A. Barrie Pittock
This paper provides an introduction to the possible impacts of the enhanced greenhouse effect on agriculture, including the many uncertainties. It stresses the need to take into account changes in seasonal rainfall, temperature, soil moisture and ambient carbon dioxide. It also emphasizes that present GCMs do not give reliable regional and local predictions of climate change, but that some GCMs do a better job than others. Until better quantitative models are available incorporating all these effects, and taking into account relevant soil conditions and agricultural practices, the balance of gains and losses cannot be assessed. Nevertheless, first estimates tend to suggest that the changes could be large.
Since the first accurate measurements of the concentration of carbon dioxide in the atmosphere, at the Mauna Loa observatory in Hawaii in 1958, the global average concentration has increased from about 315 parts per million (ppm) to over 350 ppm, as shown in Fig. 1. Other greenhouse gases, which allow sunlight in to the Earth's surface but absorb some of the outgoing heat radiation, are also increasing, including methane, nitrous oxide, and the artificial chlorofluorocarbons (CFCs). Moreover, measurements of air trapped in bubbles in ice from Antarctica and Greenland show that these increases began early in the nineteenth century. The preindustrial concentration of carbon dioxide was only about 270 ppm (Pearman et al., 1986; Barnola et al., 1987), while methane concentrations have more than doubled in the same time (Chappellaz et al., 1990).
We know from calculations of the temperatures of Earth, Mars and Venus that increases in greenhouse gases do lead to a warmer surface. In the case of the Earth, this effect is magnified by the effect of any initial warming, which causes greater evaporation, so that more water vapor enters the atmosphere. Water vapor is an efficient greenhouse gas. This effect roughly triples the warming due to increases in the other greenhouse gases.
Increasingly elaborate global climate or "general circulation" models (GCMs) have been used to establish the likely climatic consequences of the enhanced greenhouse effect (Schlesinger and Mitchell, 1987; IPCC, 1990), and the potential impacts for particular regions, industries or fields of interest have been examined (e.g. for possible Australian impacts, see Pearman, 1988). This is not the place for a full discussion of the physics of the GCMs, nor of their detailed results. Suffice to say that the GCMs generally agree that an effective doubling of carbon dioxide (i.e. including the effects of other greenhouse gases as if they were carbon dioxide) should lead to a global average surface warming of about 1.5 to 5deg.C if and when equilibrium is reached. In practice, there will be a delay due to the large heat capacity of the oceans (Schlesinger, 1986). The Intergovernmental Panel on Climate Change (IPCC) predicts that warming will occur over the next century at a rate of 0.2 to 0.5deg.C per decade. The warming is expected to be greater at high latitudes, especially in the northern hemisphere in winter, due to a positive feedback (or reinforcing effect) resulting from the melting of sea ice and land snow cover.
Related to the global warming, there are also expected to be regional changes which will be far from uniform due to changing cloud cover and the effects of land-sea contrasts and topography. Moreover, rainfall changes will be far more complex, with highly local changes in amount, seasonality, and the frequency of extremes. There will also be a rise in global average sea level due to thermal expansion of the ocean waters, the melting of glaciers, and changes in the mass balance of the polar ice sheets.
There are considerable uncertainties in trying to predict the precise magnitude of any global warming, and more particularly the related local and regional effects on rainfall and other atmospheric variables. There are a number of reasons for these uncertainties, and it is in order to narrow the uncertainties that more research is needed.
Even if we knew the precise climatic effect of any given increase in greenhouse gas concentrations, there would be uncertainty because we cannot predict exactly how rapidly the concentrations of the greenhouse gases will increase. This is because greenhouse gas emissions are subject to changes in human behavior due to economic or technological change, or deliberate policy decisions. There is also some uncertainty as to the exact sources and sinks of the various greenhouse gases, which makes extrapolation beyond a few decades, even if we knew the future industrial emission rates, rather uncertain (Tans et al., 1990). Despite these uncertainties, we are nevertheless confident that the warming potential of all such gases will be equivalent to a doubling of carbon dioxide, relative to preindustrial values, by about 2030 to 2050 AD. This assumes no drastic reduction in total greenhouse gas emissions over the next several decades.
Another major cause of uncertainty is the effect of possible changes in cloud cover. Global warming will result in an increase in the water vapor content of the atmosphere, and an associated change in cloud cover. While clouds as a whole make the Earth cooler (Ramanathan et al., 1989), the effect of any change in cloud cover may be either to cool or to warm the Earth, depending on the cloud altitude, latitude and other properties, the season, and the reflectivity of the underlying surface. It is thus no simple matter to estimate the overall effect of a change in cloudiness, and it is not surprising that different ways of representing clouds in climate models give different answers. A recent international comparison of some fourteen different climate models showed that the models gave very similar global average warmings for doubled carbon dioxide when the clouds were held constant, but that warming estimates ranged from about 2 to 5[[ring]]C when the different variable cloud schemes were included (Cess et al., 1989; Levi, 1990). Thus the major cause of uncertainty in average warmings was due to clouds, even though all the models gave significant warmings.
At the local and regional level, the major cause of uncertainty is the fact that the GCMs calculate climatic variables only at points some 500 to 1000 km apart, and use similarly coarse representations of the topography and coastlines. When we recall that rainfall can vary rapidly from one side of a mountain range to the other, or with distance from a coastline, it is clear that we cannot expect to get realistic local and regional rainfall distributions from such models. This can be overcome by a number of strategies:
We could go to higher resolution GCMs, which calculate variables at points much closer together, but this is very expensive with the present generation of computers, and requires more detailed descriptions of smaller scale weather systems. This is being pursued in our laboratory and elsewhere, but it is going to take several years and lots of funding.
We can use limited area models which calculate climate at points much closer together over the area of interest. Such a model must, however, be driven by global conditions generated in a global model in order to simulate changed climates. Again, we are following this course in CSIRO, and hope to have preliminary results for Australia in a year or two.
We can use various other schemes, which take topography and coastlines into account, to interpolate results between the coarse grid of points in the global model results. We have already applied one such scheme for extremes of temperature and their variation with local topography in Australia (Pittock and Hennessy, 1989). In many parts of the world with complex topography a particularly useful method may be to use GCM-generated near-surface pressure or wind fields for an enhanced greenhouse effect climatology, and to infer resulting rainfall patterns from these fields using relationships derived from historical data, or limited area models.
The GCMs are very crude in how they represent surface processes including soil hydrology and the role of vegetation. As a result, estimates of soil moisture and runoff arc likely to be unreliable in present models. Again, in CSIRO and elsewhere, improvements are being made to the models in these areas.
Another source of uncertainty is the role of the oceans. Because of the large heat capacity of the oceans, they will certainly introduce a delay in the global average warming by at least a decade or two, and possibly more (Schlesinger, 1986). Moreover, changes in the oceans, including surface currents and the deep ocean circulation, could introduce local and regional anomalies (Washington and Meehl, 1989; Manabe et al., 1990). One such mechanism of great significance to climate in southern Asia, the western United States, Australia, and parts of South America is the behavior of the El Nino-Southern Oscillation system (ENSO). When there is an El Nino year we generally experience drought conditions in much of southern Asia and in northern and eastern Australia, as in 1982-83, while wet conditions usually occur in the same regions during anti-El Nino years such as 1974-75 and 1989 (Ropelewski and Halpert, 1987; Philander, 1990). At present we do not know how ENSO will change with global warming, and this is a major uncertainty. We need more evidence of past ENSO behavior under warmer and colder conditions (Enfield, 1989), and better interactive ocean-atmosphere models, to remove this uncertainty. Many scientists around the world are working on it.
Different models give different results for the climate changes due to the enhanced greenhouse effect, and indeed for the present climate (Schlesinger and Mitchell, 1987; IPCC, 1990). Figs. 2 (a) and (b) show model simulations of the latitudinally averaged percentage change in precipitation for doubled effective carbon dioxide concentrations, compared with the simulations of the present climate. Note firstly that there is considerable disagreement between models in the low latitudes of the northern hemisphere in January, although all tend to show increased rainfall in low southern latitudes. In July there is agreement on a moderate increase in rainfall in the 10 to 30 N region, but considerable disagreement south of 10 N as far as 40 S. Note also that these changes are averaged over land and sea, and so they may not be representative of land areas. They do, however, give a warning that it would be most unwise to take the results of any one model that happens to be available as if it were reliable or representative. Indeed, plots of the absolute rainfall rates calculated by the various models for the present climate (not shown here) show quite large differences between models: although they all tend to show maximum rainfall rates in the vicinity of the Intertropical Convergence Zone, they range from zonal averages of around 5 mm/day to more than 11 mm/day in one model.
We have taken the view that a minimum, if not sufficient, condition for taking model results seriously is that the model does a good job in simulating the present climate in the region of interest. Thus we have compared model-generated fields of surface pressure, temperature and rainfall for both January and July under present conditions (the "control" climate) with observed values over Australia (Whetton and Pittock, in preparation) for seven different simulations which were available to us at the time. For Southeast Asia we would suggest that it would be better to base a comparison primarily on the stimulated control near-surface wind field, since this should be driven more by the large-scale monsoon circulations and less by local topography and land-sea distribution, which will strongly affect rainfall and temperatures.
Although it is difficult to be totally objective about the relative merits of the various model simulations when comparing their performances for a number of different variables, it was clear to us that in the Australian region only two of the seven control simulations which we had available were acceptable in both January and July - those of the CSIRO4 model and the UKMO model (Whetton and Pittock, in preparation). Despite a cruder representation of the surface topography, and giving too much rain, the CSIRO4 model was in fact marginally better in our view. An intercomparison of model results in the Australian-New Zealand region by some visiting colleagues from the New Zealand Meteorological Service (Mullans and Renwick, in preparation) also found the CSIRO4 model simulation to be the best, with a simulation from the Geophysical Fluid Dynamics Laboratory at Princeton next best.
In what follows I will therefore present results for temperature and rainfall changes over Australia, due to an effective doubling of carbon dioxide, from the CSIRO4 model only. Within the next year we hope to have further results from an improved version of the CSIRO4 model, and improved results from several overseas GCMs, notably some with higher horizontal resolution, should be available (see IPCC, 1990).
The temperature changes given by the CSIRO4 model for a doubling of carbon dioxide are shown in Fig. 3 for (a) July and (b) January. We see that in both winter and summer the warmings are about 2 to 3deg.C in northern coastal areas, 3 to 4deg.C in southern coastal areas, and 4 to 5deg.C in many inland areas. This is close to the average values found by other models.
These are, for reasons discussed above, far less reliably given. However, for what it is worth, the results from the CSIRO4 model are presented here. Fig. 4 shows the winter and summer rainfall regions of Australia as defined by the control (1 x CO2) simulation of the CSIRO4 model. Fig. 5 and Table 1 show the simulated changes in average rainfall over these two regions as given by the model. These show a general tendency for the rainfall in those areas of Australia dominated by summer rainfall to increase, while areas affected more by winter rainfall show decreases. Although the large-area average changes reported here are less than 10%, it must be borne in mind that higher resolution results would probably show larger percentage changes in those regions where the changed circulation interacts more with the topography, which is greatly smoothed in this version of the CSIRO4 model. When coupled with increases in evaporation, typically in the range of 5 to 15% for a 3deg.C warming, these results suggest increased aridity in many areas of Australia, and especially in the southwest.
I would greatly hesitate to quote these results were it not for the fact that two other lines of argument point in the same general direction. One is that, in an earlier study (Pittock, 1983), which looked at changes in rainfall so far this century during a period of gradual warming by about 0.5deg.C, trends of the same sign were observed. The other is that a recent review of paleo-climatic evidence for the warmer period which occurred in Australia some 6000 years ago (Wasson et al., in preparation) also indicated wetter conditions in the north and east, and drier conditions in the southwest. This warm epoch is at least a partial analog of conditions several decades from now.
While a similar rainfall tendency is suggested by these three lines of evidence, at this stage we have reservations about each, and would like to see agreement between more reliable results from several different climate models before placing too much faith in the results. We are pursuing this possibility in close cooperation with overseas modelling groups. In the meantime, we do have some pointers in what we hope is the right direction, and these may be useful for sensitivity studies, as discussed below.
These have been explored so far for extreme hot and cold days in summer and winter respectively, on the simple assumption of a uniform warming of one, two or three degrees. In Victoria, a 3deg.C warming would result in a drastic reduction in days with overnight minima below freezing in winter, and in many locations a doubling in the number of days in summer which have maxima over 35deg.C.
Fig. 6 shows maps of the average number of days each winter in which the overnight minimum screen temperature is at or below 0deg.C under (a) present climatic conditions, and (b) with a 3deg.C warming. Fig. 7 shows similar maps for days with maximum screen temperatures greater than or equal to 35deg.C. It is apparent that the frequency of such extreme events changes rapidly with changes in the mean. The probability of runs of days of extreme temperatures changes even more frequently.
Such results have serious implications for agriculture. This is especially so for crops requiring winter chill conditions or "vernalisation". Seasonal chill units under the present and warmer conditions at a number of locations across Australia have been calculated, and other agriculturally relevant statistics can readily be derived for various warming scenarios.
It must be stressed that these calculations are made on a number of rather crude assumptions, and so should not be taken as any more than a guide to the potential for impacts on agriculture. Nevertheless, if our assumptions are even approximately correct, the viability of various horticultural crops is threatened in a number of key areas. For example, most stone fruits would no longer be viable in the Goulburn Valley, which is a leading fruit-growing area in Victoria, and grape growing would be seriously at risk in the Mildura area. Horticultural crops are even more at risk in the southwest of Western Australia, which is a far more marginal area. It should be noted that it is not so much the average conditions which are critical, but the change in the year-to-year risk of not having an economic crop, and this too can be calculated using appropriate assumptions.
Selection or breeding of low-chill varieties, and changes in agricultural practices, may to some extent reduce the potential impact of the expected climatic changes.
At the high temperature end of the frequency distribution, the results may be almost as disturbing, with a much increased frequency of days over 35deg.C and of runs of such days. Table 2 shows, for various stations in Victoria, the effect of a 3deg.C warming on the average recurrence interval of a run of 5 consecutive days with maximum temperatures greater than or equal to 35deg.C at least once each summer.
These are days of extremely high potential evaporation, which put great stress on most field crops. Given a dry spell they also imply a rapid drying of the soil and vegetation, with a much greater risk of wildfire. Warmer conditions earlier in the growing season may also lead to earlier maturation of crops, with reduced time for grain filling and lower yields.
We can also calculate the number of degree-days, or thermal time available for various stages of crop growth, and how these may change with a warming. Different crops have different thermal time requirements. Again, even a relatively small increase in the average temperature can mean a large increase in the available thermal time, which may mean that certain crops with large requirements may be grown further polewards or at higher elevations under warmer conditions. Reductions in frost frequency may have a similar effect.
As more results become available, especially on a daily basis, from limited area and global climate models, we will be able to further explore this question of changes in the frequency and severity of extreme temperature, rainfall and other events, and their possible impact on crops.
One of the most important types of extreme event in Australia and in many Pacific rim countries is the tropical cyclone. In Australia these produce extreme wind, rainfall and sea level events along our tropical and sub-tropical coastline, and flood rains in much of inland Australia.
Tropical cyclones are too small in area to be seen or predicted in the present generation of global climate models, although there has been much speculation as to the possible effect of global warming on tropical cyclones (Evans, 1990). We are tackling their likely behavior initially through the use of limited area models which calculate weather at points as close as 20 km apart. Our strategy is first to attempt to model observed tropical cyclones well in a limited area model, and then to explore systematically how such storms respond to expected or possible changes in large-scale conditions such as sea surface temperature, vertical stability, and wind structure. This will tell us how sensitive such storms are to changes in some key climate variables. This is no easy task, as present models do not reproduce the behavior of tropical cyclones well. Particular attention is being paid to improving the model representation of cumulus convection.
At a later stage, we hope to study tropical cyclone frequency, location, and intensity in a limited area model covering the whole Australian region, "nested" inside, or driven by the results of a global climate model. Responses to changes in monsoon behavior and to cross-equatorial flow will also be examined. This work is being done in close cooperation with scientists at the Bureau of Meteorology Research Center and overseas.
In general, increased carbon dioxide is beneficial to the growth of plants such as wheat which have what is called a "C3" carbon fixation metabolism, but not to plants such as sugar cane and sorghum which are "C4" plants. In C3 plants, the rate of carbon fixation is unsaturated with the present level of carbon dioxide in the atmosphere and will increase in the condition of elevated carbon dioxide concentration without an increase in water loss; therefore, the efficiency of water use for carbon fixation will also increase. However, the carbon fixation rates of the C4 plants are already saturated with the present carbon dioxide concentration in the atmosphere, and so any further increase in the ambient carbon dioxide will not directly affect their carbon fixation and water use.
The actual response in the field is complicated because of competition between plants, changes in plant characteristics such as leaf area, and limitations due to other factors such as moisture or nutrient supplies. So far no experiments have been done to study crop response to high carbon dioxide concentrations in the field.
Our approach has been to recruit a biological modeller who is currently developing crop and ecosystem models which respond both to climatic variations and to changes in carbon dioxide concentration. (This work is described in Chapter 3 by Dr Yingping Wang.) He is attempting to do this with models based on a mechanistic understanding of the biology, rather than empirical correlations, in the hope that such models will hold up better when climate and carbon dioxide are both varied over large ranges. A major problem is that very little is known about the response of many crop cultivars and native plants to changes in carbon dioxide concentration in the laboratory, let alone in the field.
Potential changes in plant water use efficiency with increasing atmospheric carbon dioxide concentration, together with possible changes in interspecies competition, plant morphology and phenology, and possibly in the behavior of plant pests, may require significant changes in farm management. Our present state of relative ignorance on these issues should be a matter for considerable concern.
Another factor which ought to be considered is that agriculture may be adversely affected by any decrease in stratospheric ozone, which would cause an increase in biologically damaging ultraviolet radiation (UV-B). We know even less about the potential impact of increased UV-B radiation on relevant species than we do about ambient carbon dioxide effects, but we know that, in general, the effects will be deleterious. Table 3 summarises the effects of increases in both carbon dioxide and UV-B on plants in general, and is based on Teramura (1983).
The future depletion of stratospheric ozone due to CFCs and other global pollution is at present uncertain, as the outcome will depend in large measure on how well the Montreal Protocol on the protection of the ozone layer is adhered to by the international community. Hopefully we will not experience more than a 10% decrease in stratospheric ozone over Australia in the next 40 or 50 years, which would mean no more than an additional 20% UV-B. How much this would affect plant productivity is at present uncertain, although it is well within the normal seasonal variation and the variation with latitude within Australia. As the ozone layer is normally thinnest in summer and autumn and at low latitudes, ozone depletion may be an additional factor favouring the planting and harvesting of cereal and other crops earlier in the season and at higher latitudes.
The enhanced greenhouse effect due to the increasing concentration of infrared-absorbing gases in the atmosphere is expected to lead to a significant warming of the earth's surface, and related changes in precipitation and other climatic conditions. Combined with the increased concentration of carbon dioxide, which directly affects plant growth, these changes could profoundly affect agricultural production in many parts of the world.
At present the detailed regional climatic changes to be expected are rather uncertain, with the global-scale general circulation models (GCMs) of climate not having the horizontal resolution necessary to adequately describe local and regional climates. However, rapid progress is being made, and the uncertainty has been unnecessarily exaggerated by the common uncritical acceptance of all GCM results as of equal validity.
The time has come for agricultural modellers to conduct sensitivity experiments on productivity using the best possible GCM climate change scenarios and factoring in the effects of increasing carbon dioxide. Large changes in seasonal soil moisture, loss of winter chilling in many milder climates, increased heat stress in warmer climates, and earlier maturation due to higher temperatures may cause reductions in yield which may or may not be compensated for by the beneficial effects of increased carbon dioxide. Quantitative estimates of all these effects must replace qualitative arguments, which too often reflect disciplinary biases and vested interests in particular outcomes.
It must be stressed, however, that until better GCMs are developed, and especially ones which include a coupled dynamic ocean, any conclusions must be regarded as interim, useful for assessing alternative strategies and especially research directions, but not providing definitive predictions of the future.
Barnola, J.M., Raynaud, D., Korotkevich, Y.S. and Lorius, C. (1987): Vostok ice core provides 160,000-year record of atmospheric CO2 Nature 329, 408-414.
Cess, R.D., Potter, G.L, Blanchet, J.P. and 17 other authors (1989): Interpretation of cloud-climate feedback as produced by 14 general circulation models. Science 245, 513-516.
Chappellaz, J., Barnola, J.M., Raynaud, D., Korotkevich, Y.S. and Lorius, C. (1990): Ice-core record of atmospheric methane over the past 160,000 years. Nature 345,127-131.
Enfield, D.B. (1989): El Nino, past and present. Rev. of Geophys. 27, 159-187.
Evans, J.L (1990): Envisaged impacts of enhanced greenhouse warming on tropical cyclones in the Australian region. CSIRO Division of Atmospheric Research Technical Paper No. 20, 31 pp.
IPCC (1990): Climate Change. The IPCC Scientific Assessment (ed. J.T. Houghton, GJ. Jenkins and JJ. Ephraums), Cambridge Universiy Press, Cambridge.
Levi, B.G. (1990): Climale modellers struggle to understand global warming. Physics Today February, pp.17-19.
Manabe, S., Bryan, K. and Spelman, MJ. (1990): Transient response of a global ocean-atmosphere model to a doubling of atmospheric carbon dioxide. J. Phys. Oceanog. (in press).
Mullans, A.B. and Renwick, J.A. (in preparation): Climate change in the New Zealand region inferred from general circulation models. New Zealand Meteorological Service, Wellington, NZ.
Pearman, G.I. (ed.)(1988): Greenhouse: Planning for Climate Change, CSlRO/EJ.Brill, Leiden, 752 pp.
Pearman, G.I., Etheridge, D., de Silva, F. and Fraser, PJ. (1986): Evidence of changing concentrations of atmospheric CO2, N2O and CH4 from air bubbles in Antarctic ice. Nature 320, 248-250.
Philander, S.G. (1990): El Nino, La Niria, and the Southern Oscillalion, Academic Press, San Diego, 293 pp.
Pittock, A.B. (1983): Recent climatic change in Australia implications for a CO2-warmed earth. Climatic Change 5, 321-340.
Pittock, A.B. and Hennessy, KJ. (1989): Regional Impact of the Greenhouse Effect on Victoria, Victorian Ministry for Planning and Environment. Melbourne, 11 pp,
Ropelewski, C.F. and Halpert, M.S (1987): Global and regional scale precipitation patterns associatcd with the El Nino/Southern Oscillation. Mon. Wea Rev. 115, 1606-1626.
Schlesinger, M.E (1986): Equilibrium and transient climatic warming induced by increased atmospheric CO2. Climate Dynamics 1, 35-51.
Schlesinger, M.E and Mitchell, J.F.B. (1987): Climate model simulations of the equilibrium climatic response to increased carbon dioxide. Rev. of Geophys. 25, 760-798.
Tans, P.P., Fung, I.Y. and Takahashi, T. (1990): Observational constraints on the global atmospheric CO2 budget. Science 247, 1431-1438.
Teramura, A.H. (1983): Effects of ultraviolet-B radiation on the growth and yield of crop plants. Physiol. Plant. 58, 415-427.
Washington, W.M. and Meehl, G.A. (1989): Climate sensitivity due to increased CO2: experiments with a coupled atmosphere and ocean general circulation model. Climate Dynamics 4, 1-38.
Wasson, R.J., Donnelly, T.H. and Fleming, P.M. (in preparation): Palaeoclimatic reconstructions for the last 30,000 years in Australia - a contribution to prediction of future climate. CSIRO Division of Water Resources Technical Paper.
Whetton, P.H. and Pittock, A.B. (in preparation): Australian region intercomparison of the results of some greenhouse general circulation modelling experiments. CSIRO Division of Atmospheric Research Technical Paper.