Predicting Shortwave Cloud Feedback in Models and Observations Based on the Seasonal Cycle of Liquid Water Path

Thursday, 18 December 2014: 4:45 PM
Paulo Ceppi, Daniel McCoy and Dennis L. Hartmann, University of Washington, Department of Atmospheric Sciences, Seattle, WA, United States
Shortwave cloud (SWcld) feedbacks in the midlatitudes are primarily caused by changes in cloud optical depth, in turn related to liquid water path (LWP). However, large uncertainties in the LWP response to climate change cause inter-model spread in SW feedbacks. Here we present a novel method to compare extratropical shortwave cloud feedbacks in models and observations based on the sensitivity of liquid water path to temperature and humidity in a reference climate.

We first regress monthly-mean LWP onto 850 hPa temperature and relative humidity at each point across the seasonal cycle in the historical climate of CMIP5 models and in observations, and multiply the regression coefficients by the RCP8.5 temperature and humidity response to yield a “predicted LWP response.” This compares very well with the actual response (0.70 < r < 0.95) across the extratropics (35°-70° N and S) in the annual and zonal mean. We then determine the relationship between historical monthly-mean LWP and SW cloud-radiative effect (CRE) for each model, latitude, and month, excluding land gridpoints. For observations, the LWP data are combined with CERES retrievals for 2000-2008. These LWP-SWCRE relationships then allow us to predict the SW response for a given LWP change.

The predicted SWcld feedback is generally well-correlated with the actual SWcld feedback in the Southern Hemisphere (0.45 < r < 0.75), while the prediction has less skill in the Northern Hemisphere. In the Southern Hemisphere, the SW feedback prediction has a mean negative bias of about 1 W m-2, likely due to changes in cloud microphysics not accounted for by our model. Importantly, however, the SWcld feedback predicted from observations agrees very well in magnitude and meridional structure with the multi-model mean predicted SWcld feedback. Both models and observations tend to predict a meridional dipole in SWcld feedback with a transition from negative values in the midlatitudes to positive values in the subtropics, similar to the actual SWcld feedback in RCP8.5 simulations.