Stratocumulus Sensitivity to Aerosols and Dynamics: Evaluating Aerosol-Cloud Parameterizations

Global changes in cloud properties have the potential to significantly impact the Earth's energy balance. Due to the strong cooling effect of stratocumulus clouds, it is of particular importance to quantify their sensitivities to changing aerosol and dynamical forcings. Prior observational studies have shown instantaneous correlations between aerosols and cloud properties, but have generally been unable to test if these correlations reflect a true causal relationship. Mauger and Norris (GRL, 2007) recently presented a new technique for separately quantifying the impacts of aerosol and dynamical forcings on clouds. The method uses HySPLIT back trajectories to control for the influence of meteorological history on cloudiness. By combining MODIS observations with ECMWF operational analyses, Mauger and Norris found that covariation between aerosol optical depth and lower tropospheric stability (LTS) during the previous 48 hours led to an overestimate of the cloud sensitivity to aerosols. Controlling for variations in LTS reduced the estimated sensitivity by 54%. The present work extends the analysis by estimating partial derivatives of cloud properties with respect to aerosols and meteorology, and by applying the technique to model evaluation. Both GFDL and NCAR have recently implemented interactive aerosol-cloud schemes in their GCMs. Prior validation studies have typically focused on comparison of mean fields. By instead examining cloud response on daily time scales, this method provides new diagnostic information on model performance. Specifically, model and observational sensitivities are computed by estimating partial derivatives of cloud properties with respect to aerosol and meteorological forcings. Partial derivatives are estimated by compositing data into high, low, and middle terciles, and considering variations in one variable while holding others constant. The results provide a set of statistically robust estimates of stratocumulus sensitivities, obtained from both observations and model output. Comparison of the two provides unique information for model validation and for diagnosing sources of model error.