SCOPE 29 - The Greenhouse Effect, Climatic Change, and Ecosystem

10.1 INTRODUCTION

Potential responses of forests to the direct effects of increasing CO₂ and climatic change should be considered in the context of forest management and the trends in wood supply and demand. The total amount of carbon stored in terrestrial ecosystems has diminished over the past several centuries as a result of anthropogenic actions, especially forest clearance (see Bolfin, 1977; Woodwell et al., 1978; Houghton et al., 1983). In the last half of the 1970s, about 2400 million m³ of wood were consumed annually (FAO, 1982). About half of this wood was used as fuelwood and the other half for industrial purposes (lumber, plywood, chipboard, paper, etc.). For the remainder of this century, FAO (1982) projects an increasing rate of forestclearance, based on forecasts of rising demands for wood. If this occurs, the present area of forest may be reduced by as much as 20% by the year 2000. Problems of wood scarcity could become critical in regions such as Africa and Asia where fuelwood is the primary source of energy for domestic heating and cooking. Whether increasing CO₂ and climatic change will exacerbate or diminish the problems that loom on the horizon is a fundamental issue which concerns the entire global community.

In this chapter we examine the possible responses of forests to changes in climate and the direct effects of increased atmospheric CO₂ concentrations. Change or stress can induce responses in a wide range of frequencies from a forest ecosystem; these are reviewed in Section 10.2. We review the mechanisms involved in the direct CO₂ effect in Section 10.3. Since prediction of the responses of forests may well involve the use of models, the types of models available are reviewed in Section 10.4. The results of applications of such models and other methods to assess the impacts of climatic change are presented in Section 10.5.

10.2 TIME SCALES AND THE RESPONSE OF FORESTS

Forest ecosystems contain a complex web of interactions among physical, chemical and biological processes. Because of this interactive complexity, direct changes in a given process can be attenuated or amplified, and the responses elicited from a forest will be manifested on many different time scales. In this section, we discuss the responses of forests at various scales of time and space as a necessary preamble to considering the possible effects of increasing CO₂ concentrations and climatic change.

10.2.1 Fast Responses

Several of the important processes in forest ecosystems that have very rapid response times are influenced by CO₂ and climatic variables. Among the most important are those involving the exchanges of water, heat and carbon dioxide between the leaf surfaces and the environment. For example, stomata (microscopic pores on the leaf surface) allow the inward diffusion of carbon dioxide used in photosynthesis and, at the same time, allow the loss of transpired water. An increase in the ambient CO₂ concentration could reduce the opening of the stomata required to allow a given amount of CO₂ to enter the plant and might thus reduce the loss of water from a tree. This could increase the efficiency of water use and raise the productivity of forests in certain circumstances. Changes in the radiation input, temperature or humidity above a forest canopy could also produce almost instantaneous responses in CO₂ uptake and water use by the forest. The rapid, direct responses to CO₂ are considered in more detail in Section 10.3.

The question that arises immediately is, to what degree would these fast responses change the amount of photosynthate and ultimately the amount of production over the course of a year in a forest? Tree growth results from the amount of photosynthate produced and the allocation of this photosynthate within the tree. The growth of trees has been modelled using `mechanistic' representations of physiological processes, but these models have rarely been used to predict responses over periods longer than a year. The complexity of the problem is illustrated in Figure 10.1, which shows how opposing environmental conditions (high temperature and low precipitation; low temperature and high precipitation) may produce the same response: the production of a narrow annual ring in trees growing in arid sites. It is important to note: (1) that the web of interactions is complex and the details of some of the interactions are not well known, and (2) that the response of the whole plant to a stress may be non-linear across the possible range of that stress. In general, the fast processes that operate in forest ecosystems can only be predicted over the longer term with a considerable degree of uncertainty.

There have been several attempts to develop highly detailed `mechanistic' models of natural ecosystems including forests. These models are useful as heuristic tools for integrating studies of different ecological processes, but are much less useful for predicting long-term ecosystem behaviour. At the Third Conference on CIAP (Cooper et al., 1974), results were presented on the performance of several of such models (developed during the International Biological Program for different types of ecosystems) when subjected to a change in climate. It was noted that 'The models are not designed to predict long term successional trends or changes in plant or animal composition... The simulation runs are therefore limited to a 5-year period and emphasize changes in primary production and in associated aspects of community structure. It is questionable how far beyond this 5-year period the results can validly be extrapolated.' (Cooper et al., 1974). These cautionary comments still apply to current models of ecosystem processes. 

10.2.2 Intermediate Time-scale Responses

Environmental change can affect the growth rates of individual trees and thereby have a cumulative effect in changing the amount of living material in the forest system. However, the relation between the rate of individual tree growth and the rate of forest increase (or yield) is more than a simple additive effect. Relatively low levels of stress on trees can produce large changes in the dynamics and composition of forests. Furthermore, the interactions between the populations of trees and phytophagous insects (and other pests and diseases) are, in many cases, mediated by climate and thus are of importance in assessing the possible effects of climatic change on forests. For example, the oak wilt disease in the USSR appears to be dependent on the decreased ability of the trees to resist leaf-eating insects during drought (Israel et al., 1983).

The prediction of yield from growth has been an important topic in modern forestry (Fries, 1974). Forest yield is a consequence both of the growth of individual trees and of the rates of recruitment and death of trees in the forest stand. For example, Figure 10.2 illustrates the relation between stand biomass and the age of stands of Picea glauca forests (Yarie and Van Cleve, 1983). Over the time period indicated by the different ages, the growth increment of the trees was constant (diameter increase = 0.11 cm/year; r² of regression = 0.87) but the rate of increase of the total biomass clearly declined as a function of age. In contrast, the rate of biomass change of a single tree, which was enlarging with a constant diameter increment, increased with the size of the tree (because tree biomass is a function of the diameter). Thus, the change in biomass growth rate shown in Figure 10.2 was opposite to that of the individual trees that comprised the forest. Effects such as these prevent direct extrapolation of short-term changes in trees to predict the longer term responses of the forest.

Figure 10.1 Diagrams representing some of the relationships that cause climate variables to lead to the formation of a narrow ring in trees growing in arid sites. (a) The effect of low precipitation and high temperature during the growing season. (b) The effect of low precipitation and high temperature before the growing season. (c) The effect of high precipitation and low temperature (from Fritts, H.C. 1976).

Figure 10.2 The relationship of above ground stand biomass to stand age for white spruce (Picea glauca) from the Interior of Alaska (from Yarie and Van Cleve, 1983). A fully stocked stand has a site density index (SDI) of 1000, a stand with one half this density of trees has a SDI = 500

The modern theoretical concept for understanding the intermediate time scale response of a forest is to consider the forested landscape to be a mosaic with each element of the mosaic scaled in relation to the dominant canopy tree (ca 0.1 ha, depending on the size of the trees). This concept was initially developed by Watt (1925, 1947) and has been the topic of a series of papers and books (Raup, 1957; Whittaker and Levin, 1977; Bormann and Likens, 1979a,b; Shugart, 1984). The concept, in brief, is that the dynamic response of the forest occurs at an areal scale of a large canopy tree and on a time scale that relates to the longevity of the tree, as follows.

'Following the death of a large tree and its fall, a canopy gap forms. The area below this gap becomes the site of increased regeneration and survival of trees. Trees grow, the forest builds, the canopy closes, and the gap disappears. Eventually, the now mature forest in the vicinity of the former gap suffers the mortality of a large tree and a new gap is formed and the cycle is repeated (Shugart 1984).'

The dynamics of a forest are the aggregated dynamics of a large number of such individual gaps. When considered at intermediate time scales (ca 100 years), the pattern of dynamics of a forest can be seen as a cycle of recruitment, death and growth processes; environmental change alters the pattern within the cycle (Figure 10.3). The regeneration phase of the forest cycle and the exact timing of the death of a canopy tree that produces an opportunity for regeneration are highly stochastic processes. This is particularly the case in the regeneration phase when the mortality of small trees is very high. It is in this stochastic part of the cycle that climatic change and variability can have the largest effects in producing change in the forest. For example, a dry, hot year could remove the seedlings of drought-intolerant tree species at the local site of regeneration. It might then be several hundred years (the longevity of a canopy tree) before the eliminated species would again have a reasonable chance to occupy the same area. During the growth, competition and thinning phases of the forest cycle, the dynamics of the forest are much more determinate and the change of the forest in this part of the cycle in response to a climatic change has considerable inertia. In' summary, the forest system has great inertia but features short periods when relatively small outside perturbations can greatly alter the eventual path of the system.

Figure 10.3 Simulation of the long-term regeneration cycle of forests using the BRIND model (Shugart and Noble, 1981) for the Australian Eucalyptus forest. The determinism in the system is great during the growth, competition and thinning phases but more stochastic in death and regeneration phases

Figure 10.4 Migration maps for three species. The numbers refer to the radio-carbon dates (in thousands of years before present) of the first appearance of the species in a site, as evidenced by an increased pollen abundance or the presence of macrofossils at a site. Isopleths connect points of similar age and represent the leading edge of the expanding population since the Wisconsin glaciation. The dotted area is the modern range of (a) Castanea dentata, (b) Fagus grandifolia, (c) Tsuga canadensis. (From Davis 1981)

10.2.3 Long-term Responses

At the intermediate scale, the successful regeneration of the species suited to particular climatic conditions depends on an adequate supply of seeds. However, if a change in climate was sufficiently large to extend the climate favourable for a species far beyond its present limits, the rate of response to the change would become dependent on the migration rate of the species. Such delays in response have been inferred from data on past climatic changes. The changes in geographical range of different species over time following the last glaciation have been calculated from fossil pollen data. In Figure 10.4, the time required for a species to move between the lines that indicate range boundaries was 1000 years. There is considerable variation in the rates of migration that have been recorded for various taxa (Table 10.1), so the influence of migration in delaying the local response of a forest is potentially large.

Table 10.1 Migration of European fossil pollen taxa (after Huntley and Birks, 1983)

It is, of course, possible to reduce large, migration-related lag effects by actively moving plant material (seeds and seedlings) to appropriate locations. Suitable propagation techniques to accomplish such a task are better developed for commercially important tree species than for non-commercial species. Such a strategy probably would be prohibitively expensive if pursued on a large scale with native forests. The actual movement of tree species to compensate for a change in climate could most likely be attempted for commercial tree crops only, in a forest management context.

Change in the rate of soil development could also greatly delay the response of a forest to a climatic change. The soil at a given location derives its characteristics from the parent material (the geology at the site), the vegetation and the climate. If both the climate and the vegetation at a given location were to change, there might be considerable delays in the development of the soils and, hence, the forests which one would expect ultimately to develop at the site.

10.2.4 Responses to Climatic Changes in the Past

A very large amount of evidence shows that the world's forests have responded, often dramatically, to changes in climate on various time scales. Even such a small fluctuation as the high-latitude period of warm summers during the 1930s, when mean annual temperatures were typically no more than 1 °C above the average temperatures for previous and subsequent decades, caused a burst of regeneration in boreal forest trees near their polar and altitudinal limits, with measurable advances of the timberline (e.g. Hustich, 1958; Morisset and Payette, 1983). The 'Little Ice Age' was a longer period, roughly between 1400 and 1800, during which there were glacial advances in many mid- and high-latitude regions, with recorded or reconstructed mean temperatures in Europe and North America of ca 12 °C below present values. The `Little Ice Age' had substantial effects on forest growth and composition in many northerly areas. For example, in the prairie-forest border region of Minnesota, an increase in precipitation relative to evaporation caused a reduction in the frequency of fires; this, in turn, favoured a forest with more big hardwoods than the oak forest which it replaced. This change took only about a century to accomplish (Grimm, 1983).

Larger climatic changes have taken place since the peak of the last major glaciation ca 20,000 years ago. These changes caused whole biomes to be replaced. The evidence for long-term changes in vegetation comes from analysis of fossil pollen in sediments and other research techniques. Maps based on fossil pollen data show the major patterns of change (e.g. Huntley and Birks, 1983; Peterson, 1983). Some documented changes have been remarkably rapid. For example, the transformation from spruce-dominated to jack pine-dominated forests in eastern Minnesota ca 10,000 years ago took a few hundred years at most, and perhaps less than one hundred years (Wright 1984).

10.3 THE DIRECT IMPACTS OF INCREASING CO₂ CONCENTRATION 

10.3.1 The Base of Experimental Evidence

There has been a sufficient number of studies on CO₂ exchange of the dominant tree species and some of the understorey species in northern forests to support the assertion that all the species of major ecological and economic importance are likely to have photosynthesis of the typical C₃ pattern. Therefore, we expect the main effects of rising CO₂ concentrations to result from changes in photosynthesis and stomatal action as has been shown for other C₃ species (see Section 9.2). One can not rule out other direct effects of CO₂ on, say, root and leaf development or CO₂ assimilation in the dark by roots, but little is known about such responses and they are most probably of secondary importance. The likely effects of changes in photosynthesis, stomatal conductance and leaf growth on the growth of trees are indicated in the network diagrams in Figure 10.1.

There is a great dearth of information on the effects of CO₂ concentration on the physiology and growth of particular, relevant forest species. Only a few short-term experiments have been conducted on single leaves, shoots or small plants in assimilation chambers where the response of CO₂ uptake and stomatal action has been characterized in relation to ambient CO₂ concentrations and, in even fewer cases, to mean intercellular space CO₂ concentration. Many measurements have been made on leaves and shoots and sometimes on whole trees using cuvettes in the field, but in these experiments CO₂ has not usually been a controlled variable and higher than normal CO₂ concentrations have not usually been used. There have also been several series of measurements of CO₂ influx from the ambient atmosphere to stands of conifers and broad-leaved species using micrometeorological methods but these have, of course, all been at normal ambient CO₂ concentrations.

A very small number of experiments at elevated CO₂ have been performed on the growth and water use of seedlings and small trees over periods of weeks in controlled environment facilities or in glasshouses, plastic tunnels and open-topped chambers in the field. However, as far as we are aware, there have been no long-term experiments in the field on the response of trees, stands or other parts of forest ecosystems to elevated CO₂ concentrations. To say anything at all about the likely responses of stands or ecosystems to elevated CO₂ concentrations, the only approach available to us is to rely on information concerning short-term responses of processes in leaves and shoots (see Section 10.2.1), and, by using models, to extrapolate up the scales of time and space. This approach offers some hope for the future, but there are formidable problems involved in deriving realistic estimates of forest growth with a 'bottom-up' modelling approach (as discussed in Section 10.3.4).

In the following sections we review the information available concerning responses of CO₂ assimilation, stomatal action and growth of forest species to elevated CO₂ concentrations. We shall begin with CO₂ assimilation and stomatal action at the leaf level and then consider the effects on growth on longer time scales and larger area scales.

10.3.2 Fast, Short-term Responses at the Leaf Level

In short-term experiments, the rate of CO₂ assimilation (A) has been shown to increase more or less linearly over the range of ambient CO₂ concentrations (C_a) from 20 to ca 400 ppmv (e.g. Zelawski, 1967; Brix, 1968; Kriedemann and Canterford, 1971; Ludlow and Jarvis, 1971; Luukkanen and Kozlowski, 1972; Regehr et al., 1975; Gross, 1976; Watson et al., 1978; Masarovicova, 1979; Beadle et al., 1981) in both conifers and broad-leaved species, before starting to level off (Figure 10.5). The onset of significant levelling off is variable, occurring anywhere between 300 and 600 ppmv in different species and may be followed by a sharp decline at ca 1500 ppmv (Koch 1969).

In some conifers, stomatal conductance (g_s) is relatively insensitive to the ambient CO₂ concentration (C_a) compared with the general response of g_s to C_a in other species that use the C₃ photosynthetic pathway (e.g. Beadle et al., 1979; Morison and Jarvis, 1983). This may explain in part why A continues to increase steeply up to high values of C_a in conifers rather than levelling off at lower C_a as in many other C₃ species (Figure 10.6).

We may expect varied responses of A to C_a depending on the capacity of the photosynthetic apparatus to fix CO₂, relative to the rate of supply of CO₂ to the reaction sites. The rate of assimilation depends on two factors. The first is the activity of the photosynthetic enzymes, particularly ribulose-bisphosphate carboxylase, and the enzymes in the carbon reduction cycle that limit the regeneration of the ribulose-bisphosphate acceptor. The second factor is the concentration of CO₂ in the air bathing the mesophyll cells (C_i), since that determines the rate of supply of CO₂ to the reaction sites. C_i is, of course, related to C_a depending on the resistance in the CO₂ transfer pathway between the bulk atmosphere overhead and the intercellular spaces, particularly the resistance of the stomata. These relationships are evident in the curve describing the relationship between A and C_i, the so-called A/C_i curve (Figure 10.7). Through the A/C_i curve we may seek an explanation of the varied responses of A to C_a.

Figure 10.5 The relationship between assimilation rate (A), stomatal conductance for CO₂ (g_s) and ambient CO₂ concentration (C_a) for unstressed Populus deltoides. Quantum flux density 2000 µmol m^-2 s^-1; temperature 30 °C; water vapour saturation deficit 1.7 kPa. Vertical bars indicate one standard deviation. Recalculated from Regehr et al. (1975)

Figure 10.6 The relationship between assimilation rate (A), stomatal conductance for CO₂ (g,) and ambient CO₂ concentration (C_a) for unstressed Picea sitchensis. Quantum flux density 1000 µmol m^-2 s^-1; temperature 20 °C; water vapour saturation deficit 0.6 kPa. Each point is based on at least five measurements. Recalculated from Beadle et al. (1979, 1981)

Figure 10.7 The relationship between assimilation rate (A) and mean intercellular space CO₂ concentration (C_i) for unstressed Picea sitchensis. Quantum flux density l100 µmol m^-2 s^-1; temperature 20 ^°C; water vapour saturation deficit 1.0 kPa. The arrow indicates the value of C_i at the current atmospheric concentration of C_a. The line has been fitted in two sections to a total of 72 data points. Unpublished data of the authors

In Figure 10.8 the point X is the rate of assimilation corresponding to the indicated ambient CO₂ concentration (C_a)and the intercellular space CO₂ concentration (C_i). The line Z represents the `demand function' of assimilation for CO<sub>2</sub>: the line C<sub>a</sub>D represents the `supply function' (Jones, 1973; Raschke, 1979). If the A/C_i curve is approximately linear up to the operating point X, the demand function can be represented by

A = g_m(C_i - )

where g_m is carboxylation efficiency or a mesophyll conductance. (If the relationship is not more or less linear over this range, a more complex treatment is requiredsee Farquhar and Sharkey, 1982; Jones, 1983). Above the operating point, A becomes a function of the capacity to regenerate ribulose-bisphosphate and reaches an asymptote (A_max). The supply function is represented by

A = g_c(Ca - C_i)

where g_c is the overall conductance of the CO₂ transfer pathway from bulk atmosphere to intercellular spaces and is approximately equal to the stomatal conductance, g_s, for a well-ventilated, isolated leaf.

Figure 10.8 A diagram to show the interrelationship between the 'demand function' and the 'supply function' of assimilation. The line Z is a relationship between assimilation rate (A) and mean intercellular space CO₂ concentration (C_i) of the kind shown in Figure 10.9. The other lines show how A may vary depending on changes in the demand function (e.g. X ®D or C), the supply function (e.g. C_a X ® C_a W or C_a Y), or the ambient CO₂ concentration (e.g. C_aX® C_a' Z) or combinations of these three variables (e.g. C_a® C_a' leading to decreases in both the demand and the supply functions but X rising to X')

These functions have the following consequences (assuming for the moment that only one variable changes at a time):

• if the carboxylation efficiency changes, A will change along the line CD; if g_m increases to g'_m, A will increase from X to D.

• if the stomatal conductance changes, A will change along the line WXY; if g_s decreases to g'_s, A will decline from X to W.

• if the ambient CO₂ concentration changes, A will change along the line WXYZ; if C_a increases to C'_a, A will increase from X to Z.

If, for example, there are no changes in g_s or in the demand function, we would expect an increase in C_a to lead to an increase in A along the line XYZ. This is very much the case in Picea sitchensis where g_s is largely unaffected by CO₂, so that the response of A is almost linear for C_a between 75 and 600 ppmv (Beadle et al., 1979, 1981). If, however, an increase in C_a were to change the supply function, a smaller increase in A might result. For instance, if the stomata were to close somewhat in response to the increase in C_a, C_i and, hence, A would increase less than in the previous example (e.g. only to Y). In short, it is evident that the increase in A depends on the sensitivity of the stomata to CO₂. Therefore, we would expect the largest, short-term increase in assimilation to occur in plants such as the conifers, which have stomata that are relatively insensitive to CO₂. Smaller, shortterm increases in assimilation would be expected to occur in species like Populus deltoides that have stomata which are quite sensitive to CO₂.

Other environmental variables may also influence g_s and lead to changes in C_i. For example, in many tree species, both broad-leaved and conifers, the stomata are sensitive to the water vapour saturation deficit of the air (D) (e.g. Appleby and Davies, 1983). As a result, the ratio of internal to ambient CO₂ concentrations (C_i/C_a) decreases with increasing D (Morison and Gifford, 1983). The climatic changes associated with the global increase in CO₂ are very speculative but it seems likely that middle latitudes may become less humid. If an increase in D is correlated with the increase in CO₂, there may be very little change in assimilation rate. Environmental stresses, such as low light (e.g. Morison and Jarvis, 1983), low water (e.g. Beadle et al., 1981; Jones and Fanjul, 1983) or low nutrients (e.g. Wong, 1979) may displace the A/C_i curve downwards, reducing either or both mesophyll conductance (g_m) and A_max (see Figure 10.9) and thus reducing the opportunity for A to respond to enhanced CO₂.

It is clear from the preceding discussion that evaluation of the response of assimilation to a change in CO₂ requires knowledge of the A/C_i curves for a range of environmental and physiological conditions. This information is hardly available at the present time for plants grown in controlled environments, let alone for plants grown in the field. One notable exception is a recent experiment in which the CO₂ uptake of a large Eucalyptus maculata tree growing on a weighing lysimeter in a forest was measured in relation to a range of CO₂ concentrations by enclosing the tree in a plastic tent (S.C. Wong, personal communication). Assimilation, stomatal conductance and transpiration of the tree responded to CO₂ in much the same way as would a leaf in an assimilation chamber in the laboratory.

All the experiments discussed above investigated short-term responses of assimilation and stomatal conductance in plants that were not given the opportunity to acclimate to higher CO₂ concentrations. There are very few published experiments in which the same responses have been studied in forest species that have been allowed to grow for extended periods in different CO₂ concentrations (e.g. Rogers et al., 1983). What experiments there are suggest that the response of A and g_s in plants acclimated to elevated CO₂ is qualitatively similar to the responses in plants grown in normal air, i.e. A increases and g_s decreases up to CO₂ concentrations of at least 1000 ppmv.

Figure 10.9 The relationship between assimilation rate (A) and mean intercellular space CO₂ concentration (C_i) for leaves of water-stressed apple trees cv James Grieve. The figures by the curves are the leaf water potentials. Quantum flux density 15002000 µmol m^-2 s^-1; temperature 1520 °C; water vapour saturation deficit 0.51.0 kPa. The arrows indicate the values of C; at the current atmospheric concentration of C_a. Each point is the mean of at least four measurements. From Jones and Fanjul (1983)

Studies of the quantitative effects of acclimation are rare. Data for plants other than forest species indicate that acclimation to elevated CO₂ leads to a reduction in certain photosynthetic enzymes (see review by Berry and Downton, 1982) and thus changes the demand function. As a result, A and g_s in air containing 300 ppmv CO₂ become less than they were before a period of acclimation to 600 ppmv (e.g. Wong, 1979). Nonetheless, the relevant observations are that long-term exposure to elevated CO₂ does result in higher A and lower g_s than in plants grown at current normal concentrations.

The combination of the effects of CO₂ on A and g_s may lead to a substantial increase in assimilation per unit of water transpired, the so-called water-use efficiency (i.e. A/E), for plants in enclosures. In Liquidambar stryaciflua, for example, a doubling of the CO₂ concentration more than doubled the water-use efficiency (Rogers et al., 1983; Tolley and Strain, 1985). Such a response may, of course, vary from species to species depending on the relative sensitivity of A and g_s to CO₂. Whether this response is necessarily likely to occur in field conditions is discussed in Section 10.3.4.

10.3.3 Intermediate Time Scale Effects on the Growth of Trees

There are very few reported experiments in which tree species have been grown for extended periods at elevated CO₂ concentrations. What experiments there are have been confined to the whole or part of one growing season (e.g. Sionit et al., 1985). We are not aware of any published experiments in which woody perennials have been grown at elevated CO₂ concentrations over several growing seasons. This is an important point because quite small differences in relative growth rate, if they persist for several years, lead to large differences in plant size.

The limited evidence available suggests that growth in height, leaf area and dry weight of temperate trees is increased by higher than normal CO₂ concentrations (Yeatman, 1970; Funsch et al., 1970; Krizek et al., 1971; Tinus, 1972; Laiche, 1978; Canham and McCavish, 1981; Lin and Molnar, 1982; Rogers et al., 1983; Tolley and Strain, 1984a and b). However, the results reported so far are extremely variable and in some species no growth response was elicited (Lin and Molnar, 1982). There are too few reliable data from which to generalize about the growth response.

On the basis of results with other species (e.g. Morison and Gifford, 1984a and b) and limited data for tree seedlings (Tolley and Strain, 1984b), we expect the primary consequence of an increase in A to be larger leaf areas that will increase interception of radiation and thus amplify the initial effect of enhanced CO₂. That is not to say that we suppose a direct effect of elevated CO₂ on leaf development; rather, we expect more rapid leaf initiation and lamina expansion as a result both of higher carbohydrate status and possibly higher turgor, as well as greater longevity of leaves. These feedback effects, we suppose, lead to the more rapid attainment and sustention of larger leaf areas and hence to higher growth rates. If leaf growth is inhibited, there may be no growth response to elevated CO₂. For example, ethylene, which inhibits cell division in expanding leaves, when found as a contaminant of artificially synthesized CO₂, completely negated the effect of doubling the ambient CO₂ concentration on plant growth (Morison and Gifford, 1984c). As far as we know, there are no clearly established initial effects of CO₂ on leaf initiation and development in herbaceous plants (e.g. Hurd, 1968; Hurd and Thornley, 1974; Sionit et al., 1981; Morison and Gifford, 1984b) or in trees.

There is little relevant information on the effect of elevated CO₂ on partitioning of assimilate. The evidence available suggests that partitioning is largely unaffected (Tinus, 1972; Rogers et al., 1983). More detailed studies on a number of herbaceous species (e.g. Morison and Gifford, 1984b) have demonstrated a consistent decrease in specific leaf area associated with increase in the number and size of cells in leaves, and some changes in leaf weight ratio but not in root to shoot ratio and somewhat similar responses seem to be found in tree seedlings. In a recent study, Thomas and Harvey (1983) found increases in leaf thickness, apparently resulting from the presence of both more and larger cells, in leaves of Liquidambar stryaciflua and Pinus taeda grown at elevated CO₂ concentrations.

10.3.4 Long-term Impact on Ecosystems

The possible effects of climatic change on the dynamics of species relationships in forest ecosystems are treated in Section 10.4. As far as the direct effects of CO₂ are concerned, we are less able to comment on the effects of an increase in CO₂ concentration on processes of regeneration, competition and species composition because of the lack of data. We know of no experimental studies on relevant species in which the interference between species has been investigated at elevated CO₂ in either field or laboratory. Long-term exposure of undisturbed vegetation to elevated CO₂ is certainly feasible. Unenclosed areas of agricultural crops have been exposed to controlled concentrations of SO₂ throughout an entire growing season in at least two localities and the same could be done with CO₂ in crops, artificial mixtures and natural vegetation, given adequate resources. However, at the present time such projects have not progressed beyond the discussion stage. As is always the case with forest vegetation, the problems are more formidable, both because of the larger scale and the close coupling between forest canopies and the bulk atmosphere.

Recently, two papers have appeared in which annual tree growth of trees in stands in the field has been analysed in relation to the age of the tree over the past 50 or so years and growth was found to be more than expected (LaMarche et al., 1984; Arovaara et al., 1984). It was suggested by the authors that this might be attributed to the increase in CO₂. Such conclusions must be regarded as highly speculative, since growth responds to many environmental variables.

An increase in rate of photosynthesis (and leaf development, if it occurs) would be expected to increase the rate of production of biomass (i.e. the productivity) and the standing crop in forest systems, like plantations, that have not yet achieved a steady state. In forests that have achieved a steady state there may also be increases in the rate of growth of individuals as a result of an increase in CO₂, but this may well not lead to an increase in the biomass of the standing crop. At a steady state the overall, average annual net exchange of CO₂ is zero, although the instantaneous rate of assimilation by some components is very high. This is because the assimilation of CO₂ is balanced by losses of CO₂ in respiration of the living biomass and through mortality and respiratory turnover of leaves, fine roots, branches and entire individuals (Jarvis and Leverenz, 1983). How a rise in ambient CO₂ concentration will affect the checks and balances in this complex system (see Figure 10.1) is very uncertain.

There are two major problems in trying to scale up to the growth and productivity of even comparatively simple stands such as coniferous plantations. First, we lack adequate physiological information to scale up from measurements of physiological processes made over periods of minutes or hours to estimate growth over periods of months or years. We are able to simulate adequately the CO₂ influx to a plantation, in the short term, using data on leaf photosynthetic properties and canopy structure in a model of canopy processes (e.g. Jarvis et al., 1985). However, we still lack sufficient physiological knowledge about processes such as assimilate allocation to make the jump from canopy CO₂ exchange, whether measured or predicted, to growth of individual trees or of the stand. Whilst we can run the same model with higher ambient CO₂ concentrations as input, and predict a new CO₂ influx, we lack sufficient knowledge concerning the feedbacks between CO₂ supply, leaf growth, nutrition and tree growth, to be able to say whether photosynthesis would continue at a higher rate and whether this would lead to a long-term increase in growth of the stand. It is very doubtful that we could, at the present time, even predict the long-term growth responses of young potted trees to elevated CO₂ that have been observed in growth-room experiments, from the short-term measurements of CO₂ uptake that have been made on leaves or shoots.

Second, we have scarcely sufficient micrometeorological knowledge about coupling between vegetation and the atmosphere to scale up from a leaf to a stand, let alone to an ecosystem on a regional scale. A leaf in an assimilation chamber or growth room is, by design, tightly coupled to an environment that is imposed upon it. The natural feedback loops between the leaf and the atmosphere are deliberately broken, so that the leaf has little influence over its own environment. In the field, the degree of coupling of vegetation to the atmosphere varies, depending on the area scale and the aerodynamic roughness of the vegetation. As the area scale increases in size, the degree of coupling between vegetation and atmosphere decreases, with the result that transpiration becomes less influenced by a change in stomatal conductance and more strongly dependent on the radiation input. On the regional or global scale the overall rate of transpiration is set by radiation and is not affected by changes in stomatal conductance. Consequently CO₂ cannot be regarded as a global antitranspirant (Jarvis and McNaughton, 1985).

Nonetheless, transpiration from aerodynamically rough vegetation (that is well-coupled to the atmosphere, like a tree canopy) will be reduced by stomatal closure, but at the expense of transpiration from poorly-coupled vegetation which will increase (McNaughton and Jarvis, 1983; Jarvis and McNaughton, 1985). Similarly, we expect CO₂ uptake of well-coupled vegetation such as trees to tend to increase in response to an increase in global CO₂ concentration, depending also on the response of the stomata, whereas the CO₂ uptake of largely decoupled vegetation, such as grasslands and field crops, and probably also the forest understorey, is likely to respond to a lesser extent because of feedback between the local ambient CO₂ concentration and the rate of CO₂ uptake.

Because of these problems in scaling up with respect to both time and space, we regard speculation about the likely effects of rising CO₂ on carbon assimilation and water use by forests as particularly unreliable at present. While empirical experimentation is, in principle, possible, the logistical problems render this impracticable with mature forests. Therefore, there is an urgent need to improve the basis for scaling up from leaf to region and from minutes to years.

10.3.5 Further Research Directions

It is clear from the foregoing survey that a shortage of data at all scales seriously limits an assessment of the likely effects of rising CO₂ on forests. Short-term, fast response data are the most readily obtainable and some progress may be made by scaling up such data with the use of appropriate physiological, canopy and micrometeorological models. For this purpose we need far more extensive characterization of the responses to CO₂ of assimilation, stomatal conductance and transpiration for plants that have been acclimated to high concentrations of CO₂, and for a much wider range of relevant species, than is available at present. However, because of the gaps in our physiological knowledge concerning the linkage between assimilation and transpiration of leaves and the growth and water use of whole plants, it is very necessary that the response of growth and water use to high CO₂ (and other variables such as nutrient and water stress) should be analysed over extended periods of time covering several growing seasons. Thus, we see a need for many more growth experiments on tree species. Finally, we would emphasize that the forest is a functional unit that is likely to respond to high CO₂ in a rather different way to the response of isolated individual leaves or trees. Thus, we identify a need for a much better understanding of the micrometeorological processes that determine the exchanges of CO₂ and water between forests and the atmosphere.

While the use of models offers a means to scale up in both time and space, our present state of knowledge about the processes involved is insufficient to allow this to be done with any real confidence in the results. Consequently, we see a need for the concurrent development of models and empirical studies of the physiological and micrometeorological processes that determine the response to CO₂. Given the paucity of our present knowledge, such empirical studies are needed at each spatial and temporal scale.

10.4 MODELS FOR ASSESSING THE IMPACTS OF CLIMATIC CHANGE

10.4.1 Forest Simulation Models

Determination of the response of a forest to a climatic change involves evaluation of a complex system with many levels of response. One means of attempting to handle this complexity is by the use of quantitative models of forest dynamics as investigative tools. There are several hundred extant models of forest dynamics that simulate the growth of individual trees to determine the temporal response of a forest. These models seem to be most appropriate to `intermediate time scales' discussed earlier. There have been several reviews of the types and performance of detailed models of forest dynamics (Munro, 1974; Shugart and West, 1980; Shugart, 1984).

The models that have had the greatest success in duplicating the responses of forests over 10- to 40-year time scales have been developed by foresters using empirical calibration of tree growth and form based on large, spatially extensive data sets. Examples of such models include the HUGIN Project (Hägglund, 1981) in Sweden and the STEMS Model of the Lake States in the United States (Belcher et al., 1982). These models are responsive to climatic variables. For example, Hägglund (1981) reports a 10% variation in the yield from plots over a simulated growth period as being attributable to variation in weather. The statistical accuracy of predictions of forest yield from such models is high. These models could, in principle, be used to assess the effects of small changes in climate. However, the models are restricted to conditions within the calibration range of the set of parameters, and the developers of the models are outspoken in their caution against using the models to extrapolate outside this range (Hägglund, 1981; Belcher et al., 1982).

Shugart (1984) categorized several forest simulation models with respect to the structure of the model, and to the age and the diversity of the forest simulated by the model. The models function by simulating the growth (and often death and recruitment) of individual trees in mixed-aged or even-aged and mixed-species or mono-specific forest stands. One important consideration is the form of the competition equations used in a model to simulate the interactions among individual trees spatially (2- and 3-dimensional models), vertically (l-dimensional models called gap models) or nonspatially. A tabulation of some of these models and a brief indication of their advantages and disadvantages in estimating forest response to climatic change are given in the Appendix. Of these models, only the gap models have been used extensively to simulate the response of forests to climatic change.

10.4.2 Gap Models of Forests

Gap models simulate the diameter increase of each of the trees growing on a small plot (described in Section 10.2 as the gap in the forest cycle). The trees increase in size annually. Mortality of the trees on the plot and the recruitment of new trees are stochastic functions that are applied annually. The models are not constructed to simulate mechanisms on a time scale of less than one year (although monthly temperature and precipitation data are used to compute indices for calculating the regeneration, growth and death of trees). The fundamental approach to developing these models was outlined by Botkin et al. (1972) and has been reiterated by several authors since that time. Shugart (1984) provided a detailed explanation of the assumptions in gap models and related the behaviour of such models to modern ecological concepts of forest dynamics. The approach that underlies the development of gap models is to represent the dynamics of the forest by general equations that are parameterized from basic physiology, morphology or forestry. In general, gap models do not require elaborate data sets for parameter estimation, and the data available for the forests in a given locality are often reserved for tests of the models. The models have been used to reconstruct the response of forests to climatic changes associated with the period since the last glaciation (Solomon et al., 1980; Solomon and Shugart, 1984; Shugart, 1984). This application is indicative of the potential of these models to predict the response of forests to future changes in climate.

Since these models can be used to simulate the dynamics of each of the patches comprising a forest mosaic, a two-level model can be used to simulate the dynamics of a region (e.g. Shugart et al., 1973). Such a model functions by using the detailed sub-model to simulate the individual patch dynamics and the higher level sub-model to maintain the inventory of the number of patches that could be ascribed to each of several forest types. Israel et al. (1985) have successfully used this approach to simulate the response of production of forest biomass to an increase in temperature.

10.4.3 Dynamic One-dimensional Forest Models

An alternative approach to gap models is to use dynamic, one-dimensional forest models. This approach was followed by Korzukin et al. (1986a) to accommodate age distribution in multi-species plant communities, and differs from gap models with respect to averaging procedures. In gap models, most of the dynamic effects derive from the nonlinear equations upon which they are formulated; Monte Carlo procedures are used to generate simulations of gap development by averaging the trajectories to produce the dynamics of plant community behaviour. In the dynamic, one-dimensional model, mathematical averaging is applied to each tree of the community, and then equations are derived to yield an emergent community structure through time. Furthermore, gap models assume constant growth of the taller trees with the growth of shorter trees being reduced by shading, whereas the one-dimensional model does not assume constant growth rates.

Using such a dynamic one-dimensional model, Korzukin et al. (1986b) adequately simulated succession in a two-species, post-fire cedar forest in western Siberia. The model described the wave-like, non-stable age dynamics that actually occur in the Siberian cedar forest for 200 years following fires. This type of behaviour is similar, to that shown in Figure 10.3 and to that simulated by gap models.

Following the ideas presented by Shugart et al. (1973), Antonovsky and Korzukin (1986) developed a three-level (individual-plant community-regional) model that could be used in global-biospheric modelling. This model allows the investigation of the simultaneous effects of climatic factors on forests at each of the system levels. For example, a simulated increase in temperature in the boreal forests of the cold latitudes stimulated the growth of individual trees, increased community biomass (but to a lesser degree than if the various species had not interacted competitively) and increased the frequency of regional fires. The net effect on forest biomass at the regional level can be either positive or negative. As shown by Korzukin et al. (1986b) in the case of post-fire succession (mentioned above), predictions of declining total regional biomass are obtained for a 1 °C warming when the difference between the relative acceleration of individual growth (Wm) and the relative increase in fire frequency (W_n) was set to values of 3W_m - W_n < 0. At values of 3W_m - W_n > 0, total regional biomass increased with a  l °C warming. The importance of processes occurring at all levels of forest systems-including the critical role played by fireare highlighted by this modelling approach.

10.4.4 Evaluation of Models

The use of a model to predict the response of a forest to climatic change necessarily requires extrapolation of one degree or another and for this reason is a procedure that involves some measure of uncertainty regarding the reliability of the estimates. Because it is impossible to conduct long-term experiments on today's forests with respect to climatic change over 10- to 100-year time scales, there will always be a degree of uncertainty in model predictions. It is, of course, the difficulty of conducting such experiments on forests that draws one to use models for this purpose in the first place. Careful evaluation of the models can reduce the uncertainty and thus give greater confidence in the estimates of the models.

Table 10.2 Tests of gap models showing structural and functional responses (after Shugart 1984)