TREE vol. 5, no. 9, September
changes in their physical environment due to global warming process studies, surveys through and the application of remotesensing techniques to the global range of marine ecosystems. Such work will show whether changes in the strength of the ocean CO, sink due to biological activity can be expected to have a major influence on the future rate of global warming and the likelihood of achieving long-term climatic stability. On the basis of previous changes in Earth’s climate, the magnitude of response by the ocean carbon cycle may be at least as great as any changes that we can make by curbing CO, emissions from fossil fuels.
Reichle, D., edsl, pp. 403-424, SpringerVerlag 2 Moore, B. and Bolin, B. ( 19861 Oceanus 29, 9-15 3 Steele. I.H. (19891 Oceanus 32, 4-9 4 Holligan, P.M. I19891 Adv. Bat. Res. 16, 193-252 5 Longhurst. A.R. and Harrison, W.G. (19891 Prog. Oceanogr. 22, 47-l 23 6 Eppley, R.W. ( 19891 in Productivity of the Ocean: Present and Past (Berger. W.R., Smetacek, V.S. and Wefer, G., edsl, pp. 85-97, john Wiley & Sons 7 Shanks, A.L. and Trent. I.D. t 19801 DeepSea Res. 27, 137-l 43 8 Sugimura. Y. and Suzuki, Y. II9881 Mar. Chem. 24, 105-131
Acknowledgements We thank our colleagues within the NERC Biogeochemical Ocean Flux Study and the IGOFS programme who have contributed to the practical and theoretical development of these concepts.
t 19901 Science
I Bolin. 6. I19861 in The Changing Cycle: A Global
ITrabalka, J.R. and
Predicting ecosystem response to climate change is a dynamic version of the classic problem of understanding vegetationclimate interrelations. Computer models can synthesize current knowledge and are important tools for understanding possible ecosystem dynamics under changed condifions. Models based on individual plant biology and natural history have been tested with respect to their ability to simulate vegetation response to changed climate, and are being applied to assessthe potential effectsof future climate change.
At the turn of the century, it was noted that there is a high degree of convergence of plant form in similar climates. In some ways, these similarities appear to transcend taxonomic relations. For example, in Mediterranean climates in southern Africa, Chile or Australia, the plants and vegetation resemble those in similar climates in France or Italy, even though the taxonomic affinities of many of the components of H.H. Shugart is at the Dept of Environmental Sciences,The University of Virginia, Charlottesville, VA 22903,USA. (< 1990
9 Toggweiler, I.R. (1989) in Productivity of the Ocean: Present and Past (Berger, W.R., Smetacek. VS. and Wefer, G., edsl, pp. 65-83, lohn Wiley & Sons IO Angel, M.V. (19891 Prog. Oceanogr. 22, l-46 I I Lampitt, R.S. ( 19851 Deep-Sea 885-897
I2 Tans, P.P., Fung, I.Y. and Takahashi. T. 247, 1431-1438
I3 Turner, D.R. et a/. ( 19891 in Third International Conference on Analysis and Evaluation of Atmospheric CO, Data pp. Present and Past (Extended Abstracts), 135-139, World Meteorological
14 Walsh, 1.1.(I9891 in Productivity of the Ocean: Present and Past (Berger. W.R., Smetacek, V.S. and Wefer. G., edsl, pp. 175-191, fohn Wiley&Sons
I5 Dickson, R.R., Kelly, P.M., Colebrook, j.M., Wooster, W.S. and Cushing, D.H. 11988) /. Plankton Res. IO, I5 I-I 69 I6 Stouffer, R.I., Manabe, S. and Bryan, K. II9891 Nature 342,660-662 I7 Folland. C.K. and Parker, D.E. 11990) in Climate-Ocean interaction (Schlesinger, M., ed.), pp. 21-52, Kluwer Academic Publishers I8 Berger, W.H., Smetacek. VS. and Wefer, G. II9891 in Productivity of the Ocean: Present and Past (Berger, W.R., Smetacek, VS. and Wefer, G., edsl, pp. l-34, lohn Wiley G Sons 19 Shackleton, NJ. and Pisias, N.G. (19851 Geophys. Monogr. 32,303-3 I7 20 Mix, A.C. ( 19891 Nature 337, 541-544 21 Muller. P.I. and Suess, E. (19791 DeepSea Res. 26A, 1347-l 362 22 Barnola, I.M., Raynaud, D.. Korotkevich. Y.S. and Lorius, C. II9871 Nature 329, 408-4 I4 23 Broecker, W.S. and Peng, T-H. (19891 Global Biogeochem. Cycles 3,2 15-239 24 Broecker, W.S. and Denton, C.H. (1989) Geochim. Cosmochim. Acta 53,2465-2501 25 Prentice, K.C. and Fung, I.Y. (1990) Nature 346,48-51 26 Berger, W.H. II9891 in Productivity of the Ocean: Present and Past I Berger, W.R.,
Smetacek, V.S. and Wefer, C., edsl, pp. 429-455, John Wiley&Sons 27 Dymond, j. and Lyle, M. (19851 Limnol. Oceanogr. 30.699-7 I 2 28 Bakun, A. (I9901 Science 247, 198-201 29 Charlson, R.I., Lovelock, I.E., Andreae, M.O. and Warren, S.G. (19871 Nature 326, 655-66 I
UsingEcosystemModelsto Assess PotentialConsequences of Global ClimaticChange H.H.Shugart the vegetation are different. Explaining these patterns inspired the earlier plant geographers - Drude, Graebner, Warming, and Schimper to postulate relationships between climate and plant structure whose causalities are still being explored today at both the plant’ and vegetation2-4 level. The classic question of why the vegetation in widely separated places with similar climates is alike has been recast as a topical scientific problem, namely the evaluation of global climate change. The rising level of CO, (and other’greenhouse’ gases) in the atmosphere, and its consequences, have been called an ‘uncontrolled experiment’ on the earth’s geophysical and biotic sys00
tems5. The level of uncertainty about the response of the earth’s atmosphere to this change is very high6. Nevertheless, the prediction of the response of the earth’s vegetation to global climate change is a remarkably rich scientific challenge that hearkens back to the classic plant geographers. The evaluation of the effects of change in the earth’s climate with respect to terrestrial ecosystems can be broken down into three scientific problem areas. Recognizing that the concentrations of CO, and other gases in the atmosphere are climatic variables, sensu stricta, one can ask whether, under conditions of a climate change, one would expect: 303
TREE vol. 5, no. 9, September
uniquely determined by the ecosystems involved? (31 Development of new ecosystems not currently in evidence. We know that in the recent geological past (within the past 30 000 years) and over extensive areas, there have been ecosystems that are not found today7. Could one expect some of the present ecosystems to disappear or new ecosystems to form under a future climate change? The answers to these questions are to a degree dependent upon the scheme that is used to classify ecosystems. In general, the classifications used in global studies are relatively broad. They are often based on the life forms of the dominant plants. Thus, the questions posed as examples above would be asked with regard to ecosystem units such as tropical moist forest, subtropical savanna, temperate deciduous forest, etc. Evaluation of responses of ecosystems to climate change poses questions at large spatial scales and over long time scales relative to traditional ecological studies (studies conducted over a few hectares for two or three years). Also, the experimental protocols that attend climate-change problems are not well developed. The result is that considerable use of ecological models is made to synthesize knowledge gained from laboratory and field studies, and to project the consequences of reasoned hypotheses as to how ecosystems and their components interact with changes in the environmenP”.
( I I Fundamental changes in the functioning of given ecosystems. For example, would one expect the productivity of the boreal forests of the world to increase or decrease? (2) Changes in the locations of boundaries between major ecosystems. What causes the boundaries between certain ecosystems to be sharp and those between other ecosystems to be relatively gradual? Are the positions of boundaries 304
Static approaches Perhaps the simplest approach to predicting climate-induced changes in vegetation is to apply the plantgeographers’ algorithms relating vegetation and climate to a predicted climate change (see Box II’2. Assessing the consequences of GCM (general circulation model) predictions based on increased atmospheric CO, using this approach, one finds that about 30% of the earth’s terrestrial surface might be expected to change from being occupied by one broad vegetation type (such as temperate grassland or deciduous forest) to being occupied by another. As mentioned above, this result is sensitive to the degree of specificity in the veg-
etation classification systems’3,‘4. The more categories that are recognized, the greater the terrestrial surface area that changes categories under a given climate-change scenario. Furthermore, since these changes are based on observed correlations between vegetation and climatic conditions, the mechanisms that might actually cause the projected responses are not explicitr4. Thus, there is a need to augment these predictions with methods based on a better understanding of the mechanisms that might be involved’5. One such method is to apply ecological models of forests and other ecosystems at certain locations: this provides an indication of the local reliability of the global predictions. Several models that are being developed for this purpose are based on the responses of individual organismsi5. These models are furthest developed for forest ecosystems’6.
Forest models based on individual organisms Ecosystem models can be used to synthesize what is known about the mechanisms that could affect the responses of ecosystems to novel environmental conditions - both the direct effects of an altered ambient CO, concentration, and the potential effects of other climatic changes induced by increased CO, in its role as an atmospheric greenhousegas. Models that simulate the individual plant - its growth, birth and death-have been used to predict the dynamics of many forests and some grassland ecosystems. Huston et a/.r7 point out that one advantage of such models is that two implicit assumptions associated with the more traditional statevariable approach used in ecological modeling of populations are not necessary. These are the assumptions that ( 1I the unique features of individuals are sufficiently unimportant that the individuals in the population can be assumed to be identical; and (21 the population is perfectly mixed so that there are essentially no local spatial interactions. These assumptions seem inappropriate for vascular plants, which are sessile, and in particular for trees, which vary greatly in size over their lifespanI Ix.
TREE vol. 5, no. 9, September
Perhaps for this reason, computer models that simulate the dynamics of a forest by following the fate of each individual tree were developed, initially in the mid- 1960~‘~. The earliest such model was put forward by Newnham in 1964’6,‘9. Foresters at several institutions proceeded in the late 1960s and early 1970s to develop additional models of forest change16,19. Digital computer programs that produced dynamically changing maps of the size and position of each tree in a forest became increasingly used as more computer power became available to ecologists16,19. One class of individual-based tree models that has been widely used in ecological (as opposed to traditional forestry) applications is the so-called ‘gap’ models’6~20. These models simulate the establishment, diameter growth and mortality of trees over usually annual intervals and on a simulated area of defined size (about 0.1 ha). This general approach has been applied to a wide variety of forested systems around the world (Fig. 1). Most of the models indicated in Fig. 1 are variants of the FORET model*‘. ShugartZo provides a more detailed discussion of these models, their derivation and testing. Whitmore ** further elaborates theories about the dynamics of forests (particularly tropical forests) that share a common basis with assumptions in this class of models. The FORET model was originally derived from the JABOWA model23 which, in turn, is an ecological version of the earlier forestry models based on individual tree dynamics16.‘9. The JABOWA model does not consider the two- and threedimensional geometry of tree-totree interactions as did the earlier forestry models; this reduction in detail is balanced by a simplification in estimating the model parameters20,2’.
Testing the models In applying models to simulate the effects of a changed climate on vegetation, one looks to the ways in which the models have been tested to increase confidence in the results20. Conducting field experiments to inspect predictions of responses over tens or hundreds of years is not a realistic option’4.‘5.
Fig. I. Calibration sites for a series of functionally similar individual-based models of vegetation. Black circles indicate locations for published ‘gap’ models of forest vegetation: gray circles indicate sites where work on forest models is in progress. The lightest circles indicate calibration sites for models of vegetation dynamics using an individual-organism approach for vegetation other than forests (sites include both grasslands and savannas).
However, gap models have been subjected to a variety of tests that provide insights into their reliability in predicting the effects of climate change’6,20. Principal among these tests are the following. ( I ) Simulation of natural gradients. One can determine whether the forests predicted by a simulation model when given different environmental conditions agree with observed patterns of forests. Included in such tests are simulations of vegetation change along altitudinal gradients20z2’26. One can also predict the forest that should occur at several locations with different conditions, and test the predictions against observed patterns (see Box 2)20,27. (2) Prediction of the results of natural experiments. In some cases, chance events create conditions that resemble experiments. Usually such events are not as well designed or as controlled as one would like, but they do provide data for model testing. Box 3 illustrates how a blight that eliminated a major tree species was used as a natural experiment to test the FORET model”. (3) Hindcast validation. In some cases, the prediction of past forests from past conditions can be used as a model test8,20,28.Individual-based tree models have been used to predict the forests under ice-age climates20 and at other times in the past8.9r20,28. The ability to predict the
forests of the past from the climates thought to have occurred then is an indicator of the ability of a model to predict future forests from future climates. In general, the results from testing individual-based models of forests are encouraging with respect to using them to evaluate the effects of future climate change on forests.
TREE vol. 5, no. 9, September
The models are largely untested as regards the direct effects of changes in atmospheric CO, levels29, but there have been extensive tests in terms of using the models to predict forest change under the influence of changes in temperature and moisture8,9,20*28. Model tests that involve inspecting the predicted average patterns of forests in comparison to observed average conditions over areas of 1O-100 ha have been more successful than those at finer spatial resolution20. The models can indubitably be improved by the addition of increased biological detail (particularly plant and plant-animalphysiology30 microbe interactions). Applications of individual-basedmodels to large-scale problems The importance (or at least the potential importance) of inter306
actions at the spatial scale of an individual tree in understanding forest-environment feedbacks implies that these ought to be considered in interfacing global change issues with forest ecosystem dynamics. Indeed, there are several promising starts in applying these models at the continental level. Drawing from the performance of the FORET model’b,*O in predictas well as other ing historical*” changes (see Box 31 in the deciduous forests of North America, Solomon used an extended version of the model (FORENA) to predict the change in forest vegetation in eastern North America in response to a doubled-CO, climateR. The FORENA model was subsequently used to inspect the response of forests to a change in disturbance regime’ associated with a climate change. A similar model was used to look at changes in nitrogen cycles and forest patterns under an altered climate”. A boreal forest simulator2’, tested in predicting important boreal forest patterns (see Box 21, has been applied to inspect the complex interactions among soil processes, climate change and vegetation dynamics for the boreal forests near Fairbanks, AlaskalO. The JABOWA3’ model has been used to quantify the inertia in the response of a forest to climatic change. A modified version of the FORET model has demonstrated the possibility of hysteresis - the condition that the response to change is history dependent - in the response of forests to a changing climate3*. There has been some application of these models to inspect direct effects of CO, on forest dynamicslO, but the addition of mechanistic physiological responses in the models is needed to make these applications more than conjectural. The expression, ‘unable to see the forest for the trees’, implying an over-attention to detail in the face of the obvious, comes to mind when one is attempting to predict largerscale dynamics from small quantum behavior. This trepidation is allayed to some degree by the impressive power of modern digital computers. Nevertheless, individual-organism models are limited by their fundamental spatial scale, and probably will continue to be used in coniunction with other models of forests in
global ecosystem studies. Earlier, I referred to such models as a basis of a special theory of forest dynamics because there isa limitation on generality that attends the underlying model assumptions20. Model scaling issues The current discussions of biospheric dynamics and global ecology I4 ‘5,33come at a time when there is a renewed interest in temporal and spatial scales in ecological systems34-3h. An appreciation of scales is a clear prerequisite to unifying the dynamics of atmospheric and oceanographic processes with the dynamics of ecosystems on the terrestrial surface. Of particular importance is a knowledge of the patterns of dominance (in the sense of importance in controlling responses) of particular causal factors at particular scales. The interweaving of models of differing fundamental scales is a problem of considerable difficulty3’. In many fields using dynamic models, so-called stiff system problems (in which the time constants for important processes span several orders of magnitude) can exceed the capacity of modern digital computers. Interestingly, both production and decomposition modeling can lead to stiff systems of equations if fast processes that influence production and decomposition (e.g. biophysical responses of the leaf surfaces with response times of seconds’ or minutes’; microbial growth or soil chemical kinetics with similarly fast responsivitiesl are coupled with slow processes (e.g. tree mortality; soil organic matter dynamics, soil genesis]. As a numerical problem, stiff systems are sometimes solved by separating the components into one system of ‘fast’ variables and another of ‘slow’ variables that can be evaluated separately. In ecological modeling this same procedure is applied in something of an ad hoc manner, in the assumptions regarding what processes can be included in or excluded from a given model formulation. One would hope that the relatively fast-response models can be interfaced to some degree with relatively slow-response models. Such an interfacing seems essential to evaluate the direct effects of increased CO, on vegetation over
YREE vol. 5, no. 9, September
large areas. For ecological models, this is also in the interests of climate modelers, who would like to have dynamic models of the response of vegetative canopies at a spatial scale (about 100 x 100 km) that exceeds that of physiologically based models, but on the temporal scale of many physiological models (minutes and hours]. While it is clearly important that, in developing vegetation models, ecologists have an initial interest in understanding the time and space scales of ecological phenomena, we must also realize that scientists in other fields (e.g. atmospheric sciences, oceanography) are increasingly posing modeling problems for ecologists that are in unfamiliar parts of the space and time domain. Acknowledgements I thank the US National Aeronautics and Space Administration (Grant NAG-5-10181, the US Environmental Protection Agency (Grant CR-816267-01-O) and the US National Science Foundation (Grants BSR-8702333 and BSR-88078821. The data for Box I were provided by Dr Rik Leemans of the International institute for Applied Systems Anafysis, Laxenburg, Austria.
I Givnish, T.). II9861 On the Economy of Plant Form and Function. Cambridge University Press 2 Holdridge, L.R. 119671 Life Zone Ecology, Tropical Science Center, San Jose. Costa Rica and f/ant 3 Box, E.O. I I98 I I Macroclimate Forms: An Introduction to Predictive Modeling in Phytogeography, funk Publishers
4 Woodward, Distribution,
Ft. (1987) Climate and Plant Cambridge University Press
5 Baes, C.F., Ir, Goeller, H.H., Olson, J.S. and Rotty, R.M. 11977) Am. Sci. 65, 3 IO-320 6 Bolin. B., Do&, B.R., lager, f. and Warrick. R.A.. eds i 1986) The Greenhouse Effect, Climatic Change and Ecosystems fohn Wiley 7 COHMAP Members 11988) Science 241. 1043-1052 8 Solomon, A.M. f 1986) Oecologia 68, 567-579 9 Overpeck, J.T., Rind, D. and Goldberg, R. II9901 Nature 343, 51-53 IO Bonan, C.B., Shugart, H.H. and Urban, D.L. (19901 Clim. Change l6,9-31 II Pastor, f. and Post, W.M. I I9881 Nature 334, 55-58 12 Mitchell, I.F.B.. Senior, C.A. and Ingrahm, W.f. 11989) Nature 341, 1558 I3 Emanuel, W.R., Shugart, H.H. and Stevenson, M.P. t 1985) Qim. Change 7, 29-43 I4 Warrick, R.A., Shugart, H.H. and Antonovsky, M.Ya., with Tarrant, f.R. and Tucker, Cf. ( 19861 in The Greenhouse Effect, Climatic Change and Ecosystems (Bolin, B., Does, B.R., lager. I. and Warrick, R.A., eds), pp. 363-392, John Wiley I5 Shugart, H.H., Antonovsky, M.Ya., Jarvis. P.G. and Sandford, A.P. II9861 in The Greenhouse Effect, Climatic Change and Ecosystems IBolin, B.. Doas, B.R., lager, 1. and Warrick, R.A., eds), pp. 475-52 I ( fohn Wiley I6 Shugart, H.H. and West, D.C. II9801 Bioscience 30. 308-3 I3 I7 Huston, M.A.. DeAngelis. D.L. and Post, 38,682-692 W.M. ( 19881 Bioscience 18 Shugart, H.H. and Urban, D.L. f 1989) in Toward a More Exact Ecology IGrubb, P.J. and Whittaker, J.B., eds), pp. 249-274, Blackwell I9 Munro, D.D. (I9741 in Growth Mode/s for Tree and Stand Simulation (Fries, 1.. ed.), pp. 7-21, Royal College of Forestry, Stockholm 20 Shugart, H.H. f 1984) A Theory of forest Dynamics: The Ecological Implications of Forest Succession Mode/s, Springer-Verlag
21 Shugart, H.H. and West, D.C. ( 1977) 1. Environ. Manage. 5, 161-170 22 Whitmore. T.C. (1982) in The P/ant Community as a Working Mechanism (Newman. E.I., ed.), pp, 45-60, British Ecological Society and Blackwell Scientific Publications 23 Botkin, D.B., lanak. l.F. and Wallis, l.R. I 197211. Ecol. 60,849-873 24 Shugart. H.H. and Noble, I.R. f 1981) Aust. 1. Eco/. 6, I 49- I64 25 Kienast, F. I 1987) FORECE - A forest Succession Model for Southern Central Europe, Oak Ridge National Laboratory fORNUTM 10575) 26 Kercher, 1.R. and Axelrod, M.C. ( 1984) Ecology 65, 1725-l 742 27 Bonan, G.B. II9891 Ecol. Mode/. 3, Ill-130 28 Solomon, A.M., West, D.C. and Solomon, I.A. ( I98 I 1in forest Succession Concepts and Application (West, D.C., Shugart, H.H. and Botkin. D.B., eds), pp. 154-l 57. Springer-Verlag 29 Strain, B.R. 11985) Biogeochemistry I, 219-232 30 Landsberg. 1.). II9861 Physiological Ecology of Forest Production, Academic Press 31 Davis, M.B. and Botkin, D.B. (1985) &rat. Res. 23, 327-340 32 Shugart, H.H., Emanuel, W.R., West, D.C. and DeAngelis, D.L. 11980) Math. Biosci. 50, 163-170 33 Risser, PG. II9861 Spatial and Temporal Variability of Biospheric and Geospheric Processes: Research Needed to Determine interactions with Global Environmental Change, International Council of Scientific Unions Press 34 Smith, T.M. and Urban, D.L. (1988) Vegetatio 74, 143-l 50 35 Defcourt, H.R., Delcourt, P.A. and Webb, T., Ill f 1983) Quat. Sci. Rev. I, 153-l 75 36 O’Neill, R.V., DeAngelis. D.L., Waide, 1.B. and Allen, T.F.H. 11986) A Hierarchical Concept of the Ecosystem, Princeton University Press 37 Urban, D., O’Neill, R.V. and Shugart. H.H. II9871 Bioscience 37, 119-127