Using In Situ Observations from the SOCRATES Field Campaign to Evaluate Satellite Retrievals of Low Clouds over the Southern Ocean

Abstract

Satellite retrievals of cloud properties have been widely used to study clouds over the Southern Ocean, but our confidence in these retrievals has been limited by a lack of verification studies due to a lack of in situ observations. In this study, cloud properties observed from aircraft during the Southern Ocean Cloud Radiation Aerosol Transport Experimental Study (SOCRATES) in January and February of 2018 are used to evaluate cloud properties from three satellite-imager retrieval datasets that are based on the bi-spectral (Nakajima-King) technique: (i) The operational Moderate Resolution Imaging Spectroradiometer (MODIS) level 2 (collection 6.1) cloud optical properties product, (ii) The CERES-MODIS Edition 4 cloud product, and (iii) the NASA SatCORPS Himawari-8 (geostationary) cloud product. Satellite retrieved cloud optical depth, effective radius, liquid water path, and cloud droplet number concentration are evaluated. The analysis focuses on the use of vertical profiles of cloud properties constructed from individual aircraft penetrations through overcast stratocumulus. Overall the satellite retrievals compare well with the in situ data for these conditions, with little bias (no statistically significant bias at the 95% level of confidence) and modest to good correlation coefficients, when considering all aircraft profiles for which there are coincident MODIS observations. The SatCORPS Himawari-8 product does, however, show a statistically significant mean bias for re of about 1.2 μm when applied to a larger set of profiles with coincident Himawari-8 observations. A close examination of the data shows that the low overall mean-bias in the retrievals is due in part to compensating errors between cases (profiles) that are non- or lightly-precipitating, with cases that have heavier precipitation (precipitation water path > 10 gm-2). The effective radius is slightly biased high (by about 0.5 to 1.0 μm) for non- and lightly-precipitating cases and biased low by a large amount (about 3 to 4 μm) for more heavily-precipitating cases. The bias in non- and lightly-precipitating conditions is due (at least in part) to having an assumed drop size distribution (DSD) in the retrieval that is too broad. For heavily-precipitating cases, large biases occurred when significant precipitation was found near cloud top. These biases in the effective radius ultimately propagate into the retrieved liquid water path and number concentration.