There is current debate among climatologists on the response of climate to internal or external stimuli, such as changes in the CO₂ content of the atmosphere or of the solar constant. Much of this discussion centers around the effects that two radiative forcing agents, greenhouse gases and solar radiation, have on cloud cover. It is thought that cloud cover exerts two opposing influences upon climate. On the one hand, clouds reflect solar radiation which tends to having a cooling effect on the earth's climate. On the other hand, because cloud top temperatures are usually colder than the temperature of the earth's surface, cloud cover reduces the effective temperature for outgoing terrestrial radiation, thus contributing to climate warming.

These feedback mechanisms related to clouds are truly complex. As expected, multiple studies have been done in order to determine if clouds act as a positive or a negative feedback to global warming trends. In 1967, Manabe and Wetherald suggested that low clouds have a strong net cooling effect on climate because of the high surface albedo and relatively high cloud-top temperature. Their results also indicated that high clouds have a net heating effect due to small albedo and relatively low cloud-top temperatures. Overall, their study determined that total cloud cover cools the climate. Likewise, Smagorinsky, in 1978, speculated that the increase in downward radiative flux, due to higher levels of carbon dioxide, enhances evaporation, increases the amount of low cloud, and, therefore, cools the climate.

But others dispute these findings. In a 1978 study, Roads indicated that cloud variations may have a positive feedback on the sensitivity of the global mean climate. He suggested that an increase in sea surface temperature leads to stronger vertical velocities and the removal of moisture from the atmosphere through more intense precipitation. This results in a lower relative humidity, less cloudiness, a smaller reflection of solar radiation, and, thus, further warming of the earth's surface and the air above. Schneider concurred with the positive feedback effect in his 1978 study.

Manabe and Wetherald (1980) undertook a study of the radiative effect of changes in both the amount and height of cloud cover using two versions of a simple atmospheric general circulation model (GCM). For the first version, climate data is obtained for various values of the solar constant over a model run time of 1200 days in which cloud cover is allowed to respond to the radiative forcing. The second version of the model is carried out using the same solar constant values but with cloud cover distributions kept at a fixed level. Temperature, precipitation, and wind data of the two model runs were then compared to determine the influence of cloud feedback on model climate sensitivity.

The first model version, called "variable cloud" (VC), shows that the meridional temperature gradient in the lower troposphere is significantly reduced in response to the increase of the solar constant. This reduction is caused by the poleward retreat of highly reflective snowcover and the increase in the poleward transport of latent heat, both of which result from the general warming of the atmosphere. With increases in insolation of 2, 4, and 6% from the norm, mean surface air temperature also increases, along with precipitation and evaporation rates. The precipitation rate is especially large poleward of 50 degrees latitude, whereas the evaporation rate does not vary much with latitude. The latent heat transport is responsible for these results, which show precipitation rates as much as 24% higher in response to the 6% increase in the solar constant.

VC run results also show that the intensity of the zonal wind is greatly reduced from 15 degrees to 40 degrees latitude when insolation is at a higher level. This is in response to the decrease of the meridional temperature gradient and how that affects the thermal wind relationship. It is also thought that the reduction in the gradient may cause the weakening of the direct circulation of the Hadley and polar cells and the indirect Ferrel cell.

The cloud response to the solar constant increase indicates that cloud amount generally diminishes in most of the troposphere with the exception of the layer near the earth's surface. Here the amount of low cloud increases poleward of 50 degrees and in the subtropics but decreases in the tropics and the middle latitudes. In the lower stratosphere cloud amount tends to increase, especially in high latitudes. The reduction of cloud cover in the middle and upper troposphere is responsible for not only the decrease of total cloudiness but also the lowering of the effective cloud top height in lower latitudes. Because of the compensation between the change in total cloudiness of the high latitudes versus the low latitudes, the total cloud amount over the entire model domain changes little in response to the increase of the solar constant.

When the second version, or "fixed cloud" (FC), of the model run is compared to the VC run, the responses are very similar. The comparison reveals that the magnitudes of atmospheric warming, precipitation, and wind velocities differ little between the VC and FC model runs. These results seem to indicate that, for the cloud parameters used in this study, cloud feedback mechanisms have a relatively small effect on climate sensitivity despite a strong influence of cloud cover on solar and terrestrial radiation. A short time later, in 1980, Manabe and Wetherald conducted another test using the same model, this time varying the CO₂ concentrations in the atmosphere instead of solar constant values. The results of the later study mirrored those of the first; CO₂-induced climate change is hardly affected by the cloud feedback mechanism.

Yet once again, Manabe and Wetherald (1988), prompted perhaps by disagreements with their earlier findings, studied cloud feedback processes with a GCM. This investigation, undertaken in 1987, was a continuance of their prior work in 1967, 1975, 1979, 1980, and 1986. It utilizes two versions of the model, again one identified as VC, which contains the cloud feedback process, and the other, FC, which does not. A numerical integration for each model version is run over a 40 year period, first with a CO₂ concentration of 300 ppm and then at a doubling of 600 ppm. By comparing the alterations of the climates of the two versions, the influence of the cloud feedback process upon the sensitivity of climate can be shown.

This time the comparison between the two model runs shows an increase in global mean surface air temperatures, with the VC run to be considerably greater than the FC run (4.0 degrees C compared to 3.2 degrees C), except in high latitudes of the Southern Hemisphere where temperatures are very similar. On the other hand, the decrease in stratospheric temperatures is very similar for both model runs, which implies that cooling in the stratosphere due to additional amounts of CO₂ is relatively unaffected by cloud feedback mechanics. The VC run shows the following cloud patterns: a reduction of cloud amount in the upper troposphere in middle and low latitudes, an increase of cloud amount around the tropopause for all latitudes, and an increase of cloud amount in the stable region near the surface in higher latitudes. Altogether, total cloud amount decreases equatorwards of approximately 40§ latitude in both hemispheres and increases polewards of this latitude. In short, model results show that the change in cloud cover increases the net incoming short-wave radiation while it reduces the outgoing long-wave radiation. This interaction between cloud cover and radiation enhances the CO₂-induced warming of climate and is a positive feedback process.

The question that must be asked is what conditions or circumstances were different between the CO₂ doubling experiments done by the authors in 1980 and 1988 that caused them to alter their suggestion that the cloud feedback process has very little influence upon climate sensitivity? In both studies, cloud amount increases around the tropopause and is reduced in the upper troposphere at most latitudes in response to an increase of atmospheric CO₂. The magnitudes of these changes, however, are quite significant between the two studies. For example, the increase in cloud amount around the tropopause in low latitudes is very small in the first test, but is a much higher magnitude in the second test. Also, the increase of cloud amount in the surface layer of high latitudes in the earlier study is larger than that of the later study. In addition, the 1988 experiment shows low cloudiness is actually reduced in summer when the insolation is at a maximum, which tends to increase surface temperatures.

Another reason for the change in findings between the two studies is the relative primitiveness of the first GCM. Many simplifications, primarily dealing with momentum, heat, moisture, and cloud development, were made in the construction of the earlier model, and may have led to the differences in the study results. The time integration differences between the two models, 1200 days versus 40 years, may also have been a determining factor.

In conclusion, Manabe and Wetherald's latest model output shows that the increased high cloud amounts at all latitudes tends to warm the climate and produce positive feedback. This overshadows the smaller cooling effect or negative feedback of the low cloud increase at high latitudes. It is the smaller magnitude of this negative feedback, which was substantially higher for the earlier model runs, that accounts for the total positive feedback results for climate sensitivity now put worth by the authors. But with the number of studies previous done and those yet to be done on this subject by seemingly countless researchers, the debate of whether the cloud process constitutes positive or negative feedback is bound to continue for some time.

References

Wetherald, R.T. and S. Manabe, 1980: Cloud cover and climate sensitivity. Science, 37, 1485-1510.

Wetherald, R.T. and S. Manabe, 1988: Cloud feedback processes in a general circulation model. Science, 45, 1397-1415.