Sensitivity of cloud property retrievals to differences in radiative transfer simulations

Accurate Radiative Transfer Model (RTM) simulations are the basis for retrieval of cloud properties from satellite radiances. Since the relationship between cloud radiances and cloud properties is not linear small differences in simulated radiances can result in large differences in retrieved cloud properties. This paper presents the sensitivity of cloud optical thickness and droplet effective radius retrievals to typical differences in radiative transfer simulations. Four widely accepted RTMs for multiple scattering calculations: Monte Carlo, MODTRAN4v2r0 (beta release), DAK and SHDOM, are intercompared to assess the differences in radiative transfer simulations. For two wavelengths (0.63 and 1.61 m) plane parallel water cloud simulations are compared for a wide range of cloud properties and viewing geometries. The results show that radiative transfer simulations differ between 3% and 10%, due to differences in model parameterizations, number of streams, scattering phase function and treatment of the forward scattering peak.

The sensitivity of cloud property retrievals to differences in radiative transfer simulations is examined for NOAA16-AVHRR cloud properties retrievals. The retrieval algorithm used is the one planned in the Climate Monitoring SAF of EUMETSAT for meteorological satellites. The sensitivity study reveals that the differences in cloud property retrievals are very sensitive to viewing conditions and cloud characteristics. 3% error in simulated radiances at 0.63 m accounted for differences in optical thickness up to 40% with increasing optical thickness. 3% error in the simulated radiances at 1.61 m accounted for about 2 m differences in the droplet effective radius for clouds with optical thickness greater than 5. The retrieval of cloud optical properties was most uncertain for clouds with optical thickness lower than 5 and greater than 50.
.