Page 111
13
Model Performance
Adequate prediction of global surface warming caused by increasing greenhouse gases will require coupled models of the atmosphere, oceans, ice sheets and snow fields, and biosphere. These coupled models will have to provide time-dependent simulations that adopt, as input conditions, scenarios for future emissions of greenhouse gases. And, like the proverbial chain, the validity of these coupled models is governed by the weakest of the component models. The IPCC scientific assessment (Intergovernmental Panel on Climate Change, 1990) reviewed current results in these and related topics. To date, most emphasis has been placed on the development and testing of atmospheric models. Even though policymakers' would like them, current capabilities do not allow credible projections of regional effects. The panel finds that these and other aspects of climate models have even greater uncertainty than those associated with global mean temperature projections. However, for purposes of assessing their limits for policy decisions, the primary focus of the examination here is on global mean temperature.
Considerable effort has been focused on atmospheric GCM experiments in which the CO2 concentration of the atmosphere is instantaneously doubled and the models are then allowed to achieve a new equilibrium climate. Although these simulations do not provide information on time-dependent (or "transient") climatic changes that would accompany more realistic greenhouse gas accumulation scenarios, they do allow a means of testing, understanding, and comparing atmospheric GCMs. The IPCC scientific assessment (Intergovernmental Panel on Climate Change, 1990) provided a convenient summary of these simulations. For present purposes the only simulations considered are those that utilize computed clouds; i.e., that incorporate cloud feedback.
Page 112
The equilibrium global warming (i.e., dTs) produced by 18 different CO2 doubling simulations 1 is summarized in Figure 13.1. The simulation numbers are in order of increasing projection of global warming. Multiple simulations have been performed by five of the seven involved GCMs; these serve as sensitivity studies for a specific model. As an example, simulation numbers 4 and 5 (UKMO), and 8 through 10 (GFDL), respectively, proceed to a finer horizontal resolution. Note that neither model indicates a significant influence of horizontal resolution on the model-predicted global warming. The UKMO GCM produced both the greatest (5.2°C (9.4°F)) and the smallest (1.9°C (3.4°F)) global warming, and this notable variation is the consequence of differences in assumptions about cloud parameters (Mitchell et al., 1989).
The horizontal solution technique used in the seven GCMs is either finite difference (UKMO, GISS, OSU) or spectral (GFDL, NCAR, CCC, BMR). The spectral models are in much better agreement (dTs= 3.5° to 4.0°C (6.3° to 7.2°F)) than the finite difference models (dTs = 1.9° to 5.2°C (3.4° to 9.4°F)). This is probably coincidental. Of the 19 models in Figure 12.2, eight are finite difference and eleven are spectral, and here neither group is
Page 113
found to exhibit better agreement than the other. It should be noted that no persuasive comparison of model results with global warming observations has yet been constructed.
It is important to realize that the global warming results in Figure 13.1 include snow-ice feedback, whereas the sensitivity parameters in Figure 12.2 do not, and this partially explains differences between Figures 12.2 and 13.1. For example, simulation numbers 10 (GFDL) and 14 (GISS) in Figure 13.1 produce quite comparable global warming (4.0° and 4.2°C (7.2° and 7.6°F), respectively). The same GFDL and GISS models are, however, models number 12 and 19 in Figure 12.2. Relative to clear skies, Figure 12.2 shows rather modest positive cloud feedback in the GFDL model, whereas there is a very strong positive feedback in the GISS model.
That these two models agree well in Figure 13.1 is at least partially due to compensatory differences in snow-ice albedo feedback. Both modeling groups have provided feedback diagnostics so that individual feedbacks may be progressively incorporated, and this is demonstrated in Table 13.1. The two GCMs produce similar warming in the absence of both cloud feedback and snow-ice albedo feedback. The incorporation of cloud feedback, however, shows that this is a stronger feedback in the GISS model, as is consistent with Figure 12.2. But the additional incorporation of snow-ice albedo feedback largely compensates for their differences in cloud feedback. Thus, while the two models produce comparable global warming, they do so for quite different reasons.
It is emphasized that Table 13.1 should not be used to estimate the amplification factor due to cloud feedback, because feedback mechanisms are interactive. From Table 13.1 the cloud feedback amplifications for the GFDL and GISS models might be inferred to be 1.2 and 1.6, respectively, but only in the absence of snow-ice albedo feedback. If snow-ice albedo feedback is incorporated before cloud feedback, then the respective amplification factors are 1.3 and 1.8. These larger values are due to an amplification of cloud feedback by snow-ice albedo feedback.
The change in precipitation that would be concurrent with global warming is of considerable importance. Global mean precipitation change, for
TABLE 13.1 Comparison of Global Warming (°C) for the GFDL and GISS GCMs with the Progressive Additions of Cloud Feedback and Snow-Ice Albedo Feedback
|
Page 114
the same models as in Figure 13.1, is summarized in Figure 13.2 (Inter-governmental Panel on Climate Change, 1990; no value is reported for simulation number 7). Although there is considerable variability among the simulations, with reference to Figure 13.1 there is an obvious correlation between precipitation change and global warming. This is to be expected, since global warming enhances surface evaporation and hence precipitation.
To be more specific on this point, Figure 13.3 provides a scatter plot of precipitation change versus global warming. Although 17 simulations are represented here, there are four coincident points (one finite difference and three spectral). Clearly, differences among global warming simulation models are the cause of much of the variance in the predictions of global precipitation change. Note that the finite difference models exhibit a considerably stronger correlation, and there is no obvious explanation for this. The primary point of Figure 13.3 is that global precipitation change and global warming are, as would be expected, strongly coupled.
Of far more practical importance than the global averages are the regional patterns of changes in both precipitation and soil moisture. While global precipitation, for reasons discussed above, will increase with global warming, one would anticipate that there would be geographical regions where just the opposite occurs. Frequently, arid subtropics are a consequence of the descending branch of the tropical Hadley cell, and a shift in
Page 115
its location should bring with it a reduction of precipitation. Regional changes in soil moisture are even more difficult to predict, because soil moisture is the difference between precipitation and evaporation plus run-off, and errors in the change of any of these quantities will produce magnified errors in the fractional change of soil moisture.
As noted at the beginning of this chapter, persuasive projections of future climatic changes will require the availability of reliable coupled atmosphere, ocean, cryosphere, and biosphere models. As of now, in most models, the role assigned to the oceans does not include any horizontal transport. In many studies, in fact, the ocean temperature has been postulated as a boundary condition, often to provide a surrogate parameter from which potentially useful insights can be drawn. Recently, more elaborate models of the ocean have been coupled to GCMs in climate change simulations (Stouffer et al., 1989; Manabe et al., 1990), but efforts at this level are in their infancy. Clearly, there is a need here for model improvements, particularly with respect to cloud-climate interactions. And, of course, the other component models, as they evolve, will have to undergo similar scrutiny.
Page 116
Note
1. Not the same set as cited in Figure 12.2.
References
Intergovernmental Panel on Climate Change. 1990. Climate Change: The IPCC Scientific Assessment, J. T. Houghton, G. J. Jenkins, and J. J. Ephraums, eds. New York: Cambridge University Press.
Manabe, S., K. Bryan, and M. J. Spelman. 1990. Transient response of a global ocean-atmosphere model to a doubling of atmospheric carbon dioxide. Journal of Physical Oceanography 20:722–749.
Mitchell, J. F. B., C. A. Senior, and W. J. Ingram. 1989. CO2 and climate: A missing feedback? Nature 341:132–134.
Stouffer, R. J., S. Manabe, and K. Bryan. 1989. Interhemispheric asymmetry in climate responses to a gradual increase of atmospheric CO2. Nature 342:660–662.