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

	9.1 INTRODUCTION
	9.2 THE DIRECT EFFECTS OF INCREASED CO2
		9.2.1 Biochemistry and Physiology of CO2 Responses
		9.2.2 Biological Feedbacks
		9.2.3 Growth and Yield Under Good Environmental Conditions
		9.2.4 Interaction with Growth-limiting Environmental Factors
		9.2.5 Summary
	9.3 PERSPECTIVES ON CLIMATE IMPACTS
		9.3.1 The Slow Change View
		9.3.2 The Shift-in-risk View
		9.3.3 Blending the Views: Adaptation and Adjustment
	9.4 THE IMPACTS OF CLIMATIC CHANGE
		9.4.1 Crop Impact Analysis
		9.4.2 MarginalSpatial Analysis
		9.4.3 Agricultural Sector Analysis
		9.4.4 Historical Case Studies
	9.5 SUMMARY AND CONCLUSIONS
		9.5.1 The Possible Impacts of Increased CO2 and Climatic Change
		9.5.2 Further Considerations
		9.5.3 Some Next Steps
	NOTE ON AUTHORSHIP AND ACKNOWLEDGEMENTS
	9.6 REFERENCES

9.1 INTRODUCTION

As discussed in the previous chapter, climate plays a major role in determining the yield levels, the year-to-year variability and the spatial patterns of global agriculture. Increases in atmospheric CO₂ concentrations and changes in climate could thus have far-reaching implications for international food production and security. From a global perspective the issues vary from the temperate zones to the tropics, from the core crop regions to the margins of production, and from the developed to the developing countries. The purpose of the present chapter is to gather together what we know (and point out what we do not know) about the effects of CO₂ and climatic change, and to elucidate the broad approaches for addressing these issues.

This review is selective in its treatment while attempting to maintain a global perspective. The emphasis is placed on food crops, particularly grains, to the neglect of livestock, grasslands and fibre crops. This is not only because grains account for the vast bulk of global food production and serve to link world regions through international trade, but also because existing models and studies of the agricultural impacts of CO₂ and, especially, climatic change have focused largely on these crops. As a whole these studies jump scales to address relevant research questions at different levels of organization ranging from the plant to global food trade; as a consequence, so does this chapter.

The chapter begins by reviewing the evidence for the direct effects of higher CO₂ concentrations on individual plant processes, growth and yield. In Section 9.3 we turn to the subject of the agricultural impacts of long-term climatic change. In this section we consider alternative ways of formulating the problem in the light of the complexities created by short-term climatic variability and the response capabilities of agriculture itself. This is followed in Section 9.4 by a discussion of the major approaches to impact assessment and the results of specific analyses. In the concluding section we summarize the main findings and suggest some general directions for future research.

9.2 THE DIRECT EFFECTS OF INCREASED CO₂

In contrast to climatic changes, the rise in atmospheric CO₂ itself will be comparatively smooth and continuous with little year-to-year variability. How will higher CO₂ concentrations affect crop yields in the future in the absence of consideration of climatic change?

Much of the discussion of direct biological effects focuses on the impact of a CO₂ doubling. Unfortunately, there has been no consistency as to the control or base concentration used for purposes of comparison. Although the global average is now 343 ppmv (in 1984), the exact concentration varies with season and latitude (Figure 9.1a), with height above ground and with time of day (Figure 9.1b). Base CO₂ levels used in experiments have varied from 270 to 340 ppmv, and since the response of plant photosynthesis to CO₂ concentration is a saturating one (Figure 9.1c), the reader is cautioned that the quantitative effect of a CO₂ doubling can vary for that reason alone.

The reduction of CO₂ to carbohydrates by photosynthetic carbon fixation accounts for about 90% of the accumulation of plant dry matter. That CO₂ concentration is a limiting factor is shown by the numerous experiments in which higher CO₂ enhances photosynthesis and crop growth. The response has been found to hold even for plants grown under a variety of stressful conditions, despite a frequently repeated generalization that relates to the so-called `law of limiting factors'. This idea is that when other environmental factors such as water shortage, low light, mineral shortage or excess, and non-optimal temperature limit yield, then higher CO₂ concentration will have little or no effect. Although the generality of that concept has been challenged (Gifford, 1977, 1979a, 1979b, 1980a; Pearcy and Bjorkman et al., 1983) the idea persists (e.g. Kramer, 1981; Liss and Crane, 1983; Tolbert et al., 1983). Indeed, in certain stressful environments the relative photosynthetic response of plants to CO₂ enrichment is actually increased (as is noted below).

Predictions of crop growth and yield are incomplete if based solely on the photosynthetic response at the level of the primary CO₂ fixation mechanism. Other primary and secondary responses (like stomatal conductance and morphological development) and feedbacks interpose between photosynthetic metabolism and crop yield and must be taken into consideration in assessing the effects of higher atmospheric CO₂. Our current understanding of these processes and their effects is reviewed below, building from the underlying biochemical and physiological responses through to field crop yield, with and without limitations imposed by other environmental restraints.

Figure 9.1 Some examples of variations in CO₂ concentrations (a) seasonally, (b) diurnally in an alfalfa crop at Mead, Nebraska, U.S.A., and of (c) net photosynthetic response curves to CO₂ concentration of a C₃ species (wheat) and a C₄ species (maize)

9.2.1 Biochemistry and Physiology of CO₂ Responses

Plants are grouped photosynthetically into three groups-'C₃', 'C₄' and 'CAM'-according to biochemical distinctions in the mechanism of primary CO₂ fixation. Pineapple is the only representative among commercial crops of CAM plants and would not be expected to respond appreciably to higher CO₂, so it will not be discussed further. At least 95% of the world's biomass is of the C₃ category as are most crop species (Figure 9.2).

The distinctions between the groupings derive largely from the enzymes involved in photosynthetic fixation. Only two, perhaps three, enzymes are known to be of significance in the response of plants to CO₂ enrichment: 'rubisco',¹ 'PEP carboxylase'² and perhaps carbonic anhydrase. Rubisco is the primary enzyme for photosynthetic fixation in the C₃ plants and the CO₂ fixation rate per unit leaf area is typically related positively to the amount of this enzyme per unit leaf area. Carbon dioxide itself (together with Mg⁺⁺) activates rubisco by binding at a non-catalytic site on the enzyme protein (Jensen and Bahr, 1977). Rubisco may not always be fully activated in vivo for leaves in normal air (Perchorowicz et al., 1982; Vu et al., 1983), but it is not known whether higher CO₂ concentrations increase the level of activation in the long term (i.e. over the seasonal growth cycle).

Rubisco is not only responsible for catalysing the reduction of the CO₂ to carbohydrates, but, in the presence of light, catalyses a reaction with oxygen. The metabolite produced by the oxygenation releases recently fixed CO₂ during its metabolismthe process of photorespiration. The same catalytic site on rubisco binds both O₂ and CO₂. In C₃ plants the photorespiration rate is high and is determined in part by the relative proportions of CO₂ and O₂ in the leaf. Some of the response of photosynthesis to higher CO₂ concentration is believed to derive directly from the improved competitive advantage of CO₂ molecules over O₂ molecules for the active sites on rubisco. The reduced carbon flow through the photo respiratory cycle leads to less photorespiratory CO₂ loss as well. Even in the absence of competing O₂, however, the fully activated enzyme in leaves at 340 ppmv CO₂ is probably operating only at about half to three-quarters of its substrate-saturated capacity (Edwards and Walker, 1983).

In contrast, the primary carboxylase in C₄ plants is PEP carboxylase which is not competitively inhibited by O₂. Photorespiration is therefore negligible. PEP carboxylase has a higher effective affinity for CO₂ than does rubisco in the absence of O₂, so the enzyme is close to CO₂-saturation at the present atmospheric CO₂ concentration. Therefore one would not expect a significant enhancement of C₄ crop growth from increased CO₂ in so far as the primary carboxylase properties are concerned.

¹ Ribulose 1,5-bisphosphate carboxylase-oxygenase (EC4.1.1.39 or 'rubisco').

²Phosphoenolpyruvate carboxylase (EC 4.1.1.31 or 'PEP carboxylase').

Figure 9.2 Examples of C₃ and C₄ crops and their global annual production (fresh weight) as reported by FAO Production Yearbook, 1980 (adapted from Swaminathan, 1984)

Leaf surfaces are covered with microscopic pores, or stomata, through which gaseous exchange occurs. The aperture of the pores varies, striking a balance between inwards diffusion of CO₂ and outward diffusion of water vapour (transpiration). Higher atmospheric CO₂ concentration reduces stomatal aperture, thereby reducing transpiration. Hence the efficiency of water use in photosynthetic carbon fixation (`water use efficiency') is increased. The biochemical mechanism of stomatal response to CO₂ is unknown.

There is no difference between C₃ and C₄ plants with respect to the sensitivity of stomatal conductance to change in CO₂ concentration (Morison and Gifford, 1983). This view contradicts a common assertion that C₄ species display greater sensitivity. The assertion derives from leaf chamber studies from which are observed considerable variations among species in the absolute value of stomatal conductance in 'normal' CO₂ for particular combinations of genotype, development stage and environmental conditions. If, however, one focuses on the relative sensitivity of stomatal conductance in control- vs increased-CO₂ experiments, there appears to be little difference between the plant groups (Morison 1985). A reasonable approximation is that for most species and environmental conditions, a CO₂ doubling will cause about a 40% decrease in stomatal conductance, at least in the short term.³

Experiments on the effects of high CO₂ concentrations on dark respiration show mixed results. It has been proposed that mitochondrial respiration may increase in plants under high CO₂ in response to sucrose accumulation in leaves (e.g. Tolbert et al., 1983). A mechanism for this is thought to act via the 'alternative pathway of respiration', a normal mechanism that may function to dissipate excess photosynthesized energy (Lambers, 1982). This proposal is consistent with the findings of Hrubec et al. (1984), for example, who reported increased respiration rates of soybean leaves grown in high CO₂. However, the converse result was found for wheat (Gifford et al., 1985); plants grown continuously in 590 ppmv CO₂ experienced half as much whole-plant respiration by night (per unit net carbon fixed by day) as did plants grown in normal air. Whether a primary or secondary response, any inhibition of respiration will contribute to the stimulating effect of high CO₂ on net carbon gain while increased respiration will detract from it.

With respect to morphology and development, some species grown in high CO₂ experience greater leaf area expansion and advanced time of flowering (e.g. Hand and Postlethwaite, 1971; Goudriaan and de Ruiter 1983). Although such effects are presumably often a response to improved photosynthate supply, there are also indications of a less direct CO₂ effect. In C₄ species that do not respond photosynthetically to high CO₂, leaf area has been observed to increase. For example, growth analysis of both maize (Imai and Murata, 1978) and itchgrass (Patterson and Flint, 1980) showed that leaf area increased while net dry weight (DW) gain per unit leaf area (`net assimilation rate') was unaffected by CO₂ enrichment to above 600 ppmv. Similarly, with a doubling of normal CO₂, Morison and Gifford (1984b) observed increases in leaf area of the C₄ species Amaranthus edulis (15%), Sorghum bicolor (29%) and Zea mays (40%) grown on declining soil water content. At the same time, the efficiency of conversion of intercepted radiation into dry matter was unchanged by the high CO₂. Thus the increase in growth caused by higher CO₂ in these C₄ species was attributable to greater interception of light because of bigger leaf area, not to increased photosynthesis per unit leaf area. This implies that CO₂ was acting on leaf area development in some way other than via CO₂ effects on photosynthesis rate.

³ Morison (1985) plotted conductance at 660 ppmv CO₂ against conductance at 330 ppmv for 80 observations from the literature covering a wide range of species (C₃ and C₄), conditions and methodologies. There was linear correlation through the origin over a 20-fold conductance range, with conductance at 660 ppmv being 0.59 ± 0.04 (99% confidence limit) of conductance at 330 ppmv.

Effects of CO₂ on flowering time are usually minor but not necessarily solely due to change in photosynthate supply. For example, a slowing in the rate of flower development in sorghum without any change in dry weight growth (Hesketh and Hellmers, 1973; Marc and Gifford, 1983), seems indicative of some more direct influence of CO₂ on flowering.

9.2.2 Biological Feedbacks

Although leaves of C₃ species photosynthesize faster when transferred to an atmosphere containing higher CO₂ concentration, this initial response may not necessarily persist. Over the growing cycle of the plant, biological feedbacks can come into play acclimating enzymatic activities and leaf photosynthetic rates to the CO₂-enriched environment. But reports on the subject of photosynthetic acclimation offer no consistency as to the direction of change. For example, leaf photosynthetic capacity in high CO₂ concentration has been shown to be higher than (e.g. Bishop and Whittingham, 1968), the same as (e.g. Gifford, 1977), and lower than (e.g. von Caemmerer and Farquhar, 1984) the capacity of plants grown in normal air.

These differences may arise from the interaction of multiple feedback mechanisms. Understanding of photosynthetic acclimation to high CO₂ is hindered by the fact that most reports do not permit separation of effects operating just at the enzymatic and leaf levels, from the longer-term effects emanating from changes in the 'source:sink balance' in the whole plant. Maintenance of enhanced photosynthesis rates and, eventually, yield depend on an adequate sink (or storage organ, like the grain) for the photosynthates. If the growth of sinks does not respond to higher photosynthate supply, then photosynthesis can be depressed a negative feedback. This occurs, perhaps, by build-up of photosynthetic products in the leaf (Madsen, 1968; Herold, 1980), although the exact mechanism behind this feedback is still unknown (Gifford and Evans, 1981).⁴

Despite limited information, such results suggest that several feedback mechanisms can develop within the CO₂-enriched plant, and that the balance between them varies during plant development and determines the photosynthetic acclimation to higher CO₂. A clearer understanding of this process will depend in part on the ability of further research to separate the influences of these feedback mechanisms on photosynthetic response over time.

An increased rate of senescence (aging) is another possible feedback effect of CO₂ enrichment. Accelerated senescence has been observed in two winter annual species (St. Omer and Horvath, 1983) and in cotton (Chang, 1975), the latter exhibiting concurrent decline in carbonic anhydrase activity in the leaves. Although the observed senescence effect is minor, and is not always detected (e.g. no effect in wheat (Krenzer and Moss, 1975; Gifford, 1977)), it could possibly be pervasive due to increase in ethylene, a natural growth regulator in plants which accelerates senescence. High CO₂ concentrations caused sunflower plants to produce more ethylene, for instance (Dhawan et al. 1981). In addition, the CO₂ source for enriching the air might also contain unsuspected traces of ethylene which could promote early senescence. Some Australian Sources of CO₂, for example, contained traces of ethylene which were sufficient to hasten senescence of some species (e.g. tomato) but not others (e.g. maize) (Morison and Gifford, 1984a).

If some CO₂-stimulated DW growth were invested in leaf area expansion, a positive feedback effect could be established. Expanded leaf area would allow greater light interception which would promote further DW growth, additional leaf area expansion, and so on, until the leaf canopy becomes dense enough for full interception of incident radiation. Experimental results vary. Soybean and sunflower grown at twice normal CO₂ did not develop more leaf area in some experiments (Carlson and Bazzaz, 1980; Marc and Gifford, 1984), but did in others (Rogers et al., 1984; Morison and Gifford, 1984b). Rice frequently does not increase leaf area appreciably under CO₂ enrichment even though DW growth responds (Yoshida, 1972; Imai and Murata, 1978; Morison and Gifford, 1984b). Conversely, several C₄ species that did not show a response of net CO₂ fixation per unit leaf area or per unit of intercepted radiation, nevertheless responded with an increase in leaf area (as noted above).

The mechanisms involved in CO₂-stimulated leaf area expansion have not been widely investigated. They could be expected to vary, however, since it is known that, depending on the species, the component of leaf area increase under CO₂ enrichment varies between axillary growth (branching; Johnston, 1935), faster rate of leaf emergence (Hofstra and Hesketh, 1975) and development of larger leaves (Goudriaan and de Ruiter, 1983).

⁴ It is interesting to note, however, that in one experiment in which the sink:source rate in soybean was lowered surgically, enhanced leaf photosynthesis persisted for many weeks at high CO₂ levels (as compared to control-CO₂), despite the inability of the remaining sinks to accept more photosynthetic assimilate (Peet, 1984).

9.2.3 Growth and Yield Under Good Environmental Conditions

Under favourable growing conditions, what can we say about the net effect of all the aforementioned responses and feedbacks on plant growth and yield? One attempt to summarize CO₂ enrichment experiments showed mostly positive effects (and a few negative effects) across all groups of C₃ species. Kimball (1983) interpolated 134 observations from the CO₂ enrichment literature published over 64 years to ascertain the average increase of DW growth and yield of a variety of species in response to double `normal' (330 ppmv) atmospheric CO₂ (Table 9.1). Most of the experiments were conducted under 'good' conditions of nutrient and water supply. For the average of all C₃ species investigated, economic yield increased 26% and immature shoot dry weight increased 40%.

One particularly interesting finding concerns the growth response of small grains. Immature DW (biomass) generally exhibits greater response to high CO₂ than the final economic yields, but this is not so for small grain cereals like wheat. As shown in Table 9.1, the high (36%) increase in grain yields with a CO₂ doubling is nearly twice the increase in biomass of immature crops (20%), a finding that is also supported by work of Goudriaan and de Ruiter (1983). The effects of high CO₂ on wheat seedlings is small (Neales and Nicholls, 1978) compared to the effects once tillering and grain formation occur (Gifford, 1977; Sionit et al., 1981a). This might be a reflection of the powerful influence of CO₂ enrichment before ear emergence on tillering and sink size (ear number) in small-grain cereals (Gifford et al. 1972; Cock and Yoshida, 1973). This result with cereals belies the tentative generalization (Kramer, 1981) that determinate species (i.e. those for which leaf development ceases after flowering, as in cereals) respond less to CO₂ enrichment than do indeterminate species. Given the central role that small grains play in world food production and trade (see Chapter 8), this finding could prove to be of special importance in a CO₂-enriched future.

For C₄ species, the results are mixed. Some growth experiments (Marc and Gifford, 1984; Gifford and Morison, 1985) confirm the biochemically derived expectation of no appreciable growth response of well-watered plants to high CO₂. However, other examples (cited by Kimball, 1983; Morison and Gifford, 1984c) show substantial CO₂ effects on growth and yield. There are two routes whereby this could occur: by some unknown non-photosynthetic effect of high CO₂ on leaf area expansion and hence on light interception (see Section 9.2.1), or via an interaction with water stress, as discussed below. 

Table 9.1   Mean predicted growth and yield increases for various groupings of C₃ species for a doubling of atmospheric CO₂ concentration from 330 ppmv to 660 ppmv (adapted from Kimball, 1983). the errors indicated are 95% confidence limits

	Footnote	Immature crops		Mature crops
		No. of  records	% increase of biomass	No. of  records	% increase of marketable yield
Fibre crops	1	5	124	2	104
Fruit crops	2	15	40	12	21
Grain crops	3	6	20	15	36
Leaf crops	4	5	37	9	19
Pulses	5	18	43	13	17
Root crops	6	10	49
C3 weeds	7	10	34
Trees	8	14	26
Av. of all C3		(83)	40 ± 7	(51)	26 ± 9
Footnotes: The species represented are:
1. cotton (Gossypium hirsutum);
2. cucumber (Cucumis sativus), eggplant (Solanum melongena), okra (Abelmoschus
esculentus), pepper (Capsicum annuum), tomato (Lycopersicum esculentum);
3. barley (Hordeum vulgare), rice (Oryza sativa), sunflower (Helianthus annuus), wheat
(Triticum aestivum);
4. cabbage (Brassica oleracea), white clover (Trifolium repens), fescue (Festuca elatior),
lettuce (Lactuca sativa), Swiss chard (Beta vulgaris);
5. bean (Phaseolus vulgaris), pea (Pisum sativum), soybean (Glycine max);
6. sugar beet (Beta vulgaris), radish (Raphanus lativus);
7. Crotalaria spectabilis, Desmodium paniculatum, jimson weed (Datura stramonium),
pigweed (Amaranthus retroflexus), ragweed (Ambrosia artemisiifolia), sicklepod (Cassia
obtusifolia), velvet leaf (Abutilon theophasti);
8. cotton (Gossypium deltoides).

9.2.4 Interaction with Growth-limiting Environmental Factors

Under controlled-environment conditions, the percent enhancement of growth owing to high CO₂ concentration has been found to be greater with restricted water supply than with unlimited watering. Since higher CO₂ reduces stomatal conductance (by about 40% for a CO₂ doubling), water use efficiency in the production of dry matter (WUE) increases with CO₂ concentration, even for C₄ species. However, the relative reduction of transpiration rate per unit leaf area is not as great as that for stomatal conductance because, with reduced evaporative cooling, leaf temperature increases, thereby increasing the driving force behind transpiration (viz. the leaf-to-air vapour pressure difference) (Morison and Gifford, 1984c). Thus doubling the CO₂ concentration reduced stomatal conductance of sorghum by 40%, but transpiration rate by only 15% (van Bavel, 1974).

Under growth-limiting water supply, growth of C₃ crops responds to higher CO₂ because of both photosynthetic and stomatal effects (Gifford, 1979a; Morison and Gifford, 1984c), while growth of C₄ species responds because of stomatal effects alone. Thus for both C₃ and C₄ species, the less the availability of water, the greater the percent increase (`relative enhancement') of growth by high CO₂ concentrations.

However, in scaling up from controlled environment to a wide expanse of vegetation in the field, other attenuating phenomena come into play to determine the rate of transpiration. In circumstances where boundary layer conductance is low relative to stomatal conductance (i.e. non-windy conditions), not only is the role of stomata] conductance in controlling transpiration attenuated, but also leaf temperature is higher than under windy conditions and atmospheric humidity close to the crop surface can increase. All these aspects would reduce the impact of CO₂-induced stomatal closure on transpiration. Furthermore, and perhaps even more powerfully, soil water content may be a more important determinant of rate of transpiration on a time scale of days to weeks than is stomatal response to attributes of the aerial environment. For example, for 16 species, the time-course of depletion to exhaustion of stored soil water by individual plants was little affected by twice normal CO₂ because the high CO₂-induced leaf area increases compensated for the reduction in transpiration rate per unit leaf area (Morison and Gifford, 1984b).

Simulation of the CO₂ effect under optimal supply of water also indicated a compensation of decreased leaf transpiration by increased leaf area, resulting in a practically constant transpiration rate per ground area. In other words, the increase in overall water use efficiency approximately equalled the increased growth rate (Goudriaan et al., 1984). In this sense, higher atmospheric CO₂ concentrations may not reduce the frequency of agricultural droughts, as some have claimed. Rather, droughts will still occur but at a higher level of biomass.

Carbon dioxide enrichment increases crop growth and yield at low light intensity which is itself severely growth limiting. The relative enhancement of growth can even be greater than at high light level, as has been found for wheat (MacDowell, 1972; Gifford, 1979a). There are at least two aspects to the mechanism of growth response to CO₂ under photosynthetically limiting light intensities. One is that the quantum yield of leaf photosynthesis close to the light compensation point (i.e. the light intensity at which CO₂ uptake by a leaf is just balanced by respiratory CO₂ release) is CO₂-dependent in C₃ species, but not in C₄ species (Ehleringer and Bjorkman, 1977). High CO₂ increases C₃ species' quantum yield because it suppresses photorespiration. The extent to which the effect on quantum yield manifests itself as plant growth is dependent on the second pertinent aspecthow whole plant (dark) respiration responds. If whole plant respiration is less under high CO₂, then the light compensation point is lowered and some growth is achieved at light intensities that otherwise would prove insufficient for growth to occur. The larger relative enhancement of growth in wheat (reported to show reduced whole-plant respiration in high CO₂) in low light compared to high light intensities might be explained on this basis. For other species such as soybean, which has shown increased respiration under high CO₂, the relative enhancement of growth by high CO₂ appears equal at low and high light (Sionit et al., 1982).

With the prospect of warmer average global temperatures in the future, the response of CO₂-enriched plants under higher temperatures is pertinent. Based on limited information, it appears that in general the positive effect of higher CO₂ in stimulating photosynthesis is increased with higher temperature. However, this effect tends to be counteracted by negative feedback effects over the growth cycle of the plant. For example, for two C₃ species, Berry and Raison (1981) found that the ratio of short-term leaf photosynthesis at 1000 ppmv to that at 330 ppmv CO₂ increased sharply from 1.15 at 15 °C to 3.5 at 50 °C. This is explicable on the basis of the kinetic properties of rubisco and perhaps also on the declining solubility of CO₂ (relative to O₂) with increasing temperature (Jordan and Ogren 1984). However, temperature is important in determining the rate of growth of metabolic sinks (such as developing fruits). Sufficiently high temperatures can adversely affect sink growth (see also Section 9.3.1) and thereby feed back onto leaf photosynthesis and modulate the CO₂ response. This could possibly be one reason why soybeans, grown at supra-optimal temperature (above 30 °C) did not express the large potential CO₂ responsiveness of photosynthesis in enhanced growth (Hofstra and Hesketh, 1975; Hofstra, 1984).

At very low temperatures the inherent capability of sinks to grow is low and not limited by photosynthate supply. Thus even if photosynthesis were highly responsive to CO₂ enrichment at low temperature, it might not have much effect on sink growth that is itself temperature limited. However, high CO₂ can reduce the minimum temperature at which a plant grows and completes its life-cycle. The tropical vegetable okra (Abelmoschos esculentor) was unable to complete its life-cycle in normal CO₂ at temperature below 23 (day)/17 °C (night), while plants grown in 1000 ppmv CO₂ at 20/14 °C matured and produced fruit (Sionit et al., 1981b). Thus there was an infinite relative response (i.e. from nothing to something) of fruit yield to high CO₂ at sub-critical temperatures.

Response of plant growth to high CO₂ under nutrient deficiency or surfeit varies with both the nutrient and the species concerned. Low nitrogen supply reduces growth of all species, but with C₃ non-legumes, DW growth of both N-deficient and N-sufficient plants is increased by doubling normal CO₂ concentration. For instance, the weight of cotton plants almost doubled, irrespective of whether they had received 2 mM or 24 mM nitrate in the nutrient solution (Wong, 1979). Although not as pronounced, perennial ryegrass, wheat and soybean also achieved high per cent increases of dry weight growth from CO₂ enrichment under N-deficiency (Sionit et al., 1981a; Goudriaan and de Ruiter, 1983). The implied improvement in N-use efficiency may emanate from reduced investment in photosynthetic machinery (which has a high N-requirement) per unit of photosynthetic assimilate produced. In nodulated legumes such as soybeans or peas, high CO₂ leads to greater biological nitrogen fixation (Hardy and Havelka, 1974). This effect can be attributed to the production of more nodules on a bigger root system, rather than to greater specific activity of nodules (Phillips et al., 1976; Finn and Brun, 1982). In short, the CO₂ effect is positive under nitrogen stress.

In contrast, Goudriaan and de Ruiter (1983) were unable to show a growth response to CO₂ in phosphorus deficient plants of several species (with the exception of P-deficient bean (Vicia faba) plants which were even more responsive to high CO₂ than were plants grown with adequate P).

Potassium is another major nutrient but there is little information on its interaction with atmospheric CO₂. In potato, Goudriaan and de Ruiter (1983) noted a negative effect of increased CO₂, probably associated with higher demand for potassium.

Sodium is an essential element for C₄ photosynthesis. Growth of sodium deficient plants of two C₄ species which do not normally respond to CO₂ enrichment was greatly enhanced by high atmospheric CO₂ (1500 ppmv) (Johnston et al., 1984). While sodium deficiency is uncommon in the field, sodium excess (salinity) is common and causes reduced yield or, at greater excess, toxicity symptoms. Schwarz and Gale (1984) have shown that for diverse species, tolerance of saline conditions is increased by CO₂ enrichment to 2500 ppmv. This effect was ascribed to a shortage of photosynthate in plants suffering salt stress, but it might also be associated with the reduced demand for saline water because of CO₂-reduced transpiration.

9.2.5 Summary

Based on limited experimental results, we can expect a doubling of atmospheric CO₂ concentration from 340 to 680 ppmv to cause a 0 to 10% increase in growth and yield of C₄ crops (such as maize and sugarcane) and a 10 to 50% increase for C₃ crops (such as wheat, soybean and rice), depending on the specific crop and prevailing growing conditions. For C₃ species, the principal source of this response is at the level of the primary carboxylaseoxygenase enzyme, but stomatal, respiratory and morphological responses may also be involved. The latter three effects appear to be the principal sources of growth response for C₄ plants, where response occurs.

Table 9.2   Effects of increased CO₂ on crop response and feedbacks: a tentative compilation

	C3	C4
Photosynthesis	+ +	0
Photorespiration		NA
Nitrogen fixation	+	NA
(leguminous species)
Transpiration
Dark respiration	M	?
Leaf area development	+?	+?
(non-photosynthetic
CO2 response)
Photosynthetic acclimation	M	M
(at leaf level and via sink:source ratio)
Senescence	M	0?
Leaf area expansion	0 to +	0
(via greater photo synthesis)
+ + = strongly positive
+ = positive
= negative
0 = no effect
NA = not applicable
M = mixed response (positive or negative)
? = not known or uncertain

There are numerous feedbacks operating within the plant that serve both to accentuate and to attenuate the effect of the primary responses. An important positive feedback is the increase in the proportion of incident radiation that is intercepted by leaves because of more rapid leaf expansion in a stand of CO₂ enriched plants. Whereas in C₃ plants this stimulation of leaf expansion is likely to result mainly from the CO₂-stimulated growth itself, in C₄ species circumstantial evidence suggests that it may be a non-photosynthetic effect of higher CO₂. Important negative feedbacks can develop from the build-up of photosynthetic products on the leaves and from changes in the source:sink ratio of the plant. The effects of CO₂ enrichment on the basic biochemical and physiological plant processes and feedbacks are summarized in Table 9.2.

Table 9.3 Relative effects of increased CO₂ on growth and yield: a tentative compilation¹

Interactions between the effects of atmospheric CO₂ and other growthlimiting environmental variables on plant growth are complex and not amenable to simple generalization from the so-called 'law of limiting factors'. For example, high CO₂ concentration can reduce the deleterious impacts on growth of water-shortages, low light intensity, temperature extremes or certain mineral deficiencies, notably nitrogen deficiency. On balance, the results of experimental studies indicate that in most conceivable circumstances, the effects of increased CO₂ are beneficial (rarely detrimental) to plant growth and yield, as indicated in Table 9.3.

On the other hand, the scientist's glass-house is not the same as nature's laboratory. Carbon dioxide enrichment also stimulates the growth of weeds, for instance, which compete with crops for available moisture, light, and nutrients in actual field situations. Field studies of CO₂ enrichment have been attempted (e.g. Rogers et al., 1981), but suffer from poorer control of environmental conditions than can be obtained in growth chambers, and have not yet progressed to the stage of studies on competition in mixed plant communities. One important goal of experimental work is to contribute to the development of simulation models (discussed below). In lieu of field studies such models allow one to examine how the biochemical, physiological and environmental factors interact dynamically in the presence of high CO₂ to influence plant growth and yield. Some models have been used in this regard (e.g. Baker and Lambert, 1980; Goudriaan et al., 1984) but greater progress in model development is required before we can place high confidence (at, say, the 5% level) in the results. Such progress is being made.  

9.3 PERSPECTIVES ON CLIMATE IMPACTS

How is agriculture affected by changes in climate in the absence of direct CO₂ effects? There are myriad answers: crop varieties are switched, cultivation techniques are modified, plant development is retarded or accelerated, yields become more or less variable, production trends are altered, cropped area expands or contracts, and so on. Ideally, we should like to know the aggregate of these effects, but in the short term this is impracticable. We are forced to be selective.

The selection of research questions depends, in large part, on one's perspective concerning the effects of climate and climatic variation on agriculture. The perspectives vary widely. For instance, Bach et al. (1981) expand upon two related but opposing themes: climate as a `resource' and as a 'hazard'. Riebsame (1985) distinguishes two additional perspectives: climate as 'setting' (the background for agriculture) and climate as `determinant' (the cause of agricultural patterns and practices). Glantz (1979) identifies yet another dichotomy: CO₂-induced climatic change as an `event' (a focus on the doubling) and as a 'process' (a focus on the gradual, accumulating environmental change). There is considerable overlap between these perspectives.

Two additional perspectives can be discerned from the rapidly mounting literature on the agricultural effects of climatic change. One view holds that the potential problems (or benefits) for agriculture arise from slow, gradual changes in average climate. The other view portrays the problem as one of slow shifts in climatic risks. The way in which agricultural impacts of climatic change are assessed takes a slightly different twist according to which view predominates. Let us characterize them. 

9.3.1 The Slow Change View 

The 'slow change' view is implicit in most impact studies. It derives, in part, from the way in which the entire problem of increasing greenhouse gases and climatic change has been analysed. As reflected in this volume, the analysis begins with estimates of past and future emissions of greenhouse gases, followed by estimates of their rates of accumulation in the atmosphere and by predictions of their effects on climate. In general, the changes in climate variables are presented in terms of their central tendencies, based on the equilibrium response of climate models. It is not unexpected to find an extension of this chain of analyses to assess the impacts on agriculture and other ecosystems. The slow changes in average temperature or precipitation predicted in the previous step are quite literally assumed to be the potential problem faced by agriculturalists: a gradual, long-term, cumulative alteration of climate and, consequently, a slow deterioration (or enhancement) of the growing environment.

The specific research questions derived from this slow change perspective are cast in a similar mould. For example, 

• How would average wheat yields be affected by a 2 ^°C rise in average annual temperature? 

• How might the boundaries of North American corn production slowly recede from its present semi-arid margins with a gradual regional drying and warming?

• In the light of growing global food demands, how might world-regional grain production trends and trade balances be affected by gradual climatic changes?

• How would (or could) agriculturalists perceive and adapt to the slow changes in their environment?

This last question concerning response is particularly vexing from the slow change view. On the one hand, it has been argued that the magnitudes of some estimated changes in climate are so large that they are unprecedented in recorded history and thereby fall outside the realm of human experience. In order to adapt, agriculture may have to devise wholly unique and imaginative strategies for dealing with the effects (Cooper, 1978; 1982). On the other hand, it has also been argued that because the climatic changes will occur in a slow, cumulative fashion, agriculture has plenty of time and can most likely adapt in pace (e.g. Wittwer, 1980). The rate of change should be slow enough to allow farmers to perceive the changes in their growing environment and to switch crops, to adopt more suitable varieties, and to modify their farming practices accordingly. In short, the transition, while formidable, is eased by the luxury of time. These appraisals hinge upon assumptions concerning adaptive capacity and rates of response, to which we shall return shortly.

9.3.2 The Shift-in-risk View

The CO₂ problem as a shift in climatic risks presents an interesting contrast. According to this view, the potential agricultural problems arise mainly from changes in the frequencies of unusually disruptive (or beneficial) climatic events. There is no denial that changes in regional climates may occur slowly and gradually. It is argued, however, that the long-term changes in average temperature or precipitation per se are of relatively little importance to agriculturalists. Rather, the year-to-year risks from climatic events such as droughts, frosts, or excessive moisture are more important (e.g. Fukui, 1979). The impacts of these relatively infrequent events on crop yields cause financial (or other social, economic or human) stress and play a large role in determining agricultural viability. With a change in climate, it is likely that the frequencies of such events would shift.

In part, this 'shift-in-risk' view is a result of reversing the chain of analyses described earlier, that is, by commencing with agriculture and working toward changes in climate resulting from increasing concentrations of greenhouse gases. The questions are posed: What are the processes or resources (climatic or otherwise) critical to agricultural activities and yields? How might changes in climate affect these processes or resources? Parry and Carter (1984) call this an 'adjoint approach'. This approach, incidentally, creates greater complexity and uncertainty in climate analysis and places the onus on the climate modeller to provide more impactspecific climatological detail at relevant scales of resolution-hence the existing gap between needed and available information for impact studies (WMO, 1984).

The shift-in-risk view leads one to formulate specific research questions rather differently. For example, 

• What climatic conditions lead to particularly poor yields of sorghum and millet in the semi-arid sub-Saharan tropics? How would the probabilities of their occurrence shift with a change in climate? 

• How would the boundaries of the wheat belt in the drought-prone margins of North America be altered from increased drought risk that may result from a warmer and drier climate? 

• What would be the impacts on global grain production from simultaneous poor harvests in North America and the Soviet Union? How might the frequencies of such occurrences change if the climate of the Northern Hemisphere were to change? 

In these questions, the emphasis falls on the interannual variability of climate. 

The focus on year-to-year events rather than long-term means is partly grounded in assumptions concerning agriculturalists' perception of climatic change. In the absence of scientific information, agriculturalists may encounter considerable difficulty in perceiving and reacting to changes in mean trends. Indeed, atmospheric scientists themselves can only identify a trend in climate if a sufficiently long period of record is available to separate the `signal' from the `noise'. Instead, in the face of changing climate, the impacts from the occurrence of particularly unfavourable (or favourable) growing seasons are, in effect, the principal stimuli to which agriculturalists can, and do, react (as in changing crop type or variety, migrating elsewhere or adopting different technologies or cultivation techniques).

Thus, from the shift-in-risk view the environmental cues to which agriculture will respond are not unprecedented (even though the magnitude of the climatic change itself may be unprecedented). The cues are the same year-to-year events that agriculturalists now experience, albeit at different frequencies of occurrence. Therefore, many tactics and strategies already exist for dealing with these familiar climatic risks, although different levels of adoption, changes in farm structure and organization, or possibly new risk strategies may be warranted. This viewpoint is implicit, for example, in the remarks by Clark (1982) who states that `...what we would be doing if we were certain about CO₂ predictions is what we should be doing anyway to cope with droughts, heat waves, etc.' (p. 3).

9.3.3 Blending the Views: Adaptation and Adjustment 

While we have dichotomized the slow change view and the shift-in-risk view for purposes of explication, they are not mutually exclusive. This is evident in many statistical crop impact studies (Section 9.4.1) in which the effects of climatic extremes are reflected in mean yields; the mean and variance are inextricably bound in their long-term effects on yield trends (Mearns et al., 1984).

Furthermore, it has been hypothesized that long-term adaptation to climate results from the aggregate of short-term responses to risk (Parry, 1978, 1985; Whyte, 1981; de Vries, 1980). As agriculturalists strive to achieve the best returns and to build resiliency in the face of interannual climatic variability, the agricultural system becomes best fitted to the most frequently occurring climatic conditions over the long runthat is, those described by measures of central tendency. For example, both the heart and the spatial extent of the Australian wheat growing regions may be largely a reflection of the spatial gradients of drought risk, but may be well described by mean rainfall.

In this respect, the two views just simply may be emphasizing two separate bands of the same response spectrum. We illustrate this notion in Figure 9.3, partly adapted from Fukui (1979). Figure 9.3a displays a hypothetical curve of precipitation, conveniently distributed normally (in reality precipitation is usually described better by non-Gaussian curves). Let us assume that an agricultural system is perfectly centred on the mean, so that any deviation from the mean has negative effects on yields. Superimposed on the distribution in Figure 9.3b are three categories of agricultural response (after Burton et al., 1978; Kates et al., 1985).

First, for frequently occurring amounts of precipitation there exists a set of responses that are labelled adaptations. From one year to the next, agriculturalists expect mildly wet or dry conditions to occur. It is this expectation of weather that, to the agriculturalist, is `climate' (Hare, 1985). Individuals and organizations accumulate a large mix of cultural, technological or behavioural measuresadaptationsto accommodate this expected variation. They are reflected broadly in the timing of farm operations like planting and harvesting, the migration routes of pastoral nomads, or the spatial patterns of major agricultural systems like livestock grazing or dryland wheat farming. Adaptations evolve over the long term (greater than several generations) and may not be consciously recognized as having any relationship at all to climatic or environmental fluctuation. Adaptations allow agriculture to interact freely with the expected environment without disruption or inhibiting stress. Within this `band' of adaptation, climate is a resource.

Figure 9.3 Schema of adaptation and adjustment to climate and climatic change (adapted partly from Fukui, 1979). (a) Frequency distribution of a climatic element, upon which are superimposed 'bands' of adaptation and adjustment (b). (c) A change of mean (X₀ to X₁) requires a shift in adaptation (A₀ to A₁) and adjustment (B₀ to B₁) in order to compensate for higher frequencies of dry (hatched) and extreme drought (cross-hatched) events

Third, at the far tails of the distribution are the very extreme, rare events-for example the 1-in-500 year flood or drought. Few specific adjustments are contemplated, either because they would be too costly, no viable alternatives are perceived, the events are considered too rare, or some combination thereof. Herein lurks the potential for catastrophe: a decade of drought in the North China Plains or five consecutive years of monsoon failure in South Asia.

In this sense, the future impacts on crop yields and production depend on the dynamics of agriculture and society as well as the stimulus of environment. For example, it has been asserted (Burton et al., 1978) that, in many developing countries struggling with transition to modern agriculture, the bands have shrunk rapidly. Traditional adjustments and adaptations have been displaced or discarded, while the more technological or market-oriented mechanisms that are characteristic of the developed world have not, as yet, been satisfactorily adopted. This creates situations of high vulnerability to the vagaries of climate, as evidenced by the high tolls exacted by the occurrence of extreme climatic events (Kates, 1980).

What are the effects of climatic change? In Figure 9.3c we have superimposed a change in climate, a slow change in mean from ₀ to ₁-drier conditions-assuming no change in variability. For conditions described by the new mean (₁) and the expected deviations immediately around it, the change results in lower crop yields more often than before. But these yield changes fall well within the existing band of adaptation; so, on a year-to-year basis they are not really unexpected or disruptive, and there is no lack of mechanisms to accommodate them. The potential problem is rather to recentre on the central tendency over the long run. The flexibility afforded by the adaptive capacity of the agricultural system will most likely allow it to calibrate fairly easily, closely in pace with the changing climate, as through gradual alterations of planting dates, planting densities, or allocations of irrigation watera fine tuning of the system.

Greater difficulties are encountered further left on the curve in Figure 9.3c. Climatic events that were once infrequent enough to be considered hazards (i.e. moderate droughts) now occur with troublesome regularity. Agriculturalists may begin to perceive them as part of the expected weather. If agriculturalists wish to maintain the levels of acceptable risk previously attained, the higher probability of loss or disruption associated with these events (represented by the hatched area) becomes intolerable. In effect, agriculture is under-adapted. An expansion of the band of adaptation from A₀ to A₁ will be required. This might be accomplished, for example, through a continuous adoption of stress-tolerant crop varieties, an expansion of farm sizes, or a switch to diversified farm operations that are more suitable to the new climatic conditionsan alteration of the system.

At the far left tail of the curve in Figure 9.3c are the extreme droughts to which agriculture is largely unadjusted. By virtue of the climatic change, the occurrence of these events has become more probable. Again, if previous levels of acceptable risk are to be maintained, the band of adjustment must be expanded accordingly, from B₀ to B₁ (the cross-hatched area in Figure 9.3c). But, in many cases, the alternatives may be so severely limited or prohibitively costly, and the impacts so disruptive in terms of crop yields and socio-economic consequences, that the only perceived recourse may be abandonment and migration. The Dust Bowl migrations from the U.S. Great Plains during the 1930s (Worster, 1979), the abandonment of cereal and hay production in Iceland with the Little Ice Age (Ogilvie, 1981), or, perhaps, the present situation in vast areas of drought-stricken Africa are illustrative. The long-term effect could be a change in land use and agricultural landscapea change of system.

Figure 9.4 The sensitivity of extreme climatic events to changes in the mean, based on normal distribution and constant standard deviation (see text for explanation) (from Wigley, 1985)

Of course, on the right 'wet'side of the curve the frequencies of occurrence have been reduced. For these events it could be argued that agriculture is over-adapted and over-adjusted in relation to apparent levels of acceptable risk. 

Three additional points should be emphasized with respect to climatic change and risk. First, the frequency of occurrence of extreme events can be very sensitive to relatively small changes in the mean (Mearns et al., 1984; Wigley, 1985). This relationship is illustrated in Figure 9.4 (from Wigley, 1985). The abscissa shows the change in the mean (X) as a multiple of the standard deviation (S), while the ordinate shows the resulting change in the probability of extreme events with initial probabilities (PI) of 0.1, 0.05, and 0.01 (like the previous figure, this diagram is based on an assumed normal distribution and constant standard deviation). For example, if the annual mean precipitation over England and Wales (approximately 920 mm) fell by 100 mm (approximately 0.9 standard deviationsan amount, by the way, projected by some GCMs with a CO₂ doubling), the initial 1-in-100 year drought (P1 = 0.01) would become roughly 7.5 times more frequent in any given growing season (P2/P1 = 7.5, point A).

Second, as pointed out by Parry (1985; also see Sakamoto et al., 1980), individual farmers and agricultural systems may be especially vulnerable to consecutive years of poor yields, and the probabilities of consecutive occurrences of extreme climatic events could increase dramatically with a change in climate. For instance, while in the previous example the initial 1-in-100 year drought became 7.5 times more frequent, the chances of two consecutive years of drought of this magnitude would increase by over 56 times (assuming independent events). The potential for a catastrophic succession of poor harvests, particularly in areas already sensitive to drought, could escalate rapidly, even if the change in climate itself (as measured by the central tendency) were glacially slow.

Third, it is likely that the agricultural response would not be smooth and gradual. The disruptive climatic events are already infrequent, so considerable time might pass before farmers could perceive that the probabilities were changing. In the absence of credible scientific information, response would come about through direct experience, as through a rash of particularly severe years of unfavourable weather (if, indeed, climatic change is for the worse). In this way, agricultural response is apt to occur in an abrupt, step-like manner as human perception catches up with physical reality. In the meantime, the adverse impacts could be severe.

We have attempted to show that the slow change view and the shift-in-risk view just simply emphasize different aspects of the same problem of climatic change. The climatic effects of increased concentrations of greenhouse gases, although commonly described in terms of long-term, large-scale averages, can be manifested in many ways across a wide range of spatial and temporal scales. In the global context, one danger is that the problem of climatic change may be defined too narrowly. For instance, people who represent the interests of developing countries sometimes claim that the problem of a slow, long-term change in climate is quite secondary to immediate problems of interannual yield variability, and is therefore of limited interest (WMO, 1984). This is unquestionably a valid point from the slow change view. However, from the shift-in-risk perspective the potential agricultural impacts of climatic change could be interpreted as an exacerbation of existing yield variabilitya problem which could be felt acutely, abruptly, and possibly in the not-so-distant future. 

9.4 THE IMPACTS OF CLIMATIC CHANGE 

Four broad approaches to assessing the agricultural impacts of climatic change can be identified. Crop impact analysis concentrates directly on estimating the primary effects of environmental variables on crop yields. Marginal-spatial analysis examines the possible spatial shifts in cropping patterns (or other characteristics of agriculture) that might result from changes in climate at the margins of production. The third approach, agricultural sector analysis, focuses on estimating the range of impacts within and between agricultural regions, with an emphasis on the positive and negative feedback mechanisms that, in a dynamic fashion, reduce or enhance the primary impacts on crop yields and production. Finally, historical case studies ask, What does past experience tell us about the agricultural impacts of climatic change? Let us examine each approach.

Figure 9.5 Crop impact analysis. The approach is largely unidirectional and sequential and seeks to estimate the primary, first-order impacts of changes in the growing environment on crop responses and yields as a result of increasing atmospheric concentrations of greenhouse gases

9.4.1 Crop Impact Analysis

 The first approach is presented schematically in Figure 9.5. Crop impact analyses seem to isolate and to quantify the effects of climate variables (including the direct CO₂ effects, treated separately in Section 9.2) on crop response and yields. In applications to problems of climatic change, such analyses have attempted to estimate the 'before-and -after' yield effects, usually assuming an instantaneous change from one climate state to another. Although frequently unstated, rather constrictive boundary conditions are required, and the results of most crop impact analyses should include the following caveats:

• If no shifts in spatial cropping patterns take place to adapt to changes in regional climate.

• If no changes in perception and managerial response occur.

• If the technologies and cultivation practices that affect crop-climate relationships remain constant over time.

•  If no feedbacks to yields and production from market forces or government policies occur. 

Of course, these are big `ifs', and, as we shall see, subsequent approaches (Sections 9.4.2, 9.4.3 and 9.4.4) progressively relax these constraints by setting the boundary conditions to include wider aspects of the problem. 

Crop-climate models 

Most crop impact analyses have relied on three methods for assessing the possible effects of climatic change, each of which has its advantages and drawbacks. In empirical-statistical, multiple regression models, some aspect of productionusually commercial yieldsis explained by some set of 'independent' climate variables, like monthly values of precipitation and temperature, plus a term to account for any long-term trends in yields that are usually attributed to 'technology'. The constants in the regression equation are determined empirically, and the observations for regression fitting are taken from historical records of agricultural production and climate data. The more explanatory variables included in the regression equation, the larger the number of empirically derived constants. This, in turn, requires a long historical record to provide a sufficient number of observations to derive statistically significant equations and avoid spurious resultsa major constraint in many countries where reliable historical records are short. Even where records are sufficiently long, changes in crop varieties, management or technology can alter cropweather relationships and, in effect, make historical data 'outdated' (Robertson, 1983). This is a serious drawback to using such models to predict the long-term effects of changes in climate on yields.

Regression techniques are not particularly suitable for understanding the interacting physical, biochemical and physiological processes underlying crop growth and yield. They skip the stage marked `plant response' in Figure 9.5 and attempt a direct link between environmental change and reported yield. This is the 'black-box' criticism frequently levelled at regression models (e.g. see Katz, 1977). Furthermore, differences in crop varieties, management practices and soil conditions are difficult to include as explanatory variables in regression equations (this would also increase the number of constants). Thus regression models tend to be site specific, and it is commonly accepted (but frequently ignored) that they should not be applied outside the region or data range from which they were constructed.

With the advent of computers, it has been possible to construct crop-growth simulation models which combine the mathematical equations that describe the physical, chemical and physiological mechanisms and their interaction. Such models focus explicitly on plant processes such as photosynthesis, transpiration and respiration. Data requirements for simulation models are, as a rule, demanding. The simulation time-step can vary from weeks to minuteshourly is commonand data on radiation, minimum and maximum temperatures, and soil moisture are required at those same time intervals.

One major advantage of simulation models for assessing the impacts of climatic change is their potential 'transportability'. In principle, if the processes of plant growth are described accurately and integrated correctly, the specific region of application should be of little consequence, since the model itself will demonstrate the limiting factors for growth (Baler, 1977). The effects of different management practices or environmental sensitivity can then be examined systematically.

With assumptions about management, soil conditions and planting densities, area-wide yields can be estimated using simulation models. However, Monteith suggests (WMO, 1985) that, despite their complexity and processorientation, computer simulations have not been conspicuously more successful than simpler models in making predictions of crop yields. In fact, attempts to be comprehensive have sometimes increased the size and complexity of models to the point where confusion eclipses illumination.

Intermediate to the regression and simulation approaches are simpler, deterministic mathematical functionsor mechanistic schemes (cf WMO, 1985)that relate individual climate variables to particular crop growth processes over the stages of plant development. Such schemes are especially useful for analysing the effects of a specific climate variable with respect, say, to its limiting or optimal conditions. However, their simplicity contributes to their principal drawback: the failure to consider the correlation and interaction of elements, the adaptation of plants to stress over the period of growth, and the growth restrictions imposed by nutrient deficiencies, pests or other factors (Monteith, 1981). In short, mechanistic schemes lack comprehensiveness and dynamismthe fundamental rationale for building simulation models. Mechanistic schemes provide the building blocks for process-based simulation models. ⁵ 

The strengths and weaknesses of crop-climate models are summarized in Table 9.4.⁶ In general, a common deficiency of all three types of model is the lack of rigorous validations. Ultimately, all models should be tested on independent data (not used in model construction or parameter estimation), a criterion which applies to 'process-based' simulation models and mechanistic schemes, as well as to regression models (Robertson, 1983; Haun, 1983). It is likely that many models that are potentially useful for crop impact analysis in various regions of the world have not been adequately validated, although the extent of the situation has yet to be determined (WMO, 1985).

Despite their deficiencies, crop-climate models have been used to examine the possible impacts of climatic change. What have we learned?

⁵  it is instructive to note that mechanistic schemes are the outcome of laboratory and field measurements of plant processes fitted to mathematical functions based on the laws of physics and physical chemistry. As such, they contain statistical summaries of experimental workand, thus, so do simulation models. Therefore, although we make the distinctions between statistical regression models, mechanistic schemes and crop-growth simulation models, the distinctions are somewhat artificial.

 ⁶  For reviews of crop-climate models, see WMO (1982; 1985), Baier (1977; 1983), Robertson (1983), Biswas (1980), CIAP (1975), Sirotenko (1983), or Nix (1985).

 Table 9.4   The (a) uses and (b) criticisms of types of crop-climate models: statistical relations (SR), mechanistic schemes (MS) and crop-growth computer simulations (CS) (after WMO, 1985)