only Drosophilu spend SOY0 of their time at the top. There is a consistent bias here and in all other experiments towards the bottom of the aquarium. All other experiments were run with fish that had been exposed to both prey at regular intervals over several weeks. There is a problem in that a given fish a t a given point in time tends to hunt either mainly

PREDATION AND POPULATION STABILITY

76

at the top or mainly at the bottom, even if it is fed only fish food. Since some of the fish food remains on top and some sinks to the bottom, this behaviour may reflect the recent pattern of reward. At any rate, it means that fish were never “neutral” with respect to top and bottom at the start of an experiment. I n the subsequent experiments, therefore, we made certain that each fish had fed at both top and bottom each day for several days before the start. The switching experiment was run as follows. Each fish was tested once a day when it was allowed to eat 20 prey or until it stopped searching. This always took less than one h. The fish was tested at each prey ratio for three days, then moved to the next ratio. There were four ratios available, ranging from 4:1, Drosophila to Tubificids, to 1:4, so the experiment ran for 12 days. Each fish started at one extreme ratio and moved in steps to the opposite extreme ratio, thus modelling directly a situation in which the prey relative abundance5 are changing through time. Guppies are temperamental and get “spooked” and refuse to eat (they change colour and remain in a corner at the bottom of the aquarium), so it is difficult to get a large number that complete the course perfectly. We present data (Figs 18 and 19) for the first five fish to finish successfully. This is probably the clearest and most convincing example of switching so far. I n looking at Fig. 18, especially, the reader should pick out the sequence of diets for each fish separately since fish were variable. A remarkable phenomenon is that the fish seem to be able to distinguish between reward rates that are not very different (at ratios 3: 2 and 2: 3). Without further experimentation and analysis we cannot describe the mechanism that produced switching, but our guess is that the guppies respond to two kinds of information. First, when they are being frequently rewarded, they tend to remain where they are and sample the alternative habitat infrequently, i.e. they respond to the absolute reward rate. Second, they do sample the alternative habitat and can distinguish between the two reward rates and return to the better sub-habitat. We do know that as the proportion that the bottom food contributes to the diet increases, so does the proportion of the total time spent a t the bottom (Fig. 20). Figure 21 shows that, even in this simple situation where the cost of sampling is so small, the predator gains by selecting a restricted diet as a result of switching. Thus, the selective advantage of switching is that prey are encountered more frequently. Both the mechanism and its selective advantage in the guppies thus seem to be different from those in the snails and birds discussed earlier. It may be, too, that the different mechanisms have different consequences for absolute attack rates, and therefore for the predator’s

w. w. MURDOCH and

76

-

'.O

A. OATEN

0

(0)

A

O

-

0.8

0

*

A

-

OC

A

A

0 0 A 0

-

0.6

0

!-

A

A 0

-

0.4

- 0

-0 A

0.2

-

0

0

.

0

A

A

--.-0.-o

lo-

08

M

(b)

K

I

0

O 0

0

0

-

0 0

0.6D

0 0

0.40

0

0

0.2 - -o 0

0

0

0

-0-0

I

2

I

3

'

4

1

5

I

6

I

7

I

8

'

9

I

10

I

II

'

12

FIQ. 18. Switching in guppies attacking Tubificids (T) on the bottom of an aquarium and Droaophda (D)at the top. Experimentsstarted on day 1 and lasted 12 days. Each point is the diet of a given fish (three fish in top graph, two in bottom graph) on a given day. The short lines are the proportions available for each three-day period.

77

PREDATION AND POPULATION STABILITY

ability to stabilize the prey. I n the sort of systems modelled by the snails and birds, we suggested that switching is more likely to occur at high prey densities. By contrast, the guppies may be more likely to switch when total prey density is low, because there should be strong selection for a predator to maximize its reward rate especially

0 2

04

0.6

0.8

5 FIQ.19. Switching in guppies. F , is the proportion of Tubificids in the available food, P , is their proportion in the diet; each point is the mean diet of a given fish (five fish) based on three days’ feeding. The line is the null hypothesis with no preference.

when the prey are scarce; i.e. when both prey are rather scarce, the predator will benefit by concentrating on the more abundant species. But when all prey are dense the predator might not respond much to differences in prey densities, either because it is feeding at a maximum rate in both sub-habitats, or because both sub-habitats reward it above the threshold rate that keeps the predator in the area. More

w. w.

78

MURDOCH

and

A. OATEN

experiments need to be done in which total prey density (and cost of sampling) is varied in order to answer these questions. But this discussion illustrates that; some notion of the mechanism involved is necessary in order to go from switching to absolute predation rates. Ware (1971) independently developed the idea of a necessary rate of reward, below which the predator will tend to stop searching in one place and move to another. He introduced a trout into an aquarium

t 08

6

1

1

S 0

0 0 0

.

0

0

I

0

I 1

0.2

I

04

I

I

I I

0.6

0.0

1.0

T

FIG.20. The proportion of the time spent on the bottom by guppies increases a8 the proportion of bottom prey, tubificids (T), in the diet increases. Each point represents the results from one day for one fish (there are five fish).

that had a species of amphipod on the bottom and showed that the trout would concentrate on the amphipods, but after some time would begin to shift to searching, or at least swimming, in the water column. There seemed to be a threshold rate of reward, above which the predator remained hunting for the amphipods and below which it started to shift its attention to the water column. The threshold rate was, of course, not deterministic but had a large stochastic element. The threshold for one prey species when i t was alone appeared to be constant regardless of how hungry the predator was, but for the other prey species it appeared to increase with the trout’s hunger. There was no alternative prey in these experiments, so we cannot use them to

79

PREDATION AND POPULATION STABILITY

2 o L L 0.2

0.4

0.6

0.0

1.0

T

FIG.21. Percentage efficiency of guppies is higher at extreme diets, when the proportion of tubificids (T) in the diet is either high or low. Each point is the diet and efficiency of one fish on one day. The maximum number of prey eaten per minute for a given fish is called 100 and all other days are expressed aa a percentage of that maximum. There are five fish.

examine the switching mechanisms proposed here, but they do support the general and obvious hypothesis that fish change their searching behaviour in response to reward rate. c. Evidence for the null hypothesis

Snails The first two sets of experiments done with the snails provided a good fit to the null model of no switching (Murdoch, 1969). The predators were given two similar prey species (mussels) but showed very strong preferences a t equality (c = 10 or more). Figure 22 shows one set of results from one of the pairs of mussel species. A good deal of unsuccessful effort was spent trying to break this strong preference, by feeding the snails a pure diet of the less preferred species, by making it more easily available, and so on. But for the following reasons this seems to be a real “built-in” preference: the less preferred species was more common (and was sometimes attacked) in the predator’s natural environment; when they each were presented alone the rate of feeding upon each species was about the same and both prey were encountered

80

w. w.

MURDOCH and A. OATEN

roo -

I

0

20

I

40

I

I

60

80

0

5 FIG.22. Snails fit the null model. F, is the percentage of the mussel M. e4ldio in the food available, PI is its percentage in the diet; the alternative prey is M. ccdqornkznw. Each point is the mean of three replicates; the barn indicate the range. The curve is the null hypothesis with strong preference.

in direct proportion to their abundance. The preferred species (MytiZw edulis) is thinner-shelled, and all predatory seashore species examined in the Santa Barbara area prefer it to the other mussels. These data, and those snails exhibiting weak preference, led to the prediction that predators with a strong preference at equality, a preference that is consistent among different predator individuals, should not switch. Protozoa The only other example of moderately strong preference at equality that we have found is from Rapport’s experiments with Stentor (Rapport, 1974). We described his experimental procedure above. Stentor showed a moderate preference (c = 3 to 6) for the ciliate Tetrahymena over the flagellate Euglena. When prey ratios were varied by varying the density of Euglena while holding that of Tetrahymena constant, preference remained remarkably constant (Fig. 23). Our prediction in this case is that there should be little variability among diets when the two prey are equally available. Unfortunately, there is no equal abundance treatment. Also, few Euglena were eaten in the treatment closest to equality (2.5 Euglena: 4.0 Tetrahymena), and in some replicates close to zero were eaten. This introduces variability simply owing to the numbers’ being small. However, the data in Fig. 13 still fit the prediction of consistent diets. Thus, so far,

PREDATION AND POPULATION STABILITY

81

when the predators show strong consistent diets at equality there is a good fit to the model. A somewhat riskier prediction is that when different predators have both weak preference at equality and similar diets, no switch should occur. The notion here is that if all predators show a similar indiscriminate choice between the prey, this suggests that feeding upon one of the species does not produce any “conditioning” or training

FIU.2 3 . Stentor attacking Euglena and Tetrahymena fit the null model. F , is the proportion of Euglena in the food available, P , is its proportion in the diet. Each point is the mean of 12 observations; the bars represent & 2 S.E. The curve is the null hypothesis with strong preference for Tetrahymena. Data from Rapport (1974).

effect to that species. We think that such a training effect is needed to produce switching.

Ladybirds Figure 13 shows the diets of ladybirds given a choice between equal numbers of two aphid species, Acirthosiphon pisum and Aphis fabae (Murdoch and Marks, 1973). The ladybird diets show a strong central tendency, and are approximately binomial as would be expected if each diet consisted of a sequence of independent meals, like a series of coinflips. The ladybirds were therefore used to test the prediction that weak but consistent preference at equality indicates that no switch will occur at unequal ratios.

a2

w. w.

MURDOCH

and

A. OATEN

The ladybirds’ behaviour differed from that of the snails discussed above. The snails examined and touched many more prey individuals than they attacked and seemed to be highly selective. They can also detect their prey from a distance (Wood, 1968). By contrast, the ladybirds appear to attack all prey they encounter while in an “attacking phase” (Marks, 1970), i.e. they are not selective, and can only detect

FI

FIQ.24. Ladybirds attacking two aphids, A . pi8um and A . fabae, fit the null model. P , is the proportion of A . pisum in the food available, PI is its proportion in the diet. Each point is the average of a t least six observations. Some ladybirds had been trained to fabae, some to pi8Um and some to a third species (control). The lines are not different from the null hypothesis with no preference.

prey by touching them. The ladybirds could not be trained to prefer one or other prey, even when they were reared exclusively on it. When a switching experiment was done the ladybirds simply ate prey in direct proportion to their abundance and did not switch (Murdoch and Marks, 1973). These results are summarized in Fig. 24. The data provide a very good fit to the no-preference, no-switching model (r2 values 0-85to 0.97), and when sequences of meals were examined, it was shown that meals were a series of independent triala,

PREDATION AND POPULATION STABILITY

83

The ladybird results were obtained both in a very simple universe (a leaf in a petri dish) and on bean plants, on which they live in the field. We tried, unsuccessfully (Fig. 24), to train the ladybirds to each of the two prey and to detect whatever cues were correlated with the prey’s distribution on the plant, so our results probably are applicable to the field. The ladybird behaviour presumably is the simplest type that will be found when a predator is faced with two prey species. Fish-bluegill Reed (1969) gave small bluegills, taken from a local pond in Santa Barbara, a choice between mosquito larvae and midge larvae of the same size. The experiments were run in a 22-gallon aquarium which had black vertical slats to reduce the distance from which the fish could see the prey. Prey ratios were varied from 1:5 to 5:1, with a total of 60 prey in each treatment. Prey were not replaced, and in each trial between 10 (17%) and 27 (45%) of the prey were eaten (we exclude from the analysis one fish out of 30 which ate fewer than 10 prey). This proportion is larger than optimal but probably does not affect the conclusions. Each fish was run twice at a given density but we treat each run as a replicate observation. Reed ran some treatments with gravel and/or cloudy water, but here we discuss only the “clear water, no gravel” situation. The two prey behaved rather differently. The midge larvae spend most of their time on the bottom, while the mosquito larvae spend some time on the bottom but also sometimes moved to the surface to breathe. No measurements of these differences were made, or of where the fish caught their prey. There was, then, large but not complete overlap in the distribution of the prey. The fish hunted throughout the aquarium. The fish showed no preference for either prey when the prey were equally abundant (average percentage of midges in the diet was 49.6%). When individual diets are analyzed (Fig. 13) there is a clear central tendency, rather than a marked tendency for fish to have extreme diets. When the ratios were varied no switch occurred (Fig. 25), and the data fit the null model (Reed, 1969). Thus the bluegill provide another example that fits the prediction of weak preference at equality, low variability among predators and no switch. Figure 25 illustrates an interesting minor point. At both extreme diets the proportion of the rare species in the diet is higher than expected. The difference between observed and expected is significantly different only in the low midge treatment (x2 analysis). Similar, small, deviations were observed in some of the snail experiments, and there may be some biological

84

w. w.

MURDOCR

and

A. OATEN

significance in this, although the mechanism has not been studied in these cases. d. Summary of tests of switching. The predictions concerning switching, based on data from seashore snails (Murdoch, 1969), have been confirmed here for a wide range of predators. When average preference at equality is weak but also variable among predators, switching is found in protozoa, snails, fish and pigeons. In two other sets of data (from quail and fish) we were unable to measure the variability of the preference at equality though weak preference at equality was associated with

0

0.5

1.0

FI

FIG.26. Bluegill (fish) attacking midge and mosquito larvae fit the null model. F, is the proportion of midge larvae in the food available, P, is their proportion in the diet. Each point is the mean of 12 (in one case 10) observations; the bars represent & 2 S.E. The line is the null hypothesis with no preference. Data from Reed (1969).

switching. When individual predators have similar diets given equal prey abundances, whether their preference at equality is weak or strong, they do not switch but fit the null model, Strong preference at equality was found in snails and a protozoan; weak consistent preference at equality was found in fish and ladybirds. Clearly the occurrence of switching is unrelated to the predator’s phylogenetic position; switching has been found from protozoa to birds, and, even more striking, the same predator or similar predators either switch or fit the null model depending upon the circumstances. It is not yet clear just what determines that the preconditions for switching (weak variable preference at equality) will occur. Both the fish and the snail data suggest that environmental heterogeneity and its affect

PREDATION AND POPULATION STABILITY

85

on the spatial distribution of the prey increase the likelihood that switching will occur.

e. Other data. Three other sets of data are commonly referred to as providing evidence on relative attack rates as a function of relative prey abundances, or on “search image”. These are in papers by Tinbergen (1960), Mook et al. (1960) and Holling (1959a, 1965). We discuss these in detail below in Section V A2 on “Search Image”, but note here that their data are not in a form that allows us to analyze whether or not the predators switched. I n the first two papers there are no good data on the prey ratios available, though the ratio in the diet is given. In Holling’s work both the diets and food available are presented as the density of the prey species of interest, not relative densities. Finally, Royama (1970)has data from birds in the field, but we have again not been able to get the appropriate information from his tables. However, he does state that the birds’ diets do not seem to correspond much to the abundances of prey in the field. f. Eztrapolation to field studies. The consistent relation that has been found between weak, variable preferences at equality and switching is relevant to extending the study of switching to the field. It is probably difEcult to do switching studies in nature because ol” the difficulties in observing predators’ diets over long periods and in estimating prey densities as these vary. But it should be reasonably easy to determine preferences, or at least to determine whether or not diets vary greatly among predators that are faced with the same prey species. If diets are variable among predators then we may guess that they also vary through time in response to changes in the relative frequency of different prey species, and that switching may be common where diets are variable. Bryan and Larkin (1972) used the ingenious device of pumping out the stomach contents of individual trout to show that different trout had significantly different diets, even though they were all in a similar environment. They did the analysis on the same fish on up to six occasions over a period of more than a year and showed that, especially over short intervals of up to six months, prey types were eaten in proportions that were characteristic for each trout. Hobson (1968), in his analysis of the feeding behaviour of fish from the Gulf of California, noted that the guts of individual Microlepidotus (a schooling grunt) contained either exclusively midwater crustacea or exclusively benthic molluscs, even though all the fish fed at the same time of the day. Paraques (a croaker) as a population fed upon several species of crustacea, but the seven individual guts examined each contained only one prey species. D

86

w. w.

MURDOCH

and

A. OATEN

g . Another switching mechnism-$locking. The switching mechanisms examined above arise from the independent behaviour of individual predators. We suggest that the likelihood of switching will be enhanced in predators that are labile in their diets, as individuals, but whose diets are also influenced by those of other predators close to them. The snails possibly illustrated a simple mechanism for causing this. It appears that snails will feed upon a prey that has been killed and is being eaten by another snail (Connell, pers. comm.). I n Murdoch’s (1969) experiments, groups of snails often were used together and there was some evidence that snails in a group had significantly similar diets. This could have happened because snails became conditioned to the prey species being eaten by their neighbour when they joined them at the same trough, so to speak. We expect predators that have a social organization and, in particular, those that feed in flocks or schools, to show switching at the group level. An analogous situation might be the “band wagon” effect that operates before elections when people swing to the candidate who appears to be more popular. Consider a flock in which there are no dominant individuals, or “leaders”. A rather loose intuitive argument suggests that switching should occur. Suppose in a given area with two prey, species 1 constitutes 75% of the available food. When the flock arrives the first few meals taken by the group, which are independent events (i.e. the first few predators to feed), are expected to consist of 75% of species 1. Now, however, the average predator about to feed is influenced by three others feeding on species 1 for every one feeding on species 2. Thus, at any given moment, the probability that a predator will eat species I, given that it will eat something, is 0.76 plus some other fraction that will depend upon the precise assumptions that we make about how predators influence each other. The probability will then be greater than 0.75 that the average predator will choose species 1. As the relative prey abundance5 become reversed through time, there will be a lag aa the predators switch to the newly abundant prey, but switching should still occur.

2 . Search image One of the most seminal studies of predation is Tinbergen’s (1960) work on predation by birds, especially the Great Tit, Parus major, in pinewoods in The Netherlands. Tinbergen obtained estimates of the densities of different prey insects (mainly caterpillars) in the forest and compared this with the food brought by adult birds to their nestlings. He and his co-workers tried to examine the effect of factors such as size and crypsis on the degree of risk run by Werent prey species. He observed that frequently there was a sudden increase in the con-

PREDATION AND POPULATION STABILITY

87

tribution made by a species to a bird‘s diet soon after that species became abundant in the environment. The increase in the diet seemed to lag a few days behind the initial increase in density, it occurred in different birds at different times, and it appeared to be greater than one would expect simply from the observed increase in the prey’s density. After considering alternative explanations, he hypothesized that the abrupt change in behaviour was caused by the birds’ developing a “specific search image”. He thought that birds used cues specific to a prey when searching, that they overlooked a species when it k s t appeared at low densities, and that when they began to encounter a prey frequently enough, they abruptly “turned on” to it. Thus, “Specific searching images are adopted only when the species in question has exceeded a certain density . . . increase in risk seems to be restricted to a rather narrow density range . . .” (Tinbergen, 1960, p. 331, italics added). Adopting a specific searching image has been called “learning to see” (Dawkins, 1971), and this seems to be an apt description of what Tinbergen had in mind. Clearly it is dii5cult to get inside a bird’s head and discover whether such a process is going on, but at least for chickens there is evidence that birds can indeed learn to see a new prey (Dawkins, 197 1). Furthermore, regardless of the physiological explanation (and behaviourists have cast doubt on Tinbergen’s notions in this area), the consequences of such a hypothesis for predation rates can be examined and are of interest at the level of the population. In this section we explore in detail Tinbergen’s paper on search image in birds, and that of Mook et al. (1960). The reader not interested in this particular discussion may skip it, since it is not crucial to the rest of the paper. We consider it important, however, since these papers have been quite widely misinterpreted, perhaps owing to the relative obscurity of the journal in which they are published, but no doubt also because the data in Tinbergen’s paper are presented rather obscurely. I n addition, confusing the issue further, Tinbergen and Mook used different types of analysis to test for the existence of search image. There are two kinds of misinterpretation, the first concerning what search image means, and the second concerning what Tinbergen’s data demonstrate. Briefly, with respect to the first point, “search image” refers either t o an abrupt change in predation behaviour in response to an increase in the absolute density of the prey that is of interest (Tinbergen),and/or to a sudden change over time as a new prey appears (Tinbergen and Mook); “search image” specifically does not refer to a change in predation behaviour in response to a change in the relative density of one prey with respect to other prey. The search image hypothesis is not equivalent to switching or apostatic selection. With

w. w.

88

MURDOCH

and A. OATEN

respect to the second type of misinterpretation, Tinbergen’s data do not provide evidence for an S-shaped functional response. The rest of this section provides a detailed discussion of these and other aspects of the search image. First we discuss the concept as it relates to absolute and relative prey abundance. Tinbergen considered that the abundance of “other” prey species affected the birds’ diets in two ways. First, when alternative species are abundant they are more frequently encountered and therefore should contribute more to the diet, simply because they are encountered frequently. But that does not imply search image (or switching), and would arise from a randomencounter model. Second, he thought that the risk (probability of attack for a given individual) of any one species would vary with the density of other prey species (p. 297), but he believed this is simply owing to the fact that when several prey species are abundant, differences in risk between species will break down and the birds will be more indiscriminate (p. 303). Thus he is suggesting that when food is plentiful the various abundant species will be attacked at roughly the same rates. (Actually the opposite might be expected because predators could “afford” to be more choosy when food is plentiful.) I n fact, in his analysis of search image, Tinbergen tried to exclude the effect of scarcity or abundance of alternative prey by dividing the data into two series, one of which consists of observations made when the alternative prey were rare, the other when they were abundant. Within each series the alternative prey are treated as though their densities were constant. If search images indeed occur, one might ask if a bird can have multiple search images. We have not been able to find comments on this question in Tinbergen’s paper. Thus, if the development of a search image is a response to the prey’s exceeding a critical density, what happens if two or more prey species exceed their critical densities? It is not clear whether search images are then no longer used-and all the abundant prey are t a k e n - o r whether the bird has a search image for all of them. Tinbergen derived the algebraic form of the search image model explicitly from the Lotka-Volterra assumption of a linear increase in number of attacks with increasing prey density, i.e. N , = a,H,T. (In Tinbergen’s notation this is N, = R,D,t). This is an unsatisfactory basic assumption, which fortunately is essentially lost in the development of the model. Where the density of the species of interest is H I and the combined density of all other species is H,, we have

N,/N,

=

- and where P, is the percentage that species 1 forms of aao)

the actual diet, we obtain

PREDATION AND POPULATION STABILITY

P, =

a,H,

89

100

a1H1 +adlo

where a , is the average attack rate on all other prey species and a{ (Tinbergen’s Ri)is the probability that a given prey individual belonging to species i will be attacked per unit time; it is the attack rate of all the other models discussed here (see also Royama, 1971), and is Tinbergen’s “risk index”. Tinbergen then rewrites the equation as

-

Y, =

100

. . a&Il

1

so that the percentage of species 1 in the diet can be written as a function of its absolute density in the environment (H,). An “expectation” curve is then drawn, i.e. PIas a function of H,, by substituting a value for a, in the equation and assuming that a, and H , are constant over all values of H I . The null hypothesis, of course, is that a, is constant over all values of H,. The search image hypothesis, by contrast, predicts that a, is high at high values of H , and low at low values of HI. An example is shown in Fig. 26, taken from Mook et al. (1960), in which various expectation curves are drawn by assuming that a,, H , and a, have various constant values. The search image hypothesis predicts that observations will lie below the curve at low H,, and above the curve at high H,. Tinbergen’s own data provide a less than convincing test of the model, though they are widely quoted as illustrating search image. His results are beset by the usual problems of field data: some prey species were not sampled; many prey densities were measured very indirectly (by estimating frass) and this method gave inconsistent results (see Appendix by de Ruiter in Tinbergen’s paper); some estimates are based on very small numbers in the original counts; both dependent and independent variables have unmeasured (or undescribed) statistical variation. For only one prey species (Acuntholyda) were the results adequate for statistical analysis, and only in this one cme, out of six examined, was there moderately convincing evidence for the hypothesis (Tinbergen, 1960, Fig. 21). Also, for some unexplained reason, only some of the broods were included in the analysis. The analysis of diets of titmice from Tinbergen’s study area by Mook et a2. (1960) is more informative than Tinbergen’s, and probably more appropriate to his search image idea. The main prey was Bupalus, the Bordered White moth, Unfortunately these authors dso da n9t have

90

w. w.

MURDOCH

and

A. OATEN

good estimates of the density of alternative prey. They show b t that because of the enormous scatter in the data and the consequent diflCiculty of choosing a value for a, the risk index, it is very difficult to calculate the expected curves (i.e. the null hypothesis of no search image) in order to plot the percentage of Bupalua in the diet versus its density (Fig. 26). No clear conclusion can be drawn from that analysis. However, they then analyze changes in diet with respect to time,

HB

FIG 26. The percentage ( P B )that Bupalua forms of the diets of Great Tits versus its density (HE) in the environment. Each curve is a null hypothesis for Tinbergen’s “search image” model with constant values for a,, a~ and H, (see text). Reproduced with permission from Mook et al. (lQ60).

and here the evidence is somewhat clearer. They test for an increase in the “risk” of Bupalus with time, after its appearance in the habitat. They assume that H , and a,, the density and attack rate for all alternative prey, are constant. They show: 1. That there is sometimes a sudden jump in attack rate, generally a day or so after Bupalus has become moderately dense. 2. In six out of 16 birds, there is a statistically significant increase in Bupalus risk with time. This is caused by the sudden increase in risk. However, the estimated risk is highly variable from day to day, and

PREDATION AND POPULATION STABILITY

91

during the periods both before the jump and after the jump there is no clear relation between risk and density. Fortunately Mook provides his original results (see our Appendix 11) and we can therefore analyze the relationship between risk and density. In Appendix 11, taken from Mook’s Tables 111-VI, we have listed the density of Bupalus ( H B ) ,the number of prey brought to the nest (N) and the number of those that were Bupalus (B). From equation (28),

P B

lOOB

, is given by N a~H~l00 = ~BHB

the percentage of Bupalus, PB=

__

+

then so

a,H, -- - HB(100 - P B ) aB

P B

- HB(N- B) B Still, assuming that a, and H , were constant, Mook et al. investigated H B ( N - B ), and therefore of Ilae, against time. Asignificant the trend of B downward trend indicates that Bupalus was at a greater risk as time went on. As we noted above, six out of 16 birds showed such a trend. But in Tinbergen’s algebraic hypothesis, risk should increase with density ( H B )rather than with time. We have done regression analyses of Bupalus risk versus Bupalw density and found that risk increased with density in six cases out of 16. However, only three of these were the same birds that showed a positive trend against time. Another six birds showed a decrease in risk versus density, and the remainder showed no trend a t all. One of the birds that had demonstrated a positive trend against time showed a negative trend against density. (It should be borne in mind that the assumption of H, fixed is shaky, and that the estimates of H B were often based on very small numbers in samples.) In summary, Tinbergen’s data, analyzed against prey density, are moderately convincing for some broods for only one of the six prey species examined. Mook‘s data convincingly demonstrate a sudden increase in risk through time, for a minority of the predators, and for some birds the increase is also correlated with density. For the majority of the birds in both studies there is no evidence for search image turning

92

w. w.

NURDOCH

and

A. OATEN

on, either as a function of time or of prey density; indeed the bulk of the evidence suggests that risk is not related to fluctuations in absolute density of the prey in question. The data are not available that would let us examine switching, i.e. the influence of the density of other prey in relation to that of Bupalus. This analysis serves to illustrate the difficulty in giving a precise interpretation of Tinbergen’s model. Mook’s contention that the search image is a change in behaviour over time is quite consistent with part of Tinbergen’s verbal account of the idea. And so long as density increases with time during the period of observation, risk will show an overall increase against both time and density in those birds that show the shift in behaviour. But Bupalus density in fact often did not show a clear trend with time, except when it first appeared in samples. On the other hand, the algebraic version of Tinbergen’s model quite clearly implies that risk is a continuous variable, fluctuating in response to the prey’s absolute density. Tinbergen’s concept therefore remains ambiguous. The search image model presents other difficulties. It seems to fall between two stools in analyzing the percentage of species 1 in the diet versus its absolute density in the environment. I n the first place, Tinbergen was interested in actual attack rates, yet we cannot assume that the number of a prey species eaten increases linearly with its percentage in the diet unless the total number of prey items eaten of all prey species remains roughly constant. I n fact Mook‘s data show that the total number of prey in the diet was variable and very variable in some cases, and that in some birds the number of Bupalus in the diet showed only a loose correlation with its percentage in the diet. I n the second place, although these authors estimated the percentage of species 1 in the diet, we still do not get information on the effect of relative prey abundances on relative risk, since the independent variable they used is the density of only species 1. We turn now to the broader implications of this detailed treatment of Tinbergen’s and Mook’s papers. Tinbergen seems to have had two objectives. The first was to demonstrate a behavioural phenomenon, namely that birds “learn to see” a new prey species over a short period of time when its density increases above a certain threshold. This hypothesis can be examined by short-term experiments whose outcome will have mainly behavioural implications. To evaluate its populational consequences (Tinbergen’s second objective) the concept would need to be further refined: for example, does the search image “switch off” when the prey returns to a density below the threshold? i.e. what does the verbal hypothesis predict about the relation between risk and density when time is abstracted? We noted above that the existing

PREDATION AND POPULATION STABILITY

93

algebraic formulation of this relationship is unsatisfactory, and the evidence is generally against the hypothesis. What happens when there are several prey species above their threshold densities? Finally, the effect of relative prey abundances on risk is not taken into account in the hypothesis, and this is a serious defect for application to populational problems. Tinbergen’s second objective WM to test the idea that the risk run by a prey individual increased as the density of that prey increased. (He thought it also decreased at very high densities.) But the model is poorly framed to examine this hypothesis since the variable measured is the percentage of a given species in the diet, and not actual number eaten. I n order to examine the risk of a given prey species (or whether or not predation upon it is density-dependent), the framework proposed by Holling (1959a, b, 1965) in his analysis of the different types of functional response is more direct, more appropriate and much more useful. On the other hand, for studies of the consequences of changes in the relative prey abundances upon attack rates, we prefer the concept and algebraic forms for switching. We suggest, therefore, that Tinbergen’s formulation of the search image model, which continues to be referred to in the literature, often incorrectly, is not a particularly useful one. For population studies preferable alternatives exist. However, the model may prove useful in behavioural studies if it can be refined to take account of extra parameters such as the effects of the abundance of alternative prey species. Because the search image model is mathematically similar to the switching model, we clarify the relationship between them. Considering only two prey species, for notational convenience, it can be seen that the non-switching model is another form of expression (28), the null hypothesis of the search image model, provided we assume that c = al/az.Thus, the null case of no switching can be written as

though this is not the form in which it has been used. Now setting c = ul/a2,

a2

which is expression (28) divided by 100. Thus, the null hypothesis for search image and switching can be rearranged to be algebraicly identical, but the hypotheses themselves

94

w. w.

MURDOCH

and A. OATEN

are different. I n the search image hypothesis, prey risk for the main prey species (and therefore the relative risk that Merent prey run) is a function of the absolute density of the main prey. If the density of the main prey remains constant, while that of the alternative prey varies, the composition of the diet will change only because relative encounter rates will change; but the relative risks will remain constant. By contrast, the switching hypothesis predicts that if the main prey density remains fixed and the alternative prey density changes, the composition of the diet will change because (1)relative encounter rates change and (2) relative risks (preferences) also change. We now summarize the clarificatory points made about search image, both in this section and in Section V A1. 1. A sigmoid relationship between P, and H, in the observed data does not imply density-dependent mortality upon species 1 during the accelerating part of the curve. This is not a graph of functional response and no conclusions can be drawn about functional response unless information about absolute predation rates is provided. We can be certain mortality is density-dependent in these circumstances only if the total number of all species eaten is constant. The bird data show that the number eaten is highly variable. Holling’s type of analysis is preferable for this purpose. 2. Data lying appropriately above and below expectation curves do not substantiate the search image hypothesis unless a, and H, are constant as EI, vanes, or unless the curves are calculated with a. and H o varying as discovered by observation. 3. A sigmoid curve of P, versus H, does not illustrate switching or apostatic selection unless total prey density (H, + H,) is fixed. We would need to recalculate the relationship between P, and the percentage of species 1in the food available, to analyze Tinbergen’s data for switching or apostatic selection. 4. Tinbergen’s verbal account of the model, and subsequent analyses by Mook, actually suggest that search image is a change in time, as well as, or instead of, with prey density, though time is extracted from the model. To test the verbal account of the idea, it is really necessary to look at risk through time. Recent studies on behaviour relating to the search image idea and not reviewed here are discussed by Krebs (1973). Holling’s field data on predation upon sawflies (Holling, 1969a) has been interpreted as illustrating search image, but in the absence of data on the abundance of alternative prey species, these results unfortunately cannot be used to examine the search image hypothesis. In

95

PREDATION AND POPULATION STABILITY

his laboratory studies (Holling, 1959a, 1965)he gave Peromyscw varying densities of sawfly cocoons and kept the density of alternative food (dog biscuits or sunflower seeds) fixed. The total number of items eaten by one mouse was roughly constant (Holling, 1965, Fig. 4.2), so the sigmoid curve of number of sawflies eaten probably confirms Tinbergen's algebraic model: the percentage sawflies in the diet is sigmoid with respect to sawfly density. Another experiment (Holling, 1959a, Fig. 6) also seems to fit the model, a t least where the mouse was given a highly palatable alternative food (sunflower seeds).

B. F U N U T I O N A L 1. Experimental results

RESPONSE-TWO-PREY

SPECIES

Information about relative attack rates does not lead directly to general conclusions about the number of each species killed as a function of its absolute density. To make this transformation we need to build models of switching that incorporate assumptions about absolute attack rates, which we do in the next sub-section (V B2). Alternatively we can do a wide range of factorial experiments in which both absolute and relative densities are varied, though it would be very time-consuming to look a t a broad range of situations. However, there are some data both from the field and the laboratory, which give us a look a t a limited set of circumstances (Table 111). TABLE TI1 Fwctional reaponae when mare than one prey apeciea .is available Functional response Predator

Acanthina (snail) Thab (snail) sk?&&(PrOtOZO8)

Stentor Cocciwlla (ladybird)

Quppies (fish) Qreat Tit Deermice (infield) Deermice (in lab)

type

Switch

Mussels and barnacles Mussels Euglena Chlamydamonaa Euglena Tetrahymena Aphids

3 2 3

yes no yes

Murdoch (1969) Murdoch (1969) Rapport (1974)

3 1

no no

Various invertebrates Midge larvae Mosquito larvae Droeophila Tubi6cida Various insects S a d i e s and other prey Sawflies and biscuits or sunflower seeds

3 1

no

Repport (1974) Murdoch and Marks (1973) Ivlev (1961) Reed (1969)

Prey ~

Source

~~

3

?

2, 3

yes P

3P

?

3

?

This paper Mook et al. (1960) Holling (1959a) Holling ( 1 959a, 1965)

96

w. w.

MURDOCR

and

A. OATEN

The laboratory studies discussed earlier (Section V A), in which relative attack rates are examined when the densities of two prey species are varied, provide some limited information concerning functional response. I n the snails (Murdoch, 1969), mortality was type 2 when there was strong preference, no switching and the density of the alternative prey was held constant. When preference was weak, total prey density was held constant and switching occurred, percentage mortality on both prey was increasing over the whole density range. Thus the conditioning mechanism led to a response that was essentially type 3, under these specific conditions. The Protozoan Stentor, it will be remembered (Section IV), has a type 3 response to its prey when only one prey species is available. A similar slight acceleration can be observed at the lower prey densities when an alternative species is available a t a fixed density. This is true both when switching does and does not occur (Section V A). Our interpretation is that the S-shaped response in the two-prey species system is caused, not by the interaction between the prey species, but by the same mechanism that operates when only one prey species is present. This may be a response to extra metabolites in the medium causing the feeding rate to increase over the lower range of densities. I n the case where switching occurred it may have enhanced the effect. Neither Reed’s fish (Reed, 1969) nor the ladybirds switched. I n the fish the functional response was type 2 to both prey species. I n the ladybirds mortality was density-independent upon both prey species (type 12). I n the guppies it is again difficult to extrapolate to the field because total density was fixed at all prey ratios. However, in this case we have additional information that facilitates the interpretation: we measured the amount of time the fish spent hunting, and therefore have an estimate of absolute attack rate. From the point of view of the stability of each prey species, the important variable is the attack rate per unit time during the experiment. Figure 27 shows that for both prey species the number killed per unit of time accelerated at least over the lower densities presented. (The curves must pass through the origin, which would accentuate the density-dependent effect.) There is also a small amount of evidence from field studies. Tinbergen’s (1960) and Mook et aZ.’s (1960) papers have been discussed in some detail above (Section V A2). We noted that Tinbergen’s results were not presented as “number eaten” versus density. However, we have re-analyzed Mook’s results from Bupalus in terms of numbers eaten in an area versus density available there, which varied in time (Appendix 11). There is a great deal of scatter in the data, which makes interpretation difficult. I n addition, there is no very good and consistent way

PREDATION AND POPULATION STABILITY

97

of analyzing the data from all birds, since some birds show no evidence of “satiation” (i.e. number eaten does not level off), and other birds even show no increase in number eaten versus density. I n the absence of a sound statistical method a good deal is left to the curve-fitting imagination. We fitted the data with linear regression analysis to help us in this respect. Our conclusions are that, of 16 birds, four showed clear type 3 (initially accelerating) responses, and another four probably had type 3. Four showed type 2. Two showed a flat response, and since the curve must pass through the origin, these are also best designated type 2. One set seems best fitted by a straight line through the origin (type 12) and one showed a negative correlation between number eaten and Bupalua density.

Prey densily

FIQ.27. The number of prey eaten per minute (attack rate) by guppies verms prey density; tubificids open circles, Drosophila closed circles.

Thus Mook’s data provide moderately convincing evidence for a type 3 response in some birds in the field preying on several prey species, and mainly type 2 for the remainder. We also analyzed the proportion killed versus density (proportion here is B / H ) . No birds showed a significant positive relationship between percentage kill and Bupalus density, six showed significant inverse (negative) densitydependence and ten showed no relationship. We can summarize the major results of Mook’s study as follows. Some birds showed a type 3 response to increasing Bupalus density, so that the mortality was potentially stabilizing over the low and intermediate range of Bupalus density. In order to know if the mortality actually caused Bupalus numbers to stabilize we would need to know much more about the system, including the relation between the amount of mortality and the rate of increase of Bupalus in its absence (Section I11 F). Some of the type 3 curves arose because of a sudden change in behaviour

98

w. w.

MURDOCH

and

A. OATEN

with time; but it is not clear whether the birds were changing behaviour in response to the absolute density of Bupalus or to its relative abundance (Section V Al). This s6udy is the only example we have found of a predator in the field showing a type 3 function in response to changes in prey density through time. Half of the birds showed type 2 or other responses. Such variability among individual birds was also obvious in all other aspects of their predatory behaviour that Mook and ourselves have analyzed. It is also typical of other predators, especiallythose that show switching behaviour (Section V A). While this variability makes analysis and generalization more dif6cult, it is to be expected in predators that show the kind of labile behaviour that leads to switching. A more recent study of predation by the Great Tit (Royama, 1970) failed to show the density-dependent relationships noted above. Royama concluded that there seemed to be no clear relationship between the abundance of prey and their abundance in the birds’ diets. Finally, we must mention the field and laboratory observations made by Holling. Holling (1959a)suggested that deermice (Peromyscus)in pine plantations had a type 3 functional response to the density of sawfly pupae. Each point in the graphs came from different areas, not from the same area at M e r e n t times. There is a problem here of deciding whether or not the data are actually sigmoid (his Fig. 1, 1959a), and evaluating this is even more difficult when the data came from the field. For example, it is notoriously difficult to get good estimates of the numbers of small mammals, and the techniques used (Sherman and snap traps) were not especially free from error. It may be better to suspend judgement on these data. I n his laboratory experiments Holling (1959a, 1965) presented a deermouse with different densities of sawfly cocoons, buried in sand. Alternative prey consisted of a constant (excess)amount of dog biscuits or sunflower seeds. The curves were more clearly type 3, though in some cases (1959a, Fig. 2) the acceleration was very slight and then the curves probably were not significantly non-linear. These results from birds and small mammals are the only field evidence we have found concerning the form of the functional response in general predators. This is altogether too meagre a basis for making generalizations, though the bird data in particular show some clear examples of both type 2 and type 3 responses. No evidence is available about invertebrate general predators in the field.

2. Hodels We can take at least two approaches in going from information about switching, or relative attack rates, to models of the absolute attack

PREDATION AND POPULATION STABILITY

99

rate on each prey species. The first is t o modify existing models of attack rates to include two prey species and then to insert switching or some other assumption about relative attack rates. This was discussedbrieflybyMurdoch(1973).The second is to startwith adescription of a specific switching mechanism and to build from this a model of attack rates on each of two or more species of prey. This is done by Oaten and Murdoch (1974a’ b). We prefer the second approach, but briefly discuss each in turn. Murdoch (1973) generalized the Holling disc equation for one prey species, (32) to the case for k prey species. This is

N5

=

where Nj is the number eaten from the 5th species in T time units. At is the attack rate on the ith species, vt is the handling time for the ith species and H is prey density. For the case with only two prey species the model is

AaTHa 1 + Ai~iHi+ AaTaHa (This generalization had been done independently by several authors.) F’rom our point of view the model has the advantage that dividing one equation in (34) by the other gives the null case of no switching: N, =

where Ai/Aa = C. The model fits one’s intuitive notion of how a predator should operate if it is attacking each species independently of the other and at a rate proportional to the abundance of the species (except that hunger is omitted). The only way that one species interferes with predation on the other species is by requiring the predator to spend time handling it. Notice that relative handling time does not figure in the “preference” parameter, c. Unfortunately, to explore the conse-

100

w. w. MURDOUH and

A. OATEN

quences of this model we now have to deal with four parameters that can vary, in addition to the fact that two densities can vary. If H , and H , are varying independently, there is no way of predicting a general relationship between Ne and Hi, which can be done if H , + H , is fixed equal to K , or if the density of one species is held constant. When total density is fixed, the first derivative of expression (34) for N, as a function of H , is always positive. The second derivative is negative if Alql > h2r),, and is positive if hlvl < A,r),. Thus, for simplicity letting the handling times r), and r), be equal, the mortality is densitydependent (second derivative positive) upon the less preferred prey and inversely density-dependent on the preferred prey. The above, however, is a very special case. The more general case can be examined by differentiating expression (34)when one of the prey species, say species 2, has a fixed density. I n that case the first derivative (dN,/dH,) is always positive but the second derivative is always negative, so the attack rate on the first species is inversely density-dependent, i.e. a type 2 functional response on species 1 for El, fixed. We can then calculate such functional responses for species 1, with species 2 fixed, for various densities of species 2, and obtain an “attack surface” for species 1 by graphing these responses in three dimensions with axes H,, H , and N,. Any cross-section of this surface taken parallel to the H , axis will be concave down. However, if H , changes as H , changes, the relevant cross-section is not parallel to the HI axis, but follows the path traced out by the point (B,, H,). As we go out along the H , axis, the height of the surface decreases; so if species 2 decreases rapidly as species 1 increases initially, so that the cross-section runs in the direction of increasing H , but rapidly decreasing H,, the attack rate on species 1 will be held down at low densities of species 1 and a type 3 response to species 1 can result. I n summary, the generalized disc equation usually gives type 2 responses to both prey, but under some circumstances can give type 3 or density-dependent responses to one of the species. The model can then be modified to incorporate switching. Its analysis then becomes more complex. I n addition we must add to the assumptions already embodied in the disc equation additional assumptions about the relationship between preference and the relative abundance of the two prey, and about the effect of preference on the attack rates. For these reasons we prefer the second approach discussed above, and simply indicate briefly here how one might proceed in general with the disc equation model. The simplest way to incorporate switching into the above model is to let At increase linearly with the proportion of species i available, although this is rather a strong assumption. Let A{ = a$‘(, where

PREDATION AND POPULATION STABILITY

101

P( = Ht and, where to maintain the original average attack rate, Hl + H ,

act is twice the original A(, since Ft varies from 0 to 1 and on the average

is 0.5. Then the model becomes

Notice now that

which is the function graphed in Pig. 11, and also yields Eqn (27). I n the special case where the total density is fixed and v1 = vz, the functional response to both prey is type 3. Where the density of one species is fixed and vl = v2 the response to the other species is type 3. But if a three-dimensional graphical analysis were done as above, the attack surfaces obtained would be complex, making generalization difficult. It seems, though, that switching usually yields type 3 curves to at least one species (often the preferred species), but sometimes at least one prey species receives destabilizing mortality. I n the second approach we begin with a switching mechanism and explore the consequent functional response for two species. We present a rather simple model of switching, in which the switching is caused entirely by variable preference which, in turn, is taken to depend only on the most recent meal. We give here only the assumptions and conclusions; the derivation is somewhat lengthy and is given in a separate paper (Oaten and Murdoch, 1974a, b). Suppose a single predator feeds off two prey species, species A whose population (to reduce the number of symbols) we take to be A , and species B whose population is B. We assume that A and B do not change appreciably over the time in which functional response is considered. We also assume the two prey species are randomly and independently distributed in a single area (i.e. no patches) and that the predator searches randomly (as in Section IV) and at a constant speed. When the predator encounters a prey, however, he will not necessarily attack it. Whether he does will depend on such things as his recognition of the prey as food, his taste for this particular prey, his hunger and his assessment of his ability to make a successful attack. We assume that all these factors, except hunger, depend only on the last meal: that he is more likely to attack an encountered prey if his last meal was of the same species than if it was not. It is possible to allow for hunger too, but this leads to difficulties in achieving an explicit solution so we do not comider it here.

102

w. w.

MURDOCH

and

A. OATEN

I n developing this model, we have thought particularly of the snail, Acanthina, feeding on mussels and barnacles (see Section V A). The snail seems to move about aimlessly, encountering prey randomly. It inspects encountered prey-perhaps to “decide” whether it is possible and worth while to drill through the shell-and frequently decides not to attack. Thus it appears to fit our assumptions fairly well. It is obvious that quite different models are needed for predators, like the guppies (Section V A), whose prey live in different places. We have not yet completed work on these models. Turning to the details of the model, we let V A and V B be the times taken to handle prey from species A and B respectively. If h is the proportion of the total habitat that the predator searches per unit time, then the probability he will search for more than t time units without encountering any prey is P ( T > t ) = exp { - h(A+ B)t). What he will do with an encountered prey depends on what it is, what the last meal was, and how long ago the last meal was. However, for simplicity, we w u m e that the last of these factors, which is essentially hunger, is negligible, and that the probabilities are constants, independent of the time since the last meal. They are perhaps most easily given in a table: Species of encountered prey:

A

B

Species of last meal: A B

a A

PA

aB

PB

When switching is occurring, a A > aB and PB > PA. As in the development of our model for functional response for a single, patchily distributed, prey species, and our general discussion of this situation, we make use of the following result from renewal theory: suppose events occur at random times such that the lengths of time between successive events are independent and identically distributed with mean p ; then the expected number of occurrences of these events in time T is approximately TIP, i.e. the rate at which the events occur is l/pper unit time (see, for example, Cox, 1970, Chapters 4 and 9). Our derivation of the functional response for this model takes these events to be the capture of prey of species A; we obtain the mean time between these captures, and take the reciprocal to obtain the functional response (rate of capture of prey of species A) when there are A members of species A and B of species B:

P(A,B ) =

103 I n considering whether this functional response is stabilizing at ( A , B ) we will, as in Section IV, adopt the criterion: PREDATION AND POPULATION STABILITY

3F

1

F is stabilizing at ( A , B ) if - ( A , B ) > - P ( A , B ) . aA A That is, F is stabilizing at ( A , B ) if it is stabilizing for changes in A with B held constant. Applying this criterion we have, after differentiating and some cancelling and rearranging, that F is stabilizing if PA(aA- aB)B- XaA2aBvAA2- ~ ~ A ~ B P A ~ A A B

+XPA{( aA P B - a B P A ) v B -

aBPA7A}B2

> 0 (37)

Although (37) is rather involved, we can draw some reasonable conclusions from it. First, it is clear that, with B fixed, (37) cannot hold for very large A . This is clear, also, from (36), since P ( A , B ) - + ~ A - ~ as A + m , as one might expect (this would be the case when the predator is doing nothing but eating A , taking time Y A for each meal). Next, if we take A to be small, so that X A is well below 1, the two negative terms are essentially zero, so stability depends on the first and last terms. If a A < a B and /?B < P A (i.e. if switching goes in the “wrong” direction, so the less abundant prey is preferred), stability is not possible, since both terms are negative. If a A < a B and P B > P A sufficiently so that O(APB> a B P A

( + 3, 1

-

then there will be stability in A only for large

values of B (i.e. when XB is large). If

a A > a B but a A P B < a B P A

( +3, 1

then stability in A occurs only for small values of B. However it is for small values of A and B that our assumption, that hunger does not affect the probability of the predator attacking encountered prey, is weakest: in this situation, when there is a long time between meals, the predator seems less likely to discriminate on the basis of his last meal, so a A is likely to be close to ag, and PB to PA. Finally, if both a A>a B

3

and a A P B > a B P A 1 + - , then stability results (for small A )

at all levels of B.

(

I n summary then, our results and criterion suggest that, in a situation in which the tendency for stability depends on the predator being more likely to eat a prey if his previous meal was of the same species, stability will more likely occur if (i)the effect that the previous meal has is large, so that the differences between a A and ag and between PB and P A are great; (ii) the handling time for species A (the species whose stability

104

w. w.

MWRDOCH

and

A. OATEN

we are concerned with) is small compared to that of species B (of course this means species B is less likely to be stabilized); (iii) (if (ii) does not apply) the handling times are small (they are not for snails) and h is small. It is not unreasonable to expect h to be quite small, since we are concerned here not with patches but with all prey in the habitat which (e.g. in the case of the snails) might be much larger than the predator could cover in a lifetime.

c. S U M M A R Y I n Section V we have sumeyed experimental results from a wide range of predator-prey systems and confirmed the following relationships between the occurrence. of switching and the preference shown by predators when given equal densities of the two prey species (preference at equality): fist, predators that show weak average preferences at equality but whose preferences vary greatly among individuals subsequently switch when presented with unequal prey ratios. When preference at equality is consistent among predators, and is either weak or strong, no switch occurs and the data fit the null model of no switching. These relationships may make it easier to discover switching in field situations. Three different mechanisms of switching have been suggested: a variable rejection rate, a variable “ignoring” rate, and an ability to evaluate varying reward rates in different sub-habitats. We suggest that flocking or schooling may also promote switching. Tinbergen’s (1960) idea of search image, and Tinbergen’s and Mook et al.’s (1960) data on this idea are analyzed. We conclude that the concept is ambiguous, the data are not very convincing, and that other approaches to the analysis of predation are more useful. We present two methods of modelling the functional responses that arise from a situation with two prey species, and especially from switching. Generalizing Holling’s disc equation (Holling, 1959b) and inserting switching into it does yield some results, but the model is rather awkward and encumbered by assumptions. Type 3 functional responses sometimes, but not always, result from this model. If no switching is incorporated the resulting functional response is usually type 2, but is not always, and sometimes is type 3. We looked at the alternative approach of modelling directly a switching mechanism, having results from seashore snails in mind. This yields a rather complicated model which predicts that type 3 responses are most likely to occur when the predator’s tendency to eat a contacted prey is greatly increased if its last meal was of the same species, and when the

PREDATION AND POPULATION STABILITY

105

species under consideration is not abundant and the alternative prey is.

VI. LEARNINGA N D FUNOTIONAL RESPONSBI A rather common generalization about predation, in the recent literature, is that predators that can learn have an S-shaped functional response (type 3) while “non-learning” predators have a type 2 (or type 1) response. A frequent extension is that vertebrate (or intelligent) predators can learn and invertebrate predators cannot. This latter generalization may have arisen from a misreading of Holling’s work (1965, 1966), where in fact he stated that probably some invertebrates could have a type 3 response. It should be clear from the earlier material presented in this paper that the major generalization-learning predators have a type 3 response, non-learning predators a type 2 or type 1 response-is at best too oversimplified. Variables in the system other than the predator’s ability to learn determine the form of the functional response, and we illustrate this briefly in the next few paragraphs. (In this context, it seems to have escaped general notice that in the paper that stimulated the generalization, Holling’s invertebrate predator was given only one species of prey, while the vertebrate predators were simultaneously given two types of prey (cocoonsand dog biscuits)). In addition, the issue is further confused by the fact that “learning” seems to be ill-dehed, the problem we address second in this section. The earlier sections in this paper show that the predator’s phylogeny or “intelligence” is a poor predictor of the type of functional response: some protozoa have a type 3 response; some birds have a type 2 response (Tables I and 111).I n fact, the form of the response is determined by the behavioural options open to the predator and by the characteristics of the prey to which it is exposed-the number of prey species, whether or not they occur in the same small area, the way in which their numbers vary, etc. I n particular, type 3 responses can arise in at least the following ways: (i)One prey species present 1. The predator is stimulated to spend more time hunting when the prey are more abundant, perhaps in response to increased prey odour or metabolites. Probably this mechanism is illustrated by the work of Burnett (1964), Takahashi (1968) and Rapport (1974). I n all three caaes the predators are non-vertebrate. 2. The prey occur in patches of different densities and the predators

106

w. w.

MURDOCH

and

A. OATEN

feed more often in patches having greater densities of prey (see Section IV). Presumably such behaviour is widespread among predators.

(ii) Two or more prey species present The most likely cause of a type 3 response when two prey are present is that the predators switch, although a type 3 response is by no means an inevitable consequence of switching. Switching can be caused by several mechanisms (Section V): the predator may change its preference towards the abundant prey as it eats it more frequently, by chance; it may largely ignore the rare prey; or it may concentrate its searching in the most rewarding areas. The absolute densities of the prey are important, and indeed in some peculiar situations type 3 may arise without switching because the prey abundance8 vary in just the right way (Section V B2, and Murdoch, 1973). No doubt other mechanisms can also lead to type 3 responses. For example, the presence of refuges for the prey can produce a type 3 response. If prey individuals differ in their vulnerability to the predator and the relative frequency of different prey types varies with density, this could lead to type 3. If the average prey becomes more exposed because crowding increases with density this could lead to type 3. The second issue, namely the consequences of learning, is an awkward one. The problem is that “learning” seems to have a very broad meaning for behaviourists, but probably ecologists use it more narrowly and, indeed, its meaning may vary among ecologists. Its meaning for behaviourists may be: any change in behaviour caused by experience. But in the predation literature it is often used to imply that the predator has learned how to do something (e.g. acquired a perceptive ability or skill for catching or handling prey). Tinbergen (1960) thought that, when the frequency of accidental contacts with a prey species exceeded a threshold, the bird “learned to see” the prey, i.e. it learned to associate the characteristics of the prey with the fact that it was food and therefore was able more frequently to distinguish it from its background. Holling (1966) demonstrated that small mammals could learn where to hunt for sawfly cocoons in sand when they had previously never been exposed either to cocoons or to food buried in sand. We showed above that other variables than this “ability to learn how to do something” determine the form of the functional response. I n the remainder of this section we want to show fist that this sort of learning will usually lead to a type 2 response when only one prey species is available. Secondly, we discuss a study that shows a type 3 response in a predator feeding upon two prey species, where that sort of learning does not seem to be the cause. Reed (1989) gave bluegill fish varying densities of a single species of

PREDATION AND POPULATION STABILITY

107

mosquito larvae in a 22-gallon aquarium (see Section V A). The fish were naive to the experimental situation and probably also to the prey. Prey were replaced as they were eaten to keep the density constant. The number of larvae eaten in 30 min is shown in Fig. 28. This seems clearly to be a type 3 response. Learning indeed caused this curve, and this would seem to run counter to our argument. We therefore pursue the matter further. Reed watched the h h and timed meals. The fish behaved aa follows. At the start of the experiment, each fish swam around the aquarium

Prey density

FIG.28. The number of mosquito larvae eaten in 30 minutes by bluegill (fish) shows a sigmoid relation with prey density. Each point is the mean based on 10 fish; the bars represent 2 S.E.Reproduced with permission from Reed (1969).

attacking any small object-pieces of dirt, excreta and so on. Sooner or later it would meet and attack a larva, and aa soon as it had done so, successfully, the fish's behaviour changed; search rate increased and the fish vigorously attacked larvae from a distance of several om aa soon as the larvae moved (occasionally two meals were needed to trigger this change). The fish had clearly learned to recognize the larvae as prey. The curve in Fig. 28 accelerates with increasing density because, aa density increased, by chance the fish found its first larva earlier (Fig. 29); thereafter, it hunted faster, and it then hunted speciJically for mosquito larvae for a longer fraction of the half-hour. Some fish at the

w. w.

108

MURDOCH

and

A. OATEN

lowest prey densities never found a larva, and therefore never learned, and others learned very late. Thus, the curve is accelerating because the length of time each fish was in the “learned” state was a function of prey density. Now generally, the length of time in which there are unlearned fish must be very short. (In this experiment almost all fish had become learned within 30 min.) I n nature, a predator that feeds on only one prey species either learns very quickly or dies. The obvious prediction is that when all predator individuals are “learned”, a type 2 response will result. 26 *

24 -22 -

18 -

20

tb

0

FIG.29. The time taken by bluegill to eat the first meal (closed circles) and the first two meals (open circles) at different densities of mosquito larvae. Each point is the mean of 10 fish; the bars represent f 1 S.E. Data from Reed (1969).

This prediction was tested by Landenberger (1973), who showed that this is precisely what occurred in starfish feeding on a turban snail (Tegula) in the laboratory. The starfish were na’ive to these prey in the experimental set-up. Each predator was kept a t a fixed density of prey. The number of prey eaten was recorded every day over a period of three months. The predators started out with a type 3 response, but during the experiment they all “learned” and then the response became type 2 (Fig. 30). It is not clear what sort of curve will result in predators that feed almost exclusively on only one prey species at any one time, but also

109

PREDATION AND POPULATION STABILITY

move from one prey species to another as the season progresses. The result will depend on how the relative abundances of the prey change, whether the predator in fact “forgets” a prey species as it becomes rare, and how quickly it does so. I n any case, for information on the transitional periods we would need to do experiments with both prey present and would not necessarily be able to predict the outcome from experiments using only one prey species. The second situation we want to examine is Holling’s laboratory study of mouse predation upon sawfly cocoons and biscuits, referred to above (Section V). A type 3 response to cocoon density was obtained.

0

40 C

5

30

L

0

n

5 20

z

10

2

4

S

16

I .

I

32“

6d

Prey density

FIQ. 30. The number of turban snails (Tegula) eaten by starfish in 10 days at different prey densities. The dotted sigmoid curve is drawn through data from weeks 2 and 3 in the experiment (open circles); the solid curve is drawn through data from weeks 10 to 12 (closed circles). Each point is a single starfish. Reproduced with permission from Landenberger (1973).

Holling was able to show in a different experiment that the mice learned how to find the cocoons: the feeding rate of the mice increased over a four-day period following their first exposure to the situation; after this four-day period the feeding rate remained at a steady maximum for the the remainder of that experiment. Holling then showed (1965, Figs 4 and 5 ) that another predator (a shrew) that had learned to look for the cocoons gradually hunted less and less for them when the shrew was in a container with no cocoons. The hunting rate declined approximately exponentially: about 50% in the first 30 h, 75% in the first 60 h and so on. If mice and shrews are similar we can conclude that they can learn to hunt in specific ways and can also either “forget” or learn not to hunt.

110

w. w. MURDOCH and

A. OATEN

However, in spite of the above evidence, it seems doubtful that this sort of learning and forgetting behaviour explains the type 3 responses that were obtained. Thus, in the functional response experiment (Holling, 1966) the mouse had already been exposed to cocoons in the experimental enclosure for 20 days before the experiment started, i.e. it was in a “learned” state at the start of the experiment and was by then eating a constant maximum number of cocoons. The experiment then ran for only eight h at each cocoon density (biscuits were supplied in excess), which is not enough time to allow for a significant amount of the “forgetting” demonstrated above. I n addition, within each eight h period the sawfly-hunting behaviour was reinforced by sawfly meals; even a t the lowest s a d y density, between three and 13 sawflies were taken by the mouse (Holling, 1966, Table 4.1). We suggest the following alternative explanation for Holling’s Sshaped response: the mouse tended t o concentrate on that feeding pattern that was more rewarding. Since cocoons were preferred to dog biscuits, the reinforcement from a cocoon meal should be greater than that for a biscuit meal, and so one would expect searching for cocoons to persist even when cocoons were rare. But when cocoons were sui3lciently rare the mouse switched to biscuits. This explanation is the same as that proposed for switching in the guppies (Section

V).

There is a problem here in that “learning” is so ill-defined that “doing the activity that is rewarded more” might be construed as learning. However, the mechanism seems t o be clearly different from (a) the guppies learning to recognize food, or ( b ) the mice learning that by digging in sawdust they could find prey, and it seems preferable not to gloss over such differences. Also, we expect that such behaviour of playing the best bet, where it is not prevented by other constraints, must be universal in animals. Finally, we describe a study in which a predator learned to search for its prey in a novel situation, and again no type 3 response occurred. Taylor (1972) has studied in detail the effects of learning on a hostspecific parasitic wasp, Nemeritus. He demonstrated experimentally that Nemeritwr, in the laboratory, learns to search for its host (larvae of the flour moth Anagasta) in a novel situation in which the host larvae are placed in a petri dish covered by bolting silk. The dish was kept in a large plexiglass container. Wasps that had not learned searched the entire surface of the box, rarely palpated the surface with their antennae and rarely tried to probe with the ovipositor. Eventually each wasp appeared to get some stimulus from the larvae and searched the cloth more intensively. As soon as it had had a successful probe (piercing a larva) it changed to a characteristic hunting behaviour: the

111

PREDATION AND POPULATION STABILITY

wasp then spent almost half of its time on the cloth, palpated the cloth much more frequently, and a large proportion of palpations led to probes and hits (Table IV). Naive wasps were exposed to different densities of hosts for six h. The cumulative number of prey attacked per hour is shown in Fig. 31. Since learning takes some time, and might be expected to occur sooner at higher than at lower densities (as in Reed’s fish), it is somewhat surprising that the curves from early in the experiment are not accelerating. Instead, the data from the h t four h axe statistically indistinguishable from straight lines passing through the origin. It may be recalled that TABLEIV Cmparieon of the behuvbur of hunting and rwn-hunting waepe, derived from obeemation 8esaiOne. A non-hwnting waap Tar& probes or hi&. From Taylor (1972)

1. Proportion of time spent on bolting cloth 2. Frequency with which the wasps walk over a larva while on the bolting cloth 3. Proportion of the times a wasp passes over a larva leading to palpation of bolting cloth with antennae 4. Proportion of palpations leading to probing 5. Proportion of bursts of probing leading to hits

Non-hunting wasps

Hunting wasps

0.03

0.47

162F

176/h

0.13

0.7 1

-

0.7 1

-

0.23

Takahashi (1968) found accelerating curves in this wasp. In Fig. 31 there is a slight suggestion that the slope for the first h increases, and perhaps this would have occurred to a si&cant extent at higher prey densities. In any case, these results are a further demonstration both that learning can occur in a wide range of predators and that its consequence is to produce type 2 (or type 1, which is an extreme case of type 2) functional response curves in predators attacking a single species of prey. Taylor (1972) was able to show that a stochastio model in which the predator learns (a)to find the host’s habitat and (b) to find hosts within that habitat, is in good agreement with the experimental data.

w. w.

112

MURDOCR

and

A. OATEN

50 I

Prey density

FIG.31. The cumulative number of ovipositions made in 1 to 6 h by the parasitic wasp Nemeritw, given different densities of flour moth larvae. Reproduced with permission from Taylor (1972).

V I I . OTHER R E S P O N S E SB Y PREDATORS I n this section we touch briefly upon the other components of predation that combine with functional response to give the predator population’s total response to changes in the density of its prey. We do this for completeness, to set the functional response in perspective, and because some of the concepts need to be clearer. Solomon (1949) noted that the predator population could respond either by changing in numbers (numerical response) or by the average individual’s eating more or fewer prey per unit time (functional response). Some confusion exists concerning the numerical response. Predator numbers in any area can respond to prey density via the birth or death rates or by movements in and out of the area. Taking movements first, it is obvious that one can include these either in the

PREDATION AND POPULATION STABILITY

113

numerical response (Buckner and Turnock, 1966) or as the “aggregative” response to prey patchiness-as we did in Section IV. Hassell (1966) suggested the name “aggregative response” and this response can be easily subsumed under the functional response. The latter course seems better in general, but if the movements are over a large distance, for example birds arriving from a habitat several miles away, it might be better to include this movement in the numerical response. The problem is partly a question of scale, and the decision as to where to include movements is to some extent arbitrary, depending upon how large an area one thinks is covered by the prey and predator populations being studied. In our discussion here we categorize all movements as the aggregative response and assume that they therefore are included in the overall functional response (summed over all patches of prey) as we discussed in Section IV. Murdoch (1971) noted that predators (but not parasites) have an additional response, namely the developmental response. The predator’s growth rate and rate of development are geared to prey density, and in turn the predator’s size-distribution (for example) affects its attack rate: in general, larger predators will eat more prey. Furthermore, a point not made in that paper, larger predators tend to eat larger prey, for example, large back-swimming bugs (notonectids) prey selectively on the later instars of mosquito larvae (Murdoch, unpublished data), which might have additional consequences for the prey population; since the larvae are pre-reproductive, the larger notonectids are thereby selectively killing prey individuals with high reproductive value (Fisher’s V,) and are therefore having a disproportionate effect on the future density of the prey. Fox and Murdoch (1974) have experimental data illustrating some of the relationships that make up the developmental response of notonectids to mosquito larvae. The existence of developmental response unfortunately complicates the bookkeeping needed to keep track of the predator’s functional response. The most simplified system we might consider, where only one predator individual is examined and only one age class of prey is vulnerable, is shown in Fig. 32, which uses a notonectid bug attacking a single age class of mosquito larvae as an example. Where the prey and predator both have age distributions that interact, a very complicated bookkeeping system would be necessary. We would also need to keep track of the predator’s numerical response. For example, Fox (1973) has shown that the cannibalism rate of notonectids is a function of prey density, predator density and predator age-distribution. It is possible that important insights would be drowned in the welter of information needed for such an integrative analysis, but there may be ways of extracting crucial answers from even such complicated analysis.

114

w. w.

MURDOCH

and

A. OATEN

An important consideration we have essentially neglected throughout the paper is the question of interaction between predators which would cause the functional response of each predator to be a function of predator density. This has been discussed by Holling (1969a), Hassell (1971), Hassell and May (1973) and Griffiths and Holling (1969).Hassell and May (1973) have suggested from their work modifying the Nicholson-Bailey model of parasitism that such interference may be a MOSQUITO RESERVOIR

10 day development A(A)-

- -

-7

-I

e.g. f , = P = K(I-e-OM),where

-

K-atbW,

P =#

prey eaten

M =#

prey available

WN= weight of notonectid A W N = f,(F,W,)

W =,

weight of mosquito

F =

weight of prey eaten

S = proportion of prey sucked

FIG.32. A diagram of a book-keepingsystem for a simple computer model of a single notonectid predator feeding and growing aa it attacks five-day-oldmosquito larvae. The values of the variables chango each day (day is subscripted i). A is the number of adult mosquitoes.

more important source of stability than is a stabilizing functional response. A central problem here is the question of whether or not predator densities in nature are often high enough for such interaction to be important. We turn now to problems of integrating these responses into a total response. Hassell (1966) suggested that we distinguish between intragenerational responses (functional and aggregative) and the intergenerational relationship, and we modify the latter here to the intergenerational (numerical) response, The advantage of making this

PREDATION AND POPULATION STABILITY

116

distinction is that the inter-generational response can be integrated with the intra-generational response simply by multiplying the two functions once per generation (Holling, 1959a). However, this scheme is roughly appropriate only for parasites, where attacks and reproduction occur only a t the start of each parasite generation. Even here, the numbers of adult parasites, when the adult actually feeds on prey exudate, may vary within each generation in response to prey density. The dichotomy is inappropriate for predators whose numbers can always vary in response to prey density because of their continuous intragenerational death rate, Indeed, there seems to be a tendency to forget that the predator’s death rate is just as responsive to prey density as is its birth rate. For predators in particular, then, the numerical response will have to be combined with the other responses by a process of continuous, or at least continual, integration of the components through time. I n fact, we do not yet have a description of these combined responses for any predator. Suppose we know the functional, P ( H ) ,and the numerical response, N ( H ) ,the no. of predators. We might obtain an approximation to the total response, the no. of prey killed by the entire predator population as a function of prey density, by multiplying these two functions, P ( H ) N ( H )= J ( H ) . Of course we will almost never be able to describe predation in this simple way, because of the complications noted above. However, for parasites, which have no developmental response, this simple multiplication may not be too bad an approximation. I n addition, the analysis of the product gives us some insight into the potential consequences of variously shaped functional and numerical responses. This analysis was fist made by Holling (1959a, Fig. 8). We pursue the analysis here since it is again an area in which some misconceptions have arisen; in particular it has been claimed by various authors that any increasing numerical response function N ( H ) will produce an initially accelerating total response function J ( H ) which will therefore be potentially stabilizing (Section 111).This is not the case. Some combinations of numerical and functional responses do produce density-dependence in J ( H ) . Consider the three types of functional response in Fig. 8 (Section 111).If the numerical response, N ( H ) ,passes through the origin and is either linear or initially accelerating, then J ( H ) is initially accelerating for all 3 types of P ( H ) . If N ( H ) passes through the origin but is decelerating (a “type 2” numerical response), J ( H ) is initially accelerating when P ( H ) is type 1 or type 3. The uncertain cases occur (i) when P ( H ) is type 2 and N ( H ) is decelerating (“type 2”) and (ii) when F ( H )is type 2 and N ( H )does not pass through the origin.

116

w. w.

MURDOUH

and

A. OATEN

(i).Clearly, J ( H ) = P ( H ) N ( H )need not accelerate when both functions in the product decelerate. Suppose, for the sake of simplicity, that F ( H ) = Ha, N ( H ) = Hb, and a+ 13 1. Then J ( H ) = Ha++, which is a decelerating function. The important point here, then, is that when predators have a type 2 functional response, the numerical response must rise rapidly if the product, F ( H ) N ( H ) ,is to be accelerating. The criterion is that J " ( H )> 0 for an interval beginning at the origin, i.e. F"N + 2 FIN' + FN" > 0. However, this is simply a restatement of the

-=

H FIG.33. Numerical responses N ( H ) , solid curves, and total responses J ( H ) , dashed curves, as a function of prey density H . To get total response N ( H ) is multiplied by P ( H ) = Ha.&.The difference between ( a )and ( b ) is explained in the text.

verbal criterion and is not of much general use. The solution will have to be found for each particular F ( H ) and N ( H ) .

(ii).Suppose F ( H ) is type 2 and N ( 0 ) is positive, i.e. N intersects the ordinate (Fig. 33). All of the numerical response curves examined by Holling (1959a, Fig. 8) intersected the ordinate. This is not an unreasonable assumption since presumably in many situations in nature there will still be predators present when the prey species in question has disappeared, as is assumed here: general predators will turn to feeding upon alternative prey; many invertebrate predators can survive for long periods without food. The general effect of a positive Y-intercept is to decrease the likelihood that the total response will be initially densitydependent, We illustrate this in Fig. 33. Suppose F ( H ) = H0.5. If N ( H ) = Ho*6 (i.e. passes through the origin), the resulting J ( H ) is accelerating. However, when N ( H ) is of the same form but cuts the ordinate, N ( H ) = H 0 . 6 + p , (the solid curve (a) in Fig. 33), then the resultant J ( H ) is decelerating, i.e. inversely density-dependent, initially (dotted curve ( a ) in Fig. 33).

PREDATION AND POPULATION STABILITY

117

On the other hand, suppose the numerical response curve intersects the abscissa. Again, this is a reasonable assumption since some species of predators are likely to disappear locally before their prey are driven extinct. For example, Connell (1971) has suggested that predators are more sensitive than are their prey to bad weather-seashore snails seem to suffer more than do barnacles from a severe winter; predatory insects are notoriously more sensitive to insecticide than are their prey; highly mobile predators may leave an area before the prey there are depleted; in seasonal species the prey usually becomes active before the predator does, so that at initially low prey densities there will be no predators. We illustrate this effect by first letting N ( H ) = H0.4.Then clearly J ( W ) = ( H 0 * 4 ) ( H 0 9is decelerating. Now, make N ( H ) intersect the Xaxis by setting N ( H ) = HO'"- 8, (solid curve ( b ) in Fig. 33), then J ( H ) becomes accelerating initially (dotted curve ( b ) in Fig. 33). Thus, by moving the numerical response so that it no longer passes through the origin, we can change the total response from stabilizing to destabilizing, and vice versa.

VIII.

CONCLUDINQ

REMARKS

We have covered a wide range of approaches to the study of predation, and have discussed the major points at the end of each section, so in this concluding section we will simply raise briefly a number of issues concerning the broad significance of some approaches and the kinds of new studies that are needed most. First, the motivation for much of the work we have done here, and for earlier work by the senior author, is the existence of an apparent paradox: many predator-prey systems in nature seem to be stable; that is, they seem to persist without noticeably large fluctuations, yet when we add reality (e.g. time lags, predator satiation) to simple models of predation the models become unstable. Furthermore, in the laboratory, functional responses to a single-prey species seem almost universally to be destabilizing. The paradox can be resolved,as we discussed in the first two sections of this paper, by adding other complications to the models: density-dependence (e.g. resource limitation) in the prey; refuges for the prey; an invulnerable class of prey; spatial heterogeneity; and indeed these complications exist in some populations in nature. Perhaps these ancillary aspects are both the necessary and sufficient stabilizing mechanisms in real systems. However, we have tried to see if exploring different aspects of the functional response exposes other mechanisms that can turn the response from a source of instability to a source of stability, and within this framework have examined both the effects E

118

w. w.

MURDOCH

and A.

OATEN

of prey patchiness and the predator’s response to it, and the existence of alternative prey species. Our tentative conclusion is that the latter is more likely to be stabilizing than the former. But we must immediately state a caveat: our conclusions concerning patchiness derive purely from simple models. Much remains to be done to improve these models, and of course they can only suggest the effect of patchiness in the real world. Second, we spent some time examining mechanisms and conditions that can lead to density-dependent functional responses, but one might well question that the functional response can be an important source of stability in nature. Perhaps the anecdotal evidence from biological control is relevant: van den Bosch (pers. comm.) and others have noted that, while specific parasites are important in controlling introduced pests, at least in cotton in California the major control of potential pests is carried out by a complex of predators-especially Heteroptera. These predators have generation times equal to or larger than those of the pests, suggesting that functional response, and perhaps the developmental response, are at least as important as the numerical response. (The potential pests, incidentally, are controlled mainly while they are on alternative crops such as safflower and alfalfa and before they can migrate on to the cotton.) This is pretty unsatisfactory evidence, but in fact we know surprisingly little about natural predator-prey systems. Hassell and May (1973) have suggested that the functional response may be less important for stability than are interactions among predators, which vary as a function of predator density. It is hard to evaluate this claim considering the simplicity of current models and the absence of good field data on the relative importance of these two aspects of predation. Third, perhaps the most striking characteristic of the current status of predation studies is the marked unevenness between the development of models (including laboratory studies) of predation, and field analyses. The models are in a fairly healthy state: we have reasonably adequate models of general predator-prey systems, and these general models are able to incorporate the main stabilizing features discovered in field studies- refuges, spatial heterogeneity-or probably can be modified to incorporate them, e.g. invulnerable classes of prey. Recent work with multi-species systems (especially May’s papers) has provided insight into the nature of mathematical generalizations. At a lower level of generality also, models of predation are rapidly becoming adequate. Beginning with the work of Holling, both mathematical and laboratory analyses have been developed for functional response and for interference between predators, for example. The effect of patchiness has been studied in this paper and elsewhere (Royama, 1970; Hassell and

PREDATION AND POPULATION STABILITY

119

May, 1973), as has the effect of having more than one prey species (e.g. Murdoch, 1969). (The numerical response is less well studied, as is developmental response.) This is not to say that such models cannot be improved; no doubt they should be and will be. But their development has vastly outstripped the study of the real situations they are meant to illuminate. Field studies themselves are uneven. Particularly from Connell’s work on the seashore, we have exemplary analyses of the stabilizing consequences of refuges. The importance of an invulnerable class of prey is also illustrated by his work and is suggested by other studies, such as the moose-wolf system on Isle Royale, although the dynamics of this mechanism must vary from system to system and it needs much more study. By contrast, we really have no satisfactory field analyses of the importance of those components of the predator-prey interaction that have been analyzed so intensively mathematically and in the laboratory: functional, developmental and numerical responses; interactions among predators as a function of predator density; prey patchiness as it affects functional response, predator aggregation and the interaction among coupled sub-systems. There is again some suggestive work; for example Hassell’s (1968) analysis of aggregation in the parasite Cyzenis, and key-factor analyses of predation (e.g. Varley and Gradwell, 1968) suggest that some components of predation are stabilizing. But there is a great need for field studies that will look experimentally at different examples of predator-prey systems to determine which components are stabilizing and which are destabilizing, and how these components actually operate. The value of the mathematical and laboratory analyses is that, by exploring possible relationships and mechanisms, they point to those features of field systems that should repay study. Finally, since we have been concerned with the question of the stability of predator-prey systems, various of the mechanisms we have discussed above bear upon the current ecological faith that diversity enhances stability. We can recognize two sorts of diversity: species diversity and spatial heterogeneity, and note in passing that the existence of an alternative prey, combined with predator switching, may lead to greater stability, which is relevant to the claim about species diversity and stability. Of those mechanisms we want to mention, the remainder fall into the spatial heterogeneity class. A brief list ie as follows: 1. Refuges for the prey. 2. Differences in space leading to a mosaic of sub-systems that are

mutually out of phase, and fluctuate at different frequencies. 3. Barriers to the predator’s dispersal.

120

w. w.

MURDOCH

and

A. OATEN

4. Patchiness of the prey tending to increase the stabilizing ability of

the functional response. 5. Differences in the environment causing two-prey species to occur in

sub-habitats, thus encouraging switching. This is not the place for a lengthy disquisition on diversity and stability, but it is perhaps worth noting that this analysis is consistent with a general claim that the major aspect of diversity that leads to stability in model and simplified systems (such as the laboratory and agriculture) is spatial heterogeneity rather than species diversity. Since natural systems with few species can be very stable, we wonder whether our claim about spatial heterogeneity in simplified systems may not also apply to natural systems.

ACKNOWLEDGEMENTS We are grateful to J. H. Connell and P. S. McNulty for commenting on a draft of some sections. S. Avery provided excellent technical assistance. The following people kindly gave us access to unpublished manuscripts, theses or data: J. C. Allen, J. H. Connell, L. R. Fox, M. P. Hassell, J. R. Krebs, D. E. Landenberger, L. Luckinbill, R. M. May, R. J. Marks, R. K. Murton, D. J. Rapport, R. C. Reed, J. L. H. St Amant, R. J. Taylor and D. M. Ware. Our own research described here was supported by grants from N.S.F.

REFERENCES Allen, J. C. (1974). Mathematical models of spocies interactions in time and space. I n prep. Bailey, V. A,, Nicholson, A. J. and Williams, E. J. (1962). Interaction between hosts and parasites when some host individuals are more difEcult to find than others. J. theor. Biol. 3, 1-18. Baldwin, N. S. (1964). Sea lamprey in the Great Lakes. Can. Audubon Nov.-Dec. 1964, pp. 2-7. Banks, C . J. (1957). The behaviour of individual coccinellid larvae on plants. Br. J. Anim. Behav. 5, 12-24. Bartlett, M. S. (1957). On theoretical models for competitive and predatory biological systems. Biometrika 44, 27-42. Brooks, J. L. and Dodaon, S. I. (1965). Predation, body size, and composition of plankton. Science, N.Y. 150, 28-35. Bryan, J. E. and Larkin, P. A. (1972). Food specialization by individual trout. J . Fish. Res. Bd Can. 29, 1615-1624. Buckner, C. H. and Turnock, W. J. (1965). Avian predation on the larch sawfly, Priatiphora erichsonii (Htg.). Ecology 46, 223-236.

PREDATION AND POPULATION STABILITY

121

Burnett, T. (1954). Influences of natural temperatures and controlled host densities on oviposition of an insect parasite. Phy8iol. 2061. 27, 239-248. Burnett, T. (1964). Host larval mortality in an experimental host-parasite population. Can. J. 2001.42, 745-765. Clark, L. R. (1963). The influence of predation by Syrphua sp. on the numbers of Cardiaqira albitextura. Aust. J . 2001.11, 470-487. Clarke, B. C. (1962). Balanced polymorphism and the diversity of sympatric species. I n “Taxonomy and Geography” (Ed. D. Nichols).Syst. Assoc. Publ. 4, Oxford. Clarke, B. C. (1969). The evidence for apostatic selection. Heredity 24, 347352.

Connell, J. H. (1970). A predator-prey system in the marine intertidal region. I. Balanua glandula and several predatory species of Thaie. Ecol. Monogr. 40, 49-78.

Connell, J. H. (1971). On the role of natural enemies in preventing competitive exclusion in some marine animals and in rain forest trees. Proc. adv. Study I m t . D y n a m h Numbers Pop&. (Oosterbeck, 1970), 298-312. Connell, J. H. (1972). Community interactions on marine rocky intertidal shores. Ann. Rev. Ecol. and Sys. 3, 169-192. Connell, J. H. (1974). A predator-prey system in the marine intertidal region. 11. Balanus cariosw and its predators. In prep. Cox, D. R. (1970). “Renewal Theory.” 142 pp. Methuen, London. Dawkins, M. (1971). Perceptual changes in chicks: another look at the “search image” concept. Anim. Behav. 19, 566-574. Dixon, A. F. G. (1959). An experimental study of the searching behaviour of the predatory coccinellid beetle Adalia dtxempmtata (L). J . Artim. Ewl. 28, 259-281.

Dixon, A. F. G. (1966). The effect of population density and nutritive stat- of the host on the summer reproductive activity of the sycamore aphid, Drepanosiphum platanodes (Schr.).J . Anim. Ewl.35, 105-112. Elton, R. A. and Greenwood, J. J. D. (1970). Exploring apostatic selection. Heredity 25, 629-633. Fleschner, C. A. (1950). Studies on searching capacity of the larvae of three predators of the citrus red mite. Hilgardia 20, 233-265. Fox, L. R. (1973). Food limitation, cannibalism and interactions among predators: effects on populations and communities of aquatic insects. Ph.D. Thesis, Univ. of California, Santa Barbara, California, U.S.A. Fox, L. R. and Murdoch, W. W. (1974). Developmental response in Notonectids. I n prep. Gause, G. F. (1964). “The Struggle of Existence.” 163 pp. Hafner, New

York. Gilpin, M. E. (1972). Enriched predator-prey systems: theoretical stability. Science, N . Y . 177, 902-904. Griffiths, K. J. (1969). T h e importance of coincidence in the functional and numerical responses of two parasites of the European pine sawfly Neodiprion sertifer. Can. Ent. 101, 673-713. Griffiths, K. J. and Rolling, C. S. (1969). A competition submodel for parasites and predators. Can. Ent. 101, 785-818. Hall, D. J., Cooper, W. E. and Werner, E. E. (1970). An experimental approach to the production dynamics and structure of freshwater animal communities. Limnol. Oceanogr. 15, 839-928.

w. w.

122

MURDOCH

and

A. OATEN

Harcourt, D. G. and Leroux, E. J. (1967). Population regulation in insects and Man. Am. Scient. 55, 40CL415. Hassell, M. P. (1966). Eveluation of parasite or predator responses. J. Anim. Ecol. 35, 65-75. Hassell, M. P. (1968). The behavioural response of a tachinid fly (Cyzenk albicana (Fall.)) to its host, the winter moth (Operophtera brumata (L.)). J . Anim. Ecol. 37, 627-639. Haasell, M. P. (1969). A study of the mortality factors acting upon Cyzenk albicane (Fall.), a tachinid parasite of the winter moth (Operophterabru& (L.)). J . Anim. Ecol. 38, 329-339. Hassell, M. P. (1971). Mutual interference between searching insect parasites. J. Anim. Ecol. 40, 473-486. Hassell, M. P. and May, R. M. (1973). Stability in insect host-parasite models. J. Anim. Ecol. 42, 69S726. Hobson, E. S. (1968). Predatory behaviour of some shore fishes in the Gulf of California. Bureau of Sport Fisheries and Wildli,fe,Research Report8 73.91 pp. Holling, C. S. (1959a). The components of predation as revealed by a study of small-mammalpredation of the European pine sawfly. Can. Ent. 91,293-320. Holling, C. S. (1959b). Some characteristics of simple types of predation and parasitism. Can. Ent. 91, 3 8 6 3 9 8 . Holling, C. S. (1966). The functional response of predators to prey density and its role in mimicry and population regulation. Mem. ent. Soc. Can. 45, 1-60. Holling, C. S. (1966). The functional response of invertebrate predators to prey density. Mem. ent. SOC.Can. 48, 1-86. Huffaker, C. B. (1958). Experimental studies on predation. 11.Dispersion factors and predator-prey oscillations. Hilgardia 27, 343-383. Huffaker, C. B. and Kennett, C. E. (1959). A 10 year study of vegetational changes associated with biological control of Klamath weed species. J. Range Mgmt 12, 69-82.

Ivlev, V . S . (1961). “Experimental Ecology of the Feeding of Fishes.” Yale University Press, New Haven, Connecticut, U. S. A. Jordan, P. A., Shelton, D. C. and Allen, D. L. (1967). Numbers, turnover, and 7, 233-252. social structure of the Isle Royale wolf population. Am. 2002. Kitching, J. A. and Ebling, F. J. (1967). Ecological studies a t Lough Ine. Advances in ecol. Rea. 4, 197-291. Krebs, J. R. (1973). Behavioural aspects of predation. I n : “Perspectives in Ethology” (Eds P. P. G. Bateson and Peter H. Klopfer). Plenum Press, New York. Landenberger, D. E. (1968). A study of predation and predatory behaviour in the Pacific Starfish, Pisaater. Ph.D. Thesis, Univ. of California, S m t a Barbara, California, U.S.A. Landenberger, D. E. (1973). Predation by starfish. Unpublished MS. Leslie, P. H. (1948). Some further notes on the use of matrices in population mathematics. Biometrika 35, 213-245. Leslie, P. H. (1967). An analysis of the data for some experiments carried out by Gause with populations of the Protozoa, Paramecium aurelia and Paramecium caudatum. Bwmetrika 44, 314-327. Leslie, P. H. (1958). A stochastic model for studying the properties of certain biological systems by numerical methods. Biometrika 45, 16-31. Leslie, P. H. and Cower, J. C. (1960). The properties of a stochastic model for the predator-prey type of interaction between two species. Biometrika 47, 219-234.

PREDATION AND POPULATION STABILITY

123

Lotka, A. J. (1925). “Elements of Physical Biology.” Williams and Wilkins, Baltimore, Maryland, U.S.A. Luckinbill, L. S. (1974). Coexistence in laboratory populations of ’Paramecium aurelia and its predator Didinium nasutum. In prep. MacArthur. R. H. (1970). Species packing and competitive equilibrium for many species. Theor. pop. Biol. 1, 1-11. Manly, B. F. J., Miller, P. end Cook, L. M. (1972). Analysis of a selective predation experiment. Am. Nat. 106, 719-736. Marks, R. J. (1970). A study of interactions in an invertebrate predator-prey system. M.Sc. Dissertation, University of London, London, England. May, R. M. (1971). Stability in model ecosystems. Proc. ecol. SOC.A w t . 6, 18-56. May, R. M. (1972). Limit cycles in predator-prey communitics. Science, N . Y . 177, 90&902.

May, R. M. (19734. Time delay versus stability in population models with two and three trophic levels. Ecology 54, 318-325. May, R. M. (1973b). “Stability and Complexity in Model Ecosystems.” Princeton University Press, Princeton, New Jersey, U.S.A. Mech, L. D. (1966). “The Wolves of Isle Royale.” 27.8. Nat. Park Sew. Fauna Nat. Parks US.,Fauna Series No. 7. Messenger, P. S. (1968). Bioclimatic studies of the aphid parasite Prwm emoletum. I. Effects of temperature on the functional response of females to varying host densities. Can. Ent. 100, 728-41. Mook, J. H., Mook, L. J. and Heikens, H. S. (1960). Further evidence for the role of “searching images” in the hunting behavior of titmice. Archa nderl. 2001.13, 448-465.

Mori, H. and Chant, D. A. (1966). The influence of prey density, relative humidity, and starvation on the predacious behaviour of Phytoseivlua persirnitis Athias-Henriot (Acarina: Phytoseiidae). Can. J. 2001.44, 483-491. Morris, R. F. et al. (1956). The population dynamics of the spruce budworm in eastern Canada. Tenth intern. Congr. Entomol., Proc. 4, 137-149. Morris, R. F. (Ed.)(1963). The dynamics of epidemic spruce budworm populations. Mem. ent. SOC.Can. 31, 332 pp. Murdoch, W. W. (1966). Aspects of the population dynamics of some marsh Carabidae. J. Anim. Ecol. 35, 127-156. Murdoch, W. W. (1969). Switching in general predators: Experiments on predator specificity and stability of prey populations. Ecol. Monogr. 39,335364.

Murdoch, W. W. (1971). The developmental response of predators to changes in prey density. Ecology 52, 132-137. Murdoch, W. W. (1973). The functional response of predators. J. appl. Ecol. 14, 335-341.

Murdoch, W. W. and Marks, R. J. (1973). Predation by coccinellid beetles: Experiments on switching. Ecology 54, 160-167. Murdoch, W. W., Avery, S. L. and Smyth, M. E. B. (1974). Switching in predatory fish. I n prep. Murton, R. K. (1971). The significance of a specific search image in the feeding behaviour of the Wood Pigeon. Behaviour 40, 10-42. Oaten, A. and Murdoch, W. W. (1974a). Functional response and stability in predator-prey models. Am. Nat. I n press. Oaten, A. and Murdoch, W. W. (1974b). Functional response, switching and stability. Am. Nut. I n press.

w. w.

124

MURDOCH

and

A. OATEN

Oaten, A. and Murdoch, W. W. (1974~).The effect of an invulnerable age class on the stability of predator-pray models. In prep. Oaten, A., Murdoch, W. W. and McNulty, P. S. (1974). Prey patchiness, functional response and stability. I n prep. Popham, E. J. (1941). The variation in the colour of certain species of Arctocoriaa (Hemiptera, Corixidae) and its significance. Proc. 2001. SOC.Lond. (A) 111, 136-172.

Quesada, J. R. (1969). Population biology of the cottony-cushion scale, Icerya parchmi Maskell (Homoptara: Coccidae), and its natural enemies. Ph.D. Thesis, Univ. of California, Biverside, California, U.S.A. Rapport, D. J. (1974). Response of S. coerukna to changes in prey density. In prep. Reed, R. C. (1969). An experimental study of prey selection and regulatory capacity of bluegill sunfish (Lepomia macroohirms). M.A. Thesis, Univ. of California, Santa Barbara, California, U.S.A. Richards, 0. W. (1961). The theoretical and practical study of natural insect populations. A . Rev. Ent. 6, 147-162. Rosenzweig, M. L. (1972). Comment on Gilpin (1972). Science, N.Y. 177, 904.

Rosenzweig, M. L. (1973). Exploitation in three trophic levels. Am. Nat. 107, 275-294.

Royama, T. (1970). Factors governing the hunting behaviour and selection of food by the great tit (Parzcs major L.). J . Anim. Ecol. 39, 619-659. Royama, T. (1971). A comparative study of models for predation and parasitism. Rea. popln. Ecol. Kyoto. Suppl. 1, 1-91. Salt, G. W. (1967). Predation in an experimental protozoan population ( WoodruffiaParameoium). Ecol. Monogr. 37, 113-144. Sandness, J. N. and McMurty, J. A. (1970). Functional response of three species of Phytoseiidae (Acarina) to prey density. Can. Ent. 102, 692-704. Smith, F. E. (1972). Spatial heterogeneity, stability, and diversity in ecosystems. I n “Growth by Intussusception, Ecological Essays in Honor of G. Evelyn Hutchinson.” Trana. Conn. Acad. Arta Sci.44, 30%335. Solomon, M. E. (1949). The natural control of animal populations. J. Anim. E c o ~18, . 1-35. St Amant, J . (1970). The mathematics of predator-prey interactions. M.A. Thesis, Univ. of California, Santa Barbara, California, U.S.A. Takahashi, F. (1968). Functional response to host density in a parasitic wasp, with reference to population regulation. Rea. popul. Ewl. 10. 5468.

Taylor, R. J. (1972). The role of learning in insect parasitism. Ph.D. Thesis, Univ. of California, Santa Barbara, California, U.S.A. Tinbergen, L. (1960). The natural control of insects in pine woods. I. Factors influencing the intensity of predation by songbirds. A r c h nderl. 2001.18, 265-343.

Varley, G. C. and Gradwell, G. R. (1963). The interpretation of insect population 142-156. changes. Proc. Ceylon Aaa. Advmt. Sci. 18 (D), Varley, G. C. and Gradwell, G. R. (1968). Population models for the winter moth. I n “Insect Abundance” (Ed. T. R. E. Southwood). Symp. R . ent. Soc. London 4. Ware, D. M. (1971). The predatory behaviour of rainbow trout (Salmo cairdnen‘). Ph.D. Thesis, Univ. of British Columbia, Vancouver, Canada.

PREDATION AND POPULATION STABILITY

125

Whittaker, R. H. and Feeney, P. (1971). Allelochemics: chemical interactions between species. Science, N . Y . 171, 767-770. Wood, L. (1968). Physiological and ecological aspects of prey selection by the marine gastropod Uroaalpinx cinerea (Prosobranchiat:Muricidae). Malacobgia 6, 267-320.

126

W. W. MURDOCH

and

A. OATEN

A P P E N D I XI Solutions to predator-prey equations All the predator-prey equations we deal with in Section I11 can be reduced, by ignoring second and higher order terms of factors assumed to be small, to systems of simultaneous first-order (i.e. involving only first derivatives) linear differential equations, special cases of the system ax&& = aalx,+a~p,+... +aakxk, for i = 1, 2, ..., k, where the atj's are given constants. Such systems are solved, and their solutions discussed, most easily by the use of matrix methods (St Amant, 1970; May, 1971). However, less elaborate methods suffice when only two equations are involved, e.g.

ah

- =

at

rh+sp;

at at

=

uh+vp

We consider only cases where both s and u are non-zero. If both are zero, the equations are unconnected and have solutions h = Aert and p = BeVt where A and B are arbitrary constants; if, say, s = 0 but u # 0 , then h = Aert and the equation for p can be rewritten, multid plying throughout by e-vt, as - {e-Vtp(t)} = Aue(*-v)t so that at

Although these cases do not interest us further, they suggest an approach to our problem. From the form of the solutions to these simple cases, we guess the solution of our more general case (s # 0, u # 0) to be h = yeAt and p = Sept. To find the constants y, 8, A and p, we substitute these solutions into ( A l ) and get Ayeat = rye*t+s8eflt;BSept = uyeAt+vseFt

(A21

These equations hold for all t ; dividing the first by eAt and the second by ebt, we see this is possible only if A = p (since s and u are non-zero). If this is so, then after the division we find that ( y , 8) must be the meeting point of the straight lines

sS = ( A - r ) y and (A-v)S = uy

(A3)

But these lines obviously meet at zero. If we are to have any solution to (Al), except the trivial p = h = 0, these lines must'also meet

127

PREDATION AND POPULATION STABILITY

somewhere else; since they are straight lines, they can only meet A-r u somewhere else if they are the same line. This requires - = -or s A-v (A-r)(A-w)-us

= 0

This is the characteristic equation of the matrix

:I.

[;

(A4)

Its solutions

are

A = . - - r+v +${(r-v),+ 4su}1’2 2 We will call these two solutions A, and A,, with A, the larger. If A takes on either of these values, the two lines in (A3) are identical and there are infinitely many points-all

1

points ( y, 6) with 6 = - ( A - r )y-solving 8

both. For a given h (either A, or ha), any suitable pair ( y , 6) will do, since it is easily seen from ( A l ) that, if h(t) and p ( t ) are any solutions of ( A l ) then so are Ah(t)and A p ( t )for any number A . Thus we can take = 1 and

6,

=

1

- (A1-r) to get that h

= eAlt and

p = S,eAlt is one

8

1

solution to (A1); similarly, with y = 1 and S, = - (A,-r), h = 8

eAet

and

p = sseAat is another solution. We can also see easily from (Al) that if h = h,(t) and p = pl(t)is any set of solutions, and h = h2(t) and p = p,(t) is any other set, then h = Ah,(t)+Bh,(t) and p = Ap,(t)+ Bp2(t)is also a solution. Thus we have the general solution h(t) = AeAit+BeI9Pt p ( t ) = A6,e*it + BS2eA2t (A6)

It can further be shown that this is the only solution: i.e. if h(t)and p ( t ) is a solution of ( A l ) then there are constants A and B so that h(t) and p ( t ) satisfy the formulae above. In our case, the A and B are given by the initial conditions: if we know h and p when the system is started (i.e. at t = 0) then we find A and B by solving A + B = h(0) and AS,+ BS, = p ( 0 ) . (A7) The behaviour of the solutions in (A6) as t varies, depends on A, and A,, the two values given in (As). We list the possibilities: If

(r-v),> - 4 s u

(A8)

w. w.

128

MURDOCH

and

A. OATEN

then both solutions are real and the system is (a) unstable ( h and p increase indefinitely) if either r + v > 0 or su > rv; (b) stable (h and p decrease to zero) if r + v < 0 and su < rv; C

c

r

where c r+v' p3ir+vy depends on the initial conditions) if r+v< 0 and YU = rv. If (r - v), < - 4su (A91

(c) convergent to a new equilibrium (h+-

then both solutions are complex; in this case the solutions oscillate and the system is (a) unstable if r+v > 0 (the amplitude of the oscillations increases indefinitely) ; (b) stable if T + v < 0 (the oscillations are damped-their amplitude

tends to zero); (c) oscillatory if r + v = 0 (the oscillations have constant amplitude). We now discuss this list briefly. Suppose fist that (r - v), 2 - 4su so both solutions are real. Since A, >= A,, it is then clear from (A6) that whether h and p increase indefinitely, converge to a constant, or decrease to zero, depends on whether A, is positive, zero or negative respectively. This assumes 6, # 0, which follows from s u # 0, and assumes A # 0 which is true unless p ( 0 ) = Sah(0), a rather strange condition for the initial perturbation to satisfy (if it happens, the behaviour of the solutions depends on A,). T+V

If ( r - v ) , ~ - k u , then A, and A, are al+ia2, where a1 = a,

=

4-1

2 '

${-(~-v)~-&u}~'~ and i = as usual. Recalling that ez(cos y + i sin y), we get from (A6),

ez+@ =

h(t) = ealt{(A+ B ) cos a,t

+ i ( A- B ) sin a2t}

and

p ( t ) = ea1t{(A6,+B6,) cos a2t+i(A6,-B6,) sin a2t}. From (A7) we get

A + B = h(O), A6,+ B6,

=

p(O), A - I3 = i{(r- a,)h(O)-sp(O))/a,

and

AS,- B6, = ia,h(O)+ ( A- B)al; eubstituting these in, we see that all terms become real and, using

PREDATION AND POPULATION STABILITY

129

standard formulae for cosines of sums of angles, that the solution can be written as h(t) = ealrRz cos (a,t+

el), p ( t ) = ealrR,COB (a&+ O,),

where R,, R,,8, and 8, are functions of A , B , 6, and SZ. The “cos” causes the solutions to oscillate, the “a,” causes the period of the oscilla2n

tion to be -, and the

“ealt”

determines the amplitude of the oscilla-

a2

tions. (A9) follows from these observations.

A P P E N D I XI1 Census data of Bupalus density and composilion of food between 7 and 10 a.m..,from Mook et al. (1960) H E = Calculatad number of Bupalua moths on 10 000 sq. m. N = Number of prey brought to nestbox between 7 and 10 a.m. B = Number of B u p d w , moths brought to nestbox between 7 and 10 a.m. ~

Date

19,

N

N-B 7 . H B

B

__

1952 May 23 24 25 26 27 28 29 30 31 June 1 2 3 4 5 6 1954 May 26 27 28 29 30 31 June 1

71 48 74 37 222 111 111 37 37 407 333 259 111 296 259 0 125 167 83 292 42

0

N

75 82 73 85 104 74 91

N-B -.HB

N

N-B B.H,

B

N

B

N-B ~ .

~

13 P

13 d 48 47 50 60 70 65 65 27 44 50 60 58 36

B

2 1 8 4 31 12 5

3 13

5 5 1 1 29 10 17 17 35 27 4

14

1702 1702 1165 1554 140 163 444 3256 795 2331 1221 16872 9065

53 68 43 88 129 129 89 161 119 116 81 123 86

0 478 550 118 833 735 0

44 42 53 67 73 85 33

d

7 1 1 18 60 6 13 133 86 21 24 12 6 29 ? 2 9 16 17 22 8

4

486 2479 9324 432 128 759 216 86 128 1171 263 2738 3453

35 53 42 37 38 35 66 56 35 60 53 65 61 76 98

36 5 0 0 3

d

1

0 38

11 5 33 37 28 17 40 63

432 03 03

419 8214 03

83 151 222 333 144 342 287 266 144

67 68 51 56 47 45 63 81 66 80 88 86 110 126 85

18 ?

18 d 458 386 244 677 404 0

68 54

13 30

36 ? 3 1536 0 00 1 3700 654 3 10212 1 1 4884 33 49 9 296 85 20 57 164 164 59 52 169 53 119 358 57 156 53

178 0

48 83

2 11

966 0

H

B

1954 June

7 8 9 10 11 12 13 14 15

1955 May 30 31 June 1 2 3 4 5 6 7 8 9 10 11 12

292 208 250 292 125 625 500

126

37

625 500 222 83 389 167 306 222 250

64 81 83 70 81 58 41 86 59

33 24 70 55 26 53 42 25 56 22

44

26 13 27 30 41 50 107 57 61 102 132

111 89 156 89 333 311 400 734 356 244

I

102 109 123 116 I17 158 105

702

6

4 d 1 2 2 2 2 6 20 8 51 47 58

1042 79 113 140 206 64 196 119 420

75 63 61 58 79 66 80 103 97

1125 1332 1113 2184 1736 2442 1353 2450 144 417 311

17 25 35 27 61 53 117 121 102 75 168

33 61 53 29 17 53 25 42 49 33

0

4

? 0 0 0 3

1 2 16 75 91 39 59

143 94 246 200 191 274 277 245 485

94 102 104 90 103 85 80 102 93

cc

78 126 95 79 72 119 158 43 59 62 88

38 6 73 9.1

85 70 86 65 64 82 79

co 1248 5340 8492 1963 245 89 329 45 1

~

180 43 50 24 77 51 77 54

182 143 121 147 195 19'7 106

'

76 89 76 70 69

44

6

22 o?

735 274 116 97 332 135 173

29 47 84 87 32 130 78

2 1 0 1 0 10

3 5 22 18 36

1710 13875 co 12168 W

3630 16068 3040 1234 870 352

I

55 52 55 57 77 88 81 42 62 60 62

4539 3510 470 533 1398 2307 341

11 8 42 52 16 42 63

38 16 7 29 34 47 40 47 39 43

9

22

?

1484 2928 360 134 240 175 85 176 151

3 1 1 2 2 38 24 21 47 28 19

780 5661 4806 4290 3338 438 739 400 234 407 552

40

495

ce

! 208

142

511

c

%

This Page Intentionally Left Blank

Mathematical Model Building with an Application to Determine the Distribution of DursbanB Insecticide added to a Simulated Ecosystem G. E. BLAU

Computation Research, T h Dow Chemical Company, Midland, dlichigan 48640, U.S.A.

and W. BROCK NEELY

Ag-Organic8 Product Department, The Dow Chemical Company, Midland, Hichigan 48640, U.S.A. I. Introduction

-

-

-

11. Model Building Techniques

-

. -

-

-

-

. -

-

-

-

. .

-

-

-

-

133 134 134 136 138 139 143 144 146 146 147 148

-

149 149 149 161 160 162 162

-

Types of Mathematical Models Model Building Procedure The Design Problem and the Analysis Problem The Likelihood Approach to Model Discrimination Example of Model Discriminetion by Likelihoods Parameter Estimation Procedures G. Testsof Model Adequacy 1. Goodnewof Fit 2. Residual Analysis H. Conclusion III. The Environmental Fate and Distribution of DURSBANB Added to en . Ecosystem A. Introduction B. Description of the Ecosystem C. Building theMode1 D. Discussion ofResults E. Conclusion References A. B. C. D. E. F.

-

~

-

-

~

-

-

~

~

-

-

-

-

-

-

-

-

~

-

I. INTRODUCTION Mathematical models of different types and different levels of sophistication have been widely used in the chemical industry. These models have ranged from large plant models, used to determine the optimum operating conditions which maximize or minimize some economic criterion, to process models which predict the steady state operation of processes or the dynamio response of the process to 133

o.

134

E. BLAU

and w.

BROCK NEELY

disturbances. A plant model usually consists of several process models. In order to obtain these process models it is necessary to have some understanding of the underlying physical and chemical mechanisms involved. This gives rise to mechanistic or phenomenological modelling. These mechanistic models should be capable of describing the basic physical and chemical steps so that a process may be both designed and operated properly. Apart from these conventional engineering applications, mechanistic models provide valuable insight into the behaviour of any system in which chemical reactions are taking place. For example, it is vitally important to know the environmental impact that a chemical may have when added to an ecosystem. Many of these systems are very complex, so that selection of a suitable model is by no means transparent before or after data have been collected. I n fact, one can frequently postulate several models which, superficially at least, represent experimental kinetic data. The problem then is to determine the constants and if possible choose between these candidate models. This report presents a procedure for building a mechanistic model which represents an experimental reaction system. Starting with one or more plausible models, the principle of maximum likelihood is applied to the data collected in order to estimate the constants in the model and choose the best model among those originally postulated. Then, conventional statistical techniques are used to determine the suitability of this “best” model. If the model is inadequate, a technique is presented for identifying the specific limitations. Then the model builder must postulate additional physical meaningful models to accommodate this limitation and the procedure is repeated. There are two parts to this report. In the first part, the model building procedure is developed from elementary statistical principles. The important concept of maximum likelihood is introduced and illustrated with an example. The need for proper experimental design and an iterative experimentation-analysis program is presented with examples. The second part of the paper illustrates the model building procedure by finding a model which describes the fate and distribution of DURSBANQ insecticide added to a laboratory system which simulates a pond of water.

11. M O D E L BUILDINGTECHNIQUES

A.

T Y P E S OF MATHEMATICAL MODELS

I n theory, it is possible to represent all the phenomena occurring in any physical system by a precise mathematical model. To do this

MATHEMATICAL MODELS AND INSECTICIDE DISTRIBUTION

135

requires a complete description of the true scientific mechanism of each phenomenon. I n practice, however, a complete description of this mechanism is not available, so approximations must be made. The extent of these approximations classifies the mathematical model representation as mechanistic, empirical or regression. For example, if one is concerned with the basic steps that take place when a chemical is introduced into an ecosystem for different conditions of the system, e.g. amount of chemical added, temperature etc., a phenomenological or mechanistic model must be used. Here each term or group of terms represents some specific phenomenon such as the formation of a metabolite or the transfer of a chemical from one compartment to another. Obviously, development of this type of model requires an extensive and carefully designed testing program. Suppose, however, that considerable data has been gathered either in the laboratory or in the field on, say, the decomposition of an insecticide with time. Then, a regression model may be used to condense or organize this data so that it is readily accessible. No attempt is made to add any physical significance to the individual terms of the regression models, which are simply multivariable polynomials of different degrees. A compromise between the mechanistic model and regression model is the empirical or, as it is sometimes called, the qwi-mechanistic model. I n such a model, some physical meaning is attributed to the potential selection of terms for the model, although no attempt is made to identify the basic steps in the process being modelled. These models are widely used where the biological variation is high and/or the testing is minimal. A typical example is an attempt to characterize the biodegradation of chemicals by their 112 life. Here, the physical principle assumed is that a chemical biodegrades exponentially with time. However, the steps involved in this degradation process are left unspecified. This paper presents statistical methods for developing mechanistic mathematical models. Many of the methods employed, however, are valid for the other model types and in most cases were originally derived from techniques for regression models. Tests will be given to help guide the model-building practitioner in deciding whether his data is of sufficient quality to justify using these more meaningful, albeit more mathematically complex, mechanistic models.

B. M O D E L

BUILDING PROCEDURE

It is frequently possible to postulate several physically meaningful mathematical models describing the particular system being studied. Ueually these models are based on theoretical principles or intuitive insights from observations taken on analogous systems. I n general, the

136

G . E. BLAU

and w.

BROCK NEELY

degree of sophistication of these models will range from complex multiparametered models to simple one-parameter models. Model discrimination is the statistical procedure which chooses or distinguishes among the various postulated models to find the model or models which best describe the system studied. Note that this discrimination only takes place among the set of postulated models. That is, the model selected by model discrimination may be the best of the originally postulated models but totally inadequate in describing the actual physical system. Using statistical residual analysis on the data collected, it is frequently possible to identify specific inadequacies in this “best” model. The model builder should then be able to suggest other plausible models which include one or more additional terms to accommodate the inadequacies in the original model system. Then discrimination is carried out on the new models and the process is repeated, It is apparent from the foregoing discussion that model building is, in general, an iterative procedure. The steps may be summarized as follows: 1. Postulate one or more models to describe the physical system studied. 2. Use model discrimination techniques to identify the best model among those postulated in step 1 from experimental data collected on the system. 3. Determine whether the model identified in step 2 adequately describes the experimental data generated. If it does the procedure is terminated. 4. Use residual analysis to identify the specific inadequacies of the model selected in step 2 and suggest a new model or models to accommodate these inadequacies. Return to step 2.

This model building procedure is continued until a suitable model is found and the procedure is terminated at step 3. As an example, consider the problem of building a model to describe the appearance and disappearance of a chemical B with time where B is formed from A . Suppose concentration-time data is available for component B only. I n the absence of any prior knowledge of the chemistry of the process, the simplest model to postulate corresponds to an irreversible reaction ki

A+B

Ml

where k, is a reaction rate constant. By adding an additional parameter, k,, describing the reversible reaction between A and B, one obtains

MATHEMATICAL MODELS AND INSECTICIDE DISTRIBUTION

137

Choosing between these two models is equivalent to determining whether or not the reverse reaction rate constant k, is greater than zero, i.e. k, > 0. M, is said to be more complex than M, since it has an additional parameter. The effect on the model discrimination method of adding k, to M, to form M, is analogous to the physical chemistry phenomenon of changing the degrees of freedom in a system. That is, there is twice as much flexibility in making M, explain the data as M,. This increased flexibility is reflected in the statistical criterion used to discriminate the models. For example, if M, and M, “explain the data to the same extent”, the additional parameter k, is indeterminate and M, is said to adequately represent the data. Suppose the concentration-time data for this example exhibited a maximum. Then both M, and M, would be inadequate. It would be necessary to postulate different models to explain the data and recycle through the model building procedure. Some typical models which could account for such a maximum are k,

ks

A+ B+C

Generally, some of these models can immediately be eliminated by physicochemical reasoning. The most suitable of those remaining can be identified by the discrimination methods discussed below. If the model selected is still inadequate, additional ones can be postulated and the procedure continued until an adequate model or models is found. Usually little difficulty is experienced in generating a variety of models of varying degrees of sophistication. A good rule to follow in choosing models is to keep them as simple as possible (i.e. minimal number of parameters and degrees of freedom).I n fact, the best approach is to progress from the simplest model to progressively more complex models until no further increase in complexity is warranted by experimental uncertainties in this data. This principle of going from the simple to the complex is called Ockham’s razor (Solberg, 1972) or the principle of parsimony (Kittrell, 1970). A good example of this principle is the stepwise add procedure of multilinear regression analysis (Draper

138

a.

E. BLAU

and w.

BROOK NEELY

and Smith, 1966). Blau et al. (1970, 1972a, b) have demonstrated the utility of this technique in a wide variety of model building applications.

c. T H E

D E S I G N PROBLEM A N D T H E A N A L Y S I S PROBLEM

I n the procedure described above, it is assumed that the available data collected on the system is sufficient to choose between different models. I n many cases this is not true. Consider the problem of choosing between the following two chemical reaction models

k,

ka

A+B&

M2

ki

where A , B and C represent three chemical species and k,, k, and k, represent reaction rate constants. Concentration-time data is available for component B as shown in Fig. 1. Chemically speaking, to choose

time, t

FIG. 1. Inadequate concentrat,iontime data.

between these models it is necessary to decide whether the disks

appearance of B occurs irreversibly, B+C (Model l),or whether B is ks

in equilibrium with C , B+C (Model 2). The key to distinguishing the kr

models is the rate constant k,. That is, Model 1 is best if k, is zero while Model 2 is best if k, is nonzero. This can be expressed in statistical terms by saying, discrimination between the models is equivalent to testing the null hypothesis k, = 0. The data of Fig. 1 does not allow us to make this distinction between models. Even doubling the number of points between t = 0 and t = t, would shed no new light on the value of k,. What is needed, of course, is some data at times greater than t,. Figure 2 shows two situations which might arise. If Model 1 is correct the concentration of B would drop off to zero for tat,. Conversely, an

MATHEMATICAL MODELS AND INSECTICIDE DISTRIBUTION

139

equilibrium concentration greater than zero, Be > 0, would be observed if Model 2 were correct. Note that only one or two additional data points may be necessary to distinguish these models provided they are located or “designed” properly, i.e. tStu The foregoing illustrates that there are two aspects to the application of the model discrimination phase (step 2 of the model building procedure). The first is the design problem, i.e. choosing the experimental conditions in such a way that discrimination is possible. The second is the analysis problem, i.e. analysing the data to assess how much discrimination has been achieved. The design problem is the more xxx =M2 OOOr M I x ox xoxo ox

m X0

L

0

X

E

2

-

+

k! B e

0

OX$

lOnX

X0

Y

_,

”

v

c

X

0 c

I

0 X

0 C

V

X

0

O O 0

0

0 0

X

time,t---

tu

FIG.2. Adequate concentration time data.

fundamental one. If for some reason the analysis of the data is faulty it may be repeated. However, the damage of poor design is irreparable and invalidates subsequent data analysis regardless of its level of sophistication. Considerable research effort by the scientific community has recently been expended on this design problem (Box and Hill, 1967; Reilly, 1970; Hsiang and Reilly, 1971). The methods developed rely heavily upon efficient optimization algorithms implemented on high-speed computers. It is beyond the scope of this paper to discuss these methods in detail, and the interested reader is referred to the literature. In the remainder of this paper it will be assumed that adequate designs have been employed so that the problem in choosing among models is only one of analysis.

D.

T H E L I K E L I H O O D A P P R O A C H TO M O D E L D I S C R I M I N A T I O N

Suppose that a set of models has been postulated and experimental data has been collected. In this section the statistical methodology for

140

o.

E. BLAU

and w. BROCK NEELY

using the experimental data to discriminate among the models will be presented. The methods to be discussed here are intended to be applied to models which are nonlinear in the parameters, and are not recommended for models linear in the parameters. The two most commonly used approaches are the likelihood approach and the Bayesian approach. The latter is based upon a subjective interpretation of probability (Bayes, 1763), a measure of the degree of belief that an event will happen rather than the objective interpretation in which the probability of an event is a long-term relative frequency. The Bayesian approach is readily embraced by scientists and engineers who advocate using knowledge other than that contained in the data. On the other hand, likelihood methods are claimed to have an advantage in objectivity in that they “let the data speak for themselves”. Since the purpose of this paper is not to compare discrimination methods, the simpler likelihood method will be discussed. This is not an in indictment against the Bayesian approach. The interested reader may wish to compare the two methods in the excellent paper by Reilly (1970). Suppose p ( z , 8) is a probability function which when given values of one or more parameters 8, allows the probability of any outcome to be calculated. For example, the bionomial probability function

becomes the following function of x alone

which is the probability of obtaining x heada in five tosses of a true coin. In this case n is the number of tosses or trials, 2 the number of heads, and 8 the probability of a. head in one toss. Suppose now that the coin is being tested for trueness and therefore 8 is unknown. If it is tossed five times and a head turns up once, this available information may be substituted into the probability function to obtain

The terminology L(0) is used to emphasize that this is a function of 8 only and is called the likelihood function. If a value of 8, say 8,, is substituted into L(8),it gives the probability that the event which actually happened (one head in five tosses) would have if the value of 8 were 8,. Comparing the values of the likelihood function for two different

MATHEMATICAL MODELS AND INSECTICIDE DISTRIBUTION

141

values for 0 gives the relative plausibilities of those two values in t,he light of the data. The comparison is carried out by examining their ratio, L(0,)/L(d,), sometimes called the odds ratio. I n the above example, L(1/5)/L(1/2) = 2-51. This indicates that the data obtained are 2.51 times as probable if 8 = 1/5 as they are if 6' = 1/2, and the value 2.51 can be taken as the weight of the evidence against the coin being true. (1/5 is the value of 8 which maximizes L(8),while 8 = 1/2 if the coin is true.) This would not ordinarily be taken as strong evidence that the coin is not true. A likelihood ratio of 10 is ordinarily taken as showing a real difference in plausibility, while 100 denotes strong preferences for the value of one parameter over the other (Reilly, 1970; Barnard et al., 1962). This concept of a likelihood ratio for measuring the plausibilities of different parameter values can be extended to measuring the plausibilities of different mathematical models. First, consider the problem of discriminating two Models M, and M, where M,, a function of two parameters, is denoted fl(8,, d2, z),and M,, a function of three parameters, is denoted f2( el, 8,, 03, x),where x is a single independent or controlled variable. Suppose some dependent variable y is determined for n different experiments corresponding to n values of the independent variable generating the data set {(yt, xt)i = 1, ..., n}.If the Models M, and M, are to be used to predict the observed values of y, then

Yc

= fl(6'1,

Yt

= f2(0,,

$1,

Xi>

e,, e,,

+ Et xt)+ Ei

Ml M,

(3)

where E$ is the experimental error corresponding to the ith observation. For any set of parameter values, a set of differences between observed and calculated values is determined. These differences, called residuals, are given by e t ( 8 j ) = yt-fj(%,

a)j

= 1, 2

(4)

where 8, = (el, 0,) for Model M, and 8, = (el, 6',, 8,) for Model M,. Let p ( ~8j, , x; $!-J) represent the joint probability density function of all experimental errors in the observed values where

= ( E l , E2, ..., &,), x = (51, x2, ..., x,) 1c) = ($Il, $I, ..., a,bm). Here 1c) represents the parameters E

and of the probability distribution, e.g. the mean and variance for the normal distribution. Under the hypothesis that the j t h model is true, i.e. that there is no modelling error, the residuals are estimates of experimental error and may be substituted for the errors in the joint probability density function. This now gives a function depending only on the

142

(x.

E. BLAU and

w.

BROCK NEELY

parameters in the model and the form of the probability distribution which is the likelihood function L(81, +) = p(e(Or),+). If the experimental errors E( are uncorrelated from point t o point, (Draper and Smith, 1966), the joint probability density function is the product of the individual probabilities pr(ez(O j ) , q). That is

Now, if the experimental errors are independent (Draper and Smith, 1966) and normally distributed with zero means and a known variance u2,the individual probability density functions are

Substituting these values into Eqn ( 5 ) gives the likelihood function

which is valid for the j t h model. For any particular parameter values, A e.g. 0, = O j , this function gives the probability that the data set which actually was generated ({(yl,zr)i = 1, ...,n } )would have been generated A by the j t h model with parameters t9j = 0,. Since M, has two parameters, while M3 has three parameters, it is not possible t o form a likelihood ratio for the same parameter values. Consequently, some way must be found t o eliminate this “nuisance” dependence on the parameters. One way of eliminating parameters is by “maximizing them out”. Thus, form the likelihood ratio

This is a comparison of the likelihoods of the two models at their individual best. It is simply a comparison of how well the two models can be made t o fit the data, expressed in likelihood terms. To compare more than two models the maximum likelihood for each model is calculated and two-way comparisons made by examining the ratios. This is expressed by the relationship

where m is the number of different models being compared.

MATHEMATICILL MODELS AND INSECTICIDE DISTRIBUTION

143

When the models have different numbers of parameters there are inherent difficulties with any discrimination method. Using the likelihood method described here, good discrimination requires that the likelihood ratio be much higher than usual if the favored model is the one with the larger number of parameters. Now the likelihood functions Lj(0j)given by Eqn (7) are maximized by choosing 0, values which minimize

This is the familiar least-squares criterion for estimating 0,. In passing it should be noted that the justification for using the lea&-squares criterion t o obtain parameter estimates is that it maximizes likelihood function when the error distribution is normal. The maximum likelihood for the j t h model can be written

L p = Ip,"" Lj(0,) = exp ( - RSSj/2u2)

(11)

where

RSS~= ;*wj(ej) (12) is the conventional Tesidw;cl sum of sqwzrtx obtained with the optimal least-squares parameter estimates. Since only ratios are relevant between likelihoods, the constants which multiply all the likelihoods in a comparison set are irrelevant and have been dropped from Eqn (1 1). The maximum likelihood approach is in principle easy to use. For each of the models postulated, determine the least-squares parameter estimates and the associated residual sum of squares. Then select the model with the smallest residual sum of squares and calculate the likelihood ratios relative to this model. Recall that a likelihood ratio of 10 is ordinarily taken as showing a real difference in plausibility while 100 denotes a strong preference for one model over the other. These numbers assume that the number of parameters in the models are the same. Therefore, it is necessary that the likelihood ratios be somewhat higher than usual if the favored model has a large number of parameters.

E.

EXAMPLE O F MODEL DISCRIMINATION

B Y LIKELIHOODS

Consider the problem of choosing between the following three models (Reilly, 1970) M,: Yt = e l l ~+t Et M,: yz = eZl+e,,zt-fEz M3: YC = e31 exp ( o , p t ) +

ct

144

and w.

Q. E. BLAU

BROCK NEELY

where x is a single dependent variable and y is the dependent variable. Data was collected at four different values of x giving the points shown below xt

YZ

0 1 2 3

- 1.290

5.318 7.049 19.886

It is also known that the errors E( are normally distributed with means zero and variances u2 = 1. The residual sum of squares RRSj for each of the models and the maximum likelihood values are shown below Model (j)

RSSj

KL;

1 2 3

28.465 22-473 11.853

0-050

1 202.2

L,*ILj* 4000 202.2 1

The maximum likelihoods have been multiplied by a constant K in order to give them manageable values. Model 3 is obviously preferred to the other models. I n fact, the data were generated artificially using Model 3 with Oal = O,, = 1 and u = 1.

F.

PARAMETER ESTIMATION PROCEDURES

An important part of the likelihood discrimination method is the determination of those parameter values which minimize the least squares criterion of Eqn (10). That is, it is necessary to have a procedure which will find those parameter values 8* which n

minimizeS(8) = 9

n

C e2(8)2= iC [Y6-f(e,xi)]2 i=l 1

(13)

where xi = (zli, x2t, ..., xp:pz)represent the ith value of p independent variables. For models that are linear in the parameters, i.e. models of the form

the parameters are readily estimated by linear least squares (Draper and Smith, 1966). To obtain the estimates 8*, it is only necessary to solve a p x p system of linear equations, for which a unique solution is usually guaranteed. Unfortunately, most meaningful mechanistic models are nonlinear in

MATHEMATICAL MODELS AND INSECTICIDE DISTRIBUTION

145

the parameters. Here it is necessary to apply iterative parameter estimation procedures. That is, a sequence of parameter estimates 01, 02, ..., 88, ... are generated which eventually converge to the optimum. This presents numerous complications such as initial guesses of O1 to institute the sequence, efficiency and effectiveness of convergence algorithms, multiple minima in the least-squares surface, and poor surface conditioning (Rosenbrock and Storey, 1966). A discussion of these topics is beyond the scope of this paper and are mentioned only to inform the reader that these problems exist. Nonlinear least-squares parameter estimation is a nontrivial task. The paper by Bard and Lapidus (1968) discussed the merits of several of the different algorithms as they relate to maximum likelihood estimation.

G.

TESTS O F MODEL ADEQUACY

After likelihood discrimination has chosen the best model from the set of candidate models, it is still necessary to test the suitability of this model to describe the data. Then a method is needed to identify any specific limitations in the model so that the model builder may modify the existing model to overcome these limitations. Although several new methods exist (Blau et al., 1972a, b), they do not supplant the more conventional tests of model adequacy of classical statistical theory, i.e. the goodness of fit test and tests of residuals.

1. Goodness of Jit A goodness of fit test compares the amount of variability between the differences of predicted and experimental values, i.e. the residual sum of squares, with the amount of variability in the data itself. This comparison allows the model builder to determine whether the overall model is adequate. If the model being considered is correct, the residual for the ith data point using the least-squares estimates 8*, er(O*) = Y r - f (O*, x t ) , will be a measure of experimental error. A measure of the total amount of variation unaccounted for by the model is the residual sum of squares n

RSS

=

C

n

er2(O*)

i= 1

=

C

[Y~-~(O*,X~]~

i= 1

It is a direct result of the orthogonality property of linear least squares (Draper and Smith, 1966), that

f y: = f fye*,xz)+RSS

i=1

i =1

(16)

Equation (16)states that the total amount of variability in the data

a.

146

E. BLAU

and w.

BROUK NEELY

c y:, is equal to the total amount of n

called the crude sum of squures,

i- 1

variability which can be accounted for by the model, called the sum n

f 2 ( O * , xi), plus the residual sum of

of squares due to regression, i- 1

squares. Associated with each source of variation is a certain number of degrees of freedom, which is used to attribute more information to, say, 100 data points than to five data points. I n particular, if n data points are used, the crude sum of squares possesses n degrees of freedom. The predicted values estimated by the model with p parameters have p degrees of freedom while the remaining n - p degrees of freedom are possessed by the residual sum of squares. If several data points have been taken at the same settings of the independent variables, then a measure of the inherent error in the data is given by the pure-error sum of squares

where are n, repeat observations of x1 y21,y2,, ..., y z n, are n2 repeat observations of x, Yll

. .

. . . .

ykl, Y k z , ..., Y k n nare nk repeat observations a t

Xk

and gj = (yjl+ yj2+ ... +yj,,)/nj is the average of all the repeated or replicated points of xj. Since the residual sum of squares measures the amount of variability as seen by the model, and the pure error sum of squares is a true measure of error in the data, it follows that the inability of the model to fit the data is given by the difference of these two quantities which is appropriately called the lack-of-jit sum of squares

For simplicity assume there are r replications at q different settings of the independent variables, then the pure error sum of squares possesses q(r - 1) degrees of freedom (one degree of freedom being used to estimate g j ) ; while the lack-of-fit sum of squares possess n - p - k(r - 1) degrees of freedom, which is the difference between the degrees of freedom of the residual sum of squares and the pure-error sum of squares. The quotients obtained when the sum of squares discussed above are

MATHEMATICAL MODELS AND INSECTICIDE DISTRIBUTION

147

divided by their degrees of freedom are called mean squares. The pure error mean square j=lu=l

is a measure of experimental error independent of the validity of the model employed. Therefore, a test of whether a model is adequate can be made by determining the ratio of the lack-of-fit mean square, n

k

nl

to the pure error mean square. If the ratio is large, it suggests the model inadequately fits the data. Using the F statistic to quantify the magnitude of this ratio, the test of inadequacy is usually written

where (1 - N ) 100 is the confidence level in percent for rejecting the hypothesis that the model is adequate. The F statistic is tabulated in almost every statistics reference text. If an independent estimate of pure error is available, say s2 with u degrees of freedom, then the test for adequacy of the model simply becomes the ratio of the residual mean square to this measure of pure error. That is, the model is said to be inadequate if

at the (1 - a) 100 percent confidence level.

2. Residual analysis The goodness of fit test provides information about the overall ability of the model to fit the data. It can also be used to test the importance or contribution of certain terms in the model towards providing the overall fit of the data. However, these methods do not identify the specific limitations of the model. I n particular, even though the overall goodness of fit is quite acceptable, more subtle model inadequacies may exist. These inadequacies can often be detected through an analysis of the residuals of the model. As defined by Eqn (4), a residual is the difference between the observed and predicted values of the dependent variable. If the model is correct, the residual for any point is solely attributable to experimental error. Therefore, plots of this residual versus any independent

148

o.

E. BLAU

and w.

BROCK NEELY

variable should exhibit all the characteristics of this error, such as being random with zero mean. However, if the model is inadequate, the residual will not be random and possibly biased above or below zero when plotted against some independent variable. Several methods have been suggested for preparing the most revealing residual plots (Kittrell, 1970; Draper and Smith, 1966). Consider the following three typical methods:

(a).Predicted value residual plots. A plot of the residual ei(8*)versus the predicted valuef(B*, xi) can indicate whether the model truly represents the data. For example, residuals that are generally negative a t low predicted values and positive a t high predicted values indicate a model inadequacy even though it may have passed the goodness of fit test. These plots can also provide information about the assumption of constant error variance made in the maximum likelihood approach. If the residuals continually increase or decrease in such plots, a nonconstant error variance is indicated and either a weighted least squares analysis should be conducted (Kittrell, 1970) or a transformation must be found to stabilize the variance (Box and Cox, 1964). (b). Independent variable residual plots. By plotting the residuals versus the independent variable values, it is possible to identify which of the variables in the model is causing the residual trends that occur in the predicted value residual plots. The nonconstant error variance described above also is exhibited in these plots and can provide useful information for developing a weighting function.

(c).Overall residual plots. If one plots the frequency of occurrence of the rounded values of the residual against the magnitude of the residual, it is possible to assess the normality of the error if the model is correct. Also these plots test the assumption made earlier that the mean of the error distribution is zero. Basically this plot allows an approximate check on the assumptions made in the development of the least squares analysis from the theory of maximum likelihood. H.

CONCLUSION

The preceding sections have presented a methodology for building a mathematical model of some physical system from experimental data collected on the system. Again, it is important to re-emphasize that the best approach to model building is by carrying out the experimentation and analysis programs iteratively. Nothing is more frustrating than trying to obtain information about a system after the experimentation program has been terminated and the existing data are inadequate.

MATHEMATICAL MODELS AND INSECTICIDE DISTRIBUTION

149

Another important point is the importance of properly designed experiments. Certain statistical assumptions relative to the distribution of the experimental errors are inherent in applying the statistical techniques to analyse the data collected. Proper experimental design will provide some indication of the validity of these assumptions. If the assumptions are invalid the data can be transformed and this transformed data can be analysed. The importance of knowing this distribution of the experimental error or error structure cannot be overemphasized. Although it may require more experimental measurements, the probability of building a meaningless or overly sophisticated mathematical model will be minimized.

111. T H E E N V I R O N M E N TFATE A L A N D DISTRIBUTION O F D U R S B A N @ A D D E D T O A N ECOSYSTEM A.

INTRODUCTION

An important environmental problem is the determination of the ultimate fate and distribution of a chemical introduced into an ecosystem. Numerous phenomena take place simultaneously in such a situation. Hence, a true mathematical model describing each step of the process would be extremely complex. It is important, however, to try and find a suitable model to identify the most important chemical, physical and biological phenomena taking place and to predict the long-term environmental consequences. The example chosen for study in this paper concerns the addition of a chemical agent to a laboratory system which simulates a pond of water. Some of the phenomena that need to be included are the distribution and partitioning of the agent between the water and soil that may be present. I n addition to these, consideration must also be given to the absorption, metabolism and excretion of the agent by the various aquatic species.

B.

DESCRIPTION O F T H E ECOSYSTEM

Smith et al. (1966) published some studies on the distribution and fate of a new agent for the control of insects, D U R S B A N ~insecticide. The active ingredient of D U R S B A N ~ ,O,O-Diethyl0-(3,6,6-trichloro-2-pyridyl) phosphorothioate, was labelled with radioactive carbon 14C in the pyridyl ring and added at a level of 1 mg/6 gal in a 10-gallonglass jar. This aquarium contained 2 in. of soil (13.3% organic matter), plants (salvinia, anacharis, milfoil and water cucumber) and 46 goldfish. Tp

150

o.

E. BLAU

and w. BROCK

NEELY

Samples of the various components were analyzed for radioactivity at different time periods after addition of the DURSBAN@. A summary of this data taken from the paper of Smith et al. (1966) is presented in Table I and plotted in Fig. 3. The experimental setup described above was disassembled before this model building program was initiated so that additional experimentation was impossible. Therefore, knowledge of the underlying error structure must be based on existing replicate analysis and subjective interpretation of the experimentalist. From independent measurements TABLEI

Distribution of

Time after

14C DURSBAN@

in the eooeystem

Percent radioactivity in the three components of the ecosystem

D U R S B A N ~addition

(h)

Fish

0 1.5 3.0 4.0 6.0 8.0 10.0 24.0 48.0 72.0 96.0 120.0

0 15.2 19.0 19.3 20.7 23.0 24.2 21.2 23-0

22.7 20.5 17.3

Soil and plants 0 35.2 46.0 56.0 61.0 60.5 59.3 51.5 38.3 38.3 36.3 38.3

Water

100 49.7 28.3 24.5 18.3

17.0 18.2 26.5 34.5 39.5 43.0 44.5

made in the system but not reported in Table I, it may be concluded that: 1. Measurements of

14Cin the three components are independent of the different components. 2. Measurements of 14Cfor any one component are independent of other measurements of that component. 3. The measurement errors are approximately the same for each component.

If one assumes that the errors are normally distributed with zero means and constant variance for each of the components, then the single response likelihood analysis of Section I1 E is readily extended to this multiresponse case (Kittrell, 1970). Here, for example, the residual sum

MATHEMATICAL MODELS AND INSECTICIDE DISTRIBUTION

151

of squares defined by Eqn (16) is the sum of the residual sum of squares for the three components. This total residual sum of squares can be used to determine the lack of fit sum of squares. It will also be informative to analyze the residuals of the individual components. Such an analysis provides valuable insights into particular limitations of the model. The data in Table I have been transformed into percentages from the crude radioactivity measurements. Although the error structure defined in the preceding paragraphs is also transformed, the variability in the

Od

O:

40

$0

e'0

IbO

ILO

14

Hours after 14C dursban addition

FIG.3. Distribution of

14C

Dursban in the ecosystem.

original data is so small that the effects of this transformation are minimal. It will be assumed, therefore, that error properties 1-3 are still valid and that for each component, the error is normally distributed with zero means and a constant standard deviation of 1%.

c. B U I L D I N G

THE MODEL

The simplest model which can be postulated to explain the data of Table I is to assume (i) that an equilibrium exists between the chemical

152

GI. E. BLAU

and w.

BROCK NEELY

in the water and the soil and plant constituents, and (5)a direct uptake of the chemical by the fish. This model can be represented symbolically a8 follows:

A ~ B Model 1

41 " C A B C

where

= 14C in = 14C in = 1% in

the water the soil and plants the fish

and k,, k,, k3 are reaction rate constants in h-l. It is further assumed that all the steps or reactions are first order. Mathematically, therefore, the model is represented by the following differential equation system

with initial conditions z ~ ( 0 = ) 100, z ~ ( 0 = ) 0 and zc(0) = 0.Here, z ~ ( t )z, ~ ( tand ) zc(t) are the percentages at time t of A , B and C respectively with the restriction that XA(t)+2B(t)+XC(t)

(22)

= 100

Using a nonlinear parameter estimation program, it is possible to find the parameter values kr = 0.510, k i = 0.800 and k: = 0.00930 which best describe the data of Table I. Corresponding to these parameters, the overall residual sum of squares is RSS = 5374. Residuals for each of the three components measured can be calculated. Since an independent estimate of error is available, i.e. s2 = 1 for all three measurements, the lack of fit relation 19 can be applied directly with the numerator degrees of freedom n - p = 36- 3 = 33 to give R8s'(n-p) 3.82

- 5373133 = 54.3 > F,.,,(33,20) 3 ~

= 1-44

Since this ratio is considerably greater than the tabulated F value, the model is totally inadequate. By a residual analysis it might be possible to identify the specific inadequacies in the model. Figure 4 is a plot of the residuals for each of the measured components versus the independent variable time. This residual plot reveals the following

MATHEMATICAL YODELS AND INSECTICIDE DISTRIBUTION

163

discrepancies: (i) initially the model predicts a higher proportion of 14C in the water and a lower proportion in the fish, (ii) after 72 h the model predicts the opposite of (i), and (iii) the model predicts low proportions of 14C in the soil and plants throughout the experiment.

E

I 40

I 60

0

1

I

80

100

- - 10

p, , , , 1 P

A 100

120

FIG.4. Independent variable residual plot for Model 1.

The next step in the modelling process is to use this residual analysis to postulate a better model. The large negative residuals observed in predicting the 14C proportion in the fish after 80 h, confirm a major limitation of Model 1. That is, Model 1 predicts an ever increasing proportion of 14Cin the fish. To compensate for this trend, Model 2 is

154

a.

E. BLAU

and w.

BROOK NEELY

postulated where the chemical in the fish is excreted from the fish either (a) unchanged k,

A 8 B

or (b) metabolized and excreted as a new entity A’ kz

Model 2b

The results of fitting the differential equations corresponding to these models to the experimental data is presented in Table 11. First, note that both forms of Model 2 indicate a lack of fit so that additional modifications will be necessary. Secondly, the two forms of Model 2 can be compared with themselves and Model 1 by calculating likelihood ratios. The likelihoods for the different models are shown in Table 111. Obviously, Model 2b is superior to both Models 1 and 2a, indicating that the existence of the entity A’ is highly probable. Residual plots for both forms of Model 2 are shown in Figs 5 and 6 respectively. Relative to the other components, the residuals for the 14C in the fish are reasonable, although considerably larger than expected from experimental variations alone. However, the residuals for the best-to-date Model 2b indicate that the Model predicts higher proportions of 14Cin the soil and plants than indicated by the data during the first 60 h, and lower proportions during the last 40 h. Analogously, the predicted water proportions during the first 60 h are lower than indicated by the data and higher during the last 40 h. I n order to bring the proportions between soil and plants and water into agreement, the next step is to give the entity A’ access to the soil and plants. Thus two forms of a new Model 3 were examined. The first is a simple uptake of A’ by the soil and plants

A

~

B

ks lk’

C

Model 3a

lk,

k

A’+ B’ while the second postulates an equilibrium relationship

(25)

TABLEI1 Parameter estimation and lack of fit analysis

Model number

(A

Number of parameters

(P)

Optimal parameter estimates (k,*) __

Residual sum of squares (RSSj)

Lack of fit mean square (LOFMS)

LOFMS Error

1

3

kl = 0.510 k2 = 0.800

k , = 0.0093

5374

162.8

54.3

2a

4

k1 = 0.493 k2 = 0.299

k, = 0.150 k4 = 0.187

1964

61.4

20.5

4

kl = 0.286 k, = 0.0277

k , = 0.111 k, = 0-0206

848

3a

5

k1 = 0.337 k, = 0.069 k, = 0.104

k, = 0.0273

208.3

k, = 0.00833

k1 = 0.338 k2 = 0.069 k, = 0.104

k, = 0.0275 k , = 0.00927 k , = 0.00235

207.9

k1 = 0.338 k2 = 0.0515 k , = 0.136 k, = 0.0788

k, = 0.00679 k, = 0.0670 k, = 0.0254

58.6

k, = 0.336 k2 = 0.0572 k, = 0.124

k4 = 0.0521 k, = 0.00697 k . = 0.0181

6

U

M F

0)

2b

3b

2.02

26.9

8.97

2.02

2.24

2.03

*4

U

6.72

k! M

6.93

2.31

2.04

8

9 ti M Y

4n

4b

7

6

U

2-02

0.673

2.05

Ge d

i

79.4

2.64

0.880

2.04

2

a.

166

E. BLAU

and w.

BROOK NEELY

A :B 4

C lk'

Model 3b b

A'#B' n;

I n these models, a new entity B' distinct from B is assumed. The results of fitting these models to the data and the calculated likelihoods are shown in Tables I1 and 111respectively. These models exhibit a lack of fit of the data. However, the likelihood ratios L,alL,b and LgblLgb show a marked improvement by Model 3 over Model 2 in fitting the data. Since the likelihood ratio L,aIL,b is approximately unity, it is impossible to discriminate between the two forms of Model 3. I n other words, the reverse readion B'-+A' does not improve the ability of the TABLEI11 Likelihood analysis

(j)

Residual sum of squares (RSS,)

Maximum likelihood for model j (L,*)

Likelihood ratio (LZIaIL,*)

1 2a 2b 3a 3b 4a 4b

5374 1964 848 208.3 207.9 58.6 79.4

1.44 x 0.95 x 4-16x 8.37 x 8.95 x 5-73x 1.79 x

3-98 x 1 0 5 8 6 8.24 x 10137 1-37 x 1067 6.84 x 101O 6.40 x 1010 1 32.0

Model number

10-88*

10-198 10-62 10-ls 10-10 10-6

lo-''

model to explain the data so that k, = 0. The residuals for Model 3a are plotted in Fig. 7. A comparison of Figs 6 and 7 shows the striking improvement in predictability of Model 3 over Model 2. These residuals show that the consequences of bringing the proportions of chemical in the water and soil and plants into better agreement have decreased the ability to predict the proportions in the fish. Further, it appears that a low prediction of chemical in the fish is accompanied by a high prediction of chemical in the water. It may be possible to improve the distribution of chemical between the fish and water by postulating that (a) the chemical in the fish partitions into a second compartment (e.g. the flesh), or (b) the entity A' in Model 3a is in equilibrium with the fish. Modifying Model 3a to

MATHEMATICAL MODELS AND INSECTICIDE DISTRIBUTION

au

!?$

Y Y

-2-4-

20

40

60

I

I

I

a

157

80

=I -6-

-

-8

-10

t

-

52

lo:

8

6-•

4z

4-

0; nu

2-

=O

ar za z

I

I

I

I

-4 -6-

=a 0

0

10

-t

-

0

8-

I

I

I

J

include it second compartment in the fish gives

A :B B lks

c :c' lk7

Model 4a

4

A'+ B' ks

where C' ie the proportion of 14Cin the second compartment. Model 4b is obtained simply by making the step C+A' reversible

a.

158

and w.

E. BLAU

BROUK NEELY

A :B ka lka

C

Model 4b

kllt.Z.8

A'+ B' k'

These models were fitted to the data and the results are presented in Tables I1 and 111. Both of these models adequately describe the data according to the lack of fit criterion. The residual plots shown in Figs 8 and 9 do not reveal any major discrepancies in the chemical distributions among the major components, although the residuals are some10

-

I 20

40

I 20

I 40

I

60

P 80I

I 60

I 80

I @

100

?

120

L" - 6 -8

-

-10

L

-

8

nu a 2 -6-

::

FIU.6.

TI

100

120

MATHEMATICAL MODELS AND INSECTICIDE DISTRIBUTION

159

what larger for Model 4b than for Model 4a. Non-parametric statistical tests indicate that the residuals are indeed plausible estimates of normally distributed experimental error (Draper and Smith, 1966). Based on residual analysis alone, either form of Model 4 is valid and further refinement of the model to better explain the data is not

52 sz

a0 0;

64-

2-

t

-%L

-8 -10

FIG.7. Independent variable residual plot for Model 3a.

warranted. The likelihood ratio L,,/L.,b = 32.0 implies a preference for Model 4a over 4b. This is equivalent to saying that there is strong evidence that a second compartment is set up in the fish. Since the value is less than 100, however, it is difficult totally to reject Model 4b without additional experimental work.

160

and w. BROCK NEELY

0. E. BLAU 10

t

-

-

6

a#

0:, 2 2 2 4c

ZY iE ar

42

73

0 - 2 - 3

1

20

fj

I 40

I

Q

60

0

60 TIME

d

I

m i

80

100

I 80

"100

&

-4-6-

-8 -10

-

-

10

0 -2

-4

-t -8

t

I -

5J

20-

I

40

I

6

t

0 120

-10L

FIQ.8. Independent variable residual plot for Model 4a.

D.

D I S C U S S I O N OF R E S U L T S

The final model that emerges from this analysis is the following: 1. There is a rapid equilibration between the applied DURSBANBand the soil and plant system. This step was also seen in the work reported by Smith et al. (1966). 2. This is followed by a slower uptake of the insecticide by the fish. 3. Once in the fish the material is metabolized and excreted. The metabolite is probably the pyridinol which was identified in the water a t the termination of the 120 h exposure (Smith et al., 1966). 4. The liberated pyridinol in the water is again taken up by the soil and plants.

MATHEMATICAL MODELS AND INSECTICIDE DISTRIBUTION io

t

-

161

-

6 4 I

I

20

40

6

I

60

P I80

,.

"100

=I -6-8 -10

-

42 sz

f

-

8 6 '4-

::I -10

FIQ.9. Independent variable residual plot for Model 4b.

5. The best fit was obtained with Model 4a which includes a partitioning of the material between two compartments in the fish. Again the data obtained by Smith would tend to substantiate this step in that these authors demonstrated a partitioning between the viscera and the meat. 6. Finally, Models 4a, 4b and 3a all indicate that the final sink for the added Dursban is the soil and plants. This last item is very important is since Smith (1966)has shown that the 3,5,6-trichloro-2-pyridinol metabolized readily by plants and will ultimately be degraded to CO,, NH, and H,O. Such a situation would imply that there is no persistance of Dursban in this particular ecosystem.

162

a.

E. BLAU

end w.

BROCK NEELY

The fast initial absorption of the insecticide by the soil and plants has an added advantage in that this particular sink acts as a reservoir for the slow release of Dursban. This feature gives added long-term protection for the control of mosquito larvae in polluted waters. Schaeffer and Dupraa (1970) demonstrated that a similar series of events occurred in a field trial.

E.

CONCLUSION

The model building exercise in this paper has generated a picture of the distribution pattern of DURSBAN@ when added to a pond of water. Furthermore, the picture that emerges is compatible with what is known about the insecticide.

REFERENCES Bard, Y. and Lapidus, L. (1968).Kinetic analysis by digital parameter estimation. Catalysis Reviewa, 2 (l),67-112. Barnard, G.H., Jenkins, G. M. and Winsten, C. B. (1962).The likelihood inference and time series. JI R. statist. SOC.Ser. A, 125. Bayes, T. (1763).An essay towards solving a problem in the doctrine of chances. The Philoeophical Tramactiona 58. Reprinted in Biometrika 45 (1958). Blau, G. E., Klimpel, R. R. and Steiner, E. C. (1970). Equilibrium constant estimation by nonlinear optimization. I d . Eng. Chern. Fundam. 9, 334-339. Blau, G. E., Klimpel, R. R. and Steiner, E. C. (1972a).Equilibrium constant estimation and model distinguishability. Ind. Engng Chem. Fundam. 11, 372-3. Blau, G. E., Klimpel, R. R. and Steiner, E. C. (1972b).Parameter estimation and model distinguishability of physicochemical models a t chemical equilibrium. Can. J. chem. Engng 50, 324-332. Box, G. E. P. and Cox, D. R. (1964).An analysis of transformation. JZ R. stat&. SOC.Ser. B, 26 (2),211-252. Box, G.E. P. and Hill, W. J. (1967).Discrimination among mechanistic models. Technomet&, 9 (l),57-71. Draper, N. R. and Smith, H. (1966).“Applied regression analysis.” John Wiley and Sons, New York. Hsiang, T. and Reilly, P. M. (1971).A practical method for discriminatingamong mechanistic models. Can. J. chem. Engng 50, 865-871. Kittrell, J. R. (1970).Mathematical modelling of chemical reactors. Adw. chem. E w . 8, 97-183. Reilly, P . M. (1970).Statistical methods in model discrimination. Can. J. chem. Engng. 48, 168-173. Rosenbrock,M.M. and Storey, C. (1966).“Computational Techniques for Chemical Engineers.” Pergamon Press, Oxford. Schffer, C. H. and Dupras, E. F. Jr. (1970).Factors affecting the stability of Dursban in polluted waters. J. econ. Ent. 68, 701-705. Smith, G. N. (1966).Basic studies on Dursban insecticide. Down to Earth 22, 3-7.

MATHEMATICAL MODELS AND INSECTICIDE DISTRIBUTION

163

Smith, G. N., Watson, B. S. and Fisher, F. S. (1966). The metabolism of [CV] 0, 0-diethyl 0- (3, 4, 6-trichloro-2-pyridyl) phosphorthioate ( D U R S B A N ~ ) in fish.J . econ. Ent. 50, 1464-1475. Solberg, J. J. (1972). Principles of system modelling. Proc. Int. Symp. Systems Engng and Analysis, 67-74.

This Page Intentionally Left Blank

The Pressure Chamber as an Instrument for Ecological Research GARY A. R I T C H I E

Weyerhaeuser Company, Tacoma, W N 98401, U.S.A.

and THOMAS M. H I N C K L E Y

Xchool of Forestry, University of Missouri, Columbia, MO 66201, U.S.A. _ _ _ _ _ _ _ I. Introduction A. Plant Water Status B. A Brief Historical Perspective . C. Objectives 11. TheoryandMethodology A. Theoretical Considerations and Terminology B. ApparatusC. Procedures D. CalibrationE. Precautions F. Measurements of P on Conifer Needles G. Use of the Pressure Chamber to Determine Osmotic and Potentials . H. WheretoSemple III. Review of Ecological Studies A. Some Physical Relationships B. Plant Responses to Supply and Demand C. Expressionandhterpretation of Data D. Pin Relation t o Habitat E . P in Relation t o Some Plant Factors IV. Other Applications of the Pressure Chamber . A. Pathology, Entomology, Pollution Effects . B. LeafFolding inLegumes C. Water Relations of Roots D. Frost Hardiness. E. CulturalApplications F. Other Applications V. Some Unresolved Questions A. Why Does P Fail t o Meet the Gravitational Potential Gradient? B. Why is 0 B m N e v e r Achieved?~

~

-

-

~

-

-

-

~

~

~

-

~

~

~

~

~

~

-

~

-

~

~

105

-

166 166 167 169 169 169 171 173 174 183 192

193 196 200 200 202 206 210 218 229 229 230 231 232 232 233 234 234 236

166

GARY A. RITCHIE

and

THOMAS M. HINCKLEY

C. Is There Substantial Resistance to Flow Between Leaf and Stem? D. Do Plants at Night Act aa Tensiometers? VI. Concluding Statement . AcknowledgementsReferences . . . . -

~

-

~

~

~

~

~

. -

-

.

236 238 240 243 443

I. I N T B O D U C T I O N A.

PLANT WATER STATUS

Due to its unique hydrogen-bonding, dipole-dipole structure, water has life-sustaining properties not present in any other known chemical compound of similar molecular weight. All forms of terrestrial life are dependent upon their ability to extract water from their environment and to hold it above certain free energy levels within the cells in order that life processes be sustained at rates commensurate with survival. Plants are immobile and unable to escape the demands of their immediate environment. They must procure copious quantities of water from tightly held reserves in the soil and at the same time retain it within the tissue despite the enormous evaporative power of the atmosphere and the necessity for maintaining above-ground air-water interfaces for the exchange of carbon dioxide and oxygen. A qualitative model of plant water status may be outlined as a simple mass budget equation:

W

(G-E)+N (1) where W is the water status of the plant (defined and discussed in Section I1 A), 0 represents the water gained by absorption principally through the roots, E represents the water lost (transpired primarily through the leaves) and H is the water stored within the plant. Water status is generally proportional to the difference between water gained and water lost since storage is probably of secondary consequence in most herbaceous plants, although its importance in some xeric species is well documented (Oppenheimer, 1960). During daylight hours, E normally exceeds G, resulting in a progressive decrease in W , termed “water deficit” or “water stress” (Kramer, 1938; Weatherley, 1963; Cowan and Milthorpe, 1968), until stressinduced stomatal closure retards transpiration. At night, stomatal closure and reduced atmospheric demand for water retard E and thereby allow residual free energy gradients within the plant to equilibrate. Thus on a daily basis, G approximates to E if soil water is not limiting; however, soil water frequently is limiting and plants are almost always under some degree of water stress. The level of water stress in plants profoundly influences virtually all

THE PRESSURE CHAMBER I N ECOLOGICAL RESEARCH

167

physiological and metabolic functions (see, for example, Slavik, 1965; Slatyer, 1967; Kozlowski, 1968a, b; Hsiao, 1973). This physiological constraint is substantially responsible for determining plant adaptation and distribution in nature. Thus the impact of plant water status on plant ecology becomes obvious as does the necessity for its quantification and interpretation.

B. A

BRIEF HISTORICAL PERSPECTIVE

The cohesion theory of Dixon and Joly ( 1 894) remains the only viable explanation for a tree’s ability to elevate water well above heights unattainable by a vacuum pump (Scholander et al., 1966a). However, this theory remained essentially untestable due to lack of a convenient means of measuring the hydrostatic tension of the xylem sap or transpiration stream (Zimmerman and Brown, 1971). It is ironic that Dixon himself had devised a method of making such a measurement as early as 1914. I n his effort to estimate the “osmotic pressure” of the leaf cells, Dixon (1914, p. 142) developed an apparatus consisting of a glass cylinder capped with metal at both ends, gasketed with “leather washers, soaked in bees’ wax and turpentine”, capable of resisting substantial internal pressures. To this chamber he attached a source of pressurized liquid carbon dioxide; pressure changes within the chamber were measured manometrically. He experimented with numerous species of plants by excising leafy twigs and inserting them into the cylinder with their cut ends protruding through a rubber tube at the mouth of the chamber and into a beaker of water. Dixon subjected these leafy branches to constant pressures of from 3 to 16 bars (1 bar = 0.987 atm = 14.5 p.s.i.) until the leaves wilted. By weighing the beaker of water he determined whether branches gained or lost sap during the course of each experiment. Unfortunately, his studies were fraught with difficulties including explosions of the glass cylinders “fortunately attended by delay in the work only”, so that the technique was ultimately abandoned. It apparently did not occur to Dixon, gradually to increase the pressure and note the point at which water was expressed from the cut end. Had he done so, he would have devised a method not only for testing his cohesion theory but also for assessing the water status of plants. Such an innovation was to await the ingenuity of P. F. Scholander, H. T. Hammel and their colleagues of the Scripps Institute of Oceanography. They designed a pressure chamber and associated apparatus of remarkable similarity to that of Dixon’s. Their application of the device, however, was somewhat different. Following the excision of a

168

GARY A. RITCME and THOMAS M. HINCKLEY

leafy branch and its insertion into the sealed chamber, they reasoned that “one should be able to measure the hydrostatic pressure which existed in the vascular syatem prior t o the (excision) simply by observing at which external gas pressure the system starts to yield liquid”. Scholander et al. (1964) reported the results of measurements on a number of species including xerophytes and halophytes in the Proceedings of the National Academy of Sciences. These experiments seemed to confirm for the first time that indeed xylem sap tension could be successfully measured. The significance of this paper went relatively unnoticed until a subsequent article appeared in Science (Scholander et aE., 1966a) presenting evidence indicating that pressure chamber data were not only relevant to theoretical problems of water transport in plants but were also predictably related to plant water status and thus germane to studies of plant adaptation and distribution. A major contribution to the usefulness and acceptance of the pressure chamber arose out of some experiments at the Forest Research Laboratory at Oregon State University. Waring and Cleary (1967)reasoned that if the pressure chamber measures the hydrostatic tension in the xylem, and if the matric and osmotic forces therein are small, then the chamber value is, in effect, an estimate of total water potential (the relationship between the plant water status and plant water potential will be developed in Section 11). A portable pressure apparatus suitable for field use was designed and tested in situ on a number of tree species native to the southern Oregon Siskiyou Mountains. Waring and Cleary outlined procedural guidelines for attaining repeatability of measurements and demonstrated a predictable relationship between plant diatribution and intensity and duration of “plant moisture stress” as measured with the pressure chamber. Their studies established the pressure chamber as a valuable and viable ecological research tool and the designs and procedures outlined in their paper have been followed in countless subsequent ecological research efforts. As with any new technique, however, reports of errors or inconsistencies began to appear in the literature. Boyer (1967it) compared pressure chamber values with thermocouple psychrometer1 measurements of water potential and found poor agreement in some species. With yew (Tmw cuspidata) and sunflower (Helianthus annuus) the differences did not exceed & 2 bars, but with Rhododendron roseum the differences approached 4 bars under some conditions. Kaufmann (1968a) found even more serious discrepancies with red oak (Quercus 1 Thermocouple psychrometer values are considereda standard in water potential measurement (Boyer and Knipling, 1966; Boyer, 1966).

THE PRESSURE CHAMBER I N ECOLOGICAL RESEARCH

169

rubra) and white oak (Q. alba) wherein values differed by 16 bars at low water potentials. Ritchie and Hinckley (1971) found that with some species of conifers (PsewZotsuga menziesii, Abies procera, A . amabilis) pressure chamber values measured on individual needles differed from those measured on adjacent twigs by as much aa 4 bars. The reasons for these inconsistencies will be discussed in subsequent sections. Nevertheless, the pressure chamber technique has found extremely wide use in studies of plant eco-physiology. Research haa dealt not only with plant distribution and adaptation but also with such areas as plant pathology, productivity, growth, frost-hardiness, fertilization and irrigation, to name but a few. The pressure chamber technique has become the standard method for assessing plant water status in the field.

c. O B J E C T I V E S To date no complete review of this technique h a appeared in the literature, although certain aspects of its use and interpretation have been mentioned in various articles. We now present such a review with occasional inclusions of unpublished data, in an attempt at synthesis and evaluation. The review has five somewhat broad objectives. They are (1)to provide workers with a thorough and useful discussion of the effective use of the pressure chamber technique, (2) to offer a uniform terminology, (3) to assemble, interpret and evaluate published ecological studies wherein the pressure chamber has been employed, (4) to suggest areas of research where the technique is potentially useful and (5) to discuss some unknowns and problems presently associated with the pressure chamber.

11. T H E O R YA N D METHODOLOQY A.

THEORETICAL CONSIDERATIONS AND TERMINOLOQY

A satisfactory system for quantifying and expressing plant water status was not devised until Slatyer and Taylor (1960) and Taylor and Slatyer (1961, 1962) unified earlier concepts within a thermodynamic framework and provided plant scientists with a consistent and theoretically sound terminology. The reader is referred to the above papers as well as to the works of Spanner (1964, esp. Chaps. 6, 7, 13 and 15), Slatyer (1967), Dainty (1969) and Lange (1972) for discussions of the thermodynamic basis of plant water relationships. Water status will be functionally defined as equivalent to the water potential ($) in the plant system which is equal to ( , ~ ~ - p ~ ~ ) / ~

170

GARY A. RITCHIE

and

THOMAS M. HINCKLEY

where (pw- pow)is a measure of the capacity of water a t a point in the system to do work with respect to pure water, and Vw is the partial molar volume of water (Slatyer and Taylor, 1960). The water potential of pure free water has a value of 0 bars. The water potential of xylem sap ( $ w ) under isothermal conditions is influenced by several factors:

where - p g h is gravitational potential (product of density of water, acceleration due to gravity and the height above standard pure water at 1 bar pressure and standard temperature, and is equal to about

- 0.1 bar m-1

height);

[:

1

x f i ~ t is the

frictional potential (Richter et al.,

1972; Richter, 1972), which represents the sum of products from partial fluxes (fg) and partial resistances (rz) along the branched xylem conduit from soil (S)to point (P)in the plant; T is the solute potential; and T is the matric or surface potential. Scholander et al. (1965b) asserted and Boyer (1967a) demonstrated that the pressure chamber measures only the gravitational and frictional potentials. Thus a pressure chamber value is related to xylem water potential by a modification of Eqn (2): *w = p- * y e m (3) where P is the combined gravitational and frictional potentials and is the combined solute and matric potentials (Boyer, 1969). This component is not measured by the pressure chamber but its value in the xylem sap of most plants is regarded as either constant or negligible (Boyer, 1969; Duniway, 1971a), so that: *w

z P

(4)

Therefore, pressure chamber determinations are estimates of the total water potential of the xylem sap. The relationship between P and leaf water potential (+t)has been the subject of considerable research, which will be reviewed in detail in sub-section D. Many different terms have been used to express pressure chamber values. Waring and Cleary (1967) and Love and West (1972) used “plant moisture stress”, which was designated as a positive value in atmospheres of pressure. Klepper and Ceccato (1968), Begg and Turner (1970) and Sucoff (1972) prefer “water potential”, apparently accepting the fact that the total water potential is not actually measured. Other terms such aa “xylem pressure potential” (Kaufmann, 1968a), “sap stress” (DeRoo, 1969b), “xylem sap pressure” (Turner

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

171

and Graniti, 1969), “internal moisture deficit” (Pierpoint, 1967), “negative hydrostatic pressure” (Waggoner and Turner, 1971) and “Saugspannung” or suction-tension (Richter and Rottenburg, 1971; Halbwachs, 1970, 1971) have been offered. We recommend establishment of consistency in terminology and units and suggest that the designation “P”, with appropriate prefixes and subscripts, be used to designate pressure chamber values, and that P be expressed (and graphed) as a negative function in units of bars (Boyer, 1967a, 1968, 1969; Boyer and Ghorashy, 1971; Ritchie and Hinckley, 1971; Duniway, 1971a; Tyree et al., 1973a). Thus (Pstem) designates measurements with a pressure chamber on a cut stem, and (P,,,f)designates measurements made on a detached leaf. This terminology will be used throughout the present review, as will the expression “xylem pressure potential”, which is thermodynamically correct. The advantages of this approach are (1) it is simple, consistent and flexible, (2) it is compatible with thermodynamic concepts and (3) it is consistent with Eqn (3), thereby expressing the actual quantity measured.

B. A P P A R A T U S The pressure chamber concept was introduced by Scholander et al. (1964, 1965a) but the apparatus itself was only briefly discussed. Waring and Cleary (1967) were the first to present details on the assembly of the device and its associated components. Turner et al. (1971) have given a detailed description of the instrument as well as a list of sources for purchasing accessories necessary for its construction. Figure 1 is a, schematic diagram of the pressure chamber system (modified from Waring and Cleary, 1967, and Turner et al., 1971). A tank of compressed nitrogen (A) provides the source of gas pressure. A desirable modification for field use is a smaller cylinder (B), which can be filled from a large stationary cylinder and then easily carried to the sampling sites. The pressure gauges (C and D) associated with the spring-loaded safety regulator (E) record the pressure of nitrogen in the cylinder and delivery pressure respectively. A metering valve (F) regulates the rate of pressure increase in the chamber (I) and a gauge (H) monitors the chamber pressure. Some systems utilize two such gauges, one giving high sensitivity over narrow pressure ranges and the other responding to a broader pressure range with commensurate reduction in sensitivity (e.g. Turner et al., 1971). A bleed-off valve (J) allows nitrogen to be purged rapidly from the system following a determination.

172

GARY A. RITCHIE

LB

and

THOMAS M. HINCKLEY

E

F

FIG. 1. Schematic diagram of a pressure chamber system showing (A) reservoir tank of nitrogen gas, (B) portable nitrogen tank, (C, D, H) pressure gauges, (E)safety regulator, (F)metering valve, (G) shutoff valve, (I)pressure chamber, (J)bleed-off valve.

F

FIG.2. Diagram of 8 typical pressure chamber. Foliage sample (A) is inserted through rubber gland (B), which is placed in ohamber top (C) and sealed with silicone grease. Top is then affixed to chamber body (D). Pressurized nitrogen enters chamber through fitting (E) and leaves through fitting (F) on way to 1971.) bleed-offvalve. (Modified from Wiebe et a!.,

THE PRESSURE CHAMBER I N ECOLOGICAL RESEARCH

173

The pressure chamber itself is depicted in Fig. 2. It comprises a cylindrical metal chamber capable of withstanding 200 to 276 bars of internal pressure (D). The chamber is fitted with a screw-on or bayonetmounted top (C). A rubber stopper (B), its center perforated with a hole of appropriate diameter, is fitted tightly into the top and sealed with silicone lubricant (hydrocarbon lubricants should not be used). Pressurized nitrogen gas enters the chamber through fitting (E) and is vented through fitting (F). A tight seal is provided by a rubber O-ring between the chamber top and b0dy.l Tobiessen (1969) desoribes a pressure chamber which can be homebuilt for under $60 and is capable of safely withstanding internal pressures of 100 bars. Turner et al. (1971), Ritchie and Hinckley (1971), Goode (1968), Johnson and Nielson (1969) and Gifford (1972) describe modifications which permit measurements on individual conifer needles, blades of grass, or single leaves.

c. P R O C E D U R E S The procedure for determining the xylem pressure potential (P) of a leaf or stem with the pressure chamber is relatively simple and has been outlined by many authors (Scholander et al., 1965a; Waring and Cleary, 1967; Barrs, 1968; Boyer, 1969; Zimmerman and Brown, 1971). A twig or leaf is excised from the specimen plant or, in the case of small seedlings, the entire plant may be decapitated above the root collar. If a conifer or hardwood stem is used, the phloem and bark may be peeled back far enough to allow for the insertion of the twig into the rubber gland. Silicone grease is often used on the gland to insure an adequate seal between stem and rubber. The chamber top is then mounted on the chamber body. With the bleed-off and metering valves securely closed, pressure from the storage cylinder is gradually applied to the chamber at a constant rate. A t the instant water appears at the cut end of the leaf petiole or twig, the balancing pressure is read from gauge (H). The metering valve is then closed and the system vented through the bleed-off valve. When small twigs, leaf petioles or conifer needles me used, it is useful to mount a dissecting scope atop the apparatus. With larger material, a l o x magnifier is usually adequate. Pressure chambers are manufactured commercially in the United States by

P.M.S.Instrument Company, Corvallis, Oregon 97330; in Great Britain by Chas. W. Cook and Sons, Ltd., Perry Bar, Birmingham; and in D;ermany by R.oth Geriitebau, D-8523 Baiersdorf, Blumenstrasse 5.

174

GARY

A. RITCHIE

D.

and

THOMAS M. HINCKLEY

CALIBRATION

Waring and Cleary (1967) were apparently the first to investigate the relationship between P and other measures of plant water status. They used a vapour equilibration technique (Slatyer, 1967) for estimating leaf water potential ($L) and compared these values against P from Douglas-fir (Pseudotsuga menziesii) foliage over a range of from - 5 to -20 bars. Agreement between the techniques was reported to be k 1 bar. This preceded a number of similar comparisons wherein P values were calibrated against leaf water potential as measured with the vapour equilibration technique (thermocouple psychrometer), the freezing point depression technique (Cary and Fisher, 1969, 1971) P (-bars) 20

10 0

c

m

k

cd

10

Q

20

FIG.3. Theoretical relationship between P and and - 2 bars (B).

$L

when

$:ylern

is negligible (A),

and the density method (Shardakov, 1948; Knipling, 1967). Here we will summarize these reports as well as reports on comparisons of P and other measures of water status.

1. P versus $L (psychrometer) Figure 3 shows the theoretical relationship between P and $L. I n line A the osmotic and matric components of $L in the transpiration stream (I@'"") are 0. With an inclusion ( - 2 bars) of solutes or surface (matric) effects in the xylem sap, and assuming its constant concentration throughout a range of from 0 to -20 bars water potential, the calibration curve would resemble line B. This situation is considered to exist in many plants (Boyer, 1969).

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

175

Calibrations of P versus $L, as measured with the thermocouple psychrometer, have been made for a number of species including woody plants and trees, herbaceous dicots, and monocots (Table I). In most of these calibration curves a consistent relationship is evident. At high water potentials, $L tends to be more negative than P,but as water potential decreases, P becomes more negative than +L. (For brevity this will be referred to as the “X” relationship because of the X-shape of the curves.) Location of the crossover point (where P = $L) varies greatly with species, being as high as - 3 bars in loblolly pine (Kaufmann, 1968a) and as low as -20 bars in yew (Boyer, 1967a). The predicted relationship shown in Fig. 3 (line B) is not often realized. Neglecting matric or osmotic forces would tend to produce $L values which are more negative than pressure chamber values. However, the opposite effect is apparent in most published curves. Two causes of error may be in operation: in the curves for northern red oak and white oak (Kaufmann, 1968a), giant sequoia (Sequoiadendron giganteum) (Tobiessen et al., 1971) and rhododendron (Boyer, 1967a), a rather consistent pattern is evident. At high water potentials there is relatively good agreement between P and $L, but as water potentials decrease this relationship is severely distorted. Boyer speculated that this would occur if voids or non-conducting xylary elements were dry before measurement. During application of pressure the filling of these voids would require excess pressure, resulting in erroneously low P values. When a leafy branch is excised from a plant, tension in the transpiration stream is relieved, transpiration rate increases and “water stress” (low &) results (Slatyer, 1967). Kaufmann (1968a, b) used this procedure with yellow poplar (Liriodendron tulipifera), white oak, northern red oak, and “Washington” Naval and “Valencia” orange foliage to induce water stress. Klepper and Ceccato (1968)used it again with pear, apricot and grape leaves, Barrs et al. (1970) with tomato (Lypersicon esculentum) and Boyer and Ghorashy (1971) with soybean (Glycine m a z ) . Their procedures were essentially similar. Leaf or branch specimens were removed from trees and allowed to desiccate. Throughout the desiccation period, P and $L were measured and compared. Inherent in this approach is the assumption that the P vs. +L relationship is the same before and after excision. Duniway (1971s) questioned this assumption on theoretical grounds and West and Gaff (197 1) presented experimental evidence that the assumed relationship is not valid, at least not with apple (Pyrus malus) leaves. They compared P and +L on leaves which were desiccated in situ on the tree and those which were excised and allowed to desiccate for 0, 10, 20 and 30 minutes. Their data are shown in Fig. 4. The imposition of water stress

c 4 Q,

TABLEI Summary of calibration ohta for peaawe chamber versua water potent* determined by themnocouple paychrometer ( T P ) ,density technique (D) and freezing point depression technique ( F D P )and the leaf thermocouple Hygrometer ( L T H ) Species Sequoia (SequoiadendrongGanteum) White pine (Pinuu stmbus) Loblolly pine (Pinua taeda) Ponderosa pine (Pinus ponderosa) Sitka spruce (Pice0 &hen&) White 6r (Abiea coneoh) “Washington” Naval orange “Washington” Naval orange Valencia orange Yellow poplar (Lirwdendron tdipiyera) Rhododendron ro~eum Grape (Vitia sp.) Pear (Pynu, wmmunak) Chihpsi.4 l i m r i e Lawea divaricata Encilia farinam Tobacco (Nicotiana tabaoum) Tomato (Lywperaioon euculentum) Sorghum (Sorghum bieolor) Yew ( T a m ~ l s p i d d a ) Radiata pine (Pinuu &&a) Pear Apricot Greasewood (Scarwbatuu vermiculatwr)

Description conifer conifer conifer conifer conifer conifer woody perennial-tree woody perennial-tree woody perennial-tree woody perennial-tree woody perennial-ahrub woody perennial-vine woody perennial woody deeert shrub woody desert shrub woody desert shrub annual dicot aMUal-yOWlg leaves annual monocot conifer conifer woody perennial-tree woody perennial woody desert halophyte

Degree Type Technique scatter* curve*** TP TP TP TP TP TP TP TP TP

TP TP** TP TP TP TP TP

TP TP TP

TP** TP TP

TP D

+++ ++ + + ++ + ++ + + + + +++ ++ + ++ ++ ++ + +++ + +++ +++ ++ ++

A A A A A A A A A A A A A A A A A A A B B B B B

Reference

Tobiesson et d. (1971) K a u f m a ~(1968s) ~~ Kaufmann (1968a) Barker (1973) Hellkvist et al. (1974) Barker (1973) Klepper and Gccato (1968) Kaufinann (1968b) Kad‘mann (1968b) Kaufmann (1968a) Boyer (1967a) Klepper and Ceccato (1988) Kaufmann @en.comm.) Oechel (pers. comm.) comm.) Oechel (pen. Oechel (pers. comm.) &Roo (1970) Barn et al. (1970) Blum et al. (1973) Boyer (19678) Rook (1973) Klepper and Ceccato (1988) Klepper and Ceccato (1968) Detling and Klikoff (1971)

Perennial aeepweed ( 8 d ufruticoeu) Sunfiower (Hdiccnthw annuus) Soybean (Mycine maz) Tomato (Lywperaicon eSculen$um) Tomato (Lywpersicon eaculentum) Pepper (CapaiMlm unnuum) Snap bean (Viciafdo) Cotton (Gosqpium h i r a u m ) Cowpea ( VGna 8inena-k) Sorghum (Sorghum b i c o h ) Sorghum (Sorghum vdgure) Wheat (Triticum wtiwum) Wheat (Triticum uestkmm) Wheat (Triticum aestivum)

woody desert halophyte annual dicot annual dicot annual dicot-old leaves annual dicot annual dicot annual dicot annual dicot annual dicot annual monocot annual monocot annual monocot-tilleringstage annual monocot-headingstage annual monocot

TP TP TP TP TP TP TP TP TP TP LTH

CORl (ZeCC M y 8 ) Northern red oak (Qzcercus &a) White oak (Quercw dba) Engelmann spruce (Pice0 engdmunnii) Chrysanthemum morvolium Apple ( P y m &us) Pinto bean ( P h e o l u s vdguri.9) Sunflower (Heliccnthw annuw) Alfalfa (Medicago edivu) Sugar beet (Bcta vulgaria) Russet potato (Sokanum tubemsum)

annual monocot woody perennial-tree woody perennial conifer annual dicot woody perennial-tree annual dicot annual dicot annual dicot annual dicot annual dicot

TP TP TP TP TP TP FPD FPD FPD FPD FPD

* + indicates low degree of scatter in data points, + + + indicates high degree. ** corrected for #~"'""

***

me Fig. 6, p. 183.

D TP**

TP

+ + + + + + + + + ++ + + + + + + ++ ++ ++ ++ +++ +++ +++ +++ +++

B B B B B B B B B B B B B B B C C D E

F

-

-

-

Detling and Klikoff (1971) Boyer (19678) Boyer and Ghorashy (1971) Bans et d. (1970) Duniway (19718) Gee et d.(1973) h m (1972) Jordan (pers. comm.) Jordan (pers.comm.) DeRoo (1969b) Jordan (pers. comm.) Frank and Harris (1973) Frank and Harris (1973) Campbell and Campbell (1974) DeRoo (1969b) Keufmann (19688) Kaufmann (19688) Kaufmann (19688) Spomer and Langhans (1972) West and Gaff (1971) Gary and Fieher (1971) Cary and Fisher (1971) Cary and Fisher (1971) Cary and Fieher (1971) Cary and Fisher (1971)

H

B

8

m W

c:

E

F B

1 G M

d

0

!*3 E

F

W

td

6

178

amy

A. RITCHIE

and

THOMAS M. HINCKLEY

had a greater effect on P than on +L; consequently the relationship between the two terms was altered. West and Gaff explain the discrepancy on the basis of the difference when the xylem was filled with sap in the artificially versus naturally desiccated material. Upon excision, xylem sap is lost to evaporation through the cut end as well as to exchange with leaf mesophyll cells. When this stem is then subjected to a chamber measurement, more pressure is required to fill the dried xylem tissue than is required with a naturally desiccated leaf. Thus the P values are erroneously low and the P vs. # L relationship is modified. P (-bars) 48

32

16

0

FIU.4. The relationship between leaf water potential ( 4 ~and ) xylem pressure potential ( P ) in apple leaves. Line (A: 0 ) : leaves stresvcd on the tree and measured within 60 sec after leaf excision ( 1GW = - 9.02 + 0.709 P , r = 0.8900). Line (B): leaves measured after 10, 0 20 and * 30 min following excision (1GW = 6.63+ 0.912 P , r = 0.7069). The dashed line is the line of equal potential. (Reproduced with permission from West and Gaff, 1971.)

It is possible that the pattern evident in calibration curves generated in this manner is a procedural artifact. West and Gaff, however, only demonstrated this effect in apple. W. C. Oechel (pers. comm.) has kindly provided us with calibration curves for three desert species : Larrea divaricata, Encilia farinosa and Chilopsis linearis. Curves for L. divaricata and C. linearis are linear and nearly identical down to about - 50 bars. I n E . farinosa the relation is curvilinear. At high water potentials there is fairly close agreement between P and t,b~in all three species, but as water potentials become very negative (below -30 bars) P values are substantially more

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

179

negative than corresponding $L values. His procedure allowed for desiccation of tissue to occur in situ. Boyer and Ghorashy (1971) found good agreement between P and $L with soybean despite the fact that leaves were desiccated for up to an hour after excision to achieve low water potentials. With chrysanthemum, Spomer and Langhans (1972)found a steep depression in $L with decreasing water potential, showing a marked departure from the “X” relationship. Their measurements were made directly on leaves which were dehydrated in situ. They also measured $:yIem, which, when added to P, produced a different relationship, P being increasingly more negative than $L with decreasing water potential. No explanation was given. Duniway (1971a) compared P and $L in leaves of healthy and Pwarium-infected tomato plants where leaves were allowed to desiccate in situ. Close agreement between P and $L throughout a range of 0 to - 16 bars was observed. If the measured value of approximat,ely - 0.5 bars were added, agreement would be even closer. The “X” relationship is again apparent with tobacco (DeRoo, 1970), although natural desiccation of leaves occurred before measurement. In tomato (Barrs et al., 1970) the “X” relationship is seen with young leaves but not with old. I n these tests, desiccation occurred after excision. P was more negative than $ b in all cases. The data were not corrected for osmotic potential; had they been, z,hW values would have been even more negative. The authors suggest that the negative P values could have resulted from increased resistance to flow of water through the xylem during pressure increase due to compression of xylem tissues. Boyer (1967a) found good agreement between P and $L with sunflower leaves after he corrected for $;ylem. P values were slightly more negative than $L (0 to 1.5 bars) but the relationship was linear with very little scatter in data points. In calibrations carried out on three monocots, corn ( Z e a mays), sorghum (Sorghum bicolor) (DeRoo, 196913) and wheat (Triticum aestivum) (Lawlor, 1972), success was variable. I n corn the agreement between P and $L was very close. I n sorghum the “X” relationship was evident although desiccation occurred in situ. Correction for $:ylem (average -0.48 bars) would have again made P more negative than $L. Wheat gave a linear calibration with appreciable scatter. In summary, it seems that the error reported by West and Gaff (1971) might be more serious in woody than in herbaceous plants. This is by no means consistent, however, since the “X” relationship exists in some curves when desiccation occurred in situ (see, for example, Boyer and Ghorashy, 1971). Generally, P values are more negative

180

GARY A. RITCHIE

and

THOMAS M. HINCKLEY

than predicted by Eqn (3). Often the correction of P for #:Flern compounds this error (Barrs et al., 1970; Spomer and Langhans, 1972). The suggested explanations for the low measured values of P generally implicate voids in xylem tissue requiring increased chamber pressures, compression of xylem during pressure application causing increased flow resistance, the failure of the system to equilibrate during measurement, and the use of excision to induce water stress. More definitive work is clearly needed in this area.

2. P versus #L (density technique) There exist other techniques besides the psychrometer for estimating water potential of leaves (Table I). One of these is the dye or density technique described by Shardakov in 1948 and more recently in English by Knipling (1967). With this technique, leaves are immersed in a graded series of solutions of known water potentials. The leaf water potential ($L) is assumed to lie between the solutions in which leaf samples absorb water and those in which they lose water. Details of the method are given by Knipling (1967), who indicates that others have found agreement between the density method and psychrometer estimates of $L usually within three bars, but that differences as great as 5-8 bars have been reported in certain species. Detling and Klikoff (1971) compared the density method with the pressure chamber in estimating #L of two desert halophytes, greasewood (Sarcobatus vemiculatus) and perennial seepweed (Suueda frutiwsa), growing in their native habitat. I n each species the regression equations for the calibration lines were not significantly different (0.01 level) from the lines of equal potential. Detling and Klikoff stressed a belief that this does not necessarily suggest that both methods give accurate estimates of water potential. Rather, they listed known sources of error in both techniques and suggested that these might offset one another. Nevertheless, with these two species, comparable values might be expected from both methods over a wide range of water potentials.

3. P versw

#L

(freezing point depression technique)

Cary and Fisher (1971) compared P and #L as measured with a freezing point depression meter (Cary and Fisher, 1969) in five herbaceous dicots: sunflower, pinto bean (Phaseolus vulgaris), sugar beat (Beta vulgaris), alfalfa (Medicago sutiva) and potato (Solanurn tuberosum). All points fell on or below the equal potential line. The authors indicate that the relationship would have been 1 : 1 only if the plants were at equilibrium at the time of measurement and if the pressure chamber and freezing point meter both measured #. Clearly, at least

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

181

one of these assumptions was not met. The freezing point meter is believed to measure both the pressure and solute potentials of t,bm, while the pressure chamber measures only the pressure potential of zjw.Furthermore, the meter measures potential of water held in the cell walls and the intracellular spaces, whereas the chamber only measures potential of water in the xylem conduits. Both of these factors would cause freezing point meter values to be more negative than pressure chamber values, as is the case in these data. The authors expressed some difficulty in reconciling their measurements with other reported data showing close agreement between pressure chamber and psychrometer measurements.

4. P versus relative water content Hodges and Lorio (1971) sought a correlation between P and leaf relative water content ( R W C ) of loblolly pine needles. RWC was computed by the method of Weatherley (1950) where:

RWC

=

Fresh weight-dry weight x 100 Turgid weight-dry weight

(5)

P (-bars) 24

r

-

8

1F I

I

I

0 I

I

100

LOBLOLLY PINE

90

5 0

-

-

e,

P

0

-

80

%

FIG.5. The relationship between P and leaf relative water content (RWC) for loblolly pine needles. (Reproducedwith permission from Hodges and Lorio, 1971.)

Needles for R WC determination were taken from the same branch to be used in the pressure chamber determination. Data were collected from a single tree throughout the summer and are shown in Fig. 5 . Correlations were statistically significant at the 0.01 level (r = 0.91). Sankary and Barbour (1972a) made similar comparisons with the

a

182

GARY A. RITCHIE

and

THOMAS M. HINCKLEY

desert shrub Atriplez polycarpa in California. Their data were nearly linear with a RWC of lOOyo corresponding to - 2 2 bars P and 69% corresponding to - 6 5 bars P. Similar results were noted in Salsola vemiculata var. villosa, except that P was - 69 bars at 68% (Sankary and Barbour, 1972b). A relationship between RWC and P has been described for sorghum (Ritchie and Jordan, 1972) and for cotton (Lawlor, 1969), while Campbell and Pase (1972) found no relationship between R WC and P in Cercocarpus betuloides. A closed hysteresis loop was obtained by Jordan and Ritchie (1971) throughout the course of a summer day when hourly comparisons of P and RWC were made on cotton (Qossypium hirsutum). Namken et al. (1971) apparently have observed similar results. Comparisons such as these may be useful in delineating differences between species in their ability to maintain given levels of water stress under various degrees of tissue dehydration (Lopushinsky, 1969). Hodges and Lori0 (1968, 1971) also compared P with oleoresin exudation pressure (OEP),which was measured manometrically, and found a significant correlation (0.01 level; r = 0.82). Because OEP is a function of turgor pressure in the epithelial cells and since xylem sap and cell turgor are in a dynamic equilibrium, these results are not surprising. 5. summary Of the nearly 40 different calibration curves we have examined, all but about 10 conform to one of two basic patterns (Fig. 6; Table I). Fifteen fall into category (A), which has been described previously as the “X” relationship. At high potentials, the psychrometer yields more negative values than the pressure chamber, but as potentials decrease this tendency reverses. The next most common pattern is depicted in line (B). Here the calibration is nearly linear and tends towards a 1 : 1 relationship throughout the measurement range. Thirteen species fall into this category. There are, however, instances when one species has fallen into both categories (e.g. tomato and sorghum). This may be explained by differences in methodology or in characteristics of the tissue used. These considerations are discussed below in sub-section E. Four other types of curves have been generated, but none is represented by more than two species. Calibrations attempted with the freezing point depression technique have been disappointing and difficult to explain, but the psychrometer and density techniques have generally yielded relatively consistent data, except with diffuse porous hardwoods. The necessity for calibration exists when the pressure chamber is being used to estimate water potential. However, when P values are

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

183

P o r P i - + , xylem (-bars)

0

I

I

I

I

FIG.6. Calibrations for most species between pressure chamber and psychrometer fit one of the above curves. Types (A) and (B) are the most common. See text and Table I. (Type A is the so-called “X”relationship.)

used as relative indicators of plant “water stress” and not for absolute or comparative purposes, calibration may not always be necessary.

E.P R E C A U T I O N S Despite the simplicity of measurement, numerous sources of error have been reported with the pressure chamber. It is important that these errors be recognized and that mitigative procedures be followed. Here we discuss several common errors and how they can be avoided.

1. Recutting the stem The initial step in making a P determination is the excision of a stem, twig or leaf from the plant. It is important that this first cut be made with a sharp knife and that no subsequent cutting be done (Barrs, 1968; Wiebe et al., 1971). Scholander et al. (1965s) noted that in many species with long xylem vessel elements, an appreciable quantity of sap withdraws from the cut surface and restores some turgor to the leaves. Subsequent trimming of the stem to produce a better fit in the chamber or to clean the xylem face to facilitate endpoint determination, can result in decreased balancing pressures and, therefore, erroneously high P readings. Klepper and Ceccato (1968), however, used this procedure on

184

GARY A. RITCHIE

and THOMAS

M. HINCKLEY

petioles of citrus leaves, recutting them routinely to standardize length and to insure that the cut ends had not dried out. The scatter in their data points may or may not result from this practice. Scholander et al. (1965a) suggested that in coniferous trees with short tracheids or in monocots with mixed vascular bundles, this error may not be important. Richter et al. (1972) report that excessive trimming of twigs from a giant sequoia branch had considerable influence on subsequent P values 20 min after excision of the branch.

2. The amount of stem inside and outside the chamber Waring and Cleary (1967) indicated that, in their studies on Douglasfir, variability in the length of stem protruding from the pressure chamber top was a source of variation in measured P values. With 50 cm of stem protruding they noted that P values were 10 bars too negative. Conversely they indicated that the amount of stem within the chamber was not critical. Boyer (1967a) and Kaufmann (1968b) have noted, however, that with some species the amount of stem inside the chamber is critical. I n studies with Rhododendron roseum, Boyer compared P with water potentials ($L) in long-stemmed (10-12 cm) and short-stemmed (3-5 cm) samples taken from the same plant and having approximately the same number of leaves. For a wide range of water potentials and comparable lengths of emerging stem, the average difference between P and $ L for the short-stemmed samples was 2-2 bars and for the longstemmed samples 3.6 bars. The difference was highly significant. Boyer concluded that during application of pressure, voids in the stems of rhododendron require filling with water, which increased the required balancing pressure from 0.2 to 0.3 bars (only explaining part of the difference). This effect was present in sunflower but not in yew. The degree of discrepancy is apparently dependent on the length of the stem: the longer the stem the greater the difference between P and 41,.

3. Time elapsed following excision As a transpiring branch is excised, tension in the xylem is relieved, resulting in stomata1 opening and an increased transpiration rate (Slatyer, 1967; Kramer, 1969). If water loss occurs during the time elapsed between excision and measurement, this will result in erroneously low P values. Scholander et al. (1965a) recognized this effect and urged that post-excision evaporation from the twig be kept at a minimum. Waring and Cleary (1967), however, reported that in Douglas-& no decrease in P resulted, despite a 5-min delay between excision and measurement. Ritchie and Hinckley (1971) found no greater than a -0.2 bar change in P of detached needles of five species of conifers

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

185

following 5-min storage periods in petri dishes lined with wet filter paper. If Sitka spruce (Picea sitchensis) twig samples were tightly rolled in polythene bags, kept cool and out of the sun, 6- and 26-h storage caused only 0.5 and 1.0 bar errors respectively (Hellkvist et al., 1974). Goode (1968),however, found that with petioles of apple (Pyrus sp.), currant (Ribes sp.) and black raspberry (Rubus sp.) there was a noticeable P decrease within min after excision. Jordan (1970) measured P in cotton leaves under greenhouse conditions, and P decrease following leaf excision was about 5 bars in 5 min. It is possible that conifers, whose leaf resistance to water loss is generally several times greater than deciduous plants (Holmgren et al., 1965; Gates, 1968), and plants at low water potentials whose stomata tend to be closed, are less susceptible to this error. Nevertheless, it seems desirable to ascertain the magnitude of error for each species studied, t o minimize the time lapse between detachment and measurement, and to store material in a humid chamber prior to measurement (provided sample rehydration does not occur).

+

4. Failure to achieve pressure equilibrium Scholander et al. (1965b) indicated that the pressure chamber measures the negative hydrostatic pressure in the xylem ducts as a “straightforward null measurement” implying the establishment of an equilibrium condition between P and the positive chamber pressure. Boyer (1967a) and Boyer and Ghorashy (1971) actually demonstrated that the quantitative relationship given in Eqn (3) is valid only when the system is at equilibrium. Concerned that water lost from the foliage during sampling, potential gradients within the leaf and twig, and resistance to water flow between xylem bundles and mesophyll cells may disrupt this equilibrium, Boyer (1967a) exposed twigs to balancing pressures for 5-30 min. When no change in pressure was required to maintain balance in sunflower, rhododendron and yew, he assumed that equilibrium potentials in mesophyll cells and xylem ducts had occurred rapidly. Waring and Cleary (1967) reported that if pressure is applied to the chamber too rapidly in Douglas-fir twigs, significant errors occur apparently due to a failure of the system to equilibrate. Under these conditions, P values are often erroneously low (balancing pressures high). They suggest use of a constant, moderate rate of pressure increase (about 0.7 bars sec-I) and note that a higher rate could be used up to within 6 or 7 bars of the balancing pressure in material with low xylem pressure potentials. Others have followed this advice. Kaufmann (1968a, b) used a rate of about 0.32 bars sec-l, West and Gaff (1971) and Haas and Dodd

186

GARY A. ICITCIIIE

and THOMAS

M. HINCKLEY

(1972)used about 0.7 bars sec-l, while Jordan (1970) used 0.2-0.3 bars sec-l and Campbell and Pase (1972) used 0.36 bars sec-l. Oechel et al. (1972a, b) used the higher rate of 1-2 bars sec-l, while Duniway (1971a) and Hong and Sucoff (1971) used the relatively low rates of 0.02-0-07 bars sec-1. Blum et al. (1973) examined the influence of two M e r e n t ratea of pressure increase on the P vs. $L relationship in sorghum (Fig. 7). The higher rate of 0-38 bars seo-l yielded a better estimate of $L than the lower rate of 0.33 bars sec-I. Both rates, however, would be considered low within the context of the range given above. I n contrast, P (-bars) 20

10

0 0

20

FIG.7. The relationship between P and

+L in five varieties of sorghum at several stages of growth aa influenced by two different rates of pressure increme. (Redrawn with permission from Blum et al., 1973.)

Hellkvist et al. (1974) found that the rate of pressure increase in Sitka spruce had no effect on the endpoint pressure. Related to the error associated with the rate of pressure increase is the equilibrium imbalance brought about by loss of water from sample foliage during its enclosure in the pressure chamber (Scholander et al., 1965s; Boyer, 1967a; Duniway, 1971a). Some authors consider this error to be substantial and have taken such precautions as bubbling the incoming nitrogen through water (Boyer, 1967a; Boyer and Ghorashy, 1971; Goode and Higgs, 1973) or enclosing the leaf samples in plastic bags before inserting them into the chamber (Klepper and Ceccato, 1968; West and Gaff, 1971). Our experience with coniferous material (Abies, Pinw, PsewEotsuga) indicates that the bubbling procedure may not be required with these

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

187

genera, possibly reflecting their previously mentioned high leaf resistances. Other workers have expressed agreement and do not take mitigative precautions (Kaufmann, 1968a, Doley, 1970; as well as E. D. Schulze, P. V. Biscoe, W. Koch, P. E. Kriedemann: pers. comm.).

5 . Heat build-up in the chamber during pressure application One would expect that application of pressure would result in an increase in temperature in a chamber of constant volume. J. S . Boyer (pers. comm.) indicated that he observed temperature increases of up to 16°C in chamber contents when pressure was applied rapidly. Recently

..\

--

e

-

Gas

D

Entry

I

1

1

1

1

1

FIU. 8. Temperature changes recorded within the pressure chamber during a measurement sequence. (A) increase in temperature with initial increase of gas pressure, (B) gradual decline in temperature with increased gas pressure, (C) cooling of chamber at final (constant) pressure, (D) drop in temperature with gas release. (Reproducedwith permission from Puritch and Turner, 1973.)

Puritch and Turner (1973) studied this phenomenon. Instrumenting a pressure chamber with a 26-gauge copper-constantan thermocouple, they measured the changes in internal temperature as pressure increased and decreased. Their findings are shown in Fig. 8. As pressure was applied the internal temperature rose precipitously (A). The magnitude of this rise was directly related to the rate of pressure increase with high rates producing temperature increases of as much as 30°C. As pressure continued to rise, temperature reached a maximum value, then declined (B). Upon cessation of pressure application, temperature stabilized at its initial value (C) until the pressure was re-

188

GARY A. RITCHIE

and

THOMAS M. HINCELEY

leased (D). At this point, chamber temperature dropped to subzero levels. When plant material was enclosed in the chamber, temperatures were moderated due to the absorption of heat energy by the foliage. The effects of these temperature fluctuations on plant water potential were discussed. An increase in temperature may increase the osmotic potential as the two are directly related:

+, = CSRT

(6)

where $, is the osmotic potential, C, is the molar concentration of the solute, R the gas constant and T the Kelvin temperature (Slatyer, 1967). An increase in solute concentration may also result from elevated metabolic rates at higher temperatures, and alterations in the turgor and matric potentials are suggested. If measurements are of long duration, elevated temperatures could also increase the rate of transpiration (see above). Another problem may result from the rapid cooling of foliage after the release of pressure. Several workers (Boyer, 1967a; Klepper and Ceccato, 1968; Barrs et al., 1970; Duniway, 1971a; West and Gaff, 1971) have calibrated P values against a thermocouple psychrometer by fist determining P in the pressure chamber, then removing the leaf and inserting it into the psychrometer for measurement. Duniway (1971a) and West and Gaff (1971) obtained close agreement between P and t , h ~ and Boyer (1967a) found discrepancies of only f 2.5 bars in rhododendron. However, Richter (pers. comm.) found that rapid decompression was associated with the formation of brown spots on leaf margins, indicating mechanical tissue damage. Hence, the possibility of tissue disruption resulting from rapid decompression and its subsequent effect on +L should be recognized.

6 . Recognition of endpoint The principal of operation of the pressure chamber provides that an endpoint occurs when the external balancing pressure equals the xylem pressure potential. At that instant water should appear on the cut face of the twig. For many species this endpoint is sudden and easily recognized. For others it is not. Two types of difficulties arise. Klepper and Ceccato (1968) report that with petioles of grape (Vitis sp.) leaves, bubbling on the cut surface during pressure application was observed before balancing pressure was achieved, due to the passage of air out of the xylem carrying with it xylem contents. They were able to distinguish between this false endpoint and the real endpoint by subjectively determining when bubbling more or less gave way

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

189

to fluid exudation. They indicated that some experience was needed to recognize this distinction. Jordan (1970) reported that, with cotton leaves, bubbles frequently emerged from the cortical and pith tissues well before fluid appeared on the cut surface of the xylem. Mistaking the two would cause serious errors. Experience and reduced rates of pressure increase seem to enhance endpoint recognition. Another source of difficulty occurs in coniferous genera which have resin ducts in the mature xylem (e.g. Pinus, Pseudotsuga). Kaufmann (1968a) noted that his endpoint determinations were complicated by resin exudation in loblolly pine (Pinus taeda) and eastern white pine ( P . strobus). With white pine it was necessary to continue to wipe resin off the cut face throughout the period of pressure increase. With loblolly pine he was able to detect the endpoint by observing the rate of bubbling of water up through the resin film. Our own experience (Ritchie and Hinckley, 1971) with several genera of conifers indioates that with pines and Douglas-fir, resin exudation can be (but is not always) a problem, whereas with the true firs (Abies sp.) it is not. Occasionally premature water bubbling occurs with true fir samples but experience enables us to define endpoints. Again, reduced rates of pressure increase (0.1 bars sec-l) are desirable. The problem with resin exudation seems to be open to two solutions. Experiments with conifer needles or fasicles removed (not cut) from the stem and placed directly into the pressure chamber (Ritchie and Hinckley, 1971; Gifford, 1972) indicated that endpoints obtained in this manner are sudden and clear and not complicated with resin exudation if the needle bases are not damaged. Secondly, Richter and Rottenburg (1971) have devised a small battery-powered conductivity meter for determining endpoints. The fine electrodes are placed on the cut xylem face and as water reaches the surface an abrupt increase in conductivity is observed.

7. Osmotic and matric potential of the xylem sap A word is needed concerning the common assumption that the osmotic and matric potentials of the xylem sap ( I);Y'~"') are negligible and need not be quantified when using the pressure chamber to estimate xylem water potentials. With many species this assumption seems valid. Boyer (1967a) and Scholander et al. (1966) have demonstrated that the osmotic potential of xylem sap is generally greater than - 3 bars in species with which they have worked. For Sitka spruce, values were always greater than -0.2 bars (Hellkvist et al., 1974) and for cotton, greater than -0.8 bars (Jordan, 1970). Some precaution regarding this assumption is warranted, however.

190

QBRY A. RITCHIE and THOMAS M. HINCKLEY

J. S. Boyer (pers. comm.) has cited unpublished data on rhododendron growing in mannitol solutions in which the pressure chamber gave measurements of - 12 bars, an analysis of #;Ylern indicated a value of - 9 bars, and the psychrometer gave # L values of - 11 bars. Boyer of xylem sap in woody warns against assuming negligible @'Iern plants growing in saline medium. His warning has been echoed by Kappen et al. (1972)) who measured @lem values of - 6.2 bars in the desert shrub Artemisia herba-alba. There is also evidence that #:Y1ern varies with changes in #'. Boyer iii L (-bars) 30

20

r -/ -

/*

-RHODODENDRON - - - SUNFLOWER -a-

YEW

FIQ. 9. Xylem osmotio potentials measured at various leaf water potentials (#L)in sunflower, yew and rhododendron. (Redrawn with permission from Boyer, 1967a.)

(1967a) found such a relationship in yew and rhododendron (Fig. 9). As t , h ~varied from about - 5 to -30 bars, #:ylern varied from about -0.4 t o - 1.3 bars in yew and -0.5 to -2.5 bars in rhododendron. Spomer and Langhans (1972) found a change in t,hiylern of more than 8 bars associated with a decrease in # L of from - 5 to - 2 0 bars in Chrysanthemum morijbrum. Where the pressure chamber is employed to estimate # L , measurements of @Iern are clearly warranted.

8. Effect of tissue age or developmental stage Leaf age has been shown to influence a number of plant water relationships. Leaf resistance, for example, increases with age in corn (Turner, 1969), cotton (Slatyer and Bierhuizen, 1964) and red pine

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

191

(Pinus resinosa) (Waggoner and Turner, 1971). Begg and Turner (1970) demonstrated a higher petiole resistance in older leaves of tobacco than in younger leaves. This is apparently reflected in a decrease in water loss with leaf age in some species (Pazourek, 1968). This effect was enhanced in sitka spruce needles by an accumulation of epidermal wax which increased with age (Jeffree et d., 1971). catskg (1962) observed a preferential movement of water toward younger leaves in two Brassica species under conditions of water stress. Similar phenomena have been observed in soybean by Stevenson and Shaw (1971). P (-bars) 12

8

4

0

4

12

FIG 10. The effect of leaf age on the relationship between P and leaf water potential (+L) in tomato. Young leaves were sampled on 23 March 1969; older leaves were sampled on 23 May 1969. (Reproduced with permission from Barn et al., 1970.)

Therefore, it is not surprising that leaf age might influence the relationship between #L and P. Kaufmann (1968b) observed different slopes in the regressions from young and older leaves of “Valencia” and “Washington” Navel oranges and suggested ohanges in membrane permeability and wall strength as possible mechanisms. Barn et al. (1970) noted a similar age effect with tomato foliage (Fig. 10) and Frank and Harris (1973) found that the stage of growth influenced the pressure chamber-psychrometer regression in wheat leaves (Fig. 11). It appears that, in at least some species, the age or stage of development of sample tissue can influence the relationship between P and #L, and that care should be taken to account for this source of error.

192

GARY A. RITCHIE

and

THOMAS M. HINCKLEY

P (:bars)

FIQ. 11. The relationship between P and leaf water potential (#L)in wheat as affected by developmental stage. (Reproduced with permission from Frank and Harris, 1973.)

F.

MEASUREMENTS O F

P

ON C O N I F E R N E E D L E S

After some reports of unsuccessful attempts at measuring P on conifer needles directly (e.g. Kaufmann, 1968a), Johnson and Nielson (1969) reported the results of some successful measurements. For a pressure gland they used a soft surgical rubber disc approximately 5 mm thick and slitted along the radius. The fasicle sheath was stripped from pine needles and the tiny protruding xylem traces were cut crosswise to provide a smooth viewing surface. A metal disc perforated in the center was screwed down over the rubber gland to form a tight, leakproof seal. They reported excellent results with this technique. (Turner et al. (1971) reported good results with this technique on grass blades using a similar apparatus.) When the needle bases were not damaged, resin exudation was minimal and endpoints were readily distinguishable. P values measured on needles (Pn) were compared with P values measured on stems for red pine, white pine and Austrian pine (Pinus nigra). With each species the P vs. Pn relationship was linear and nearly equal throughout a range of from - 2 to - 40 bars. Johnson and Nielson (1969) concluded that measuring Pn on these species was as reliable or more reliable than measuring P and had the advantage of smaller tissue requirements. Hong and Sucoff (1971) measured Pn on individual red pine fasicles from plantation trees in Minnesota. They commonly found a range of f 0.7 bars among sets of four measurements. Occasionally, however,

THE PRESSURE CHAMBER I N ECOLOGICAL RESEARCH

193

two readings from the same branch differed by as much as 2.1 bars during midday. They indicated that errors can be kept within f 1.0 bar if care is taken. Gif€ord (1972) has devised a highly simplified chamber for measuring P in pine needles. The device consists of a hollow, cylindrical tube approximately 20 cm long and 1-5 cm in diameter, sealed at one end. He reports success with fasicles of radiata pine (P. radiata) which yielded sudden, clear endpoints through a range of from - 5 to - 35 bars. No calibration details were given. Ritchie and Hinckley (1971) reported on P vs. P n comparisons in lodgepole pine (P. contorts), Jeffrey pine (P.jeffreyi), Douglas-fir, noble fir (Abies procera) and Pacific silver fir. Comparisons in both pine species gave results very similar to those of Johnson and Nielson (1969). With Douglas-& and the true firs, however, P was always more negative than P n . I n noble and silver fir, a 4 bar P vs. P, difference occurred a t - 14 bars P;in Douglas-fir the difference waa about 2.5 bars. Similar results have been noted in Abies balsamea (C. H. A. Little, pers. comm.). The fact that less pith was present in the stems of pines than firs led to the tentative conclusion that the P vs. P,,discrepancy resulted from the pith of the firs filling with xylem sap during pressure application, causing excessively high balancing pressures. Where practical, measurements of P n may be preferable to measurements of P. Although estimates of P are apparently more sensitive to diurnal fluctuations in water stress, Pn is probably a better indicator of the true water potential in the photosynthesizing and transpiring tissues. The advantages of smaller tissue samples, less gas required to fill smaller chambers, and clearer endpoints, also speak in favor of the Pn technique.

G.

U S E O F T H E P R E S S U R E C H A M B E R TO D E T E R M I N E OSMOTIC A N D MATRIC P O T E N T I A L S

Scholander and colleagues (1964, 1965a) proposed a procedure for measuring osmotic potentials ( n ) with the pressure chamber. A leafy shoot is excised, placed in the chamber and subjected to increasing levels of internal chamber pressure. When the internal pressure exceeds the balancing point (P),xylem sap is expressed from the shoot’s cut end. Pressure is incrementally increased and the amount of sap expressed with each increment is recorded. When a good linear relationship has been established, the shoot is weighed, dried and reweighed. This weight difference added to the total amount of sap expressed is an estimate of the original volume of sap in the shoot. These data are then plotted as a “pressure-volume curve” where the

194

GARY A. RITCHIE

and

THOMAS M. HINCKLEY

volume of water removed is a function of 1 / P (Fig. 12). The following quantitative relationship is believed to exist : I/P =

v- v,

-

(7)

RTn

where P is the chamber pressure in bars (assumed equal and opposite to the leaf water potential), V is the volume of cell sap that has been removed by the pressure (cm3),V, is the original volume of cell sap in turgid leaf, n is the solute content (moles), R is the gas constant and T the Kelvin temperature (Boyer, 1969). Tyree and Hammel (1972) have -0.1

vl

2i a

‘=

water outside

0 0

100

VOL‘JME OF S.4P REMOVED

-v0

(9;) I

I Vt I

FIU. 12. Typic2 “pressure-volume curve” for a leafy shoot. The .-mar portion of the curve is an estimate of the osmotic potential at that volume. Extrapolation to the left produces an estimate of the original osmotic potential, and to the right gives an estimata of the volume of water outside the cell protoplasts. (Reproduced with permiasion from Boyer, 1969). V , is the volume of free water; V t is the total volume of water in the tissue obtained by oven-drying the tissue (see Hellkvist et al., 1974).

provided a thorough quantitative examination of this relationship and found it to be valid. A “pressure-volume curve” (Fig. 12) typically has two portions. I n portion (A) turgor pressure and osmotic potential are combined. As turgor pressure falls to zero with increased chamber pressure, the relationship becomes linear (B) and represents only n. Here the leaf cell n is balanced only by the hydrostatic tension in the xylem and extracellular (apoplastic)sap. Since the osmotic potential is proportional to the concentration of active materials in the cell, and since the concentration is inversely proportional to the amount of cell water, the

THE PRESSURE CHAMBER IN ECOLOQICAL RESEARCH

195

cell T and hydrostatic tension of the xylem sap are inversely proportional to the water removed from the cell. The “pressure-volume curve” is the zero turgor line and is the osmotic potential for any volume. Therefore, it intersects the ordinate at the reciprocal of the initial r of the leaf cells (Hammel, 1968). I n addition, the intersect of the curve with the abscissa is an estimate of the volume of water outside the leaf protoplast because line B apparently represents behavior of protoplasts with little effect from cell walls. This technique has been reviewed by Boyer (1969) and has found some use in the detection of damaged membranes in diseased plants (see Section IV A). Generation of pressure-volume curves has been restricted to theoretical rather than eco-physiological studies because of the long sampling times involved and the lack of a coherent theoretical examination, until that of Tyree and Hammel (1972). However, by modifying the lid of a pressure chamber, Hellkvist et al. (1974) were able to take simultaneous measurements from several Sitka spruce twigs. They examined pressure-volume curves for various times of the year and at various heights in the 10.5 t o 11 m study trees. Solute potentials (osmotic plus matric) decreased with height and from early to late summer. I n addition, the mean water potential at incipient plasmolysis (point at which turgor pressure becomes zero, i.e. where curve A intercepts line B, Fig. 12) was - 21 bars in early summer and -33.7 bars in late summer. Similarly, bound water content (determined from the ratio of {Vt- V,}/Vt) reached very high values in the winter and declined through the summer. Obvious plant adjustments in both solute potential and percentage bound water would affect physiological processes such as frost and drought hardening, stomatal activity, etc. ; however, measurement of P alone would not yield information on these quantities. The importance of osmotic adjustment in maintenance of positive or constant turgor under conditions of decreasing leaf water potential for stomatal opening (Biscoe, 1972; Goode and Higgs, 1973) and plant growth (Meyer and Boyer, 1972) has been emphasized. This expanded use of the pressure chamber in eco-physiological studies by Hellkvist et al. (1974) represents a major development. Matric potentials (T)have also been estimated with a pressure chamber (Boyer, 1967b). The technique involves freezing and thawing leaves, which destroys the membranes and converts leafy tissue into disorganized masses of cell walls containing small volumes of cell sap. Boyer argues that with the hydrostatic component of water potential thus removed, chamber pressures must be balancing matric forces. These are not clearly defined but are associated with forces exerted by adsorbed water, colloids and surface tension. Thus they vary with

196

GARY A. RITCHIE

and

THOMAS M. HINCKLEY

water content; and therefore the volume of sap expressed from frozenthawed tissue at varying pressures is believed to generate a curve of matric potentials at decreasing water contents. Tyree and Hammel (1972) have questioned this approach, however, suggesting that much of what Boyer called matric potential is actually a mechanical resistance of cell walls to compression and would be more properly considered M turgor pressure.

H.

WHERE TO SAMPLE

Any attempt at characterizing the behaviour of a system as complex and variable as a living plant by sampling at only a few preselected points within the system requires that such sampling points be carefully and wisely selected. The value of P varies both spatially and temporally in plants, and knowledge of the patterns of variation are prerequisite to the formulation of an adequate sampling program. The problem of when to sample (temporal variability) will be dealt with in Section I11 B. Here we examine the phenomenon of spatial variability of P within a single plant at any point in time. Three sources of spatial variability of P exist in plants: (1)variation with height above the ground due to hydrostatic pressure and xylem resistance or friction effects, (2) variation with tissue age and (3) variation with differences in the absorption-transpiration balance. The relationship of friction and gravity to water potential is shown in Eqn (2). I n theory, a decrease in water potential should accompany an increase in height in a plant due to both the gravitational and frictional effects. Therefore, in tall trees this is an appreciable source of variability under both equilibrium and non-equilibrium conditions. Scholander et al. (1965a)measured P changes with height in Douglas-fir and redwood (Sequoia sempervirens) and found the expected pressureheight relationship in both species. Tobiessen et al. (1971) measured the P gradient in a 90 m-tall giant sequoia tree with similar results. This effect is negligible in smaller trees and plants and is usually ignored. Kaufmann (1968b) and Barrs et al. (1970) have demonstrated that the P vs. t , b ~ relationship varies with foliage age in orange and tomato leaves. Leaf age also influences such factors as stomata1 resistance and reactivity, dry matter content and wilting resistance. Therefore, it is not surprising that Sucoff (1972) found substantial P differences between 1966, 1967 and 1968 needles sampled from the main axis of a branch of the 4th whorl of a red pine tree. The crowns of most plants growing in natural environments are constantly experiencing internal changes in water supply and external changes in microenvironment. These external changes are brought

197

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

about by such factors as sun angle, wind speed and direction, proximity to solid objects, etc. I n addition, crowns of larger plants or trees can influence their own microenvironment to an appreciable extent through mutual shading. Because P levels are a function of supply and loss of water and because loss of water depends to a large extent upon crown microenvironment, it follows that within-crown variations in P should be in evidence at any point in time. Waring and Cleary (1967) were early to demonstrate this on Douglas& (Table 11).They compared variability in P values under different soil and atmospheric moisture conditions and at different levels of sampling. Variability increased with sampling level and degree of soil and atmospheric moisture stress. TABLEI1 Maximum variation in pressure chamber readings on Douglas-jZr under diflerent conditions of atmospheric stress ( A S )and soil moisture stress (SMS). (Reproduced with permission from Waring and Cleary, 1967. Copyright 1967 by the American Association for the Advancement of Science.)

Low AS Low SMS

High A S Low SMS ~- ~ . . . _ _ _ _

f 0.5 bar f 0.5 bar

f 1.0 bar

1.0 bar

Within a branch f 0.5 bar Shade versus exposed branch f 1.0 bar Within a tree f 1.5 bars Among trees 3.0 bars

_ _

High A S High S M S ~

k 1.0 bar k 1.5 bars

+ 2.5 bars f 10.0 bars

Microclimate varies with height above the ground; thus one might expect to find P varying with height in a plant (neglecting gravitational effects). Although Goode (1968) was unable to detect such variation in 2.5 m-tall apple trees, Jordan (1970) noted marked variability with point of leaf insertion (node) in greenhouse cotton plants. At 2200 h, means of four samples from each of eight insertion points varied from about - 3 to -4.5 bars, while at 1300 h the range of means increased to from - 7 bars to - 14 bars (Fig. 13). Waggoner and Turner (1971) compared diurnal P values in upper and lower crown foliage of plantation red pine. P was always more negative in the upper foliage than the lower, the difference being as great as 4 bars during midday. Hinckley and Scott (1971) were able to define different zones of xylem pressure potentials within crowns of 6 m-tall Douglas-fir saplings. Begg and Turner (1970) used the pressure

198

and

GARY A. RITCHIE

THOMAS M. HINCKLEY

chamber to determine P in leaves and stems of field-grown tobacco plants. At 1300 h (EST) on a clear sunny day, P values of -9.2 to -3.8 bars were recorded for the top (90 cm high) and bottom (20 cm high) leaves respectively. P measured on covered leaves (estimate of Psiem) varied with height from - 6.6 to - 1.8 bars. Hinckley and Ritchie (1 970) measured within-crown variability of P in a large field-grown Pacific silver fir tree on two separate days during the summer. They found that on both days a reverse gradient existed. P tended to be more negative in the lower crown than in the middle or upper crown. There was evidence for more efficient stomata1 control of

id 4

0

4

8

12

16

P(-bars) FIG.13. The relationship between P and the position of leaf insertion on the main stem of cotton above the cotyledonary node at 2200 h (left) and 1300 h (right) of the same day. The horizontal lines represent the standard error of the mean of four determinations made at each leaf position. (Reproduced with permission from Jordan, 1970.)

transpiration in the upper branches, possibly accounting for the reverse gradient. Hellkvist et al. (1974) also noted inverted P gradients within the crowns of plantation Sitka spruce; however, they attributed these gradients to differences in xylem resistance to water flow between upper and mid-crown branches and bole. Variability in crown microenvironment associated with orientation is perhaps even more pronounced than that accompanying height. Zaerr (1971) was unable to detect any such variation within crowns of 6.4 to 10 m-tall Douglas-fir trees; however, Kaufmann (1969) noted that with orange trees, aspect waa a source of considerable P variability. During the afternoon, P was lower on the south side of the crown and

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

199

higher on the shaded north side, whereas morning values were uniform throughout. This pattern has since been observed in a number of studies (Pierpoint, 1967; Klepper, 1968; Hinckley and Scott, 1971; Jones, 1972). From 0900 to 1800 h (PDT) on a July day, P differed between north and south crown aspect of two large Pacific silver fir trees by as much as 4 bars (Hinckley and Ritchie, 1970). There was statistical evidence to implicate not only external factors such as net radiation, vapor pressure deficit and air temperature but also internal factors such as transpiration rates, stomata1 reactivity and possible redistribution of

8

; 12 14 08

12

1'6

20

TIME O F DAY (hours) FIG.14. Xylem pressure potential ( P ) compared between the second and sixth whorl from the top of a red pine tree on 19 June. (Reproduced with peimission from Sucoff, 1972.)

water within the crown. Haas and Dodd (1972) studied the desert shrub, honey mesquite (Prosops~sghndulosa var. ghndulosa), growing in natural communities in Texas. They noted that P on the sunny sides of the shrubs was always more negative than P on the shady sides and that the difference was greatest in mid-morning and mid-afternoon. Sucoff (1972) examined the variability in P within the crowns of 7 to 8 m-tall red pine trees in northern Minnesota. He was concerned with (1) variation among whorls, (2) variation among major branches within a whorl and (3) variation among branch orders within a major branch. Among whorls the uppermost usually had more negative P values than the lower a t midday, but at 0800 and 2200 h agreement between the two was within 1bar (Fig. 14). He attributed these results to differences in solar radiation. Daily mean differences between branch

200

GARY A. RITCHIE

and

THOMAS M. HINCKLEY

pairs from the third whorl were always less than 0-7 bars. The maximum difference recorded was 1.2 bars. P values within any major branch were within 1 bar. Theory predicts and data support the existence of appreciable spatial variability of P within crowns of herbaceous plants as well as large trees, Thus the formulation of adequate sampling schemes must include careful preliminary testing and rigorous consideration of research objectives.

111. REVIEWO F ECOLOQICAL STUDIES All ecological inquiry is directed basically towards explaining distribution patterns of organisms (Spomer, 1973).I n terrestrial plants, distribution is often regulated more by the availability of water during critical periods than by any other single environmental factor (Kozlowski, 1968a). The availability of water at any given time is reflected in the water status of the plant. It follows that assessment and interpretation of plant water status in relation to environmental factors should be a useful approach in interpreting distributional patterns of plants. However, assessment and interpretation of plant water status using single factor measurements, or efforts to relate a single factor measure of plant water status to a physiological response using correlation analysis, may not be satisfactory. This concept will be examined. I n this section we will review the status of the pressure chamber technique in current ecological and eco-physiological research and suggest what appear to be promising new directions and applications. Our emphases will be on synthesis, summarization and evaluation rather than on a lengthy and detailed discussion.

A.

SOME PHYSICAL RELATIONSHIPS

I n Eqn ( l ) , a conceptual model of plant water status was presented. If we can assume that P,as an estimate of $, is a reasonable index of water status ( W in Eqn (l)), it follows that P should reflect imbalance between the supply of available water in the soil and the demand for water by the atmosphere (AED)l if water storage (I?)is negligible or constant. The physical mechanisms of water movement through the soil-plant-atmosphere continuum (SPAC)(Cowan, 1965)have been the subject of numerous articles and reviews (e.g. Slatyer, 1967; Kramer, Atmospheric Evaporative Demand ( A E D ) reflects both the heat load at the leaf surface and the diffusion gradient across the stomata1 pathway and is roughly equivalent to net radiation, relative humidity, vapor pressure deficit or measured evaporation.

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

20 1

1969; Dainty, 1969) and may be summarized as follows. Water is supplied through mass flow in the liquid state and seems to behave as a series of steady-state fluxes (van den Honert, 1948; Elfving et al., 1972) through various components of the SPAC. Flow from the soil-root interface through the root cortex and endodermal tissues can be expressed as :

G = Jv = LP{ - p - (%d, - %y,) - T,ou} (8) where (7 is water gained and is proportional to volume flow (Jv)and Lp is the hydraulic conductivity of the system (Dainty, 1969). Root resistance (l/Lp) as well as the soil resistance is variable, is often the major resistance to liquid water movement (Boyer, 1969), and depends on a,$root, root metabolism and soil temperature (Kramer, 1969). Movement through the xylem occurs (Dainty, 1069) by:

JV

=

Lp(pgAh+ A P )

(9)

where both p and g are constant and represent density of water and acceleration due to gravity respectively. The height above the water table, or h, is constant when samples are taken at the same vertical position in a plant and negligible in small plants. Xylem resistance (l/Lp) is generally constant (Dimond, 1966); however, it may vary with position in the canopy and age (Begg and Turner, 1970). Supply of water to the leaf is (Dainty, 1969) simply:

JV

=

AT)

(10)

AT is relatively constant over short time periods (hours), although Lp generally increases with hydration of leaf tissues. Loss of water can occur only when two sets of conditions are met: sufficient energy must be available to convert liquid water to water vapor (latent heat of vaporization), and the specific humidity gradient across the leaf and boundary layer must be sufficient to overcome the resistances to diffusion. Hence:

where E is the increased latent heat content with increase of sensible heat content of saturated air at the temperature of the ambient air (T), H‘ is the net rate of absorption of radiant heat by the leaf, X is the latent heat of vaporization of water, Pa is the density of air, q‘(T) and q are the saturation and actual specific humidities respectively of the ambient air, and Ra and R L are the external (boundary layer) and internal (stomata1 and mesophyll) resistances of the leaf (Cowan and Milthorpe, 1968).

202

GARY A. RITCHIE

and

THOMAS M. RINCELEY

For most purposes, E, h and pa are constant, while all other variables in this relationship depend on environmental or plant conditions. Changes in net or solar radiation will leadto changes in H , q, q’ and T,while changes in wind speed are accompanied by changes in Ra. Thus, any factors of the environment which alter soil temperature or moisture will influence P through their effect on supply (Eqns (8), (9) and (lo)),and any factors which change the energy budget, diffusion gradients or resistances across the leaf will affect P through their impact on demand (Eqn (11)). Interpretations of P values relative to environmental factors must be made within the framework of these physical relationships.

B. P L A N T

RESPONSES T O SUPPLY AND DEMAND

The physical relationships outlined above can be distilled into some rather simple working concepts. Three factors must be considered : supply, demand and control. Supply factors are related to the availability of soil moisture as influenced by soil temperature, solute and matric potential. Demand factors (AED)include leaf energy load and diffusion gradient. The primary control exerted by the plant over water The diurnal fluctuation in P loss is resistance at the leaf surface (RL). is a function of the water imbalance resulting from the excess of transpiration over absorption (Eqn (I)).Assuming that moisture availability does not fluctuate widely over short time periods (hours), transpirational flux is proportional to demand divided by resistance (Elfving et al., 1972): flux

AED UP RL

u-

I n theory, when soil moisture is readily available, stomata remain open, RL remains low and P responds directly to flux, which is a function of AED.Conversely, with decreasing soil moisture availability and stomata1 closure, RL dominates so that P reflects leaf resistance more closely than it reflects AED. At night, AED is very low and R L tends to be high so that flux is generally negligible. Hence P tends to reflect supply factors. These concepts have implications in the interpretation of diurnal and seasonal patterns of P . When soil moisture availability is adequate, diurnal fluctuations in P should reflect the diurnal march of AED,as was observed by Klepper (1968) in a variety of woody angiosperms and by Cary and Wright (1971) in a number of crop species. Similar results

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

203

have been noted in a number of studies (Goode, 1968; Scholander and Perez, 1968; Namken et al., 1969, 1971; Stark, 1970; Pearcy et al., 1971; Haas and Dodd, 1972; Sucoff, 1972; Elfving et al., 1972). I n each of these studies, P closely followed the daily trend of AED whether expressed as radiation, vapor pressure or humidity. On cloudless, warm days the pattern shows a decline in P during the morning followed by a distinct peak around noon and then an increase (Fig. 15, Type I curve). Such a curve seems to be indicative of high soil moisture availability and generally low leaf resistance. Under these conditions, flux, hence P,is a function of AED.

24

6

TIME OF DAY (hours) 12 18

24

“t

I

I

I

I

I

FIG.15. A generalized curve depicting the diurnal course of atmospheric evaporative demand (,4ED) and xylem pressure potential ( P ) ,The Type I curve occurs when soil moisture is not limiting. The Type I1 curve is typical of plants undergoing moisture stress sufficient to close stomata and restrict transpiration, allowing for some recovery of P .

As soil moisture becomes limiting, however, diurnal curves assume a different character (Fig. 15, Type I1 curves). With these curves, P normally begins at a lower pre-dawn value and decreases sharply, reflecting increasing AED.But rather than a peak, a plateau reflecting the influence of RL occurs and may persist for several hours. Finally, P again increases to a level often somewhat lower than the initial predawn value. Cam (1971/1972) observed that with “dry” soils, P in tea plants (Camellia sinesis) generally decreased until 1000 h and plateaued, then began to increase at 1600 h. This plateauing phenomenon has been observed in a wide variety of species (Cleary, 1968; Pearcy et al., 1971;

204

GARY

A. RITCHIE

and

THOMAS M. HINCKLEY

Zaerr, 1971; Haas and Dodd, 1972; Halevy, 1972; Jones, 1972). Occasionally Type I1 curves display a distinct midday increase followed by a secondary minimum in later afternoon (Hinckley and Ritchie, 1970; Jones, 1972; Sucoff, 1972; 0. L. Lange, pers. comm.). Both Type I and Type I1 curves seem to be characteristic of mesic species and tend to reflect the supply conditions and not the species under consideration. Hence a given species might display a Type I curve until soil moisture becomes limiting, at which time it would gradually change to a Type I1 curve. I n general, the degree of correlation between P and AED depends on availability of water supply-the greater the supply, the stronger the correlation. A number of diagnostic properties can be observed in the two curves. For example, when water supply is not limiting, night-time P values tend to approach pre-dawn values for that day. As the moisture supply becomes limiting, however, re-equilibration of P with the pre-dawn value requires a longer time period or may not occur at all. Also, as moisture supply dwindles the diurnal variation in P tends to decrease. This was observed by Sucoff (1972) in red pine and Gardner and Nieman (1964) in pepper. However, all species do not display this pattern, as will be shown later. Of additional interest is the level of P at which the plateau occurs. If this relaxation of water stress is indeed caused by increased leaf resistance, it may follow that resistance to CO, diffusion is also increased. Although it is well known that internal impedance to CO, assimilation involves metabolic as well as physical resistances, simple stomata1 closure alone might limit net assimilation. Species able to assimilate CO, under severe water stress conditions may have adaptive advantages over those which are not. On the other hand, it has been suggested (Lopushinsky, 1969) that the ability of certain pines to limit water loss through RL may be related to their persistence in drought-prone environments. Another type of curve which seems to be characteristic of some arid zone species has not been depicted in Fig. 15. Kappen et al. (1972) measured P in Artemisia herba-alba in the Negev Desert of Israel and reported that, a t 0400 h on 7 September, P in the xylem was about - 100 bars. By 1200 h it had decreased to - 118 bars and only gradually increased to about -112 bars by nightfall. Oechel et al. (1972b) reported similar data for Larrea divericuta, a desert shrub widely distributed in the southwestern United States, Mexico and South America. Whether this form of curve is an extreme variant of the Type I1 curve resulting from excessively low soil moisture, or a distinct type typical of arid zone species regardless of moisture regime, is not presently clear.

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

c. E X P R E S S I O N

205

A N D INTERPRETATION O F DATA

Perhaps the single most remarkable attribute of the pressure chamber is the ease and rapidity with which it can be used. An estimate of P can usually be obtained within minutes after excision of a twig or leaf. I n addition, the technique is readily adaptable to a wide spectrum of plant forms, from grasses (Teal and Kanwisher, 1970; Turner et aE., 1971) to succulent herbs and ferns (Hickman, 1970) to deciduous (Kaufmann, 1968a, b) and coniferous trees (Waring and Cleary, 1967; Waggoner and Turner, 1971). With a technique of such convenience and simplicity, it may be tempting hastily to begin data collection before a careful experimental design has been formulated. Once it has been established that P varies diurnally and seasonally in a predictable fashion and that species differences do indeed occur (whether in fact or artifact), it then becomes necessary to apply meaningful and imaginative strategies for obtaining and analyzing field data. I n this sub-section we outline several such strategies, largely founded on the physical concepts outlined in sub-sections A and B above, which have been shown to be of value in interpreting ecological patterns and relationships based upon measurements of P. 1. Base-P and other parameters

If one were to measure P continuously throughout a clear, calm summer day when #8'so,, is relatively high, a curve resembling Fig. 16A would be generated. P measured just before dawn, or base-P (BP), is a function of soil moisture if it is assumed that during the night, low atmospheric demand for water and stomatal closure prohibit transpiration. Therefore, the water potential gradients which existed in the plant during the previous day have been equalized and equilibrium has been established with the soil (Waring, 1969a; Begg and Turner, 1970; Hickman, 1970). This assumption also depends upon the ability of the soil to supply water to the plant at night in sufficient quantities to recharge the dehydrated plant tissues. Several factors affect these processes, such as (1) soil temperature, (2) night-time vapor pressure deficit, (3) night-time stomatal closure and (4) soil water potential. The justification of this assumption will be discussed in detail in Section V D, but for now it will be considered valid. D P is the depression of P below BP at any time during the day (Hinckley and Ritchie, 1973). It is, therefore, a function of the ratio of supply to loss of water and depends upon the evaporative power of the atmosphere and the internal and external resistances to the flux of

206

GARY A. RITCHIE

and

THOMAS M. HINCKLEY

water within the plant; the former affects supply, the latter affects loss. Therefore, at any time during the day

P = BP+ DP (13) The minimum P value (P"'") and the maximum P depression (DPmax) occur simultaneously at the same P value and

Pdn = B P + DPmaX

(14)

I

I

I

I

I

I

I

24

04

08

12

lti

20

24

TIME OF DAY (hours)

FIG.16. (A) An idealized diurnal curve showing the relationship between base-P ( B P ) and depression-P (DP). The highest D P value (DPmux) occurs at the lowest P value (P'"'").(B) All P values summed for the day equal CP, which is the sum of Z B P and ZDP.

On cloudless days DPmaXtypically occurs from 1300 to 1500 h. On intermittently cloudy days, however, DPmaxmay occur at any time. Total daily P (ZP)(Fig. 16B) is the integrated area under the diurnal P curve (Haas and Dodd, 1972). Net daily P (Pnet)is the difference between XBP and ZP:

Pnet= ZP-ZBP

=

CDP

(15)

Various workers have used these concepts and expressions in attempts

THE PRESSURE CHAMBER IN ECOLOGICllL RESEARCH

207

to quantify or explain the relationships between P and environment in a number of species. Waring and Cleary (1967) reported a relationship between BP and Pdn in a plant undergoing drought (Fig. 17A). BP reflects the equiDECREASING 3 SOIL

-

JULY

AUGUST

m ld

n

-10 -20

-30

-40 0 -20

-40

-60

-80

I

0

E -20 -40

-GO

-

-

Polygonuni kelloggii

-

-

-80

Fra. 17. (A) The theoretical relationship between B P and P m i n aa drought increases. (Reproducedwith permission from Waring and Cleary, 1967. Copyright 1967 by American Association for Advancement of Science.) (B-F) The actual measured relationship between B P and aa drought increases (during July and August) for five different species. (Reproduced with permission from Hickman, 1970.)

librated plant-soil water potential and, therefore, gradually decreases as drought persists. As water stress builds up in the foliage, stomata1 restriction of transpiration tend8 to increase until, at some level of BP,BP- Pmin = 0.

208

QARY A. RITCHIE

and

THOMAS M. HINCKLEY

Hickman (1970) measured BP and Pminon a wide range of plant forms in the southern Oregon Cascades throughout the summer of 1967 during which no rainfall was recorded; hence drought gradually increased ($80ii gradually decreased). Thus Hickman was, in effect, testing the relationship proposed by Waring and Cleary. He found that five seasonal patterns persisted among the 44 species tested, although there were some intermediates (Fig. 17B-F). The most common pattern is typified by the curves from Polygonum cascademe, an annual endemic (Fig. 17B). Hickman suggested that this pattern is probably typical of most plant species in areas with modified Mediterranean climates. A more unusual pattern was evident in the curve for Ribes binominatum, an inhabitant of forests and dense meadows (Fig. 17C). The curve typified by Lotus nevadensis var. doouglasi showed no change in BP or Pminthroughout the season (Fig. 17D). Neither did that of Sedum stenopetalum, a succulent. Polygonum kelloggii, an ephemeral annual of vernal-wet open ground, displayed the curves shown in Fig. 17E. The plants desiccated and died rapidly when a critical Pmlnwas reached. Nimulus breweri is even less longlived (Fig. 17F). Hickman discusses these patterns and draws an admittedly arbitrary distinction between “conformers” (Fig. 17B, E) and “regulators” (Fig. 17C, D, F), the degree of conformity increasing with the negative slope of the Pmln curve. I n conformers, B P and Pmlnreflect $soil and E respectively, due to a relative inability to control water balance. Hence the reported good agreement between calculated and measured P values for Pacific silver fir and noble fir reported by Hinckley and Ritchie (1973) indicates that these species are conformers under conditions of low to moderate soil water stress. On wet sites conformers are able to maintain moderately high P levels, but on drier sites they tend to follow one of two strategies. Physiological adaptations allow some species to sustain low P values with minimal tissue damage (e.g. Polygonum cuscadense), while others complete their life cycles early in the spring before reaches intolerably low values (e.g. Polygonum kelloggii). Most annuals, as well as many perennials, are conformers, generally having only moderately well-developed root systems and lacking xeromorphic characteristics. Regulators are able, by various means, to regulate P levels despite soil or atmospheric stresses. Such species are represented by Fig. 17C, D and F. In Ribes binominatum, BP reflects decreasing but Pmin remains essentially constant. This possibly reflects good stomata1 control and an efficient root system. Polygonum kelloggii and Mimulus breweri have similar morphologies but very different strategies. Both are conformers in the sense that they have adapted phenologically to drought, but Mimulzcs is more able to regulate Pminand B P up to a

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

209

critical point, whereas Polygonurn is an extreme conformer up to the same point. From these preliminary indications, it seems that the curves reported by Waring and Cleary (1967) may not be common in nature. However, the usefulness of B P and Pmin,which they were early to recognize, has been substantiated, leading to wide employment of these concepts in numerous studies (e.g. Cleary, 1968; Cary and Wright, 1971; Oechel et al., 1972a; Sucoff, 1972; Haas and Dodd, 1972; Lassoie and Scott, 1972; Schulze et al., 1973; Rundel, 1973; Griffin, 1973; Halvorson and Patten, 1974). b

0

-5

1

'.

20

-?

'. 40

'..

I

El

I.

M.b.

I I

'..

'

60

FIG.18. A graphic portrayal of five species' responses to depletion of soil moisture at relatively constant atmospheric evaporative demand (only clear, warm days). B P is P measured before dawn where B P wm assumed t o be a reflection of soil moisture content, DPmaX is the maximum diurnal depression of P as shown in Fig. 16. L.d. is Larrea divaricata (Oechel et al., 1972b), R.b. is Riber, binominaturn, M.b. is Mimulua breweri, P.c. is Polygonurn cascadense, P.k. is P . kelloggii (Hickman, 1970).

Another method of graphically portraying species responses to environmental water regimes would be to consider the relationship between BP and DPm""across a gradient of soil moisture conditions. In Fig. 18, five species have been so characterized. Polygonum cascadense and P . kelloggii, two of Hickman's conformers, show a negative slope, while the regulators Ribes binominaturn and Mirnulm breweri show positive slopes. Larrea divaricata, a widely distributed desert shrub, is a regulator, but is also able to tolerate severe levels of water stress (Oechel et al., 1972b). It is tempting to compare species based upon these curves using both the slope and intercept as indicators of species differences. Such comparisons may or not be warranted, however, because the pressure chamber as used in these studies provides only an estimate of water

210

GUY

A. RITCHIE

and

THOMAS M. IIINCKLEY

potential. Therefore, species comparisons may require calibration with the psychrometric technique. Also, pressure chamber values can be affected by the tissue on which the measurement is made, as discussed in Sections I1 E8 and I1 F. Nevertheless, such curves do suggest tendencies and may be a starting point for classifying vegetation types relative to their drought adaptability. Honey mesquite, reported to be an extravagant user of readily available soil water, persists in semi-arid areas where rainfall may be less than 16 mm per year. Haas and Dodd (1972) meaeured diurnal patterns of P in this species at approximately weekly intervals from T a m 111 Results of a d y & for four multiple regreaaiona comparing aix variablee and three effect8 for four expeaaims of water strea8 (see FG. 16). (Summarized from Haas and Dodd, 1972.) Variables

BP

Leaf exposure vpd at max T soil T (30 cm max) soil H,O (% max) wind speed (max) C solar rad. (ly day-')

-

Coef. of determ. (P)

**

** **

CP

XDP

Pmin

**

**

-

-

** ** **

-

**

-

0.64

0.67

-

-

** ** **

0.72

0.74

-

**

**

-

indicates significance at the 1% level

April to October to determine the influence of selected environmental variables on P . From the diurnal values, they computed the integrated total daily P (W), BP and net P (ZDP; Eqn (15)). These values were then compared to measured environmental variables with multiple regression techniques (Table 111). They found XDP to be generally more highly correlated with environmental factors than the other expressions of P . Some of their other results, however, are surprising. BP was not significantly correlated with soil moisture determined gravimetrically adjacent to the study trees, but was with soil temperature and vapor pressure deficit (vpd) at maximum daily air temperature. Sucoff (1972) also found poor correlation between BP and soil moisture in a red pine plantation until approximately 70-80% of the available soil moisture in the root zone was depleted. Following this was a nearly linear relationship between soil moisture and BP. R. H. Waring and Betty Klepper (pers. comm.) and Lawlor (1973) suggest that soil moisture

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

21 1

may be depleted initially in zones of high $soil, depletion then proceeding until all zones are essentially depleted. At this point, BP decreases substantially. The physical relationship between BP and soil temperature involves root resistances to water flux and will be discussed later. The high correlation with vpd measured later in the day is difficult to explain and may underscore the inability of regression models to represent a system as complex as the soil-plant-atmosphere continuum. Nevertheand Z P as expressions of diurnal and seasonal less, the usefulness of Pnet plant water status as it relates to environment is apparent.

2 . Concept of limiting conditions Elfving et al. (1972) proposed a model of the dynamic soil-plantatmosphere continuum based on the premise that the system may be adequately visualized as consisting of a series of steady states. Accordingly, the steady state flux of water through the system is assumed to be directly proportional to the water potential gradient and inversely proportional to the sum of the flow resistances: flux =

$a011

- $leaf

(16)

raoil to leaf

and, rearranging: $leaf

=

$soil

- (flux)* soil to leaf)

(17)

Hence as $soil decreases, or flux or resistance increases, should become more negative (Hanan, 1972; Hailey et al., 1973). Figure 19 shows results of studies testing the relationship between $L (P was measured in the field and then converted to $I,using P vs. $L calibration curves) and flux during non-limiting environmental conditions on “Valencia” orange trees. The data indicate that $L is specifically related to water flux in a curvilinear fashion. The fact that data from widely different climatic areas, soil types, different aged and sized trees and different seasons did not vary in this relationship, lends confidence to the validity of the model. I n this case, non-limiting conditions were defined as: $aoll > - 0.3 bars; soil temp. > 16°C. When the relationship between flux and 41, was or soil temperature, $I,was always tested under conditions of lower lower than predicted. This departure resulted from modification of flow resistance caused by lowered soil temperature, and by modification of both resistance and water potential in the soil by decreasing soil moisture. Elfving et al. (1972) visualize this model as a useful tool for defining base-line conditions with which to compare $I,values and environ-

GARY A. RITCHIE and THOMAS M. HINCKLEY

212

mental variables. I n “Valencia” orange trees, +L deviated from predicted values when tensiometer readings went below -0.3 bars and soil temperature fell below 15°C. The authors tentatively use these values to delineate limiting edaphic conditions for this species. Presumably this model should be applicable to other species. Once 2

I “

1

I

I

va 1e nc ia ” orange -8.28 - 9.23 log (x+ 0.5)

1I

1

r=-0.92

G h

m

k

cd

P I

I 14

t

.Desert *Intermediate OC 001, Coastal

--

\ \

1

0

Vpd’RLeaf

(Hgmm . s

*

I.

2 -1

.cm)

FIG.19. The relationship between I/JL and the ratio of vpd/RL under non-limiting environmental conditions ( I/Jsoil > - 0.3 bar; soil temp. > 15OC). Dashed lines represent 95% confidence limits. Data from three different climatic regions are shown. (Reproduced with permission from Elfving et al., 1972.)

the $L vs. flux relationship has been established and non-limiting conditions defined experimentally, the model could be used to “diagnose” a soil-plant-atmosphere continuum to determine whether conditions are limiting or non-limiting. This approach could also be used to compare species’ responses to environmental stresses.

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

213

3. Reference-species We have seen how interspecies comparisons can be made by ascertaining behavioural differences between species coexisting in similar environments; now we ask if it is possible to characterize different environments in terms of the responses produced in one “referencespecies”. The reference-species concept based on pressure chamber measurements was introduced by Waring and Cleary (1967) and later expanded by Cleary and Waring (1969), Waring (1969a), Waring and Youngberg (1972) and Waring et al. (1972) and used by Rundel (1972). The principle is straightforward: a given site or environment is defined based on the response it produces in one selected species. A suitable reference-species must be (1) responsive to water supply and demand, (2) widely distributed and (3) perennial. Clearly, a species which shows little or no predictable response to a changing environmental water regime would be of little or no value as an indicator of water regimes. I n order to make comparisons over a wide range of habitats using one reference-species, it is necessary that this species occur in each of the habitat types in question. Therefore, it must be widely distributed edaphically and geographically. It is also desirable to characterize sites at various times during the year; thus annual plants are of limited utility. Waring and Cleary (1967) selected Douglas-fir and Shasta fir (Abies magni$ca var. shastensis) as reference-species in their studies of environments in the Siskiyou Mountains of southern Oregon and northern California. They noted that saplings 1-2 m tall were more sensitive to moisture stress than larger trees and P varied little within crowns. Results comparing BP at the peak of summer drought for a variety of sites are given in Table IV. BP at peak of drought was not related to elevation, but was strongly related to stand density. The authors point out that stands growing on very shallow soils at timberline may be under considerable water stress (Stand 8), at the same time that nearby stands on deeper soils are subject to little water stress (Stand 1). Waring (1969a)used this approach in combination with an assessment of optimum air and soil temperature for Douglas-fir seedling growth (optimum temperature day, OTD) (Cleary and Waring, 1969), established by field and laboratory studies, to develop a two-dimensional grid of environments of the eastern Siskiyous (Fig. 20). On the basis of this grid, Waring et al. (1972) recognized seven somewhat overlapping environment-vegetation types, each characterized by a specific range of OTD’s per year and BP at peak of drought. Each type is associated with a rather distinct dominant and subordinate flora and is often II

214

GARY A. RITCHIE

and

THOMAS M. HINCKLEY

related to soil conditions. This overall concept was later integrated into a Biome model (Waring et al., 1972) wherein reference-species are used to characterize environments. 4. The stress-day index The stress-day index ( S D I )is defined aa the plant water stress that is responsible for a reduction of yield below the potential for that plant or crop, and was developed by Hiler (1969) and Hiler et al. (1972) as a TABLEIV Siakiyou oegetation in relation to internal water stresa of two reference-species (Douglas-fir, Shastafir)at peak of drought ( 1Sept. 6 6). (Reproducedwith permission from Waring and Cleary, 1967. Copyright 1967 by the American Association for the Advancement of Science.) Stand No.

Elevation

1 2 3 4 5 6 7 8 9 10

2040 1400 1920 1680 760 1500 1280 2130 1700 790

(4

Vegetation

B P at peak of drought ( -bars) 3.0 6.5 7.0 11.0 12-5 14-0 17.5 19.0 19.0 28.0

BO, black oak (Quercw, kelloggii); DF, Douglas-fir (Paeudotsuga menzieaii); ES, Engelmann spruce (Picea engelmannii); IC, incense cedar (Libocedmce decurrena); JP, Jeffrey pine (Pinus jeffreyi); MH, mountain hemlock (Tauga mertensiana); PP, ponderosa pine (Pinua ponderosa); SF, Shaata fir (Abiea magnifica var. shaatenai8); SP, sugar pine (Pinw, lambertiana); WF, white fir (Abiea concolor);WP,white pine (Pinua monticola); Y, yew (Taxua brevifolia)

means of establishing irrigation schedules. SDI is composed of two elements: a crop susceptibility factor (CS)and a stress-day factor (SD), which is a measure of the degree and duration of plant water deficits, hence

SDI =

2 (CStxSDt)

n=1

(18)

where i designates the growth stage from I to n and n is the number of growth stages considered. CS must be established experimentally for each species. Hiler et al. (1972) conducted numerous tests with southern pea (Vigna sinesis var. Burgundy) on the effects of water

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

216

stress levels (measured with a pressure chamber) on growth and yield at different growth stages according to

CS,, =

X-ia

X

where CS, is the susceptibility factor at growth stage I and treatment a, in which a specified level of P was allowed to develop, x is the growth or yield of fully irrigated controls, and ia is the growth or yield obtained

I 10

20

30

BPATPEAKOFDROUGHT (-bars) FIG.20. Distribution of the major forest vegetation types in the eastern Siskiyou Mountains of Oregon, based on moisture and temperature gradients. Mixed conifer type consists of white fir, Douglas-fir, sugar pine and ponderosa, pine. (Reproduced with permission from Waring, 1969a.)

from treatment a with stress developing at growth stage I only. Thus the susceptibility factor quantifies the species’ sensitivity to a given water deficit at a particular stage in its development. Production in pea plants, including fruits, was closely related t o S D I , and SDI readily allowed for determination of an irrigation schedule which would minimize water use and maximize production. Such an approach would seem to have value not only in crop studies, but also in ascertaining the dependency of wild annual vegetation on water

216

GARY A. RITCHIE

and

THOMAS M. RINCKLEY

status. The relationship of such phenological events as onset of height growth, flower initiation, seed maturation and dispersal (all critical in the competitive interactions of pioneer annuals) to plant water status could be quantified, and species comparisons made as an aid in interpreting the dynamics of early, secondary plant succession (see Section I11 E 5). D. P

I N RELATION T O HABITAT

Scholander et al. (1965a) made numerous measurements of P in plants growing across a wide range of habitats and they recognized that these values could be used to stratify communities according to plantenvironment water stresses (Fig. 21). The highest (least negative) values were generally found in plants growing in or at the edge of fresh water. Forest plants of both the understory and overstory were the next highest, while plants of the seashore consistently had values from - 35 to -60 bars (T of sea water N -25 bars). P of desert plants ranged from higher than -20 bars for “tank” plants such as cactus (Opuntia tejaho), which can store water, and for plants such as cottonwood and sycamore which inhabit desert washes, to lower than -80 bars for those such as creosote bush and juniper growing in apparently more arid regions. Despite limited information on the time and conditions under which measurements were made and limitations in interspecific comparisons using P values, the consistency of their data is striking. There have been numerous subsequent studies of xylem pressure potentials in relation to habitat. I n some cases these investigations have tended to confirm the earlier findings of Scholander et al. (1965a) and Waring and Cleary (1967), but in other studies definitive trends were not always established. Habitat studies have been conducted in salt water environments (Scholander et al., 1964, 1965a, 1966; Teal and Kanwisher, 1970), in fresh water environments (Scholander and Perez, 1968; Dickson and Broyer, 1972; Small, 1972) and in xeric environments (Detling, 1969; Love and West, 1972; Kappen et al., 1972; Oechel et al., 1972a, b; Sankary and Barbour, 1972a, b; Campbell, 1973; Detling and Klikoff, 1973; Halvorson and Patten, 1974). Investigators have frequently used measures of P to differentiate between environments or to describe the distribution of different species within an environment. Some have been successful (Waring and Cleary, 1967; Dickson and Broyer, 1972; Sankary and Barbour, 1972b; DePuit and Caldwell, 1973; Wambolt, 1973) while others have not (Kuramoto and Bliss, 1970; Love and West, 1972; Small, 1972; Dina et al., 1973; Griffin, 1973). The varied ability of pressure chamber values per se to distinguish

217

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH P I

0

n Y 0

I

P

0

I

l

a, 0

D -4

Rhuop hor o Avicennia Laguncularia

s

Bolts Dislichlis Salicornio

-%.

% -,;

Cotton wood Sycamore Desert witlo w Smoke tree Mesquite Tomaris k OCOtlllO Solvia Ence Ii a Cots claw A triplex Creosote bush Juniper

7 4

0 7r

Douglas fir Redwood Hemlock Maple I Maple Il Darlingtonla L a d y slipper Thuja OXOliS

Dryopteris Pteris Blechnum Adiantum

Spiraea Dogwood Solix Myric a Comarum Hemlock Menyonlhes Ver o n ico Polygonum

FIG.21. Xylem pressure potentials (P)in a variety of flowering plants, conifers and ferns. Most measurements were taken during the daytime in strong sunlight. Night values in all cases were apt to be several barn less negative. (Reproduced with permission from Scholander et a,?.,1965a. Copyright 1965 by American Association for Advancement of Science.)

218

GARY A. RITCHIE

and

THOMAS M. HINCKLEY

between habitats in these studies may be due to a number of factors. Differences in soil characteristics and stand density can often obscure topographic moisture gradients. Griffin (1973) demonstrated that water stress in some California oaks was higher on densely vegetated north-facing slopes than on sparsely vegetated ridgetops and upper south-facing slopes, and that the presence of deep groundwater reserves had a profound influence on P values. I n addition, he indicated that under dry conditions, seedlings and saplings showed lower xylem pressure potentials than dominant trees, apparently due to different rooting effectiveness. Dina et al. (1973) also emphasize that moisture stress patterns alone may not be sufficient to determine vegetational distribution and habitat adaptation because of wide ranges of interspecies tolerances to water stress. Finally, the validity of making direct species comparisons based on non-calibrated P values must again be seriously questioned.

E. P

I N R E L A T I O N TO S O M E P L A N T FAUTORS

Until now we have dealt primarily with the relationships between P and environmental factors and have found that often such comparisons alone do not successfully explain distributional patterns of plants or plant communities. This approach does not account for the fact tha€ different species have varying levels of physiological tolerance to water stress and that various physiological processes are affected by water stress to differing degrees and through different mechanisms. Hence, to have ecological significance, pressure chamber data should be accompanied by an understanding of the influence of P on pertinent facets of the physiology of the species in question. I n the discussion to follow, we will examine some relationships which have been observed between P and such physiological factors as stomatal activity, transpiration, net assimilation, growth etc., in an effort more clearly to interpret the influence of water stress on plant distribution. We have confined our review to those studies where the pressure chamber has been employed to assess plant water stress for two reasons: (1) a complete review of the effects of plant water potential on physiological processes is far beyond the scope of this report and (2)) as we have shown, P and water potential are not necessarily the same quantity. 1. P in relation to stomatal activity The dependence of stomatal function on leaf water balance is well known (Heath, 1959a, b; Meidner and Mansfield, 1968)and is illustrated in Fig. 22. Some pertinent questions to which the pressure chamber

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

219

technique may be applied include whether the P vs. stomata1 resistance relationship is a threshold, linear or curvilinear function. If a threshold function exists, what is the critical P value! How do these relationships vary with species, plant material and environmental conditions? How does the history of a plant affect its response to stress conditions? Such questions have been addressed, directly or indirectly, in a

ATMOSPHERIC hIOISTURE C O N T E h T (X,)

Tr

:x = 2 3 Transpiration F l u x

+ RL 0

LEAF WATER CONTENT (X,)

-x

P

Flux X * X = 1' 2 Absorption by

P

roots & storage LWC

0

SOIL AND PLANT WATER CONTENT (X,)

FIG 22. A generalized flow diagram showing the interdependence of water fluxes throughout the soil-plant-atmosphere continuum and the xylem pressure potential ( P ) and leaf diffusive resistance (RL). Small graphs depict the typical nature of relationships between elements of the diagram. These functions are not always linear or curvilinear as shown, and often vary between species and with tissue age or environmental conditions (see text).

number of studies. Duniway (1971b) compared RL and P in tomato plants and found a strong curvilinear relationship from - 4 to - 13 bars (Fig. 23). A similar relationship was noted in yellow poplar (Liriodendron tulipifera) by Richardson et al. (1973) and in pea by Clark and Hiler (1973). The same comparison for tea plants (Carr, 1971/1972) and tomato and black locust (Robinia pseudoacacia) (Hinckley, 1973) yielded a linear correlation. Kriedemann and Smart (1971) observed a threshold relationship between P and R L with Vitis vinifera and orange

220

GARY A. RITCHIE

and

THOMAS M. HINCKLEY

(Citrus sinensis) where no change in RL was observed until P reached a certain value. Threshold relationships (Kaufmann, 1968a; Turner and Waggoner, 1968; Lopushinsky, 1969; Hiler et al., 1972; Kassam, 1973; Hellkvist et al., 1974) seem to be more common than linear or curvilinear relationships. Caution must be exercised, however, in attempting to make direct comparisons of P between species and between different growth stages

4

G

8

10

12

14

P(-hrs) FIG.23. The diffusive resistance of the adaxial and abaxial leaf surface in tomato plotted as functions of decreasing leaf water potential ( P ) .(Reproduced with permission from Duniway, 1971b.)

and tissues within the same species, because anatomical differences can influence pressure chamber values (Kaufmann, 1968a; Ritchie and Hinckley, 1971).Waggoner and Turner (1971) observed that no change in RL occurred in one-year-old red pine foliage as P decreased from - 4 to - 15 bars. However, RL in newly emerged needles changed from 7.5 to 15.0 sec cm-1 over the same range. Carr (1971/1972)found different P vs. RL curves for tea plants from different seed sources. In addition, Hellkvist et a2. (1974) point out that site quality can appreciably influence xylem resistance which indirectly affects balancing

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

221

pressures (Fig. 22). Further, preconditioning or history can affect the P vs. RL relationship (Hinckley, 1971; Jordan and Ritchie, 1971; Hinckley and Ritchie, 1972) as well as whether the P vs. R L relationship was generated under controlled or field situations (Jordan and Ritchie, 1971; Ritchie, 1973; Turner and Begg, 1973). Environmental conditions can also affect the P vs. RL relationship (Schulze et al., 1973). Four desert species, Zygophyllum dumosum, Artemisia herba-alba, Hammuda scoparia and Reaumuria negevensis, and one cultivated tree, Prunus americana, were subjected to a range TABLEV Approximate values of P at which stomata1 closure occurs aa inferred from plateaux in diurnal P curve%from clear days when soil moisture waa not limiting (see Fig. 15) Critical P (-bars)

Species

Experimental conditions

source

~~

Ponderosa pine Grape vine Apricot Pear Red pine Honey mesquite Honey mesquite Engelmann spruce Douglas-& Valencia orange Apple Apple Douglas-fir Douglas-fir White oak Corn Sorghum Tobacco Cotton

15 17.5 16.5 17 18 27 26 13.5 13 17.5 19 24

16 18 26 17 19 13 <28

seedlings-field field trees-field trees-field treesplantation field (sun foliage) (shade foliage) seedlings-field seedlings-field trees-field field (July) field (Sept) treep-stand tree tree-stand field field field field

Cleary (1968) Klepper (1968) Klepper (1968) Klepper (1968) Sucoff (1972) Haas and Dodd (1972) Haas and Dodd (1972) Jones (1972) Jones (1972) Elfving and I(aufmann (1972) Goode end Higgs (1973) Goode and Higgs (1973) Zaerr (1971) Waring (pers. comm.) Hinckley (unpubl.) Turner (1972) Turner (1972) Turner (1972) Jordan and Ritchie (1971)

of air temperatures (25 to 40%) and water stresses. At low water stress (as measured by the pressure chamber) RLdecreased with a temperature increase in all species if vapor pressure gradient between mesophyll wall and ambient air was held constant. A t high water stress, the relationship was reversed, but the P value at which the reversal occurred varied between species. By using the relationship illustrated in Fig. 15, which gives diurnal P curves from clear days when soil moisture is not limiting, it is possible to infer from plateau values the P levels at which stomata1 closure occurred. Table V is a compilation of such values taken from the literature. These data are relatively consistent with the experimental

222

O

~ A.Y RITCHIE

and

THOMAS M. HINCKLEY

data of Lopushinsky (1969), whose range of critical P values varied from - 14.6 to - 25.1 and the median - 16.9 bars. The range was from - 14.6 to - 18.0 bars for pines and Engelmann spruce. For Douglas-fir, however, the range was much broader. From this Table, there seems to be little or no consistent relationship between critical P values and the habitat preference of the species tested. The theory that stomata of drought-resistant species tend to close a t higher P values (more sensitive to water stress) is not strongly supported by these data. From the limited evidence presented, it seems that the correlation between P and stomatal activity (expressed as RL)can be linear, curvilinear or threshold, depending upon species, type of tissue and environmental conditions. Threshold values apparently vary widely between species and probably between growth stages and sites. Therefore, a great deal of caution must be used in interpreting plant adaptation or distribution on the basis of only limited data.

2 . Transpiration, sap velocity and stern diameters Through the effect of P on stomatal behaviour, an indirect but profound effect can be exerted on other aspects of water transfer within the plant system (see Fig. 22). Transpiration rates, for example, should be closely related to P,if P is closely related to stomatal aperture, which seems to be the case in many species (Kochenderferand Lee, 1973). The relationship between P and transpiration has been described for a number of species from a diversity of habitats. These include Atriplex confertifolia and Eurotia lanata (Moore et al., 1972), southern pea plants (Hiler et al., 1972), “Valencia” orange trees (Elfving et al., 1972), balsam fir (Abies balsarnea), subalpine fir ( A . lasiocarpa), grand fir ( A . grandis) and Pacific silver fir (Puritch, 1973), Pacific silver fir (Hinckley and Ritchie, 1970, 1972) and noble fir (Hinckley and Ritchie, 1972). The P vs. transpiration relationship was closely linked in these studies; however, field results with a Pacific silver 6r tree indicated that P was not only influenced by transpiration in nearby foliage, but also by transpiration in other parts of the crown (Hinckley and Ritchie, 1970). The possibility of internal redistribution of water was suggested. Reasoning that transpiration rates should be correlated with sap velocity ( S V ) in the stem xylem, Hinckley (1971) and Hinckley and Ritchie (1972) used a modified heat-pulse velocity technique (Skau and Swanson, 1963) to measure SV in Douglas-fir, Pacific silver fir and noble fir saplings growing in a controlled environment. Under such conditions, good correlations were obtained between transpiration and S V and between SV and P. Field studies have produced a predictable relationship between P,S V , RL (or transpiration) and microenvironment under conditions of high &n,; however, with low t,bSoil, the

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

223

relationships were poor (Hinckley and Scott, 1971; Lassoie and Scott, 1972; Swanson, 1972; Lassoie, 1973). Comparisons between stem circumference or diameter (SD)and P have been a productive area of study. Indeed, with the use of sensitive linear variable displacement transducers ( L V D T ) to measure SD,a continuous, accurate estimate of P may be possible (Namken et al., 1969). Early comparisons between P and SD in red pine and Callitris 0

I

I

I

I

I

Day 180

5

Day 230

-

7-

-

20 I

18

30

&

0-0-’0--0 11 13

12.3

I

15

11

13

/w I

I

I

12.7

I

d

13.1

Stem Circumference (mm) FIG.24. Hourly diurnal changes in xylem pressure potential ( P )and stem circumference on a “wet day” (day 180) and a “dry day” (day 230) for a 19 m white oak tree. P was measured on four individual leaf petioles and averaged. Numbered points indicate the hour (in solar time) of measurement. (Hinckley and Bruckarhoff, unpubl.)

cupressiformis under controlled environmental conditions yielded a strong linear correlation (Worrall, 1966). Field examinations in such species as cotton, red pine, Douglas-fir and white oak have indicated a more complicated relationship: ( 1 ) a lag between SD and P existed (Namken et al., 1969, 1971; Zaerr, 1971; Waggoner and Turner, 1971; Worrall, 1973), (2) a hysteresis loop was observed between stem circumference and P (Fig. 24) (Namken et al., 1971; Klepper et al., 1971; Hinckley et al., 1974), (3) P reached a critical value before stem dehydration and shrinkage occurred (Namken et al., 1971; Jordan and Ritchie, 1971) and (4) stem dimensional changes did not occur in mature xylem, but were confined to the phloem, cambium and region

224

GARY A. RITCHIE

and

THOMAS M. HINCKLEY

of immature xylem cells (Dobbs, 1966; Molz and Klepper, 1973; Molz et al., 1973). A theoretical basis for this hysteresis effect has been provided by Molz and Klepper (1972, 1973) using a passive diffusion analysis. Their calculated values compared closely with values measured on a cotton plant, suggesting that the effect is at least partially a function of the diffusion processes between phloem and xylem. They also postulated that transpiration-induced changes in leaf water potential may not be instantly propagated to the water potential of stem xylem, supporting findings of Waggoner and Turner (1971) but conflicting with predictions of Spomer (1968). Molz and Klepper (1972) suggested using the relationship between S D and P as a means of continuously monitoring P.

3. Net assimilation, growth and yield Growth depends upon cell division and elongation and ultimately on a supply of carbon compounds from the photosynthetic tissues; both processes are intimately tied to the plant’s water balance (Hsiao, 1973). Measurements of #Boll are not suEcient to determine effects of water supply on plant processes (Kramer, 1969); thus a direct measure of water status of the plant itself is essential. Therefore, we feel compelled to discuss a t least briefly the relationship between P and these processes, especially in light of the demonstrated applicability of the pressure chamber to field or ecological research. We do not, however, intend to review the entire field of water deficits and plant growth, and will confine our observations to studies in which the estimate of plant water status has been based upon pressure chamber measurements of P. Certain questions are of immediate interest: (1) how are P and growth or yield related, (2) how are P and net CO, assimilation rate ( N A R ) related, and (3) how can these relationships be applied to studies of plant adaptation and distribution? Growth of ponderosa pine (Waring, 1969b), red pine (Sucoff, 1972), Douglas-fir (Cleary, 1968), cotton (Fig. 25; Jordan, 1970) and the growth and yield of pea (Hiler et al., 1972) were reduced as P decreased. Cell elongation and enlargement, dependent upon adequate levels of positive turgor pressure, were far more sensitive to decreasing P than cell division (Meyer and Boyer, 1972; Hsiao, 1973). I n yellow poplar, a reduction in P from - 6 to - 7 bars caused a 70% reduction in stem elongation (Doley, 1970) while some cell division occurred a t - 18 bars in Douglas-fir (Waring and Cleary, 1967). Cellular processes, such as cell division and protein synthesis, which are probably not affected directly by changes in turgor but which are dependent either directly or indirectly upon a large number of other processes, vary in their sensitivity to decreasing P. These other pro-

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

225

cesses include net carbon fixation, storage, conversion and translocation, relative rates of enzymatic and nucleic acid synthesis and degradation, membrane structure and permeability and cytoplasmic viscosity. Hence relationships between P and physiological processes may or may not be apparent. It is tempting to interpret direct relationships as 0

c

10

L

0

9 I

\

a

\

20

-

30

2

8

I

I

I

1

E

7

VI) 0 0

?!

E

4

c

r

F

I"

0

..-.

N

- 200 E

u

0

: E 0

e

100

L

Stress period (days)

FIG.25. Effect of the length of a stress period on daily maximum xylem pressure potential ( P ) ,height increase, and leaf area increase of cotton seedlings. Control plants ( 0 )were watered daily while stressed plants (0) were deprived of water beginning at day zero. The vertical lines represent the standard error of the mean of 10 determinations. (Reproduced with permission from Jordan, 1970.)

causative when at best they may be only correlative. The measurement of several processes simultaneously with decreasing P may well help to establish causative relationships. Brix (1972) showed that N A R in Douglas-fir was closely related to P and that this relationship was independent of the nitrogen nutrition of the needles. At - 10 bars, N A R began to decline and completely ceased at - 35 bars. Work with the vine Vitis vinifera (Kriedemann and

226

GARY A. RITCHIE

and

THOMAS M. HINCKLEY

Smart, 1971), with balsam fir, subalpine fir, grand fir and Pacific silver fir (Puritch, 1973) and with black locust and tomato (Hinckley, 1973) has indicated that parallel decreases in N A R and P were strongly linked to changes in RL. However, changes in N A R in pear (Pyrus communis) (Kriedemann and Canterford, 1971) and in Pacific silver fir and noble fir (Hinckley and Ritchie, 1972) did not always coincide with changes in RL, thereby suggesting the influence of a mesophyll and/or carboxylation resistance. Recent work by Brandle et al. (1973) represented an effort to link decreases in P to changes in NAR, dark respiration, the ratio of polysomes to monomer and levels of ribonuclease in black locust. Though definitive relationships were not clearly established, it became evident that species and processes vary considerably in their reaction to plant water deficits. Research on desert species has indicated an ability to carry on positive N A R even at extremely low values of P . Atriplex hymenelytra (Pearcy et al., 1971), Larrea divaricata (Oechel et al., 1972a), brigalow (Acacia hurpophylla) (van den Driessche et al., 1971) and Artemisia herba-alba (Kappen et al., 1972) carried out positive N A at - 45, - 55, - 6 0 and -100 bars, respectively. Such data also suggest an ability of xerophytes to assimilate at lower P values than other plant species. This was tested by Dins and Klikoff (1973) in several streamside and scrub oak communities in Utah. Inhabitants of the xeric oak community had mean seasonal P values lower than those of the mesic streamside habitat. With box elder (Acer negundo), a mesic species, a predicted inability to assimilate CO, at lower P values was evident. For the other species tested, however, the relative N A R and dark respiration response bore little relation to their habitat preferences. Such evidence is surprising but certainly not conclusive. Attempts at modeling N A R based upon measurements of P have met with some success. Helms (1972) used P as an index of environmental stress in his studies of photosynthesis in ponderosa pine. Richardson et al. (1973) developed a linear regression model to predict N A in yellow poplar. Many variables were used in the model including stomata1 resistance, photo-pigments, nitrogen and radiation. He was able to obtain accurate predictions only when the model incorporated some index of water status such as P. Future models will hopefully proceed beyond multiple regression techniques and strive for recognition and delineation of plant mechanisms and plant-environmental interactions (Reed and Waring, 1974).

4 . Studies with fruit The pressure chamber has been used to work on problems in pomology or studies of fruit. Kaufmann (1969), searching for an expedient field

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

227

method of estimating P for irrigation purposes in orange groves, noted that when orange fruits are cut halfway through a plane perpendicular to the axis, a gap forms due to turgor within the orange. He measured the widths of these gaps and compared them to concurrent pressure chamber measurements from adjacent branches. He found a close curvilinear relationship between slice gap width and P from - 5 to - 15 bars. McMichael et al. (1973) examined the effects of P on abscission of leaves and bolls of cotton plants in the greenhouse. Abscission rates of both structures increased linearly as P decreased from - 10 to 24 bars, and younger bolls were apparently more sensitive to water stress than older bolls. Additional studies by McMichael et al. (1972), Jordan et al. (1972) and Lipe and Morgan (1973) demonstrated that decreased P was associated with a stress-induced hormone imbalance involving ethylene, which may trigger the abscission process. Early studies by Hodgson (1917), Bartholomew (1926) and others with excised fruit-bearing branches suggested that fruits may act as reservoirs supplying leaves with water during transpirational periods. Klepper and Ceccato (1968) suggested the use of the pressure chamber directly on fruits, and Klepper (1968) successfully produced the first data with pears. The technique was used to test the above suggestion. Klepper found that diurnal changes in Plea,(P,) and Pfruit (P,) on orchard pear trees were closely related, and that changes in Pf were correlated with changes in fruit diameter. Pears which were coated with vaseline to inhibit transpiration showed little difference in diurnal P curves from uncoated pears. This was taken as evidence that water leaving the fruits had apparently passed through the petiole. P measured in leaves adjacent to fruit, however, was not significantly different from P measured in isolated leaves. Elfving and Kaufmann (1972) measured changes in “Valencia” orange diameters relative to plant water relations and environmental factors. From their experiments they concluded that the dependency of P, on P,may be more apparent than real because transpirational losses of water from the two organs are affected by the same environmental factors. They were unable to demonstrate unequivocally that fruit can act as a midday reservoir for transpiring leaves.

-

5. Phenology The importance of recognizing and accounting for phenological development in plants in relation to ecological studies has been emphasized by Lieth (1970). Phenological studies delimit periods of the year and developmental stages of plants during which environmental stresses such as water deficit are most or least critical. There has been surprisingly little use made of the pressure chamber

228

GARY A. RITCHIE

and

THOMAS M. HINCKLEY

to describe water relations during specific growth processes such as flowering, stem growth, fruiting etc. Cleary (1968),in some observations on Douglas-fir seedlings, found that times of bud burst and bud set occurred at widely differing levels of BP and Pmin. Bud burst occurred earlier in the spring when moisture supplies were apparently high. Hiler et al. (1972) divided the growth of southern pea into three periods: (1) vegetative, (2) flowering and early pod formation and (3) pod development. Experiments showed that the species was most sensitive to low P during the flowering and early pod formation stages and least sensitive during the pod development stage. Waring et al. (1972) characterized the water regimes of three types of forest communities using a reference-species (young Douglas-fir). TABLEVI The effect of site moisture regime on time of leaf development and on B P at the time of full leaf development and ceasation of cell division in Dough-jir seedlings growing on three aitea. The spmcce type haa the lowest annual phnt moieture stress, the oak type the higheat (after Waring et al., 1972)

Type ~

Days since bud swell

2 4

0 0

5

0

Full leaf development BP Days (-bars) 31 37 40

Cessation of cell division BP Days (-bars)

3

*

5 9

93

~~

Spruce Pine Oak

*

Initial BP ( -bars)

80

~~

18 18 18

Never reached.

Some of their observations dealt with phenological events. For example, they found that the time interval required for full leafing-out to occur wm greater and the BP level was lower in drier habitats (oak type) than in more mesic habitats (spruce, pine types) (Table VI). I n some ephemeral species, desiccation and death appear to occur when P reaches a critical level (Hickman, 1970). Polygonurn kelloggii populations vanished within a day or two following attainment of a mean daily P level of about - 18 bars. When DPmax reached - 10 bars, plants of Mirnulus breweri desiccated within hours. A rather detailed study of BP in relation to phenology in Lurreu divaricatu from southern California was carried out by Oechel et al. (1972a). During the study, the dependency of growth on BP varied phenologically. Reproductive growth only occurred with high BP. Stems were formed throughout the year but were initiated at the highest rate in winter when BP was high. Elongation rate of stems for

THE PRESSURE CHAMBER I N ECOLOGICAL RESEARCH

229

each sample period was correlated with BP at the beginning of the period (r = 0.71). The rate of node production was higher during periods of high BP, and nodes formed under low B P conditione often did not expand until BP increased. I n addition, leaves were formed at a higher rate and persisted longer during periods of higher BP. Studies such as these coupled with data on physiological processes will aid in developing models of strategies of species-environmental interactions. By expanding studies to include a number of species which occupy either the same sere or successive seres, modeling may be taken to the succession level.

I V . OTHERAPPLICATIONS OF

THE

PRESSURE CHAMBER

The pressure chamber has been employed in many studies having only tangential bearing on the field of ecology or “applied ecology”. For completeness we will attempt briefly to review this work with an emphasis on application rather than results.

A.

PATHOLOGY, ENTOMOLOGY, POLLUTION EFFECTS

I n a sense, P in plants is analogous to body temperature in homeothermic animals. Under certain conditions, a knowledge of “nbrmal” P values may lead to recognition of abnormal values, thus indicating physiological problems. The activities of fungi, bacteria, insects etc., which interfere with roots or vascular systems so as to disrupt the absorption and translocation of water within the plant, may be reflected in abnormal water stress conditions (DeRoo, 1969b; Turner, 1972; Dimond, 1972). I n addition, Mason (1969), Hodgea and Lori0 (1971) and Helms et al. (1971) have pointed out that success of insect attacks on trees may be influenced by plant water status. I n either case, the pressure chamber would seem to have numerous applications in the fields of plant pathology and entomology. DeRoo (1969~)employed the pressure chamber to diagnose pathological wilts in Rhododendron catawbiense by comparing transpiration and P in one-year-old container-grown plants. He found that the disease could be detected even in the early stages when pressure chamber comparisons with healthy plants were made. Duniway (1971b) employed the pressure chamber to ascertain the causes of wilting in Fusarium-infected tomato plants. He was able to show that wilt was caused by high water stress induced by increased stem resistance, and not by an alteration in transpirational behaviour. Helms et al. (1971)

230

GARY A. RITCHIE

and

THOMAS M. HINCKLEY

found that stomata1 closure, net photosynthesis and transpiration decreases were associated with increases in foliar water stress in Verticicladiella-infected ponderosa pine seedlings growing under field conditions. The pressure chamber has also shown that snag-tops in giant sequoias, which have been fire damaged, are a response to high water stresses in the uppermost tree crowns (Rundel, 1973). It is apparently possible to detect membrane damage with the pressure chamber (Turner, 1972; Dimond, 1972; Turner and Dimond, 1972). By plotting pressure-volume curves (as described in Section I1 G ) for diseased and healthy plants, it was shown that the P-V curves were linear and parallel, with greater pressures required to produce given volumes of xylem sap in diseased plants. Non-parallel curves would have suggested membrane damage in the leaf parenchyma. With oat leaves, Turner and Dimond (1972) found that membrane damage was detectable only when most of the cells in a diseased leaf had been damaged. The dwarf mistletoe (Arceuthobium americanum) is an extremely destructive parasite of lodgepole pine in western North America. Mark and Reid (1971) observed that P values of dwarf mistletoe stems affixed to lodgepole pine branches were always more negative than branch values, thus favoring movement of water from host to parasite even when host plants are under considerable water stress. The effects of air pollution on foliar injury may be suitable for investigation with the pressure chamber. Halbwachs (1970, 1971) measured diurnal patterns of P in the crowns of fluorine-damaged spruce (Picea dies), pine (Pinus sylvestris), willow (Salix caprea), birch (Betula pendula) and alder (Alnus glutinosa) trees, operating under the hypothesis that a water deficiency may be responsible for top die-back in affected trees. He found that when the water balance of the injured trees was “even” (reflective of low A E D), there was no significant difference between P in healthy and diseased trees. But with “uneven” conditions (high A E D) P was abnormally high in the tops of the diseased trees. This surprising increase of P with height paralleled the degree of needle or leaf injury. B.

LEAF FOLDING I N LEGUMES

The pinnate-compound leaves of numerous Leguminosae are known to fold or “sleep” at night or during and after rain storms. Diurnal changes in water status would seem to be implicated in this phenomenon. Scholander and Perez (1968) observed diurnal patterns of P in the legumes Parkia and Pithecolobium. They found that as evening approached, leaf folding in Parkia occurred at a P of - 5 bars and in

THE PRESSURE CHAMBER IN ECOLOOICAL RESEARCH

231

Pithecolubium at - 8 bars. I n submerged Pithecolubium plants, the leaves remained folded continuously. The pressure chamber would appear to offer promise for studying not only the folding of leaves at night but also the dramatic movements of Mimosa leaves in response to physical stimuli.

c. W A T E R

RELATIONS O F ROOTS

The various components of water potential are not as well understood in roots as they are in shoots, and the contribution of the hydrostatic component to total water potential may not be the same in roots as it is in shoots (DeRoo, 1969a). Field studies on tobacco of Pshoot (Pd), showed that Ps behaved in a predictable diurnal Proot( P r ) and fashion as it responded to changes in AED (DeRoo, 1969a).P , showed little diurnal change and increased at night to within 2 bars of Ps. However, when #so,, was compared diurnally to P,, a striking inverse gradient was observed, with P , often being as much as 10 bars higher than #soil. Similar studies under conditions of low with potted tobacco plants (DeRoo, 1969a) and with bean (Phaseoh vulgaris) (Taerum, 1973) supported these initial findings. As an explanation, DeRoo (1969a) postulated that roots underwent elastic contraction, forming a vapor gap (Bernstein et al., 1959; Huck et al., 1970) between themselves and the soil. This provided an insulation, allowing the root system to retain water while evaporation was occurring from the exposed soil. I n contrast, Taerum (1973) concluded that ion uptake by the root was responsible for the inverse gradient, while root damage during sampling may be solely responsible (Gee et al., 1974). The work of Huck et al. (1970) on changing root diameters in cotton plants has indirectly suggested that Pr undergoes considerable diurnal change. Hellkvist et al. (1974) with Sitka spruce trees, however, found that fluctuations in P , were very small and were not clearly related to the time of day or weather (only a one bar difference was noted between a clear and an overcast day). P , increased with increasing soil temperature and increased from the trunk towards the root tip. Another interesting finding of DeRoo (1969a) was that when a potted plant was severed at the root collar and Ps and Pr measured, the root system was normally at a higher P value than the shoot system despite the fact that measurements were made at exactly the same location on the stem. The observed difference seemed to indicate a rapid equilibration of P throughout the wetter root system and the drier shoot system. This was tested by placing plants and their exposed root systems in a dark moist room overnight to equilibrate. The next day, the plants were

232

GARY A. RITCHIE

and

THOMAS M. HINCKLEY

severed and P measured. P8 and Pr were nearly equal on most plants. I n a few cases the shoots had higher P values than their root systems, possibly reflecting superior water-retention capability of the leaves than of the exposed soil-root systems. In summary, these studies demonstrate that (1) Pr values do not seem to display strong diurnal patterns (DeRoo, 1969a; Hellkvist et al., 1974) despite conflicting indirect evidence of Huck et al. (1970), (2) in transpiring plants P8 is always more negative than Pr (DeRoo, 1969a; Taerum, 1973), (3) in non-transpiring plants (equilibrium) the values tend to be equal (DeRoo, 1969a; Kaufmann and Eckard, 1971) and (4) as soil moisture stress increases, the difference between Pr and P8 becomes smaller (Sankary and Barbour, 1972a) and an inversion between and Pr may occur (DeRoo, 1969a; Taerum, 1973). These preliminary findings suggest that the pressure chamber generates valid estimates of root water potential, although more experimentation is needed. D.

FROST HARDINESS

Brown et al. (1972) employed the pressure chamber to ascertain degrees of damage in two-month-old black locust seedlings exposed to freezing. They reasoned that when cells are frozen intracellularly, membrane rupture is likely to occur followed by a rapid loss of cellular water. Hence an increase in P following freezing should indicate cell rupture. Testing this theory on cold-hardened and frozen vs. control seedlings, they found a correlation between P and later survival of treated seedlings, and suggested that the pressure chamber may find w e in the determination of seedling susceptibility to freezing stress and degree of hardiness. Implications of these h d i n g s in the study of alpine or high-latitude plant ecology are obvious (Larcher, 1972).

E.

CULTURAL APPLICATIONS

An obvious cultural application of the pressure chamber technique is the development and implementation of effective irrigation programs for agricultural plants and forest trees. Two areas seem directly susoeptible to study: determination of the timing of irrigation schedules (Klepper and Ceccato, 1968; Cleary, 1968; Goode, 1968; Kaufmann, 1969; Hanan, 1972; Halevy, 1972; Hiler et al., 1972) and the effects of irrigation on plant processes (Woodman, 1971; Hiler et al., 1972; Hinckley and Ritchie, 1972; Bordovsky et al., 1974). Traditional irrigation practices rely upon estimates of soil moisture to determine when to apply water and how much to apply (Hailey

THE PRESSURE CHAMBER IN ECOLOQICAL RESEARCH

233

et al., 1973). Replacement of this scheme with a “physiological indicator” (Kriedemann and Smart, 1971; Elfving et al., 1972; Halevy, 1972; Clark and Hiler, 1973)is desirable, and an estimate of plant water stress or water potential would seem a most appropriate index. It would also be useful to delineate experimentally which stages of plant growth are most limited by water stress (Hiler et al., 1972), and to what degree and by what mechanisms irrigation tends to relieve these effects (Hinckley and Ritchie, 1972). Studies with cotton have demonstrated that when irrigation schedules were based upon an estimate of plant water stress coupled with growth stage-drought sensitivity measurements, savings in water supplies were realized (Bordovsky et al., 1974). The problem of when and where to fertilize forest trees to produce economic returns is one which may be subject to study with the pressure chamber (Brix, 1972; Waring and Youngberg, 1972). Fertilization, like irrigation, must be timed properly so that its availability to the plant occurs when nutrients are limiting. Other operations in agriculture, forestry and horticulture could be improved with knowledge of their effects on plant water status. Measurements of P made before and after transplrtnting citrus trees (Davenport et al., 1972b), for example, showed that anti-transpirants are valuable in reducing water stress during this operation. Radiata pine seedlings must be “hardened off” by using restricted watering schedules to minimize transplanting shock (Rook, 1973). Havranek and Tranquillini (1972)and Tranquillini (1973) have used measurements of P to analyze the cause of transplanting shock in larch (Laris decidua) seedlings and, based on their findings, handling procedures have been standardized. Similar investigations have been made with Picea abies (Lupke, 1973a, b). Measurements of P have also been used in evaluating the use of anti-transpirants on plantation red pine (Turner and Waggoner, 1968) and on orchard-grown peach (Prunus persicu) trees (Davenport et al., 1972a). Cleary (1968) indicates that, in forestry, the most important field application of the pressure chamber may be to evaluate differences in planting sites and methods on seedling survival in reforestation programs.

F.

OTHER APPLICATIONS

There are still other applications to which the pressure chamber is well suited. One of these is the extraction of xylem sap from woody stems to derive pressure-volume curves (Hammel, 1967) and for various types of sap analysis (e.g. Fenn et al., 1970). Havis (1971) placed cut stems of holly ( I l e z opaca) into a flask of dye and placed the flask into a pressure chamber with the stem protruding through the chamber neck.

234

GARY A. RITCHIE

and

THOMAS M. HINCKLEY

By applying a pressure of 2 bars, he was able to observe the movement of dye through the stems during freezing. Graham and Ulrich (1972) employed the pressure chamber to estimate changes in xylem resistance (by forcing a solution through the stem) in normal and potassiumdeficient sugar beet plants. To ascertain the differential effects on growth of stress induced by pressure vs. root desiccation in soybean, Meyer and Boyer (1972) placed cotyledons and hypocotyls in a pressure chamber. Other imaginative applications of the technique are likely to be forthcoming.

V. S O M EUNRESOLVED QUESTIONS A.

WHY DOES

P

F A I L TO M E E T T H E G R A V I T A T I O N A L

POTENTIAL GRADIENT?

The total water potential in the xylem sap is the algebraic sum of the solute, matric, frictional and gravitational potentials (Eqn (2)). The pressure chamber is considered to measure the sum of the gravitational potential (the hydrostatic head or weight of the water column, which has a value of about -0.1 bar m-l) and the frictional potential. I n plants of limited height, the gravitational potential is negligible and the pressure chamber measures only the friction potential. The gradient due to the weight of the water column ( A P ) may be calculated knowing the density of the xylem sap ( p , g/cm3),the acceleration due to gravity (9, cm/sec2), the height in centimeters above the water table (h),and a constant ( a ) to convert from dynes/cm2 to bars, using the following equation:

AP(notlux, = aPgAh (20) This relationship would apply only under static conditions (no transpirational flux). Under flux conditions, however, steeper gradients would be generated due to the frictional shear forces between the xylem sap and flow conduits and between the torus and xylem sap in conifer pits. Changing the character of the conduits or the rate of flux would change the shear force and hence would change the pressure gradient necessary to overcome the increased resistance. Applying an analogy of Darcy’s Law, Ptl, =

rate of flux -viscosity .distance xylem permeability

+ apgAh

(21)

Dixon (1914) estimated that under flux conditions, P would equal 2 (apghh) or about - 0 . 2 bars m-l. Zimmerman and Brown (1971, p. 206) estimated that under conditions of high flux, P would be

-

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

235

- 0-15 bars m-l. Most observations indicate that actual pressure gradients are equal to or more negative than the theoretical gradients (Hellkvist et al., 1974). It is therefore of some interest that Scholander et al. (1965a) observed P values in tall Douglas-fir and redwood trees at midday to closely approximate to the pure gravitational potential gradient ( - 0.1 bar m-1). I n a later study, Tobiessen et al. (1971) confirmed these findings in a 90 m-tall giant sequoia tree, observing that under conditions of low AED, P was about -0.08 bar m-1. Zimmerman and Brown (1971, p. 204) have discussed this problem, suggesting that anatomical differences between upper and lower crown twigs may be implicated in the P discrepancies. They reasoned that, if this were true, the apparent P in a felled (horizontal) tree should be slightly negative under non-flux conditions. To test this theory they measured P on twigs which had been bagged the day before to retard transpiration at heights between 0.8 and 15.5 m on a white spruce (Picea glauca) tree. On the standing tree, P values were steeper than the gravitational gradient as expected, but after the tree was felled P turned slightly negative. Tobiessen (pers. comm.) argues against this evidence because the tree was transpiring at the time of measurement. If the lower portion of the crown had greater transpirational area, hence greater transpiration rates, P could be more negative in the lower branches. This argument would be valid only if water loss from the unbagged branches caused a reduction of P in the bagged branches. There is limited evidence that this phenomenon may occur in some conifers (Huber and Schmidt, 1936; Hinckley and Ritchie, 1970) as well as angiosperms (Fraxinus) (Daum, 1967). Plumb and Bridgman (1972) postulated that a concentration gradient of filamentary monomolecular chains (or gel), one end of each chain affixed to the xylem vessel wall, occurs in trees. The chemical activity (matric potential) of the xylem sap thereby supports the water column and the hydrostatic pressure is constant throughout the stem. Some indirect evidence for the existence of such a gel is cited. This model proposes that the pressure chamber measures not the pressure potential but the matric forces associated with the hypothesized xylem gel. This theory would account for the findings of Scholander et al. (1965a) and Tobiessen et al. (1971) relative to the failure of P to meet the hydrostatic gradient in tall trees, but it has come under some rather sobering criticism from a number of authorities (Levitt and Storvick, 1973; Richter, 1973; Scholander, 1973; Hammel, 1973). A more defensible explanation could be developed around Richter’s (1972) argument that “a true gradient of activity or pressure is established only along a coherent conduit in the xylem”. Frequently,

236

GARY

A. RITCHIE

and

THOMAS M. HINCKLEY

measurements on water potential gradients in tall trees fail to insure that pressure chamber readings are taken from a coherent conduit. In these studies, readings are taken on branches growing from the main bole. It is in these partially independent conduits that frictional potentials may be of considerable magnitude (1.0 bars/m) and, therefore, cause error in the estimation of the gravitational or hydrostatic gradient in the main bole (Richter, pers. comm.). For example, Hellkvist et al. (1974) observed in Sitka spruce that secondary branches had resistances of two and four times greater than those found in primary branches and the main stem respectively. I n addition, resistances may increase dramatically under low flow conditions (Elfving et al., 1972). Such an explanation seems to accommodate the observations of Scholander et at?. (1965a) and Tobiessen et al. (1971). B.

W H Y IS

0 BARS NEVER

ACHIEVED?

Consider it plant rooted in a saturated soil (y5soil N 0 bars). If this plant is kept in a dark, humid environment and no transpiration occurs, then an equilibrium water potential should be established throughout the soil-plant system. Under these conditions, P values measured near the soil surface should equal 0. However, many workers have reported that under such conditions P is consistently more negative than y5so,l (e.g. Sucoff, 1972). Elfving et al. (1972) suggest that a - 4 to -6 bar difference might be expected between P and in woody plants under equilibrium conditions. Sucoff (1972) found differences of - 2.6 to - 3.7 bars in red pine, and we found from - 1.5 to - 3 bars difference in noble fir, Douglas-fir, Pacific silver fir and eastern red cedar (Juniperus virginiana), and differences as low as -0.8 bars in dogwood (Cornus JEorida) and white oak, and -0.3 bars in soybean (unpublished). Several explanations have been offered for this phenomenon. Elfving et al. (1972) implicate the development of very high flow resistances from soil to leaf a t low fluxes. J. S. Boyer (pers. comm.) suggests that the demand for water through growth process may be sufficient so that P never reaches zero, while R. B. Walker (pers. comm.) notes the possibility of significant xylem osmotic and/or matric potentials. To our knowledge, this question has not been fully resolved.

c. I S T H E R E S U B S T A N T I A L R E S I S T A N C E

TO

FLOW B E T W E E N LEAF A N D STEM?

Neglecting gravitational and temperature effects, if there were no resistance to the flow of water through the soil-plant system, water potential gradients would approach zero in plants under equilibrium or

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

237

non-equilibrium conditions. Clearly, however, there are resistances to flow and there are gradients in water potential. Whether or not the resistance existing between leaf and stem is appreciable, however, has not been clearly demonstrated, and has substantial bearing upon interpretation of pressure chamber data. Begg and Turner (1970)applied the pressure chamber t o an examination of this problem in tobacco plants. They enclosed leaves from different insertion levels in plastic and aluminium foil bags and allowed them to equilibrate with the water potential in the adjacent stem tissue. They then measured P of enclosed leaves (assumed proportional to P8 at that insertion level) and unenclosed leaves (Pi) diurnally, comparing the results to those obtained with control (unbagged)plants. Differences in P occurred along the stem, suggesting that resistances to flow existed within the stem; differences between Pl and Pa a t the same insertion points on the plants were interpreted as a result of substantial resistances to flow in the petiole. Hellkvist et al. (1974) applied Begg and Turner’s (1970) approach to the examination of potential gradients along the main bole and along primary branches of Sitka spruce trees. They estimated that the water potential gradient in secondary branches was four and two times that of the main bole and primary branches respectively. However, the relative contribution of primary and secondary branches to the overall resistance to water transport through the plant as compared to the main bole was almost negligible because of their relative lengths. Boyer (1967a) suggested that differences in water potential exist in the xylem and leaves of plants due to the resistances to water flow between these points. He found, however, that equilibrium between water potential in stem and leaf occurred very quickly in the pressure chamber, and suggested that flow resistances must have been relatively small to allow this to happen. I n studies with hemlock (Tsuga canadensis), however, several hours were required for equilibrium to occur between leaf and stem water potential (Tyree et al., 1973a, b; Tyree and Dainty, 1973). M. T. Tyree (pers. comm.) suggests that the rapid establishment of equilibrium found by Boyer and others may reflect the fact that apoplastic water potential changes rapidly with volume changes. These small volume changes barely affect the symplast water potential and therefore go unnoticed. In a more recent study, Tyree, Caldwell and Dainty (unpublished) found that the xylem network of hemlock, up to but not including the leaves, contributes about half of the resistance to water flow in whole shoots of 20 to 30 g fresh weight. Leaves near the cut end of the shoots seemed to experience smaller xylem resistances than leaves near the apex. It is quite possible that such internal resistances vary considerably

238

GARY A. RITCHIE

and

THOMAS M. HINCKLEY

with environmental and plant conditions and that different species display different characteristics. We anticipate that imaginative application of the pressure chamber will in the future shed considerable light on these and related questions.

D.

D O PLANTS AT NIGHT ACT AS TENSIOMETERS?

A number of methods exist for measuring bulk soil moisture content expressed as water potential, matric potential, percent moisture, etc. However, rooted plants do not respond directly to soil moisture content or bulk water potential because of differences in root density and root structure and gradients of soil moisture content occurring throughout the rooting profile. It is these rather than point conditions which the plant actually “feels” (Gardner, 1960). The integrated soil water potential at the total absorbing surface of the root system has more physiological or ecological meaning than a bulk moisture content or potential value from somewhere in the soil matrix (Newman, 1969). Veihmeyer and Hendrickson (1950) expressed the situation thus:

. . . neither soil sampling nor measurements of soil properties which are related to soil moisture contents, made with physical instruments inserted into the soil, will give reliable records of the actual moisture content of the soil in contact with the absorbing portion of the roots. value for the Slatyer (1967, p. 122) comments: “. . . the effective root system as a whole is probably impossible to obtain by direct methods”. I n addition, Holmes and Robertson (1959) observed: It is clear that any accurate description of the moisture status of a soil will not be simple. The plant is the only true indicator of this factor and at the present time it is not possible to measure plant moisture stress, per se. The implication is clear that a measure of “plant moisture stress” is in effect a measure of “soil moisture stress”. I n concert with this thinking, Tinklin and Weatherley (1968) proposed the following relationship: $ L = 1Clg+.fr

where +L and ah,s are total water potential of the leaf and soil respectively, f represents a transpirational flux and r a resistance to transpiration. Thus they reasoned that $ L of a transpiring plant is the sum of two terms, the !JI,~ and a term related t o transpiration rate and resistance. When transpiration is zero, then: *L = *a

(23)

THE PRESSURE CHAMBER I N ECOLOGICAL RESEARCH

239

and a pressure chamber estimate of +h is essentially an estimate of the “effective soil moisture” or the soil moisture potential which the plant is actually experiencing regardless of the bulk soil water content. A number of authors (Klepper, 1968; Waring, 1969a; Jordan and Ritchie, 1971 ; Sucoff, 1972) have recommended such a bioassay-type approach for assessing soil moisture rather than more traditional gravimetric, psychometric or radiometric methods. Before a BP measure can be valid, an equilibrium condition must have been attained during the night, so that at dawn P is neither increasing nor decreasing. I n other words, total recharge of the plant at night is essential. This has been tested by Elfving and Kaufmann (1972) who found, with citrus trees, that total recharge occurred over a rather wide range of soil moisture and temperature conditions. Klepper (1968), Hiler et al. (1972) and Elfving et al. (1972) report similar results. Ritchie (1971) and Hinckley and Ritchie (1973) found that Pacific silver fir seedlings underwent sufficient night transpiration in both field and laboratory to prevent an equilibrium condition from occurring. However, when the seedlings were covered with plastic bags to retard transpiration, an equilibrium condition was reached within 5 h after dark when BP was greater than - 5 bars. Conversely, Sucoff (1972) found that a steady-state did not always occur at night with red pine needles. Lassoie and Scott (1972) observed an inverse relationship between seasonal trends in B P and calculated soil moisture in vine maple from 6 July to 24 September 1971. Kappen et al. (1972) noted that Artemisia herba-alba plants were unable to resaturate themselves at night, and Haas and Dodd (1972) found no statistical correlation between B P and percentage soil moisture in honey mesquite. Although the existence of steady-state BP values would seem to indicate that an equilibrium between plant and soil had been established, this may not always be the case due to two mechanisms. The findings of DeRoo (1969a) with tobacco roots suggest that at - 12 bars soil moisture potential a vapor gap formed around the roots, due to root shrinkage, and effectively insulated the roots from the soil. Thus an equilibrium would only be expected to occur over a very long period of time. Recent data of Taerum (1973) with bean tend to confirm this. Also, it is well known that hydraulic conductivity of soils decreases as z,hsoi, decreases (Slatyer, 1967, p. 106). Such a mechanism could greatly increase the time required for equilibrium or even prevent an equilibrium from occurring. It seems reasonable to conclude tentatively that at high soil water potentials equilibrium tends to be established at night if transpiration is sufficiently retarded and thus BP is closely related to the effective

240

G

~ A.Y RITCHIE

and

THOMAS M. HINCKLEY

soil moisture. As soils dry out, however, the equilibration time increases until BP is no longer a steady-state phenomenon and becomes less indicative of soil moisture. However, it is indicative of the level of water stress at which the plant begins the day and, therefore, remains a useful estimate of effective soil moisture. Such an approach has promise in ecological studies due to its simplicity and physiological relevance. The BP-soil moisture relationship undoubtedly varies with species and soil type and perhaps other factors. This is clearly a fruitful area for further research. VI. CONCLUDINGSTATEMENT The pressure chamber has become the standard technique for assessing plant water status in the field. Its reliability, repeatability and rapidity of memurement may also recommend it as a substitute for the traditional but more laborious methods used in the laboratory. Nevertheless, it is incumbent upon the investigator to educate himself not only to the advantages of the technique, but also to its procedural and interpretational pitfalls. It has been our objective in this review to identify, describe and interpret many aspects of the pressure chamber method. I n this brief concluding statement we will summarize some of the more salient concepts. a. It has been established that the pressure chamber measures only the gravitational and the frictional components of the total water potential in the xylem or transpiration stream and does not measure the osmotic or matric components. The combined gravitational and frictional potentials are referred to as P or the “xylem pressure potential”.

b. A number of workers have compared the pressure chamber to other techniques of assessing plant water status. Most calibration curves generated in these studies have taken one of two basic forms. I n the most common form, the pressure chamber yields lower (more negative) values at high water potentials than the other techniques, but as water potential decreases this tendency is reversed. I n the second form, the calibration is nearly linear, tending towards a 1 : 1 relationship throughout the measurement range. When the pressure chamber is used to estimate the total water potential, calibrations with the thermocouple psychrometer are desirable. However, when P values are used only as relative indicators of water status, calibration may not be necessary, as P itself is a meaningful index of plant water status.

A number of procedural errors are associated with the pressure chamber method. Recutting the stem after the initial excision has been

c.

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

241

made or using excision to induce water stress in foliage are not recommended. The amount of tissue inside and outside the chamber should be standardized for all measurements. Determinations should be made as rapidly as possible following excision. The system must be permitted to come to equilibrium during the course of a measurement; therefore, low rates of pressure application (from 0.1 to 0.3 bars sec-l) are desirable. I n some species the recognition of the endpoint has been difficult due to bubbling of water or resin onto the cut xylem surface during pressure application. Remedies may include the use of individual leaves, standardization of procedure and observational methods, or instrumentation of the system to allow for electronic endpoint determination. Appreciable solute potentials have occasionally been recorded, further arguing for calibration when total water potential values are sought. d. With suitable apparatus, it is apparently possible, perhaps even desirable, to make P determinations directly on individual leaf petioles, conifer needles or grass blades. e. Use of the pressure-volume curve permits estimates of solute potentials, mean water potential a t incipient plasmolysis, bound water content and relative osmotic adjustments as well as the xylem pressure potential.

f. P typically varies spatially in plants growing under natural conditions. This variability can be accounted for by variations with height above the ground through friction-resistance or gravitational effects, differing tissue age or developmental stage and differences in the absorption-transpiration balance, which are strongly associated with crown orientation. It is imperative that these sources of variability be recognized and accounted for when selecting sampling locations, especially when P values are used to compare the behaviour of one individual or species to another. g. When soil moisture is adequate, P measured diurnally in the field generally reflects the atmospheric evaporative demand curve. As soil moisture becomes limiting, however, and resistances to water flux increase, the diurnal curve tends to plateau near midday and even to display a brief recovery. Diurnal curves from arid zone species tend to begin a t low P values and gradually decrease throughout the day, only recovering slightly at night. h. Four strategies for expressing and interpreting P data were explored. The first considers any value of P to be a predawn or base level P (BP) and a depression below that base level ( D P )to be brought about by the daily transpiration-absorption imbalance. The lowest P recorded

242

GARY A. RITCHIE

and

THOMAS M. HINCKLEY

during the course of a day is called Pmin. By comparing these expressions of P for different species over a range of water potentials, it is possible to infer species’ ability or inability to regulate water absorption and loss. A second approach is the concept of limiting conditions. Laboratory or field studies are used to delineate the relationship between P and transpirational flux under “non-limiting” soil moisture and temperature conditions for each species. Examination of the P vs. flux relationship in the field then permits recognition of limiting conditions for various species. A given site or environment can be described by noting the manner in which one selected reference-speciesreacts to it. The selected species is then monitored over a range of environments and its responses used as an indicator of site conditions. The fourth concept embodies a stress-day index, which is defined as the plant water stress level responsible for a reduction of growth or yield below the potential for that species. This approach provides a basis for quantifying relations between water stress and physiological processes or phenological events.

i. Pressure chamber values per se are not always consistent enough to distinguish between species’ behaviour patterns in different habitats, because wide interspecific ranges of tolerance to water stress exist.

j. The relationship between P and various physiological processes or plant factors has been reviewed for selected species. The relationship between P and stomatal resistance can be linear, curvilinear or a threshold function depending on species, environmental conditions and plant materials. Transpiration rates, being strongly influenced by stomatal resistance, are closely related to P. The velocity of sap in the xylem tends to be correlated with transpiration and P. I n order to correlate P with stem diameters, it has often been necessary to plot a lag-time due to an hysteresis effect. Net assimilation rate generally declines as P decreases; however, this relationship varies widely with species, habitat and plant conditions.

k. The pressure chamber may also be used effectively in studies of plant pathology, entomology and pollution effects. Activities of pathogenic agents which disrupt vascular function or cause membrane damage can often be detected. Productive use of the pressure chamber technique in forestry, agriculture and horticulture in such operations as irrigation, transplanting and fertilization has proved successful. 1. Some perplexing questions pertinent to the pressure chamber technique are as yet unanswered: (1) why do P values in some tall trees fail

THE PRESSURE CHAMBER I N ECOLOQICAL RESEARCH

243

to reflect the gravitational potential gradient predicted by theory, (2) why are P values of zero bars never measured, (3) is there a substantial resistance to water flow between leaf and stem xylem and (4) do plants equilibrate with soil water potential at night so that a measure of P is an estimate of soil water potential!

ACKNOWLEDGEMENTS Through the preparation of this manuscript, the authors have relied heavily upon the suggestions and contributions of colleagues throughout the world. These individuals include Drs D. Y. Alexandre, Fakhri A. Bazzaz, Udo Benecke, P. V. Biscoe, John S. Boyer, John W. Cary, David Doley, John M. Duniway, James Ehleringer, Neil Fettell, Robert H. Haas, W. M. Havranek, H. Hilscher, L. D. Incoll, H. A. P. Ingram, Paul G. Jarvis, Merrill R. Kaufmann, Betty Klepper, Lionel G. Klikoff, Werner Koch, V. Kozinka, P. E. Kriedemann, C. H. Anthony Little, M. M. Ludlow, I. J. McCracken, Harold A. Mooney, Russell T. Moore, Walter C. Oechel, George S. Puritch, Hanno Richter, Joe T. Ritchie, David A. Rook, E.-D. Schulze, Edward Sucoff, Peter Tobiessen and Neil C. Turner. We are especially indebted to Drs Brian D. Cleary, H. T. Hammel, James P. Lassoie, Wayne R. Jordan, Stephen D. Ross, P. F. Scholander, Melvin T. Tyree, Richard B. Walker, Richard H. Waring and James N. Woodman for cogent and thoughtful reviews of the manuscript. Funds were provided by McIntyre-Stennis Grant 7008-1690. Contribution from the Missouri Agricultural Experiment Station Journal Series Number 6836.

REFERENCES Barker, J. E. (1973). Diurnal patterns of water potential in Abiea concolor and Pinua ponderoea. Can. J . For. Res. 3, 556-564. Barrs, H. D. (1968). I n “Water Deficits and Plant Growth” (Ed. T. T. Kozlowski), Vol. I, pp. 236-368. Academic Press, London and New York. Barrs, H. D., Freeman, B., Blackwell, J. and Ceccato, R. D. (1970). Comparison of leaf water potential and xylem water potential in tomato plants. Auat. J . biol. Sci. 23, 485-487. Bartholomew, E. T. (1928). Internal decline of lemons. 111. Water deficit in lemon fruit caused by excessive leaf evaporation. Am. J . Bot. 13, 102-117. Begg, J. E. and Turner, N. C. (1970). Water potential gradients in field tobacco. Pl. Phyaiol. 46, 343-346. Bernstein, L., Gardner, W. R. and Richards, L. A. (1959). Is there a vapor gap around roots? Science, N . Y . 129, 1750-1753.

244

GARY A. RITCHIE

and

THOMAS M. HINCKLEY

Biscoe, P. V. (1972).The diffusion resistance and water status of leaves of Beta vulqark J . exp. Bot. 13, 930-940. Blum, A., Sullivan, C. Y. and Eastin, J. D. (1973).On the pressure chamber technique for estimating leaf wat,er potential. in sorghum. Agron. J . 65, 337-338. Bordovsky, D. G., Jordan, W. R., Hiler, E. A. and Howell, T. A. (1974).Choice of irrigation timing indicator for narrow row cotton. Agron. J . 66, 88-91. Boyer, J. S. (1966).Isopiestic technique: Measurement of accurate leaf water potentials. Science, N . Y . 154, 1459-1460. Boyer, J. S. (1967a).Leaf water potentials measured with a pressure chamber. Pl. Physiol. 42, 133-137. Boyer, J. S. (1967b).Matric potentials of leaves. PI. Physiol. 42, 21%217. Boyer, J. S. (1968).Relationship of water potential to growth of leaves. PI. Physiol. 43, 1056-1062. Boyer, J. S . (1989).Measurement of the water status of plants. A . Rev. Pl. Physiol. 19, 351-364. Boyer, J. S . and Ghorashy, S. R. (1971).Rapid field measurement of leaf water potential in soybean. Agron. J . 63, 344-345. Boyer, J. S. and Knipling, E. B. (1965).Isopiestic technique for measuring leaf water potentials with a thermocouple psychrometer. Proc. natn. A d . Sci. U.S.A. 54, 1044-1051. Brandle, J. R.,Schnare, P. D., Hinckley, T. M. and Brown, G. N. (1973).Changes in polysomes of black locust seedlings during dehydration-rehydration cycles. Physiologia Pl. 29, 406-409. Brix, H. (1972).Nitrogen fertilization and water effects on photosynthesis and earlywood-latewood production in Douglas-fir. Can. J . For. Rw. 2, 467-478. Brown, C . N.,Hinckley, T. M. and Bixby, J. A. (1972).Xylem sap pressure relationship to seedling survival after freezing. Cryobiology 9,314. Campbell, C. J. (1973).Pressure bomb measurements indicate water availability in a southwestern riparian community. U.S.D.A. Forest Serv. Res. Note RM-246,1-4. Campbell, C. J. and Pase, C. P. (1972).Pressure bomb measures changes in moisture stress of birchleaf mountain-mahogany after partial crown removal. U.S.D.A. Forest Serv. Rw. Paper RM-221,1-4. Campbell, G. S. and Campbell, M. D. (1974). Evaluation of a thermocouple hygrometer for measuring leaf water potential in situ. Agron. J . 66, 24-27. Cam, M. K.V. (1971/1972).The internal water status of the tea plant (Camellia sinensis):Some results illustrating the use of the pressure chamber technique. Agric. Metewol. 9,447-460. Cary, J. W. and Fisher, H. D. (1969).Plant moisture stress: A portable freezingpoint meter compared with the psychrometer. Agron. J. 61, 302-305. Cary, J. W. and Fisher, H. D. (1971).Plant water potential gradients measured in the field by freezing point. Phyaiologia PI. 24, 397-402. Cary, J. W. and Wright, J. L. (1971).Response of plant water potential to the irrigated environment of southern Idaho. Agron. J . 63, 691-695. Catsky, J. (1962).Water saturation deficit in the wilting plant. The preference of young leaves and the translocation of water from old into young leaves. Biologia Pl. 4, 306-314. Clark, R. N. and Hiler, E. A. (1973).Plant measurements as indicators of crop water deficit. Crop Sci. 13, 466-469. Cleary, B. D. (1968).Water stress measurements and their application to forest regeneration. Wwtem Reforestation, 1968, 1-3.

THE PRESSURE CHAMBER IN ECOLOGICAL RESEARCH

245

Cleary, B. D. and Waring, R. H. (1969). Temperature: Collection of data and its analysis for the interpretation of plant growth and distribution. Can. J . Bot. 47, 167-173. Cowan, I. R. (1965). Transport of water in the soil-plant-atmosphere system. J . appl. Ecol. 2, 221-239. Cowan, I. R. and Milthorpe, F. L. (1968). I n “Water Deficits and Plant Growth” (Ed. T. T. Kozlowski), Vol. I, pp. 137-193. Academic Press, London and New York. Dainty, J. (1969). I n “The Physiology of Plant Growth and Development” (Ed. M. B. Wilkins), pp. 419-452. McGraw-Hill, London. Daum, C. R. (1967). A method for determining water transport in trees. Ecology 48, 425-431.

Davenport, D. C., Fisher, M. A. and Hagan, R. M. (1972a). Some counteractive effects of antitranspirants. PI. Physiol. 49, 722-724. Davenport, D. C., Martin, P. E. and Hagan, R. M. (1972b). Antitranspirants for conservation of leaf water potential of transplanted citrus trees. HortSci. 7, 511-512.

DePuit, E. J. and Caldwell, M. M. (1973). Seasonal pattern of net photosynthesis of Artemisa tridentata. Am. J . Bot. 60, 426-435. DeRoo, H. C. (1969a). Water stress gradients in plants and soil-root systems. Agron. J . 61, 511-515. DeRoo, H. C. (1969b). Leaf water potentials of sorghum and corn, estimated with the pressure bomb. Agron. J . 61, 969-970. DeRoo, H. C. (1969~).Sap stress and water uptake in detached shoots of wiltdiseased and normal rhododendrons. HortSci. 4, 51-62. DeRoo, H. C. (1970). Leaf water potentials of tobacco, estimated with tho pressure bomb. Tobacco Sci. 14, 105-106. Detling, J. K. (1969). Photosynthetic and respiratory response of several halophytes to moisture stress. Ph.D. Dissertation, Univ. of Utah. Detling, J. K. and Klikoff, L. G. (1971). Comparison of two field techniques for determination of water potential in two halophytes. Am. Midl. Nat. 85, 236238.

Detling, J. K. and Klikoff, L. G. (1973). Physiological response to moisture stress as a factor in halophyte distribution. Am. Mi&. Nat. 90, 307-318. Dickson, R. E. and Broyer, T. C. (1972). Effects of aeration, water supply, and nitrogen source on growth and development of tupelo gum and bald cypress. Ecology 53, 626-634. Dimond, A. E. (1966). Pressure and flow relations in vascular bundles of the tomato plant. PI. Physiol. 41, 119-131. Dimond, A. E. (1972). In “Phytotoxins in Plant Diseases” (Eds R. Wood, A. Baillio and A. Graniti), pp. 289-306. Academic Press, London and New York. Dina, S. J. and Klikoff, L. G. (1973). Carbon dioxide exchange by several streamside and scrub oak community species of Red Butte Canyon, Utah. Am. Midl. Nat. 89, 70-80. Dina, S. J.,Klikoff, L. G. and Keddington, M. B. (1973). Seasonal water potential patterns in the Mountain Brush Zone, Utah. Am. Midl. Nat. 89, 234239.

Dixon, H. H. (1914). “Transpiration and the Ascent of Sap in Plants.” MacMillan, New York. Dixon, H. H. and Joly, J. (1894). On the ascent of sap. Phil. Trane. R . SOC. (London) Series B 186, 663-676. I

246

GARY A. RITCHIE

and

THOMAS M. HINCKLEY

Dobbs, R. C. (1966). The intraseasonal development of longitudinal growth patterns in Douglas-fir (Pseudotsuga menziesii [Mirb.] Franco). Ph.D. Dissertation, Univ. of Washington, Seattle. Doley, D. (1970). Effects of simulated drought on shoot development in Liriodendron seedlings. New Phytol. 69, 655-673. Duniway, J. M. (1971a). Comparison of pressure chamber and thermocouple psychrometer determinations of leaf water status in tomato. PI. Physiol. 48, 106-107. Duniway, J. M. (1971b). Water relations of Fusarium wilt in tomato. Physiol. Plant Pathol. 1, 537-546. Elfving, D. C. and Kaufmann, M. R. (1972). Diurnal and seasonal effects of environment on plant water relations and fruit diameters of citrus. J . Am. SOC.Hort. Sci. 97, 55&570. Elfving, D. C., Kaufmann, M. R. and Hall, A. E. (1972). Interpreting leaf water potential measurements with a model of the soil-plant-atmosphere continuum. Physiologia Plant. 27, 161-168. Fenn, L. B., Oertli, F. T. and Bingham, F. T. (1970). Specific chloride injury in Persea americana. Soil Sci. SOC.Am. Proc. 34, 617-620. Frank, A. B. and Harris, D. G. (1973). Measurement of leaf water potential in wheat with a pressure chamber. Agron. J . 65, 334-335. Gardner, W. R. (1960). Dynamic aspects of water availability to plants. Soil Sci. 89, 63-73. Gardner, W. R. and Nieman, R. H. (1964). Lower limit of water availability to plants. Science, N . Y . 143, 1460-1462. Gates, D. M. (1968). Transpiration and leaf temperature. A . Rev. PI. Physiol. 19, 211-238. Gee, G. W., Liu, W., OlvBng, H. and Janes, B. E. (1973).Measurement and control of water potentials in a soil-plant system. Soil Sci. 115, 336-342. Gee, G. W., Liu, W., OlvBng, H. and Janes, B. E. (1974). Use of pressure bomb measurements to estimate root water potentials. Agron. J . 66, 7578. Gifford, H. A. (1972). Simplified apparatus for determining leaf water potentials in pine needles. N . 2.Jl Forest Sci. 2 , 284-286. Goode, J. E. (1968). The measurement of sap tension in the petioles of apple, raspberry and blackcurrant leaves. J . hort. Sci. 43, 231-233. Goode, J. E. and Higgs, K. H. (1973). Water, osmotic and pressure potential relationships in apple leaves. J . hort. Sci. 48, 203-215. Graham, R. D. and Ulrich, A. (1972). Potassium deficiency-induced changes in stomata1 behavior, leaf water potentials, and root system permeability in Beta vulgaris L. PI. Physiol. 49, 105-109. Griffin, J. R. (1973). Xylem sap tension in three woodland oaks of central California. Ecology 54, 152-159. Haas, R. H. and Dodd, J. D. (1972). Water stress patterns in honey mesquite. Ecology 53, 674-680. Hailey, J. L., Hiler, E. A., Jordan, W. R. and van Bavel, C. H. M. (1973). Resistance to water flow in Vigna sinensis L. (Endl.) a t high rates of transpiration. Crop Sci. 13, 264-267. Halbwachs, G. (1970). Vergleichende Untersuchungen iiber die Wasserbewegung in gesunden und fluorgeschadigten Holzgewachsen. Cbl. ges. Forstwesen 87, 1-22. Halbwachs, G. (197 1). Physiologische Probleme der Vegetationschadigung durch gasformige Immissionen. Ber. Dtsch. Bot. Ctes. Bd. 84, 507-514.

THE PRESSURE CHAMBER IN ECOLOOICAL RESEARCH

247

Halevy, A. H. (1972). Water stress and the timing of irrigation. Hort. Sci. 7, 113-114.

Halvorson, W. L. and Patten, D. T. (1974). Seasonal water potential changes in Sonoran desert shrubs in relation to topography. Ecology 55, 173-177. Hammel, H. T. (1967). Freezing of xylem sap without cavitation. PI. Physiol. 42, 55-66.

Hammel, H. T. (1968). Measurement of turgor pressure and its gradient in the phloem of oak. PI. Physiol. 43, 1042-1048. Hammel, H. T. (1973). On the ascent of sap. Science, N . Y . 179, 1248-1249. Hanan, J. J. (1972). Repercussions from water stress. Hort Sci. 7, 108-112. Havis, J. R. (1971). Water movement in woody stems during freezing. Cryobiology 8, 581-585.

Havranek, W. and Tranquillini, W. (1972). Untersuchungen uber den Versetzschock bei der Llirche Wachatum und Wasserhaushalt nach dem Versetzen. Forstl. Bundem. in Wein 96, 111-135. Heath, 0. V. S. (1959a). I n “Plant Physiology-A Treatise” (Ed. F. C. Steward), pp. 193-250. Academic Press, London and New York. Heath, 0. V. S. (1959b). I n “Handbook of Plant Physiology” (Ed. W. Ruhland), 17, 416-464. Springer-Verlag, Berlin. Hellkvist, J., Richards, G. P. and Jarvis, P. G. (1974). Vertical gradients of water potential and tissue water relations in Sitka spruce trees measured with the pressure chamber. J . appl. Ecol. (in press). Helms, J. A. (1972). Environmental control of net photosynthesis in naturally growing Pinus ponderosa Laws. Ecology 53, 92-101. Helms, J. A., Cobb, F. W. and Whitney, H. S. (1971). Effect of infection by Verticicladiella wagnerii on the physiology of Pinus ponderosa. Phytopath. 61, 920-925.

Hickman, J. C. (1970). Seasonal course of xylem sap tension. Ecology 51, 10521056.

Hiler, E. A. (1969). Quantitative evaluation of crop drainage requirements. Transactions of the ASAE 12, 499-505. Hiler, E. A., van Bavel, C. H. M., Hossian, M. M. and Jordan, W. R. (1972). Sensitivity of southern peas to plant water doficit a t three growth stages. Agron. J . 64, 60-64. Hinckley, T. M. (1971). Estimate of water flow in Douglas-& seedlings. Ecology 52, 525-528.

Hinckley, T. M. (1973). Responses of black locust and tomato plants after water stress. Hort Sci. 8, 405-407. Hinckley, T. M. and Ritchie, G. A. (1970). Within-crown patterns of transpiration, water stress and stomata1 activity in Abies a m b i l k Forest Sci. 16, 490492.

Hinckley, T. M. and Ritchie, G. A. (1972). Reaction of mature Abies seedlings to environmental stresses. Trans. Missouri A d . Sci. 6, 24-37. Hinckley, T. M. and Ritchie, G. A. (1973). A theoretical model for calculation of xylem sap pressure from climatological data. Am. Midl. Nat. 90, 56-69. Hinckley, T. M. and Scott, D. R. M. (1971). Estimates of water loss and its relation to environmental parameters in Douglas-& saplings. Ecology 52, 520-524.

Hinckley, T. M., Chambers, J. L., Bruckerhoff, D. N., Roberts, J. E. and Turner, J. (1974). The effect of mid-day shading on stem diameter, xylem pressure potential, net assimilation rate and leaf surface resistance in a white oak sapling. Can. J . For. Res. 4 , 296-300.

248

GARY A. RITCHIE

and

THOMAS M. HINCKLEY

Hodges, J. D. and Lorio, P. L. (1968). Measurement of oleoresin exudation pressure in loblolly pine. Forest Sci. 14, 75-76. Hodges, J. D. and Lorio, P. L. (1971). Comparison of field techniques for measuring moisture stress in large loblolly pines. Forest Sci. 17, 220-223. Hodgson, R. W. (1917). Some abnormal water relations in citrus trees of the arid southwest and their possible significance. Univ. Calif. Publs agric. Sci. 3, 37-54.

Holmes, R. M. and Robertson, C. W. (1959). A modulated soil moisture budget. Mon. Weath. Rev. 97, 1-7. Holmgren, P., Jarvis, P. G. and Jarvis, M. S. (1965). Resistance to carbon dioxide and water vapour transfer in leaves of different plant species. Physiologia PI. 18, 557-573. Hong, S. G. and Sucoff, E. I. (1971). Measuring water potentials in pine: Reevaluation of thermocouple psychrometer and pressure chamber. Univ. Minnesota, School of Forestry. (Unpublished Manuscript.) Hsiao, T. C. (1973). Plant responses to water stress. A . Rev. PI. Physiol. 24, 5 19-570.

Huber, B. and Schmidt, E. (1936). Weitere thermoelektrische Untersuchen uber den Transpirationsstrom der Baume. Tlzarandt.forstl. Jb. 87, 369-412. Huck, M. G., Klepper, B. and Taylor, H. M. (1970). Diurnal variations in root diameter. PI. Physiol. 45, 529-530. Jeffree, C. E., Johnson, R . P. C. and Jarvis, P. G. (1971). Epicuticular wax in the stomata1 antechamber of Sitka spruce and its effect on the diffusion of water vapour and carbon dioxide. Planta 98, 1-10. Johnson, N. E. and Nielson, D. G. (1969). Pressure chamber mcasurements of water stress in individual pine fasicles. Forest Sci. 15, 452-453. Jones, J. R. (1972). Moisture stresses in Arizona mixed conifer seedlings. U.S.D.A. Forest Serv. Res. Paper RM-86, 1-8. Jordan, W. R. (1970). Growth of cotton seedlings in relation to maximum daily plant-water potential. Agron. J . 62, 699-701. Jordan, W. R. and Ritchie, J. T. (1971). Influence of soil water stress on evaporation, root absorption, and internal water status of cotton. PI. Physiol. 48, 783-788.

Jordan, W. R., Morgan, P. W. and Davenport, T. L. (1972). Water stress enhances ethylene-mediated leaf abscission in cotton. PI. Physiol. 50, 756-758. Kappen, L., Lange, 0. L., Schulze, E.-D., Evenari, M. and Buschbom, U. (1972). Extreme water stress and photosynthetic activity of the desert plant Artemisia herba-alba Asso. Oecologia 10, 177-182. Kassam, A. H. (1972). Determination of water potential and tissue characteristics of leaves of Vicia faba L. Hort. Res. 12, 13-23. Kassam, A. H. (1973). The influence of light and water deficit upon diffusive resistance of leaves of T7icia fabia L. New Phytol. 72, 557-570. Kaufmann, M. R. (196%). Evaluation of the pressure chamber technique for estimating plant water potential of forest tree species. Forest Sci. 14, 369374.

Kaufmann, M. R. (196813). Evaluation of the pressure chamber method for measurement of water stress in Citrus. Proc. Am. SOC.hort. Sci. 93, 186190.

Kaufmann, M. R. (1969). Relation of slice gap width in oranges and plant water stress. J . Am. SOC.Hort. Sci. 94, 161-163. Kaufmann, M. R. and Eckard, A. N. (1971). Evaluation of water stress control with polyethylene glycols by analysis of guttation. PI. Physiol. 47, 453-456.

THE PRESSURE CHAMBER IN ECOLOQICAL RESEARCH

249

Klepper, B. (1968). Diurnal pattern of water potential in woody plants. PZ. Physiol. 43, 1931-1934. Klepper, B. and Ceccato, R. D. (1968). Determinations of leaf and fruit water potential with a pressure chamber. Hort. Res. 9, 1-7. Klepper, B., Browning, V. D. and Taylor, H. M. (1971). Stem diameter in relation to plant water status. PI. Physiol. 48, 683-685. Knipling, E. B. (1967). Measurement of leaf water potential by the dye method. Ecology 48, 1038-1040. Kochenderfer, J. and Lec, R. (1973). Indexes to transpiration by forest trees. O e w l . Plant. 8, 175-184. Kozlowski, T. T. (Ed.) (1968a). “Water Dcficits and Plant Growth.” Vol. I. Academic Press, London and New York. Kozlowski, T. T. (Ed.) (1968b). “Water Deficits and Plant Growth.” Vol. 11. Academic Press, London and New York. Kramer, P. J. (1938). Root resistance as a cause of the absorption lag. Am. J . Bot. 2 5 , 110-113. Kramer, P. J. (1969). “Plant and Soil Water Relationships-A Modern Synthesis.” McGraw-Hill, New York. Kriedemann, P. E. and Canterford, R. L. (1971). The photosynthetic activity of pear leaves (Pyrus communis L.). Aust. J . biol. Sci. 24, 197-205. Kriedemann, P. E. and Smart, R. E. (1971). Effects of irradiance, temperature, and leaf water potential on photosynthesis of vine leaves. Photosynthetica 5, 6-15. Kuramoto, R. T. and Bliss, L. C. (1970). Ecology of subalpine meadows in the Olympic Mountains, Washington. Ecol. Monogr. 40, 317-347. Lange, 0. L. (1972). 11. Wasserumsatz und Stoffbewegungen. Bericht uber die Arbeiten zum Wasserumsatz in den Jahren 1969-1971. Fortschr. Botan. 34, 91-112. Larcher, W. (1972). Der Wasserhaushalt immergruner Pflanzen im Winter. Ber. dt. bot. Qes. 85, 315-327. Lassoie, J. P. (1973). Diurnal dimensional fluctuations in a Douglas-fir stem in response to tree water status. Forest Sci. 19, 251-255. Lassoie, J. P. and Scott, D. R. M. (1972). I n “Proceedings-Research on Coniferous Forest Ecosystems-A Symposium” (Eds J. F. Franklin, L. J. Dempster and R. H. Waring), pp. 265-270. U.S.D.A. Forest Sew., Pacific Northwest. Lawlor, D. W. (1969). Plant growth in polyethylene glycol solutions in relation to the osmotic medium and the leaf water balance. J . exp. Bot. 20, 895-911. Lawlor, D. W. (1972). An automatic multichannel thermocouple psychrometer based on an operational amplifier. J. appl. Ewl. 9, 581-588. Lawlor, D. W. (1973). Growth and water absorption of wheat with parts of the roots a t different water potentials. New Phytol. 72, 297-305. Levitt, J. and Storvick, T. S. (1973). On the ascent of sap. Science, N . Y . 179, 1250. Lieth, H. (1970). I n “Analysis of Temperate Forest Ecosystems” (Ed. D. E. Reichle), pp. 29-46. Springer-Verlag, Berlin. Lipe, J. A. and Morgan, P. W. (1973). Ethylene, a regulator of young fruit abscission. PI. Physiol. 51, 949-953. Lopushinsky, W. (1969). Stomata1 closure in conifer seedlings in response to leaf moisture stress. Bot. Gaz. 130, 258-263. Love, L. D. and West, N. E. (1972). Plant moisture stress patterns in Eurotia lanata and Atriplex confertifololia.Northweat Sci. 46, 44-51.

250

GUY

A. RITCHIE

and

THOMAS M. HINCKLEY

Lupke, B. von (1973a). Wachstum junger Fichten in Abhiingigkeit von ihrem Frischezustand vor der Pflanzung. Allg. Forst. u.Jagdztg 144, 141-146. Lupke, B. von (1973b). Wasserhaushalt junger Fichten nach dem Verpflanzen. Foratw. Cbl. 92, 311-327. Mark, W. R. and Reid, C. P. P. (1971). Lodgepole pine-dwarf mistletoe xylem water potentials. Forest Sci. 17, 470-471. Mason, R. S. (1969). A simple technique for measuring oleoresin exudation flow in pine. Forest Sci. 15, 56-57. McMichael, B. L., Jordan, W. R. and Powell, R. D. (1972). An effect of water stress on ethylene production by intact cotton petioles. P1. Physiol. 49, 658-660.

McMichael, B. L., Jordan, W. R. and Powell, R. D. (1973). Abcission processes in cotton: Induction by plant water deficit. Agron. J . 65, 202-204. Meidner, H. and Mansfield, T. A. (1968). “Physiology of Stomata.” McGraw-Hill, London. Meyer, R. F. and Boyer, J. S. (1972). Sensitivity of cell division and cell elongation to low water potentials in soybean hypocotyls. Phnta 108, 77-87. Molz, F. J. and Klepper, B. (1972). Radial propagation of water potential in stems. Agron. J . 64, 469-473. Molz, F. J. and Klepper, B. (1973). On the mechanism of water-stress-induced stem deformation. Agron. J . 65, 304-306. Molz, F. J. Klepper, B. and Browning, V. D. (1973). Radial diffusion of free energy in stem phloem: An experimental study. Agron. J . 65, 219-222. Moore, R. T., White, R. S. and Caldwell, M. M. (1972). Transpiration of Atriplex confertifolia and Eurotia lanata in relation to soil, plant, and atmospheric moisture stresses. Can. J . Bot. 50, 2411-2418. Namken, L. N., Bartholic, J. F. and Runkles, J. R. (1969). Monitoring cotton plant stem radius as an indication of water stress. Agron. J . 61, 891-893. Namken, L. N., Bartholic, J. F. and Runkles, J. R. (1971). Water stress and stem radial contraction of cotton plants (Qosaypium Airautum L.) under field conditions. Agron. J . 63, 623-627. Newman, E. I. (1969). Resistance to water flow in soil and plant. 11.A review of experimental evidence on the rhizosphere resistance. J . appl. E d . 6, 261-272.

Oechel, W. C., Strain, B. R. and Odening, W. R. (1972a). Tissue water potential, photosynthesis, W-labeled photosynthate utilization, and growth in the desert shrub Lamea divaricata Cav. Ecol. Monogr. 42, 127-141. Oechel, W. C., Strain, B. R. and Odening, W. R. (197213). Photosynthetic rates of a desert shrub, Larrea divaricata Cav., under field conditions. PhotoMjnthetica 6, 183-188. Oppenheimer, H. R. (1960). Adaptation to drought: Xerophytism. UNESCO Arid Zone Rea. 15, 105-138. Pazourek, J. (1968). Effect of age on water loss of leaves. Biologia PI. 10, 27-30. Pearcy, R. W., Bjorkman, O., Harrison, A. T. and Mooney, H. A. (1971). Photosynthetic performance of two desert species with C, photosynthesis in Death Valley, California. Carnegie Instit. Year Book 70, 540-550. Pierpoint, G. (1967). Direct measurement of internal moisture deficit in trees. For. Chron. 43, 145-148. Plumb, R. C. and Bridgman, W. B. (1972). Ascent of sap in trees. Science, N . Y . 176, 1129-1131.

Puritch, G. S. (1973). Effect of water stress on photosynthesis, respiration, and transpiration of four Abiea species. Can. J . For. Rea. 3, 293-298.

THE PRESSURE CHAMBER I N ECOLOGICAL RESEARCH

251

Puritch, G . S . and Turner, J. A. (1973). Effects of pressure increase and release on temperature within a pressure chamber used to estimate plant water potential. J . exp. Bot. 24, 342-348. Reed, K. L. and Waring, R. H. (1974). Coupling of environment to plant response: A simulation model of transpiration. Ecology 55, 62-72. Richardson, C. J., Dinger, B. E. and Harris, W. F. (1973). The use of stomatal resistance, photopigments, nitrogen, water potential, and radiation to estimate net photosynthesis in Liriodendron tulipqera L.-a physiological index. EDFB-IBP-72-13Oak Ridge Natnl Lab., Oak Ridge, Tenn. Richter, H. (1972). Wie entstehen Saugijpannungsgradienten in Biiumen?. Ber. dt. bot. Ctea. 85, 341-351. Richter, H. (1973). On the ascent of sap. Science, N . Y . 179, 1249. Richter, H. and Rottenburg, W. (1971). Leitfahigkeitsmessug zur Endpunktanzeige bei der Saugspannungsbestimmung nach Scholander. Flora 160, 440-443. Richter, H., Halbwachs, G. and Holzner, W. (1972). Saugspannungsmessungen in der Krone eines Mammutbaumes (Sequoiadendrongiganteum). Flora 161, 401-420. Ritchie, 0. A. (1971). Transpiration, water potential and stomatal activity in relation to microclimate in Abiea amabilia and A . procera in a natural environment. Ph.D. Dissertation, Univ. of Washington, Seattle. Ritchie, G. A. and Hinckley, T. M. (1971). Evidence for error in pressure-bomb estimates of stem xylem potentials. Ecology 52, 534-536. Ritchie, J. T. (1973). Influence of soil water status and meteorological conditions on evaporation from a corn canopy. Agron. J . 65, 893-897. Ritchie, J. T. and Jordan, W. R. (1972). Dryland evaporative flux in a subhumid climate: IV. Relation to plant water status. Agron. J . 64, 173176. Rook, D. A. (1973). Conditioning radiata pine seedlings to transplanting by restricted watering. N . Z . For. Sci. 3, 54-69. Rundel, P. W. (1972). Habitat restriction in giant sequoia: The environmental control of grove boundaries. Am. Midl. Nat. 87, 81-99. Rundel, P. W. (1973). The relationship between basal fire scars and crown damage in giant sequoia. Ecology 54, 210-213. Sankary, M. N. and Barbour, M. G. (1972a). Autecology of Atriplex polycarpa from California. Ecology 58, 1155-1 162. Snnkary, M. N. and Barbour, M. G. (1972b). Autecology ofSalaola vermiculata var. villoaa from Syria. Flora 161, 421-439. Scholander, P. F. (1973). On the ascent of sap. Science, N . Y . 179, 1248. Scholander, P. F. and Perez, M. (1968). Sap tension in flooded trees and bushes of the Amazon. P1. Phyaiol. 43, 1870-1873. Scholander, P. F., Hammel, H. T., Hemmingsen, E. A. and Bradstreet, E. D. (1964) Hydrostatic pressure and osmotic potential in leaves of mangroves and some other plants. Proc. natn. A d . Sci. U.S.A. 52, 119-125. Scholander, P. F., Hammel, H. T., Bradstreet, E. D. and Hcmmingsen, E. A. (1966a). Sap pressure in vascular plants. Science, N . Y . 148, 339-346. Scholander, P. F., Hammel, H. T., Bradstreet, E. D. and Hemmingsen, E. A. (1965b). Sap pressure in plants. Science, N . Y . 149, 920-922. Scholander, P. F., Bradstreet, E. D., Hammel, H. T. and Hemmingsen, E. A. (1966). Sap concentrations in halophytes and some other plants. PI. Phyaiol. 41, 529-532.

252

GARY A. RITCHIE

and

T H O I d S M. HlNCKLEY

Schulze, E.-D., Lange, 0. L., Kappen, L., Buschbom, U. and Evenari, M. (1973). Stomata1 responses to changes in temperature at increasing water stress. Planta 110, 29-42. Shardakov, V. W. (1948). New field method for the determination of the suction pressure of plants. (In Russian). Dokl. Akad. Nauk. S S R 60, 169-172. Skau, C. M. and Swanson, R. H. (1963). An improved heat pulse velocity meter as an indicator of sap speed and transpiration. J . geophys. Res. 68, 47434749. Slatyer, R. 0. (1967). “Plant-Water Relationships.” Academic Press, London and New York. Slatyer, R . 0. and Bierhuizen, J. F. (1964).Transpiration from cotton leaves under a range of environmental conditions in relation to internal and external diffusive resistances. Aust. J . biol. Sci. 17, 115-130. Slatyer, R. 0. and Taylor, S. A. (1960). Terminology in plant-soil-water relations. Nature, Lond. 187, 922-924. Slavik, B. (1965). In “Water Stress in Plants” (Ed. B. Slavik), pp. 195-202. Dr. Junk Publishers, The Hague. Small, E. (1972). Water relations of plants in raised sphagnum peat bogs. Ecology 53, 726-728. Spanner, D. C . (1964). “Introduction to Thermodynamics.” Academic Press, London and New York. Spomer, G. G. (1968). Sensors monitor tensions in transpiration streams of trees. Science, N . Y . 161, 484-485. Spomer, G. G. (1973). The concepts of “interaction” and “operational environment” in environmental analysis. Ecology 54, 200-204. Spomer, L. A. and Langhans, R. W. (1972). Evaluation of pressure bomb and dye method measurements of tissue water potential in greenhouse Chrysanthemum. Hort Sci. 7, 412-413. Stark, N. (1970). Water balance of some warm desert plants in a wet year. J . Hydrol. 10, 113-126. Stevenson, K. R. and Shaw, R. H. (1971). Diurnal changes in leaf resistance to water vapor diffusion at different heights in a soybean canopy. Agron. J . 63, 17-19. Sncoff, E. (1972). Water potential in red pine: Soil moisture, evapotranspiration, crown position. Ecology 53, 681-686. Swanson, R. H. (1972). Water t,ranspired by trees is indicated by heat pulse velocity. Agric. Meteorol. 10, 277-281. Taerum, R. (1973). Occurrence of inverted water potcntial gradients between soil and bean roots. Physiologia P1. 28, 471-475. Taylor, S. A. and Slatyer, R. 0. (1961). Water-soil-plant relations terminology. P,roc. int. Congr. Soil Sci. 1, 394-403. Taylor, S. A. and Slatyer, R. 0. (1962). Proposals for a unified terminology in studies of plant-soil-water relations. UNESCO Arid Zone Res. 16, 339349. Teal, J. M. and Kanwisher, J. W. (1970). Total energy balance in salt marsh grasses. Ecology 51, 690-695. Tinklin, R. and Weatherley, P. E. (1968). The effect of transpiration rate on the leaf water potential of sand and soil rooted plants. New Phytol. 67, 605-615. Tobiessen, P. (1969). An inexpensive pressure chamber for xylem water tension measurements. Ecology 50, 726-728. Tobiessen, P., Rundel, P. W. and Stecker, R. E. (1971). Water potential gradient in a tall Sequoiadendron. PI. Physiol. 48, 303-304.

THE PRESSURX CHAMBER IN ECOLOGICAL RESEAFLCH

253

Tranquillini, W. (1973). Der Wasserhaushalt junger Forstpflanzen nach dem Versetzen und seine Beeinflussbarkeit. Cbl. gee. Fwstweeen 90, 46-52. Turner, N. C. (1969). Stomatal resistance to transpiration in three contrasting canopies. Crop Sci. 9, 303-307. Turner,N. C. (1972). In “Phytotoxins in Plant Diseases” (Eds R. K. S. Wood, A. Ballio and A. Graniti), pp. 407-412. Academic Press, London and New York. Turner, N. C. and Begg, J. E. (1973). Stomatal behavior and water status of maize, sorghum, and tobacco under field conditions. I. At high soil water potential. PI. Physiol. 51, 31-36. Turner, N. C. and Dimond, A. E. (1972). Water balance and membrane damage studies with the pressure chamber technique. Phytopath. 62, 501. Turner, N. C. and Graniti, A. (1969). Fuaicoccin: A fungal toxin that opens stomata. Nature, Lon& 223, 1070-1071. Turner, N. C. and Waggoner, P. E. (1968). Effects of changing stornatal width in a red pine forest on soil water content, leaf water potential, bole diameter and growth. PI. Physiol. 43, 973-978. Turner, N. C., DeRoo, H. end Wright, W. H. (1971). A pressure chamber for the measurement of plant water potential. Conn. Ag. Exp. Sta. Special SoiL Bull. XXxIII. Tyree, M. T. and Dainty, J. (1973). The water relations of hemlock (Tsuga canadensis). 11. The kinetics of water exchange between the symplast and apoplast. Can. J. Bot. 51, 1481-1489. Tyree, M. T. and Hammel, H. T. (1972). The measurement of the turgor pressure and the water relations of plants by the pressure-bomb technique. J. exp. Bot. 23, 267-282. Tyree, M. T., Dainty, J. and Benis, M. (1973a). The water relations of hemlock (Tsuga canaden&). I. Some equilibrium water relations as measured by the pressure-bomb technique. Can. J. Bot. 51, 1471-1480. Tyree, M. T., Benis, M. and Dainty, J. (1973b). The water relations of hemlock (Teuga canadensis). 111.The temperature dependence of water exchange in a pressure bomb. Can. J. Bot. 51, 1537-1543. van den Honert, T. H. (1948). Water transport in plants as a catenary process. Discuss. Farahy SOC.3, 146-153. Veihmeyer, F. J. and Hendrickson, A. H. (1950). Soil moisture in relation to plant growth. A . Rev. PI. Physiol. 1, 285-304. Waggoner, P. E. and Turner, N. C. (1971). Transpiration and its control by stomata in a pine forest. Conn. ag. Exp. Sta. Bull. 726. Wambolt, C. L. (1973). Conifer water potential as influenced by stand density and environmental factors. Can. J. Bot. 51, 2333-2337. Waring, R. H. (1969a). Forest plants of the eastern Siskiyous: Their environmental and vegetational distribution. Northweet Sci. 43, 1-17. Waring, R. H. (196913). I n “Regeneration of Ponderosa Pine” (Ed. R. K. Hermann), pp. 54-61. Oregon State Univ., Corvallis, Oregon, U.S.A. Waring, R. H. and Cleary, B. D. (1967). Plant moisture stress: Evaluation by pressure bomb. Science, N . Y. 155, 1248-1254. Waring, R. H. and Youngberg, C. T. (1972). Evaluating forest sites for potential growth response of trees to fertilizer. Northweat Sci. 46, 67-75. Waring, R . H., Reed, K. L. and Emmingham, W. H. (1972). In ‘‘ProceedingsResearch on Coniferous Forest Ecosystems-A Symposium” (Eds J. F. Franklin, L. J. Dempster and R. H. Waring), pp. 79-90. U.S.D.A. Forest Serv., Pacific Northwest.

254

GARY A. RITCHIE

and

THOMAS M. HINCKLEY

Weatherley, P. E. (1950). Studies in the water relations of the cotton plant. I. The field measurement of water deficits in leaves. New Phytol. 49, 81-97. Weatherley, P. E. (1963). I n “The Water Relations of Plants” (Eds A. J. Rutter and F. H. Whitehead), pp. 85-100. Blackwell, London. West, D. W. and Gaff, D. F. (1971). An error in the calibration of xylem-water potential against leaf-water potential. J . u p . Bot. 22, 342-346. Wiebe, H. H., Campbell, G. S., Gardner, W. R., Rawlins, S. L., Cary, J. W. and Brown, R. W. (1971). Measurement of plant and soil water status. Utah Ag. Exp. Sta. Bull. 484. Woodman, J. N. (1971). Is there a future for irrigation in the management of forests? Proc. ASAE 71-179. Worrall, J. (1966). A method of correcting dendrometer measures of tree diameter for variations induced by moisture stress change. Forest Sci. 12, 427-429. Worrall, J. (1973). Seasonal, daily, and hourly growth of height and radius in Norway spruce. Can. J . For. Res. 3, 501-511. Zaerr, J. (1971). Moisture stress and stem diameter in young Douglas-fir. Forast

Sci. 17, 466-469. Zimmerman, M. H. and Brown, C. L. (1971). “Trees, Structure and Function.” Springer-Verlag, New York, Heidelberg and Berlin.

The Ecology of Serpentine Soils J O H N PROCTOR

Biology Department, University of Stirling, Stirling, Scotland

and S T A N L E Y R. J. WOODELL

Botany School, University of Oxford, Oxford, England I. Introduction. . 11. Geologyand Soils . A. Geology . . . B. Weathering and Pedogenesis C. Clay Minerals D. A Generalview of Serpentine Soils 111. TheVegetationofSerpentine IV. TheReasonsfor SerpentineInfertility A. Physical Properties of Serpentine Soils B. Low Levels of Nitrogen, Phosphorus and Potassium in Serpentine Soils C. Nickel, Chromium and Cobalt in Serpentine Soils 1. Nickel. . 2. Chromium . . . 3. Cobalt. D. CalciumandMagnesium 1. Soil Calcium and Magnesium 2. Calcium and Magnesium in Plants 3. CalciumDeficiency . 4. Magnesium Toxicity . 5. Calcium-Magnesium Interactions 6. A High Magnesium Requirement by Serpentine Plants 7. Interaction between Magnesium, Calcium, and other Ions 8. The Relationship between Serpentine and Maritime Plants 9. The Ecological Significance of Calcium and Magnesium on Serpentine E. Other Unusual Chemical Features of Possible Importance t o Plants V. Animalson Serpentine Soils . VI. Fungiand Bacteriain Serpentine Soils VII. Evolutionon Serpentine A. Ecotypic Differentiation B. Endemics . C. The Exclusion of Serpentine Endemics from other Soils . D. Plants Showing Disjunct Distribution on Serpentines . E. Morphological Differences shown by Serpentine Plants F. Speciation ~

~

~

~

~

~

~

~

~

~

~

~

~

~

~

~

-

~

~

-

256

-

~

266 266 256 259 261 262 263 271 271 273 276 277 296 308 314 315 316 316 331 332 334 335 336 336 336 338 338 340 340 342 343 345 346 347

256

JOHN

VIII. Conclusions AcknowledgementsReferences .

PROCTOR

-

-

-

and -

-

STANLEY R. J . WOODELL

-

~

-

-

-

-

~

-

-

-

-

-

-

-

349 350 350

I. I N T R O D U C T I O N “Serpentine” is used by biologists to describe a group of ultramafic rocks and the soils derived from them (discussion of the definition of “serpentine” is given in Section I1 A). Such rocks and soils have interested botanists much more than zoologists. There is a very large body of work on the plant ecology of serpentines but very little on their animals. Serpentines frequently hold very many attractions for the botanist. They often harbour rare species, sometimes endemic to one or a few serpentine outcrops. Their vegetation is usually sharply contrasting with that of surrounding areas. Such features of serpentines have been documented throughout much of the world since the middle of the 19th century. The causes of this peculiar vegetation have been speculated upon and investigated for many years. Various investigators came to different conclusions and tried to generalize to all serpentines. This has led to unnecessary controversy, for, as we are at great pains to emphasize in this review, serpentines are exceedingly variable. They have many chemical and physical peculiarities, all of which often vary from site to site. The ecology of serpentine was comprehensively reviewed by Krause (1958). Since then a great deal has been discovered about serpentines, and they can now be seen in a much clearer light. 11. G E O L O G YA N D SOILS

A.

GEOLOGY

The origin of the term “serpentine” is discussed by Baust and Fahey (1962). I n its strictest sense “serpentine” refers to a small group of minerals, of similar chemical composition, which are products of a hydrothermal alteration. Frequently “serpentine” is used for rocks consisting mainly of serpentine group minerals (although the term “serpentinite” proposed by LodoEnikov (1936) is much better for this). Biologists have their own use of “serpentine” which has come to include most ultramafic rocks, i.e. those rich in ferromagnesian minerals even though they may not contain any minerals of the serpentine group sensu stricto. “Serpentine” used in this way is a convenient shorthand and is used throughout this review as such. It must always be under-

THE ECOLOGY O F SERPENTINE SOlLS

257

stood, however, that the term may give a misleading impression of the mineralogy of the rocks so described. Ultramafic or serpentine rocks vary greatly in chemical and mineralogical composition, but can be expected t o be composed of various combinations of the minerals olivine, orthorhombic and monoclinic pyroxenes, horneblende and the secondary alteration products of these minerals, such as the serpentine group minerals, fibrous amphiboles and talc. Various accessory minerals such as iron oxides, chromite, spinel and biotite are common. The geology of ultramafic rocks has recently been reviewed in detail in Wyllie (1967). It is appropriate t o give a brief summary of the important details of the chemistry of the minerals frequently present in ultramafic rocks. Much of the information below is from Ernst (1969). Olivine. The olivine group minerals form an isomorphous series between forsterite Mg,SiO, and fayalite Fe,++SiO, and are commonly represented as (Mg, Fe),SiO,. Rocks composed purely of olivine are called dunites, whilst rocks composed of olivine and pyroxene are peridotites. Pyroxenes. Apart from sodic varieties, two main series of pyroxenes can be distinguished. One group is calcium-rich whilst the other has only a low content of calcium. Both groups are encountered in ultramafic rocks, but the low calcium group is of particular importance. The calcium-poor orthopyroxenes are members of a two-component solid solution series extending from enstatite, MgSiO,, through hypersthene (Mg, Fe++)SiO, toward ferrosilite Fe++SiO,, although natural occurences of the pure ferrous iron end member have not been reported. Calcium-rich pyroxene end members are diopside CaMg(SiO,), and hedenbergite CaFe++(SiO,),; intermediate and subcalcic members of this series are known collectively as augite. Amphiboles. Magnesium-iron amphiboles are often important constituents of ultramafic rocks. Anthophyllites, represented by the formula (Mg, Fe++),(Si,O,,),, are reported by Ernst (1969) as “especially characteristic of metamorphosed, hydrated dunites and related rocks (serpentinites), as in Pennsylvania, North Carolina, the Hebrides, Italy and Egypt”. There are monoclinic polymorphic equiva,lentsof the orthorhombic anthophyllites, the cummingtonites which tend to be enriched in ferrous iron. Calcic amphiboles, such as tremolite Ca,Mg, (Si,Oll),(OH),, actinolite Ca,(Mg, Fe++), (Si401J2 and horneblende NaCa, (Mg, Fe ++), (Al, Fe+++)(Si,AIOll), (OH), are of importance in some ultramafic rocks. Serpentines. These minerals are of a group of layer silicates represented by the formula Mg,Si4010(OH),. Significant quantities of substitution

258

JOHN PROCTOR

and

STANLEY R. J. WOODELL

by other elements can occur, e.g. Ni, Co and Fe++frequently substitute for magnesium. Faust et al. (1956) divided serpentine minerals into two classes with regard to their minor element content. Class A serpentines are derived from ultramafic igneous rocks by hydrothermal alteration and have a relatively rich content of minor elements, particularly nickel, chromium, cobalt and scandium. Class B serpentines are derived by alteration of magnesium-rich rocks such as various types of metamorphic dolomites and limestones, and have a much lower minor element content. They are also likely to have lower iron content. The mode of origin of serpentines is of importance to biologists, since toxic trace elements are thought to be of great consequence ecologically. Moreover, it seems that class B serpentines are much more likely to be associated with calcium-containing minerals and to contain higher levels of calcium than class A serpentines. Calcium is well known to have an ameliorative effect both on the toxicity of heavy metals and on magnesium, as will be discussed later. Three principal polymorphic forms of serpentine are recognized : chrysotile, antigorite and lizardite (Whittaker and Zussman, 1956). There is evidence that these minerals are not true polymorphs (e.g. Faust and Fahey, 1962; Page, 1968). However, the serpentine minerals are at least similar in composition and in the basic arrangement of atoms in the crystal structure. Aumento (1970) has proposed a broader classification of serpentine minerals which includes some more recently discovered variants of polymorphs.

Accessory minerals. The following have been commonly reported in ultramafic rocks: ( a ) Members of the spinel group, magnetite Fe304, spinel proper MgA1,0, and chromite Fe++Cr204. ( b ) Biotites, micas of the general formula K (Mg, Fe++),Si,A10,,(OH)2, which are intermediate members of the two component series extending from phlogopite KMg,Si3A10,,(OH), to annite KFe,++Si,AlO,,(OH),. ( c ) Talc, a sheet silicate of formula Mg,Si,O,,(OH),. The above list is not exhaustive; other minerals have been reported from serpentine rocks. The proportions of these minerals are different from rock to rock and the minerals themselves are not of the exact chemical composition represented by the end member formulae. The result is that the chemical composition of serpentine rocks varies greatly, and this confirms the impression gained from the appearance of these rocks in the field. Specimens of a wide range of colours and variations are known, e.g. red, green, blue and nearly black. (Serpentines are easily worked and some of the more attractively coloured

THE ECOLOGY OF SERPENTINE SOILS

259

examples have long been used for ornamental purposes.) The only generalizations that can be made with confidence concerning the chemical composition of serpentine rocks are that iron and magnesium are always relatively high and silicon relatively low. Nevertheless serpentines very commonly have other features of their chemical composition which may exert unusual influences on the soils and living things associated with them. Many have a low calcium, potassium, phosphorus and molybdenum content, for example, and also a relatively high nickel, chromium and cobalt content (Krause, 1958). B.

WEATHERING A N D PEDOGIENESIS

Many authors (Pichi-Sermolli, 1948; Rune, 1953; Krause, 1958) have maintained that serpentine rocks are characteristically resistant to weathering. de Sequeira (1969) considered this doubtful since the main constituents of the rocks, viz. olivine, pyroxene and serpentine, are generally accepted as easily weathered minerals (e.g. by Mitchell, 1964). However, much must depend on the extent t o which the rocks are broken up by physical weathering to offer a larger surface area for chemical weathering. We have seen serpentines which weather slowly and others which weather more rapidly even in the same climate. For example, in Sweden the almost total lack of rock fragments and soil, together with the massive humped shape of the sites at Junsterklumpen and Atoklinten (Rune, 1953; Proctor, 1969), indicate a great resistance to weathering. By contrast, at Kittelfjiill and Bunnerviken (Rune, 1953; Proctor, 1969) the serpentine rocks have weathered to give a, stony ochreous soil. Clearly serpentines weather at different rates and this is likely to be of considerable importance for organisms. Proctor and Woodell (1971) and Spence (1957) have commented on the ecological significance of weathering rate within a single serpentine area in Shetland, Great Britain. Over the past decade there have been important laboratory experimental studies on the weathering of serpentine rocks. PBdro and Bitar (1966a, b) distinguished two important processes by which serpentine rock could be weathered in conditions of free drainage. When subjected to leaching with pure water, or water containing carbon dioxide, silica was removed more rapidly than magnesium, and magnesium compounds of a non-silicate type remained in the weathered crust. Iron oxides tended to accumulate and the process corresponded to a laterization. On the other hand, if acetic acid was used as the leachate there was an intense removal of magnesium compared with silica. n e e amorphous silica was formed on the rock surface and the process corresponded to a

260

JOHN PROCTOR

and

STANLEY R. J. WOODELL

podzolization, although iron was resistant to mobilization, apparently whilst magnesium was being removed. PBdro and Bitar emphasized that drainage is a most important factor in the weathering process, and showed, using a neutral or slightly acid leaching solution, that below a certain level of drainage the rate of leaching of silica relative to magnesium reversed, and magnesium was eliminated more rapidly than silica. The relative rates of leaching of silica and magnesium have also been shown to depend on temperature (PBdro and Delmas, 1971). Wildman et al. (1968a) leached serpentine rocks from four different sites with solutions containing different concentrations of carbon dioxide. The rates of dissolution of both magnesium and silica increased with increasing carbon dioxide concentrations. At the higher carbon dioxide concentrations more magnesium was removed than silica, which agrees with the observations of PBdro and Bitar since carbon dioxide concentration is related to pH. Differences between serpentine5 in the relative proportions of magnesium and silicon released on treatment with the solutions containing CO, were also demonstrated. It seems that relative rates of silica and magnesium removal can be altered quite easily by changes in conditions that are likely to occur in soils. Differences in relative rates of removal have important effects on the properties of the soil. On the overall changes in composition during soil development on serpentine rocks in the field, most records indicate a loss of magnesium relative to silica and a loss of both magnesium and silica relative to aluminium and ferric oxides. Such has been shown, for example, by Butler (1953) in Cornwall, England; Bogatyrev (1958) in Albania; Harada (1953) and Kanno et al. (1965b) in Japan; Hoyos de Castro (1960) in southern Spain; Veniale and Van der Marel (1963) in northern Italy; and Wildman et al. (1968b) in California. The data of Sasaki et al. (1968) for a serpentine podzol in Japan is an exception to this. I n this soil there was a marked increase in SiOz relative to A1,0, and Fe,O,. MgO was lost relative to SiO, and Al,O, but increased relative to Pe,O,. de Sequeira (1969) has suggested that both the weathering processes postulated by PBdro and Bitar (1966a, b) can occur simultaneously within the same soil profile. He found in north-east Portugal that in the lower horizons of well drained serpentine soils the rate of leaching was MgO > SiO, > R,O,. This corresponds to the ‘(laterization” or “sialitization” observed by PBdro and Bitar. I n the upper horizons of deeper soils the order of leaching was reversed R,O, > SiO, > MgO, and probably corresponds to a “podzolization” of PBdro and Bitar. de Sequeira went on to suggest that the two weathering processes of “laterization” and “podzolization” go on simultaneously in the same horizons. This

THE ECOLOGY OF SERPENTINE SOILS

261

would imply that there are pH differences within horizons, and de Sequeira confirmed this by measurement. The most dramatic changes of composition from rock to soil occur in the tropics, where removal of silicon and magnesium, and the accumulation of metal oxides, reaches its greatest extent. Frasche (1941) described a weathered serpentine area in the Philippine Islands which was comparatively barren, and had extensive areas of bare red soil. Analysis of 2000 samples showed an average of 47.8% iron in the soils. The silica, content throughout most of the soil profiles was less than 2%. Similar changes have been observed by Bennett and Allison (1928) in Cuba, Robinson et al. (1935) in Cuba and Puerto Rico, and Schellman (1964) in Borneo, and by Hotz (1964) in some extremely weathered soils on partially serpentinized peridotites in northern U.S.A. It is clear that as a result of these weathering processes the chemical composition of serpentine soils not only differs markedly from the chemical composition of serpentine rocks, but does so in a way that depends on a number of factors.

c. C L A Y

MINERALS

The proportion of weathering products (minerals synthesized or substantially altered during weathering) and of weathering residues (minerals present in the parent rock and remaining unaltered in the soil) in serpentine soils is variable. There have been reports of soils on serpentine rocks in which the clay fraction appears to consist largely of weathering residues. Iberg (1954) found in the clay complex of an alpine Swiss serpentine soil only antigorite and chrysotile, i.e. forms of serpentine. Wildman (1967) considered that the clay minerals on northern Italian serpentines reported by Malquori and Cecconi (1956) and Veniale and Van der Marel (1963) are all residual. Most serpentine soils that have been investigated appear to contain weathering products as well as residual minerals in the clay fraction. Apart from serpentine group minerals the clay fraction of serpentine soils usually contains at least one of the groups listed below.

Kaolinite group, e.g. kaolinite and halloysite. These are 1 : 1 type minerals composed of alternate layers of silica and alumina sheets. Montmorillonite group, e.g. niontmorillonite, beidellite, nontronite and saponite. These are 2 : 1 type minerals made up of units of two silica sheets with an alumina sheet sandwiched between them. Chlorite group, e.g. chlorite. These are typically made up of alternate talc (similar to a montmorillonite crystal unit) and brucite (Mg(OH),) layers. Large quantities of magnesium are present in the “alumina sheet” of the talc layer. The basic crystal unit thus consists of two silica

262

JOHN PROCTOR

and

STANLEY R. J. WOODELL

sheets and two magnesium sheets and is said to have a 2 : 2 structure. Veriniculife group, e.g. vermiculite. The structure of vermiculite is similar to that of chlorite except that in the crystal unit a small number of highly hydrated magnesium ions take the place of the brucite layer. Various non-crystalline hydrous oxides, e.g. gibbsite (A1,0,. 3H,O) and goethite (Fe,O,. H,O). Wildman (1967) related the weathering products of serpentine soils to the weathering intensity. He regarded the loss of silica (irrespective of other changes in chemical composition) as a key factor. Wildman envisaged that in situations of low weathering intensity, clay mineral weathering products are likely to be those with a higher silica content, e.g. montmorillonite, followed with increasing weathering intensity by secondary silicates of lower silica content, e.g. chlorite and kaolinite. The ultimate stage is reached in lateritic soils which have high proportions of metal oxides. It is true to say that where clay mineral analyses of serpentine soils have been made, the results usually agree with these views (e.g. Kelley et al., 1939; Butler, 1953; Bogatyrev, 1958; Litvinenka, 1962; Hotz, 1964; Schellman, 1964; Kanno et al., 1965a; Wildman et al., 1968b; Wilson, 1969; Haantjens and Bleeker, 1970). There is a great difference in the quantity of clays in serpentine soils. Some are exceedingly rocky and stony, others have a very high clay content, a,nd still others show an intermediate condition.

D.

A G E N E R A L V I E W O F S E R P E N T I N E SOILS

The processes by which serpentine rocks develop into soils depend on climate, time, relief and biotic factors as well as on the chemical composition of the parent material. As a result many types of soil occur on serpentines. Some of the soil types that have been recorded are given below : podzols (Nogina, 1948; Ragg and Ball, 1964; Sasaki et al., 1968); laterites (Bennett and Allison, 1928; Frasche, 1941; Birrell and Wright, 1945; Hotz, 1964); brown ranker, eutrophic ranker, eutrophic braunerde, pseudogley and anmoor (Coombe and Frost, 1956a, b); eutrophic braunerde, slope gleys, and ground water gleys (Wilson, 1969); “smonitsa”, i.e. dark coloured solonetz-like soils (Bogatyrev, 1958); very immature skeletal or lithosolic soils (Rune, 1953; Spence, 1957; de Sequeira, 1969; Kruckeberg, 1969a; Proctor and Woodell, 1971). This list is by no means exhaustive. Although serpentine soils differ

THE ECOLOGY O F SERPENTINE SOILS

263

greatly from one another, they are frequently distinctive in some way, and are regarded as intrazonal or azonal in the region where they occur. The many types of serpentine soil have the common property of frequently being covered by a distinctive vegetation. We shall briefly survey the characteristics of this serpentine vegetation before going on to discuss the factors which control it.

111. T H E VEGETATION O F SERPENTINE Caesalpino (1583) described a plant restricted to the “black stones” (“sassi neri” is still the Tuscan term for serpentine rocks) of the upper Tiber Valley. It was “Lunaria quarta alias Alysson” and the description corresponds to that of Alyssum bertolonii, which is still abundant in the area. This appears to be the first published record of a serpentine plant. Modern interest in serpentine vegetation began in the mid 19th century with Amidei’s (1841) list of plants restricted to serpentine in a locality in Italy, and PanEiE’s (1859) description of the vegetation of a serpentine site in Serbia. The numerous subsequent accounts of Serpentine vegetation are indicative of the interest in its distinctive character. There are good reviews by Rune (1953), Whittaker (1954b) and Krause (1958). Table I lists many papers containing descriptive accounts of serpentine vegetation region by region. It seems certain that the distribution of papers on serpentine vegetation is correlated with the presence of interested botanists. For instance there are many papers on serpentine in Italy, but TABLE I Selection of works which contain descriptive studies of serpentine vegetation Europe Albania Austria

Balkans Czechoslovakia Finland France Germany Greece

Markgraf (1925, 1932) Eggler (1954, 1955); Hayek (1923); Kretschmer (1931); Liimmermayr (1926, 1927, 1928a, b, 1930, 1934); Maurer (1966); Nevole (1926) Adamovio (1909); Beck von Mannagetta (1901); Krause and Ludwig (1956, 1957); Krause et al. (1963) DvofAk (1935); NovAk (1928, 1937); Suza (1921, 1928, 1930); Zlatnik (1928a, b) Kotilainen (1944); Kotilainen and Seivala (1954); Mikkola (1938); Rune (1953) Duvigneaud (1966); LeGendre (1919) Dahl (1912, 1915); Gauckler (1954); Krapfenbauer (1967) Boydell (1921);Krause (1962);Krause and Klement (1962); Turrill (1929)

264

JOHN PROCTOR

and

STANLEY R. J. WOODELL

TABLEI (continued) So6 (1934); Zolyomi (1936) Amidei (1841); Bargoni (1943); Becherer (1969); Fiori (1914, 1919); Gismondi (1953); Martino and Orsino (1969); Messeri (1936); Negodi (1941); Pampanini (1903); Pavarino (1912, 1914, 1918); Pichi-Sermolli (1948); Rigotti (1930); Vergnano (1953, 1959b) Bjorlykke (1938); Knaben (1952); Svenonius (1883) Norway Macko and Sarosiek (1961); Sarosiek and Sakowska Poland (1961); Sarosiek (1964) Portugal Pinto da Silvn (1965, 1970) Spain Palacios (1936); Rivas-Goday (1969) Sweden Rune (1953) Switzerland Beger (1922, 1923); Braun-Blanquet and Jenny (1926); Braun-Blanquet (1932, 1951); Zollitsch (1927) United Kingdom Coombe and Frost (1956a, b); Ferreira (1959); Halliday (1960); Hunter and Vergnano (1952); Marshall (1959); Proctor and Woodell (1971); Spence (1957, 1958, 1969, 1970); West (1912) Yugoslavia Krause and Klement (1958); Novak (1928); PanEiE (1859); PavloviE (1953, 1955, 1962, 1964); Ritter-StudniEka (1956, 1963, 1964, 1970); Ritter-Studnicka and Klement (1968) Asia Indonesia Lam (1927) Kitamura (1950, 1952a, b); Kitamura and Momotani Japan (1952); Kitamura and Murata (1952); Kitamura et at!. (1950, 1953); Nagano et al. (1966); Taniguti (1958); Yamanaka (1952, 1954, 1959, 1967) Igoshina (1966); Iljinski (1936); NovBk (1926); SoEava Russia (1927) North America Fernald (1907, 1911); Low (1884); Raymond (1950); Rune Canada (1953, 1954); Scoggan (1950) Braun (1950); Harshberger (1903, 1904); Pennell (1910, U.S.A. East 1913; 1930); Radford (1948); Shreve (1910); Wherry (1963) U.S.A. West Kruckeberg (1950, 1951, 1964, 1967, 1969a, b); Maas and Stuntz (1969); Mason (1946); Walker (1948a); Whittaker (1954a, b, 1960) Central America Beard (1953); Carabia (1945); Seifrie (1940, 1943) Cuba Holdridge (1945) Puerto Rico Hungary Italy

Australasia New Caledonia New Zealand Africa Rhodesia Ghana ~~

Birrell and Wright (1945); Daniker (1939) Betts (1918, 1919, 1920); Cockayne (1928); Davidson el al. (1969); Lyon et al. (1971) Wild (1965, 1970, 1974a, b) Woodell and Newton (1974)

THE ECOLOGY O F SERPENTINE SOILS

265

virtually nothing on Australia and Africa (except Rhodesia), though there is serpentine in those regions. The existence of the reviews mentioned makes it unnecessary to describe serpentine vegetation in detail. We shall attempt to summarize its characteristics, referring to more recent work where it has added information to existing descriptions. Serpentine vegetation is extremely variable; the surprising divergence of types within the small area of Great Britain (Proctor and Woodell, 1971) illustrates this. Although this variability compels caution in making generalizations, botanists have undoubtedly concentrated on the more interesting types of serpentine, which show several features in common, and which are the type referred to here unless otherwise stated. Serpentine vegetation has two major characteristics: 1. Physiognomic differences from the vegetation of surrounding rocks. 2. Rare species and combinations of species.

I n this section we shall consider the first characteristic; the second is discussed in Section VII. The relative openness o f the plant cover has impressed many biologists. Braun-Blanquet (1932) stated: “The serpentine ridges of the Alps in their dark deathly hardness are among the most depressing lonely phenomena of nature, and the popular reference of “Tote Alp” (dead Alp) is quite appropriate.” Kruckeberg (1969a) described the situation thus: “Stark contrast between the barrenness of ultramafics and the comparative luxuriance of adjacent non-ultramafic sites, and the pronounced differences in species composition, are familiar and striking features of this discontinuity in vegetation wrought by geology”. He went on: “In varying degrees, plant life on ultramafics all over the world casts the same spell, conjures the same bleak images, and excites the analytical mind”. There are good illustrations of this barrenness and contrast in Krause (1958), Whittaker (1954b), and especially the photographs in Markgraf (1932), Rune (1953)and Kruckeberg (1969a). There are few quantitative data for barren areas. One site which has been cited in this respect is the Keen of Hamar, Unst, Shetland. West (1912) noted that the upper part of this site was “90% bare”. Spence (1959) referred to cover varying from 1% up to about 16% on the debris, and Proctor and Woodell (1971) measured a cover of 6.3% on the same site. Proctor and Woodell found total cover varying from 2.1 to 20.7y0 on eleven of the sites they surveyed in Scotland. It should be emphasized that these figures refer to the relatively barren debris areas; all of these sites had more complete cover in parts. Serpentine sites often carry more xeric vegetation than adjacent areas, and this is frequently of a type characteristic of higher altitudes,

266

JOHN PROCTOR

and

STANLEY R. J. WOODELL

or of earlier stages in succession. One of the best examples of the contrasts between serpentine and non-serpentine vegetation in a single area is that of the Siskiyou Mountains, in the western United States (Whittaker, 1954b, 1960), and we shall make extensive use of it as a basis for this descriptive section. Whittaker compared adjacent serpentine and diorite outcrops in the Siskiyou Mountains of southern Oregon. The “typical” vegetation of the area is on quartz diorite and other rocks. Whittaker noted that the vegetation of these rocks and serpentine were “so different . . .that one might imagine that one had travelled to another part of the continent”. The diorite mountains are covered by dense evergreen forests, mainly Douglas Fir (Psezdotsuga tazi;foZia),Port Orford Cedar (Chamaecyparis lawsoniana) and sclerophyllous shrubs, and these forests are valuable timber sources. There is a gradient on diorite from the more mesic sites to xeric areas on southerly slopes where scattered Douglas Firs occur above a canopy of sclerophyllous shrubs. On the serpentine the same two tree dominants occur on the more mesic sites, but they form open stands together with Pinus monticoZa. Beneath them are shrubs, and an assemblage of rare serpentine herbs and commoner species of bogs and marshes. I n less mesic sites, there are mixed stands of various conifers, with a number of sclerophyllous shrubs; many of these are shrubby varieties of the sclerophyllous trees of diorite sites. I n the more mesic sites, shrub cover may be up to SO%, but this is much reduced on xeric sites, where the shrubs are in scattered patches, interspersed with grassland. On the most xeric sites the whole vegetation is grassy, with scattered dwarfed pines. Whittaker described this open savannah as “pine steppe”. He summed up the differences (1954b)between diorite and serpentine vegetation in a diagram (Fig. 1) and remarked that much of the contrast between the vegetation of the two rock-types is a consequence of the exclusion from one vegetation-type of growth forms important on theother. The broad-leaved trees are absent, or represented only as shrubs, on the serpentine; the pines are absent or of minor importance on the diorite. I n the coast ranges of southern California Woodell et al. (1974b) have noted a mosaic of vegetation types, including oak woodland, digger pine (P.sabiniana) woodland, chaparral and grassland, the last occurring on the serpentine. Birrell and Wright (1945) found xerophyll scrub vegetation a few feet high in New Caledonia, in an area where tropical rain forest is the typical cover on non-serpentine soils. Kruckeberg (19694 has many photographs of this type of contrast. Life-form spectra are also illustrative of the differences between serpentine and other vegetation (Table 11). The spectrum for the quartz diorite of the Siskiyous is similar to that of mid-temperate

FIG.1. Transect from mesic to xeric regions, on quartz diorite and serpentine, showing vegetation

patterns. (Reproduced by permission Duke University Press. From Whittaker, 1954b.)

, Other

herbs

Mesic broadleaf trees

? Incense cedar

Y Other shrubs

Bracken

y Vaccinium

Sclerophyllous trees

Douglas fir

A

Achlys

8 Sclerophyllous shrubs

Port Orford cedar

Jeffrey pine

$

DIORITE

2000 f e e t

Grass

1

gradient at l o w elevations,

SERPENTINE

moisture

f Arctostaphylos

ON

ON

along the

Western white pine

KEY:

Change of vegetation

VEGETATION PATTERNS IN THE SISKIYOU MOUNTAINS

TABLEI1 Life form spectra of serpentine and non-serpentineflora

Habitat

% of Phanerophytes

Chamaephytes

Hemicryptophytes Geophytes

Therophytes No. of species

__

1. Quartz diorite,

Siskiyou Mountains, Oregon, U.S.A. 2. Serpentine, Siskiyou Mountains, Oregon, U.S.A. 3. Diorite-mesic, Siskiyou Mountains, Oregon, U.S.A. 4. Serpentine-mesic, Siskiyou Mountains, Oregon, U.S.A. 6. Diorite-xeric, Siskiyou Mountains, Oregon, U.S.A. 6. Serpentine-xeric, Siskiyou Mountains, Oregon, U.S.A. 7. Mid temperate mesophytic forest 8. Serpentine, Tuscany, Italy 9. R.aw serpentine slopes, Washington, U.S.A. 10. Serpentine debris, Scotland

32

12

30

24

20

19

44

15

35

14

28

21

2

84

2

72

ct

E

@

27

19

35

18

1

88

31

8

42

13

6

47

v)

2

z: ?J

15

17

43

14

11

76

34

8

33

23

2

-

11 0

9 14

40 80

15 6

26 0

405 -

0

7

12

58

77-5

Sources: 1-8, Whittaker (1960);9, Kruckeberg (19698);10,Proctor and Woodell (1971)

3.5

4

E

THE ECOLOGY OF SERPENTINE SOILS

269

mesophytic forest, computed by Whittaker (1960). That for serpentine shows fewer phanerophytes and more hemicryptophytes, and in xeric areas more therophytes. Kruckeberg (1969a)gave an extreme example, from raw serpentine slopes in Washington, where the flora is 80% hemicryptophytes. Whittaker also discussed the Clementsian view that distinctive vegetation types on special soils are seral, and that given time for mature soils to develop, they will converge to climatic climax. He found no evidence for this from the Siskiyou Mountains. There is no indication that the vegetation from diorite is invading serpentine, and the serpentine stands show the characteristics of self-maintenance to be expected of stable vegetation. The soils are no more immature than those on diorite, and the floristic and vegetation similarity on serpentine sites indicates that the vegetation there is stabilized. Whittaker could find no support for the view that serpentine vegetation is a stage in a sere, and interpreted it as a complex and fully developed climax pattern in its own right; we consider this to be true for many other areas also. Whittaker’s work also shows that the types of plants which occur on the Siskiyou Mountains are similar to those found on many serpentines. Coniferous trees are able to tolerate serpentine. (Griffin (1965) has indicated that Pinws sabiniana is preadapted to serpentine without development of special ecotypes. Other conifers may be similar, although Jenkinson (1966) has demonstrated serpentine ecotypes in Pinus ponderosa.) All but one of the coniferous trees of the Siskiyou diorite also occur on serpentine, and deciduous shrubs appear to be less successful than evergreens on this substratum. Whittaker goes so far as to state that “most serpentine vegetations throughout the world appear to be dominated by some one or other combination of these three adaptable growth-forms-coniferous trees, sclerophyllous shrubs, and grass-like plants”. If among grass-like plants he includes forbs, we agree with this. The Siskiyou Mountains are representative rather than exceptional. We have already referred to the Californian Coast Range vegetation, where serpentine often bears species-rich grassland. Scattered pines similar to those in the Siskiyous occur on serpentine in Europe (e.g. Martino and Orsino, 1969; Ritter-StudniEka, 1970)and elsewhere in the U.S.A. (Kruckeberg, 1969a). Heathy vegetation can occur in areas of forest or herb-rich scrub (Pinto da Silva, 1970). Poor savannah or grassland occurs in areas of dense savannah or tropical forest (Wild, 1965; Woodell and Newton, 1974). Stunted scrub can occur in areas of dense beech forest (Lyon et al., 1971). The stunting of growth which occurs on serpentine has been referred to by many authors, and enhances the peculiar appearance of serpentine

270

JOHN PROCTOR

and

STANLEY R. 3. WOODELL

vegetation. For instance in the Siskiyou Mountains, trees which are large elsewhere are dwarfed on serpentine. They are also scattered, and shade-intolerant tree species can grow between them. Broad-leaved trees off the serpentine become shrubs on it and a much more complete field layer can develop. The paradoxical result is that the serpentine vegetation is often richer in numbers of species than the surrounding areas, although this is not an invariable feature (e.g. Rune, 1953; Proctor and Woodell, 1971). Del Moral (1972) highlighted this relatively high diversity on serpentine when he calculated diversity patterns on serpentine and nonserpentine soils in the Wenatchee Mountains, Washington, U.S.A. On soils where mature forest can develop, habitat heterogeneity is reduced and diversity is lower. On serpentine there is not a continuous canopy, and the more heterogeneous habitat results in greater diversity. He concluded that maximum diversity occurs at the degree of environmental harshness which prevents a closed canopy but is sufficiently mild to permit specialization; as environmental rigour increases, diversity declines, and at the same time more “generalist” species are selected. Whittaker’s data (1960) support this. In the Siskiyou Mountains the highest species diversity occurs on serpentine sites intermediate between the most mesic and xeric. It would be misleading to suggest that this clear difference between serpentine and surrounding vegetation is universal. Proctor and Woodell (1971) discussed several British serpentine areas, sometimes with very shallow soils, which do not bear a distinct vegetation. This difference from other serpentines may result from the fact that some are partly covered with superficial deposits, and also their weathering does not result in debris formation; it is also related to the problem of defining serpentine, as discussed at the beginning of this review. More work on these sites would be helpful. I n the conclusions to his paper (1954b) Whittaker cited the work of Billings (1952), who had pointed out that all environmental factors must be considered as part of the ecosystemic or total environment. This idea is now a commonplace in ecological thinking. It is explicit in Whittaker’s paper, and implicit in many later accounts of serpentine vegetation, which stress the complexity of the controlling factors. Similar environmental conditions, especially in exposed and edaphically severe situations, produce vegetation having some characteristics in common with that of serpentine. Whittaker referred to accounts of vegetation on gypsum, magnesites, dolomites and limestones, shale barrens and other soils deficient in nutrients. Polunin (1948) cited similar examples in the Canadian Arctic, and Kruckeberg (1969b) discussed similar vegetation types of abnormal edaphic conditions.

THE ECOLOGY OF SERPENTINE SOILS

27 1

Spence (1970) found that debris scree and crevice vegetation on serpentine had a good deal in common with high-altitude sites in the eastern Scottish Highlands, and with low-level sites in the western Highlands. He traced affinities between these and the vegetation of west Scandinavian base-rich talus, and Icelandic and Faroese fellfield. Proctor and Woodell (1971) refer to the barrens on the Cairnwell limestone in Scotland. Similar vegetation exists on other Scottish fellfields, such as those on quartzite in Sutherland, when there is extreme exposure. Such areas can be found in many parts of the world, at high altitudes or high latitudes. The special interest of serpentine lies in its combination of so many adverse characteristics, among which are low levels of nitrogen, phosphorus and potassium; low calcium and a high level of magnesium giving an unfavourable Ca/Mg balance; low molybdenum, relatively high nickel, chromium and cobalt in the soil, and unfavourable physical factors. These adverse features will be discussed in detail in the next section; they are mentioned here t o emphasize that the total environment provided by serpentine is uniquely unfavourable. IV. T H E R E A S O N S F O R S E R P E N T I NIE NFERTILITY Serpentine soils have often been referred to as “infertile”, meaning that they are inimical to the growth of most plants. (“Infertility” is not entirely satisfactory but it serves as a useful shorthand.) What causes this “infertility” has been the source of controversy, because some workers have failed to appreciate the complexity of the factors involved.

A.

PHYSICAL PROPERTIES O F S E R P E N T I N E SOILS

Many authors have commented on the unfavourable physical properties of serpentine soils. I n temperate regions there have been many observations of such soils which are shallow and stony, especially on steeply sloping ground (e.g. Rune, 1953; Spence, 1957; Krause, 1958; Kruckeberg, 1969a; Rai et al., 1970). Stony shallow soils have also been reported from regions of tropical rainforest (e.g. Lam, 1927; Krause, 1958). The coarse stone content and shallowness must result in a low water-holding capacity and a restricted soil depth for root penetration, and plants growing in these soils are likely to suffer from drought. The frequent scantiness of vegetation cover will exacerbate drought problems, since the plants are more exposed to excessive wind and insolation. Bradshaw (pers. comm.) has pointed out that the barrenness of some soils may be a response to a lack of vegetation cover. The soil-plant interaction is ‘Yiwo-way” and the presence of plants is necessary for

272

JOHN PROCTOR

and

STANLEY R. J. WOODELL

soil formation and stability. If vegetation is removed from a mature soil, as it has been, for example, by being killed by smelter effluents in the Lower Swansea Valley, Wales, the soil is rapidly eroded and the remaining debris often resembles that found on many naturally occurring debris areas. Walker (1954) considered “at least in western United States, that plants which are widespread on serpentine soils possess drought resistance as well as tolerance for serpentine soils’). Sarosiek (1964) thought that xerophytic plants are most successful in withstanding serpentine site conditions in lower Silesia. Certainly, many plants of serpentine soils show xeromorphic characters, but as we shall discuss, such morphological effects can also be the result of growth in conditions of low nutrient supply. Whittaker (1960) and Whatley (1965) both measured soil moisture in the western U.S.A., and neither found significant differences between serpentine and adjacent non-serpentine soils. Many serpentine soils are not very shallow or stony, particularly those developed on level ground; for example, deep weathered serpentine soils have been observed under forest in central Europe (NBmec, 1951a). Maljuga (1947) found that serpentine is able to form black earths and brown steppe soils in Russia. Ellis (1951), although pointing out that serpentine soils tend to be more shallow than those found on other rocks in Rhodesia, still reported depths of about 120 cm. The lateritic serpentine soils that have been reported from the tropics, e.g. by Bennett and Allison (1928) in Cuba and by Frasche (1941) in the Philippines, are very deep. Some authors have thought that poor drainage might be an important factor. Robinson et al. (1935) considered that in some serpentines in eastern U.S.A. there is not enough coarse material to give good internal drainage. Rune (1953) stated that many serpentine soils are very poor in coarse material, and considered that this may result in a poor drainage at least within the lower horizons. de Sequeira (1969) reported poor or slow internal drainage in four out of five Portuguese serpentine soils. Wilson (1969)found glei soils in imperfectly drained situations on serpentines in Scotland. There are indications that poor plant growth might be caused by soil instability, which has been reported by many authors, e.g. Kanno et al. (1965a),Kruckeberg (1969a),de Sequeira (1969) and Proctor and Woodell (1971). Rune (1953) has recorded a pronounced tendency of some Swedish serpentine soils to solifluction and (1954) reported an example of frost movement on serpentine on Mt Albert in Canada. He regarded the serpentine bedrock, which is close to the surface, as acting like perma-frost in delaying the disappearance of water, and hence enhancing the frost movements. In this context, it is interesting that Bennett

THE ECOLOGY OF SERPENTINE SOILS

273

and Allison (1928) reported a “remarkable resistance to erosion” in a lateritic serpentine soil in Cuba. Saline properties may be imparted to soils by high levels of magnesium (Gedroits, 1916; Usov, 1937). Bogatyrev (1958) considered that excess magnesium imparts salinity to the “smonitsa” soil of Albania. He regarded a number of soil properties to be an expression of this, e.g. lumpiness and high mechanical stability of the lumps, compactness, viscosity and stickiness when wet, vertical cracking when dry, and poor water permeability. Clearly, many serpentine soils show some physical peculiarities which are usually regarded as unfavourable for the growth of plants. It is pertinent to mention the work of Richardson and Greenwood (1967), who investigated the cause of lack of vegetation on coal-mine spoil heaps (likely to have many similar physical features to rocky serpentine outcrops) in north-eastern England. They concluded that low soil moisture was the main cause of lack of vegetation and they were unable to find any evidence of toxic chemical effects. This work illustrates that physical factors in the substratum can account for almost total lack of vegetation, even in a moist temperate climate. However, most ecological studies have emphasized the importance of chemical factors in accounting for the characteristic vegetation of serpentines. There have been a number of observations of serpentine soils showing no unusual physical characteristics yet appearing highly unfavourable to plant growth; for example, Robinson et al. (1935) observed that the very infertile Conowingo Loam of eastern U.S.A. had a similar mechanical composition to fertile agricultural soils. We agree with Kruckeberg (1969a): “I believe that physical properties alone do not account for the floristic uniqueness of ultramafic rocks. Soil chemistry provides the discriminating character.” The influence of physical factors will vary from site to site, and they must be regarded as part of the “serpentine complex’’ of factors unfavourable to plant growth, even though they are rarely of overriding importance.

B.

LOW L E V E L S O F NITROGEN, PHOSPHORUS A N D POTASSIUM I N S E R P E N T I N E SOILS

Gordon and Lipman (1926) regarded low major nutrient levels (particularly nitrogen and phosphorus) as being very important in the infertility of these soils. Their conclusion, which they supposed applied to all serpentine soils, and which they considered “reverses all the teaching on this subject heretofore given, especially that as regards the toxicity of magnesium in serpentine soils”, belongs to the age before the complexity of such soils was fully appreciated. Moreover, their conclu-

274

JOHN PROCTOR

and

STANLEY R. J. WOODELL

sions were based largely on analysis of distilled water extracts, an experimental procedure which Rune (1953) among others has regarded as inadequate. Nevertheless, other workers have agreed that many of the serpentine soils investigated by them contain low levels of at least one of the nutrients nitrogen, phosphorus and potassium, e.g. Minguzzi and Vergnano (1953); Walker (1954); Spence and Millar (1963); Sarosiek (1964); Griffin (1965); Rai et al. (1970); Proctor and Woodell (1971); Ferreira and Wormell (1971). Krause (1958) presented rock analyses which show that on average serpentine rocks contain less phosphorus and potassium than granites, gabbros and sandstones. Plant analyses have usually confirmed soil analyses in demonstrating a low major nutrient content. NBmec (1951a) showed that the leaves of two conifers on serpentine contained only about half the quantity of phosphorus and potassium as on non-serpentine. Birrell and Wright (1945)reported from New Caledonia that Araucaria mulleri growing on a serpentine soil took up only about one-sixth as much phosphorus as Araucaria cookii growing in a greywacke soil. Duvigneaud (1966) commented on the low major nutrient content of plants growing in serpentine soils. Sarosiek (1964) presented data for nine species which showed that the plants growing in serpentine soils had a lower phosphorus and potassium content in the leaves than did the same species growing in non-serpentine soils. The leaf nitrate content was higher in the serpentine plants, however, despite a lower total soil nitrogen content. White (1971) analysed approximately one hundred nonserpentine and serpentine plant samples for total nitrogen content, and with a few exceptions the non-serpentine samples were higher in nitrogen than those from serpentine. Serpentine soils are not always low in major nutrients. Peligek (1939) reported normal levels of major nutrients in a serpentine rendzina from Moravia. Ishimoto (1958) compared the chemical composition of some serpentine and non-serpentine soils in central California, and found that the serpentine soils contained greater quantities of nitrogen, phosphorus and potassium. Proctor (1971a) showed that a soil on serpentine from Cornwall, Great Britain, is not particularly deficient in major nutrients. Experiments in which plants have been grown in fertilized serpentine soils have yielded variable results, sometimes supporting but frequently inconsistent with views on the overriding importance of low major nutrient levels. Many authors have reported no satisfactory improvement in crop plant growth in serpentine soils fertilized with NPK, e.g. Blackshaw (1921); Vlamis (1949); Johnson et al. (1952); Walker (1954); Soane and Saunder (1959); Proctor (19718). On other serpentine soils, however, a large increase in plant growth has been observed after

THE ECOLOGY OF SERPENTINE SOILS

275

fertilization with major nutrients, e.g. by Spence and Millar (1963) and Proctor (1971a). Ferreira and Wormell (1971) applied NPK to the sparse vegetation on unstable serpentine soil on an extremely exposed ridge on the Isle of Rhum, Scotland. They recorded a marked increase of cover of mature plants, and a change in the species composition, and concluded that at this site nutrient deficiency is a major factor limiting the development of closed plant communities. Proctor (1969) noted that where a farmer had fertilized a serpentine debris soil on Unst, Shetland, the native species showed greatly increased vigour. The differences in response of plants to fertilization of serpentine soils are one of the best demonstrations of their variability. Few workers would now subscribe to the view of Gordon and Lipman (1926)that low nutrient levels largely account for the peculiar characteristics of the vegetation on serpentine soils. Indeed, it has been pointed out a number of times that many other soils have levels of nutrients equally low or lower, and yet bear no “serpentine flora”. Robinson et al. (1935), referring to serpentine soils, noted “the calcium, potassium and phosphorus were generally low, but not lower than in many soils that are successfully farmed”. The possibility must be borne in mind of interactions between heavy metals and phosphates which might exacerbate the effects of a phosphate shortage. Soane and Saunder (1959) envisaged the possibility of interference with phosphorus absorption by chromium. They pointed out that Russell (1954) considered that the presence of iron and aluminium ions may cause a reduction in the ability of plant roots to translocate phosphorus from soil to leaves. Soane and Saunder postulated that chromic ions might have the same effect, and observed that phosphorus deficiency symptoms were a feature of plants suffering from chromium toxicity. Vergnano ( 1959a) investigated plants growing on serpentine soils in Italy and noted: “It seems particularly that the Cr contents of the plants are correlated with the phosphate contents, since with increasing Cr concentration a drop in the P concentration shown by these plants is frequently observed.” Jeffrey (1971) concluded that the rarity Kobresia simpliciwcula in a non-serpentine site in Upper Teesdale, England, depended for its existence on low available soil phosphate. On adding phosphorus the cover of Kobresia diminished, and it was out-competed by grasses showing a positive phosphate response. He considered that the high lead levels of this site might be a factor in reducing the availability of phosphate and hence favouring Kobresia, and mentioned that a serpentine rarity, Arenaria humifusa in Norway, may be similarly restricted to serpentines by a low phosphatelheavy metal interaction.

276

JOHN PROCTOR

and

STANLEY it. J . WOODELL

Hunter and Vergnano (1953) commented that an effect of excessive chromium in plant nutrient status was the low level of nitrogen in the plants, and the production of nitrogen deficiency. Such an interaction might increase the difficulties of plants growing in serpentine soils of low nitrogen status. The relationship of nutrient deficit and toxic ions has been largely unexplored, and needs further investigation. Differences between plants in their response to soil nutrient levels are an important consideration in explaining presence or absence of species in the serpentine vegetation. It seems likely that species requiring high phosphate, for example, will be totally excluded from most serpentine soils. Griffin (1965) in California demonstrated the ability of Pinus sabiniana to tolerate low nutrient levels in the soil, and his data suggest that non-serpentine plants of this species are preadapted to serpentine soils. Intraspecific races of plants adapted to growing at low nutrient levels are known for other soils (e.g. Bradshaw et al., 1964), and almost certainly exist on serpentines. Proctor (1971a) observed a relatively slow growth rate in fertile soils of races of plants collected from serpentine sites. Such might be regarded, at least in part, as an adaptation to growth at low nutrient levels.

c. N I C K E L ,

CHROMIUM A N D COBALT I N S E R P E N T I N E SOILS

Robinson et al. (1935) studied serpentine soils from Cuba, Puerto Rico and the U.S.A.,and sought to apply an all-embracing explanation of their infertility. They indicated the toxic nature of nickel, chromium and cobalt in culture solutions and pointed out that large total quantities of these elements were the only fa,ctors common to the different soils that they investigated. They concluded that “the presence of comparatively large quantities of chromium and nickel, and perhaps cobalt, are the dominant causes of infertility in serpentine soils in whichthe physical conditions are favourable for plant growth”. This conclusion was hardly justified from their data. They admitted that “in 11 of the 15 soils examined Loew’s theory that an excess of magnesia over lime is responsible for the toxicity to plants may be taken as a satisfactory explanation of the observed toxicity”, and of the 4 remaining “the quantities of both lime and magnesia are so low that they may be limiting factors of plant growth”. Moreover, the quantities of heavy metals in the plants and extractable from the soils they analysed are low compared with results obtained by many later workers (see Tables I11 and VI). However, their paper remains a landmark in the study of serpentine soils since it directed attention towards the possible importance of heavy metals, although it was not the first suggestion that

THE ECOLOGY O F SERPENTINE SOILS

277

heavy metals were important. West (1912) had, for example, mentioned chromium as a possible toxic element on serpentine. We now know that the situation is complex. I n some soils these metals appear to exert no effect, in others they may be partially toxic to some plants, whilst in others nickel at least is probably very toxic to many plants. There are a number of reasons why heavy metal toxicity is so variable as an ecological factor in serpentine soils. ( 1) Although serpentine rocks frequently have a relatively high nickel, chromium and cobalt content, this is not always so, as discussed earlier, since heavy metal content depends on geological origin and weathering. (2) An unknown proportion of the total quantity of heavy metal will be available to plants. This proportion varies from soil to soil and again within a soil depending, for example, on changes in pH and redox potential. (3) As Ernst (1972) has pointed out, not only the amount of heavy metal available to the plant but also the chemical status of the available heavy metal must be considered. He cited evidence (Ernst, 1968) that organic metal compounds are taken up less than inorganic ones. (4) The likely toxicity of metals is greatly modified by the presence of other elements. (5) There is a great variability of interspecific and intraspecific response of plants to the metals. The literature on heavy metals in these soils has now accumulated to such an extent that it is necessary to consider nickel, chromium and cobalt separately.

1. Nickel (a) Soil Nickel. Non-serpentine soils normally have 6-500 pprn total nickel (Swaine, 1955). Vinogradov (1938) gave 30 ppm, and Maljuga (1950) 40 ppm as the average nickel content of soils. I n contrast to this, serpentine soils frequently have a much higher nickel content, of the order of hundreds to several thousand pprn (Table 111). There is little work on the chemistry of nickel in serpentine soils, although there are relevant data for other soils. McLean (1966) investigated in detail the behaviour of nickel when added to a range of nonserpentine soils. He explained the reactions of nickel with the mineral part of the soil in terms of normal exchange (although some evidence for specific absorption was obtained), and precipitation by reaction with soil anions. From studies on the amount of nickel in solution, McLean considered that the most important “sinks” for nickel were possibly nickel phosphates (although there is no evidence of amelioration of K

t.3 4 00

TABLEI11 Nickel content of serpentine soils

Locality New Caledonia

Remarks (Measurements refer t o sample depth Nickel ppm in cm where this is known) total 0-13

“Available” nickel ppm extractant NH,COOCH, dilute acid

Author

1590

70

n.d.*

Birrell and Wright (1945)

1972 2239 218 20 1 1i.d. n.d.

n.d. n.d. n.d. n.d. n.d. n.d.

40.38 40.44 3-23 2.07 8.70 0.47

Fernandez et al. (1965)

Halstead (1968)

~~

Spain

Guatemala Rhodesia

Soil No. 186: 186: Soil No. 754: 754: Soil No. 187 Soil No. 726

0-10 10-60 2-25 25-55

0-10 10-20

11.d.

n.d.

259 383

n.d. n.d.

0-23 23-46

n.d. n.d.

9.4 5-6

31 23

cd d

Y

C

w

E

Hunter (1954) ~~

Scotland, Whitecairns, Aberdeenshire Scotland, Hill of Towanreef Green Hill Keen of Hamar Sweden, Kittelfjiill

Basin soils Hill slope soils

n.d. n.d.

26-61 22-49

114-289 49-403

0-15 0-15 0-15

n.d. n.d. n.d.

8 8 4

47 62

0-15

n.d.

2

24

56

Hunter and Vergnano (1952)

0 0

U

Proctor (1972) ~

~

_

_

Scotland, Rhum

Ard Mheall, A, horizon peat podzol B horizon C horizon

Cuba, Central Soledad

Soil No. 2523: 10-41

Eastern U.S.A., Dublin, Maryland Dublin, Maryland Belmont, Maryland Cherry Hill, Maryland Oxford, Pennsylvania Jermantown, Virginia Hunting Hill, Maryland Alberene, Virginia

Soil No. Soil No. Soil No. Soil No. Soil No. Soil No. Soil No. Soil No.

-

~

4722: 4725: 5829: 6178: 6236: 6241: 9456: 9889:

0-15 0-15 1-20 0-10 0-30 0-11 0-18 0-20

200 200 2000

n.d. n.d. n.d.

55 8 6

Ragg and Ball (1964)

25

1-21

n.d.

Robinson et al. (1935)

24 166 640 269 16 459 158 482

3.52 3.23 0.88 1-59 1.15 2.23 1.76 0.62

n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d.

Western U.S.A., Josephine Co., Oregon

Soil No. 9932: 5-25

trace

Poland, G6ra Radunia

Soil 1:

85 91 62 38 76 45 55 18 62 260

Wzg6rze nefrytowe

Soil3: Soil 4:

*

n.d. "not determined"

r

W M

-~

~

Soil 2:

0

0-5 5-18 0-5 5-15 15-27 5-12 12-22 0-5 5-27 27-35

10.91

n.d.

n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d.

3.9 5.1 2.7 1.5 3.6 1.7 2.2 1.4 3.0 88

M

z Sarosiek (1964)

z 3 W

i3 W

Is

TABLEI11 (continued)

Locality

Remarks (Measurements refer to sample depth Nickel ppm total in cm where this is known)

00

0

“Available” nickel pprn extractant NH,COOCH, dilute acid

Author 4

Portugal

Rhodesia

SoilEM36: SoilEM42:

0-25 0-15 15-30 30-70 Soil EM41: 0-18 18-30 25-50 50-80 30-80 Soil EM44: 0-5 5-20 20-40 Soil EM38: 0-13 13-45 45-75

Soil A Soil B Soil C Soil D Soil E Soil F Soil G Soil H Soil I

520 1100 1400 1400 n.d. n.d. n.d. n.d. n.d. 9000 25400 67700 1260 1340 940 4040t 1440t l060t

lolot

3620t 3520t 3900t 700t n.d.

n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d.

137 245

70 15 24 11 40 39 23 8 4

n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d.

de Sequeire (1969)

0

W Z

2

159 70 38 45 16 19 27 33 58 11 32 23 24

0 Q

H

Q 5

G

?a

F

M

cc 5d

4 Some and Saunder (1969) 0 U M

F F

Scotland, Keen of Hamar, Unst, Shetland

Mineral soil from pioneer phase Soil with recognizable humus fraction Ochreous associated debris Unaasociated debris Rooting layer under heath vegetation

Scotland, Keen of Hamar, Unst, Shetland Scotland, Green Hill, Aberdeenshire

n.d.

n.d.

312

n.d.

n.d.

289

n.d.

n.d.

220

n.d. n.d.

n.d. n.d.

90 90

Spence (1957)

m M 0

0 F

n.d.

n.d.

329

Spence and Millar (1963)

0

r 3288 2007

37.0 17.7

206.6 95.5

Taasoulas (1970)

Australia

n.d.

n.d.

184

Williams (1967)

Rhodesia, Nor0 Chrome Mine Tipperary Claims (a) Tipperary Claims (b)

5500 9375 5850

n.d. n.d. n.d.

175 273 92.5

Wild (1974b)

n.d. n.d. n.d. n.d.

55.2 12-4 40.8 36.8

320 2 125 15

Wiltshire (1972)

Rhodesia

t

30

Leslie series Green Hill series

Soil W2 Soil W3 Soil W4 Soil W5

E M

z

8

M

B

m

extracted with 21% hydrochloric acid t.3 OD c1

282

JOHN PROCTOR

and

STANLEY R. J. WOODELL

nickel toxicity on adding phosphate to a soil), nickel silicates, and perhaps double hydroxides with other elements, e.g. aluminium. Nickel hydroxide and nickel carbonate he regarded as being relatively soluble and unlikely to be of importance. I n serpentine soils some nickel is undoubtedly in exchangeable form, but which (if any) of McLean’s “sinks” account for the bulk of inorganically bound nickel in serpentine soils is unknown. A variety of means of estimating “plant available” nickel has been employed. Many workers have used total levels of nickel as an estimate of this quantity, and sometimes this may be adequate. For example Timperley et al. (1970, 1972a, b) showed that the nickel content of the ashed leaves of Nothofagus fusca and N . menziesii correlated well with the total nickel content of the soil. Other workers, with much theoretical justification, have preferred to estimate a nickel fraction soluble in dilute acid or exchangeable with ammonium ions (Table 111). Very few workers have attempted an objective assessment of the best method of estimating “plant available’’ nickel, Soane and Saunder (1959) investigated a number of extractants in Rhodesian serpentine soils. They used neutral normal ammonium acetate, 21% HCl, dithizone, and a cation exchange resin (zeocarb 215). They related plant uptake (measured as the nickel content of dried leaves) to the various extracted fractions. The best correlations were found for the ammonium acetate extractable fractions. White (1971) found a positive correlation between exchangeable nickel removed by normal ammonium chloride extractant and “total nickel” removed by 25% nitric acid. There is evidence that dilute acid extracts might lead to serious overestimation. Spence and Millar (1963) found no toxic symptoms in oats on a Scottish serpentine soil containing 329 pprn dilute acetic acid extractable nickel. Yet Hunter (1954) reported nickel toxicity symptoms in oats grown in a Rhodesian soil containing less than 10 ppm ammonium acetate extractable nickel. Proctor ( 1972) considered that “plant available” nickel is likely t o be difficult to assess, because he obtained evidence that the more labile forms of nickel are in different forms in different serpentine soils. He extracted one Swedish and three Scottish serpentine soils with dilute acetic acid and ammonium acetate solution. Theacetic acid removed much more nickel than the ammonium acetate in all the soils. For example, a soil from Kittelfjall, Sweden, had only 2 ppm ammonium acetate exchangeable nickel but 55 ppm soluble in acetic acid. A soil from the Hill of Towanreef in Scotland, on the other hand, had 7.5 pprn ammonium acetate exchangeable nickel and only 24 ppm soluble in acetic acid. The difference between the two extractants in this respect was much greater for nickel than for calcium, magnesium, sodium and potassium.

THE ECOLOGY O F SERPENTINE SOILS

283

There are likely to be important reactions between nickel and soil organic matter. McLean (1966) considered that in the non-serpentine soils with which he worked such reactions were complex and involved a variety of reactions of nickel with coordinating and chelating functional groups. Crooke (1956) investigated an unusually peaty serpentine soil and his results indicated that a large part of the nickel existed as a complex with soil organic matter. Surprisingly, the nickel in this complex appeared to be freely available to plants even though it was not extractable by acetic acid or ammonium acetate. Halstead (1968) and Halstead et al. (1969) drew attention to the effectiveness of organic matter in ameliorating nickel toxicity in a serpentine soil, and regarded organic matter as being able to complex with nickel and render it unavailable to the plant. It seems possible that the degree of nickel toxicity in a serpentine soil might depend on the organic matter content. We would welcome some attempts to measure the concentration of nickel in the soil solution of serpentine soils. This would be a direct way of assessing a possible toxicity since such measurements could be related to evidence from water and sand culture. Of course, the possibility of contact exchange of nickel to roots from soil surfaces without a solution phase must also be borne in mind. There is much evidence showing that nickel is toxic when added in quantities of a few ppm or less to nutrient solutions, e.g. Haselhoff (1893); Cotton (1930); Scharrer and Schropp (1933); Brenchley (1938); Hunter and Vergnano (1952); Crooke and Inkson (1955). Most of this work has concerned acute nickel toxicity in crop plants, and has been particularly relevant where specific symptoms of nickel poisoning have subsequently been observed for these plants grown in serpentine soils (see page 296). However, there is a shortage of experiments carried out in culture solutions of approximately similar nutrient status to those of serpentine soils, about acute or chronic nickel poisoning of naturally occurring plants. It is such evidence which one would like to see combined with estimates of nickel in soil solutions to give a better assessment of nickel toxicity.

( b ) Plant response to nickel QUANTITIES IN PLANTS: Vanselow (1966b) assembled data, mainly for crop plants, which showed that on non-serpentine soils 0.1 to 6 ppm nickel in the dry matter is usual. Lounamaa (1956) analysed a large number of native plant species growing on non-serpentine soils in Finland and found an ashed tissue content of about 20 to 70 ppm. On serpentine soils, plants usually contain much higher quantities of nickel. For example, Lounamaa found average values of about 600 to

284

JOHN PROCTOR

and

STANLEY R. J. WOODELL

over 10 000 ppm in plant ash. Data showing the range of nickel content in plants growing on serpentine soils are given in Tables IV and V. There are marked differences between species in their nickel uptake, e.g. the data of Hunter and Vergnano (1952) in Table V and that of Lyon et al. (1971)and Wild (1970, 1974b)in Table IV. Lyon et al. found that Notothlaspi australe from a soil with 3900 pprn total nickel had 640 ppm in its ashed foliage, whilst Pimelea suteri from a soil with 3340 pprn total had 5860 pprn in its ashed foliage. Wild observed that the nickel indicator plant Dicoma niccolifera ( = D . macrocephala Auct.) had a dry matter content of 1401 ppm when growing in a soil with an average content of 7375 ppm. On the other hand Securidaca longepedunculata had a content of only 29 ppm when growing in soils with an average content of 5464 ppm. Although there is a lack of critical soil data concerning “available” nickel at the exact site of the plant collection, it does seem that such differences in uptake genuinely reflect a very different response between species to nickel. Wild’s (1974b) data in Table I V axe particularly valuable since they include analyses of roots and hence show differences in nickel distribution patterns within the whole plant. Barleria aromatica and Convolvulus ocellatus var. plicinervius had more nickel in their roots, whilst Andropogon gayanus and Barleria kirkii had similar amounts in stems, leaves and roots. Pearsonia metallifera had more nickel in its leaves and Indigofera setiJlora more in its stems. Wiltshire (1972) found that for crop plants and weeds grown in serpentine soils, the concentration of nickel in roots always exceeded that of shoots. His data for Eragrostis viscosa are included in Table V. Tables IV and V show that species differ greatly in their capacity for exclusion, concentration and withinplant distribution of nickel. Scharrer and Schropp (1933) showed differences in nickel tolerance between species of cultivated plants grown in sand culture. Of the plants they tested, peas ( P i s u m sativum) were the most resistant t o nickel toxicity. With four cereals investigated the following order of resistance to nickel poisoning was observed: oats (Avena sativa), rye (Secale cereale), wheat ( Triticuin sp.), barley (Hordeum vulgare). Large differences between species of cultivated plants in response to nickeliferous serpentine soils have been observed by Hunter and Vergnano (1952) and Mizuno (1968). The implications of this work are that naturally occurring plant species also are likely to have differences in nickel tolerance and hence capacity for growth in a nickel-rich serpentine environment. Ernst (1972) has carried out the most intensive investigation into heavy metal tolerance in serpentine plants. He has suggested that NICKEL TOLERANCE IN PLANTS:

TABLE IV Nickel content of native plant% in serpentine soila ~~

~~

~

~

Plant nickel ppm Species

Plant Pa*

ash

matter

Pancheria glabroea

leaves

n.d.*

86

Araucaria mulkri

leaves

n.d.

20

1590 total, 70 PPm ammonium acetate extractable aa above

Myosotis monroi Notothlaspi auatmle Pimelea suteri Leptospemnurn awparium

foliage foliage foliage foliage foliage foliage foliage foliage foliage foliage foliage foliage

860 640 6860 2550 1730 430 1425 1330 225 1700 1880 440

n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d.

total 2570 3900 3340 3300 1930 1160 2900 2050 1200 2900 2030 1200

stems

970

290

total 2600

flowers

230

42

2600

(4 (b)

(4

Cassinia vauvillieraii (a) (b)

(4 Hebe odora

(8)

(b)

(4 Cunhurea panic-

-

dry

Soil nickel PPm

Locality

Author

New Caledonie

Birrell and Wright (1945)

New Zealand

Lyon et al. (1971)

!

0 4

ca

M

8

3

3

Italy

Minguzzi and Vergnano (1953);Vergnano (1958)

Is 00

01

TABLEIV (continued) *.\ L"

Species

Alyssum bertolonii

Euphorbia nicaeensis

Helichlrjsum italicum

Plant Pa* stems, leaves flowers, fruits stems, leaves flowers, fruits stems stems, leaves flowers fruits

Plant nickel ppm dry ash matter

00

Soil nickel PPm

14 400

2264

2600

55200

20405

2600

160

26

2600

510

84

2600

2100 690

484 197

2600 2600

440 860

88 142

2600 2600

Q,

Locality

Author

total

Pinus sylvestris

leaves

0-2108

0-9.2

439

Abies alba

leaves

360-440

1P15

n.d.

stem wood stem bark leaves wood

230-610

0.8-2.3

n.d.

150-480

2.2-6.6

n.d.

310-360 97011 000

54-6.4 3.2-22

n.d. n.d.

60707550

195

260-2600 ppm

Pinus sylvestris

Czechoslovakia, NBmEice (a) Czechosovakia, NBmFice (b)

NBmec (1951a)

Czechoslovakia, ChlumeEek

NBmec (1954)

-~ -

total

Pinus sylvestris

leaves

stem bark stem wood leaves

Betula sp.

stem wood stem bark leaves stem bark stem wood leaves bark

Salix caprea

Sorbwr aucuparia

wood ~

390-750

n.d.

1901120 22 03023 210 21703030 6801170 2670 9001480 11301600 700 1101010 8401600

n.d.

~~~

Dianthw carthusianorum

n.d. n.d. 200 n.d.

260-2600 ppm

n.d. 66 n.d.

260-2600 pprn

11.d.

leaves

4

n.d.

n.d.

leaves

5

n.d.

n.d.

young leaves Yowl3 leaves young leaves

8.6

n.d.

4.7

n.d.

85 total, 3.9 dilute acid soluble aa above

25

n.d.

as above

~~

T.pulegioides

260-2600 ppm

~

Blackjack oak (Quercwr marilandica) Red oak (Quercw rubra) Thymus serpyllum

895

Eastern U.S.A., Maryland

Robinson et al. (1935)

Poland, G6ra Radunia

Sarosiek (1964)

~

TABLEIV (continued) ~~

Species

Veronica spicata L i n a k &garis Silene injlatu

Qalium verum Lathyrua hterophyllua V k i a cracca Alyssum serpyllifolium ssp. lwritanicum

Plant part

Plant nickel ppm dry ash matter

Soil nickel PPm

Locality

Author

n.d.

as above

siz

22.0

n.d.

as above

Cd

45.0

n.d.

aa above

1.5

n.d.

as above

0.76

n.d.

aa above

0.97

n.d.

as above

leaves

n.d.

5160

stem

n.d.

3210

230000

10000

245 acetic acid soluble

c1

:: H

El

a

Portugal

de Sequeira (1969)

leaves

670

leaves leaves leaves leaves

n.d. n.d. n.d. n.d.

1401 26 1 380 325

td

3 Western Australia

Severne and Brooks

Rhodesia

Wild (1970)

(1972)

total

D i c m macrocephala Becium obovatum Vellozia equinetoidea Blepharis baineaii

sb3 ! I-'

total

Hybanthwr jloribundua

4

5.0

young leaves young leaves young leaves young leaves Young leaves Young leaves

7373 3563 5927 6250

0

tr

M

F

Becium homblei Indigofera setima The& triandra LowEetia simplex Hyparrhnia sp. Securidada longepedunculda

leaves leaves leaves leaves leaves leaves

Andropogon gayanua

tops roots

Barleria aromatiea: Convolvulua ocellatua var. plicinerviue

Dicoma nhli,fera Pearsonia nZetdi,fera Sutera fodina Combreturn m U e

(a)

n.d. n.d. n.d. n.d. n.d. n.d.

171 101 54 56 14 29

610 620

n.d. n.d.

leaves stems roots leaves

1930 2800 10 000 3100

n.d. n.d. n.d. n.d.

stems roots leaves stems leaves StemS roots leaves st0m

2700 10 300 3300 2800 153 000 146 000 102 600 5200 2900 2600 710

n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d.

370

n.d.

roots leaves, twigs

5622 5750 4338 5600 1650 5464

5500 total, 175 dilute acetic acid soluble as above as above

aa above

n.d.

as above aa above 9375 total, 273 dilute acetic

acid soluble roots

Rhodesia, Wild (1974b) Noro Chrome Mine

Tipperary Claims

E3 (0

0

TABLEI V (continued) Plant Part

Combretum molle

(b)

leaves, twigs

Plant nickel ppm dry ash matter 650

Soil nickel PPm

n.d.

Locality

Author

5850 total, 92.5 dilute acetic

Dicomu niccolifem Loudetia simplex

tops

370 7600

x

Z

cd

Tt

acid soluble roots stems, leaves

0

0

0

n.d. n.d.

Y

aa above 720

n.d.

3500 4400 7050 6400

n.d. n.d. n.d. n.d.

9375 total, 273 dilute acetic

acid soluble

Indigofera setiflora

roots leaves stems roots

~-

as above ..-

8

total

Clethra barbinervis Quercw crispula Deutzia crenata Caatanea crenaia Cynoxyylon japonica Lindera umbellata

leaves leaves leaves leaves leaves leaves

140 42-2 24.0 31.8 24.6 26.7

n.d. n.d. n.d. n.d. n.d. n.d.

1125 1005 1005 1005 1006 1006

Japan

Yamagata and Murakami

8

(1958)

b M F

r

TABLEV Nickel content of crop plants grown on serpentine soils

Species

Plant part

Oat (Awena sativa)

Plant nickel ppm dry ash matter

Soil nickel PPm

890

n.d.*

Barley (Hordeumvulgare) Wheat (Triticum sp.)

Rye Grass (Lolium perenne) Clover (Trifolium sp.) Turnip, yellow (Brassica rapa) Turnip, swede (Brasahnaps)

*

n.d. “not determined”

fully expanded leaves straw grain straw grain fully expanded leaves straw grain fully expanded leaves whole plant minus roots leaves flowers and peduncles fully expanded leaf laminae fully expanded leaf laminae

n.d.

42

n.d. n.d. n.d. n.d. n.d.

26 47 4 7 4

n.d. n.d. n.d.

3 17 48

n.d.

39

n.d. n.d.

58 26

as above as above

n.d.

99

above

n.d.

73

as above

Locality

259, ammonium acetate exchangeable

Guatemala

114-289 dilute acid soluble and 26-61 ammonium acetate exchangeable as above

Scotland, Whitecairns, Aberdeenshire

Author Halstead (1968)

I3

E

___

-

Oat (Avena sativa)

_ _

.*

Hunter and Vergnano (1952)

0

s 0

4

0

r as above

as above

z w

ul

s F,

N

CD

TABLEV (continued)

Species

Plant part

Plant nickel ppm dry ash matter

t s

Soil nickel PPm

Locality

Author LI

-

Potato

tubers

n.d.

7

as above

fully expanded leaf laminae

n.d.

90

as above

S00dS

n.d.

59

leaves and shoots

n.d.

10-20

leaves and

n.d.

(Solanurntuberoaum)

Bean

0

r? cd c1 c3 0

td

Oats (Avena aativa) Oats (Avena d i v a )

4 sites in

Proctor (19714

340

70ammonium

Rhodesia

acetate exchangeable

leaves

88 33

37.0 17.7

Soane and Saunder (1959)

Scotland

Tassoulas (1970)

(Eragrostia vkcoaa)

Oats (Avena d i v a )

t

Weed species

shoot root shoot root

n.d. n.d. n.d. n.d.

1.3 3.4 7.4 37.8

n.d.

91

v1

y

P

? ! F

ammonium acetate exchangeable

t

P,

I 3 G

Great Britain

SteUlS

Rye Grass ( Lolium perenne)

1000-3000 total

4

13.2 ammonium acetate exchangeable 55.2 ammonium acetate exchangeable

Rhodesia

184 dilute acid soluble

Australia

Wiltshire (1972)

3 0

U

u

F Williams (1967)

THE ECOLOGY OF SERPENTINE SOILS

293

some species might tolerate high nickel by concentrating it in their leaves and discarding them. He also investigated the solubility of metals in different fractions of the root, and his data support his hypothesis that roots resist toxicity by rendering a fraction of the metal insoluble and by rapidly translocating the remainder. There is not much evidence for the existence of serpentine races of wide-ranging species specifically adapted for nickel tolerance. Proctor (1971b), using a simple rooting technique developed from Wilkins (1957), demonstrated that Agrostis spp. collected from serpentine sites nearly all showed some degree of nickel tolerance. However, there is lack of supporting evidence that nickel is important at these sites, and also there are doubts about the specificity of nickel tolerance. For example, nickel-tolerant Agrostis plants have been collected from the Keen of Hamar, Unst, Shetland (Proctor, 1971b). Yet Spence and Millar (1963) have grown oats, and Proctor (1971a) oats and nontolerant Agrostis stolonifera, in the Keen of Hamar soil with no evidence of any nickel toxicity. Proctor (1972) grew oats (Awena satiwa) and beet (Beta vulgaris) (allegedly very sensitive to nickel toxicity) in soils from the Keen of Hamar and Green Hill, Scotland (another site from which nickel-tolerant clones have been collected), and again found no evidence of acute nickel toxicity. Moreover, of the metal tolerances investigated by Jowett (1958) and Gregory and Bradshaw (1965), i.e. to copper, lead, zinc and nickel, nickel tolerance seems to be the least specific. For example, Jowett demonstrated that Agrostis from a nickel mine soil in the Black Forest (1150 ppm total nickel) was highly nickel-tolerant. However, Gregory and Bradshaw (1965)showed that plants from soils at Trelogan (with only 13 ppm nickel) are almost equally nickel-tolerant. The best evidence of a specifically nickel-adapted serpentine race is that provided by Ernst (1972). He used the method of comparative protoplasmatology and tested the plasmatic resistance of various populations of Indigofera setifera against graduated solutions of nickel nitrate. He showed that the population of I . setifera from the nickeliferous serpentine at Tipperary Claims is highly resistant to nickel, but not to any other heavy metals tested (zinc and copper). This result is backed by other evidence that this site is highly nickel-toxic. Even so, other data from Ernst suggest some non-specificity for nickel tolerance, since a non-serpentine population of I . setifera from a site rich in lead but low in nickel showed a considerable nickel tolerance. Although the work of Nielsen and Sauberlich (1970) on chicks has indicated that nickel may be essential for animals, only a little work has been done to establish whether nickel is essential, or at least has some physiological function, in plants. THE ROLE OF NICKEL IN PLANTS:

294

JOHN PROCTOR

and

STANLEY R. J. WOODELL

Minguzzi and Vergnano (1948) envisaged that nickel counteracts high magnesium levels within the nickel accumulator, Alyssum bertolonii, a role that would normally be fulfilled by calcium. However, there is still no unequivocal evidence that nickel is an essential element for this species (Vergnano, pers. comm.). Severne and Brooks (1972) gave evidence that nickel might be an essential nutrient for Hybanthus jloribundus. They pointed out that the 23% nickel in the leaf ash of this species “represent the highest values recorded for any element in the ash of land plants”. Their evidence that nickel is carrying out a physiological function in this plant is based on the pattern of accumulation at different external nickel concentrations. Timperley et al. (1970) had shown that graphs of relative accumulation against the concentration of an element in the substrate are usually hyperbolic for essential elements and linear for non-essential elements. Severne and Brooks demonstrated that a graph of the relative accumulation of nickel in the leaf ash of H . jloribundus is a hyperbolic function of the concentration of the element in the soil, and they indicated that a limiting value of about 10% in the leaf ash corresponded to the plant’s requirement for nickel. However, they admitted that this does not prove that nickel is essential to the plant, and they also showed that samples of H . Jloribundus from another region contained nickel in much smaller quantities. We look forward to further work on this intriguing plant.

(c) Some factors which injluence nickel toxicity pH: A number of workers have demonstrated that crop plants growing on serpentine soils show a reduction in nickel uptake and nickel toxicity symptoms on the addition of calcium carbonate or calcium hydroxide (Hunter and Vergnano, 1952; Crooke, 1956; Halstead, 1968; Halstead et al., 1969). This reduction in uptake was due mainly to pH increases, since calcium sulphate was ineffective (Hunter and Vergnano, 1952; Halstead, 1968) whereas sodium carbonate was shown to be as effective as calcium carbonate (Crooke, 1956). There is no doubt that nickel solubility increases with decreasing soil pH, and nickel toxicity can be expected to be more likely in more acid serpentine soils. The only British serpentine site where nickel has been shown to be acutely toxic (Hunter and Vergnano, 1952) is very acid (pH 4.5) for a serpentine. However, a t least one serpentine is so nickeliferous that oat plants have been found to suffer from nickel toxicity and to take up considerable quantities of nickel, even when the soil was raised to pH 8.2 by the addition of calcium carbonate (Soane and Saunder, 1959). CALCIUM: Calcium status per se influences nickel toxicity. Crooke and Inkson (1955) found in water culture maintained at pH 5.5 that calcium

THE ECOLOGY O F SERPENTINE SOILS

295

was effective in reducing nickel toxicity necrosis. Jowett (1958), Gregory and Bradshaw (1965) and Proctor (1971b) used 1 or 0.5 g/litre Ca(NO,), in their nickel tolerance tests t o reduce the extreme toxic effects of nickel. The influence of calcium on nickel toxicity is not surprising in view of the ameliorative effects of calcium for toxic ions in general (Wyn Jones and Lunt, 1967). It must be expected that acute nickel toxicity will be more likely in serpentine soils of lower calcium status. IRON, MANGANESE AND COPPER: Some workers have suggested that nickel toxicity involves an induced iron deficiency, and considered the plant Fe/Ni ratio as important (e.g. Crooke et al., 1954; Mizuno, 1968). Williams (1967) reported an intensification of nickel toxicity symptoms in oat plants grown in an Australian serpentine soil, caused by high levels of manganese. These plants had a normal iron status, and he envisaged some interaction between manganese and iron within the plant which effectively decreased the Fe/Ni ratio. The implication of this work involving iron and manganese is that nickel in serpentine soils with much soluble iron may be less toxic, and in soils with much soluble manganese more toxic, than otherwise. There is a lack of relevant ecological data on this, however. Hunter and Vergnano (1953) and Mizuno (1968) have provided evidence that nickel toxicity also involves an interference in copper metabolism. Again the ecological relevance of this is obscure, although high amounts of available copper have been suggested by de Sequeira (1969) to account for the non-serpentine character of two serpentine soils in Portugal. He not only referred to the evidence that nickel and copper appear physiologically antagonistic but also pointed out the antagonism between these two metals in organic matter complexes. The higher stability of copper complexes favours the leaching of nickel, and possibly accounted for the low level of nickel found in these soils.

Crooke and Inkson (1955) found that an increase in the rate of supply of nitrogen (as nitrate), potassium and magnesium decreased the toxic effects of nickel on oats grown in sand culture. Hunter and Vergnano (1952) reported that nickel toxicity is greater when soil magnesium and nitrogen are low. de Kock (1956), Crooke and Inkson (1955) and Crooke et al. (1954) observed an association between nickel-induced chlorosis and high phosphorus levels in sand-water culture, which they interpreted as involving an interaction with iron in which leaf iron content was reduced by the high phosphorus. This was compounded with the nickel-induced iron deficiency discussed in the last section. Serpentine soils are usually low in nitrogen, phosphorus and potasMAJOR NUTRIENT ELEMENTS AND MAGNESIUM:

296

JOHN PROCTOR

and

STANLEY R. J. WOODELL

sium, and it seems unlikely that any significant amelioration or enhancement of nickel toxicity by these elements occurs in nature. On the other hand the evidence of an amelioration of nickel toxicity symptoms by magnesium is of particular significance in view of the high levels of magnesium frequently found in serpentine soils. There is evidence from work on microorganisms that the toxic effect of nickel on growth can be counteracted by supplementing the culture medium with magnesium (Abelson and Aldous, 1950; Adiga et al., 1961). The field of Ni/Mg interactions in serpentines would repay further study.

( d ) A bioassay for nickel toxicity. Nickel toxicity gives rise to specific symptoms, at least in oats (Avena sativa). The characteristic symptom is a leaf necrosis in which the necrotic tissue is present as white longitudinal stripes (Hunter and Vergnano, 1953). Nickel toxicity symptoms in oats growing in serpentine soils have been seen by Hunter and Vergnano (1952), Vergnano (1953) in Scotland, Hunter (1954), Soane and Saunder (1959), Wiltshire (1972) in Rhodesia, and Halstead (1968) in a Guatemalan soil. Williams (1967) found nickel toxicity symptoms in oats growing in a serpentine soil in New South Wales, Australia. He concluded, however, that the severity of the toxicity in this case was enhanced by the presence of large quantities of manganese. I n contrast, Spence and Millar (1963) and Proctor (1971a) have detected no evidence of nickel toxicity in oats grown in serpentine soils from Great Britain. Proctor (1971c, 1972) obtained similarly negative results with oats grown in a Californian and a Swedish serpentine soil. Hunter and Vergnano (1952, 1953) and Vergnano (1959b)also obtained negative results with some serpentine soils from the Upper Tiber Valley, Italy. ( e ) The ecological importance of nickel toxicity in serpentine soils. We have presented data which demonstrate the highly variable nickel content of these soils, the differing responses of plant species to nickel, and some of the factors which could influence nickel toxicity. There is good evidence for many soils and plants that nickel is an important factor, and equally good evidence in other cases that nickel is of little or no importance.

2. Chromium (a) Soil Chromium. Swaine (1955)states that total chromium in normal soils varies from 5 to 1000 ppm, whilst Vinogradov (1938) gives 100 ppm as the average chromium content of soils. Serpentine soils frequently have total chromium contents which are very much higher than these, up t o over 100 000 ppm (Table VII).

TABLE VI Chromium content of serpentine soils ~

Locality

New Caledonia

~~

~~~

Remarks (Measurements “Available” chromium ppm refer to sample depth Chromium ppm extractant in cm where this is known) total NH,COOCH, dilute acid

0-14

Rhodesia, Tipperary Claims Kingston Hill Czechoslovakia, N6mEice

Czechoslovakia, Bystha

*

n.d. “not determined”

Soil a. 10-20 20-30 30-40 40-50 Soil b. 10-20 20-30 30-40 40-50

Author

33 500

9

n.d.

Birrell a.ndWright (1945)

730 1130

30 13

n.d. n.d.

Ernst (1972)

3057 2914 2269 2462 299 884 1096 626

n.d.* n.d. n.d. n.d. n.d. n.d. n.d. n.d.

2-69 2-04 2.11 2-24 1.22 1-16 0-68

N6mec (1951a)

md

0 r(

0

0-75

2 8

51

9 ti

10-30 30-50 50-70 70-90 90-120

1816 2033 1884 1489 1034

n.d. n.d. n.d. n.d. n.d.

3.9

N6mec (1951b)

ra

4.0 3.7 3.2 2.9 f3

W

41

E3

TABLEVI (continued)

Locality

(0 Q)

Remarks (Measurements “Available” chromium ppm refer to sample depth Chromium ppm extractant in ern where this is known) total NH,COOCH, dilute acid

Author 4

Scotland, Rhum

0

Ard Mheall, A,, horizon peat podzol B horizon C horizon

31

1000 1000 1000

11.d.

n.d. n.d.

0.6 2.4

Ra.gg and Ball (1964)

1.1

8Y

~-

Puerto Rico, Mayaguez Mayaguez Mayaguez

Soil No. 9338: 0-20 Soil No. 9343: 0-20 Soil No. 9346: 0-15

4148 3604 4352

0.52 0.24 0.86

n.d. n.d.

Soil No. Soil No. Soil No. Soil No. Soil No. Soil No. Soil No. Soil No. Soil No.

1156 1632 2856 1088 2244 2312 1224 204 2176

2.0 0.34 0.68 1.0 0.68 0.34 0.34 0.34 0.34

n.d. n.d. n.d. n.d. n.d. n.d.

i4

Robinson et al. (1935)

11.d.

s

EP rn

Eastern U.S.A., Dublin, Maryland Dublin, Maryland Belmont, Maryland Cherry Hill, Maryland Oxford, Pennsylvania Jermantown, Virginia Hunting Hill, Maryland Potomac, Maryland Alberene, Virginia Western U.S.A., Josephine Co., Oregon

Soil No. 9932: 5 2 5

381

1.56

n.d.

Poland, G6ra Radunia

Soil 1:

640 740 530

n.d. n.d. n.d.

29 54 17

4722: 4725: 5829: 6178: 6236: 6241: 9456: 9460: 9889:

0-15 0-15 1-20 0-10 0-30 0-11 0-18 0-23 0-20

5

1

*

M

w

4

s0

11.d.

n.d. n.d.

0

G

M

Soil 2:

0-5 5-18 0-5

tt-

Sarosiek (1964)

Wzg6rze nefrytowe

5-15 15-27 5-12 12-22

Soil 3: Soil 4:

0-5

5-27 27-35 ~~~

Portugal

510 1500 330 510 390 660 1900

_____

Soil EM36, A, horizon Soil EM38, A,,, horizon AI2, horizon AC, horizon

1010 n.d. n.d. n.d.

Soil A Soil B soil c Soil D Soil E Soil F Soil I

39 000 20 000 8000 8500 46 000 41 000 550

n.d. n.d. n.d. n.d. n.d. n.d. n.d.

20

13 12 20 21 33 460 ~

~~~

~

-

n.d. n.d. n.d. n.d.

1.2 1.6 1-2 1.4

de Sequeira (1969)

M d C

r

~-

Rhodesia

O*Sf

0.57 o-9t 1.17 0.97 0.4f

n.d. n.d. n.d. n.d. n.d. n.d. n.d.

Soane and Saunder (1969)

$

* C

w m M

E M

z

2

Scotland, Keen of Hamar, Unst, Shetland Scotland, Green Hill

0.97

2

M

Z

ad.

n.d.

1.0

Spence and Millar (1963) M m

10138 6923

trace trace

1.18 0.88

Tassoulas (1970)

n.d.

n.d.

0.390.76

Vergnano (1953)

i

m Leslie series Green Hill series

Scotland, Whitecairns, Aberdeenshire

t

N

extracted with Zeocarb 215, a cation exchange resin

W W

300

JOHN PROCTOR

and

STANLEY R. J. WOODELL

The quantities of chromium in serpentine soils are even more variable than those of nickel, because unlike nickel, which is an integral component of olivine and other minerals in serpentines, chromium is in an accessory mineral, chromite, which is distributed very erratically (Faust and Fahey, 1962). Total chromium analyses are of limited value to the biologist since chromite, the chromium mineral likely to be present in the greatest quantities, is so very insoluble. Little information exists on the more labile forms of chromium in the soil, and even the preponderant oxidation state is unknown. Soane and Saunder (1959), comparing the chromium extracted with cation and anion exchange resins from serpentine soils of pH 6 (approximately), found that only the former absorbed appreciable quantities of chromium over a two-month incubation period. However, from other evidence they concluded that some of the chromium was present as chromate. pH is likely to have a controlling influence on the oxidation state, an increase resulting in a change from the chromic to the chromate (Goldschmidt, 1954). Breeze (1973), using data from Harrison (1972), has indicated that chromic ion concentration in the soil is likely to be governed by the very insoluble &(OH), which has a solubility product of 1.0 x 10-30. He pointed out that the solubility of the chromic ion decreases from 1660 ppm to 1.66 ppm between pH 4-5 and 5-5. Serpentine soils usually have a pH greater than 6.0 and in them chromic ions must have a very low concentration. On the other hand most chromate salts are soluble at a pH of around 7.0, and perhaps most plant available chromium in many serpentine soils is in anionic form. There is no generally accepted method of assessing “plant available” chromium in the soil, although many have used ammonium acetate or dilute acids as an extractant (Table VI). Wild (1946b) found very low (0.2 ppm) dilute acid extractable chromium in some Rhodesian soils which contained up to 12.5% total chromium. The plants growing on these soils, however, contained very high levels of chromium (Table VII). This observation raises many questions about “plant available’’ chromium, and as Wild suggests, some sort of biossay technique may be the best way of estimating this quantity. Data are lacking for chromium, as for nickel, that would enable us to compare soil solution concentrations of chromium with those known to be toxic in culture solutions. Chromium is a very toxic element in solution culture (Scharrer and Schropp, 1935; Hewitt, 1953; Soane and Saunder, 1959; Proctor, 1971b), although probably less toxic than nickel at the same concentration for most plants.

TABLEVII Chromium content of native plunts in serpentine soils Plant chromium ppm dry ash matter

Species

Plant part

Pancheria glabrosa A raucaria mulleri

leaves

n.d.*

4.5

leaves

n.d.

12.6

Soil chromium PPm 33 500 total, 9 exchangeable n.d.

Locality

Author

___

New Caledonia

Birrell and Wright ( 1945)

m

-

Di c o m niccolifera Indigofera setiflora

A%!yOSOtkmonroi Notothlaspi australe Pimelea suteri Leptoapermuni scoparium Caasinia vauvilliersii Hebe odora

leaf root cortex root wood leaves stem cortex stem xylem root xylem root cortex

n.d. n.d. n.d. n.d. n.d. n.d.

n.d. n.d.

240 100-3 24-3 0 274 0.1 26.2 49.3

730 total, 30 exchangeable

0 0

*

0 4 M

_

680 510

n.d. n.d.

foliage foliage

1800 2470 1760 1300 815 330 141

n.d. n.d. n.d. n.d. n.d. n.d. n.d.

12 450 8950 8750 1 1 600 9600 13 800 8500

foliage

Ernst (1972)

ifF

v)

foliage foliage

(a) (b) (a) (b) (a) (b)

Rhodesia

as above

total 6700 12 400

foliage

2

M

_

~

New Zealand

~

~

Lyon et al. (1971)

k2

2

3 s

v)

E:

TABLEVII (continued)

cu

Species

Plant part

Centaurea paniculata Alyssum bertolonii Euphorbia nicaeensis Helichrysum italicum

stems flowers stems, leaves flowers, fruits stems, leaves flowers, fruits stems stems, leaves flowers fruits

Pinus sylvestris

Abiee alba

Plant chromium ppm dry ash matter 75 17 190 1 1760 < 20 200 620 110 350

24 3 30 trace 284 <3 47 177 20 58

leaves bark

91.1-124.4 78.2-1 00.0

1*3-1.7 0'34-1.02

stem wood root wood root bark root hairs leaves

325.7-760-2 8.2 425.0 391.7 24-42

0.88-2.45 0-34 10.9 15-6 1.0-1.4

leaves bark root bark stem wood root wood

53.7-124-4 0-90.4 2883.2 13.1-50.3 55.7

Soil chromium PPm 5600 5600 5600 5600 5600 5600 5600 5600 5600 5600 3057 total 2.69 dilute acid extractable

Locality

Author

Italy

Minguzzi and Vergnano (1953); Vergnano (1958)

Czechoslovakia, NBm5ice (a)

N6mec (1951a)

M

4

299 total, 1.22 dilute acid extractable

NlimEice (b)

1816 total, 3.9 dilute acid extractable

Czechoslovakia, BystFina

~

1.02-2.52 0-1-7 122.4 0-07-0.48 0.14

r La

~~

Pinus sylcestrie

0 N

~~~

~~

NBmec (1951b)

Blackjack oak (Quercus rnarilandica) Red oak (Quercus rubra)

leaves

4

n.d.

n.d.

leaves

3

n.d.

n.d.

U.S.A., Maryland

~~

Thymus serpyllum T . pulegiodes

leaves leaves

0.83 0.64

0-49 (a) 9.0 (b) 6.3 (a) 0.19 (b) 0.24 (a) 7.5 (b) 8.7 (a) 14.0 (b) 8-7 (a) 0.82 (b) 0.51 (a) 0.33 (b) 0.40 (a) 0.27 (b) 0.16

n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d.

(a,) (b) (a) (b)

0.48

total 640 330 530 390 530 390 530 390 640 330 640 330 530 390 530 390 640 330

extractable 29 12 17 21 17 21 17 21 29 12 29 12 17 21 17 21 29 12

Dianthus carthusianorum Veronica spicata

leaves

Linaria vulgaris

leaves

Silene infrata

leaves

Galium veruni

leaves

Lathyrus heterophyllus Vicia crmca

leaves leaves

Andropogon gayanus

tops roots

690 2800

n.d. n.d.

Barleria aromatica

leaves stems roots leaves

3500 7500 9000 8000

n.d.

125 000 total 2, dilute acid soluble as above

n.d.

as above

Convolvulus ocdlatwr var. plicinervius

leaves

Robinson et al. (1935)

~

Poland, G6ra Radunia and Wzg6rze nefrytowe

Sarosiek (1964)

M W

Bg Rhodesia, Noro Chrome Mine

Wild (1974b)

0 0

W

0 0

TABLEVII (continued)

Ip

Plant chromium ppm

dry Species

Plant part

ash

matter

Soil chromium PPm

Locality

Author

4

0

C . ocellatus (continued) Dicoma nicwlifera Pearsonia metallif era

stems roots leaves stems leaves stems roots Sutera fodina leaves stems roots Combreturnmolle (a)leaves, twigs roots Combreturn m l l e (b)leaves, twigs roots

16 000 11 000 30 000

n.d.

as above

n.d.

as above

n.d.

as above

40

n.d. n.d.

60 140

n.d. n.d.

12 000

n.d.

15 500 total 2 dilute acid extractable 6750 total 2 dilute acid extractable as above 16 000 total 2 dilute acid

34 000

4800 4400 15 000 48 000 7000 14 000 20

Dicom niccolqera Loudetia simplex

stems, leaves tops roots

290 120

n.d. n.d.

Indigofera setiflwa

leaves stems roots

390 1400

n.d. n.d. n.d.

440

extractable as above

Rhodesia, Tipperary Claims

THE ECOLOGY O F SERPENTINE SOILS

305

( b ) Plant response to chromium QUANTITIES m PLANTS: Pratt (1966) collected data, mainly for cultivated plants, which showed that on non-serpentine soils a chromium content of less than 2 ppm dry matter in the above ground parts is usual. Lounamaa (1956) analysed a large number of native plant species growing on non-serpentine soils in Finland and found an ashed tissue content of about 4 to 120 ppm. On serpentine soils plants usually, but not always, contain much higher quantities of chromium (Tables VII and VIII). As for nickel there are large differences between species in their uptake of chromium. I n New Zealand Hebe odora on a soil containing 8500 ppm chromium had only 141 ppm in its ashed foliage, whilst Leptospermum scoparium on a soil containing 8750 ppm chromium had 1760 pprn in its foliage (Lyon et aZ., 1971). From a Rhodesian soil containing 125 000 ppm total chromium, Andropogon gayanus had a content of 690 ppm in its shoots, whilst Sutera fodina had 48 000 pprn in its leaves (Wild, 1974b). Wild’s data show differences in distribution patterns of chromium between root and above-ground parts. Some species have the highest chromium concentration in their roots, others in stems, and others in leaves. These recent findings are surprising, since earlier work with crop plants had suggested that there was a general tendency for more chromium to occur in roots than in the above-ground parts (Vanselow, 1951; Hunter and Vergnano, 1953; Soane and Saunder, 1959). ( c ) The role of chromium in plants. Although chromium is established as having a physiological role in animals (e.g. Schwarz and Mertz, 1961; Mertz, 1967),there is no evidence that it has any physiological function in plants. The extreme accumulator plants Dicoma niccozifera and Sutera fodina would merit investigation from this point of view. Lyon et al. (1969a, b) investigated the distribution of chromium-51 in the sap, tissues and extracts of the chromium accumulator Leptospermum scoparium. Plants cultured in solutions containing Na,61Cr0, accumulated most of the absorbed radioactivity in the roots. About one third of the root radioactivity was soluble in 80% ethanol in the form of three complexes, the predominant one being identified as trioxalatochromate (111)ion. These complexes were also present in stem and leaf extracts. They showed that only the chromate ion was found in the xylem sap, and it seems that hexavalent chromate is absorbed from the solution and transported in the xylem throughout the plant. Metabolism of chromate possibly takes place within the leaf to give trioxalatochromate (111)ion. As Lyon et al. point out, there is great scope for extending this isotopic work.

TABLEVIII Chromium content of crops plants grown on serpentine soils

Species

-

Oats (Aoena satira)

Maize (Zeu mays)

Oats (Avena sativa)

Tobacco (Nicotiana tabacum)

Plant part

Plant chromium ppm dry ash matter

Locality

leaves, shoots leaves, shoots leaves, shoots leaves, shoots

n.d.*

< 10

2000

n.d.

nil

n.d.

< 10

3000

n.d.

nil

n.d.

< 10

3000

n.d.

nil

n.d.

< 10

4000

n.d.

nil

leaf leaf leaf leaves, stems leaves, stems leaves, stems leaves, stems Stem, leaf

n.d. n.d. n.d. n.d.

4 17 4 7

39 000 20 000 8000 39 000

0.9 0.8 0.5 0.9

NPK NPK NPK nil

n.d.

3

39 000

0.9

NPK

n.d.

11

39 000

0.9

NPK+B

n.d.

3

39000

0.9

NPK

n.d.

440

550

0.4

NPK

1.18

various

Scotland, Green Hill

Tassoulas (1970)

various

Scotland, Whitecairns, Aberdeenshire

Vergnano (1953)

(Lolium perenne)

leaves

n.d.

trace

Oats (Aveaa sat iva)

leaves

n.d.

0.1-0.65

* n.d.

soil chromium ppm fertiliser total extractable treatment

“not determined”

10 138 n.d.

0.39-0.76

England, Cornwall Scotland, Rhum Scotland, Unst Scotland, Meikle Kilrannoch

Proctor (1971a)

Rhodesia

Some and Saunder (1959)

0

zz

P tl

Q

2 w ci

+ Ca

3

0

U M P P

THE ECOLOGY OF SERPENTINE SOILS

307

( d ) Chromium tolerance in plants. Soane and Saunder (1959) comment on the much greater sensitivity of tobacco to chromium toxicity (Cr20,--) compared with maize. Scharrer and Schropp (1935) found that oats (Avena sativa), rye (Secale cereale),wheat (Triticum sp.) and maize (Zea mays) were more resistant to high levels of Cr+++ than barley (Hordeum vulgare) and peas (Pisum sativum). Oats and wheat were most resistant to Cr0,--. There is no evidence of specifically adaptive chromium tolerance in serpentine plants. Some attempt to obtain it was made by Proctor (1969, 1971b) but the results were inconclusive.

( e ) Some factors which influence chromium toxicity pH AND REDOX POTENTIAL: The proportion of anionic to cationic chromium depends on pH and redox potential, as we have discussed, a higher pH and a lower redox potential favouring an increase in the proportion of the anion. There are data which suggest that anionic chromium is more toxic than chromic ions (Scharrer and Schropp, 1935; Hewitt, 19/53), and certainly anionic chromium is very much more soluble than cationic at pH greater than 6. However, at present we know too little about chromium in serpentine soils to assess the importance of the ionic state in natmal situations. CALCIUM: Chromic ion toxicity is alleviated by calcium ions (Proctor, 1969, 1971b) and is less likely in more calcareous soils. The importance of calcium in alleviating chromium toxicity might vary with species, as the analytical data of Lyon et al. (1971) indicate. Leptospermum scoparium had a very high chromium content and also a very high calcium content, and Lyon et at?. commented that the high calcium content may serve to alleviate toxicity associated with high chromium. Hebe odora had a low chromium content and a low calcium content, and appears to exclude both these elements. Pimelea suteri had a very high chromium content but a low calcium content. It appears that these species have different tolerance mechanisms, in each of which calcium plays a different role. IRON: For chromium toxicity as for nickel, there is evidence of interference with iron metabolism; for example, Hewitt (1953) observed that the chlorosis in sugar beet caused by chromium disappeared when the leaves were sprayed with ferrous sulphate. However, there is a shortage of data concerning possible chromium-iron interactions in serpentine soils. NICKEL: Hunter and Vergnano (1953) found that 2 ppm of chromium in the nutrient solution increased the degree of specific nickel symptoms and nickel uptake associated with 2.5 ppm nickel in the solution. In

308

JOHN PROCTOR

and

STANLEY R. J. WOODELL

view of the toxic levels of nickel in some serpentine soils this chromiumnickel interaction needs further investigation.

(f) The ecological importance of chromium in serpentine soils. The same considerations about variability in plants and soils apply to chromium as did to nickel. I n general, though, there is much less evidence that chromium is important in serpentine soils. Although specific chromium toxicity symptoms have been described for oats in culture media by Hunter and Vergnano (1953), they have not been convincingly shown for oats grown in serpentine soils. The only good example of chromium toxicity in a crop plant grown on a serpentine soil is for tobacco grown on a Rhodesian soil by Soane and Saunder (1959). The best evidence that chromium is of ecological importance is from plant analyses. The very high quantities accumulated by some plants and the exclusion of chromium by others suggests some positive response to the element. I n view of the toxicity of chromium to plants we must suspect that these observed differences in uptake are manifestations of different chromium tolerance mechanisms. The situation is more puzzling in view of the very small quantities of extractable chromium in soils, and as we have remarked, some chromium accumulator plants grow in soils with a very low extractable chromium content. Finally we should mention the observation of Wild (1974b),who noted that in Rhodesia there are specific floras associated with a number of toxic metals, e.g. nickel and copper, but that no plants are characteristic of chromium sites. Chromium toxicity is a particularly tantalizing aspect of the ecology of serpentine soils.

3. Cobalt ( a ) Soil cobalt. The total cobalt content of ordinary soils is usually in the range of 1-40 ppm (Swaine, 1955).Serpentine soils, however, frequently contain much greater quantities than this (Tables IX, X). What little information exists on “available” cobalt in serpentine soils is summarized in Table IX. As for nickel and chromium, there is inadequate soil data for extrapolations to be made from the results of solution culture experiments which have shown cobalt to be very toxic (e.g. Scharrer and Schropp, 1933; Brenchley, 1938; Millikan, 1949; Vergnano and Hunter, 1953).

( b ) Cobalt in plants QUANTITIES IN PLANTS: Vanselow

(1966a) showed that in crop plants on non-serpentine soils a cobalt content of less than 5 ppm in the dry

TABLE IX Cobalt content of crerpentine so&

Locality

Remarks (Measurements refer to sample depth in cm where this is known

Cobalt ppm total

“Available” cobalt ppm extractant NH,COOCH, dilute acid ~

~

Rhodesia

Kingston Hill

285

11

n.d.

A d Mheall, A, horizon

50 150 150

n.d. n.d. n.d.

0.9 1.4 1.7

Author ~

~

~~~

Ernst (1972)

Scotland, Rhum

E i

peat podzol B horizon C horizon Poland, G6ra Radunia

Soil 1:

Soil 3: Soil 4:

Scotland, Unst

Scotland, Whitecairns, Aberdeenshire

M

d

0

F

0

Soil 2: Wzg6rze nefrytowe

Ragg and Ball (1964)

0-5 6-18 0-5 5-15 15-27 5-12 12-22

42 39 33 18 37 22 26

0-5 5-27 27-35

10 31 72

n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d.

2.0 1.8 1.6 0.83 1.8 1.1 2.3

Sarosiek (1964)

2

s rA

1 til

4 B

0.53 1.5 4.0

B

rA

Ochreous associated debris Unassociated debris Rooting layer under heath vegetation

n.d. n.d. n.d.

n.d. n.d. n.d.

9.5 8.9 12.2

n.d.

n.d.

4.8111.8

Spence (1957)

Vergnano (1953)

0 0 (0

TABLEX

2

Cobalt contetat of d i v e plants in 8erpentin.e so&

0

Plant cobalt ppm Species

Myomtie monroi NototMl8pi australe Pimelea euteri Leptospemum ecopariuna CQeeinia VauviUiermi Hebe odwa Cmtuureu panicdata

Alyawm bertolonii EuphorbiO nieaeeneis

Helichrymm

italiculn

Plfmt Pa&

dry

Soil cobalt

ash

matter

PPm

foliage foliage

42 83

n.d.* n.d.

330 390

foliage foliage foliage foliage foliage foliage foliage

236 96 69

270 400 390 380 396 366 426

Locality New Zealand

Author Lyon et al. (1971) cd

(a)

72

(b)

63 39 30

n.d. n.d. n.d. n.d. n.d. n.d. n.d.

stems flowers stems, leaves flowers, fruit

620 35 10

166 6 1-5

n.d. n.d. n.d.

300

36

n.d.

4

Stems,

230

36

n.d.

6

leaves flowers, fruit stems stem, leaves flowers fruits

n.d. “not determined”

(a)

(b) (a)

(b)

Q

Y

8

s a Italy

Vergnano (1968)

E td

h

0

160

26

n.d.

120 40

28 12

n.d. n.d.

15 18

3

n.d. n.d.

3

U

Abiea &a Pinua sylveatrie

leaves stem wood stem bark leaves wood

800-840 2700-4100 1300-2800 740-1600 680-13 000

27-33 9-13 22-42 14-26 2.644

n.d. n.d. n.d. n.d. n.d.

Czechoslovakia, NBmEice

NBmec (19618)

Czechoslovakia, ChlumeEek

NBmec (1954)

~~

Pinua sylveat&

B e t h sp. Sdix crrprea

sorbus oucuparia

leaves 14480-32670 stem bark 6060-9570 stem wood 6550-15 100 le8VeS 18 160-45 420 stem bark 7900-20 800 stem wood 34000-38490 leaves 7060 stern bark 2290-18 190 stem wood 2540-3100 leave8 5500 stem bark 1780-2850 stem wood 4850-5760 leaves

29

1883

91-231 total

n.d. n.d.

aa above

2570

H

B

aa above

U

n.d. n.d.

Q

0

529

aa above

517

as above

n.d. n.d.

i*

0

m

i9$

n.d. n.d. n.d.

leaves leaves

24 690

n.d. n.d.

leaves leaves leaves leaves leaves leavet3

82 470 730 19 8.7 1-8

n.d. n.d. n.d. n.d. n.d. n.d.

42 total, 2.0

dilute acid soluble aa above 33 total, 1.6 dilute acid soluble aa above aa above aa above as above above aa above

Poland, G6ra Radunia

Sarosiek (1964)

22 5 l

3 8 lP x

TABLEX (continued) Plant cobalt ppm Species

Thymus aerpyllurn T . pulegoidea Dianthus mrthzrsianomm Linaria vulgaria Silene inflat0 Veronica @a& Galium vemm

Lathyrua hrophyllua Andropogm gayanus Barleria a r o d h cmv01vulua

ocektus

Dicoma niccolifera

Plant Pa*

dry ash

matter

leaves

44

n.d.

leaves leaves

42 570

n.d. n.d.

leaves leaves leaves

200 450 95

n.d. n.d. n.d.

leaves leaves

11

3.3

n.d. n.d.

shoots

30

n.d.

roots leaves stems roots leaves stems roots leaves stems

30 60 120 110 100 130 90 140 160

n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d.

Soil cobalt PPm 22 total, 1.1 dilute acid soluble aa above aa above

Locality

Author

4

Poland, Wzg6ne nefiytowe

aa above aa above 10 total, 0.53 dilute acid soluble as above aa above n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d.

Rhodesia, Noro Chrome Mines

Wild (1974b)

d ! 4

3

8

Pearsonia ntetallqera SuteTa fodina Combreturn molle (a) Codreturn molle (b) Dicomcl niccol(fera LowEetia simplex Indigof era setiJora

leaves Stems

roots leaves stems roots leaves, twigs roots leaves, twigs roots stems, leaves tops roots leaves stems roots

3300 1800 1400 220 110 70 50

n.d. n.d. n.d. n.d. n.d. n.d. n.d.

n.d. n.d. n.d. n.d. n.d. n.d. n.d.

50 30

n.d. n.d.

n.d. n.d.

Rhodesia, Tipperary Claims M

d

40 200

n.d. n.d.

30 80 40 100 105

n.d. n.d. n.d. n.d. n.d.

661

n.d.

0

tc 0 0

n.d. n.d.

4

0 4

n.d. n.d. n.d. n.d. n.d.

4

total

Clethra barbinenria Quercua c & p h Deutzia crenata Caatanea crenata CylaOxY~

japonica Lindera umbellata

leaves

511

Japan

Yamagata and Murakami (1958)

leaves leaves ieavt~ leaves

2.6 5-3 3.6 2.0

n.d. n.d. n.d. n.d.

652 652 652 652

leaves

4.3

n.d.

652

M ra

8rn

314

JOHN PROCTOR

and

STANLEY R. J. WOODELL

matter is usual. Lounamaa (1956) analysed a large number of native plant species growing on non-serpentine soils in Finland and found ashed tissue content between 1 and 25 ppm. Plants on serpentine soils usually contain higher quantities of cobalt (Table X). Table X shows differences between species in their uptake and distribution of cobalt. Wild’s (1974b) data shows that Pearsonia metallifera is a cobalt accumulator and along with Szctera fodina has most cobalt in the leaves, whereas Loudetia simplex has more cobalt in the roots. Surprisingly high quantities of cobalt have been found in stunted pine and birch by N6mec (1954).

( c ) Functions of cobalt in plants. Cobalt is well known as an essential element in animals as a constituent of vitamin B12.It is required for nitrogen fixation by root nodules and appears to be an essential element for at least one non-nodulated legume and for wheat (Wilson and Nicholas, 1967). Its possible physiological role in serpentine plants is unknown.

( d ) Tolerance to cobalt toxicity. Plant species have very different tolerances to cobalt. Scharrer and Schropp (1933) showed that in crop plants, maize was the most resistant to cobalt toxicity, then barley (Hordeum vulgare), wheat (Triticum sp.), rye (Secale cereale), and oats (Avenasativa), with peas (Pisumsativum) the most sensitive. Serpentine plants have not been tested for cobalt tolerance. ( e ) The ecological importance of cobalt in serpentine soils. Although cobalt is often implicated with nickel and chromium as being of possible importance, there is frequently little evidence for this. Plants on serpentine soils often do not contain very much cobalt. Of the sites investigated, that of NBmec (1954) seems most likely to be influenced by high cobalt levels, although even here, more evidence is desirable.

D.

CALCIUM A N D MAGNESIUM

The necessity of a balanced content of the elements calcium and magnesium for the success of plants was recognized by Loew (1892). I n the light of this work and of that that followed, especially Loew and May (1901), the Ca/Mg ratio was soon regarded as a prime factor in serpentine infertility. Thus in 1906 Hilgard, in a standard textbook, in reference to serpentine soils, said that their “excess of magnesia over lime is injurious to most crops”. Since then the Ca/Mg ratio hypothesis has had a chequered history. Some authors have disregarded it, others have emphasized the importance of the calcium, yet others high magnesium. It is now clear, however, that both calcium and magnesium should have major consideration in any account of “serpentine infertility”. There are serpentine

THE ECOLOGY OF SERPENTINE SOILS

316

soils in which these elements have not been shown to exert an unusual influence, but we believe these to be in the minority.

1. Soil calcium and magnesium Table X I shows that most serpentine soils are low in calcium, with exchangeable quantities commonly ranging between 0.2 and 4 mEq/ 100 g. A few have higher amounts, up to 15 mEq/100 g or even higher at the Lizard, Cornwall, England (Halliday, 1960; Proctor and Woodell, 1971) and 19 mEq/100 g in a Rhodesian soil (Wild, 1974a). Occasionally there is so little that it cannot be measured by usual methods of analysis (e.g. two of the samples from Rhum, Scotland (Proctor and Woodell, 1971) and some New Guinean soils (Stevens, pers. comm.)). The calcium content depends upon geological origins, as discussed earlier. Some serpentine soils do have highly calcareous veins (Ferreira, 1959). Exchangeable magnesium is in very small quantity in some soils, especially in the wet tropics, but in most it is relatively high and ranges between 3.0 and 30 mEq/lOO g. Usually magnesium is in excess of calcium, and the Ca/Mg ratio is usually below 0.4 when expressed on a milliequivalent basis (Table XI). This is in contrast to most non-serpentine soils where calcium is in excess of magnesium. Most authors have used exchangeable calcium and magnesium as a measure of the availability of these elements to plants. However, there are indications that this may be an oversimplification. At high magnesium and low calcium levels it is likely that calcium is preferentially held on specific exchange sites, and the Ca/Mg ratio in the soil solution may be much less than that indicated from exchangeable analyses. Tinker (in Proctor, 1971b) demonstrated such a situation in two Scottish serpentines, using the method of Beckett (1966). I n one soil the Mg/Ca ratio likely to exist in the soil solution was over five times as high as the exchangeable Mg/Ca ratio, in the other over twice as high. These may be extreme examples but it seems likely that exchangeable quantities may often give some underestimate of the excess of magnesium over calcium faced by the plant. Serpentine has been used in a pulverized form as a source of magnesium for agricultural soils in New Zealand, Canada and the Netherlands. The interest of this to the serpentine ecologist is in Burns and Smith’s (1965) suggestion that its low calcium and high magnesium content create a favourable Ca/Mg ratio in magnesium-deficient soils where calcium is in adequate supply.

2. Calcium and magnesium in plants I n many cases serpentine plants reflect the quantities of calcium and magnesium in their soils to the extent that they contain more magnesium than calcium. It seems clear, however, that they usually take up

316

JOHN PROCTOR

and

STANLEY R. J. WOODELL

calcium more avidly than magnesium in proportion to their quantities in the soil. The results of Johnston (1974) in Table XI1 show that this is not always the case, and there are likely to be differences between species. He showed that Cerastium holosteoides, for example, takes up much more magnesium, relative to calcium, at a site in Scotland. Table XI1 illustrates that the amounts of calcium and magnesium, as well as their relative proportions in plants on serpentine soil, are extremely variable. The few data concerning non-serpentine plants on serpentine soil (Table XIII) suggest that they are less capable of selective absorption of calcium and restriction of magnesium uptake, but again the danger of generalization is indicated by the different behaviour of Lycopersicon esculentum grown on a Californian and a Polish soil; in the former the plant had a lower Ca/Mg ratio than that of the soil (Walker, 1954), whereas in the latter it was much higher (Buczekand Leonowicz-Babiak, 1971).

3. Calcium deficiency Some workers have regarded calcium deficiency as the crucial factor in serpentine infertility. The ability to respond differently to calcium levels has been clearly demonstrated in non-serpentine plants (Bradshaw et al., 1958; Clarkson, 1965). Clarkson showed that the grass Agrostis setacea, a species of acidic heath soils, is specifically adapted to low levels of calcium in the external medium. Adaptation to low calcium levels has also been demonstrated for some serpentine plants. Main (1970), for example, found that serpentine strains of Agropyron spicatum, and also Poa curtifolia, a serpentine endemic, showed a different response from non-serpentine A . spicatum to increasing calcium in culture solution. The serpentine A . spicatum had an initially greater response to calcium, a t levels up to 2 mM, but then fell off in yield, while the non-serpentine strain maintained a linear response of increasing yield up to 4 mM. Poa curtifolia showed a declining yield above 0.02 mM. Both serpentine plants took up more calcium at all levels supplied. A more avid uptake of calcium, particularly at low levels in the growth medium, by serpentine plants has been a frequent experimental observation, e.g. by Walker et al. (1955), Grover (1960) and Jenkinson (1966). Many workers have demonstrated that the addition of calcium to some serpentine soils can improve growth of crop plants (e.g. Vlamis and Jenny, 1948; Walker, 1954; Kruckeberg, 1964; Soon, 1971; Proctor, 1971a), although this effect is not universal (e.g. Soane and Saunder, 1959). Vlamis (1949) investigated the growth of Romaine lettuce on a Californian serpentine soil and found very low yields, and

TABLEXI Exchangeubb calcium and mapmiurn in serpentine aoik ~

Locality New Caledonia France, Haute-Vienne et L’Auvergne

Remarks (Measurements refer to sample depth in cm where this is known)

Ca

(a) (b) (c) (d) (e) (f) (8) (h) (i)

(1) (m)

Author

0.43 0.43

0.48 0.18

0.89 2.4

Birrell and Wright (1945)

3.6 5.1 7.4 1.6 4.1 1.7 1.5 3.1 4.0 0.9 6.3

21 21 19 22 18 31 12 12 46 12

0.17 0.24 0.39 0.07 0.23 0.05 0-13 0.26 0.09

Duvigneaud (1966)

36 18 10.5

0.18 0-28 0.22

e

6.0

2.3

R 0

tr 0

2

0 q

Go M

Y

z2

0.08

2 z u

u,

8

E:

Western U.S.A., Clear Creek, California

(a) (b) (0)

Twin Sisters Mt., Washington

Ca/Mg ratio

E

Selection from analyses in original paper

(1) (k)

Western U.S.A., Wenatchee Mts., Washington

Mg

(mEq1100g soil)

0-13 13-38

~~

0.21

1.89

0.11

Griffin (1965)

0.62 0.25 3.14 1-62

3.06 3.73 8.39

0-20 0.07 0.37 0.24

Kruckeberg (1967)

6.86

z

4

TABLEXI (continued)

Locality Weatern U.S.A., Twin Sisters Mt., Washington Western U.S.A., Nevada Co., California Amador Co., CalifO~nh Lake Co., California Sonoma Co., California Scotland, Green Hill, Aberdeenshire Blackwater, Banffshire

Remarks (Measurements refer to sample depth in cm where this is known)

Ca Mg (mEq/lOOg soil)

Author

2.6

7.5

0-36

Kruckeberg (19694

1.28 1.50 3.40 3.47

8.01 8.01 17.00 20.32

0.16 0.19 0.20 0.17

McMillan (1956)

1.5 1.6 0.9

0.12 0.10 0.16 0.17 0.06

Prootor and Woodell (1971)

0.8

13-0 15.6 5-6 6.4 27.7 20.4 10.1 8.3 17.7 13.6 11.9 10.1 13.8 12.9

0.6 0.0

0.4 0.0

1.1

Hill of Towenreef, Aberdeenshire Glenkindie, Aberdeenshire Coylea of Muick, Aberdeenshire Meikle Kilrannoch, Angus 1 Meikle Kilrannoch, Angus I1 Rhum, Inverness-shire I

Ca/Mg ratio

1.7 1.4 1.2 0.3 4.4 3.1 0.3 0.5 1.1

0.07

0.12 0.04 0.26 0.23 0.03 0.05 0.08 0.06

1.50

-

$

3

*

Y

@

Y

8

0

0

Rhum, Inverness-shire 11 Unst, Shetland I

UM, Shetland I1 Unst, Shetland III Fetlar, Shetland Glendamel, Argyll Green Hill, Aberdeenshire Hill of Towanreef, Aberdeenshire Coyles of Muick, Aberdeenshire Meikle Kilrannoch, Angus unst, Shetland I Unst, Shetland 11 Glendamel, Argyll, calcareous vein Balmahie Hill, Ayrshire Grey Hill, A p h i r e Girvan, Ayrshire Loch Lomond

soils under closed vegetation, 0-16

-

0.0 0.5 4-1 5.4 7.0 6.3 5.7 5.8 2.8 2.2 5.75 7.97

0.40 0.29 0.30 0.40 0.38 0.25 0.29 0431 0.77 0.29 0.24

3.2 2.5 1.5 2.0 6.7 5-6 0.2 0.2 5.6 7.0 5.3 7.1 10.3 11-5

18.9 19.2 24.3 18.1 17.8 14-6 2.1 2.6 16.9 20.6 15.6 21.6 11.4 8.9

0-17 0.13 0.06 0.11 0.38 0.38 0.10 0.08 0.33 0.34 0.34 0.33 0.90 1-29

7.5 8.6

8.7 8.7

0-86 0.99

6.6 3-8 0.4

2.8 1.6 0.8

2.36 2.37 0.50

0.0 0.2 1.2 1.6 2.8 2.4 1.4 1.7 1-7 1.7 1.64 1.91

0

r 0

0

Ic

u1

5

E

0 c

(0

w

TABLEXI (continued) Remarks (Measurements refer to sample depth in cm where this is known)

Locality

Ca

E3

0

Mg

( d q / 1 0 0 g soil)

Ca/Mg ratio

Author 4

-

2

Scotland-tinued Dunbartonshire Glen Urquhart, Inverness-shire England, Kynance Cove, Cornwall Clicker Tor, Cornwsll

2

(b) (a) (b)

Puerto Rico, Mayaguez Mayaguez Mayaguez Eastern U.S.A., Dublin, Maryland Belmont, Maryland Cherry Hill, Maryland Oxford, Pennsylvania Jermantown, Virginia Hunting Hill, Maryland Potomac, Maryland Alberene, Virginia

0.9 32.6 31.8

z

0.78 0.26 0.26

8

e

0

11.9 15.1 2.6

31.8 34.8 8.2

0.38 0.43 0.32

3.7

4.7

0.79

Soil No. 2623: 10-40

3.06

0-85

3-60

Soil No. 9338: Soil No. 9343: Soil No. 9346:

1.47 0.66 0.68

1.91 0.37 0.12

0.77 1.78 4-83

(a)

(b)

Polyphant, Cornwall Cuba, Central Soledad

0.7 8.5 8.4

0-20 0-20 0-16

Soil No. 4722: 0-15 Soil NO. 5829: 1-20 Soil No. 6178: 0-10 Soil No. 6236: 0-30 Soil No. 6241: 0-11 Soil No. 9456: 0-18 Soil No. 9460: 0-23 Soil No. 9889: 0-20

0.11 0-54 0.39 0.49 5.95 0.11 0.49 4-62

2.47 0.73 0.70 5.22 8.61 3.55 0.22 5.82

0.04 0.74 0.56 0.09 0.69 0.03 2.22 0.79

0

E

Q

3?Robinson et al. (1935)

z

$

0 4

8

0 0

CJ

M

r r

Western U.S.A., Josephine Co., Oregon Napa, California Napa, California

Soil No. 9932: 5-25 Soil No. C114: 0-25 Soil No. C114: 26-85

2-68 2-24 1.62

3.36 22.74 23-52

0.79 0.10 0.07

Soil 1:

0.91 0.43 0.78 0.39 0.25 0.91 0.75 1.16 0.93 0.75

11.13 13.98 7-57 9.54 12-33 6-33 8.02 3.35 7.01 11-12

0.08

Poland,

G6ra Radunia

soil 2:

Wzg6rze nefrytowe

soil 3: Soil 4:

0-5 5-18 0-5 5-15 15-27 5-12 12-22 0-5 5-27 27-35

Sarosiek (1964)

e

0.03 0.10 0.04 0.02 0.14 0.09 0.35 0.13 0-67

i3 !03 tr

0

2

0 kl

W

Portugal

Soil EM36: 0-(10-15) Soil EM42: 0-(10-15) (1&15)-(2&30) (26-30)-70 Soil EM41: 0-(5-18) (5-18)-(25-30) 25-50 50-80 Soil EM44: 0-20 6-20 2040 Soil EM38: 0-(10-13) (10-13)-(3045) (30-45)-75

2.4 3.0 0.6 trace 2.2 1.6 0.5 0.5 1.2 2.1 0-3 1.5 2.3 3.0

10.7 11.8 11.9 11.9 4-5 4.9 9.2 14.8 5.9 4.4 8.1 11.0 18.3 18.7

0.22 0.26 0.05 < 0.01 0.49 0.33 0.05 0.03 0.20 0.48 0.04 0- 14 0.13 0.16

de Sequeira (1969)

w

$u

i m

8 W

W

E

W CQ CQ

TABLE X I (continued) ~

Locality Rhodesia

Remarks (Measurements refer to sample depth in cm where this is known)

Soil A Soil B

soil c Soil D

Soil E Soil F Soil G Soil H Scotland, Unst, Shetland Papua and New Guinea, Mt. Suckling

Western U.S.A., Lake co., California M& CO., CalifOdS Chelan Co.,Washington Whatcom Co., Washington

~

~~~

0-16 0-16 0-16 0-16 0-16 0-16 0-16 0-16

Ca MI3 (mEq/IOO g Soil) 2-56 0-79 2.68 0-76 0.72 0.92 0.67 1-63

6.25 13-2 12.7 1-23 1.9 2-38 12.9 1-96

Ca/Mg ratio

Author

0-41 0.06

Some and Saunder (1969)

8

0.2 1 0-62 0.38 0.39 0.06 0.78

0 0 H

s P

g, co

c3

20.8

136.7

0.16

Spence and Miller (1963)

0.3

0.8

0.38

Stevens and Veldkamp (pel%.comm.)

2-12 2-33 3-20 2.03 1.07

12.1 19-7 11.2 6-96 9-16

0.18 0.12 0.29 0-34 0.12

Walker (1964)

3.3

13.0

026

Whittaker (1960)

Weatern U.S.A.,

Siskiyou Mts., Oregon

&F ! 4

d

0

Rhodesia, Noro Mine area Ghoko Hills Tipperary Claims Kingston Hill

W e d a Mountain Thornwood Asbestos Mine Selukwe Noel Nickel Mine shangani Miks Hill

area with no woody species area with woody species treelees with trees barren area area with normal vegetation treelees with trees sparsely wooded densely wooded tmlesa with treee grassland with trees

2.1 3.7 3-48 2.2 8-9 1.7 18.3 3.6 10.4 1.4 2.6 2-4 26.6 3-1 19.1 3-89 11.03

17.2 4.6 10.9 10-3 9.6 2.7 20.0 6.4 23.6 6.2 4.3 5.6 14-4 10.9 23.3 4.87 3.86

0.12 0.82 0-32 0-21 0.93 0.63 0.92 0.66 0.44 0.27

2.7 3.4 4.2 2.9 1.1 4.2 6.2

14.2 11.3 28.76 14.9 8-66 31.9 16.6

0.19 0.30 0.16 0.19 0.13 0.13 0.32

Wild (19748)

4

M

R0 F

0

2

0.60

0.44 1-85 0.28 0.82

0 4

m

M

0.80

2.86

Woodell et al. (1974b)

TABLEXu. w

Colciunz and magneeiuna content of native plan0 in serpentine ~ o i l e

ra

le

soil

Plant

Species Juniperua cmunia

Empehwn nipurn

Cochlaria;sp.

Minuartia v e m

ceraatiurn hlQStWidG3

CauUna

vulga7ia

Agrostia cunina

Feetuca sp.

Plant part

C8 Mg Capg (mEq/lOOgdrytissue) ratio

Ca Mg Camg ( m E q / l O O g soil) ratio

Locality

Author 4

leaves leaves leaves leaves leaves leaves leaves leaves leaves leaves leaves leaves leaves leaves leaves Young shoots Yo-g shoots Young shoots leaves leaves leaves leaves leaves leaves

(a) (b)

59.88 45.91 78.84 31.18 13.97 17-22 12-41 15.72 20.96 5.37 6-73 12-48 5.74 4.37 5.24 17-96

24-10 21.87 42.44 51.4 48.1 38.65 74.73 73.44 76.48 73.44 75.08 74.51 101.72 140.43 73-44 33.3

2.48 2.1 1.86 0.61 0.29 0.45 0.17 0.2 1 0.27 0.07 0.17 0.06 0.03 0.07 0.54

2.5 2.8 91.5 2.4 2.1 1.9 1.7 1.5 1.7 1.7 1.5 2.0 1.1 1.2 1.4 1.1

12.4 11.3 19.6 10.2 11.7 12.3 10.2 10.2 11.7 12-4 12-1 17.4 6.4 8.6 7-1 11.6

0.2 0.25 4.67 0.24 0.18 0.15 0.17 0.15 0.15 0.14 0.12 0.11 0-17 0.14 0.20 0.09

35-68

24.1

1.45

1.7

12.4

0-14

26.70

34.79

0.77

1.7

10.8

0.16

5.24 5.74 4.24 3.99 5-24 4.24

56.17 49.84 48.11 48.11 33.3 34.79

0.09 0.12 0.09 0.08 0.16 0.12

0.9 1.7 1.2 1.4 1.1 1.2

7.7 9.5 7.0 10.1 8.5 7.7

0.12

0.09

0.18 0.17

0.14 0.13 0.16

Scotland, Hill of Towanreef

Johnston (1974)*

8r

Agroatia atolonijera

Agroatia canina

Thymua serpyllum

*

shoots

(a)

6.49

24-67

0.26

0.8

2.8

0.29

shoots

(b)

5.99

34.54

0.17

0-2

0.6

0.33

shoots

(c)

9.48

58.39

0.16

1-2

4.1

0.29

shoots

(d)

8.98

34.54

0.26

0.1

4.2

0.02

shoots

(0)

11.98

32.9

0.36

n.d.i

n.d.

-

shoots

(f)

3.99

20.56

0.19

1.4

5.7

0.25

shoots

(g)

9.98

35-36

0.28

4.7

16.3

0.29

shoots

(a)

7.49

45-23

0-17

1.5

13.0

0-12

shoots

(b)

8.48

3’7.83

0.22

1.7

27.7

0.06

shoots

(c)

7-49

27.96

0.27

4.4

17-7

0.25

shoots

(d)

9.48

68.26

0.14

0.0

0.0

-

shoots

(e)

5-99

21-38

0.28

0.3

11.9

0.03

0.53 0.56

0.91 (3.78

11-13 7.57

0.08

leaves leaves

193.6 199-1

363.5 355.3

0.10

Sweden, Proctor (1969) Mt. Atoklinten Sweden, Mt. Gurtatj&kko Scotland, Unst, Keen of Hamar, Shetland Sweden Kittelfjiill Sweden, Mt. Klumpliklumpen Scotland, Muckle Heog, Unst, Shetland Sweden, South of Ronnbiicksj on Scotland, Green Hill Scotland, Hill of Towanreef Scotland, Coyles of Muick Scotland, Hallival, Rhum Scotland, Meikle, Kilrannoch Poland, Saroeiek (1964) G6ra Radunia

0 4

w

Johnston’s soil samples were taken from directly beneath the sample plants

E3

cn

TABLEXII (continued)

w

E3

Species

Plant part

leaves leaves leaves Dianthus carthuaianum leaves leaves Veronica leaves ep;cata leaves finark leaves vulgaria leaves Silena inJEata leaves leaves Cfalium verum leaves leaves Lathyrms heterophyllus leaves leaves Vicia cracca leaves leaves Thymus leaves aerpyllum leaves T .pdqioidea leaves leaves Dianthua carthuaianum leaves Veronica leaves leaves leaves finar.iia leaves vulgaria leaves Silene i n w leaves

T . pulegwidea

-

Ca Mg Ca/Mg (mEq/lOOg dry tissue) ratio

196.1 188.6 213.1 211.0 190.6 196.1 213.6 226.5 348.3 421.2 256.0 283.4 212.6 226.0 191.1 175.6 236.0 247.0 225.0 240.0 212.1 215.1 210.1 252.6 266.0 293-4 365.8 456.1

366.0 405.4 479.5 523.9 436.7 426-0 477.8 430.1 909.6 956.5 620.1 684.2 749.2 663.7 611.0 699.9 402.1 428-5 417.8 413.7 560.9 569.9 430.2 436.7 545.3 508.2 866-0 830.6

Q,

soil

Plant

Ca Mg Ca/Mg (mEq/lOOgsoil) ratio

0.54

0.91

0.47

0.78 0.91

0-44 0.40 0.44 0.46 0.45 0.53 0.38 0-44 0.41 0.41 0.28 0.34 0.31 0.25 0.59 0.58 0.54 0.51 0-38 0.38 0.49 0.58 0.49 0.58 0.42 0.55

0-78 0.91 0.78 0.91 0.78

0.91 0.78 0.91 0-78 0.9 1 0.78 0.91 0.78 0.91 1.16 0.91 1-16 0.91 1.16 0.91 1.16 0.91 1-16 0.91 1-16

11.13 7.57 11-13 7-57 11.13 7.57 11.13 7-67 11-13 7.57 11.13 7-57 11.13 7.57 11.13 7-57 6.33 3.35 6-33 3.35 6.33 3.35 6.33 3-35 6.33 3-36 6.33 3-36

Locality

0.08

0.10 0.08

0.10 0.08

0.10 0.08

0.10 0.08

0.10 0.08

0.10 0.08

0.10 0.08 0.10 0.14 0.35 0.14 0.35 0.14 0.35 0.14 0.35 0.14 0.35 0.14 0.35

Wzg6rze nefiytowe

Author

Gdium verum

Lathyma hterophyUus Vkiuoracca Streptanthus glanddoaw

leaves leaves leaves leaves leaves leaves

265.5 310.9 243.5 234.0 196.6 183.6

653-8 710.6 725.4 644-8 667.8 622.6

0.41 0.44 0.34 0.36 0-29 0.29

0.91 1.16 0.91 1.16 0.91 1-16

leaves

52.1

121.0

0.43

2-12

6-33 3.35 6-33 3-36 6.33 3-35 12.1

0.14 0.35 0.14 0.35 0.14 0.35 0.18

U.S.A., California

Walker (1954)

U.S.A., Lake Co., California U.S.A., chelan co., California

Walker etd. (1955)

U.S.A., Santa Barbara Co., California

Woodell et d. (1974b)

m4

Val-.

pulchellus

Y

Helianthus bolanderi

leaves leaves

finonthus andro8meus

above ground

64.8 136

38.6

100 148

48-8

0.65 0.92

0.79

2.18 7.85

2.7

13.8 16.5

14.2

0.16 0.48

0.20

P-

above

37.4

45.4

0.82

4.2

28.8

0.15

ground Pad above ground

47.4

55-6

0.85

2-9

14.9

0.19

45.8

45.4

1.01

4.2

31.9

0.13

38.3

44.2

0-87

5.2

16.5

0.32

8 tc

0

z

0

w

9

E

2

zm

0

P m

P b S

above ground

Pabove ground

P-

W

to

4

TABLEXIII Calcium and magneaiuna content of non-nativeplan& grown on serpentine aoila

Plant Species Lycopersicon eacukntuna

“Best of all” Avena sativa

P d

0

Plant

soil

Ca Mg mEq/100 g dry tissue

Ca Mg mEq/100 g exchangeable

Ca/Mg ratio

n.d.

hl

n.d.

Ca/Mg ratio

seedlings

43.9

150.5

0-29

shoot leaves

68.9 37.9

115.1 119.2

0.60 0.32

shoots

13.97

41.12

0-34

11.9

31.8

0.38

shoots

3.07

15.05

0.20

0.0

0.0

-

shoots

6.66

47.86

0.14

1a2

4.1

0.29

shoots

6.30

90.63

0.07

0.3

11-9

0.03

233

0-15

2.12

12.1

0-18

“Victory”

Lycopersicon eaculentum

m

leaves

35.5

0.06

“Marglobe” Helianthua annuua

leaves

23.5

123

0.19

2-18

13.8

0.16

leaves

44.7

121

0.37

7.85

16.5

0.48

leaves

89.2

222

0.40

7.85

16.5

0.48

Linanthw, androsaceua

shoots

17.96

80.92

0.22

2-7

14.2

0-20

Lactuca sativcs

shoots

39.89

0.42

2-7

14.2

0.20

Fagopyrunz escukntum

“Romaine”

16.86

Locality Poland

Author Buczek and LeonowiczBabiak (1971)

4

England, Cornwdl Scotland, Rhum Scotland, Unst Scotland, Meikle Kilrannoch

Proctor (1971a)

88

U.S.A., Lake Co., California

Walker (1954)

U.S.A., Lake Co., California Chelan Co., California Chelan Co., California

Walker et d. (1955)

U.S.A., Santa Barbara Co., California

Woodell et d. (1974b)

2

Y

i3 b

Q

?

3 ? F 4

3

0

U

M

F

THE ECOLOGY OF SERPENTINE SOILS

329

“Rosette” disease symptoms. The addition of NPK partially improved yields, but on addition of calcium sulphate yields improved substantially and the disease symptoms decreased. When he added Ca-Amberlite (a synthetic exchange material carrying cations in the adsorbed state) together with NPK, the yields were high and plants healthy. On the other hand the addition of Mg and K amberlites produced “Rosette” symptoms. Magnesium caused a great reduction in growth at calcium saturations below 20%, and potassium a similar reduction when calcium saturation was below 30%. Analysis showed that the calcium content of the plants was profoundly reduced by magnesium or potassium addition. I n water culture low calcium produced disease symptoms, which were exacerbated in the presence of high magnesium or potassium. Vlamis interpreted this behaviour as a response to calcium saturation, and the effects of magnesium or potassium as the result of their inhibition of calcium uptake. Kruckeberg (1954), using serpentine and non-serpentine races of native species, also found that NPK addition could not ameliorate a serpentine soil, but in contrast the addition of calcium allowed the nonserpentine strains to grow on the serpentine. Different serpentine endemics tested had different low-calcium tolerances. He concluded: “The importance of factors other than calcium level for some plants cannot be excluded, but there is good reason to stress the degree of calcium saturation of the soil as of major importance in the serpentine problem. It appears probable that an important requirement for existence of serpentine endemics on serpentine soil is their capacity to obtain calcium at low levels.” Walker (1948a, 1954) also argued that calcium deficiency was important. He leached serpentine soils with chloride solutions in which he varied calcium and magnesium, and prepared soils with added NPK in which calcium varied from 5% to 80% of the total exchangeable cations, magnesium from 94% to 19%, while potassium stayed at approximately 1%. The growth of Lycopersicon esculentum var. Marglobe was compared with that of a serpentine endemic, Streptanthus glandulosus var. pulchellum. The unaltered field soil from the serpentine had a calcium saturation of 13.5%. At and below this level the tomato grew poorly, so that at 8.3% calcium saturation its yield was only 3% of that at a saturation of 13.5%. Above this calcium saturation its yield increased dramatically. Streptanthus, on the other hand, changed its yield little with different calcium saturation levels. Streptanthus absorbed more calcium than the tomato at most levels of saturation, but at low calcium saturation levels tomato absorbed much more magnesium than the Streptanthus. Walker suggested that this excessive magnesium in the tissues could hamper the normal utilization

330

JOHN PROCTOR

and

STANLEY R. J. WOODELL

of calcium, perhaps by ion competition at uptake sites on membranes. However, he stated: “The results above lend confirmation to the proposal of previous workers . . . that the basic cause of serpentine infertility is the low calcium level, and the writer is of the opinion that in the larger picture of serpentine tolerance and intolerance this is likewise the principal factor.” These examples make it clear that the addition of calcium to serpentine soils can often ameliorate their infertility or toxicity, but none of the experiments convincingly demonstrates calcium deficiency per se. I n recent years it has become clear that the calcium requirements of plants are much lower than was thought. Provided that other potentially toxic, though often essential, ions are at a low concentration and in balanced supply, plants can be grown successfully in solutions containing only a few ppm of calcium. Tobacco (Nicotiana tabacum) and maize (Zea mays) have been grown in media with 2 ppm calcium (Wallace et al., 1966). Leaves of maize under these conditions contained only 0.011% calcium, and those of tobacco 0.08% calcium in their dry matter. Loneragan and Snowball (1969a, b) grew various crops and pasture species in flowing nutrient solutions in which the calcium concentrations were kept at levels from 0.0003 to 1.0 mM. They grew perfectly well in the range 0.0003 to 0.0026 mM, and their shoots contained much lower calcium levels than those found in the field or grown in conventional nutrient solutions. Wyn Jones and Lunt (1967) reviewed the role of calcium in plants, and pointed out that plants have lower calcium requirements than had hitherto been believed. Nevertheless calcium is essential for the maintenance of membrane integrity and mitochondria1 activity. I n addition it may be required specifically for the activation of certain enzymes, and for nucleic acid metabolism. Epstein (1972) has commented: “Since such low values are tolerable only a t low concentrations of other divalent cations, it seems that the normally higher levels of calcium serve to render innocuous otherwise toxic concentrations of these metals.’’ He regarded much of the excess uptake of calcium by plants in the field as “luxury consumption”. Many non-serpentine soils have a lower calcium status than serpentine soils yet show no hint of serpentine vegetation. Robinson et al. (1936), referring to serpentine soils, recognized that “calcium, potassium and phosphorus were low, but not lower than in many soils which are successfully farmed”. White (1971) analysed the soil changes along twenty-one transects across serpentine/non-serpentine boundaries in Oregon, U.S.A. The results did not support the view that calcium was of overriding importance. Although there was a sharp vegetation change across the boundary, (a) maximum calcium levels were in five instances found on the

THE ECOLOGY OF SERPENTINE SOILS

331

serpentine, (b) calcium was effectively constant along the entire transect in five other examples, and ( c ) serpentine calcium level waa above 4000 ppm (25% nitric acid extractable) in eleven transects. White noted that many non-serpentine soils supporting normal vegetation had calcium levels below 4000 ppm. It is evident that there are dangers in singling out one factor to explain serpentine infertility, and in neglecting its interrelationship with others. Calcium is a good example of such dangers. Nowadays there is less support for the absolute importance of calcium deficiency per se. Calcium plays a very important role in the amelioration of toxicity, however, as we have suggested for nickel and chromium (and for magnesium, as will be shown later). The importance of calcium in this respect is reflected in the fact that serpentine soils high in calcium rarely show the extreme features associated with highly toxic soils. 4. Magnesium toxicity Magnesium toxicity has been recognized since Davy (1814) pointed out the ill-effects observed in plants limed with magnesian limestone. Trelease and Trelease (1931) demonstrated deleterious effects of high magnesium in water culture, and Lyon and Garcia (1944) found that a high magnesium level depressed the development of primary phloem. Grover (1960), working with Helianthw annuus, observed that over a range of calcium concentrations, any increase in magnesium beyond 0.02 mM, a very low level, caused a decrease in yield. This extreme toxicity of magnesium was also confirmed by Proctor (1969, 1970, 1971b), who used the root extension method in water culture of Wilkins (1957), and showed that 1 ppm magnesium in distilled water was sufficient to cause a reduction in root growth of a non-serpentine strain of Agrostis stolonifera. As Proctor demonstrated, the presence of calcium ameliorated this toxicity, and it is apparent that magnesium toxicity is dependent on the calcium concentration. Many serpentine species are adapted to high magnesium. Ferreira (1964) included serpentine plants in a series of experiments on baaiphile and acidiphile species. He grew five species on various soils, one of which was a quartzite t o which magnesium carbonate had been added, and another of which was a serpentine soil. The capacity to grow on the quartzite with added magnesium carbonate was (for Hinwzrtia verma and Lychnis alpina) correlated with their ability to grow on serpentine. Species of calcareous habitats died on both soils. He concluded that high magnesium is just as toxic to calciphiles as are heavy metals. Tolerance of high magnesium in water culture was demonstrated by Madhok (1965), who showed for example that the serpentine endemic Helianthua bolanderi ssp. exilis grew well and waa quite healthy in a

332

JOHN PROCTOR

and

STANLEY R. J. WOODELL

solution containing 20 mEq/l against a background of 1 mEq/l of calcium, in contrast to H . annuus for which this level of magnesium was lethal. Main (1970) demonstrated tolerance of high magnesium by Poa curtifolia, another serpentine endemic. He also showed that magnesium tolerance was inherited to the extent that an F, hybrid between serpentine and non-serpentine races of Agropyron spicatum in Washington State was intermediate in magnesium tolerance between the parents. Proctor (1971b) tested the tolerance of several British and Swedish clones of Agrostis stolonifera and A . canina ssp. montana, and showed that with the exception of those from a singularly nutrientdeficient soil from Rhum, Scotland, all the serpentine clones showed some tolerance to magnesium at 10 ppm, a level which is lethal to nonserpentine clones.

5 . Calcium-magnesium interactions Loew and May (1901) concluded from extensive experiments that a soil Ca/Mg ratio of not less than unity was necessary for healthy growth. However, this attempt to define a universal minimum ratio has not been helpful because of the great variability in species response. Many later authors have discussed the relationship between calcium and magnesium, including Blackshaw (1921), NovAk (1928), Walker et al. (1955), Jacob (1958), Grover (1960), Sarosiek (1964), Wild (1965), Madhok and Walker (1969), Sulej et al. (1970), Proctor (1971b), Buczek and Leonowicz-Babiak (1971), Lyon et al. (1971) and Woodell et al. (1974b). All regarded the relationship as important in serpentine soils. Most of the earlier workers based their views mainly on soil, or soil and plant analyses (many of which are summarized in Tables XI and XII). More recent experimental work by Walker, Grover, Madhok and Main at the University of Washington has been very helpful concerning the relationship between these two elements. The basis of much of this work was the experiments carried out by Walker et al. (1955) on a series of crop and native plants, using the technique described earlier of leaching and reconstituting soils so as to produce a range of calcium saturations. At levels below 20% the crop plants’ yields were low, while below 10% they grew hardly at all. The serpentine species showed no yield depression even a t 6% calcium saturation. Tissue analysis showed that the crop plants absorbed less calcium and more magnesium than the serpentine species. Potassium absorption was fairly constant in spite of the wide range of Ca/Mg ratios in the soils. Walker et al. concluded that great differences existed between species in their ability to tolerate low Ca/Mg environments, and that there was no reason to assume that the serpentine plants had

THE ECOLOQY OF SERPENTINE SOILS

333

a lower calcium requirement than crop plants. They were able to get all the calcium they required from the serpentine soils, and whereas they “excluded” magnesium, the crop plants underwent “luxury” consumption of magnesium from these soils. Grover (1960) and Madhok (1965) investigated one of these serpentine species, Helianthw bolanderi ssp. exilis, and the related crop plant, H . annuw, which had shown contrasted response in Walker et al.’s experiments. Grover (1960) varied Ca/Mg ratios over a wide range in culture solutions. Yields of H . annuw increased proportionately with an increase in Ca/Mg ratio up to 0.15, above which they rose very little. Changing the concentration of either ion while the other was maintained showed that at every magnesium level used, increasing calcium increased yield. Most of the increase in total yield was attained by the very low calcium concentration of 0.04 mM. This level of calcium appeared to be critical, increase above it having little effect. Highest yields occurred at the lowest magnesium level (0.001 mM). Analysis of variance showed that the interaction of the two elements was significant in determining yield, but less so than their individual effects. H. bolanderi ssp. exilis was depressed in growth at Ca/Mg ratios above 2.0, whereas a high yield was attained at a ratio as low as 0.05 (as compared with 0.15 for H . annuw). Madhok (1965) and Madhok and Walker (1969) found that magnesium deficiency occurred a t very low levels in both species, but a t a level of 0.4 mEq/l H . annuw .yieldlevelled’off. In contrast, H . bohnderi asp. exilis increased yield up to 20 mEq/l magnesium. At much higher levels H . bolanderi ssp. exilis declined slowly in yield, though even at 100 mEq/l the yield was still 70% of that at 10 mEq/l. The internal level of magnesium for optimum growth of exilis was almost twice (25 mEq/lOO g dry wt) that of annuw (9-14 mEq/100 g) at a solution culture calcium level of 8 mEq/l. Grover (1960), Madhok (1965) and Madhok and Walker (1969) invoked competition between these two ions in their explanation of these results, together with “luxury” consumption by H . annuw of magnesium. Grover (1960) carried out experiments investigating initial and steady-state uptake. He showed that the amount of adsorbed calcium on the roots of H. bolanderi ssp. exilis was enhanced with increase in the calcium concentration of the external medium, and that increasing the magnesium concentration inhibited the adsorption of calcium; magnesium and calcium were apparently competing for sites on this exchange surface. He also showed that the steady state absorption of calcium was a function both of calcium and magnesium concentrations in the external solutions, and that magnesium competitively inhibited the uptake of calcium. These results are in contrast with those of Epstein

334

JOHN PROCTOR

and

STANLEY R. J. WOODELL

and Leggett (1964), using barley roots, in which calcium and magnesium did not compete for the same sites or “carriers”. Grover noted that in the serpentine endemic, for calcium uptake in the presence of magnesium, the rate of calcium absorption at high calcium substrate concentrations was higher than would be expected on the basis of competitive inhibition between calcium and magnesium; he suggested that calcium levels in the solution in excess of 2 mM had an autocatalytic effect on the uptake of calcium in H. bolanderi ssp. exilia. This autocatalytic effect was not apparent in H . annuua and was regarded by Grover as pointing “towards a possible basis for the apparent avid absorption of Ca by the endemic species”. Madhok (1966) obtained similar results, and confirmed that higher calcium absorption than expected was obtained at calcium concentrations of higher than 2 mEq/l in H . bolanderi ssp. exilis, and in addition to Grover’s hdings, above 4 mEq/l in H . annuus. Madhok put forward the explanation that at the magnesium concentrations used, there was a more favourable Ca/Mg ratio for calcium uptake than a t lower levels of calcium in the external solution. He considered that competition between calcium and magnesium would be more likely to show up at deficiency concentrations than a t optimum concentrations of calcium, and further suggested that there may be two separate mechanisms for calcium uptake under varying magnesium concentrations. At low concentrations of calcium in the external medium it is competitively inhibited by magnesium, whereas at optimum concentrations of calcium the magnesium inhibition is non-competitive. This might involve two carrier sites for calcium uptake. His results for magnesium uptake under varying calcium concentrations indicated the presence of both competitive and non-competitive inhibition between calcium and magnesium at low and high magnesium substrate concentrations. Madhok’s results also suggested that the lower limit of substrate (calcium or magnesium) concentration at which the competitive inhibition by magnesium or calcium becomes operative, varies in the two species. For example, the lower limit of calcium concentration at which magnesium competitively inhibited calcium uptake seems to be higher in H. annuua than in H . bolanderi exilis. Hence it seems that magnesium competes more effectivelywith calcium uptake in H . annuw than in H . bolanderi exilis, and calcium competes more effectively with magnesium uptake in the case of H. bolanderi exilis than in the case of H. annuua.

6. A high magnesium requirement by serpentine plants Madhok (1966) and Madhok and Walker (1969) reported that the serpentine endemic sunflower, Helianthw bolanderi ssp. exilis, had a

THE ECOLOGY OF SERPENTINE SOILS

335

much higher internal magnesium requirement for optimum growth (about 25 mEq/100 g dry wt) than the cultivated species, Helianthus annum (about 14 mEq/100 g dry wt). Since the serpentine plant tended to exclude magnesium compared with the cultivated sunflower the levels of magnesium in the external solution required for optimum growth were relatively higher still. This high magnesium requirement for serpentine plants may be a widespread feature and has been reported in three grasses, the endemic Poa curtifolia and serpentine races of Agropyrm spicaturn (Main, 1970) and a serpentine race of Agrostia stolonifera (Mama, 1974). Such demonstrations of a requirement of high levels of an element known to be toxic t o many species in the concentrations found on serpentine, are of considerable ecological significance. A high magnesium requirement might be of more general importance in explaining the restriction of some plants to serpentines.

7. Interaction between magnesium, calcium and other ions There are several reports in the literature (e.g. Madhok, 1965; Epstein, 1972)referring to the interference of potassium with magnesium uptake. I n addition potassium is involved in reduction of calcium uptake in some situations (Vlamis, 1949). The phenomenon of induction of potassium deficiency by application of high levels of magnesium is reported, and could be important in Serpentine since it may aggravate problems caused by low soil potassium status. Sulej et al. (1970) found the assimilation of mineral nitrogen to be adversely affected at low Ca/Mg ratios, as was subsequent nitrogen metabolism, and Buczek and Leonowicz-Babiak (1971) showed from experiments with Lycopersicon esculentum that accumulation of magnesium in the roots disturbs nitrogen metabolism and hence restricts protein production. They concluded that this is the result of the unfavourable Ca/Mg ratio. A further facet of the N/Mg interaction is that demonstrated by Morgan et al. (1972), who found that nitrate absorption into the roots of Lolium perenne was much lower when supplied as magnesium nitrate than as calcium nitrate. Again, such an interaction might be of importance in view of the low nitrogen levels found on serpentine soils. A number of workers have reported a decrease in yield of plants growing on serpentine soils on the addition of nitrate (e.g. Proctor, 1971b; Woodell et al., 1974b). This is probably caused by an increase in solution magnesium that is brought about by the addition of large quantities of soluble salt.

336

JOHN PROCTOR

and

STANLEY R. J. WOODELL

8. The relationship between serpentine and maritime plants Several species occur both on serpentine and maritime soils (e.g.

Plantago maritima, Armeria maritima and Silene maritima in Britain). Proctor (1969) pointed out that maritime and serpentine soils have in common a low Ca/Mg ratio. Albert and Kinzel (1973) analysed many species including some that occur on serpentine soils from inland saline sites in Austria, and found that calcium varied from 0.8 to 110 mEq/ 100 g and magnesium from 26.3 to 206 mEq/100 g. The Ca/Mg ratio of the plant tissues ranged from 0.013 to 0.97, very similar to the range seen in serpentine plants. Goodwin-Bailey (pers. comm.), however, in a comparison of protein synthesis and phosphate uptake in inland serpentine, non-serpentine, and saltmarsh Armeria maritima, in varying magnesium concentrations in a constant calcium environment, found more similarity between the inland non-serpentine and serpentine races than between either and the saltmarsh race, and also found that the saltmarsh population was much the least tolerant of magnesium. He has suggested that Armeria maritima and possibly other species sharing similar habitats may be generally adapted to conditions of ionic stress. More evidence is required on this.

9. The ecological signi$cance of calcium and magnesium on serpentine The importance of low calcium and high magnesium levels as factors on many serpentines is clear, as is the complexity of the interaction between them. The simple view that calcium deficiency per se is responsible for serpentine “infertility” cannot be sustained, and although magnesium is extremely toxic, the work of Grover, Madhok and Proctor has shown that its importance is greatest in low-calcium external environments. The serpentine habitat, being low in calcium in many instances, is one in which not only magnesium but also other toxic ions, as we have seen, can affect plants. There are still several aspects of calcium and magnesium metabolism that require work before their significance on serpentine soils can be fully understood. The precise way in which calcium and magnesium interact during uptake is not yet clear. Is there both competitive and non-competitive inhibition as Grover and Madhok have suggested? How do magnesium and calcium interact inside the cell? What is the biochemical mechanism of magnesium toxicity, and of the universal ability of calcium to ameliorate toxicity, not only of magnesium but of other ions?

E.

OTHER U N U S U A L CHEMICAL F E A T U R E S O F P O S S I B L E I M P O R T A N C E TO P L A N T S

Over the years various additional factors have been suggested as being of importance. Of these a low soil molybdenum status has

THE ECOLOGY OF SERPENTINE SOILS

337

attracted most attention. The initial interest in this element followed in the wake of the discovery of molybdenum as an essential element. Walker (1948b) observed molybdenum deficiency in tomatoes grown in three Californian serpentine soils which had been fertilized with NPK. Since Walker’s original observation a number of other serpentine soils have been shown to have low levels of molybdenum (Johnson et al., 1952; Walker, 1954 and pers. comm.; atsuka and Takahashi, 1963; Sarosiek, 1964; Yatazawa and Tanaka, 1965). Other authors have found no evidence for a molybdenum deficiency (Vlamis, 1949; Vergnano, 1953; de Sequeira, 1969; Proctor, 1969). The possibility of suoh a deficiency being important in any natural situation on a serpentine soil is still difficult to judge. The results for crop plants can hardly be extrapolated since wild plants are likely to have a lower molybdenum requirement. (Good evidence of the variability of molybdenum requirement between plant species was provided by Johnson et al., 1952). We can add nothing to Walker’s (1954) conclusion: “It appears reasonable to state a t present that molybdenum deficiency may contribute to the infertility of some serpentine soils but is probably not the dominant factor in any case.” I n view of the large amounts of total iron in serpentine it is not surprising that this element has attracted attention as a possible toxic factor. Ferrous iron is known t o be toxic (Martin, 1968), although ferric iron is less soluble and is less likely to be toxic. The proportion of ferrous iron relative to ferric iron depends on redox potential and pH. Serpentine soils are sometimes poorly drained, which might favour increased solubility; but on the other hand the pH of such soils is frequently around neutral, at which iron is only slightly soluble. Although iron toxicity has been suggested by several authors (Novhk, 1928; Kretschmer, 1931; Rune, 1953; Tassoulas, 1970), there is a lack of experimental evidence. Interest in iron toxicity has recently been revived by Ritter-StudniEka and Dursun-Grom (1973). They have reported some very high levels of iron, up to 13 723 ppm in the leaf dry matter of Potentilla tommsiniana plants from a serpentine site in Bosnia. Johnston (1974) has also found very high iron levels in some Scottish serpentine plants, e.g. up to 28 000 ppm in Blutcomitrium lanuginosum. I n addition there are a number of other features of some serpentines which have been discussed. Deficiencies of chloride (NovBk, 1928), copper (Spence and Millar, 1963), zinc (Otsuka and Takahashi, 1963) and sulphur (Walker, 1964; Otsuka, 1963) have been reported. The following toxicities have been suggested: boron (Sarosiek, 1964; Shkoljnik and Smirnov, 1970), barium (Tassoulas, 1970) and copper and zinc (Yatazawa and Tanaka, 1965). Finally, Gordon and Lipman

338

JOHN PROCTOR

and

STANLEY R. J. WOODELL

(1926) claimed that some serpentine soils were infertile because of a very high pH. None of these factors appear to be of widespread importance, although they might be worth further investigation in specific cwes.

V. ANIMALSO N SERPENTINE SOILS Very little work has been done on animals associated with serpentine soils. Johnson et al. (1968) observed the restriction of the butterfly Euphydryas editha to serpentine sites in the San Francisco Bay Area of California. The host plant of the butterfly larva Plantago erecta occurs on and off serpentine soils in the area, but the butterfly feeds only on those plants growing on serpentine. Johnson et al. analysed P . erecta from non-serpentine and Serpentine sites, but observed no differences between samples that could be correlated with presence or absence of Euphydryas editha. They also found that larvae fed in the laboratory on non-serpentine Plantago erecta exhibited no obvious illeffects. They concluded that the distribution of Euphydrym in the San Francisco Bay region was determined by microclimatic and possibly disease-prevalence factors, rather than by chemical features of the serpentines. Proctor and Whitten (1971) noted that there was an unusually large population of the pocket gopher (Thomomys bottae) on a serpentine soil in California. The reason for this was that Brodiaea corms which formed the principal food of the gopher were present in great abundance in the serpentine soil, much more so than in an adjacent soil (possibly because of reduced competition for light). The Brodiaea had three to four times more magnesium than calcium in its tissues and probably high contents of some heavy metals. How the gophers were adapted to this nutritionally peculiar diet is not known. Stebbins (1949) observed that certain zones of intergradation between subspecies of the salamander Ensatina eschscholtzi are related to the presence of serpentine. He attempted (pers. comm.) to demonstrate deleterious effects of serpentine soils on the salamander but obtained no evidence for this. This appears to comprise the whole body of animal work on serpentine. Why the animals of serpentine have been so little studied we do not know, but they offer L fascinating field for further work. VI, FUNGIA N D BACTERIA I N SERPENTINE SOILS There is very little information on the macrofungi. Maas and Stuntz (1969) oollected 279 species on serpentine in the Cascade Mountains of

THE ECOLOQY O F SERPENTINE SOILS

339

Washington. Of these “67 were either lignicolous or not directly dependent on soil type. Nineteen percent of the remaining 212 species were found only in Serpentine soil areas and sixty-three percent were found only in non-serpentine soil areas. Eighteen percent of the 212 species were common to both soil type areas.” There is also very limited information on microorganisms. Lipman (1926) noted a dearth of soil microorganisms in samples from the Mount Diabolo region of California. He tested the ability of the serpentine soils to nitrify ammonia or nitrite in the usual nitrification solutions. No nitrification was noted with any one of the soils even after two months’ incubation. He commented that in his “experience with hundreds of tests on nitrification with as many soils, this is the first instance of soil free from excess of salts or toxins, which contained no nitrifying bacteria, or at least in which the nitrifying bacteria, if present, were wholly inactive”. He examined the soils for Azotobaoter and found only a few cells in one of the soils, none in the others. Lipman explained the poor microflora in terms of a deficiency of suitable organic and inorganic substrate for growth. Lloyd (pera. comm.) has succeeded in isolating nitrifying bacteria from a high nickel serpentine in Rhodesia. Although low in number, the nitrifying bacteria were active. White (1967) looked at the presence or absence of nitrogen-fixing nodules in Ceanothw cuneatw in non-serpentine and serpentine soils. He found that the percentage of nodulated plants was 56% and 3% respectively. He pointed out that the few plants that were found nodulated in serpentine soils grew only in microsites where other rock types and soils had been added by alluvial or colluvial mixing. He speculated that the paucity of nodulated plants on serpentine soils may be associated with low concentrations of nitrogen, phosphorus, potassium or molybdenum, and toxic concentrations of chromium and nickel, and suggested that low available molybdenum may be of particular significance since plants which fix nitrogen require more molybdenum than those that do not. Sedova (1958) considered that if chromium is above a concentration of 1 ppm of soil, it interferes with nitrification. I n this connection an observation of Wild (1974b) is of great interest. The legume Pearsonia rnetallij’era was observed to have 15 000 ppm chromium and 102 500 ppm nickel in the root ash when growing on a serpentine soil in Rhodesia, yet the roots were well nodulated with Rhizobium which was pink and active. Plants of another endemic species, Lotononis serpentinicola, and of a more widely distributed legume, Sesbania microphylla, were similarly well nodulated with pink nodules on the mme soil. Tesic et al. (1967) and Ritter-StudniEka (1970) observed in southern

340

JOHN PROCTOR

and

STANLEY R. J. WOODELL

Europe fewer species and significantly smaller numbers of individuals in the soil flora of serpentine compared with that of dolomite. I n raw serpentine the number of microorganisms involved in ammonification was found never to exceed 300 000 per g of fresh soil. In more humusrich, mesic soils the number is greater but still only one tenth of the numbers occurring in comparable soils on dolomite. The above authors also commented on the paucity of the fungus flora in these southern European serpentine soils. Tadros (1957) suggested that the paucity of pathogenic microorganisms is a cause of the restriction of some plants to serpentine soils. He investigated experimentally the observation that the plant Emmenanthe rosea is restricted to serpentine soils in parts of California whilst the related E . pendulijlora occurs on other soils. The seedlings of E . rosea were much the more susceptible to microbial attack in the laboratory and showed greatly improved survival on sterilized soils. Tadros concluded that the soil microorganisms hinder the establishment of seedling E . rosea on non-serpentine soil. There is clearly scope for much more work on the microorganisms of serpentine soils. Apart from general information on their distribution we would also like to see them used in investigations into tolerances against various putative toxic factors. They seem ideal experimental material for this and have been used in such investigations by workers in other fields (Antonovics et al., 1971).

VII. EVOLUTIO ON N SERPENTINE Heslop-Harrison (1964) stated: “There is no particular reason for supposing that the adaptation of plants to their rooting media will involve processes or principles different from those governing adaptation to the sub-aerial environment, but it so happens that there are examples of clear-cut patterns of adaptation to soil types which should permit a more precise kind of analysis than can be given to most examples of adaptation to climatic or biotic influences.” Open habitats on base-rich soils often harbour rare plants, endemics and species of disjunct distributions, and serpentines are no exception. They can often be regarded as “museums” of plant evolution.

A.

ECOTYPIC DIFFERENTIATION

We have already mentioned examples of serpentine races of plants, adapted to chemical factors in the serpentine environment (e.g. to magnesium (Proctor, 1971b) and to nickel (Ernst, 1972)). Such adapta-

THE ECOLOQY OF SERPENTINE SOILS

34 1

tion is probably widespread, and because of the complexity of the environment is itself likely to be complex. Kruckeberg (1961, 1964) showed ecotypic differentiation on serpentine in Achillea borealis, a species in which a range of climatic ecotypes had already been demonstrated by Clausen et al. (1948). The edaphic races that Kruckeberg demonstrated are superimposed upon the climatic races, and he pointed out that to describe as ecotypes local populations which represent adaptations not only to a special local factor, but also to a complex of factors operating over the species’ whole range, would be misleading. He regarded it as better to refer to natural populations as comprising an array of ecotypic variation. Individuals of some populations of Achillea borealis from nonserpentine soils sometimes grew as well on serpentine in cultivation as those of a serpentine race. Such a situation would be expected to lead to quite rapid selection of tolerant races, since these individuals are preadapted, and indeed this does occur in the evolution of heavy metal tolerance on mine spoil-heaps (Antonovics et al., 1971). Whittaker (1954b) suggested that in the more recently glaciated and less climatically extreme serpentines of the Pacific northwest of the U.S.A.,the rate of evolution of perennials in adaptation to serpentine might be slower. Kruckeberg (1967), investigating the view that the serpentine habitats here might be less rich in serpentine ecotypes, found that in fact ecotypic differentiation occurred in 16 out of 18 species that he was able to test. He tested Achillea millefolium from a non-serpentine site adjacent to serpentine, and from one far away. The seedlings from the adjacent site showed better growth on serpentine soil, probably because of gene flow from the serpentine populration. Further, Kruckeberg noted that Prunella vulgaris and Rumex aceto8ella, which had almost certainly been introduced to the area within the last 60 t o 76 years, can be found on some serpentine outcrops. I n these cases rapid selection for serpentine tolerance appears to have taken place. The rates of establishment on different serpentines in the same area can differ markedly. We have seen serpentine sites in the Coast Ranges of southern California where introduced Mediterranean grasses of the genera Avena and Bromus occur both on and off the serpentine, whereas at Jasper Ridge, a few hundred km north, species of the same genera, common on adjacent non-serpentine soils, are almost completely excluded from the serpentine. Research on adaptation to a chemical factor in the serpentine environment, carried out in the same detail as that for the similar situations on heavy metal mines (reviewed in Antonovics et al., 1971), would be particularly welcome. M

342

JOHN PROUTOR

and

B.

STANLEY R. J. WOODELL

ENDEMICS

The occurrence of rare species on serpentine attracted the attention of several early workers. Novhk (1928) suggested that many, if not all, serpentine species were exclusive to serpentine, but this view was modified later (Novhk, 1937) and serpentine plants were classified into obligate and facultative serpentinophytes. I n fact there is little experimental evidence that a plant exists on serpentine that cannot be grown under cultivation on non-serpentine soil. Lammermayr (1926, 1927, 1928a, b, 1930, 1934)denied the existence of species growing exclusively on serpentine soil in areas of Europe. I n many parts of the world serpentines have large numbers of endemic species. Whittaker (1954b) recorded the following data in the Siskiyou mountains. Rock type Total number of species Number of endemics

diorite 101 2

serpentine 113 30

The number of endemics is correlated with both the richness and age of the surrounding flora. Thus northern serpentines are relatively poor in endemic species (Rune, 1953) because their flora has only recently arrived after the last glaciation. By contrast, in tropical and other unglaciated areas there has been no such sweeping elimination of plant cover and more endemism is to be expected. It has been suggested by Wild (1973) that endemics began their evolution on serpentine in south central Africa soon after the origin of the Angiosperms. Certainly there is a large range of endemic species on the Rhodesian serpentines. Two types of endemic occur on serpentine. They are “insular species” (Stebbins, 1942) (“serpentinophytes” of Pichi-Sermolli, 1948),which are differentiated only on serpentine and related rocks; and “depleted species” (“serpentinicolous relics” of Pichi-Sermolli), which formerly occurred over a wide area and are now restricted to serpentine. Kruckeberg (1954) has been particularly clear on the subject of serpentine endemism. He claimed to have observed biotype depletion in progress in the genus Streptanthus. S. glandulosus still has a few nonserpentine biotypes although it is primarily a serpentine species. On the other hand S. breweri and S. batrachocarpus seem to be endemic to serpentine. Kruckeberg postulated that in these latter species the gradual depletion of the non-serpentine races has led to endemism. Kruckeberg attempted a hypothesis of the development of serpentine endemism in a genus. He envisaged a species of normal soils extending into an area where serpentine occurs. Among the populations are individuals at least partly adapted to low calcium or other conditions

THE ECOLOGY OF SERPENTINE SOILS

343

of serpentine, seeds of which may fall on serpentine soils or transitional areas. Those progeny survive which are best adapted to serpentine. I n time, selection takes place of a biotype distinct from that on nearby non-serpentine soils. If then the non-serpentine plants are eliminated from the surrounding areas, the species has become a depleted serpentine endemic. As Kruckeberg put it: “It has, step by step, gone through a hypothetical sequence from serpentine exclusion to serpentine endemism which appears to exist contemporaneously among the different species of a genus (e.g. Streptanthus).” He postulated that “insular” species could arise in two ways. Once the establishment of a serpentine endemic has taken place, occasional seeds might reach other serpentine outcrops, to which they are somewhat pre-adapted; some of these new populations may develop into different biotypes and possibly into “insular” species (by, for instance, the founder effect (Mayr, 1963)). An alternative possibility suggested by Kruckeberg Is that a wide-ranging species may come into contact with several serpentine outcrops and develop different races, which eventually may become endemics. Either process could produce a cluster of closely related serpentine endemic species like those in Streptanthus. On the basis of what has occurred in other genera, the latter alternative appears to us more likely. There are numerous examples of wide-ranging species with a number of locally adapted races, and of clusters of closely-related species, not necessarily adapted to similar habitats, and it seems reasonable to suppose that they have evolved independently in their different localities. Kruckeberg cited as examples of “insular” species Quercus durata, Ceanothus jepsoni and Cupressus sargenti, which are all found on Californian serpentines. The question of whether a particular serpentine endemic is “insular” or “depleted” cannot always be easily decided. The classes are not rigidly distinct and overlap certainly occurs.

c. T H E

EXCLUSION O F S E R P E NT I NE ENDEMIUS FROM OTHER S O I L S

Two explanations have been put forward to account for this, (1) that serpentine endemics have a requirement for some factor provided by serpentine soils; (2) that serpentine endemics are poor competitors in closed vegetation. High Mg requirements have been found in the endemics Helianthus bolanderi sap. exilis (Madhok, 1965; Madhok and Walker, 1969) and P o a curtifolia and ecotypes of Agropyron spicaturn (Main, 1970) and Agrostis stolonifera (Marrs, 1974). There is some evidence of a nickel requirement in serpentine plants of Hybanthus firibundus (Severne and

344

JOHN PROCTOR

and

STANLEY R. J. WOODELL

Brooks, 1972). There are highly suggestive accumulations of nickel, chromium and cobalt in certain other serpentine plants, but good evidence of specific requirements is lacking and further work is required. Spence ( 1959) discussed observations that suggested that the elimination of pioneer species, including the endemic Cerastium arcticum ssp. edmonstonii, on serpentine on the Keen of Hamar, Shetland, was the result of seedling elimination on unsuitable soils. The precise requirement of the seedlings provided by open debris soils was not known but could be chemical or physical. There are some recent data involving the competitive ability of serpentine endemics. Forde and Faris (1962) found some populations of oaks in the Coast Ranges of northern California, where hybridization had occurred; individuals were present with characteristics of both Quercus durata, a serpentine endemic, and Q. dumosa, not found on serpentine. Usually where serpentine meets non-serpentine there is abrupt replacement of one Quercus species by the other, but there is often a scattering of hybrids in the transition zone. The nearest population of Q. durata to the hybrid mixtures in question is about 6 km away on serpentine. The hybrids grow on a mosaic of rather infertile shallow soils, which do not have serpentine characteristics. Forde and Faris suggested that the survival of these hybrids in a non-serpentine situation is the result of their being in an inhospitable total environment which is dry, acid, infertile and mineral-deficient, and presumably where competition is relaxed. Griffin (1965) has mentioned that Pinus sabiniana, in which he did not find ecotypic differentiation, is able to compete as part of the woodland vegetation on zonal soils in dry regions, whereas in more mesic regions it is confined to serpentine and cannot enter the denser forest. This suggests that some species are living on serpentine soils because they are more open and free of competition than surrounding soils, whereas elsewhere those species can grow on shallow rock soils which are also competition-free. Howard-Williams (1970) observed that the cuprophile species Becium homblei is restricted to metal-rich soils in Rhodesia by interspecific competition with the more widespread and closely related B. obovatum. B. homblei has a population on a nickel-rich serpentine soil at Tipperary Claims, Rhodesia and the situation is similar to that for the copper populations. B. obovatum does not occur on the serpentine presumably because it cannot tolerate the toxic soil, whereas B. homblei does not occur off the serpentine because it cannot tolerate competition there. These field observations are supported by some experimental work. Kruckeberg (1954) sowed a mixture of Brassica sp., Erodium sp., Lolium perenne, Medicago sp. and Avena sp., all weeds in Californian grassland, together with an endemic Streptanthus species on a Yo10 sandy

THE ECOLOGY OF SERPENTINE SOILS

346

loam and an unfertilized serpentine soil. At maturity, the Yo10 loam carried a lush growth of weed species, but no Streptanthus, whereas the serpentine had a good growth of Streptanthus with a few stunted grass seedlings. All the dicotyledonous species died or did not appear at all on the serpentine. Streptanthus did well on the loam when sown alone. Kruckeberg repeated these experiments using the same soils and a third, which was a serpentine reconstituted with calcium. The results on this soil were similar to those on the loam. As Kruckeberg admitted, it would be naive to interpret these results as confirmation of the competitive exclusion hypothesis, and even if competition is restricting the serpentine plants to the serpentine, the nature of the competition is unknown. Proctor (1971~)carried out a field experiment on competition, and his results substantiate those of Kruckeberg. I n the serpentine area of Jasper Ridge, near Stanford, California there is a great abundance of a serpentine ecotype of Plantago erecta, yet this species is absent from the adjacent sandstone area. A number of experimental plots were set up on the sandstone area, variously fertilized, and some cleared of vegetation. I n these were sown seeds of Plantago erecta. After five months, only the cleared plots contained Plantago. This indicated the importance of competition in excluding Plantugo from this soil, since the plants in the cleared experimental plots were healthy and produced seed. It seems that endemics may be restricted to serpentines in some cases by special requirements and in others by poor competitive ability. It is possible that they may operate together since lack in the soil of a special requirement could lower competitive ability. Poor competitive ability is probably also a result of an intrinsically slower growth rate, which appears to be a general feature of plants of nnfavourable environments and which has been demonstrated in serpentine ecotypes by Proctor (1971a) and Woodell et al. (1974b). D.

P L A N T S SHOWING D I S J U N C T D I S T R I B U T I O N ON S E R P E N T I N E S

There are very many examples of this; for example Duvigneaud (1966) reported that the fern Nothlaena maranthae is restricted to serpentine in the northern part of its range in France. Conversely, boreal species have southern extensions of their ranges on serpentines. Rune (1953) reported that in Scandinavia Arenaria norvegica is, south of latitude 66"N, almost restricted to serpentine. Pigott and Walters (1954) argued that the local species of base-rich habitats in Britain have survived the post-Glacial forest cover only in low-competition situations. They did not refer t o serpentine, but in Britain a few speoiee

346

JOHN PROCTOR

and

STANLEY R. J. WOODELL

are confined to this and a few other substrates, and they are mostly low-growing plants which are probably poor competitors.

E.

MORPHOLOGICAL D I F F E R E N C E S S H O W N B Y SERPENTINE PLANTS

The morphological characters by which many plants on serpentine are distinguished from their conspecifics elsewhere have been called “serpentinomorphoses” by Novhk (1928). Pichi-Semolli (1936, 1948) and Messeri (1936) recognized six such modifications: (i) reduction of the size of leaves and other organs; (ii) shrubbiness of growth and plagiotropism; (iii) stunting; (iv) much greater development of root systems; (v) greater glaucousness; and (vi) reduced pubescence. Other features observed have been increased pubescence (e.g. Proctor and Woodell, 1971) and a distinctive colouration of the plants (e.g. RitterStudniEka 1967, 1968). Many of these characteristics give an overall impression of xeromorphism. Bargoni (1940) observed in serpentine plants of Armeria denticulata that the leaves were smaller and thinner and that the leaves had a very thick palisade tissue and a lower epidermis with two layers of cells. Ritter-StudniEka (1972) compared plants of the same species growing on adjacent serpentine and non-serpentine sites. The serpentine plants had fewer stomata and more tightly-packed leaf palisade cells. Ritter-StudniEka regarded these as features which would assist survival in physically adverse sites, and added that the serpentine plants have a higher vacuolar sap content. She pointed out that other workers had noted increased succulence of plants on serpentine. This is not a universal feature, however. I n Rhodesia there is no evidence for incrertsed succulence. The number of succulent Euphorbia and Aloe species is fewer than on granite soils where the proportion of bare stony lithosols is greater (Wild, 1974, pers. comm.). The presence of this syndrome of factors has led to the assumption by authors such as Pichi-Sermolli that drought is largely responsible for the evolution of these xeromorphisms. It is likely that they are partly the result of low moisture, but a contributory cause must be low tissue production. Whittaker ( 1954b) suggested that the effects of decreased moisture are simulated by the effects of decreased nutrients, since productivity is reduced in both cases and because some of the same plant growth-forms are adaptations to both dryness and infertility. Beadle (1953) pointed this out in his account of xeromorphism in rainforests of northern Australia. Even in the wet Scottish Highlands xeromorphism occurs and it is not confined to serpentine. Shkoljnik and Smirnov (1970) observed the development in culture of morpho-

THE ECOLOGY OF SERPENTINE SOILS

347

logical changes in plants caused by high concentrations of boron, nickel, cobalt and chromium. The changes, stunted growth, stenophyllism and profusely branched shoots, were similar to those well known in many serpentine plants. “Serpentinomorphoses” are probably partly fixed genetically and partly environmental modifications. Sadeback (1887) claimed to have observed the gradual disappearance of the characteristics of the serpentine fern Asplenium adulterinum in garden culture, but his observations have never been repeated. Dvofhk (1935) explained many serpentinomorphoses as merely modifications caused by the soil. Despite some loss of distinguishing features in the British serpentine plants Minuartia verna var. gerardii and Cerastium arcticum ssp. edmonstonii on cultivation in potting compost, we have recorded that some features remain distinct. Clausen et al. (1947) crossed the very clearly demarcated serpentine race Layia glandulosa ssp. discoides with the type race and observed a striking segregation of characters in the F2 generation.

F.

SPECIATION

Normally isolation is regarded as a prerequisite for speciation, but the absolute necessity for spatial isolation has been called into question. The work of Thoday and his collaborators (Thoday, 1969; Thoday and Boam, 1959; Thoday and Gibson, 1962) using Drosophila melanogaster has shown that disruptive selection can produce racial divergence even in a situation where a high degree of outbreeding occurs. HeslopHarrison (1964) argued that for plants, the interdigitation of habitats, or the occurrence of a mosaic of habitats, raises the possibility of evolution of ecotypes, even in the face of cross-breeding in a panmictic population. This is provided that phenotypes appropriate to the conditions of one habitat survive there and are eliminated from the other, and vice versa. Woodell et al. (1974a) have found that Linanthw, undrosacew, in the San Rafael Mountains of southern California, has a flower colour polymorphism on and off serpentine, the serpentine populations being mainly purple and the non-serpentine white, and that serpentine individuals flower earlier, even in cultivation, than do those from non-serpentine sites, though serpentine and non-serpentine populations are sometimes contiguous. Dickinson and Antonovics (1973), using a deterministic computer simulation, have demonstrated that sympatric speciation is possible in a two-niche situation. Self-fertility can evolve in a population in response to the deleterious effects of gene-flow (Antonovics, 1968). Using a “mine” and a “pasture” situation, the model showed that more

348

JOHN PROCTOR

and

STANLEY R. J. WOODELL

selfing occurred on the mine than on the pasture, but the selfing gene did not reach fixation on the mine. According to the model, this situation could only arise if there were less inbreeding depression on the mine than in the pasture. I n fact, self-fertility does occur in zinc-lead mine populations of normally self-incompatible species (Agrostis tenuis, Anthoxanthum odoratum, Armeria maritima) (Antonovics, 1968; Lefhbvre, 1969) and has been interpreted by these workers as an adaptation to reduce gene flow from plants which are non-tolerant, growing adjacent to the mines. Lefhbvre (1973) has since questioned this, and as a result of the investigation of another Armeria population has suggested that self-fertility may be just an adaptation to the initial colonizing stage of an unfavourable habitat. This may be important in the evolution of serpentine races; if an isolated serpentine-tolerant plant were also self-compatible the establishment of a new ecotype would be much more rapid. It is clear that serpentine outcrops in an area of “normal” soils could provide the potential situation for disruptive selection to take place, but it must be regarded as not yet proven. There is another type of speciation: “saltational speciation” (Lewis 1962, 1966)’ in which adjacent populations, very similar in morphological and ecological adaptation, differ greatly in chromosome arrangement and sometimes in basic chromosome number. There have been few investigations of chromosome number of plants on and off serpentine. Main (1970) found no difference between serpentine and nonserpentine races of Agropyron spicatum in Washington State, U.S.A., and Proctor and Woodell (1971) also found no differences in serpentine and non-serpentine Agrostis caninu, A . stolonifera, Rumex acetosa and Silene maritima. Lewis has evidence that in several pairs of species of Clarkia, the relationship of two species is that of parent to offspring. For instance, C . franciscana, which occurs on a serpentine outcrop in San Francisco, differs from C. rubicunda by several translocations and inversions. Lewis cites this and several other examples as evidence that multiple chromosome differences can arise and become established simultaneously under certain circumstances. The situations in which he postulates that this can occur are those in which an exceptional environmental extreme, such as drought, eliminates most of the individuals in a population, leaving only those with the abnormal chromosome arrangement. After this “catastrophic selection” has occurred, the ensuing forced inbreeding produces a chromosomal barrier to gene exchange, and there results an irreversible genetic independence between parental and derived populations, Most of the examples he quotes are responses to aridity, but in the case of C .franciscana, this has been reinforced by its occurrence 011 serpentine. Once established, it

THE ECOLOGY O F SERPENTINE SOILS

349

hrts been unable to spread because i t is surrounded by non-serpentine sob. Thus the sharp environmental discontinuity provided by serpentine, and reinforced by drought, may have allowed the rapid selection of a new species. Such examples may be rare but they should be looked for. Raven (1964) has indicated that serpentines are exactly the places where catastrophic selection would be most likely to occur.

V I I I . CONCLUSIONS

A constant theme of this review has been the extreme variability of the serpentine habitats and the organisms dwelling there. Biologists include many rock types, each of varying chemical composition, under the heading “serpentine”. These rocks are subject to weathering processes which depend on many soil-forming factors. The vegetation of serpentines presents a wide range of appearances. Biologists have concentrated on the situations where serpentine vegetation is in sharp contrast with that of the surroundings. Such contrasts do not always occur and when this happens the serpentine vegetation is often not documented. The more closely studied serpentines have many vegetation features in common but the causes of these similarities can be very different. I n some cases physical factors are important, in others high magnesium and low calcium, in others high nickel and possibly chromium, and in still others low nutrient levels or even factors peculiar to one or a few sites. These factors act together but have usually been studied independently. We have discussed some studies on the interactions of factors but this is a little known field with the notable exception of Mg/Ca relationships. Mg/Ni interactions in particular should be further explored, and also the relationship between metal toxicities and low nutrient levels. At the population level, evolution in response to the serpentine environment differs from species to species, as the work on endemics and ecotypes suggests. There are clearly diverse mechanisms by which species adapt to the extreme conditions of serpentine soils. Some plants control uptake of one or more elements, but not others. Some tolerate (or may even require) high tissue levels of magnesium, nickel, chromium, cobalt and iron. The mechanisms of such tolerance vary. Some may lock toxic factors in cell walls, some may rapidly translocate toxins to leaves which are subsequently shed, others may counteract toxicities by accumulating calcium. Thresholds at which tolerance mechanisms fail differ from species to species.

360

JOHN PROCTOR

and

STANLEY R. J. WOODELL

There is much scope for further work in many fields on serpentines. Many parts of the world have had their serpentine vegetation i n d e quately described. Some aspects have been almost totally neglected, e.g. the animals, and others only slightly better studied, e.g. the microorganisms. A great deal remains unanswered about the tolerance mechanisms to toxic factors and about possible unusual requirements (of nickel and chromium in particular) in serpentine plants. It is clear that serpentines will continue to delight biologists by their intrinsic interest and beauty and at the same time offer much scope for research at all levels of organization of living things.

ACKNOWLEDGEMENTS We would like to express our thanks to Professor H. Wild for supplying much unpublished information, and to Professor 0. Vergnano Gambi for her help in tracing some of the early history of serpentines. They also, together with Professor A. D. Bradshaw, Dr T. J. King, Dr A. R. Kruckeberg, Dr P. S. Lloyd, and Dr R. B. Walker, provided very helpful comments on the manuscript. Dr R. H. Whittaker is thanked for permission to reproduce Fig. 1 from his paper.

REFERENCES Abelson, P. H. and Aldous, E. (1950). Ion antagonisms in microorganisms: interference of normal magnesium metabolism by nickel, cobalt, cadmium, zino and manganese. J . Bact. 60, 401-413. Adamovic, L. (1909). Die Vegetations-verhiiltnisse der Balkenlander (Mosische Lander). I n “Die Vegetation der Erde” (Eds Engler, A. and Drude, O.), Vol. 11,pp. 1-567. Englemann, Leipzig. Adiga, P. R., Saatry, K. S., Venkatasubramaniam, V. and Sarma, P. S. (1961). Interrelationships in trace element metabolism in Aepergillua niger. Biochern.

J . 81, 545-550. Albert, R. and Kinzel, H. (1973). Unterscheidung von Physiotypen bei Halophyten des Neusiedlerseegebietes (Osterreich). 2. Pflanzenphysiol. 70, 138-157. Amidei, G. (1841). Specie di plante osservate nei terreni serpentinosi. Atti Term Riunione Scienz. It., Firenze, 523-524. Antonovics, J. (1968). Evolution in closely adjacent plant populations. VI. Manifold effects of gene flow. Heredity, Lond. 23, 507-524. Antonovics, J.,Bradshaw, A. D. and Turner, R. G. (1971). Heavy metal tolerance in plants. Adv8 ecol. Ree, 7 , 1-85. Aumento, F. (1970). Serpentine mineralogy of ultrabmic intrusions in Canada and on the Mid-Atlantic Ridge. Bull. geol. Sum. Can., Dept. of Energy, Minee and Resourcm-Paper 69-53. Bargoni, I. (1940). Cenni sull’ anatomia della foglia di due individui di Armeria denticulata Bert. Nuovo G. bot. ital. 47, 507-509. Bargoni, I. ( 1943). Osservazioni fenologiche sulle serpentine dell’Impruneta (Firenze). Nzlovo a. bot. ital. 50, 232-251.

THE ECOLOQY OF SERPENTINE SOILS

351

Beadle, N. C. W. (1963). Some aepects of the ecology and physiology of Australian xeromorphic plants. Auat. J . Sci. 30, 348-356. Beard, J. S. (1963). The savanna vegetation of northern tropical America. E d . Monogr. 23, 149-216. Becherer, A. (1969). Serpentinfame des Tessin und des italienischen Grenzgebietes. Bauhinia 4, 66-66. Beck von Mannagetta, G. (1901). Die Vegetationsverhiiltnisse der illyrischen Llinder. I n “Die Vegetation der Erde” (Eds Engler, A. and Drude, 0.)Vol. 4, pp. 1-634. Englemann, Leipzig. Beckett, P. H. T. (1966). The cation-exchange equilibria of calcium and magnesium. Soil Sci. 100, 118-123. Beger, H. K. E. (1922). Assoziationestudien in der Waldstufe des Schadggs. Jber. W u r j . Gee. QraubUndens 61, 1-96. Beger, H. K. E. (1923). Assoziatiomtudien in der Waldstufe des Schanfiggs. Jber. naturj. Gea. QraubUndem. 62, 97-147. Bennett, H. H. and Allison, R. V. (1928). The soils of Cuba. Trop. Plant. Rea. Found. BuU., Washington D.C. Betts, M. W. (1918). Notes on the autecology of certain plants of the peridotite, Nelson. 1. Structure of some of the plants. Tram. Proc. N.Z. Imt. 50, 230-242.

Betts, M. W. (1919). Notes on the autecology of certain plants of the peridotite, Nelson. 2. Tram. Proc. N.Z. Inst. 51, 136-166. Betts, M. W. (1920). Notes on the autecology of certain plants of the peridotite, Nelson. 3. Trans. Proc. N.Z. I m t . 52, 270-314. Billings, W. D. (1962). The environmental complex in relation to plant growth and distribution. Q . Rev. BioZ. 27, 251-206. Birrell, K. S. and Wright, A. C. S. (1946). A serpentine soil in New Caledonia. N.Z. Jl. Sci. Technol. 27A, 72-76. Bjerlykke, B. (1938). Vegetasjonen pa olivinstenpa Sunmere. Nytt Mag.Naturwid. 79, 61-126. Blackshaw, G. N. (1921). Magnesia impregnated soils. S. Air. J. Sci. 17, 171-178. Bogatyrev, K. P. (1968). “Smolnitsy” (smonitsa) of Albania (cinnamon-brown meadow and meadow cinnamon brown dark-coloured magnesium solonetzlike soils). Pochvowedenie No. 4, 14, 368-364 (English translation). Boydell, H. C. (1921). The magnesite deposits of Euboea, Greece. E m . Qeol. 16, 607-623.

Bradehaw, A. D., Lodge, R. W., Jowett, D. and Chadwick, M. J. (1968). Experimental investigation into the mineral nutrition of several grass species. I. Calcium level. J . Ecol. 46, 749-767. Bradshaw, A. D., Chadwick, M. J., Jowett, D. and Snaydon, R. A. (1964). Experimental investigations into the mineral nutrition of several grw species. IV. Nitrogen levels. J . Ecol. 52, 666-676. Braun, E. L. (1960). “Deciduous Forests of Eastern North America.” Blakiston & Co., Philadelphia. Braun-Blanquet, J. (1932). “Plant Sociology.” McGraw-Hill, New York. Braun-Blanquet, J. (1951). “Pflanzensoziologie” (2nd edn). Springer, Wien. Braun-Blanquet, J. and Jenny, H. (1926). Vegetations-Entwicklung und Bodenbildung in der alpinen Stufe der Zentralalpen (Klimaxgebiet des Caricim c u m h e ) . Mit besonderer Beriicksichtigung der Verhiiltnisse im schweizerischen Nationalparkgebiet. Neue Denbchr. Schweiz. naturj. Qea. 63, 181-349.

352

JOHN PROCTOR

and

STANLEY R. J. WOODELL

Breeze, V. G. (1973). Land reclamation and river pollution problems in the Croal Valley caused by wwte from chromate manufacture. J. appl. Ecol. 10, 513-625.

Brenchley, W. E. (1938). Comparative effects of cobalt, nickel and copper on plant growth. Ann. appl. Biol. 25, 671-694. Buczek, J. and Leonowicz-Babiak,K. (1971). The effects of calcium to magnesium ratio on the increments of dry weights and protein contents in tomatoes. Acts SOC.Bot. Pol. 40, 557-667. Burns, A. F. and Smith, A. M. (1965). Pulverized serpentine as a source of available magnesium. Agric. Chem. 20, 23-26, 168. Butler, J. R. (1953). The geochemistry and mineralogy of rock weathering. I. The Lizard area, Cornwall. Geochim. comchim. Acta 4, 167-178. Caesalpino, A. (1583). G. Marescotti, Firenze. De plantis libri XVI,369. Carabia, J. P. (1945) The vegetation of Sierra de Nipe, Cuba. Ecol. Monogr. 15, 321-341.

Clarkson, D. T. (1965). Calcium uptake by calcicole and calcifuge species in the genus Agroetia. J . Ecol. 53, 427-435. Clausen, J., Keck, D. D. and Hiesey, W. M. (1947). Heredity of geographically and ecologically isolated races. Am. Nat. 81, 114-233. Clausen, J., Keck, D. D. and Hiesey, W. M. (1948). Experimental studies on the nature of species. 111. Environmental responses of climatic races of Achillea. Puble Carnegie Inatn No. 581. Cockayne, L. (1928). The vegetation of New Zealand. In “Die Vegetation der Erde” (Eds Engler, A. and Drude, O.), 2nd edn, Vol. 14, pp. 1-466. Englemann, Leipzig. Coombe, D. E. and Frost, L. C. (1956a). The heaths of the Cornish serpentine. J. Ecol. 44, 226-256. Coombe, D. E. and Frost, L. C. (1956b). The nature and origin of the soils of the Cornish serpentine. J. Ecol. 44, 605-615. Cotton, M. (1930). Toxic effects of iodine and nickel on buckwheat grown in culture solutions. Bull. Torrey bot. Club 57, 127-140. Crooke, W. M. (1966). Effect of soil reaction on uptake of nickel from a serpentine soil. Soil Sci. 81, 269-276. Crooke, W. M. and Inkson, R. H. E. (1955). The relationship between nickel toxicity and major nutrient supply. Pl. Soil 6 , 1-15. Crooke, W. M., Hunter, J. G. and Vergnano, 0. (1954). The relationship between nickel toxicity and iron supply. Ann. appl. Biol. 41, 311-324. Dahl, 0. (1912). Botaniske undereskelser i Helgeland (1). K . Norake vidensk. Selak. Kriatiana, Skr. 1911, 1-221. Dahl, 0. (1915). Botaniske undereskelser i Helgeland (2). K . Norake viderrek. Selek. Kriatiana, Skr. 1914, 1-178. Daniker, A. U. (1939). Neu-Cdedonien. Vegetationsbilder 25, V I , 1-9. Davidson, J., Grant-Mackie,J. A., Morton, J. E. and Rattenbury, J. A. (1969). North Cape (New Zealand). A scientific case for conservation. Tane 16,Ei-1 1. Davy, Sir Humphry (1814). “Elements of Agricultural Chemistry” (2nd edn). London. Del Moral, R. (1972). Diversity patterns in forest vegetation of the Wenatchee Mountains, Washington. Bull. Torrey bot. Club 99, 57-64. Dickinson, H. and Antonovics, J. (1973). Theoretical considerations of sympatric divergence. Am. Nat. 107, 256275. Duvigneaud, P. (1966). Note sur la biogeochemie des serpentines du sud-ouest de la France. Bull. SOC.r. Bot. Belg. 99, 271-329.

THE ECOLOGY OF SERPENTINE SOILS

363

DvotAk, R. (1936).Naniemi planttlrum in stepposis ad substratum serpentinaceum prope Mohelno in Moravia. Arch. Natur-u. Heim.&?chzltzM & r i a c h , L a d e 6, 1-152. (Czech. with German summary.) Eggler, J. (1964). Vegetationsaufnahmen und Bodenuntersuchungen von den Serpentingebieten bei Kirchdorf in Steiermark und bei Bernatein im Burgenland. Mitt. natuw Ver. Steierm. 84, 26-37. Eggler, J. (1966).Beitrag zur Serpentinvegetation in der Gulsen bei Kraubath in Obersteiermark. Mitt. natunu. Ver. Steierrn. 85,27-72. Ellis, B. S. (1961).The soils of Rhodesia. Rhodtwia ag&. J. 48, 182-212. Epstein, E. (1972).“Mineral Nutrition of Plants: Principles and Perspectives.” Wiley, New York. Epstein, E. and Leggett, J. L. (1964).The absorption of alkaline earth cations by barley roots: Kinetics and mechanism. Am. J. Bot. 41, 786-792. Ernst, W. (1968).Der Einfluss der Phosphatversorgung sowie die Wirkung von ionogenem und chelatisiertsm Zink auf die Zink und Phosphataufnahme einiger Schwermetallpflanzen. Phyaiologia P1.21, 323-333. Em&, W. (1972). Ecophysiological studies on heavy metal plants in South Central Africa. KirkM 8, 126-146. Ernst, W. G. (1969).“Earth Materials.” Prentice-Hall, Englewood Cliffs, New Jersey. Faust, G. T. and Fahey, J. J. (1962).The serpentine-group minerals. Prof. Pap. U.S.geol. Sum. 884-A. Faust, G. T., Murata, K. J. and Fahey, J. J. (1966).Relation of minor element content of serpentine8 to their geological origin. Beochim. comchim. Acta 10, 316-320. Fernald, L. M. (1907).The soil preferences of certain alpine and subalpine plants. Rhodora 9, 149-193. Fernald, L. M. (1911).A botanical expedition to Newfoundland and southern Labrador. Rhodora la, 109-162. Fernandez, T.C., Taboadela, M. M. and Ojea, F. G. (1966).Niquel en 10s suelos de la provincia de la Coruila. An. Edafol. Agrobiol. 24, 267-293. Ferreira, R. E. C. (1969).Some distinctions between calciphilous and besiphilous plants. I. Field data. Tram. Proc. bot. SOC.Edinb. 89,399-421. Ferreira, R. E. C. (1964).Some distinctions between calciphilous and baaiphiloue plants. 11. Experimental data. Trans. Proc. bot. SOC.Edinb. 89, 612624. Ferreira, R. E. C. and Wormell, P. (1971).Fertiliser response of vegetation on ultrabasic terraces on Rhum. Tram. Proc. bot. SOC.Edinb. 41, 149-164. Fiori, A. (1919). Contribuzione alla flora dei serpentini del Pavese. Bull. Soc. botan. Ital. 1919,39-40. Fiori, A. and Pampanini, R. (1914). La flora dei serpentini della Toscana 11, Confront0 tra la flora del M. Ferrato (serpentino) e quella della Calvana (calcare alberese). Nuovo B. bot. ital. 21,216-240. Forde, M. B. and Faris, D. G. (1962).Effect of btrogression on the serpentine endemism of Quercua durata. Evolution, Lancaeter, Pa. 16, 338-348. Frasche, D. F. (1941). Origin of the Surigao iron ores. Econ. ‘Beol. 86, 280306. Gauckler, K. (1964). Serpentinvegetation in Nordbayern. Ber. bayer. bot. Bm. ao, 19-26. @droits, K. K. (1916).Poglotitel naya sposobnost pochvy (adsorption capacity of soils). Zhur. Opyt. Agron. 14, (In Bogatyrev, 1958.)

354

:JOHN PROCTOR

and

STANLEY R. J. WOODELL

Gismondi, A. (1953). Intorno ad una particulare fitocenosi comune a1 Ferrini serpentinosi ed a1 calcare dolomitico in Liguria. Accad. figure Sci. e lett. Atti 9, 240-241. Goldschmidt, V. M. (1954). “Geochemistry.” University Press, London. Gordon, A. and Lipman, C. B. (1926). Why are serpentine and other magnesian soils infertile? Soil Sci. 22, 291-302. Gregory, R. P. G. and Bradshaw, A. D. (1965). Heavy metal tolerance in populations of Agrostia tenuia Sibth. and other grasses. New Phytol. 64, 131-143. Griffin, J. R. (1965). Digger pine seedling response to serpentinite and nonserpentinite soil. Ecology 46, 801-807. Grover, R. (1960). “Some Aspects of Ca-Mg Nutrition of Plants with Special Reference to Serpentine Endemism.” Ph.D. Thesis, Univ. of Washington, Seattle. Haantjens, H. A. and Bleeker, P. (1970). Tropical weathering in the territory of Papua and New Guinea. Aust. J . Soil Res. 8, 157-177. Halliday, G. (1960). “Studies on the Arenaria ciliata and Minwrtia verna Complexes.” Ph.D. Thesis, Univ. of Cambridge, U.K. Halstead, R. L. (1968). Effect of different amendments on yield and composition of oats grown on a soil derived from serpentine material. Can. J . Soil. Sci. 48, 301-305.

Halstead, R. L., Finn, B. J. and MacLean, A. J. (1969). Extractability of nickel added to soils and its concentration in plants. Can. J . Soil. Sci. 49, 335-342. Harada, M. (1953). The weathering of serpentine in Wakasamachi, Tottori Prefecture. J . Sci. Soil Manure, Tokyo 23, 137-40. (Japanese with English summary.) Harrison, R. D. (1972). “Book of Data.” Penguin, London. Harshberger, J. W. (1903). The flora of the serpentine barrens of south-east Pennsylvania. Science, N . Y . 18, 339-343. Harshberger, J. W. (1904). A phyto-geographical sketch of extreme south-eastern Pennsylvania. Bull. Torrey bot. Club 31, 125-159. Haselhoff, E. (1893). Versuche uber die schadliche Wirkung von nickel-haltigem Wasser auf Pflanzen. Landw. Jbr. 22, 862-867. Hayek, A. (1923). Pflanzengeographie von Steiermark. Mitt. naturw. Ver. Steierm. 59, 1-209. Heslop-Harrison, J. (1964). Forty years of genecology. Adva ecol. R M .2, 159-247. Hewitt, E. J. (1953). Metal interrelationships in plant nutrition. I. Effects of some metal toxicities on sugar beet, tomato, oat, potato, and marrowstem kale grown in sand culture. J . exp. Bot. 4, 59-64. Hilgard, E. W. (1906). “Soils.” Macmillan, New York. Holdridge, L. R. (1945). A brief sketch of the Puerto Rican flora. In “Plants and Plant Science in Latin America” (Ed. Verdoorn, F.), pp. 81-83. Chronica B o t d c a , Waltham, Mass., U.S.A. Hotz, P. E. (1964). Nickeliferous laterities in south-western Oregon and northwestern California. Econ. Qeol. 59, 355-396. Howard-Williams, C. (1970). The ecology of Becium homblei in central Africa with special reference to metaliiferous soils. J . Ecol. 58, 745-763. Hoyos de Castro, A. (1960). La Genhe du sol sur roches basiques. Trana. 7th Int. Congr. Soil Sci. 4, 413-417. Hunter, J. G. (1954). Nickel toxicity in a Southern Rhodesian soil. S. Afri J . Sci. 51, 133-135. Hunter, J. G. and Vergnano, 0. (1952). Nickel toxicity in plants. Ann. appZ. Biol. 39, 279-284.

THE ECOLOGY OF SERPENTINE SOILS

356

Hunter, J. G. and Vergnano, 0. (1953). Trace-element toxicities in oat plants. Ann. appl. Biol. 40, 761-777. Iberg, R. (1954). “Beitrag zur Kenntnis von Tonmineralien einiger schweizerischer B6den.” Diss. E. T. H., Ziirich. Igoshina, K. N. (1966). Specificity of the flora and vegetation of hyperbasic rocks in the Polar Urals (as illustrated by the example of Mt. Rai-is). Bot. Zh. SSSR. 51, 322-338. Iljinski, A. P. (1936). Zur Phytooenologie der Larchenwalder des Ilmengebirges (Siid-Ural).Ber. 8chweiz. bot. Qes. 46, 86-93. Ishimoto, T. T. (1958). “Systematic Considerations as Influenced by Certain Ecological Factors Related to Plant Distribution on Serpentine Soil in central California.” P b D . Thesis, Oregon State College. Jacob, A. (1958). “Magnesium, the Fifth Major Plant Nutrient.” Staples Press Ltd, London. Jeffrey, D. W. (1971). The experimental alteration of a Kobrk-rich sward in Upper Teesdale. I n “The Scientific Management of Plant and Animal Communities” (Eds D d e y , E. and Watt, A. S.), pp. 79-89. Blackwell, Oxford. Jenkinson, L. J. (1966). “Differential Responses of Pinw pmderoea Laws. Seedlings from Different Seed Sources in Californiato Forest Soils Formed on Ultramafic and Granitic Rocks.” Ph.D. Thesis, Univ. of California, Berkeley. Johnson, C. M., Pearson, G. A. and Stout, P. R. (1952). Molybdenum nutrition of crop plants. PI. Soil 4, 178-196. Johnson, M. P., Keith, A. D. and Ehrlich, P. R. (1968). The population biology of the butterfly, Euphydryas editha VII. Evolution, L a w t e r , Pa. 22, 422-423.

Johnston, W. R. (1974). “Mineral Uptake by Plants of Serpentine and Lead Mine Soils.” B.A. Hons Thesis, Univ. of Stirling, U.K. Jowett, D. (1958). Populations of Aqroatis spp. tolerant of heavy metals. Nature, Lond. 182, 816. Kanno, I., Onikura, Y. and Tokudome, S. (1965a). Genesis and characteristics of brown forest soils derived from serpentine in Kyushu, Japan. Part 3. Clay mineralogical characteristics. Soil Sci. PI. Nutr. 11, 225-234. Kanno, I., Tokudome, S., Arimura, S. and Onikura, Y.(1965b). Genesis and characteristics of brown forest soils derived from serpentine in Kyushu, Japan. Part 2. Genesis and characteristics of brown forest soils. Soil Sci. PI. NU&. 11, 141-150. Kelley, W. P., Dore, W. H., Woodford, A. 0. and Brown, S. M. (1939). Colloidal constituents of Californian soils. Soil Sci. 48, 101-65. Kitamura, S. (1950). Adaptation and isolation on the serpentine areas. Acta phytotm. geobot., Kyoto. 12, 178-185. (In Japanese.) Kitamura, S. (1952a). Serpentine flora of Mt. Shibutsu, Prov. Kodzuke, Japan. Acta phytotux. geobot., Kyoto. 14, 174-176. (In Japanese.) Kitamura, S. (1952b). Serpentine flora of Mt. Hayachine, Prov. Rikuchu, Japan. Acta phytotax. geobot., Kyoto. 14, 177-180. (InJapanese.) Kitamura, S. and Momotani, Y. (1952). Serpentine flora of island Sugaahima, Prov. Shima, Japan. Acta phytotux. geobot., Kyoto. 14, 118-119. (Japanese with English summary.) Kitamura, S. and Murata, G. (1952). Serpentine flora of Prov. Awa, Shikoku, Japan. A& Phytotm. geobot., Kyoto. 14, 120-122. (In Japanese.) Kitamura, S., Hiroe, M. and Nakai, G. (1960). Serpentine flora near Sekinomiya, Prov. Tazima, Japan. Actaphytotux. geobot., Kyoto 12,186-186. (InJapanese.)

356

JOHN PROCTOR

and

STANLEY R. J. WOODELL

Kitamura, S., Murata, G. and Toni, K. (1953). Serpentine flora of Mikawa Province. Acta phytotax. geobot., Kyoto. 15, 1-4. (In Japanese.) Knaben, G. (1952). Botanical investigations in the middle districts of Western Norway. Bergen Univ. Aarbok, Naturv, rekke 1950 8, 1-117. de Kock, P. C. (1956). Heavy metal toxicity and iron chlorosis. Ann. Bot., N.S. 20, 133-141. Kotilainen, M. J. (1944). Kasvit erikoislaatuisen substraatin indikaattoreina. hablVwaik. SOC.Scient. fefin. 22B, 1-18. Kotilainen, M. J. and Seivala, 0. (1954). Observations on serpentine flora a t Sunmere. A serpentinicolous form, Cerastium alpinurn. Nytt Mag. Bot. 8, 319. Krapfenbauer, A. (1967). An autoecological study of a serpentine site in the Dunkelstein forest and a vessel experiment with Pinua eylvmtrb and Pinua nigra on serpentine soil. Zentbl. ges. Foratw. 84, 207-230. Krause, W. (1958). Andere Bodanspezialisten. Handb. PJPhyaiol. 4, 755-806. Krause, W. (1962). Zur Kenntnis der Flora undvegetation aufSerpentinstandorten des Balkans. 4. Mikroprllparate von Serpentinboden aus Gnechenland. Z . PjlEmiihr. Diing. Bodenk, 99, 97-107. Krause, W. and Klement, 0. (1958). tiber die Felsflechten zweier jugoslavischer Serpentingebiete. Vegetatio 8, 1-17. Krause, W. and Klement, 0. (1962). Zur Kenntnis der Flora und Vegetation auf Serpentinstandorten des Balkans. 5. Flechten und Flechtengesellschaften auf Nord-Euboa (Gnechenland).Nova Hedwigia; 4, 189-262. Krause, W. and Ludwig, W. (1956). Zur Kenntnis der Flora und Vegetation auf Serpentinstandorten des Balkans. I. Halacsya aendtneri (Boise.) Diirf. Bot. Jb. 69, 417-428. Krause, W. and Ludwig, W. (1987). Zur Kenntnis der Flora und Vegetation auf Serpentinstandorten des Balkans. 2. Pflanzengesellschaften und Standorte im Gostovic-Gebiet (Bosnien). Flora, Jena, 145, 78-131. Krause, W., Ludwig, W. and Seidel, F. (1963). Zur Kenntnis der Flora und Vegetation auf Serpentinstandorten des Balkans. 6. Vegetationsstudien in der Umgebung von Mantoudi. Bot. Jb. 82, 337-403. Kretschmer, L. (1931). Die Pflanzengesellschaftenauf Serpentin in Gurhofgraben bei Melk. Verh. 2001.-bot. cfes. Wien 80, 163-208. Kruckeberg, A. R. (1950). “An Experimental Inquiry into the Nature of Endemism on Serpentine Soils.” Ph.D. Thesis, Univ. of California, Berkeley. Kruckeberg, A. R. (1951). Intrwpecific variability in the response of certain native plant species to serpentine soil. Am. J . Bot. 88, 408-419. Kruckeberg, A. R. (1954). Plant species in relation to serpentine soils. Ecology 85,267-274.

Kruckeberg, A. R. (1964). Ferns associated with ultramafic rocks in the Pacific Northwest. Am. Fern J . 54, 113-126. Kruckeberg, A. R. (1967). Ecotypic response to ultramafic soils by some plant species of northwestern United States. Brittonia 19, 133-151. Kruckeberg, A. R. (1969a). Plant life on serpentine and other ferromagnesian rocks in northwestern North America. Syeaia 2, 15-1 14. Kruckeberg, A. R. (1969b). Soil diversity and the distribution of plants, with examples from western North America. MudroAo 20, 129-154. Lam, H. J. (1927). En plantengeografisch Dorado. Hand2g. I V Neded I d . Natumu, Congr. 386-397. Llimmermayr, L. (1926). Materialien zur Systematik und (5kologie der Serpentinflora. I. Neue Beitrage zur Kenntnis der Flora Steirischer Serpentine. Sber. Akad. Wka. Wien. 125, 369407.

THE ECOLOGY OF SERPENTINE SOILS

367

Liimmermayr, L. (1927).Materialien zur Systematik und Okologie der Serpentinflora. 11. Drts Problem der Serpentinpfbnzen-eine kritische Gkologische Studie. Sber. Akad. Wise. W k n . 186, 25-68. Lammermayr, L. (1928a). Weitere Beitriige zur Flora der Magnesit- und Serpentinbaden. Sber. Akad. W h e . Wien. 187, 66-99. Liimmermayr, L. (1928b). Vierter Beitrag zur Okologie der Flora auf Serpentinund Magnesitboden. Sber. Akad. W k e . Wien. 187, 825-869. Liimmermayr, L. (1930). Neue floristische Ergebnisse der Begehung steirischer Magnesit- und Serpentinlager. Verh. 2001. -bot. Bes. W k n 80, 83-93. Liimmermayr, L. (1934). Ubereinstimmung und Unterschiede in der Pflanzendecke iiber Serpentin und Magnesit. Mitt. naturw. Ver. Steierm. 71,4142. LefBbvre, C. (1969). “Gtudes Biosystematiques sur le Genre A m r i a . ” Ph.D. Thesis, Univ. of Brussels. LefAbvre, C. (1973). Outbreeding and inbreeding in a.zinc-lead mine population of A m r i a mcCritim. Nature, Lond. 241, 96-97. LeGendre, C. (1919). Flore des terrains de serpentine de la Haute-Vienne. BuU. Bdogr. bot. 29, 19-34. Lewis, H. (1962). Catastrophic selection as a factor in speciation. Eprolution, Lcmcaeter, Pa. 16, 267-271. Lewis, H. (1966). Speciation in flowering plants. Science. N. Y. 152, 167-172. Lipman, C. B. (1926).The bacterial flora of serpentine soils. J. B a t . 12,316-318. Litvinenka, A. (1962). Properties of a crust of weathering of ultrabasic rocks in the Central Dneiper region. Dokl. Akad. Nauk SSSR. 147, 1444-7. (In Russian.) LodoEnikov, W. N. ( 1936). Serpentine und Serpentinite der Iltschirlagerstiitte und im allgemeinen und damit verbundene petrologische Probleme. Central Beol. and Proep. Inst. Trans. 28, 817 p. (In Faust and Fahey, 1902.) Loew, 0. (1892). ober die physiologischen Functionen der Calcium und Magnesiumsalze im Pflanzenorganismus. Flora, Jena, 50, 368-394. Loew, 0. and May, D. W. (1901). The relation of lime and magnesium to plant growth. Bull. Bur. PI. Ind. U.S.Dep. Agrio. 1, 1-63. Loneragan, J. F. and Snowball, K. (1969a). Calcium requirements of plants. A-t. J . q&. R a . 20, 466-478. Loneragan, J. F. and Snowball, K. (1969b).Rate of calcium absorption by plant roots and its relation to plant growth. Auat. J. q r k . Res. 20, 479-490. Lounamaa, J. (1966). Trace elements in plants growing wild on different rocks in Finland. Ann. Bot. SOC.2001.-bot. fenn. “ V a m ” 29 (4), pp. 190. Low, A. P. (1884). Report on explorations and surveys in the interior of the Gasp6 Peninsula. Rep. geol. Sum. Can. 1882-1884. Lyon, C. and Garcia, C. R. (1944).Anatomical response in the anatomy of tomato plants to variations in the macronutrient cation supply. Bot. Baz. 105, 441-466. Lyon, G. L., Peterson, P. J. and Brooks, R. R. (1969a). Chromium-61transport in the xylem sap of Leptoepermm scoparium (Manuka). N . Z . J1. Sod. 21, 641-646. Lyon, G. L., Peterson, P. J. and Brooks, R. R. (1969b).Chromium-61distribution in tissues and extracts of Leptoepermum ecoparium. Planta 88, 282-287. Lyon, G. L., Peterson, P. J., Brooks, R. R. and Butler, G. W. (1971). Calcium, magnesium and trace elements in a New Zealand serpentine flora. J. Ecol. 59, 421-429. Maas, J. L. and Stuntz, D. E. (1909).Mycoecology on serpentine soil. Mycologia 61, 1106-1110.

368

JOHN PROCTOR

and

STANLEY R. J. WOODELL

Macko, S. and Sarosiek, J. (1961) (Unpublished). RoBlinnoBC: wzg6rz serpentynowych w maaywie Sob6tki no Dolnym S l a s h (rekopis). Madhok, 0.P. (1965). “Magnesium Nutrition of Helianthus annuus L. and H . bolanderi Gray Subspecies exilia Heiser.” Ph.D. Thesis, Univ. of Washington, Seattle. Madhok, 0. P. and Walker, R. B. (1969). Magnesium nutrition of two species of sunflower. P1. Phy&ol., Lancuater 44, 1016-1022. Main, J. L. (1970). “A Demonstration of Genetic Differentiation of Grass Species to Levels of Calcium and Magnesium.” Ph.D. Thesis, Univ. of Washington, Seattle. Maljuga, D. P. (1947). Boden und Pflanzen und ihre Bedeutung beim Aufsuchen der Erzlagerstktten. Priroda 6, 13-17. (In Russian; in Krause, 1958). Maljuga, D. P. (1950). The biogeochemical provinces in the southern Urals. Dokl. Akad. Nauk SSSR 70, 257-259. (In Russian.) Malquori, A. and Cecconi, S. (1956). Minerali argillosi di terreni provenienti da rocce ofiolitiche. Ricerca ocient. 26, 115P1159. Markgraf, F. (1925). Botanische Reiseeindrucke aus Albanien. Repert. Spec. Novarum Regni Veg. Beih. 36, 60-82. Markgraf, F. (1932). Pflanzengeographie von Albanien, ihre Bedeutung fur Vegetation und Flora der Mittelmearlander. Biblthca bot. 105, 1-132. Marrs, R. H. (1974). “The response of serpentine and non-serpentine races of Agrostia stolonijera L. to magnesium and calcium.” B.A. Hons Thesis, Univ. of Stirling, U.K. Marshall, J. K. (1959). “Vegetation of Serpentine of Meikle Kilrannoch.” B.Sc. Horn Thesis, Univ. of St. Andrews, U.K. Martin, M. H. (1968). Conditions affecting the distribution of Mercurialis perennis L. in certain Cambridgeshire woodlands. J . Ecol. 56, 777-793. Martino, E. and Orsino, F. (1969). Flora e vegetazione delle Valli dei Torrenti Acquabona, Scorza e Lerca. Webbia 23, 397-511. Mason, H. L. (1946). The edaphic factor in narrow endemism. 2. The geographic occurrence of plants of highly restricted patterns of distribution. Madrofio 8, 141-257.

Maurer, W. (1966). Flora und Vegetation des Serpentingebietes bei Kirchdorf in Steiermark. Mittg. Zoo Bot. Landmus. Joanneum 24, Graz. 13-76. Mayr, E. (1963). “Animal Species and Evolution.” Harvard University Press. McLean, G. W. (1966). “Retention and Release of Nickel by Clays and Soils.” Ph.D. Thesis, Univ. of California, Berkeley. McMillan, C. (1956). The edaphic restriction of Cupre8surr and Pinus in the coast ranges of central California. Ecol. Monogr. 26, 177-212. Mertz, W. (1967). Biological role of chromium. Fedn Proc. Fedn. Am. Soco exp. Biol. 26, 186-193. Messeri, A. (1936). Ricerce sulla vegatazione dei dintorni di Firenze. 4. La vegetazione delle rocce ofiolitiche di Monte Forrato. Nuovo B. bot. ital. 43, 277-372.

Mikkola, E. (1938). Ultraemaksisten kivilajien vaikutus kasvillisuuteen Lapissa. Luonnon Ystava 42, 1. Millikan, C. R. (1949). Effects on flax of a toxic concentration of boron, iron, molybdenum, aluminium, copper, zinc, manganese, cobalt or nickel in the nutrient solution. Proc. R.SOC.Vict. 61, 25-42. Minguzzi, C. and Vergnano, 0.(1948). I1 contenuto di niche1 nelle ceneri di Alyosum bertolonii Desv. Memorie SOC.l o x . Sci. nat. A 55, 49-74.

THE EJCOLOGY OF SERPENTINE SOILS

359

Minguzzi, C. and Vergnano, 0. (1953). I1 contenuto di elementi inorganici delle plante della formazione ofiolitica dell’Impruneta. Nuovo Q. bot. itd. 60, 287-319.

Mitchell, R. L. (1964). Trace elements in soils. I n “Chemistry of the Soil” (Ed. Bear, F. E.), 2nd ed, pp. 320-368. Reinhold, New York. Mizuno, N. (1968). Interaction between iron and nickel and copper and nickel in various plant species. Nature, Lond. 219, 1271-1272. Morgan, M. A., Jackson, W. A. and Volk, R. J. (1972). Nitrate absorption and assimilation in ryegrass as influenced by calcium and magnesium. PI. Physkl., Lancaster 50, 485-490. Nagano, I., Shibuichi, H. and Tajino, T. (1966). On the serpentine vegetation of Mt. Kamafuse, Chichibu district, central Japan. Bull. Chichibu M w . Nat. Hkt. 11, 7-17. Negodi, G. (1941). Studi sulla vegetazione dell’Appenino Emiliano e della pianura adjacente. 11. La flora e la vegetazione dei serpentini di Varana. Annali. Bot. 22, 117-142. Ngmec, A. (19618). A contribution to the question of stunting growth in pine stands on degraded serpentine soil. L m . Pr. 30, 214-236. (Czech. with English summary.) Ngmec, A. (1951b). A study of a serpentine forest soil with regard to its chromium content and to the resorption of chromium by a stunted pine stand. S b m . 6eakoslov Akad. Zem&d.1951. 395-404. (Czech with English summary.) Ngmec, A. (1954). Wuchsstockungen und Misslingen der Waldkulturen auf Serpentinboden in Sudbohmen infolge der Intoxikationen durch Nickel und Kobalt. Prdce Vdk Ustawti L m . CSR, 6, 7-54. (Czech. with German summary.) Nevole, J. (1926). Flora der Serpentinberge in Steiermark (Oesterreich). Acta SOC.Sci. nat. nwravo-silea. 1, 59-82. Nielsen, F. H. and Sauberlich, H. E. (1970). Evidence of a possible requirement for nickel by the chick. Proc. SOC.exp. Biol. Med. 114, 845-849. Nogina, N. A. (1948). Effect of rocks on podzol formation in the mountainous part of the Central Urals. Trudg pochv. Inst. 28, 105-95. (In Wilson, 1969.) NovBk, F. A. (1926). Vegetwe na hadcovych pudach v zapadnim Srbsku. Veda Pirochni 7 , 269-274; 289-294. NovBk, F. A. (1928). Quelques remarques relatives au problhme de la v6g6tation sur les terrains serpentiniques. Prealia 6, 42-71. NovBk, F. A. (1937). Kretena a vegetace hadcovych pbd. Archiv mazu p r o cochram priorodi a domoviny v zemi Moravskoelezeke 1, 113-160. Otsuka, K. (1963). Sulphur deficiency of serpentine soils. J. Sci. Soil Manure, Tokyo 33, 465-468. (In Japanese. English s u m m a r y in annotated bibliography No. 1220, “Serpentine Soils”, Commonwealth Bureau of Soils, Harpenden, England.) Otsuka, K. and Takahashi, Y. (1963). Causes of growth injury of Satsuma orange trees on serpentine soils. 1. Some observations on affected trees and potculture experiments with barley and wheat. J. Sci.Soil Manure, Tokyo 38, 461-464. (In Japanese. English summary in annotated bibliography No. 1220, “Serpentine Soils”, Commonwealth Bureau of Soils, Harpenden, England.) Page, N. J. (1968). Chemical differences among the serpentine “polymorphe”. Am. Miner. 58, 201-215. Palacios, M. L. (1936). Algunas observaciones geobotanicas en la Serranh de Ronda. Bol. SOC.capan. Hkt. nat. 16, 39-46.

360

JOHN PROCTOR

and

STANLEY R. J. WOODELL

P a m p d , R. (1903). La flora dei serpentini della Toscana. 1. Montignoso. Nuovo B. bot. ital. 19, 464-466. PanEiE, J. (1859). Die Flora der Serpentinberge in Mittel-Sorbien. Verh 2001. -bot. Uea. Wien 9, 139-150. Pavarino, G. L. (1912). Intorno alla flora del serpentino nell’ Appennino Bobbiese. Contribuzione prima. Atti Ist. bot. Univ. Lab. crittogam. Pavia 12, 21-56. Pavarino, G. L. (1914). Intorno alla flora del calcare e del serpentino nell’ Appennino Bobbiese. Contribuzione seconda. Atti Ist. bot. Univ. Lab. crittogam. Pavia 14, 19-42. Pavarino, G. L. (1918). Intorno alla flora del calcare e del serpentino nell’ Appennino Bobbiese. Contribuzione terza. Atti Ist. bot. Univ. Lab. crittogam. Pavia 15, 89-108. PavloVi6, Z. (1953). Prilog poznavanju serpentinske flore Ozren planine kod Sjenice (I).Mw.d’Hist. nut. du Pays Serbe Bull. SBr B 5-6, 3-19. (French summary. ) Pavlovib, Z. (1965). Prilog poznavanju serpentinske flore i vegetacije Ozrena kod Sjenice (11).M w . d’Hkt. nat. du Pay8 Serbe. Bull SBr. B. 7 , 1-46. (French summary.) PavloVi6, Z. (1962). Karakteristioni elementi serpentinske flore Srbije. Qlaan. muz. D d t . Slov. 18, 3-18. Pavlovib, Z. (1964). Borove sume nea serpentinima Srbije. Ulaan. muz. D r d t . Slov. 19. PBdro, G. and Bitar, K. E. (1966a). Contribution 8. 1’6tude de la gonkse des sols hypermagn6sienes: recherches expBrimentales sur l’alteration chimique des rochee ultrabasiques (serpentinites). Annls agron. 17, 611-661. PBdro, G. and Bitar, K. E. (1966b). Sur l’influence du type chimique de la roche mere dans le developpement des phenomdnes d’alteration superficielle. Recherches expBrimentales sup 1’Bvolution des roches ultrabaaiques (serpentinites). C.r. hebd. Sdanc. Acad. Sci.,Paris. SBrie D. 263, 313-316. PBdro, G. and Delmas, A. B. (1971). Sur l’alteration de l’olivine par lessivage It l’eau et la mise en Bvidence de trois grands domaines d’6volution ghochimique. C.r. hebd. Shnc. A d . Sci., Paris, SBrie D, 273, 1543-1546. PeliSek,J. (1939). Bodendes Steppengebietes vonMohelno und kurze Characteristik der westmiihrischen Boden. Arch. Nat. u. Heimatachutz Mahrschlee. Lande Ib, 1-128. Pennell, F. W. (1910). Flora of the Conowingo Barrens of south-eastern Pennsylvania. Proc. Acad. nat. Sci. Philad. 62, 541-584. Pennell, F. W. (1913). Further notes on the flora of the Conowingo or serpentine Barrens of south-eastern Pennsylvania. Proc. Acad. nat. Sci. Philad. 64, 520-539.

Pennell, F. W. (1930). On some critical species of the serpentine Barrens. Bartvnia, 12, 1-23.

Pichi-Sermolli, R. ( 1936). Osservazioni sulk principali morfosi delle piante del serpentino. NWO U . bot. ital. 43, 461-474. Pichi-Sermolli, R. (1948). Flora e vegetazione delle serpentine e delle altre ofioliti dell’alta valle del Tevere (Toscana). Webbia 6, 1-380. Pigott, C. D. and Walters, S. M. (1954). On the interpretation of the discontinuous distributions shown by certain British species of open habitats. J . Ecol. 42, 95-116.

Pinto da Silva, A. R. (1965). Communicaci6n a1 Col6quio “Aportaci6n de las investigaciones Ecologicas y Agricolas a la lucha del mundo contra el

THE ECOLOGY OF SERPENTINE SOILS

361

hambre”. Madrid, 20 a1 26 de Octubre, 1-40. 0 s habitats serpentinicos e o seu racional approveitamento agrario primeiras achegas do caso portugu6s. Pinto da Silva, A. R. (1970). A flora e a vegetaqiio das Breaa ultrabthicaa do nordeste Transmontano. Agr. l w i t . 30, 176-364. (In Portuguese with English summary.) Polunin, N. (1948). Botany of the Canadian Eastern Arctic. III. Vegetation and ecology. Canada, Department of Minea and Reaourcea, National Muaeum Bulletin No. 104. Pratt, P. F. (1966). Chromium. I n “Diagnostic Criteria for Plants and Soils” (Ed Chapman, H. O.), pp. 136-141. Univ. of California, Div. Agr. Sciences. Proctor, J. (1969). “Studies in Serpentine Plant Ecology.” D.Phi1. Thesis, Univ. of Oxford, U.K. Prootor, J. (1970). Magnesium as a toxic element. Nature, Lond. 227, 742-743. Proctor, J. (1971a). The plant ecology of serpentine. II. Plant response to serpentine soils. J. Ecol. 69, 397410. Proctor, J. (1971b). The plant ecology of serpentine. 111. The influence of a high Mg/Ca ratio and high nickel and chromium levels in some British and Swedish serpentine soils. J. Ecol. 59, 827-842. Proctor, J. ( 1 9 7 1 ~ )(Unpublished). The distribution of Pkcntago erecta Morris and Avena jatua L. on a serpentine and adjacent soil. Proctor, J. (1972) (Unpublished). Studies on available nickel in serpentine soils. Proctor, J. and Whitten, K. (1971). A population of the valley pocket gopher (Thomomya bottae) on a serpentine soil. Am. Midl. Nat. 78, 176-179. Proctor, J. and Woodell, S. R. J. (1971). The plant ecology of serpentine. I. Serpentine vegetation of England and Scotland. J. Ecol. 59, 375-398. Radford, A. E. (1948). The vascular flora of the olivine deposits of North Carolina and Georgia. J . Eliaha Mitchell acient. SOC.64, 46-106. Ragg, J. M. and Ball, D. T. (1964). Soils of the ultrabasic rocks of the island of Rhum. J. Soil Sci. 15, 124-133. Rai, D., Simonson, G. H. and Youngberg, C. T. (1970). “Serpentine Derived Soils in Watershed and Forest Management”. Dept. of Soils, Oregon State University, Corvallis, Oregon, U.S.A. Raven, P. H. (1964). Catastrophic selection and edaphic endemism. Evolution, Lancaater, P a . 18, 336-338. Raymond, M. (1950). Esquisse phytog6ographique du Quebec. Mkm. Jard. bot. Montrhl. 5 , 1-147. Richardson, J. A. and Greenwood, E. F. (1967). Soil moisture tension in relation to plant colonisation of pit heaps. Proc. Univ. Newcastle on Tyne, phil. 80C. 1, 129-136. Rigotti, H. (1930). Significato fitogeographico delle florule dei serpentini submontani in Piemonte. Atti Congr. geogr. ital. 11, 72-74. Ritter-StudniEka, H. (1966). Beitrag zur Okologie der Serpentinflora in Bosnien. Vegetatio 7 , 89-98. Ritter-StudniEka, H. (1963). Biljni pokrov na serpentinimau Bosne. Godidinjak biol. Inat. Saraj. 16, 91-204. Ritter-StudniEka, H. (1964). Anatomske razlike izmedju biljaka sa serpentinske, dolomitne i krecnjacke podloga. BodGnjak biol. Inat. Saraj. 17, 161-197. Ritter-StudniEka, H. (1967). Uber die Glaukeszenz am Serpentinpflanzen. oat. bot. 2. 114, 101-114. Ritter-StudniEka, H. (1968). Die Serpentinomorphosen der Flora Bosniens. Bot: Jb. 88, 443-465.

362

JOHN PROCTOR

and

STANLEY R. J. WOODELL

Ritter-StudniEka, H. (1970). Die Flora der serpentinvorkommen in Bosnien. Biblthers bot. 130, 1-100. Ritter-StudniEka, H. (1972). Die erhohte Sukkulenz bei Serpentinpflanzen. Phyton (Austria) 14, 239-249. Ritter-StudniEka, H.and Dursun-Grom, K. (1973). ffber den Eisen-, Nickelund Chromgehalt in einigen Serpentinpflanzen Bosniens. Oat. bot. 2. 121, 29-49. Ritter-StudniEka, H.and Klement, 0. (1968). On lichen species and their associatiom on serpentine in Bosnia. oat. bot. 2. 115, 93-99. Rivas-&day, S. (1969). Flora serpentinicola espaiiola. Nota primeira (Edafismos endemicos del Reino de Granada). Am. R. Acad. Fam. Madr. 35, 297-304. Robinson, W. O., Edgington, G. and Byers, H. G. (1935). Chemical studies of infertile soils derived from rocks high in magnesium and generally high in chromium and nickel. Tech. Bull. U.S. Dep. Agric. No. 471. Rune, 0. (1963). Plant life on serpentines and related rocks in the north of Sweden. Acta phytogeogr. auec. 31, 1-139. Rune, 0. (1964). Notes on the flora of the Gasp6 Peninsula. Svenak bot. Tidakr. 48, 117-136. Russell, E. W. (1954). The availability of sorbed or fixed phosphates to plants. Trana. 5th Int. Cmgr. Soil Sci. 2, 308-311. Sadeback, R. ( 1887). uber die generationsweise fortgesetzten Aussaaten und Kulturen der Serpentinformen der Farngattung Aaplenium. Ber. Gea. Bot. Hamburg 3, 74-79. (In Krause, 1958.) Sarosiek, J. (1964). Ecological analysis of some plants growing on serpentine soil in Lower Silesia. Monographiae bot. 18, 1-105. (In Polish with English summary.)

Sarosiek, J. and Sadowska, A. (1961). Ekologia r6slin gleb serpentynowych. W i d . bot. 5 , 73-86. Sasaki, S., Matsuno, T. and Kondo, Y . (1968). A podzol derived from serpentine rocks in Hokkaido, Japan. Soil Sci. PI. Nutr. 14, 99-109. Scharrer, K. and Schropp, W. (1933). Sand- und Wasserkulturversuohe mit Nickel und Kobalt. 2. PjlErnahr. Diing. Bodenk. 31, 9 6 1 1 3 . Scharrer, K. and Schropp, W. (1935). Die Wirkungen von Chromi- und ChromatIon auf Kulturpflanzen. 2. PjlErnahr, Diing. Bodenk. 37, 137-149. Schellman, W. (1964). Uber die lateritische Verwitterung von Serpentinit. Ueol. Jb. 81, 645-678. Schwarz, K. and Mertz, W. (1961). A physiological role of chromium (111)in glucose utilisation (glucose tolerance factor). Fedn Proc. Fedn Am. Soca exp. Biol. 20, (suppl. lo), 111. Scoggan, H. J. (1960). The floraof Bic and the Gasp6 Peninsula. Bull. natn. Mua. Can. 115, 1-399. Sedova, P. (1968). Sanitary assessment of soil irrigation with sewage containing chromium. Gig. Sanit. 23 (No. 6), 15-19. (In Russian. English summary in annotated bibliography No. 1566, “Chromium in Soils and in Plant Nutrition”, Commonwealth Bureau of Soils, Harpenden, England.) Seifriz, W. (1940). Die Pflanzengeographie von Cuba. Bot. Jb. 7 0 , 441-462. Seifriz, W. (1943). The plant life of Cuba. Ecol. Monogr. 13, 375-426. de Sequeira, E. M. (1969). Toxicity and movement of heavy metals in serpentinic soils (north-eastern Portugal). Agronomia h i t . 30, 116-164. Severne, B. C. and Brooks, R. R. (1972). A nickel-accumulating plant from Western Australia. Planta 103, 91-94.

a.

TBE ECOLOGY OF SERPENTINE SOILS

363

Shkoljnik, M. J. and Smirnov, U.S. (1970). On possible cause8 of serpentine morphoses and morphological variations in planta i n d u d by high concentration of boron. Bot. Zh. SSSR.55, 1764-1782. (In Russian with English summary.) Shreve, F. (1910). The ecological plant geography of Maryland, Midland Zone; Lower Midland District. I n “The Plant Life of Maryland” (Shreve, F., Chrysler, M. A., Blodgett, F. H. and Beasley, F. W.), pp. 199-219. Johns Hopkins Univ. Press, Baltimore, Maryland, U.S.A. Soane, B. D. and Saunder, D. H. (1959). Nickel and chromium toxicity of serpentine soils in Southern Rhodesia. Soil Sci. 88, 322-330. SoEava, V. (1927). Botanische Beschreibung der Wiilder des Polar-Urals vom Flusse Nelka zum Flusse Hulga. Trav. Mw.bot. Acad. Sci. Ruse. 21, 1-78. Sob, R. von (1934). Vasmegyi szociol6giae 6s florisztiki niiv6nyfoldrejz&hoz. Vaai Szemli, F o l k Sabariemia (Dudntzili Szernle) 1, 106-134. Soon, Y. K. (1971). “Chemical Factors Affecting Plant Growth on Serpentine Soils, Especially Calcium Supply.” M.Sc. Thesis, Univ. of Aberdeen, U.K. Spence, D. H. N. (1957). Studies on the vegetation of Shetland. I. The serpentine debris vegetation in Unst. J. Ecol. 45, 917-946. Spence, D. H. N. (1968). The flora of Unst, Shetland in relation to the geology. Tram. Proc. bot. SOC.Edinb. 37, 163-173. Spence, D. H. N. (1969). Studies on the vegetation of Shetland. 11.Reasom for the restriction of the exclusive pioneers to serpentine debris. J. Eool. 47, 641-649.

Spence, D. H. N. (1969) (Unpublished,University Library, St. Andrewa). Coverabundance lists for related community types on serpentine and other rmka in Scotland. Spence, D. H. N. (1970). Scottish serpentine vegetation. Oikoe 21, 22-31. Spence, D. H. N. and Millar, E. A. (1963). An experimental study of the infertility of Shetland serpentine soil. J. Ecol. 51, 333-343. Stebbins, G. L. (1942). The genetic approach to problems of rare and endemic species. Madrofio 6, 241-272. Stebbins, R. C. (1949). Speciation in salamanders of the plethodontid genus Enaatina. Univ. Calq. PubL 2001. 48, 377-526. Sulej, J., Slesak, E., Leonowicz-Babiak, K. and Buczek, J. (1970). Tentative explanation of dwarfish growth of plants on serpentine soils. I. Phpicochemical and biological properties, and mineral elements of serpentine soils. Actu bot. SOC.Pol. 39, 405-419. (In Polish with English summary.) Suza, J. (1921). Xerothermni kvetena podklad&serpentinovfch na, d o h toku Jihlavky. 6m.mrav. z e m k . Mw.20, 1-32. Sum, J. (1928). Guide g6obotanique pour le terrain serpentineux p r h de Mohelno dans la Moravie du sud-ouest (TchBcoslovaquie).Rozpr. &kd A M . V6d Urnkni 37, 1-116. Sum, J. (1930). SrovnAvicf studie o liZiejnikov6 floh serpentin& (Mohelno, Gurhof a Krabath). Vergleichende Studien iiber die Flechtenflore der Serpentine. Sb. Wr.Spol. Mor. Oetrava. 6 , 231-266. Svenonius, F. (1883). Om olivinstens och serpentinforekomster i Norrland. Geol. Forr. Stockh. Forh, 6 , 342-369. Swaine, D. J. (1955). The trace-element content of soils. Tech C o m u n . Cornnzonw. Bur. SoiL. No. 48. Tadros, T. M. (1957). Evidence of the presence of an edapho-biotic factor in the problem of serpentine tolerance. Ecology 38, 14-23.

364

JOHN PROCTOR

and

STANLEY R. J. WOODELL

Taniguti, M. (1968). Phytosociological study of serpentine area of Mt. Asama, Mie Pref. Acta phytohz. geobot, Kyoto,l7,122-127. (In Japanesewith English s-v.) Tmsoules, J. A. (1970). “Comparative Studies of the Composition of Magnesium Soils with Particular Reference to the Heavy Metal (Cr, Fe, Mn, Ni) Status.” M.Sc. Thesis, Univ. of Aberdeen, U.K. Tesic, Z., Ristanovic, B. and Ritter-StudniEka, H. (1967). Contribution to the study of microflora on dolomite and serpentine habitats under Merent stages of vegetative succession in Bosnia and Herzegovina. Acta bwhgiccc I~g08h&a, Series B. Mikrobwhgija, 4, 1-19. (In Yugoslav.) Thoday, J. M. (1959). Effects of disruptive selection. Heredity, Lond. 13, 187-203. Thoday, J. M. and Boam, T. B. (1969). Effects of disruptive selection. 11. Polymorphism and divergencewithout isolation. Heredity, Lond.13, 206-218. Thoday, J. M. and Gibson, J. B. (1962). Isolation by disruptive selection. Nature, Lond. 193, 1164-1166. Timperley, M. H., Brooks, R. R. and Peterson, P. J. (1970). The signifkance of eeeential and non-essential trace elements in plants in relation to biogeochemical prospecting. J. appl. Ecol. 7, 429-439. Timperley, M. H., Brooks, R. R. and Peterson, P. J. (1972a). The improved detection of geochemical soil anomalies by multiple regression analysis of biogeochemical data. Proc. Auatralas. Inst. Min. Metall. 242, 26-36. Timperley, M. H., Brooks, R. R. and Peterson, P. 3. (1972b). Trend analysis aa an aid to the comparison and interpretation of biogeochemical and geochemical data. Ecm. Ueol. 67, 669-676. Trelease, S. F. and Treleese, H. M. (1931). Magnesium injury of wheat. BUZZ. Towey bot. Club. 58, 127-148. Turrill, W. B. (1929). “The Plant Life of the Balkan Peninsula. A Phytogeographical Study.” Clarendon Press, Oxford, England. Usov, N. I. (1937). Effect of absorbed magnesium on saline properties of soil. Prm. Cmf. Soil. Sci., S a r a h 1, 44-61. (In Chemical Abstracts 33, 8342-9; 1939.) Vanselow, A. P. (1951). Microelement research with citrus. Calif. Citogr. 37,77-80. Vanselow, A. P. (196th). Cobalt. I n “Diagnostic Criteria for Plants and Soils” (Ed. Chapman, H. D.), pp. 142-165. Univ. of California, Div. Agr. Sciences. Vanselow, A. P. (1966b). Nickel. In “Diagnostic Criteria for Plants and Soils” (Ed. Chapman, H. D.), pp. 302-309. Univ. of California, Div. Agr. Sciences. Veniale, F. and Van der Marel, H. W. (1963). An interstratified saponite-swelling chlorite mineral at3 a weathering product of lizardite rock from St Margherita Staffora (Pavia Province) Italy. Beitr. Miner. Petrogr. 9, 198-246. Vergnano, 0. (1963). L’azione fisiologica del niche1 sulle plante di un terreno serpentinoso. Nuovo a. bot. ital. 60, 109-183. Vergnano, 0. (1958). Il contenuto di elementi inorganici delle piante della formazione ofiolitica dell’Impruneta (Firenze). Nichelio, Cromo e Cobalto. Nuovo U. bot. ital. 65, 133-162. Vergnano, 0. (1969a). La nutrizione fosforia in presenza di elementi micronutritivi e tossic nel terreno. Agrochimh 3, 262-269. Vergnano, 0. (1969b). Metabolism0 minerale di piante coltivate su terreni agrari di origine ofiolitica nei Monti Rognosi (Alta Val Tiberina). Nuovo U. bot. ital. 66, 100-160. Vergnano, 0. and Hunter, J. G. (1963). Nickel and cobalt toxicities in oat plants. Ann. Bot. 17, 317-328.

THE ECOLOGY OF SERPENTINE SOILS

366

Vinogradov, A. P. (1938). Geochemistry and biochemistry. Adv. C h m . MOBCOW, 7, 646-686. (In Swaine, 1956.) Vlamis, J. (1949). Growth of lettuce and barley as influenced by degree of calcium saturation of soil. Soil Sci. 07, 453-466. Vlamis, J. and Jenny, H. (1948). Calcium deficiency in serpentine soils aa revealed by absorbent technique. Science, N . Y. 107, 549-651. Walker, R. B. (194th). “A Study of Serpentine Soil Infertility with Special Reference to Edaphic Endemism.” Ph.D. Thesis, Univ. of California, Berkeley, California. Walker, R. B. (1948b). Molybdenum deficiency in serpentine barren soils. Science, N.Y. 108, 473-475. Walker, R. B. (1954). Factors affecting plant growth on serpentine soils. Ecology

as,

269-266.

Walker, R. B., Walker, H. M. and Ashworth, P. R. (1955). Calcium-magnesium nutrition with special reference to serpentine soils. P1. P h y k l . , Lancaeter.

ao, 214-221. Wallace, A., Frolich, E. and Lunt, 0. R. (1966). Calcium requirements of higher plants. Nature, Lond. 209, 634. West, W. (1912). Notes on the flora of Shetland, with some ecological observations. J. Bot., Lond. SO, 265-275, 297-306. Whatley, J. M. (1965). “Comparative Studies on Some Components of Serpentine and Non-serpentine Vegetation.” Ph.D. Thesis, Univ. of California, Berkeley, California. Wherry, E. T. (1963). Some Pennsylvania barrens and their flora. 1. Serpentine. Bartonia, 38, 7-1 1. Whita, C. D. (1987). Absence of nodule formation on Ceanothua cunedw in serpentine soils. Nature, Lo&. 215, 875. White, C. D. (1971). “Vegetation-soil chemistry correlations in serpentine ecosystems.” Ph.D. Thesis, Univ. of Oregon, Eugene, Oregon. Whittaker, E. J. W. and Zussman, J. (1956). The characterization of serpentine minerals b y X-ray diffraction. Mineral Mag., 31, 107-126. Whittaker, R. H. (1954a). The ecology of serpentine soils. Ecology 85, 258-259. Whittaker, R. H. (1954b). The vegetational response to serpentine soils. Ecology 85, 275-288.

Whittaker, R. H. (1960). Vegetation of the Siskiyou Mountains, Oregon and California. Ecol. Monogr. 80, 279-338. Wild, H. (1965). The flora of the Great Dyke of Southern Rhodesia with special reference to the serpentine soils. Kirkia 5 , 49-86. Wild, H. (1970). Geobotanical anomalies in Rhodesia. 111. The vegetation of nickel bearing soils. Kirkia 7 , 1-62. Wild, H. (1973) (Unpublished). The evolutionary effects of heavy metals and other abnormally mineralized soils in South Central Africa. Wild, H. (1974a) Variations in the serpentine floras of Rhodesia. Kirkiu 9, 209232.

Wild, H. (1974b). Indigenous plants and chromium in Rhodesia. Kirkiu 9, 233241.

Wildman, W. E. (1967). “Serpentinite Weathering and Clay Mineral Formation in Some California Soils.” Ph.D. Thesis, Univ. of California, Davis, California. Wildman, W. E., Jackson, M. L. and Whittig, L. D. (1968a). Serpentine rock dissolution as a function of carbon dioxide pressure in aqueoue solution. Am. Miner. 53, 1252-1263.

366

JOHN PROCTOR

and

STANLEY R. J. WOODELL

Wildman, W. E., Jackson, M. L. and Whittig, L. D. (1968b). Iron-rich montmorillonite formation in soils derived from serpentinite. Proc. Soil Sci. SOC. Am. 32, 787-794. Wilkins, D. A. (1957). A technique for the measurement of lead tolerance in plants. Nature, Lond. 180, 37-38. Williams, P. C. (1967). Nickel, iron and manganese in the metabolism of the oat plant. Nature, Lond. 914, 628. Wilson, A. D. (1969). “Psdological Studies on Soils Derived from Serpentinite in North-east Scotland.” M.Sc. Thesis, Univ. of Aberdeen, U.K. Wilson, S. B. and Nicholas, D. J. D. (1967). A cobalt requirement for nonnodulated legumes and for wheat. Phytochemistry, 6 , 1057-1066. Wiltahire, a. H. (1972). Effect of nitrogen source on translocation of nickel in some crop plants and weeds. Kirkia 8, 103-123. Woodell, S. R. J. and Newton, L. E. (1974) (Unpublished). The serpentine outcrop at h u m (Volta Region) in Ghana. Woodell, S. R. J., Lewis, H. and Mooney, H. A. (1974a) (Unpublished). A serpentine-maintained polymorphism in Linanthua androsaceua (Polemoniaceae) in California. Woodell, S. R. J., Mooney, H. A. and Lewis, H. (1974b) (Unpublished). The behaviour of the annual species, Linanthua androsacew, (Polemoniaceae), on and off serpentine in California. Wyllie, P. J. (ed.) (1967). “Ultramafic and related rocks.” J. Wiley and Sons, New York. Wyn Jones, R. G. and Lunt, 0. R. (1967). The function of calcium in plants. Bot. Rev. 33, 407-426. Yamagata, N. and Murakami, Y. (1958). A cobalt-accumulator plant, Clethra barbinemis Sieb. et Zucc. Nature, Lond. 181, 1808-1809. Yamanaka, T. (1952). Studies in the vegetation and the flora on serpentine. Rea. Rep. Kochi Univ. 1, 1-8. Yamanaka, T. (1954). Sociological studies on the serpentine vegetation. III. The vegetation on Mt. Shiraga, Kochi Prefecture. Bull. Fac. Educ., Kochi Univ. 5, 47-63. (In Japanese with English Summary.) Yamanaka, T. (1959). Serpentine flora of Mt. Higashi-Akaishi, Shikoku, Japan. A c b phytotax geobot., Kyoto 18, 80-97. Yamanaka, T. (1967). The serpentine vegetation of Tanoura, Kunamote prefecture, Kyushu. Acta phytotax. geobot., Kyoto 22, 192-194. Yatazawa, M. and Tanaka, H. (1965). Levels of available micronutrient metals in some Japanese soils as measured by Aspergillw, niger with special reference to their threshold values. Soil Sci. PI. Nutr. 11, 167-163. Zlatnik, A. (1928a). gtudes ecologiques et sociologiques sur le Sederia caerulea e t le Seslerion calcariae en TchBcoslovaquie. Trav. SOC.r q . Sci. Boheme, Cl. Sci. 8. Zlatnik, A. (1928b). Les associations vegetales et les sols du terrain serpentineux prAs de Mohelno. Bull. int. Acud. tchkque Sci. 29, 333-337. Zollitsch, L. (1927). Zur Frage der Bodenstetigkeit alpiner Pflanzen unter besonderer Beriicksichtigung des Aziditats und Konkurrenzfaktors. Flora, J e w 122, 93-168. Zolyomi, B. (1936). tfbersicht der Felsenvegetation in der pannonischen Florenprovinz und dem nordwestlich angrenzenden Gebiete. Nus. Natl. Hungar. Ann. 30, 136-174.

Author Index Numbers in italics refer to the pages on which references are listed in bibliographies a t the end of each article A Abelson, P. H., 296, 350 Adamovic, L., 263, 350 Adiga, P. R., 296, 350 Albert, R., 336, 350 Aldous, E., 296, 350 Allen, D. L., 6, 122 Allen, J. C., 24, 120 Allison, R. V . , 261, 262, 272, 351 Amidei, G., 263, 264, 350 Antonovics, J., 340, 341, 347, 348, 350, 352 h u m , S., 260, 355 Ashworth, P. R., 316, 327, 328, 332, 365 Aumento, F., 258, 350 Avery, S. L., 74,123 B Bailey, V. A., 20, 21, 120 Baldwin, N. S., 4, 120 Ball, D. T . , 262, 279, 298, 309, 361 Banks, C. J., 39, 120 Barbour, M. G., 181, 182, 216, 232,251 Bard, Y., 145, 162 Bargoni, I., 264, 346, 350 Barker, J. E., 176, 243 Barnard, G . H., 141, 162 Barrs, H. D., 173, 175, 176, 177, 179, 180, 183, 188, 191, 196, 243 Bartholomew, E. T., 227, 243 Bartholic, J. F., 182, 203, 223, 250 Bartlett, M. S., 16, 16, 120 Bavel, C. H. M. van, 211, 214, 220, 222, 224, 228, 232, 233, 239, 246, 247 Bayes, T., 140, 162 Beadle, N. C . W., 346, 351 Beard, J. S . , 264, 351 Beoherer, A., 264, 351

Beck von Mannagetta, G., 263, 351 Beckett, P. H. T., 316, 351 Beger, H. K. E., 264, 351 Begg, J. E., 170, 191, 197, 201, 206, 221, 237, 243, 253 Benis, M., 171, 237, 253 Bennett, H. H., 261, 262, 272, 351 Bernstein, L., 231, 243 Betta, M. W., 264, 351 Bierhuizen, J. F., 190, 252 Billings, W. D., 270, 351 Bingham, F. T . , 233, 246 Birrell, K. S., 262, 264, 266, 274, 278, 286, 297, 301, 317, 351 Biscoe, P. V., 195, 244 Bitar, K. E., 259, 260, 360 Bixby, J. A., 232, 244 Bjorkman, O., 203, 226, 250 Bjorlykke, B., 264, 351 Blackshaw, G. N., 274, 332, 351 Blackwell, J., 175, 176, 177, 179, 180, 188, 191, 196, 243 Blau, G. E., 138, 146,162 Bleeker, P., 262, 354 Bliss, L. C . , 216, 249 Blum, A., 176, 186, 244 Boam, T . B., 347, 364 Bogatyrev, K. P., 260, 262, 273, 351 Bordovsky, D. G., 232, 233, 244 Box, G. E. P., 139, 148,162 Boydell, H. C., 263, 351 Boyer, J. S., 168, 170, 171, 173, 174,

175, 176, 177, 179, 184, 186, 188, 189, 190, 194, 195, 201, 234, 237, 244, 250 Bradshaw, A. D., 276, 293, 296, 340, 341, 350, 351, 354 Bradstreet, E. D., 167, 168, 170, 173, 183, 184, 185, 186, 189, 196, 216, 217, 236, 236, 251 367

186, 224, 316, 171, 193,

368

AUTHOR INDEX

Brandle, J. R., 226, 244 Braun, E. L., 264, 351 Braun-Blanquet, J., 264, 266, 351 Breeze, V. G., 300, 352 Brenchley, W. E., 283, 308, 352 Bridgman, W. B., 236, 250 Brix, H., 225, 233, 244 Brooks, J. L., 4, 120 Brooks, R. R., 264, 269, 282, 284, 286, 288, 294, 301, 305, 307, 310, 332, 344, 357, 362, 364 Brown, C. L., 167, 173, 234, 236, 254 Brown, G. N., 226, 232, 244 Brown, R. W., 172, 183, 254 Brown, S. M., 262, 355 Brownine, V. D., 223, 224, 249, 250 Broyer, T. C., 216, 245 Bruckeroff, D. N., 223, 247 Bryan, J. E., 8 6 , 1 2 0 Buckner, C. H., 113, 120 Buczek, J., 316, 328, 332, 335, 352, 363 Burnett, T., 37, 38, 105, 121 Burns, A. F., 315, 352 Buschbom, U., 190, 204, 209, 216, 221, 226, 239, 248, 252 Butler, G. W., 264, 269, 284, 286, 301, 306, 307, 310, 332, 357 Butler, J. R., 260, 262, 352 Byers, H. G., 261, 272, 273, 276, 276, 279, 287, 298, 303, 320, 325, 330, 362

C Caesalpino, A., 263, 352 Caldwell, M. M., 216, 222, 245, 250 Campbell, C. J., 182, 186, 216, 244 Campbell, G. S., 172, 177, 183, 244, 254 Campbell, M. D., 177, 244 Canterford, R. L., 226, 249 Cambia, J. P., 264, 352 Cam, M. K. V., 203, 219, 220, 244 Gary, J. W., 172, 174, 177, 180, 183, 202, 209, 244, 254 Catsky, J., 191, 244 Ceccato, R. D., 170, 176, 176, 177, 179, 180, 183, 186, 188, 191, 196, 227, 243, 249 Cecconi, S., 261, 358 Chadwick, M. J., 276, 316, 351 Chambers, J. L., 223, 247

Chant, D. A., 3 8 , 1 2 3 Clark, L. R., 36, 121 Clark, R. N., 219, 233, 244 Clarke, B. C., 62, 63, 121 Clarkson, D. T., 316, 352 Clausen, J., 341, 347, 352 Cleary, B. D., 168, 170, 171, 173, 174, 184, 186, 197, 203, 206, 207, 209, 213, 214, 216, 221, 224, 228, 232, 233, 244, 245, 263 Cobb, F. W., 229, 247 Cockayne, L., 268, 352 Connell, J. H., 6, 6, 7, 8, 117, 121 Cook, L. M., 62, 70, 71, 123 Coombe, D. E., 262, 264, 352 Cooper, W. E., 4 , 1 2 1 Cotton, M., 283, 352 Cowan, I. R., 166, 200, 201, 245 Cox, D. R., 44, 61, 102,121, 148,162 Crooke, W. M., 283, 294, 295, 352

D Dahl, O., 263, 352 Dainty, J., 169, 171, 201, 237,245,253 Daniker, A. U., 264, 352 Daum, C. R., 236, 245 Davenport, D. C., 233, 245 Davenport, T. L., 227, 248 Davidson, J., 264, 352 Davy, Sir Humphry, 331, 352 Dawkins, M., 8 7 , 1 2 1 Delmas, A. B., 260, 360 Del Moral, R., 270, 352 De Puit, E. J., 216, 245 De Roo, H., 171, 173, 192, 206, 253 De Roo, H. C., 170, 176, 177, 179, 229, 231, 232, 239, 245 Detling, J. K., 176, 177, 180, 216, 245 Dickinson, H., 347, 352 Dichon, R. E., 216, 245 Dimond, A. E., 201, 229, 230, 245, 253 Dine, S. J., 216, 218, 226, 245 Dinger, B. E., 219, 226, 251 Dixon, A. F. G., 39, 47, 121 Dixon, H. H., 167, 234, 245 Dobbs, R.C., 224, 246 Dodd, J. D., 186, 199, 203, 204, 206, 209, 210, 221, 239, 246 Dodson, S . I., 4, 120 Doley, D., 187, 224, 246 Dore, W. H., 262, 355

AUTHOR INDEX

Draper, N. R., 138, 142, 144, 145, 148, 159,162

Dimiway, J.M., 170,171,175,177,179, 186, 188, 219,220, 229, 246 Dupras, E. F., Jr, 162, 162 D u r n - G r o m , K., 337, 362 Dunigneaud, P., 263, 274, 317, 346, 352 Dvorak, R., 263, 347, 353

E Emtin, J. D., 176, 186, 244 Ebling, F. J., 5, 122 Eckard, A. N., 232, 248 Edgington, G., 261, 272, 273, 275, 276, 279, 287, 298, 303, 320, 325, 330,

362 Eggler, J., 263, 353 Ehrlich, P. R., 338, 355 Elfving, D. C . , 201, 202, 203, 211, 212, 221, 222, 227, 233, 236, 239,246

Ellis, B. S., 272, 353 Elton, R. A., 71, 121 Emmingham, W. H., 213, 214, 228, 254 Epstein, E., 330, 333, 335, 353 Emst, W., 277, 284, 293, 297, 301, 309, 340, 353 E r s t , W. G., 259, 353 Evenari, M., 190, 204, 209, 216, 221, 226, 239, 248, 252 F Fahey, J. J., 256, 258, 300, 353 Faris, D. G., 344, 353 Faust, G. T . , 266, 258, 300, 353 Feeney, P., 8, 125 Fenn, L. B., 233, 246 Fernald, L. M., 264, 353 Fernandez, T. C., 278, 353 Ferreira, R. E. C., 264, 274, 275, 315, 331, 353 Finn, B. J., 283, 294, 354 Fiori, A., 264, 353 Fisher, F. S . , 149, 150, 160, 163 Fisher, H. D., 174, 177, 180, 244 Fisher, M. A,, 233, 245 Fleschner, C. A., 39, 121 Forde, M. B., 344, 353 Fox, L. R., 36, 113, 121 Frank, A, B., 177, 191, 192, 246

369

Frasche, D. F., 261, 262, 272, 353 Freeman, B., 176, 176, 177, 179, 180, 188, 191, 196, 243 Frolich, E., 330, 365 Frost, L. C., 262, 264, 352 G Gaff, D. F., 176, 177, 178, 179, 185, 186, 188, 254 Garcia, C. R., 331, 357 Gardner, W. R., 172, 183, 204,231, 238, 243, 246, 254 Gates, D. M., 186, 246 Gauckler, K., 263, 353 Gawe, G. F., 11,121 cedroite, K. K., 273, 353 Gee, G. W., 177, 231, 246 Ghorashy, 5. R., 171, 175, 177, 179, 186, 186, 244 Gibson, J . B., 347, 364 GiiTord, H. A., 173, 189, 193, 246 Gilpin, M. E., 16, 32, 121 Gismondi, A., 264, 354 Goldschmidt, V . M., 300, 354 Goode, J. E., 173, 186, 186, 195, 197, 203, 221, 232, 246 Gordon, A., 273, 276, 337, 354 Gower, J. C., 16, 18, 19,122 Gradwell, G. R., 6 , 11, 12, 124 Graham, R. D., 234,246 Graniti, A., 171, 253 Grant-Mackie, J. A., 264, 352 Greenwood, E. F., 273, 361 Greenwood, J. J. D., 71, 121 Gregory, R. P. G., 293, 295, 354 Griffin, J. R., 209, 216, 218, 246, 269, 274, 276, 317, 344, 354 Griffithe, K. J., 37, 114, 121 Grover, R., 316, 331, 332, 333, 354

H Hamtjens, H. A., 262, 354 Haas, R. H., 186, 199, 203, 204, 206, 209, 210, 221, 239, 246 Hagan, R. M., 233, 245 Hailey, J. L., 211, 232, 246 Halbwechs, G., 170, 171, 184, 230, 246, 251 Halevy, A. H., 204, 232, 233, 247 Hall, A. E., 201, 202, 203, 211, 212, 222, 233, 236, 239, 246

370

AUTHOR INDEX

Hall, D. J., 4, 121 Halliday, C., 264, 315, 354 Halstead, R. L., 278, 283, 291, 294, 296, 354 Halvomon, W. L., 209, 216, 247 Hamme1,H. T., 167, 168, 170, 171, 173, 183, 184, 185, 186, 189, 193, 194, 195, 196, 216, 217, 233, 236, 236,

247, 251, 253 Hanan, J. J., 211, 232, 247 Harada, M., 260, 354 Harcourt, D. G., 11,122 Harris, D. G., 177, 191, 192, 246 Harris, W. F., 219, 226, 251 Harrison, A. T . , 203, 226, 250 Harrison, R. D., 300, 354 Harehberger, J. W., 264, 354 Heselhoff, E., 283, 354 Hassell, M. P., 11,21,32,39,54,55, 113, 114, 118, 119, 122 Havis, J. R., 233, 247 Havranek, W., 233, 247 Hayek, A., 263, 354 Heath, 0. V . S., 218, 247 Heikens, H. S . , 85, 87, 89, 90, 95, 96, 104,123 Hellkvist, J., 176, 185, 186, 189, 194, 195, 198, 220, 231, 232, 235, 237, 247 Helms, J. A., 226, 229, 247 Hemmingsen,E.A., 167, 168, 171, 183, 184, 186, 186, 189, 193, 216, 217, 235, 236, 251 Hendrickson, A. H., 238, 253 Heslop-Harrison, J., 340, 347, 354 Hewitt, E. J., 300, 307, 354 Hickman, J. C., 205, 207, 208, 209,

236,

173, 196,

228,

247 Hiesey, W. M., 341, 347, 352 Higgs, K. H., 186, 195, 221, 246 Hiler, E. A., 211, 214, 219, 220, 222, 224, 228, 232, 233, 239, 244, 246, 247 Hilgard, E. W., 314, 354 Hill, W. J., 139, 162 Hinckley, T. M., 169, 171, 173, 184, 189, 193, 197, 198, 199, 204, 205, 208, 219, 220, 221, 222, 223, 226, 232, 233, 235, 239,244, 247, 251 Hiroe, M., 264, 355 Hobson, E. S., 85,122

Hodges, J. D., 181, 182, 229, 248 Hodgson, R. W., 227, 248 Holdridge, L. R., 264, 354 Hotz, P. E., 261, 262, 354 Holling, C. S., 30, 33, 34, 36, 37, 38, 85, 93, 94, 95, 98, 104, 105, 106, 109, 110, 114, 115, 116, 121, 122 Holmes, R. M., 238, 248 Holmgren, P., 185, 248 Holzner, W., 170, 184, 251 Hong, S. G., 186, 192, 248 Hossian, M. M., 214, 220, 222, 224, 228, 232, 233, 239, 247 Howard-Williams, C., 344, 354 Howell, T. A., 232, 233, 244 Hoyos de Castro, A., 260, 354 Hsiang, T., 139, 162 Hsiao, T. C . , 167, 224, 248 Huber, B., 235, 248 Huck, M. G., 231, 232, 248 Huffaker, C. B., 9, 10, 122 Hunter, J. G., 264, 276, 278, 282, 283, 284, 291, 294, 295, 296, 305, 307, 308, 352, 354, 355, 364

I Iberg, R., 261, 355 Igoshina, K. N., 264, 355 Iljinski, A. P., 264, 355 Inkson, R. H. E., 283, 294, 295, 352 Ishimoto, T. T., 274, 355 Ivlev, V . S . , 95, 122 J Jackson, M. L., 260, 262, 365, 366 Jackson, W. A., 335, 359 Jacob, A., 332, 355 James, B. E., 177, 231, 246 Jarvis, M. S . , 185, 186, 189, 248 Jarvis, P. G., 176, 185, 186, 189, 191, 194, 195, 198, 220, 231, 232, 235, 236, 237, 247, 248 Jeffree, C. E., 191, 248 Jeffrey, D. W., 275, 355 Jenkins, G. M., 141, 162 Jenkinson, L. J., 269, 316, 355 Jenny, H., 264, 316, 351, 365 Johnson, C. M., 274, 337, 355 Johnson, M. P., 338, 355 Johnson, N. E., 173, 192, 193, 248 Johnson, R. P, C., 191, 248

AUTHOR INDEX

Johnston, W. R., 316, 324, 337, 355 Joly, J., 167, 245 Jones, J. R., 199, 204, 221, 248 Jordan, P. A., 6 , 1 2 2 Jordan, W. R., 182, 186, 186, 189, 197, 198, 211, 214, 220, 221, 222, 223, 224, 225, 227, 228, 232, 233, 239, 244, 246, 247, 248, 250, 251 Jowett, D., 276, 293, 295, 316, 351,355 K Kanno, I., 260, 262, 272, 355 Kanwisher, J. W., 206, 216, 252 Kappen, L., 190, 204, 209, 216, 221, 226, 239, 248, 252 Kaasam, A. H., 177, 220, 248 Kaufmann, M. R., 168, 170, 176, 176, 177, 184, 185, 187, 189, 191, 192, 196, 198, 201, 202, 203, 205, 211, 212, 220, 221, 222, 226, 227, 232, 233, 236, 239, 246, 248 Keck, D. D., 341, 347, 352 Keddington, M. B., 216, 218, 245 Keith, A. D., 338, 355 Kelley, W. P., 262, 355 Kennett, C. E., 9, 122 Kinzel, H., 336, 350 Kitamura, S., 264, 355, 356 Kitching, J . A., 5, 122 Kittrell, J. R., 137, 148, 150, 162 Klement, O., 263, 264, 356, 362 Klepper, B., 170, 176, 176, 183, 186, 188, 199, 202, 221, 223, 224, 227, 231, 232, 239, 248, 249, 250 Klikoff, L. G., 176, 177, 180, 216, 218, 226, 245 Klimpel, R. R., 138, 145, 162 Knaben, G., 264, 356 Knipling, E. B., 168,174, 180,244,249 Kochenderfer, J., 222, 249 Kock, P. C., de, 295, 356 Kondo, Y., 260, 262, 362 Kotilainen, M. J . , 263, 356 Kozlowski, T. T., 167, 200, 249 Kramer, P. J., 166, 184, 200, 224, 249 Krapfenbauer, A., 263, 356 Krause, W., 256, 259, 263, 264, 265, 271, 274, 356 Krebs, J. R., 94, 122 Kretschmer, L., 263, 337, 356 Kriedemann, P. E., 219, 226, 233, 249

37 I

Kruckeberg, A. R., 262, 264, 265, 266, 268, 269, 270, 271, 272, 273, 316, 317, 318,329, 341,342,344,356 Kuramoto, R. T., 216, 249 L Lam, H. J., 264, 271, 356 Lammermayr, L., 263, 342, 356, 357 Landenberger, D. E., 10, 22, 37, 38, 39, 108, 109,122 Lange, 0. L., 169, 190, 204, 209, 216, 221, 226, 239, 248, 249, 252 Langhans, R. W., 177, 179, 180, 190, 252 Lapidus, L., 145, 162 Larcher, W., 232, 249 Larkin, P. A., 85,120 Lassoie, J. P., 209, 223, 239, 249 Lawlor, D. W., 179, 182, 210, 249 Lee, R., 222, 249 LefBbvre, C., 348, 357 Le Gendre, C., 263, 357 Leggett, J. L., 333, 353 Leonowicz-Babiak, K., 316, 328, 332, 335, 352, 363 Leroux, E. J., 11, 122 Leslie, P. H., 16, 17, 18, 19, 122 Levitt, J., 235, 249 Lewis, H., 266, 323, 327, 328, 332, 335, 345, 347, 348, 357, 366 Lieth, H., 227, 249 Lipe, J. A., 227, 249 Lipman, C. B., 273, 275, 337, 339,354, 357 Litvinenka, A., 262, 357 Liu, W., 177, 231, 246 Lodge, R. W., 316, 351 Lodoenikov, W. N., 256, 357 Loew, O., 314, 332, 357 Loneragan, J. F., 330, 357 Lopushinsky, W., 182, 204, 220, 222, 249 Lorio, P. L., 181, 182, 229,248 Lotka, A. J., 15, 123 Lounamaa, J., 283, 305. 314, 357 Love, L. D., 170, 216, 249 Low, A. P., 264, 357 Luckinbill, L. S., 10, 123 Ludwig, W., 263, 356 Lunt, 0. R., 296, 330, 365, 366 Lupke, B. von, 233, 250

372

AUTHOR INDEX

Lyon, C., 331, 357 Lyon, G. L., 264, 269, 284, 285, 301, 305, 307, 310, 332, 357 M Maas, J. L., 264, 338, 357 MacArthur, R. H., 15, 123 Macko, S., 264, 358 MacLean, A. J., 283, 294, 3.54 McLean, G. W., 277, 283, 358 McMichael, B. L., 227, 250 McMillan, C., 318, 358 McMurty, J. A., 38, 124 McNulty, P. S . , 40, 46, 51, 124 Madhok, 0. P., 331, 332, 333, 334, 335, 343, 358 Main, J. L., 316, 332, 335, 343, 348, 358 Maljuga, D. P., 272, 277, 358 Malquori, A., 261, 358 Manly, B. F. J., 62, 70, 7 1 , 123 Mansfield, T. A., 218, 250 Mark, W. R., 230, 250 Markgraf, F., 263, 265, 358 Marks, R. J., 64, 81, 82, 95, 123 Marrs, R. H., 335, 343, 358 Marshall, J. K., 264, 358 Martin, M. H., 337, 358 Martin, P. E., 233, 245 Martino, E., 264, 269, 358 Mason, H. L., 264, 358 Mason, R. S . , 229, 250 Matsuno, T., 260, 262, 362 Maurer, W., 263, 358 May, D. W., 314, 332, 357 May, R. M., 4, 15, 16, 19, 21, 26, 32, 39, 54, 55, 114, 118, 119, 122, 123 Mayr, E., 343, 358 Mech, L. D., 6, 123 Meidner, H., 218, 250 Mertz, W. 305, 358, 362 Messenger, P. S., 37, 123 Messeri, A., 264, 346, 358 Meyer, R. F., 195, 224, 234, 250 Mikkola, E., 263, 358 Millar, E. A., 274, 275, 281, 282, 293, 296, 299, 322, 337, 363 Miller, P., 62, 70, 71, 123 Millikan, C. R., 308, 358 Milthorpe, F. L., 166, 201, 245

Minguzzi, C., 274, 285, 294, 302, 358, 359 Mitchell, R. L., 259, 359 Mizuno, N., 284, 295, 359 Molz, F. J., 224, 250 Momotani, Y., 264, 355 Mook, J. H., 85, 87, 89, 90, 95, 96, 104, 123 Mook, L. J., 85, 87, 89, 90, 95, 96, 104, 123 Mooney, H. A., 203, 226, 250, 260, 323, 327, 328, 332, 335, 345, 347, 366 Moore, R. T., 222, 250 Morgan, M. A., 335, 359 Morgan, P. W., 227, 248, 249 Mori, H., 3 8 , 1 2 3 Morris, R. F., 5, 9, 123 Morton, J. E., 264, 352 Muraltami, Y., 290, 313, 366 Murata, G., 264, 355, 356 Murata, K. J., 258, 353 Murdoch, W. W., 22, 32, 36, 3 7 , 4 0 , 4 6 , 51,59, 64, 65, 67, 69, 74, 79, 81, 82, 84, 86, 95, 96, 99, 101, 106, 113, 119,121,123,124 Murton, R. K., 60, 123

N Nagano, I., 264, 359 Nakai, G., 264, 355 Namken, L. N., 182, 203, 223, 250 Negodi, G., 264, 359 Nemec, A.. 272,274,286,297,302,311, 314, 359 Nevole, J., 263, 359 Newman, E. I . , 238, 250 Newton, L. E., 264, 269, 366 Nicholas, D. J. D., 314, 366 Nicholson, A. J., 20, 21, 120 Nielsen, F. H., 293, 359 Nielson, D. G., 173, 192, 193, 248 Nieman, R. H., 204, 246 Nogina, N. A,, 262, 359 Novak, F. A., 263, 264, 332, 337, 342. 346. 359 0 Oaten, A., 22, 40, 46, 51, 67, 99, 101, 123,124

373

AUTHOR INDEX

Odening, W. R., 186, 204, 209, 21 226, 228, 250

Oechel, W. C., 186, 204, 209, 216, 226, 228, 250

Oertli, F. T., 233, 246 Ojea, F. G., 278, 353 Olvhg, H., 177, 231, 246 Onikura, Y., 260, 262, 272, 355 Oppenheimer, H. R., 166, 250 Orsino, F.,264, 269, 358 &&a, K., 337, 359

R Radford, A. E., 264, 361 Ragg, J. M., 262, 279, 298, 309, 361 Rai, D., 271, 274, 361 Rapport, D. J., 36, 38, 72, 80, 81, 96, 106,124

Rattenbury, J. A., 264,352 Raven, P. H., 349, 361 Rawlins, S. L., 172, 183, 254 Raymond, M., 264, 361 Reed, K. L., 213, 214, 226, 228, 251, 254

P Page, N. J., 268, 359 Palacios, M. L., 264, 359 Pampanini, R., 264, 353, 360 Pancic, J., 263, 264, 360 Page, C. P., 182, 186, 244 Patten, D. T., 209, 216, 247 Pavarino, G. L., 264, 360 Pavlovic, Z., 264, 360 Pazourek, J., 191, 250 Pearcy, R. W., 203, 226, 250 Peareon, G. A,, 274, 337, 355 Pedro, G., 269, 260, 360 Peligek, J., 274, 360 Pennell, F. W., 264, 360 Perez, M., 203, 216, 230, 251 Peterson, P. J., 264,269,282, 284,286, 294, 306, 307, 310, 332,357,364

Pichi-Sermolli, R., 269, 264, 342, 346, 360

Pierpoint, G., 171, 199, 250 Pigott, C. D., 346, 360 Plumb, R. C., 236, 250 Pinto da Silva, A. R., 264, 269, 361 Polunin, N., 270, 361 Popham, E. J., 71,124 Powell, R. D., 227, 250 Pratt, P. F., 306, 361 Proctor, J., 269, 262, 264, 266, 268, 270, 282, 307, 332, 346,

271, 292, 316, 336, 348,

272, 274, 276, 276, 278, 293, 296, 296, 300, 306, 316, 318, 326, 328, 331, 336, 337, 338, 340, 346, 361 Puritch, G . S., 187, 222, 226, 250, 251

Q

Quesada,J. R., 9, 124 N

Reed, R. C., 38, 83,84,96,96, 106, 107, 108,124

Reid, C. P. P., 230, 250 Reilly, P. M., 139, 140, 141, 143,162 Richards, G. P., 176, 186, 194, 196, 198, 220, 231, 232, 236, 236, 237, 247 Richards, G. W., 12,124 Richards, L. A., 231, 243 Richardson, C. J., 219, 226, 251 Richardson, J. A., 273, 361 Richter, H., 170, 171, 184, 189, 236, 251 Rigotti, H., 264, 361 Ristanovic, B., 339, 364 Ritchie, G. A., 169, 171, 173, 184, 189, 193, 198, 199, 204, 206, 208, 220, 221, 222, 226, 232, 233, 236, 239, 247, 251 Ritchie, J. T., 182, 221, 223, 248, 251 Ritter-Studnioka, H., 264, 269, 337, 339, 346, 361, 362, 364 Rivas-Goday, S., 264, 362 Roberts, J. E., 223, 247 Robertson, C. W., 238, 248 Robinson, W. O., 261, 272, 273, 276, 276, 279, 287, 298, 303, 320, 326, 330, 362 Rook, D. A., 176, 233, 251 Rosenbrock, M. M., 146,162 Rosenzweig, M. L., 16, 32, 124 Rottenburg, W., 171, 189, 251 Royama, T., 40,64, 86, 89, 98, 118,124 Rundel, P. W., 176, 176, 196, 209, 213, 230, 236, 236, 251, 252 Rune, O., 269, 262, 263, 264, 266, 270, 271, 272, 274,337, 342, 346,362 Runkles, J. R., 182, 203, 223, 250 Russell, E. W., 276, 362

374

AUTHOR INDEX

S Sadeback, R., 347, 362 Sadowska, A., 264,362 Salt, G. W., 10, 124 Sandness, J. N., 38, 124 Sankary,M.N., 181, 182, 216,232,251 Sarma, P. S., 296, 350 Sarosiek, J., 264, 272, 274, 279, 287, 298, 303, 309, 311, 321, 326, 332, 337, 358, 362 Saaaki, S., 260, 262, 362 Saatry, K. S., 296, 350 Sauberlich, H. E., 293, 359 Saunder, D. H., 274,275, 280,282, 292, 294, 296, 299, 300, 305, 306, 307, 308, 316, 322, 363 Schaeffer, C. H., 162, 162 Scharrer, K., 283, 284, 300, 307, 308, 314, 362 Schellman, W., 261, 262, 362 Schmidt, E., 235, 248 Schnare, P. D., 226, 244 Scholander, P. F., 167, 168, 170, 171, 173, 183, 184, 185, 186, 189, 193, 196, 203, 216, 217, 230, 235, 236, 251 Schropp, W., 283, 284, 300, 307, 308, 314, 362 Schulze, E-D., 190, 204, 209, 216, 221, 226, 239, 248, 252 Schwarz, K., 305, 362 Scoggm, H. J., 264, 362 Scott, D. R. M., 197,199,209,223,239, 247, 249 Sedova, G. P., 339, 362 Seidel, F., 263, 356 Seifriz, W., 264, 362 Seivala, O., 263, 356 Sequeira, E. M., de, 259, 260, 262, 272, 280,288, 296, 299, 321, 337,362 Severne, B. C., 288,294, 344,362 Shardakov, V. W., 174, 180, 252 Shaw, R. H., 191,252 Shelton, D. C., 6, 122 Shibuichi, H., 264, 359 Shkoljnik, M. J., 337, 346, 363 Shreve, F., 264, 363 Simonson, G. H., 271, 274,361 Skau, C. M., 222, 252 Slatyer, R. O., 167, 169, 170, 174, 175, 184, 188, 190, 200, 238, 239,252

Slavik, B., 167, 252 Slesak, E., 332, 335, 363 Small, E., 216, 252 Smart, R. E., 219, 226, 233, 249 Smirnov, U. S., 337, 346, 363 Smith, A. M., 315, 352 Smith, F. E., 22, 24, 124 Smith, G. N., 149, 150, 160, 161, 162, 163 Smith, H., 138, 142, 144, 145, 148, 159, 162 Smyth, M. E. B., 74,123 Snaydon, R. A., 276, 351 Snowball, K., 330, 357 Soane, B. D., 274, 275, 280, 282, 292, 294, 299, 300, 305, 306, 307, 308, 316, 322, 363 Socava, V., 264, 363 Solberg, J. J., 137, 163 Solomon, M. E., 33, 112,124 Soo, R., von, 264, 363 Soon, Y. K., 316, 363 Spanner, D. C., 169, 252 Spence, D. H. N., 259, 262, 264, 265, 271, 274, 276, 281, 282, 293, 296, 299, 309, 322, 337, 344,363 Spomer, G. G., 200, 224, 252 Spomer, L. A., 177, 179, 180, 190, 252 St. Amant, J., 23,124 Stark, N., 203, 252 Stebbins, G. L., 338, 342, 363 Stecker, R. E., 175, 176, 196, 235, 236, 252 Steiner, E. C., 138, 145, 162 Stevenson, K. R., 191, 252 Storey, C., 145, 162 Storvick, T. S., 235, 249 Stout, P. R., 274, 337, 355 Strain, B. R., 186, 204, 209, 216, 226, 228, 250 Stuntz, D. E., 264, 338, 357 Sucoff, E., 170, 196, 199, 203, 204, 209, 210, 221, 224, 236, 239, 252 Sucoff, E. I., 186, 192, 248 Sulej, J., 332, 335, 363 Sullivan, C. Y., 176, 186, 244 Suza, J., 263, 363 Svenonius, F., 264, 363 Swaine, D. J., 277, 296, 308, 363 Swanson, R. H., 222, 223, 252

AUTHOR INDEX

T Taboadela, M. M., 278, 353 Tadros, T. M., 340, 363 Taarum, R., 231, 232, 239, 252 Tajino, T., 264, 359 Takahashi, F., 37, 38, 105, 111,124 Takahashi, Y., 337, 359 Tanaka, H., 337, 366 Tmiguti, M., 264, 364 Tassoulas, J. A., 281, 292, 299, 306, 337, 364 Taylor, E. J., 37, 110, 111, 112, 124 Taylor, H. M., 223, 231, 232, 248, 249 Taylor, 5. A., 169, 170, 252 Teal, J. M., 206, 216, 252 Teaie, Z., 339, 364 Thoday, J. M., 347, 364 Timperley, M. H., 282, 294, 364 Tinbergen, L.,86, 86, 87, 89, 96, 104, 106,124 Tinklin, R., 238, 252 Tobiessen, P., 173, 176, 176, 196, 236, 236, 252 Tokudome, S., 260, 262, 272,355 Torii, K., 264, 356 Tranquillini, W., 233, 247,253 Treleaee, H. M., 331, 364 Trelease, S. F., 331, 364 Turner, J., 223, 247 Turner, J. A., 187, 251 Turner, N. C., 170, 171, 173, 190, 191, 192, 197, 201, 206, 220, 221, 223, 229, 230,233, 237,243,253 Turner, R. G., 340, 341, 350 Turnock, W. J., 113,120 Turrill, W. B., 263, 364 Tyree, M. T., 171, 194, 195, 196, 237, 253

U Ulrich, A., 234, 246 Usov, N. I., 273, 364 V van der Honed, T. H., 201,253 Van der Marel, H. W., 260, 261, 364 Vanselow, A. P., 283, 306, 308, 364 Varley, G. C., 6, 11, 12, 124 Veihmeyer, F. J., 238, 253 Veniale, F., 260, 261, 364 Venkatasubramaniam, V., 296, 350

375

Vergnano, O., 264, 274, 276, 276, 278, 283, 284, 286, 291, 294, 296, 296, 299, 302, 306, 306, 307, 308, 309, 310, 337, 352, 354, 355, 358, 359, 364 Vinogradov, A. P., 277, 296, 365 Vlsmis, J., 274, 316, 336, 337, 365 Volk, R. J., 336, 359 W Waggoner, P. E., 171, 191, 197, 206, 220, 223, 233, 253 Walker, H. M., 316, 327, 328, 332, 365 Walker, R. B., 264, 272, 274, 316, 322, 327, 328, 329, 332, 333, 334, 337, 343, 358, 365 Wallace, A., 330, 365 Walters, S. M., 346, 360 Wambolt, C. L., 216, 253 Ware, D. M., 37, 78,124 Waring, R. H., 168, 170, 171,173, 174, 184, 186, 197, 206, 207, 209, 213, 214, 216, 216, 224, 226, 228, 233, 239, 245, 251, 253, 254 Watson, B. S., 149, 160, 160,163 Weatherley, P. E., 166, 181, 238, 252, 254 Werner, E. E., 4, 121 West, D. W., 176, 177, 178, 179, 186, 186, 188, 254 West, N. E., 170, 216, 249 Weat, W., 264, 266, 277, 365 Whatley, J. M., 272, 365 Wherry, E. T., 264, 365 White, C. D., 274, 282, 330, 339, 365 White, R. S., 222, 250 Whitney, H. S., 229, 247 Whittaker, E. J. W., 268,365 Whittaker, R. H., 8,125, 263, 264, 266, 266, 267, 268, 269, 270, 272, 322, 341, 342, 346, 365 Whitten, K., 338, 361 Whittig, L. D., 260, 262, 365, 366 Wiebe, H. H., 172, 183, 254 Wild, H., 264, 269, 281, 284, 288, 289, 300, 303, 306, 308, 312, 314, 316, 323, 332, 342, 365 Wildman, W. E., 260, 261, 262, 365, 366 Wilkins, D. A,, 293, 331,366

376

AUTHOR INDEX

Williams, E. J., 20, 21, 120 Williams, P. C., 281, 292, 295, 296,366 Wilson, A. D., 262, 272, 366 Wilson, S. B., 314, 366 Wiltshire, G. H., 281, 284, 292, 296, 366

Winstein, C. B., 141,162 Wood, L., 68, 82,125 Woodell, S. R. J., 259, 262, 264, 265, 266, 268, 269, 270, 271, 272, 274, 315, 318, 323, 327, 328, 332, 335, 345, 346, 347, 348,361, 366 Woodford, A. O., 262, 355 Woodman, J. N., 232, 254 Wormell, P., 275, 353 Worrall, J., 223, 254 Wright, A. C. S., 262, 264, 266, 274, 278, 286, 297, 301, 317, 351 Wright, J. L., 192, 202, 209, 244

Wright, W. H., 171, 173, 205, 253 Wyllie, P. J., 257, 366 Wyn Jones, R. G., 295, 330, 336 Y Yamagata, N., 290, 313, 366 Yamanaka, T., 264, 366 Yatazawa, M., 337, 366 Youngberg, C. T., 213, 233, 253, 271, 274, 361

Z Zaerr, J., 198, 204, 221, 223, 254 Zimmerman, M. H., 167, 173, 234, 235, 254

Zlatnik, A., 263, 366 Zollitsch, L., 264, 366 Zolyomi, B., 264, 366 Zussman, J., 258, 365

Subject Index A A b i ~ 176 , ff., 1-6, 189, 193, 286, 302, 311

amabilia, 169 magnifica, 213 ff. procera, 169 Acacia, 226 Acanthina, 37, 64 ff., 89, 95 ff., 102 Acer, 226 Acheta, 36 Acdiua, 36 Achillea, 341 ff. Age distribution: stable, 20 Aggregation, 21 Aggregative response, 113 Agropyron, 316, 332, 335, 343, 348 Aqrostia, 293, 316, 324 ff., 331 ff., 335, 343, 348

Alburnw (Bleak), 37 Alewite fish (Ahsa),4 Alfalfa, 177 Algae, 37 Almond moth, 37 Alnw, 230 Aloe, 346 Alternative crops, 118 Alternative prey, 59 Alyssum, 263,286,288,294,302,310 Amphiboles, 257 Amphipods, 37, 78 Ana;gaeta (Flour moth), 110 Analysis problem, 139 ff. Andropogon, 284, 289, 303, 305, 312 Anthoxadhum, 348 Anti-transpirants, 233 Aphids, 37, 47, 81, 95 ff. Apostatic selection, 59 ff., 62, 87, 94 Apoplastic water, 237 Apple, 175, 177, 185, 197, 221 Apricot, 175 ff., 176, 221 Amurnria, 274, 285, 301 Arceuthobuim, 230 Area of discovery, 20

Arenaria, 275, 345 Armeria, 336, 348 Artemia, 37 Artemisia, 204, 221, 226, 239 Asphiurn, 347 Assimilation; net, 224 Assimilation rate, 242 Atmospheric Evaporative Demand, 200

Atriplex, 222, 226 Atripla polycarpa, 182 Attack rates, 95 Avena, 284, 296, 328, 341, 344 Azotobacter, 339 B Bakznw, (Barnacles), 5, 6 , 19, 65, 68 m ~ m7 ,

ghndula, 5, 6, 95 ff., 102 Barium, 337 Bcm!eria, 284, 289, 303, 312 Barley (Hordeurn),291, 307, 314, 334 Barnacle, see Balanua species Bayesian function, 140 Bean, 231, 239, 292 Becium, 288, 289, 344 Beet (Beta),177, 293 Bernoulli trails, 65 Best model, 134 Beta, 177 Betula, 230 Binomial distribution, 65 Bioassay nickel, 296 Biodegradation, 135 Bionomial probability function, 140 Birch, 287, 311, 314 Blepharia, 288 Bluegill, 38, 66, 83, 95 ff., 106 Bordered white moth, 89 ff. Boron, 337 Bomina, 4 Brae&, 191, 344 '

377

378

SUBJECT INDEX

Bream, 37 Brodiaea, 338 Bromua, 341 Bugs, 113 Bupalua, 89 ff., 96, 130

Copper, 293, 295, 337 Corixa, 36, 71 Corn, 177 ff., 190, 221 C m u a , 236 Cotton, 177 ff., 185, 190, 197, 223, 224, 225, 227, 231, 233

C Cactus, 216 Calcium, 294, 305, 314 Calcium/Magnesium ratio, 3 14 Callitria, 223 Callunu, 324 Camellia, 203 Cannibalism, 113 Capsicum, 117 ff. Carp, 95 ff. Cassinia, 285, 301, 310 Caatanea, 290, 313 Catastrophic selection, 348 Ceanothus, 339, 343 Cedar (Chamaecyparis), 266 ff. Centaursa, 285, 302, 310 Cerastium, 316, 324, 344, 347 Ceriodaphina, 4 Chemical defenses, 8 Chilopsis linearis, 176 ff., 178 Chironomidae, 37 Chlamydomonas, 36, 66, 72, 95 ff. Chloride, deficiencies of, 337 Chlorite, 261 Chromium, 296, 307, 344 Chromosome number: of plants, 348 Chrysanthemum morifolium, 177 ff., 190 Chrysolina, 9 Citrus, 220, 233, 239 Clarkia, 348 Clay minerals in Serpentine, 261 Clethra, 290, 313 Clover (Trilfolium),291 Cobalt, 308 ff., 344 Coccinella, 95 ff. Cochlearia, 324 Cohesion theory, 167 Combretum, 289, 290, 304, 313 Competition, 345 Competitive exclusion, 345 Conductivity meter, 189 Conformers, 208 Conifer needles, 241 Constant Dreference curve. 62 C o n v o l v d b , 284, 289, 303; 312

Cottonwood, 2 16 Cowpea, 177 ff. Creosote bush, 216 Criteria for stability, 51 Croaker, 85 Cryptochaetum, 9 Cupressus, 343 Cushion scale, 8 Cyclops, 4 Cynoxylon, 290, 313 Cypriltus (carp), 37 Cyzenis, 11, 39, 119 D Dahlbominua, 37 Daphnia, 4, 37 Darcy’s Law, 234 Deermice (Peromy~cus), 95 ff. Degrees of freedom, 137 Demand factors (AED), 202 Density, 33 -dependence, 17, 32, 59, 116 ff. -dependent functional responses, 118 -dependent predation, 31 -independent predation, 3 1 method, 174 technique, 180 Design problem, 139 ff. Deutzia, 290, 313 Development response, 119 Dianthus, 287, 303, 311 ff., 326 Diaptomua, 4 Dicoma, 284, 288 ff., 301, 304 ff., 312 ff. Diffusion gradient, 202 Dipinium, 10 ff. Disc equation, 99, 104 Disjunct distribution, 345 Dispersal rate, 10 Disruptive selection, 347 Diversity, 119 Dogwood, 236 Douglas-fir, 174, 184, 185, 189, 196, 213,221, 224, 225, 228, 235, 236

Drosophila, 74, 76, 95 ff., 347 Dursban, 133 ff., 149

SUBJECT INDEX

E Ecotypic differentiation, 340 Effective soil moisture, 239 Elder, 226 Emmenanthe, 340 Empetrum, 324 Encarsia, 37 Enciliafarinosa, 176 ff., 178 Encounter rates, 94 Endemics, 342 Ephestia, 39 Eragrostis, 284, 292 Error structure, 149 Euglena, 36, 66, 72, 80, 95 ff. Euphorbia, 286, 302, 310, 346 Euphydryas, 338 Eurotia, 222 Exenterus, 37 Exidechthis, 37 Extinction, 4, 19 F Fagopyrum, 328 Fertilization, 233 Festuca, 324 Fir (Pseudotsuga),176 ff., 189, 193, 198, 208, 215, 222, 228, 266 ff. Fish, 84 ff. Flocking, 86, 104 Flourmoth (Anagasta), 110 Flow resistance, 211, 236 Founder effect, 343 Fractional refuge, 20, 21 Fraxinus, 235 Freezing Point Depression, 174, 180 Frequency-dependent, 59 Friction, 196 Frictional, 240 Frictional potential, 170, 234 Frost hardiness, 232 Fruit, 226 Functional response, 25, 33 ff., 95, 104ff., 112, 115 stabilizing, 26, 58 destabilizing, 26, 58 Fuaarium, 229 G Galiurn, 288, 303, 311 ff., 326 Generation time lag, 24 Glycine, 177 ff.

379

Goldfish, 149 Goodness of fit, 145 Gossypium, 177 ff. Grape, 175 ff., 176 ff. Graphical analysis of functional response, 104 ff. Grass blades, 241 Gravitational, 240 potential, 170 potential gradient, 234 Gravity, 196 Greasewood, 176 ff. Great tit, 86 ff., 95 ff. Growth, 224 stunting of, 269 Grunt (Microkpidotus), 85 Guppies, 73, 76 ff., 95 ff., 110

H Hammada, 221 Handling time, 58, 99 Heat build-up in the pressure chamber, 187 Heat-pulse velocity technique, 222 Hebe, 285, 301, 305, 307, 310 Helianthus, 177ff., 327,328,331,333ff., 343 annus, 168 Helichrymm, 286, 302, 310 Hemlock, 215, 237 Heterogeneity, spatial, 9, 22 ff. of environment and effect on switching, 84 spatial, effect on stability, 32, 117 Hierodula (Mantidae), 36 Holly, 233 Hordeurn, 284 Housefly, 36 Hybanthus, 288, 294, 343 Hyparrhenh, 289 Hypericum, 9 Hysteresis, 223 I Icerya, 8 “Ignoring rate”: a switching mechanism, 104 I h , 233 Inadequacy test, 147 Indigofera, 284, 289 ff.. 293, 301, 304, 313

380

SUBJECT INDEX

Infertility in serpentine, 271 Instability, 4 ff. Interaction between predators, 114 Internal moisture deficit, 171 Intra-generational response, 114 Inversely density-dependent mortality, 31, 33, 116

Invulnerable prey, 6, 22 ff., 117, 119 Iron, 295, 307 toxicity, 337 Irrigation, 227, 232, 242 Iterative, parameter estimation, 145 programme, 134

J Juniperua, 236, 324

Lolium, 306, 335 Lotka-volterra equation, 13 ff., 23, 25, 32, 48, 56

Lotononia, 339 Lotus nevadensis, 208 Loupetia, 289 ff., 304, 313 ff. Lycopersicon, 176 ff., 177 ff., 316, 328 ff., 335

Lychnis, 331

M Magnesium, 295, 314, 340 Manganese, 295 Maple, 239 Maritime plants, 336 Markov chain, 24 Matric potential, 170, 189, 195, 235, 240

K K (=saturation density), 17 ff. Kaolinite, 261 Key factor analysis, 11, 119 Kobrtda, 275 L Laduca, 328 Ladybirds, 39, 65 ff., 81, 84, 95 ff. Lamprey, 4 Larch (Lurk),233 Larrea, 226, 228 divaricata, 176 ff., 178, 204, 209 Latentic serpentine soils, 272 Luthyrua, 288, 303, 311 ff., 326 Lead, 293 Leaf, energy load, 202 resistance, 204 Learning and functional response, 105 ff.

Leptmpermum, 285, 301, 306, 307, 310 Lethoceroa (Bellastomatidae), 36 Life-forms, 266 Likelihood function, 140 ff. Limiting conditions, 211, 242 finanthue, 327 ff., 347 finaria, 288, 303, 311 ff., 326 Lindera, 290, 313 Linear variable displacement, 223 Liriodendron, 176 ff., 219 tulip(fera, 175 Loblolly pine, 175 ff. Locust, 219, 232

Maximum likelihood, 134 Maze (Zea),306 ff. Medicago, 344 aativa, 177 Mesquite, 199, 210, 221 Microlepidotua, 85 Midge, 83, 95 ff. Mimulua, 228 breweri, 208, 209 Minuartia, 324, 347 Mistletoe, 230 Mites, 37, 39 Model building, an iterative procedure, 136

technique, 134 ff. Models, discrimination, 136, 139 ecosystem, 134 empirical, 135 mathematical, 133 ff. mechanical, 134 ff. phenomenological, 135 plant, 133 process, 133 Moisture status of soil, 258 Molybdenum, 336 Montmorillonite, 261 Moose-wolf system, 6, 119 Mosquito, 38, 83, 95 ff., 107, 113 larvae, 162 Moth larvae, 37 Mouse, 109 Mussels (Mytdue), 10, 37, 65, 80, 95 ff., 102

SUBJECT INDEX

Myosotia, 285, 301, 310 Mytau8 edulh, see Mussels N Naval orange, 176 ff. Necessary rate of reward, 78 Negative binomial distribution, 47 ff., 49 Negative hydrostatic pressure, 171 Nerneritus (Lepidoptera), 37, 39, 114 Niche, 54 Nicholson-Bailey model, 20, 24, 65,114 Nickel, 277 ff., 293, 307, 340, 343 bioassay, 296 Nicotiana, 176 ff. Nitrifying bacteria, 339 Nothofagus, 282 Notholaena, 346 Notonecto, 36, 113 Notothkzapi, 284, 285, 301, 310 Numerical response, 112, 115 ff., 119

0 Oak (Quercw,), 168, 189, 176, 177 ff., 216, 218, 221, 223, 226, 236, 286, 287, 303, 344 Oat (Avena), 230, 291 ff., 296, 306 ff., 314 Occam’s razor, 137 Olivine, 257 Optimization algorithms, 139 Opuntia, 216 Orange, 191, 196, 211, 212, 219, 221, 222, 227 slice gap, 227 Osmotic potential, 189, 193, 240 pressure, 187 Overall functional response, 40, 44, 47 Over-browsing, 6

P Pawheria, 286, 301 Paramecium, 10 ff. Parasite aggregation index, 65 Parkia, 230 Parsimony, principle of, 137 P a w major, 86 ff. Patch density of prey, variability in, 38, 41, 60, 118

381

Pear, 176 ff., 176 ff., 221 ff., 226, 227 Pearsonkz, 284, 289, 304, 313 ff., 339 Peas (Pisurn),214, 216, 219, 224, 228, 307, 314 Pepper, 177 ff., 204 Perennial seepwood, 177 ff. Peromyecue, 95 Phase diagram, 16 Phmeolus, 177 ff., 231 Phenology, 227 Physiological tolerance to water stress, 218 Picea, 176ff., 177ff., 230, 233, 235 Pigeons, 66, 69, 84 Pimelea, 284, 285, 301, 307, 310 Pine (Pinua), 176, 189, 190, 192, 193, 199, 204, 215, 220, 221, 223, 224, 226, 230, 233, 236, 239, 266 ff., 286, 302, 311, 314, 344 Pinto bean, 177 ff. Pinua, 176ff., 186, 191, 193 Pinua pseudotauga, 189 Pitmater ( S t a A h ) , 37 Pisum, 284 Pithcolobium, 230 Plantugo, 336 ff., 345 Plant moisture stress, 170, 238 Plasmatic resistance, 293 Pbolopw, (Sawtly), 37 pmin , 242 Poa, 316, 335, 343 Pocket gopher (Thomomya),338 Podisus (Pentatomidae), 36 Poisson distribution, 47 ff., 49 Polar, 175 Pollution, effect on P, 230 Polygonurn, 228 wcadenee, 208, 209 lcelloggii, 208, 209 Polymorphism, 63 Pomology, 226 Poplar, 176 ff., 219, 224, 226 Potassium, 335 Potato (Sohnum), 177, 292 Potential, 170 Potential: base (BP), 206 ff. Potential depression (DP), 205 ff. Potentilh, 337 Prmn (Aphid), 37 Predator behaviour, model of, 41 Predator-prey equations, 128

382

SUBJECT INDEX

Predators, responses of, (see under type of response+-functional, numerical etc.) Predator switching, 59 ff. Preference by a predator, 61 ff., 104 at equality, 85 Pressure chamber, 165 ff. in entomology, 229 in pathology, 229 in pollution effects, 229 Pressure-volume curve, 193, 241 Prey, patchiness, 67 refuges, 32 species, 95 Prosopeis, 199 Protozoa, 80, 84 Prunella, 341 Prunes, 221, 233 Pseudohga, 186 menziesii, 169, 174 Psyllids, 36 “Pure” interaction, 13 Pure error, mean square, 147 Pyroxenes, 257 Pyrus, 176 ff., 185 malus, 175, 177 ff.

Q

Quail, 70 Quercus, 168, 169, 177 ff., 287, 290, 213, 325, 343 ff. R Random search, 39, 41 Raspberry, 185 Reaction rate constant, 136 Reaumuria, 22 1 Redwood, 196, 235 Reference, 213, 242 Refuge, 6, 19, 117, 119 Regression models, 135 Regulators, 208 Rejection rate, 74, 104 Relative attack rates, 59 Residual analysis, 136 ff., 147 plots, 148 ff., 152 ff. Reward rate, 78, 104 Rhacomitrium, 337 Rhododendron, 168, 175 ff., 176ff., 184, 190, 229

R i b s binominatum, 208, 209 Robinia, 219 Roots: water relations, 231 “Rosette” disease, 329 Rubus, 185 Rudd, 71 Rumex, 341, 348 Rutilw (Roach), 37 Rye grass (Lolium),291 ff. Rye (Secale),284, 307, 314

S Salamander (Ensatina),338 S a l k , 230, 287, 311 Salmon, 37 Saltational speciation, 348 Sap analysis, 233 stress, 170 velocity, 222 Satiation, 31, 97 ff. Saturation density ( = K ) , 17 ff. Saugspannung, 171 Sawfly, 37, 94, 95 ff., 109 Scarcobatus, 176 ff. Schooling, 86, 104 Search image, 71, 85 ff., 94, 104 time, 42 ff. Sea-urchins, 5 Securidaca, 284, 289 Sedum stenopetalum, 208 Seepwood, 180 Sequoia,175, 176 ff., 196, 230, 236 Sequoiadendron giganteum, 175, 176 ff. Serpentine soils, 255 ff. Sesbania, 339 Shrew, 109 Silene, 288, 303, 311 ff., 326, 348 Siskiyou mountains, 266 ff., 342 Snails, 66, 79, 84, 86, 95 ff., 104 seashore, 64, 84 Snap bean, 177 ff. Social behaviour in wolves, 8 Soil, moisture, 272 heaps, 273 resistance, 201 Solanum tuberoam, 177 Solifluction, 272 Solute potential, 170 Sorbu8, 287, 311 Sorghum, 176 ff., 177, 186, 221

383

SUBJECT INDEX

Soybean, 175, 177, 191, 234, 236 Spatial heterogeneity, see Heterogeneity Spruce, 176, 177 ff., 185 ff., 191, 198, 221 ff., 228, 231, 235 budworm, 5 Stability, 2 analysis, 13 ff. criteria, 51 Starfish, 7, 10, 38, 39, 108 Stem diameter, 222 Stentor, 36, 38, 66, 72, 80, 95 ff. Stochastic, models, 19, 61 variation, 16, 32 Stock-recruitment curves, 28 Stomatal, activity, 218 ff. resistance, 242 Streptanthus, 327, 329, 342 ff. Stress-day, 242 index (SDI), 214 Stress period, 225 Stunting of growth, 269 Suaeda frutkosa, 177 ff., 180 Suction force, 171 Sugar beet, 234 Sum of squares, crude, 146 due to regression, 146 lack of fit, 146 pure-error, 146 Sunflower, 168, 177 ff., 190 Sutera, 289, 304 ff., 313 ff. Switching, 59ff., 64ff., 70,79,82,84ff., 87, 94 ff., 104, 106, 110 Sycamore, 216 Sympatric speciation in a two-niche situation, 347 Syrphus, 36 T Tadpoles, 36 T m e cuapidata, 168, 176 ff. Taylor series, 17, 55 Tea, 203, 220 Tegula, 108 Tetrahymena, 36, 66, 80, 95 ff. Thais, 5, 6, 37, 64, 95 ff. Them&, 289 Thermocouple psychrometer, 168, 175 Threshold rate of reward, 78 Thymua, 287, 303, 311 ff., 326 Time lags, 15, 32

Tobacco (Nicotiana), 176 ff., 191, 221, 231, 237, 239, 306 ff., 330 Tomato, 175 ff., 176 ff., 177 ff., 196, 226, 329 Total response, 32, 114 ff. Training, 67, 81 Transducers, 223 Transit time, 58 Translocation of toxins, 349 Transpiration, 222, 242 Transplanting shock, 233 Tritkum, 171 ff., 284 Trout, 78, 85 Teuga, 237 Tubificid worms, 74, 76, 78 ff., 95 ff. Turban snails, 37, 38, 108 Turnip (BrmtGm),291 Typhlodromua, 37 T I

V

Vapour equilibration technique, 174 Vapour gap around roots, 239 Variability in patch densities, 52 Ve'edalia,9 VeUozia, 288 Velocity of sap, 242 Vermiculite, 262 Veronica, 288, 303, 311 ff., 326 VertkicZadkUa, 230 V k h , 177 ff., 288, 303, 311, 326 Vigna, 177 ff., 214 Vine, 221 ff., 225 Vitis, 176 ff., 188, 225 W Wasp, parasitic (Nemeritus), 110 Water, content relative, 181 deficit, 166 movement, 200 potential, 169, 170, 211 potential, total, 168 status, 166, 169 stress, 166, 207 Weathering of serpentine, 259 ff. Webworm, 36 Wheat (Tritkum), 177, 191, 192, 291, 307, 314 Whelk, see T h i a white fly, 37 Winter moth, 5, 39

384

SUBJECT INDEX

X Xerophytic plants, 272 Xylem pressure potential=P, 170 ff., 196, 205, 217, 240 Conifer needles, 192 Definition, 171 Habitat, 216 Plant factors, 218 ff. Xylem sap pressure, 1 7 0

Y Yeast, 37 Yew, 168, 176, 190, 215 Yield, 224 Z Zea mays, 177 ff. Zinc, 293, 337 Zooplankton, 4 Zygophyllum, 221

Advances in Ecological Research, Volumes 1-8: Cumulative List of Titles Analysis of processes involved in the natural control of insects, 2, 1 The distribution and abundance of lake-dwellingTriclads-towards a hypothesis, 3, 1

The dynamics of aquatic ecosystems, 6, 1 The dynamics of a field population of the pine looper, Bupaluapinbriu~L. (Lep., Geom.), 3, 207 Ecological aspects of fishery research, 7, 115 Ecological conditions affecting the production of wild herbivorous mammals on grasslands, 6, 137 Ecological implications of dividing plants into groups with distinct photosynthetic production capacities, 7, 87 Ecological studies a t Lough Ine, 4, 198 Ecology of fire in grasslands, 5 , 209 Ecology, systematics and evolution of Australian frogs, 5, 37 Energetics, terrestrial field studies, and animal productivity, 3, 73 Energy in animal ecology, 1, 69 Forty years of genecology, 2, 159 The general biology and thermal balance of penguins, 4, 131 Heavy metal tolerance in plants, 7, 2 Human ecology as an interdisciplinary concept: a critical inquiry, 8, 2 Integration, identity and stability in the plant association, 6, 84 Litter production in forests of the world, 2, 101 The method of succeseive approximation in descriptive ecology, 1, 35 Pattern and process in competition, 4, 1 Population cycles in small mammals, 8, 268 The production of marine plankton, 3, 117 Quantitative ecology and the woodland ecosystem concept, 1, 103 Realistic models in population ecology, 8, 200 A simulation model of animal movement patterns, 6, 186 Studies on the cereal ecosystem, 8, 108 Studies on the insect fauna on Scotch Broom Sarothamnue ecoparrius (L.) Wimmer, 5, 88

Soil arthropod sampling, 1, 1 A synopsis of the pesticide problem, 4, 76 Towards understanding ecosystems, 5 , 1 The use of statistics in phytosociology, 2, 59 Vegetational distribution, tree growth and crop success in relation to recent climatic change, 7, 177

386

This Page Intentionally Left Blank