Validating Machine-Learning Retrievals of Cloud Droplet Effective Radius Over Ocean that Account for 3D Radiative Transfer Effects

Le résumé fourni par la source

Bi-spectral retrievals of droplet effective radius (re) from instruments such as MODIS are widely utilized to study cloud microphysics in marine boundary layer clouds. These retrievals are known to have systematic errors due to cloud heterogeneity. Here, we develop a neural-network regression to correct for pixel-by-pixel errors in retrieved re using four features available in the MODIS L2 product. The neural-network regression is trained on 3D radiative transfer simulations of quasi-adiabatic stochastically generated clouds and corrects relative errors in re with respect to cloud-top with an r2 of 0.88. The neural-network regression produces unbiased retrievals of re against Large Eddy Simulation cloud fields where the bi-spectral retrieval has biases reaching +100% in cumuliform conditions. The neural-network regression reduces retrieval biases against airborne observations of cumulus from CAMP2Ex from +100% to +40%, and marginally improves already good consistency against stratocumulus sampled during VOCALS. A cross-comparison technique for assessing statistical remote sensing retrievals is introduced. The neural-network regression explains 63% and 91% of the variance in the differences between MODIS 1.6 µm and 2.1 µm retrievals for Overcast and Partially Cloudy Pixels (PCL) and 42% and 76% for the 2.1 µm and 3.7 µm differences, respectively. Residual spectral inconsistency is partially attributed to precipitation-sized particles using radar observations. Regional biases in the operational MODIS re are predicted that reach +50% for Overcast pixels in the tropics and are consistently +70% for PCL pixels. Errors in bi-spectral retrievals due to heterogeneity are non-random at both the cloud and climate scale.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.