Retrieval of cloud phase and ice crystal habit from multi-angle satellite data

Knowledge of cloud phase and crystal habit is important both for accurate remote sensing retrievals and for improving climate simulations. Satellite remote sensing retrievals of cloud properties, including optical depth and particle size, depend strongly on the particle phase (ice or water) and ice crystal habit (shape) via the optical properties assumed in the retrieval algorithm. New ground-based remote sensing retrievals which use radar reflectivity and Doppler velocity also depend strongly on particle shape, due to the dependence of fall velocity on particle shape and size.

The shape, size, and composition of cloud particles affect the way in which they scatter light. Consequently the radiances measured by the Multi-angle Imaging Spectro-Radiometer (MISR) instrument aboard NASA's Terra satellite, which observes light scattered from the same cloud at nine different angles, are functions of the crystal phase and habit. In principle, the measured angular pattern can be used to infer the particle phase and crystal shape. By combining the MISR multi-angular visible radiances with shortwave infrared radiances from the MODerate Resolution Imaging Spectroradiometer (MODIS) instrument also on the Terra platform, sensitivity to particle size is also incorporated, reducing ambiguity in the retrievals.

In our retrieval, we use the MODIS 2.1 um band and the nine observations in the MISR blue band (0.448 um). The MISR blue band was chosen to minimize the effect of surface reflectance on the retrieved properties over land. We have generated a look-up table of reflectance as a function of optical depth, effective radius, crystal habit, and solar and satellite geometry. The differences between the observed reflectance in each MISR camera and the MODIS 2.1 um band are minimized to retrieve the best fit cloud properties. We present statistics of our retrievals of cloud properties, including particle phase, habit, and size, over the Atmospheric Radiation Measurement (ARM) sites in Oklahoma and the tropical western Pacific. We examine the sensitivity of the retrieved properties to various assumptions in the retrieval algorithm and compare our results to the operational MODIS retrievals.