Another active area of my research is investigating how cloud horizontal and vertical inhomogeneities affect the retrieval of cloud properties from satellite. Recently launched instruments, such as the Moderate Resolution Imaging Spectroradiometer on the Terra and Aqua satellites allow information about clouds (e.g., effective particle sizes, cloud top temperature, extinction optical depths, phase of particles) to be retrieved from multi-spectral radiances. However, only one value describing the entire depth of the cloud is retrieved.

Balloon-borne replicator measurements of size and shape distributions of ice crystals have shown that clouds have definite vertical structures, with three distinct regimes. The nucleation regime at cloud top contains quasi-spherical particles that have been recently nucleated, the growth regime in the middle of the cloud contains pristine particle shapes which are growing, and the sublimation regime at cloud base contains particles that are starting to evaporate with rounded edges. Figure 1 shows derived profiles of extinction optical depth and effective radius from one of these cases. There is substantial vertical variability.

The principle behind current cloud retrieval algorithms is to compare the reflectance in a visible channel with the reflectance in a near-infrared channel to estimate effective radius and extinction optical depth. Figure 2 show an example of such a retrieval, where the reflectances are computed using observed mixtures of different crystal shapes rather than the single crystal habits that current retrieval algorithms assume. Because of considerable variability in what habits occur in different cloud types, such retrieval algorithms may have to be developed separately for different cloud types (convective, non-convective, maritime, continental) and different geographical locations (Arctic, mid-latitude, tropical). Research is proceeding on these issues.

We are also performing research to determine what parts of the cloud different wavelengths construct. Because effective particle size depends substantially on near-infrared reflectance, we do this by constructing weighting functions which describe how sensitive the reflectance is to different parts of the cloud. Such weighting functions can be constructed using vertical profiles of cloud properties observed during quasi-Lagrangian descents of aircraft during projects such as the FIRE1 project. Despite differences between days, the weighting function seems relatively constant. Investigations are continuing to construct weighting functions for other cloud types and geographical locations.