Evaluating cloud parameterizations in a climate model with satellite observations

Warm cloud microphysics is a key process in global climate models that influences representation of aerosol indirect effect in the models. Parameterizations of the microphysical processes typically contain uncertain parameters that determine how the water conversion rate depends on cloud properties and aerosol amount. Climate model representation of the aerosol indirect effect is largely dependent on how to “tune” some of these uncertain parameters. This study examines the validity of such uncertain parameterizations assumed in GFDL CM3 with satellite observations to evaluate their characteristics in terms of microphysical process representation. The model configurations examined include (i) alternate values of the threshold particle radius that triggers liquid precipitation and (ii) different representations of sub-grid cloud variability in microphysics parameterizations with varying complexities in modeling the microphysics-turbulence interaction. Given that these two types of differing configurations have a significant impact on magnitude of the aerosol indirect forcing, their process-level evaluations are critical for more reliable estimates of aerosol effects on climate. For this purpose, methodologies developed to analyze multi-sensor satellite observations are employed to construct the statistics that fingerprint process signatures of the warm rain formation. The statistics are compared among different model configurations and satellite observations in an attempt to constrain which configuration is more plausible than others in terms of microphysical process representation. One of the highlighted results demonstrates that the model predictability of twentieth-century historical temperature trends contradicts the satellite-based constraint on the threshold particle radius that has been one of the typical “tunable knobs” in climate models. This contradiction implies the presence of compensating errors at a fundamental level in the model, and underscores the importance of observation-based, process-level constraints on model microphysics uncertainties.