Climate-Cloud Feedbacks in GCMs

Matthew T. Johnson


Successful climate modeling depends heavily upon an accurate description of cloud effects and cloud-climate feedback mechanisms. The accurate reproduction of cloud effects upon climate change, however, is complicated by the high spatial and temporal variations in clouds and their shifting physical characteristics. In addition, feedback mechanisms are not fully understood. A study which intercompared 14 GCM focused on the cloud feedback mechanisms present in each of the 14 models. Intermodel variations in global climate sensitivity, produced under equivalent model runs, served as a representative measure of the variation in cloud feedback mechanisms present in each model. Concluding the model runs, a three-fold range in global climate sensitivity was found, with most of the discrepancies being coupled with the different cloud feedback mechanisms present in each model.

These results were obtained through many assumptions and simplifications, albeit necessary, in order to feasibly approach the problem and reduce computation times. Atmospheric processes were key with less emphasis placed upon cryospheric and oceanic interaction. Climate change was assumed to be a two stage process, one of forcings and then response, with all variables being globally averaged quantities.

One quantity of utmost importance for this study was the global-mean direct radiative forcing (G), which physically can be interpreted as the net radiation flux at the top of the atmosphere, measured over all wavelengths. A reduction in wavelengths corresponding to those in the CO2 absorption spectrum at the top of the atmosphere (for example) would correspond to surface warming in the model runs. Mathematically, this is represented by

G = (Delta)F - (Delta)Q

where (Delta)F represents the change in the global-mean emitted infrared radiation and (Delta)Q represents the change in net downward solar fluxes, both being measured at the top of the atmosphere. Any changes in these quantities corresponded to active cloud feedback mechanisms present in the models. Additionally, another parameter, (Lambda) is defined, which separates cloud feedback process effects from other feedback process effects.

In order to isolate cloud-feedback mechanism processes in the model comparisons, several adjustments were necessary. First, clear-sky simulations were carried out, equivalent as a control for comparing model runs. The exact process used for this is similar to the process used in the Earth Radiation Budget Experiment (ERBE). Next, in order to overcome problems created by models running with different control climates and to reduce computation time, a method of using sea surface temperature perturbations as a substitution for climate change, and an "inverse climate change simulation" were used. In essence, this can be thought of as programming a given climate change (i.e. sea surface temperature perturbation) and then running the models to calculate the forcings which would bring about such a change. A third simplification assumed a constant July situation throughout the duration of the model simulations. This reduced clutter from other feedback processes, such as those in the cryosphere created by seasonal changes.

When it comes to the explicit treatment of clouds, the 14 models are capable of describing only two types, stratiform and convective, with the clouds covering either gridboxes or percentages of gridbox areas (as in the convective clouds).

Upon completion of the model runs, results indicated that under the clear-sky simulation, the models all produced similar climate sensitivities. However, large variations occurred between models when cloud feedback mechanisms were involved. In fact, there was a difference of up to 3x in modeled climate sensitivities between the 14 models. The type of feedback process produced also varied. Three models produced situations of slightly negative feedback, whereas the other 11 models produced positive feedbacks, some of considerably large magnitude.

These discrepancies foster the need for improvements in understanding cloud feedback mechanisms, which understandably, prove to be difficult. As we know, cloud impacts are not unidirectional. Increasing cloud coverage increases the absorption of infrared radiation, but decreases incoming solar radiation reaching the earth's surface. Only as research into this area continues, can models be improved.

However, one should note that this experiment did run under many assumptions and simplifications and did not include important oceanic effects or seasonal variations in the cryosphere, among other factors. Thus, the actual feedback processes occurring in the atmosphere may not be well represented. Yet, conclusions do strongly support that large variations in cloud feedback mechanisms exist in the models and more research is needed.


Reference