Greenhouse Warming Research - Past, Present and Future

                                                       Richard T. Wetherald

                                   Geophysical Fluid Dynamics Laboratory/NOAA

                                            Princeton University, Princeton, NJ

      During the summer of 1988, one of the worst droughts in history occurred across most of the North American continent. During the subsequent winter, in the eastern United States, particularly in the mountainous watershed regions along the Appalachian range, very little snow fell. Since then, other anomalous weather events have occurred; severe flooding of the Mississippi River basin in the summer of 1993, a series of abnormally dry summers and warm winters in the eastern U.S. with little or no snow in the late 1990s and an extreme summer heat wave over all of Europe in 2003 which  caused an estimated 19,000 deaths or more. Regardless of what caused these phenomena, they serve as graphic examples of what can happen if our climate changes significantly from that to which we have become accustomed. In particular, the summer of 1988, as well as an overall tendency for global warming since then has sparked a great deal of discussion on the topic of greenhouse warming and whether or not it has actually begun.
      The Climate Dynamics Group of the Geophysical Fluid Dynamics Laboratory of NOAA, formally headed by Dr. Syukuro Manabe, began researching the greenhouse effect in the late 1960s and early 1970s. During this period, data on atmospheric carbon dioxide (CO₂) obtained by Dr. C. D. Keeling and his colleagues working at the Mauna Loa Observatory in Hawaii and Antarctica became available and indicated that atmospheric concentrations of CO₂ were, indeed, increasing and increasing at a fairly consistent rate. These observations coupled with the theoretical research work being done at the Geophysical Fluid Dynamics Laboratory (GFDL) laid the foundation for a transition of greenhouse theory from science fiction to science.
 

      Historical Background and Simple Models

      In the latter half of the nineteenth century, Tyndal and Arrhenius suggested that a climate change may be induced by a change in CO₂ concentration in the earth's atmosphere. This research work was followed with other studies by Callendar, Plass, Kondratiev, Niilisk, Kaplan and Moller. All of these earlier studies employed relatively simple radiative models based upon the radiative balance at the earth surface alone and, because of this formulation, obtained a variety of sensitivities of surface air temperature to similar increases of atmospheric CO₂. To overcome this difficulty, S. Manabe and R. T. Wetherald in 1967, employed a one-dimensional model which included both radiative processes and a temperature lapse-rate adjustment or a "radiative, convective model" which was based upon the radiation balance of the entire atmosphere as a whole, not just for the earth's surface. In doing so, they used a model which conserved heat and radiative energy throughout the entire model atmosphere and moved greenhouse research from a "back of the envelope" calculation to a computer.
      The above model was used to evaluate, among many other things, the CO₂-induced change of atmospheric temperature throughout the model atmosphere. This was a one-dimensional model and consisted of a system of equations which represented the effects of radiative transfer and vertical convective mixing upon the heat balance of the atmosphere. The mathematical formulation was based upon the following conditions: (1) the net radiative flux at the top of the model atmosphere is zero, (2)  the lapse rate (temperature decrease with height) throughout the model atmosphere cannot exceed a certain critical value (taken to be 6.5 C/km) due to convective and other dynamical processes, (3). in a convective stable layer, the net radiative flux is zero, (4) the surface has zero heat capacity which implies that the net downward flux of radiation is equal to the upward convective heat flux, (5) the vertical distribution of relative humidity rather than absolute humidity was prescribed. Three concentrations of atmospheric CO₂ were considered: 150, 300 and 600 parts per million (ppm). The model was, then, integrated until steady state or "equilibrium" temperature and moisture profiles were obtained in each case.
      Figure 1  illustrates the vertical distributions of the equilibrium temperature of the model atmosphere with the normal (300 ppm), half the normal (150 ppm) and twice the normal (600 ppm) concentration of CO₂. This figure indicates that, in response to the doubling of CO₂, the temperature of the model troposphere increases by approximately 2.3 degrees Centigrade whereas that of the middle stratosphere decreases by several degrees. In addition, it reveals that the magnitude of the warming resulting from the doubling of CO₂ concentration is approximately equal to the amount of cooling from the halving of the CO₂ concentration. This result suggests that the CO₂-induced temperature change is linearly proportional to the logarithm of CO₂ concentration rather than to the CO₂ concentration itself.
      The physical processes of the warming due to an increase in CO₂ concentration have traditionally been ascribed, in an analogous manner, to those operating in a "greenhouse". However, this analogy is not quite correct and the actual processes operating in the radiative, convective model may be explained as follows. It is well known that tropospheric temperature generally decreases with increasing altitude. From the physics of "black-body" radiation, it is also known that the amount of infrared (longwave) radiation is proportional to the fourth power of temperature in degrees Kelvin [degrees Kelvin (K) = degrees Centigrade (C) + 273.2] according to a mathematical expression called the Stefan-Boltzmann law. As the concentration of CO₂ is increased throughout the model atmosphere, so does the height of the emitting source of the infrared (or cooling) radiation to space due to the increased opacity of the model atmosphere. Since tropospheric temperature decreases with increasing altitude, this results in the reduction of the effective emission temperature for the outgoing radiation and hence the outgoing radiation itself at the top of the model atmosphere due to the Stefan-Boltzmann relationship. In order to satisfy the condition that the outgoing infrared or terrestrial radiation be equal to the incoming solar radiation (which has not changed), it is necessary to raise the temperature of the entire model troposphere underneath to prevent the reduction of the outgoing terrestrial radiation. In other words, the troposphere must become warmer to compensate for the above reduction of longwave radiation to space in order to maintain an energy balance at the top of the model atmosphere.
      In response to this warming, the absolute humidity (moisture content) of the model atmosphere also increases due to the condition of prescribed relative humidity. This causes further increases in both the opacity and height of the effective source of emission of outgoing infrared radiation just as it did for the additional CO₂ resulting in an additional increase of tropospheric temperature to compensate. This further enhancement is called "water vapor feedback" and is responsible for doubling the sensitivity as compared with the same model integrated under the condition of prescribed absolute humidity. As it turns out, this water vapor feedback is extremely important in the three-dimensional GCM studies which will be described in the following sections.
 

II.General Circulation Model Description and Experimental Procedure

      Atmospheric scientists have developed a "tool" called the general circulation model (GCM) to study a variety of problems ranging from daily forecasting to long-term climate change including the climatic consequences of increased CO₂. In simple terms, a GCM is a complex mathematical model composed of a system of partial differential equations derived from the basic laws of physics and fluid motion and modified to consider air as a fluid. These equations describe the dynamic, thermodynamic, radiational and hydrologic processes of the atmosphere. A block diagram of a GCM is illustrated by  Figure 2  where the large rectangles denote the main components of the model and the arrows joining one rectangle to another denote the interactions that take place between each component. The equations represented by this block diagram are too complex to be solved analytically, so they must be converted into an arithmetic form that is suitable for computation by a large digital computer. This alternate mathematical form is generally referred to as "finite differences". The above procedure is necessary since digital computers can only add, subtract, multiply and divide. The next step is to divide the entire three-dimensional atmosphere into a systematic series of "boxes" to which the basic equations must be applied and evaluated. In each box, the motion of the air (winds), heat transfer (thermodynamics), radiation (long and shortwave), moisture content (relative humidity and precipitation) and surface hydrology (evaporation, snowmelt and runoff) are calculated as well as the interactions of these processes among each of the boxes. The model is, then, programmed to run on this finite difference network in a series of "time steps" "until the particular forecast period is completed. A sample grid system for this purpose, viewed in two dimensions, is illustrated by
Figure 3 .  For the sake of clarity, a simple example of how this entire procedure is accomplished is given in the Appendix
      The model is integrated with current CO₂ concentrations until it reaches a steady state or a state where the global mean temperature does not increase or decrease over a long period of time. This run is called the "control" or present-day climate experiment. The above procedure is, then, repeated by assuming twice as much CO₂ with no other change. After both experiments have been completed, the two computed "climates" obtained are averaged over a sufficiently long time period to remove the natural variation present in each climate simulation (usually 10 or more model years). Finally, we compare the two climates to determine the changes caused by the doubling of CO₂. A diagram illustrating this procedure is given in  Figure 4 .
 

III.Experiments With Idealized Geography

      To initiate the research work through the use of three-dimensional models, Manabe and Wetherald used a GCM with a simplified land-sea distribution to conserve computer time and to better understand the feedback mechanisms caused by the increase of atmospheric CO₂. A diagram depicting this land-sea distribution is shown in  Figure 5 . The "ocean" was considered to be a "swamp ocean" which was, basically, a perpetually wet land surface. Also, surface hydrology was incorporated into the model in a simplistic way namely, a 15-centimeter deep soil moisture "bucket" which may be filled by rainfall or snowmelt and depleted by evaporation. Runoff occurs as any excess water in the bucket over the 15-centimeter capacity. The surface air temperature over both the ocean and land surfaces is computed under the assumption that these surfaces have no capacity to store heat. To incorporate the effect of albedo (reflectivity) change into the model, the depth of snow cover was predicted by an equation of snow budget whereas the extent of "sea ice" was determined by the temperature of the swamp surface. The albedos of snow cover and sea ice were assumed to be much larger than those of bare soil or open sea. Annual mean insolation was assumed in this model.
       Figure 6  shows the latitude-height difference of zonal mean temperature caused by a doubling of CO₂ obtained from the model described above. This distribution of temperature change is in, qualitative, agreement with that obtained from the one-dimensional radiative, convective model namely, it increases in the model troposphere and decreases in the model stratosphere. For this model, the area mean change of surface air temperature is about 3 degrees C and is somewhat larger than that obtained from the one-dimensional model. This is due to the fact that the three-dimensional GCM included the effects of the recession of snow cover in higher latitudes which was not present in the one-dimensional model. This mechanism can be described as follows. As the climate warms due to increasing greenhouse gases, snow cover at or near the average snow boundary, begins to melt uncovering bare soil with a much lower reflectivity. This enables greater absorption of direct solar radiation, thereby further heating the ground there. This, in turn, results in further removal and recession of highly reflective snow cover. This cycle repeats until a new equilibrium snow boundary is reached at a higher latitude. This process is referred to as "snow cover- albedo feedback" and is responsible for amplifying the warming of surface air temperature as shown in high latitudes of Fig. 6. Since the model atmosphere is very stable in higher latitudes, the heating due to the snow cover-albedo feedback is limited to just the surface layer.
      On the other hand, the increase of surface air temperature in the tropics is considerably less because convective processes spread the CO₂-induced surface heating throughout the model troposphere. Therefore, much less heating is available for increasing the surface air temperature there. In fact, Figure.6 indicates that, at low latitudes, temperature increases more in the upper troposphere than at the surface. This is due to increased heat release through condensation by more vigorous convection which occurs in the upper tropospheric levels of the tropics.
      The increase of CO₂ concentration not only affects the thermal structure but also the hydrologic cycle of the model atmosphere. One of these basic changes is the overall intensification of the hydrologic cycle as shown in  Figure 7  . In response to the increase of CO₂ and accompanying warming of the model atmosphere, evaporation also increases significantly, particularly over the idealized "ocean". In order to maintain an overall balance of water vapor in the model atmosphere, the increase of the global rate of evaporation stated above must be matched by a similar increase in the global rate of precipitation. These increases of both evaporation and precipitation result in an overall intensification of the entire hydrologic cycle. In this particular study, an increase of 3.7% was obtained due to the doubling of CO₂.
      One of the most important factors responsible for this intensification is the change in the surface radiation budget. There are five components of this balance namely, downward solar radiation, downward longwave radiation, upward longwave radiation, upward latent heat flux (upward motion of moisture carrying heat of vaporization) and upward sensible heat flux (upward motion of heat). For example, the increase in atmospheric CO₂enhances the downward flux of longwave radiation reaching the earth's surface. In addition, the CO₂-induced warming of the troposphere results in the increase of absolute humidity (moisture content) as discussed in the previous section on the radiative-convective model and also contributes to the increase of downward longwave radiation. Therefore, a larger amount of radiative energy is received by the earth's surface which must be removed by turbulent fluxes of latent and sensible heat. Due to the Clausius-Clapeyron relationship between the saturation vapor pressure and temperature, the saturation vapor pressure increases almost exponentially with a linear increase of temperature. This accounts for the increase of global mean rate of evaporation stated above. Therefore, the latent heat flux term becomes a more efficient means of removing heat from the surface than sensible heat.
      In a later study, an effort was made to examine the distribution of hydrologic change induced by increases of CO₂ over the idealized land surface (which has now been modified to a half land-half sea configuration). Sample results of this analysis are displayed in Figure 8  which show the zonal mean rates of precipitation (upper panel), evaporation (middle panel) and the difference between precipitation and evaporation (P-E) over the land surface (lower panel). According to these curves, precipitation minus evaporation (P-E) increases poleward of 50 degrees latitude and decreases in the latitudinal zone of 40-50 degrees in response to a doubling and quadrupling of CO₂. These results imply that wetter conditions should prevail poleward of 50 degrees whereas dryer conditions may be expected in the latitudinal zone of 40 to 50 degrees. This is shown in  Figure 9   where there is a considerable reduction of soil moisture in a zonal belt located approximately between 35 and 50 degrees latitude for both concentrations of CO₂whereas there is a small increase poleward of 50 degrees. The decrease of P-E in middle latitudes may be explained by both a decrease of precipitation and an increase of evaporation there. The decrease in precipitation in middle latitudes may be explained by a poleward shift of the middle latitude rainbelt which moves the maximum of precipitation from middle to higher latitudes (Fiigure.9a). The increase of precipitation (and hence P-E) poleward of 50 degrees latitude was found to be caused by an increased poleward transport of latent heat and moisture due to the general warming of the model atmosphere.
      These studies were followed by versions of the GCMs in which seasonal variation of solar radiation was incorporated into the model; one with idealized geography (reconfigured to two identical hemispheres of half land half sea) and one with realistic geography. Both of these models necessitated the use of a thermal conducting or "mixed-layer" 50 meters deep to simulate the "ocean" surface and its capacity to store heat energy. In the version with idealized geography, the corresponding distributions of P-E and soil moisture, shown above, now vary with season as    Figure 10   indicates. Specifically, the bottom portion of Figure.10 shows a zone of enhanced continental dryness which is centered around 35 degrees latitude during the winter but shifts poleward until it becomes centered at approximately 45 degrees latitude during the summer season. In general, the magnitude of this increased dryness over the continent is greatest in middle latitudes during summer. This summer dryness pattern was found to be caused by two factors.
      The first of these is an earlier disappearance of snow cover during the late winter, which causes an earlier beginning of relatively large evaporation from the soil. Because snow cover reflects a large fraction of solar radiation, its disappearance increases the absorption of solar energy by the land surface that is used as latent heat for evaporation. Thus, the end of the spring snowmelt period marks the beginning of the summer drying out of the soil. In the high CO₂ experiment, the period of snowmelt ends earlier, bringing an earlier start of the spring to summer reduction of soil moisture.
      The second factor involves changes in the middle-latitude precipitation pattern caused by a poleward shift of the middle-latitude rainbelt, a region associated with large-scale cyclonic disturbances. In the high CO₂atmosphere, warm moisture-rich air penetrates further north than in the normal CO₂ atmosphere. This is caused by a greater transport of moisture from lower to higher latitudes. Thus, precipitation increases significantly in the northern half of the rainbelt, whereas it decreases in the southern half. Because the rainbelt moves northward from winter to summer, a middle latitude location lies in the northern half of the rainbelt in winter and in its southern half in summer. Therefore, at middle latitudes, the CO₂-induced change of precipitation becomes negative in early summer, contributing to a reduction of soil moisture.
      These two mechanisms are illustrated by  Figure 11 , which shows the latitude-time distribution of the continental snow cover for both the normal CO₂and high CO₂ experiments and the latitude-time distribution of total precipitation amount for the normal CO₂experiment. The upper and middle portions of Fig. 11 indicate that, not only is there less snow depth, but snow cover is less extensive in the high CO₂ case as compared with the normal CO₂ case in middle latitudes. This implies that there is less snowmelt runoff during the spring season there. The snow cover also appears later in fall and disappears earlier in spring.
      In the lower portion of Figure.11, the mean position of the middle latitude rainbelt for the high CO₂ experiment (dashed line) is located poleward of its mean position in the normal CO₂experiment (solid line). Such a redistribution of the precipitation pattern results in wetter conditions to the north and dryer conditions to the south of the rainbelt in middle latitudes during the summer season.
      The summer reduction of soil moisture does not continue throughout the winter season. In response to the increase of atmospheric CO₂, soil wetness increases during the winter season over extensive continental regions of middle and higher latitudes (the lower portion of Figure.10). In middle latitudes, this increase of soil moisture is mainly due to the increase of precipitation in the northern half of the middle-latitude rainbelt. In general, total precipitation increases in both middle and higher latitudes during the winter season. Also, a larger fraction of the total precipitation occurs as rainfall rather than snowfall due to the warmer atmosphere. Both of these factors combine to cause the soil to become wetter. The lower portion of Fig. 10 also indicates that soil wetness is reduced during winter at 25 to 40 degrees latitude. The reduced rainfall in the southern half of the middle-latitude rainbelt is, again, responsible for this region of enhanced dryness during the winter season in these lower latitudes.
      Figure 12   shows the CO₂-induced seasonal variation of the various water budget components centered approximately at 45 degrees latitude. According to this illustration, both rainfall and evaporation increase during the winter months. However, as the spring season approaches, the CO₂-induced enhancement of rainfall decreases rapidly and actually changes from positive to negative (Figure.12c). A similar tendency occurs for evaporation but it does not occur as quickly and never changes sign. During summer, there is a decrease of rainfall which continues until early fall. Evaporation during summer also finally begins to decrease because there is no longer enough soil moisture to evaporate at the higher rate. These seasonal changes in rainfall and evaporation are consistent with the summertime soil dryness
( Figure 13 ) due to the earlier removal of snow cover and the poleward shift of the middle-latitude rainbelt. Changes of snowmelt and runoff indicate earlier melting of snow cover and an earlier runoff period during spring.
 

IV.Experiments With Realistic Geography

      Parallel to the studies noted above, S. Manabe and R. J. Stouffer in 1980, conducted an investigation with a GCM based upon the "spectral" method which incorporated realistic geography rather than the idealized geography used previously. Basically, the model is a variant of the ones used previously except that is a spectral model rhomboidally truncated at wave number 15 (R15) with a transform grid spacing of 7.5 degrees longitude by 4.5 degrees latitude. The "spectral" "method" is based upon the concept of performing linear operations in spectral space (expressing the horizontal distributions and their derivatives of the dynamic variables in terms of trigonometric and Associated Legrendre functions) and performing nonlinear operations in "grid point" "space (normal finite differences). All other features that were described previously are incorporated into the model such as a seasonal cycle of insolation, heat and water budgets over the land surface, "bucket" hydrology, mixed-layer ocean, etc. This model was, then integrated in the same manner as described previously, with a control experiment and an experiment in which the CO₂ concentration was quadrupled.
      The latitude-height distribution of the CO₂-induced change of zonal mean annually averaged temperature obtained from this model is shown in  Figure 14 . In qualitative agreement with the results from both the radiative convective model and the GCM with idealized geography (Figure.1 and Figure.6). the temperature of the model troposphere increases, whereas that of the model stratosphere decreases in response to the quadrupling of CO₂. As was the case in Figure.6, the increase of surface air temperature is particularly pronounced in the polar regions of higher latitudes where the poleward retreat of snow and ice cover enhances the warming due to the snow and ice/albedo feedback process described previously. In low latitudes, the CO₂-induced warming is spread over the entire model troposphere due to the effect of moist convection which causes a greater warming in the model's upper troposphere than at the surface. This feature was also identified in the earlier GCM.
      The corresponding latitude-time distribution of surface temperature change is shown in  Figure 15 . This figure is constructed in much the same way as Figures.10 and 11 except now the zonal averages are taken over the entire domain rather than just the continental regions. In low latitudes, the warming due to the quadrupling of CO₂ is quite small and is almost invariant with season whereas at high latitudes, it much larger and varies markedly with season particularly in the Northern Hemisphere. In the vicinity of the Arctic Ocean, the warming is at a maximum in early winter and is at a minimum during summer. A similar pattern of response is also evident in the vicinity of the Antarctic Ocean but the amplitude of the variation is considerably smaller than it is in the north polar regions.
      An analysis of the surface heat fluxes over both oceans indicates that the CO₂-induced reduction of sea ice is mainly responsible for the relatively large wintertime warming of surface temperature. In early winter, the upward conductive heat flux through the ice sheet in the warmer CO₂ experiment is considerably larger than the corresponding flux in the control experiment due to the reduction of sea ice thickness. Because of this the corresponding flux of sensible heat from the ice surface to the model atmosphere is also larger in the higher CO₂ experiment than in the control experiment. The winter warming is further enhanced by a stable stratification which limits the heating to the lowest layer of the model atmosphere at higher latitudes. This feature was also discussed previously.
      As Figure.15 indicates, the magnitude of the warming during summer is much less than the corresponding warming seen during the winter. Because of the reduction of sea ice, the surface albedo decreases significantly from the control to the high CO₂ experiment. However, the additional energy of solar and terrestrial radiation absorbed by the ocean surface is used mainly for heating the ice free mixed-layer ocean which has a relatively high heat capacity. Therefore, the warming of surface air turns out to be fairly small during this season. It should be noted, however, that the additional energy absorbed by the ocean either delays the appearance of sea ice or reduces its thickness, thereby increasing the conductive heat flux during early winter when the difference between the water and air temperature becomes relatively large.
      Over the continents, the situation is considerably different with regard to both mechanism and timing. As was noted in the previous study, the snow albedo feedback process was quite important in creating a relatively large warming in higher latitudes over continental regions. However because of the incorporation of seasonal variation into the current study, the recession of snow cover results in a substantial surface warming only during the spring snowmelt season when the incident solar flux is relatively high rather than in fall or early winter when it is near a minimum. However because of the particular distribution of the continents in northern higher latitudes, the increase of conductive heat flux through thinner ice during early winter also affects the continental regions there resulting in two maxima of surface temperature increase, one in higher latitudes during early winter and the other at more middle latitudes due to the recession of highly reflective snow cover during early spring.
      The discussion of the Manabe-Stouffer investigation indicates that the mechanism involved with CO₂-induced warming over the continents is significantly different from that causing the corresponding warming over the oceans in high latitudes. Over the continents, the snow cover albedo feedback mechanism dominates whereas over ice covered oceans, changes of conductive heat fluxes are the most important. This is because sea ice not only reflects a large fraction of solar radiation, it also inhibits the heat exchange between the atmosphere and the underlying ocean surface. Therefore, the CO₂-induced warming over the north polar ocean is maximum during early winter whereas the maximum warming takes place over the continents during the spring snowmelt season.
      The corresponding changes of hydrology are qualitatively similar to those described in the model with idealized geography and seasonal variation of solar radiation. As noted previously, the CO₂-induced change in the hydrologic cycle has a significant seasonal variation.  Figure 16  and  Figure 17  illustrate the time-latitude changes of both precipitation minus evaporation and soil moisture, respectively. According to Figure.17, the difference of zonal mean soil moisture in high latitudes of the model has a large positive value throughout most of the year with the exception of the summer season. As discussed previously, this increase in high latitudes is caused by the penetration of warm, moisture-rich air into high latitudes of the model. Figure.17 also indicates two zones of reduced soil wetness at middle and high latitudes during the summer season. Qualitatively similar results were obtained by the earlier study with idealized geography. The mechanisms responsible for these changes were analyzed and found to be similar to those already identified namely, an earlier ending of the snowmelt season which initiates the period of increased evaporation from the soil surface, an earlier beginning of the spring-to-summer reduction of the precipitation rate, an earlier reduction of baroclinicity (storminess) and a poleward shift of the middle latitude rainbelt. Although the middle latitude rainbelt is more difficult to identify in the model with realistic geography, the maximum belt of precipitation in middle latitudes was found to shift poleward in the high CO₂ experiment thus contributing to a reduction of soil wetness both in middle latitudes during summer and in the sub-tropics during winter. In summary, evaporation is found to increase while precipitation decreases from late spring to summer (Figure.16), thereby creating dryer conditions in mid-latitude continental regions throughout the entire period from late spring to early fall.
      Up until this point, the clouds were assumed to be prescribed and invariant with time. The next stage in our GCM development was to incorporate a simple cloud prediction scheme into the Manabe-Stouffer model. The CO₂ forcing in this case was taken to be a doubling rather than a quadrupling. This scheme basically consisted of placing cloud cover wherever the relative humidity exceeded a certain critical value (taken to be 99% in this model). In general, the patterns of zonal mean temperature and soil moisture differences obtained from this later version of the model
( Figure 18 ) are qualitatively similar to those already shown from the earlier model (Figures.15, 17). The mechanisms responsible for the temperature and soil moisture changes for this study are identical to the processes described previously. However, it is seen that the amplitudes of the changes in response to a doubling of CO₂ are comparable to those noted for the earlier study with prescribed clouds in response to a quadrupling of CO₂. This implies that the model utilizing predicted clouds is more sensitive than the model with fixed or prescribed clouds. An analysis of the CO₂-induced change of cloud cover and relative humidity ( Figure 19 ) revealed that the following two features of the CO₂-induced change of cloud cover were responsible for this increase in sensitivity: 1) low and upper tropospheric relative humidity and cloud cover reduce in low and middle latitudes, thererby decreasing the amount of highly reflective surface there. 2) relative humidity and cloud cover increase in the lower stratosphere at all latitudes which reduces the loss of outgoing longwave radiation to space without significantly increasing the reflective surface due to the relatively low reflectivity assigned to high, thin clouds. Both of these changes act to create a positive cloud feedback process to the model atmospheric system in response to increases of CO₂. These two positive cloud feedback processes are offset somewhat by increases of low cloud in high latitudes but because these regions receive a relatively small amount of insolation over a limited area as compared with the rest of the globe, the overall effect of this cloud increase is minimal. Therefore, the corresponding increase of global mean surface air temperature in response to a doubling of CO₂ is 2.0C for the prescribed cloud model and 4.0C for the predicted cloud model, respectively.
      To illustrate the geographical distributions of climate obtained from this model,  Figure 20  and  Figure 21  are presented. The geographic distribution of CO₂-induced surface air temperature change for the December-February period (Figure.20, top panel) shows a relatively large response in middle to high latitudes which is not present in the June-August period (Figure.20, lower panel). This is due to the recession of continental snow cover and sea ice during the winter and spring seasons, mentioned previously, whereas these processes are relatively inactive during the summer season. As was also previously shown, temperature changes in the tropical and sub-tropical latitudes are smaller and practically invariant with season.
      The corresponding changes of soil moisture are shown in   Figure.21. According to this figure, there is a general increase of soil moisture during December-February for most middle and high latitude regions and a decrease of soil moisture for the southern portion of the North American continent and Asia (Figure.21, upper panel). On the other hand, there was a general decrease of soil moisture for the June-August period for the entire continents of North America and Asia (Figure.21, lower panel). The magnitude of the summer dryness is particularly pronounced over the Great Plains of the United States. The only exception to this overall summer dryness pattern is an increase of soil moisture over India, which implies an increase of monsoonal rainfall there. As was previously noted, the decreases of soil moisture are caused by an earlier disappearance of continental snow cover, increased evaporation and changes of precipitation patterns associated with the middle-latitude rainbelt.
      It should be noted that, for the summertime distribution of surface air temperature change, the maximum temperature increase is centered over the upper United States. This maximum is directly associated with the region of maximum soil moisture decrease described above and is, in fact, caused by this region of increased dryness. As the soil moisture is depleted during summer, it eventually becomes too dry to support further increases of evaporation in a warmer CO₂ simulation. An analysis of the heat budget components over this region reveals that evaporation (and therefore, latent heat) actually decreases during the latter portion of the summer season. Therefore, a greater amount of the available energy at the model surface is realized as sensible rather than latent heat. Since sensible heat is a less efficient means of ventilating the surface as compared with latent heat, the surface temperature goes up accordingly as the bottom portion of Figure.20 indicates. A reduction of low and upper tropospheric cloud cover (positive cloud feedback) was also found to exacerbate the summer dryness and resulting heating in this region during the summer season.
 

V.Comparison With Other Modeling Groups

      In the 1980s to 1990s, GCMs by institutions other than the Geophysical Fluid Dynamics Laboratory (GFDL) were being used and analyzed for climate sensitivity investigations. These institutions included the Goddard Institute for Space Studies (GISS), the National Center for Atmospheric Research (NCAR), the United Kingdom Meteorological Office (UKMO), Oregon State University (OSU), Bureau Meteorology Research Centre (BMRC), the Canadian Climate Centre (CCC) and the European Centre for Medium-Range Weather Forecasts (ECMWF).
      In 1987, a detailed investigation was made by M. E. Schlesinger and J. F. B. Mitchell of the results obtained from three of these institutions; GFDL, GISS and NCAR. In addition to substantial regional differences of temperature and soil moisture, this comparison revealed that there was not universal agreement among the three models on the issue of middle-latitude continental summer dryness. A sample comparison is given in  Figure 22     which shows the latitude-time differences of soil moisture induced by a doubling of CO₂ for the three GCMs. According to this comparison, neither the GISS or the NCAR models produced a significant summer dryness pattern similar to that obtained by the GFDL model. whereas all three models yielded a tendency to produce wetter soil conditions during winter and early spring. All three models produce more consistent CO₂-induced hydrologic changes during the winter season than they do during summer.
      However, more recent GCM investigations appear to produce results that are consistent with the GFDL summertime scenario. These include models run at the UKMO, CCC and BMRC where substantial summer dryness patterns were obtained over both North America, southern Europe and Asia. These results are described in detail in the 1990 report by the Intergovernmental Panel for Climate Change (IPCC). These same models also produced, in varying degrees, the increased monsoonal conditions over India that were evident in the GFDL model
      With the aid of the researchers involved (G. A. Meehl, W. M. Washington and D. Rind), an attempt was made to discover the reasons for the apparent discrepancy between the three models. After an extensive analysis of the seasonal changes of hydrology, they determined that the amount of moisture in the soil moisture "buckets" in the control integration at the beginning of the summer season was too low to allow much further depletion to occur in either the GISS or NCAR models. If this is the case, there exist the distinct possibility that all three general circulation models would have produced dryer soil conditions in the middle latitude continental regions during the summer season provided enough moisture had been present in the soil to allow this to occur. A definitive conclusion on this issue must await future modeling studies that incorporate more realistic representations of surface hydrologic processes.
 

VI.Coupled Air-sea Models

      In the 1990s, the advent of main-frame super computers with greater speed and memory size allowed the development and integration of so-called "coupled air-sea models" (general circulation models which coupled together dynamic models of both the atmosphere and ocean). In our case, this involved the coupling of the previously described R15 GCM with a dynamical model of the world oceans which computes explicitly the three-dimensional structure of the ocean currents, large scale eddies (gyres) and the corresponding heat transports resulting from these motions. Changes in salinity (saltiness) and sea ice are also included in the ocean model. To investigate the evolution of climate with respect to greenhouse warming, we incorporated a scheme devised by J. F. B. Mitchell and his colleagues of the United Kingdom Meteorological Office which included a past, present and future estimation of both greenhouse gases and sulfate aerosols. Here, the equivalent carbon dioxide concentration (a combination of both CO₂ and other trace greenhouse gases, except water vapor) is prescribed for years 1765 to 1990 followed by a 1% per year increase compounded for years 1990 to 2065. The direct radiative forcing of sulfate aerosols is simulated by simply adjusting the surface reflectance of solar radiation without performing the radiative computations on the three-dimensional distributions of aerosols directly. The actual aerosol distribution for each model year is determined from separate loading patterns for years 1986 and 2050 following the scenario IS92a of the report IPCC-1992. It is interpolated from a zero loading pattern at 1765 to the loading pattern at 1986, linearly interpolated between the two loading patterns for years 1986 to 2050, then extrapolated from the loading pattern of 2050 to 2065. The same procedure for integrating the coupled air-sea model that was outlined at the beginning of this discussion was employed here namely: a 1000-year control integration where the equivalent CO₂ concentrations of greenhouse gases (except water vapor) and sulfate aerosols are held fixed at year 1965 levels. An integration in which both the equivalent carbon dioxide and sulfate aerosol forcings stated above are specified for years 1765 to 2065.
      As implied above, there are large uncertainties in the nature and magnitude of the thermal forcing of sulfate aerosols as well as ignorance of other anthropogenic and natural factors which may have forced climate over the past 200 years. Nevertheless, the present model, which includes in a crude parameterization both sulfate aerosols and greenhouse gases, simulates the warming trend during this century quite well as shown in  Figure 23 . Although this is not conclusive proof that the model will accurately forecast detailed changes of climate due to greenhouse warming, it can reasonably serve as a possible indicator on what type of large-scale changes might be expected to occur in the future.
      To illustrate to the reader the nature of some of these changes, the geographical distributions of annual mean surface air temperature difference at years 2000 and 2050 are presented in  Figure 24 . An inspection of Figure.24 reveals that, for a fully coupled air-sea model, the northern polar regions respond approximately as expected from the earlier studies but there is significantly less change in the southern polar regions in response to the increase of greenhouse gases. More specifically, this relatively slow response over the Southern Ocean and northeast portion of the Atlantic Ocean is due to the thermal inertia of the oceans in these regions which have a much deeper mixed-layer (and hence a much larger heat capacity) than the 50 meter oceanic mixed-layer that was assumed in the earlier studies. This feature has been noted not only in all of the recent GFDL investigations but also in many other institutions around the world where investigations have been carried out with the use of coupled air-sea GCMs. This result implies that the Antarctic region (and the accompanying ice sheet) will not experience a significant increase in surface air temperature for a very long time. It is also evident that the response of surface air temperature is considerably greater at year 2050 than it is at year 2000 although both distributions indicate responses that are considerably less than the responses would have been if the atmospheric-ocean system had been allowed to reach equilibrium with their respective radiative forcings due to the large thermal inertia of the oceans in general. This issue will be briefly dealt with in the succeeding discussion.
      With regard to the geographical distributions of hydrologic changes, these are similar in nature to those already illustrated. In particular, both the United States and southern Europe have a tendency to experience longer, hotter and dryer summers as time goes on. Winters, on the other hand, are estimated to be shorter and wetter with less percentage of total precipitation realized as snowfall. The mechanisms responsible for these hydrologic changes are identical to those already given for the mixed-layer GCMs.
 

VII.Committed Warming

      As a final topic, it is worthwhile to briefly describe our latest research efforts on the issue of committed warming. Basically, this term is defined as the difference between the realized warming at a given time and the warming of climate that would occur if the climate system had an infinitely long time to adjust to that particular radiative forcing (i.e. the gap between the equilibrium and realized temperature change for a given forcing). Two main factors that determine the magnitude of the committed warming are the amount of oceanic heat uptake (heat given off by the ocean to the atmosphere) and the climate sensitivity. The efficiency with which the deep ocean mixes with the upper ocean affects oceanic heat uptake. In a coupled model having a dynamical ocean, ocean heat uptake is influenced by ocean circulation patterns, the static stability of the ocean, the ocean model's sub-grid scale mixing parameterizations and other factors. A model's sensitivity depends upon the nature of the various feedback processes present in the model and upon the model's state before carbon dioxide or other greenhouse gases are increased.
      In this particular investigation, both a coupled air-sea model (AOGCM) and a mixed-layer (ML Model) are used. In doing this, we are building upon work done previously by J. Hansen, T. Wigley and their co-workers. The main improvement here is that three-dimensional GCMs were utilized rather than the simpler energy balance or one-dimensional models used in the earlier research efforts.
      Three separate experiments were integrated for this study: a transient AOGCM (TAOGCM), a transient ML model (TML) and an equilibrium ML model (EML). The AOGCM and ML model are identical to the respective models described earlier. Both transient experiments were started at year 1765 and integrated until year 2065. The equilibrium model was integrated until climate equilibrium was reached with the radiative forcings prescribed at 1980, 2000, 2020,2040, 2050 and 2060, respectively. The results of these experiments are summarized by   Figure 25   which illustrate the time evolution of global mean surface air temperature difference in response to the greenhouse gas and sulfate aerosol forcing for the three models. It is immediately seen that the temperature response for the TAOGCM is considerably less than that for either the TML or EML models. This is due to the relatively large thermal inertia of the ocean system particularly in the Southern Ocean and the northern Atlantic which was noted previously. On the other hand, the temperature response for both the TML and EML models are quite similar which implies that the results of the both the TML and EML models can be used as a proxy for the TAOGCM's equilibrium response on a real-time basis. This is because the TML responds almost as quickly as the EML model in its approach to climatic equilibrium. These results have implications for future climate change, both realized and unrealized. For example according to Figure.25, the "committed warming" (the difference between the EML and TAOGCM results) at year 2000 is approximately 1.0K. This is larger than the observed warming of 0.6K that has taken place since 1900 (Figure.23). At the same time, the global mean surface air temperature responses of all the numerical experiments are considerably greater at year 2060 than they were at year 2000. As the climate warms in response to the increasing greenhouse gases, the committed warming increases from about 1.0K at year 2000 to nearly 2.0 degrees at year 2060. The realized warming increases from 0.6K at year 2000 to 3.0K at year 2060. In other words, the greater the greenhouse gas forcing, the greater the warming commitment.
      There are large uncertainties in the magnitude of the radiative forcing of sulfate aerosols as well as other forcings which were not included in this study. However, it should be noted that when the TAOGCM is forced with both historical estimates of the sulfate aerosols and greenhouse gas concentrations. the observed warming of the 20th century is simulated quite well. Therefore, it is reasonable to speculate that the current study could provide a viable estimate of the magnitude of the committed global warming for both the present day and the future.
 

VIII.Summary

      Although there are many areas of disagreement between the various models, it is worthwhile to highlight the areas of agreement. With regard to temperature, the most recent evaluation of the state-or-the-art GCMs (IPCC-2001) state that: (1) a projected increase of global surface air temperature due to a doubling of CO₂ lies in the range 1.4 to 5.8C and (2) for all models, the increase of surface air temperature is much greater in higher latitudes than it is in the tropics. This polar amplification is greatest during the early winter and early spring seasons due to the conductive heat flux/sea ice and the snow cover/albedo feedback processes which operate mainly at these times, respectively. It should be noted that the most recent projected temperature increase is greater than that estimated by earlier IPCC reports (IPCC-1990, IPCC-1992, IPCC-1995) due to the lower projected sulphur emissions in the IPCC-2001 scenarios relative to the IS92 scenarios rather than a change in more recent model results. Therefore, the rate of warming is estimated to be considerably larger than the observed historical changes during the 20th century and is likely to exceed any warming which has occurred during the past 10,000 years, based upon paleoclimate data.
      With regard to hydrology, a survey of the latest GCMs reveal that continental snow cover in mid-latitudes is less extensive and shallower in depth for the higher CO₂ experiments. This implies that snow cover, in general, will appear later in fall and disappear earlier in spring and result in less spring runoff from snowmelt (although there will be greater runoff in the form of rainfall). Also, the soil surface will be exposed earlier in the winter season and, therefore, higher rates of evaporation will occur from it which will cause greater soil moisture loss from spring to summer. The same survey of recent model results also indicates a general consensus concerning the scenario of increased mid-latitude continental dryness during the summer season although the geographical details of these regional patterns vary considerably from one model to another.
      One of the largest uncertainties in climate sensitivity studies is the CO₂-induced response of precipitation over the continents during the period from early spring to late fall. Whether or not a given GCM. will produce a summer dryness scenario appears to be dependent upon a poleward shift of the middle latitude storm track (and accompanying rainbelt) and the state of the soil moisture of the control experiment for both early spring and summer. In the final analysis, a given GCM will produce a tendency for dryer summertime conditions if the projected rainfall is forecast to either decrease or remain approximately the same. Only if the rainfall is forecast to increase as much as the projected increase of evaporation will the desiccation of soil moisture be prevented.
      Other significant uncertainties include modeling of cloud processes, treatment of aerosols, inclusion of active ocean currents and the use of low horizontal resolution. For example, R. Cess and his colleagues have shown that there are major differences among the various GCMs concerning the methods of cloud formation and their corresponding effects of cloud feedback. The method by which the effects of aerosols are incorporated into the model is another factor. Currently, relatively simple parameterizations are being used rather than explicitly computing the three-dimensional distributions of sulfate aerosols and their radiative effects. In addition, the manner in which ocean currents are explicitly included can significantly alter the transient or time-dependent phase of a climate sensitivity experiment. Until recently, the use of relatively large "grid boxes" has greatly hampered the successful simulation of climate particularly on a regional scale. However, the advent of larger and faster supercomputers is making it possible for modelers to repeat their experiments with a considerably higher computational resolution and, thus, achieving a corresponding higher degree of credibility.
      In any event, it appears certain that, if the earth's climate becomes warmer, the earth's hydrology will change. If the above theoretical hydrologic responses to greenhouse warming prove to be correct, they will have serious implications for water resource management and agricultural planning. Slow melting of snow cover is a much more efficient means for recharging the water table rather than increased rainfall which is more likely to simply run off and be lost to the watershed system. At the same time, increased dryness during the growing season could place severe demands on the available water supply which would require more irrigation. These factors could easily combine to cause serious water shortages at precisely the time when water is needed the most for growing crops.
      Finally, it should be emphasized that the warming commitment discussion presented above describes the climate response for a constant radiative forcing of the earth. This requires greenhouse gas concentrations to remain constant over time at a particular level. The protocol reached by the Third Conference of Parties which met in Kyoto, Japan in 1997, limits emissions to approximately present day (year 2000) levels. As shown recently by T. M. Wigley and his co-workers, this level of emissions does not provide a stable greenhouse gas concentration by 2100 or anything close to it. In fact, the CO₂ concentration continues to rise considerably beyond the doubling of the pre-industrial values. Much more stringent controls on greenhouse gas emissions will be required to stabilize the greenhouse gas concentrations before 2100. Only after this stabilization occurs will the climate begin to come into equilibrium with that forcing, and only then will the total warming and associated climate changes be realized.
      Based upon these observations and the information contained in the latest IPCC report, two main conclusions can be drawn.

      1) Greenhouse warming and its attendant potential climate change is an issue that will not go away by simply trying to ignore it.

      2) Greenhouse warming, once initiated, can set into motion climate changes which will persist for many centuries in the future due to the thermal inertia of the oceans and the atmospheric time scale of greenhouse gases.

      We would, therefore, do well to seriously consider the possible options before serious climate changes can occur. Ultimately, each person will have to decide how to deal with the problem of greenhouse warming but to do this the public-at-large must be educated as to the issues involved and the current state-of-the-art research describing future greenhouse warming. The central question is: should we immediately use our present-day technology to act upon the scientific evidence now available or else wait until more evidence is obtained and a higher degree of technology is achieved? Both strategies outlined above have risks and costs associated with them. In the final analysis, the path taken will not be a science issue but a social one in which the decisions will be based upon our collective assessment of these risks and costs.
 

Appendix

Illustration of GCM Procedure

      The method of determining atmospheric motions can be demonstrated by a simple illustration. Suppose you have a transparent, thin box filled with air placed in the vertical position ( Figure A1 , upper left ). Further suppose that this box is heated on the right side by a Bunsen burner and cooled on the left side by ice. This pattern of heating and cooling would then set up a "convection cell" rotating counterclockwise where the air on the right side rises due to the heating and sinks on the left side due to the cooling. Finally, assume that we are able to measure both the temperature and fluid motion (horizontal and vertical motion components) with appropriate sensors without disturbing the flow. We, then, are able to determine the temperature and flow conditions within the box by varying the number of measurements in a systematic fashion.

       If we decide to take only one measurement in the exact middle of the box, we will obtain an extremely crude average of the conditions within the box. If, however, we divide the box up into two equal vertical portions and take our measurements in the middle of each separate portion (i.e two sets of measurements) we obtain a slightly better approximation to the temperature and flow conditions within the box (Figure.A1, upper-middle). On the other hand, if we now divide the box into four equal parts (Figure.A1, upper-right) we obtain an even better approximation to the flow inside the box. One can continue this procedure even further by dividing our box again into successive equal portions, eight (Figure.A1, middle-left) and sixteen portions (Figure.A1, middle), respectively. By continuing this procedure, one may obtain a better and better approximation to the temperature and flow pattern within the box - the more sub-divisions, the better the approximation (i.e. 32, 64, 128, 256 boxes, respectively; see rest of panels). This is, basically, how numerical weather forecasting is accomplished except instead of directly measuring the motions within the box by external instruments, they are calculated explicitly by certain mathematical equations which describe fluid flow. This method has been extended to actual weather forecasting except our sample box is replaced by a "grid network" of small "boxes" covering the entire earth and "stacked" vertically to provide three-dimensional coverage. For example, the horizonatl coverage of the R15 (low resolution) and R30 (high resolution) spectral models currently used at GFDL consists of 1920 (48 x 40) and 7680 (96 x 80) grid boxes, respectively. The necessary calculations are, then, performed for each small box in order to determine the future atmospheric conditions as a function of both space and time for the entire globe.

Bibliography

1. Cubasch, U., Meehl, G. A., Boer, G. J., Stouffer, R. J., Dix, M., Noda, A., Senior, C. A., Raper, S., and Yap, K. S. (2001). Projections of future climate change. Climate Change 2001: The Scientific Basis, IPCC Cambridge Univ. Press, Cambridge (in press).

2. Hansen, J., Lacis, A., Rind, D. Russel, G., Stone, P., Fung, I., Ruedy, R., and Lerner, J., (1984). Climate sensitivity: Analysis of feedback mechanisms, Climate Processes and Climate Sensitivity,.Geophys. Mono., 29,130-163, Amer. Geophys. Union.

3. Harvey, D., Gregory, J., Hoffert, M., Jain, A., Lai, M., Leemans, R., Raper, S., Wigley, T., and de Wolf, J., (1997). An introduction to simple climate models used in the IPCC second assessment report, IPCC Technical Paper II, 47 pp., IPCC Cambridge Univ. Press, Cambridge.

4. Houghton, J. T. (1997). Global Warming: The Complete Briefing. (2nd ed.) Cambridge Univ. Press, Cambridge.

5. Jones, P. D., Osborn, T. J., and Briffa, K. R., (1997). Estimating sampling errors in large-scale temperature averages. J. Clim. 10, 2548-2568.

6. Keeling, C. D., Adams, J. A., Ekdahl, C. A., and Guenther, P. R., (1976a). Atmospheric carbon dioxide variations at the South Pole. Tellus 28, 552.

7. Keeling, C. D., Bacastow, R. B., Bainbridge, A. E., Ekdahl, C. A., Guenther, P. R., Waterman, L. S., and Chin, J. S. (1976b). Carbon dioxide variations at Mauna Loa Observatory, Hawaii. Tellus 28, 538.

8. Legget, J. Pepper, W. J., and Swart, R. J., (1992). Emission scenarios for IPCC: An update, Climate Change: Supplementary Report to the IPCC Scientific Assessment, IPCC, Cambridge Univ, Press, Cambridge.

9. Kattenburg, A., Giorgi, F., Grassl, H., Meehl, G. A., Mitchell, J. F. B., Stouffer, R. J., Tokioka, T., Weaver, A. J., and Wigley, T. M. (1996). Climate models-projections of future climate, Climate Change 1995, p. 285-357. IPCC Cambridge Univ. Press, Cambridge.

10. Manabe, S., Spelman, M. S., and Stouffer, R. J. (1992). Transient responses of a coupled ocean-atmosphere model to gradual changes of atmospheric CO₂, Part II: Seasonal response. J. Clim. 5, 105-126.

11. Mitchell, J. F. B., Johns, T. C., Gregory, J. M., and Tett, S. B. F. (1995). Climate response to increasing levels of greenhouse gases and sulfate aerosols. Nature 376, 501-504.

12. Mitchell, J. F. B., and Warrilow, D. A. (1987). Summer dryness in northern mid-latitudes due to increased CO₂. Nature 330, 238-240.

13. Mitchell, J. F. B., Manabe, S., Meleshko, V., and Tokioka, T. (1990). Equilibrium climate change and its implications for the future. Climate Change: IPCC Scientific Assessment, J. T. Houghton, G. I. Jenkins, and E. J. Ephraums, Eds. Cambridge Univ. Press, 131-164, Cambridge.

14.Wigley, T. M. L. (1998). The Kyoto Protocol: CO₂, CH₄ and climate implications. Geophys. Res. Lett. 25, 2285-2288.

15. Wigley, T. M. L., Jain, L. A., Joos, F., Shukla, P. R., and Nyenzi, B. S. (1997). Implications of Proposed CO₂ Emission Limitations: IPCC Technical Paper 4, 41 pp. IPCC, Geneva.PRINTCONTENT; print $pagecontent; include "/var/www/html/core/partf";print "last modified:". date( "F d Y.", getlastmod() ); include "/var/www/html/core/partg";