Retrieval of cloud fraction and optical thickness of liquid water clouds over the ocean from multi-angle polarization observations

Abstract

We introduce a novel method to retrieve the cloud fraction and the optical thickness of liquid clouds over a water surface based on polarimetry. The approach is well suited for satellite observations providing multi-angle polarization measurements, in particular those of the Hyper-Angular Rainbow Polarimeter #2 (HARP2). Unlike commonly used methods to derive cloud fractions, our method does not depend on the spatial resolution of observations, and it does not require any threshold values for cloud detection. Based on radiative transfer simulations, we show that the cloud fraction and the cloud optical thickness can be derived from measurements at two viewing angles: one within the cloudbow and one in the sun glint region. In the cloudbow, the degree of polarization mainly depends on the cloud optical thickness. Conversely, for a viewing direction in the sun glint region, the degree of polarization depends on the clear fraction of the pixel, because here the radiation scattered by cloud droplets is almost unpolarized, whereas radiation reflected by the surface is highly polarized. Utilizing these dependencies, we developed a retrieval using a simple lookup table approach. Based on sensitivity studies, we show that prior information about wind speed and aerosol optical thickness improves the accuracy of the cloud fraction retrieval. Prior information about the cloud droplet size distribution can reduce the uncertainty of the cloud optical thickness retrieval. The prior information should be obtained by combining our method with already existing aerosol and cloud retrieval algorithms. We performed 3D radiative transfer simulations and found that the cloud optical thickness is generally underestimated due to neglect of 3D scattering effects. The cloud fraction is overestimated in cloud shadows and underestimated in in-scattering regions. As a demonstration, we apply the methodology to airborne observations from polarization cameras of the Munich Aerosol Cloud Scanner (specMACS) instrument. The high-spatial-resolution data (10–20 m) have been averaged to a spatial resolution of approximately 2.5 km to mimic satellite observations. From the average linear polarization at scattering angles of 140 and 110°, we derive continuous cloud fraction values and the corresponding cloud optical thicknesses. Comparison for cases including low, medium, and high cloud fractions shows that the retrieval, using only reflected polarized radiances at two scattering angles, provides accurate estimates of the cloud fraction for observations with coarse spatial resolution.