Cloud-clearing algorithm error vs. sensor footprint size and configuration with optional multispectral MODIS imagery

Abstract

We provide an evaluation of cloud clearing error and sensor footprint size in high spectral resolution sounding instrumentation. Input data are actual atmospheric spectra collected by the NPOESS Aircraft Sounding TestbedInterferometer (NAST-I) [see Cousins and Smith (1998), and Smith et. al., (1999, 2001]. NAST-I data is averaged to create sensor configurations of varying field of view size and array number. The cloud-clearing techniques, based on the N* approach (Smith, 1968, Chahine, 1977, and McMillin and Dean, 1982, use the linear relation between observed radiance and cloud amount to extrapolate the radiance observed for two adjacent fields of view possessing differing cloud amounts to the cloud free value. The option of including MODerate resolution Imaging Spectroradiometer (MODIS) style data was also investigated. With the MODIS filter, the assumption of cloud emissivity homogeneity is not needed because of the MODIS high spatial resolution spectral channels in which the clear air radiance can be defined for a scene. This relaxation of the need for cloud optical property homogeneity enables a higher yield of valid clear column radiance estimates. We show that the use of MODIS-like multi-spectral imagery data in the cloud clearing of high spectral resolution sounder data will minimize the dependence of the sounding retrieval accuracy and yield on instrument field-of-view size. The errors of the multi-spectral MODIS cloud-cleared spectral radiance are generally a factor of two lower than those errors associated with the use of a single window channel for the cloud clearing of radiance spectra.