Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
3 Predictions and Scenarios of Climate Changes Due to CO2 Increases DEVELOPMENT OF PREDICTIONS AND SCENARIOS A primary objective of climate-model development and related research is to improve the ability to make predictions of the response of the climate system to some internal or external change, such as increases in atmospheric CO2 concentrations. These predictions explicitly or implicitly consist of estimates of the probability of future climatic conditions due to the altered parameters or inputs and unavoidably entail uncertainties arising from model inadequacies, errors in input data, inherent indeterminacy in the climate system, and other sources. The extent of uncertainties may vary greatly with the detail of the prediction, so that, for example, regional predictions may be more uncertain than global-average predictions, and monthly predictions more uncertain than annual ones. The utility of a particular prediction depends on both the detail and the uncertainty associated with the prediction and on the use to which the prediction is put. A reliable prediction that increasing CO2 may produce global warming may be misleading to a person concerned with a particular locality if it is not accompanied by some estimate of the regional distribution of the global warming; some regions, for example, may experience no warming, a cooling, or even some combination of these depending on season or year. Similarly, reliable predictions of hydrologic changes may be needed in order to assess correctly the likely impacts of a warming on crops or on water supply. Nevertheless, internally consistent and detailed specifications of climatic conditions over space and time may be extremely useful for analysis of social 48
Predictions and Scenarios of Climate Changes 49 responses and sensitivities to climatic changes. In this report, we term such specifications "scenarios." While scenarios are naturally chosen to exclude conditions believed to be impossible, it is important to recognize that they are research toolsâsubstitutes for predictionsânot forecasts, to which objectively quantifiable degrees of credence may be attached. Assignment of relative priorities to members of a group of scenarios would, of course, constitute a probabilistic prediction. Climate models provide the opportunity to generate both predictions and scenarios of climate changes due to increasing CO2. As discussed extensively in the previous chapter, a carefully validated model should, within the bounds of the model and the validation procedure, be able to reproduce realistically the basic characteristics of the present climate or other observed climates. If the basic processes that control climate are reasonably well taken into account in the model, it should then be possible to simulate the behavior of the climate for a range of inputs and parameters. Thus, as discussed in the following section, numerical experiments with a variety of climate models have been performed that yield information on how the climate system might behave in response to a CO, increase, e.g., a doubling. In the case of global- average and perhaps zonal-average estimates of temperature changes asso- ciated with a given increase in CO2, climate models appear realistic enough to provide estimates of some reliability. However, for more detailed geography and for parameters other than temperature (e.g., windiness, soil moisture, cloudiness, and solar insolation), present climate models are not sufficiently realistic to give reliable estimates. Nevertheless, these models can still suggest scenarios of how the climate might change and provide a basis for further integration of both model and observational results. Observational studies of both present and past climates are important for both predictions and scenarios of climate changes. First, these studies can be used to check the reliability of climate models for a variety of different conditions; if a particular model is able to reproduce the characteristics of several different climates that are known to have existed, one has greater confidence that the model takes into account all relevant processes. Second, if a climate model exhibits large climatic changes due to some perturbation such as a CO2 increase, observational studies can demonstrate directly or by analogy that such large changes are indeed possible. For example, by confirming the existence of an ice-free Arctic Ocean in past epochs, paleoclimatic studies have lent some credence to model predictions of vanishing ice for greatly increased CO2. Finally, observational studies can help corroborate and fill out model-generated scenarios of climatic changes. They can, for example, suggest relationships between global or regional changes and local phenomena, between model-simulated parameters such as temperature and other potentially important parameters such as windiness
50 CARBON DIOXIDE AND CLIMATE: A SECOND ASSESSMENT and soil moisture, and between equilibrium and transient responses of the climate system as a whole or its components. These uses of observational studies are discussed more extensively in the section Observational Studies of Contemporary and Past Climates. MODEL STUDIES Numerical Experiments with Climate Models Mathematical models of climate with a wide range of complexity have been used to estimate climate changes resulting from an increase in CO2 concen- tration in the atmosphere. These models include not only 1-D RC models but also comprehensive GCM'S of the joint ocean-atmosphere system. Two kinds of numerical experiments have been conducted by use of climate models. An experiment of the first kind may be called a "CO2 transient response experiment," which seeks to investigate the temporal variation of climate caused by a continuous increase of CO2 concentration in the atmosphere. Starting from an equilibrium climate of a model with a normal CO2 concentration, a climate model is time-integrated with a prescribed CO2 concentration as a function of time. One of the key factors that control the transient response of the model climate is the thermal inertia of the oceans. Unfortunately, GCM'S of the joint ocean-atmosphere system are still in an early stage of development. A few studies of CO2 transient response experiments have begun to appear in the literature and are the subject of discussion in the section Role of the Oceans. An experiment of the second kind may be called a' 'CO2-climate equilibrium sensitivity experiment." It evaluates the total equilibrium response of the climate to a given increase of CO2 concentration by examining the difference between a model climate with normal CO2 concentration and another model climate with an above-normal concentration. In the following subsections, predictions of CO2-induced climate changes based on the results from the several numerical experiments of the second kind, i.e., CO2-climate sensitivity experiments, are reviewed. Global-Average Response Temperature. On the basis of comparative assessment of the results from the wide varieties of climate models, the Charney report estimated the equilibrium global surface warming resulting from the doubling of CO2 concentration to
Predictions and Scenarios of Climate Changes 51 be "near 3Â°C with a probable error of Â± 1.5Â°C."* The present panel has not found any new results that necessitate substantial revision of this conclusion. Table 3.1 contains predicted area-mean increases of surface air temperature obtained from various recent experiments with GCM'S of climate. The range of these results for a doubling of CO2 (i.e., about 2^Â°C) falls within the uncertainty indicated in the Charney report. (For discussion of the results from simplified models of climate, see the section Global Climate Sensitivityâ Simplified Models and Empirical Approaches.) An extensive discussion of the differences among these estimates of warming was included in the Charney report and is not repeated here. (See also World Meteorological Organization, 1979a, 1979b.) One should point out, however, that the estimates by Gates et al. (1981) and Mitchell (1979) included in Table 3.1 for completeness are much smaller than the others because the imposed condition of fixed sea- surface temperature places a strong constraint on the changes of surface air temperatures in their model atmospheres. In contrast to the warming of the troposphere, both RC models and GCM'S indicate that a cooling of the stratosphere would result from an increase of CO2 concentration in the atmosphere. The enhanced emission from the stratosphere upward into space and downward into the troposphere is responsible for this cooling. In general, the magnitude of this stratospheric cooling would be much larger than that of the tropospheric warming and would have a relatively small latitudinal variation. According to the latest study by Pels et al. (1980), the predicted cooling due to the doubling of CO2 concentration with other stratospheric conditions fixed is about 7Â°C and 11Â°C at altitudes of 30 and 45 km, respectively. Hydrology. Sensitivity studies with GCM'S suggest that the global-mean rates of both evaporation and precipitation would increase with higher atmospheric CO2 concentrations. The physical mechanism for the intensification of the hydrologic cycle was discussed, for example, by Manabe et al. (1965), Manabe and Wetherald (1975). Wetherald and Manabe (1975), and Schneider et al. (1978). It should be noted here that this does not necessarily imply an overall increase (or reduction) in the soil wetness. Table 3.2 presents the results of several sensitivity studies and shows the percentage difference in the intensity of the models' hydrologic cycles between normal CO2 concen- trations and two levels of above-normal CO2 concentrations. This, together with the results presented in Table 3.1, indicates that a model with a larger *Our understanding is that the Charney group meant this to imply a 50 percent probability that the true value would lie within the stated range.
52 at p fe . E I SL "3 .= K s ^ i- a. im tio TABLE 3.1 of CO: Con -o -o TJ -a 4> "O V V ^ "3 "O "3 ctctctcLCLcLalct -O X 4> gj a! ct t/5 C/3 < -O -O -O -O N N N N 22^ 2 2 Â« E ~ Â« h c- ~ ~ -c -c J2 =S "t^ Â« *r 4J OJ ^ **, 13 -a s ." S -^ r- Â£5 ~ "7 ".33 2 s s 2 2 s Â£ - j= i aj g I i f S S Â£ ^â â¢ â¢ TO ^ S I X â ON ^ r- . a^ c3s -D -a 8 H .5 = * P II c tu Se |. I
Predictions and Scenarios of Climate Changes 53 TABLE 3.2 Percentage Increase of Area-Mean Rates of Precipitation (or Evaporation) Resulting from Doubling (or Quadrupling) of CO, Concentration in a Model Atmosphere Reference Doubling C/5-) Quadrupling (%) Idealized Geography Manabe and Wetherald (1975)" 7 Manabe and Wetherald (1980)" 7 12 Wetherald and Manabe (1981)" 13 Wetherald and Manabe (1981) 10 Realistic Geography Manabe and Stouffer (1979, 1980) 7 Hansen etal. (1979)" 6 Hansen et al. (1979) 4 "No seasonal variation of insolation. CCK-induced warming tends to have a larger increase in the overall intensity of the hydrologic cycle. It therefore appears probable that the doubling of CO; concentrations would result in the overall intensification of the hydrologic cycle by several percent. Zonal-Average Response Temperature. Climate sensitivity studies indicate that the CO2-induced increase of surface air temperature would have significant latitudinal and seasonal variations. For example, the predicted annual-mean warming of surface air in the polar regions is 2-3 times as great as the corresponding warming in the tropical region. The surface air warming over the Arctic Ocean would be significantly greater than the corresponding warming over the Antarctic continent, where the effect of the snow-albedo feedback mechanism is relatively small. In low latitudes, the increase of surface air temperature would be relatively small because moist convection would distribute the CO2 heating over the entire depth of the troposphere. The sensitivity studies of Manabe and Stouffer (1979, 1980) also indicate that a CO2-induced increase of surface air temperature would have a large seasonal variation over the Arctic Ocean and the surrounding regions (see Figure 3.1). This Arctic warming would be at a minimum in summer and at a maximum in winter as influenced by sea ice. In low latitudes, the amplitude of seasonal variation would be small. Hydrology. The study of the hydrologic changes induced by an increase of the atmospheric CO2 concentration has just begun, and it is not yet possible
54 CARBON DIOXIDE AND CLIMATE: A SECOND ASSESSMENT Oceans and Continents 90 N 60 30 30 60 - 90Â°S . 3 -. M M J O N D FIGURE 3.1 Latitude-time distribution of the change in zonal-mean surface air temperature (Â°C) resulting from quadrupling atmospheric CO2 concentration (Manabe and Stouffer. 1979. 1980). to make reliable predictions of the latitudinal distributions of hydrologic changes. Unlike the temperature response to a CO2 doubling, the soil moisture response is of the same order of magnitude as the climatological interannual variability, and thus a lower signal-to-noise ratio is. obtained in these experiments. However, some statistically significant zonal-mean hydrologic changes have been consistently identified in a number of numerical sensitivity experiments (Manabe and Stouffer, 1980; Wetherald and Manabe. 1981: Manabe et al., 1981). These zonal-mean results are listed below for future evaluation. The uncertainties in these results do not permit more than qualitative statements.
Predictions and Scenarios of Climate Changes 55 1. The zonally averaged annual mean rate of runoff of the models increases markedly over polar and surrounding regions, where precipitation increases substantially owing to the penetration of moisture-rich, warm air into high latitudes (i.e., poleward of 60Â° N). 2. In the zonal mean, the models' snowmelt season with large runoff rate arrives earlier because of the large warming of surface air in high latitudes (i.e., north of 60Â° N), and the model's snowfall begins later. 3. During summer, the northern hemisphere zonal-mean value of soil moisture in the models decreases in middle and high latitudes (i.e., north of 35Â° N). The earlier ending of the spring snowmelt season mentioned above implies the earlier beginning of the period of relatively large evaporation rates and, accordingly, less soil moisture during summer in higher latitudes (i.e., poleward of 30Â° N). In addition, another factor contributes to the summer dryness in middle latitudes. The reduction in the rainfall rate between spring and summer occurs earlier because of the earlier beginning of the summer period of low storminess. resulting in a reduction of soil moisture. 4. The zonal-mean area coverage and thickness of sea ice over the Arctic and circum-Antarctic oceans in the models decrease in response to an increase of atmospheric CO2 concentration. Geographical Distribution of Climate Changes Local climate has a much larger temporal variability than the zonal-mean or global-mean climate. In order to distinguish a CO2-induced climate change from the natural variations in the local model climate, an extensive time integration of a climate model is required. The results from such studies are not available at present. Furthermore, the geographical distribution of hydrologic variables (i.e.. precipitation rate) as simulated by current climate models contains many unrealistic features. Because of these problems, reliable predictions of the geographical distribution of the CO2-induced climate change are not now possible. Future efforts should be directed toward the further improvement of the parameterizations of physical processes that are poorly understood at present (i.e., cloud formation, moist convection, and land-surface processes). In view of the poor performance of current climate models in simulating the distribution of precipitation in the neighborhood of major mountain ranges, the dynamical computation of the flow field over and around mountains requires improvement. To determine the geographical details of a CO2- induced climate change, it would also be necessary to develop a climate model with improved computational resolution.
56 CARBON DIOXIDE AND CLIMATE: A SECOND ASSESSMENT OBSERVATIONAL STUDIES OF CONTEMPORARY AND PAST CLIMATES Use of Observational Studies Observational studies play an important role in three areas: (1) the formulation of ideas and models of how climate operates, (2) the general validation of theories and models, and (3) the construction of climate scenarios. Both contemporary (e.g., this century) and past climatic data sets are useful in all three areas. The role of contemporary data sets in these three areas is well known and requires no elaboration. The role of past climatic data sets is perhaps less generally recognized. An example of the use of past climatic data is provided by the work of CLIMAP Project Members (1976) in developing a set of observations of the climate at the time of the last glacial maximum around 18,000 years before present. In a narrow sense, the charts of ocean- surface temperature, sea ice, land albedo, and ice-sheet topography can be viewed as a climate scenario for a glacial world. However, these observations along with related studies of the time-dependent behavior of climate before and after glacial maximum (Hays et al., 1976; Ruddiman and Mclntyre, 1979) have also led to significant advances in climate theory, and the data sets have also been useful for climate-model validation (Gates, 1976). The CO2 sensitivity studies with climate models have indicated that CO2 - induced climatic changes might become large in comparison with recent natural climatic fluctuations. Because the changes may be large, the relative importance of observational studies of past climatic data is increased. That is, the short period of instrumental climate records has been characterized by relatively small variations in climate, none of which match the magnitude of the ultimate change that might result from a doubling of the atmospheric CO2 concentrations. Only the past climate records can provide material for observational studies of atmosphere, ocean, cryosphere, and biosphere variations of the magnitude suggested by the CO2 sensitivity studies. The status of these observational studies is briefly reviewed in the following subsections. Contemporary Climatic Data The results of the Global Weather Experiment are now being analyzed and will provide a unique data base for model calibration and validation studies of one seasonal cycle (Joint Organizing Committee, 1979). Recent progress in defining the specific characteristics of certain interannual variations (in particular, the Southern Oscillation) are providing an opportunity for climate modeling experiments, for validations of short-term climatic variations, and for improved understanding of the physics of climate.
Predictions and Scenarios of Climate Changes 57 An exciting development in climate theory has been the quantification of the role of volcanic eruptions in producing short-term climatic changes (e.g., Mass and Schneider, 1977; Hansen et al., 1981; Gilliland, 1982). The interpretation of the climatic response to volcanic eruptions involves such matters as the treatment of radiative effects of aerosols and the delayed response of climate produced by the large thermal inertia of the oceans. This work provides some validation for models as one begins to treat the different set of somewhat analogous problems related to CO2 increase. Several studies have attempted to composite individual "warm" years (or seasons) for the purpose of searching for "warm-earth" climatic patterns (Namias, 1980; Wigley et al., 1980; Williams, 1980; Pittock, 1980). While these studies have been useful, they have certain problems and limitations that deserve comment: 1. The climatic variations of individual years (or seasons) are presumably due to factors other than slow changes in the CO2 concentration. 2. The composites consist of extreme "years," whereas extreme "de- cades" might provide more useful indicators of changes in the slow-responding parts of the climate system. 3. Even the composites based on extreme years do not produce global- or zonal-average climatic differences (composite minus long-term mean) that approach in magnitude the corresponding CO2-induced climatic changes that have been simulated in CO2-sensitivity studies. 4. The emphasis so far has been on studies of atmospheric parameters (such as precipitation and surface air temperature); other variables of the ocean and cryosphere need more attention. 5. The emphasis has been on observational studies of "warm" periods rather than "warming" periods; the latter are important for identifying the transient response. 6. Most of the observational studies use the CO2 sensitivity study simu- lations of large temperature increases in polar latitudes as a basis for choosing candidates for composites; if theory or models suggest other regions or variables that might be sensitive to CO2 changes, then alternative composites should be studied. To summarize this subsection, the various observational studies (only a few of which have been mentioned) have provided a useful starting point for diagnosis of climatic processes that may prove to be relevant to the CO2 problem. There are, as noted above, certain inherent weaknesses to the approach, and more attention should be given to these problems. The currently available results do not provide a firm basis for climatic assessment of possible CO2-induced climatic changes, nor should they be considered
58 CARBON DIOXIDE AND CLIMATE: A SECOND ASSESSMENT adequate at present for validation of CO2 sensitivity studies with climate models. Past Climatic Data There is mounting evidence that the varying distribution of solar radiation associated with orbital variations (Hays et al., 1976) is an important factor in climatic change on the 10J-105-year time scale. Observational studies of past climate can document these changes so that any response to changes in solar radiation forcing from model calculations can be compared with observations. These studies are aiding in the development of a theory of large-scale climatic changes in response to external changes in solar radiations. A possibly important feature of past climates is revealed by the recent work by Berner et al. (1980) and Delmas et al. (1980), which points to large changes in CO2 concentration over the past 20.000 years (more CO2 about 5000 years ago and less before 10,000 years ago). Even if these CO2 changes did not initiate climatic change, they may amplify the change and could provide extremely useful knowledge about large-scale feedback processes that operate at the glacial-interglacial scale of climate variations, a point made recently by Thompson and Schneider (1981). Several individuals and groups are beginning to examine past climatic data for the purpose of developing "warm-earth" climate reconstructions (Kellogg, 1977; Butzer, 1980; Pittock, 1980). A sequence of workshops (Kellogg and Schneider, 1981) has been proposed that would bring together a multidisci- plinary group to work on this problem area. Although there are few published papers in this area at this time, it is possible to address some of the same problems, limitations, and opportunities that apply to contemporary climatic data sets. 1. Past climate data sets can probably be identified that have the same magnitude of change as that predicted for CO2-induced climatic change. The period 5000-7000 years ago, or perhaps a previous interglacial, may be such a period, but further work will be necessary to confirm, reject, or modify this conjecture. Much effort will be required to calibrate the past climate sensors (fossil plankton, pollen, geomorphic features, for example) in quantitative terms for comparison with climate-model simulations. 2. As with contemporary data sets, past climate data sets provide the potential for observing many components of the climate system: ocean, land surface, cryosphere. Moreover, such paleoclimatic reconstructions can provide a great deal of insight into regional climate changes on a warmer Earth and notable distributions of temperature, rainfall, and soil moisture of major importance for biological productivity. However, the assembly of past climate
Predictions and Scenarios of Climate Changes 59 data sets is a major multidisciplinary effort (CLIMAP, 1976; Peterson et al., 1979). Careful dating control is essential, and this is a serious limitation of many past climate data sets at present. 3. As with contemporary data studies, there has been an emphasis on "warm" periods rather than "warming" periods. The best opportunity for studies of transients on the order of decadal time scales is probably confined to the last several thousand years, when tree-ring data, laminated lake or ocean sediment data, and laminated ice-core data provide accurate time control. 4. As with contemporary studies, it is desirable that studies should consider a variety of past climates: a narrow focus on only one or two candidate reconstructions is probably too restrictive in view of current knowledgeâin terms of both knowing what to look for (i.e., model predictions) and what to expect (i.e.. current knowledge about a specific interval of the past climatic record). Various possibilities have been suggested: (a) Comparisons of the hypsithermal climate (5000-7000 years ago) with present climate. Evidence indicates that this was the interglacial maximum and that it was warmer then than now, at least in selected locations. (b) Comparisons of the last glacial maximum (about 18,000 years ago) with the present climate. It was colder (global-average surface temperature was about 5Â°C colder than it is at present; Gates, 1976; Manabe and Hahn, 1977; Peterson et al., 1979). and it was a time of maximum ice volume. (c) Studies of various times between glacial maximum, interglacial maximum, and present conditions, coupled with information on the changing patterns of solar radiation (from orbital parameter changes) and possible changes of CO2 concentration, provide the potential for rather detailed study of the process of large climatic changes that involve the ocean and cryosphere in a major way. (d) Studies of the previous interglacial maximum (around 120,000 years ago). Sea-level evidence suggests that it might have been warmer than the current interglacial. (e) Studies of the late Tertiary period (about 3 x 106 to 12 x 106 years ago), when there was an ice-free Arctic Ocean. We recommend that interdisciplinary workshops be held to assess these possibilities (and others) and organize the work. To summarize, the past climate studies are not so advanced as the studies with contemporary data sets. This is not surprising in view of the large multidisciplinary effort that will be required to acquire the data and carry
60 CARBON DIOXIDE AND CLIMATE: A SECOND ASSESSMENT out the analysis. Nevertheless, the past climatic studies are potentially very valuable because they deal with large changes of the climate system, including the oceans and the cryosphere. because they can reveal regional patterns of climate change, and because there is knowledge of the changes in forcing that are driving the system (solar radiation and perhaps some CO2-feedback effects). As with contemporary studies, the currently available reconstructions based on past climatic data are useful starting points for further work but are not yet adequate for model validation studies or impact studies.