INDEPENDENT PIXEL AND 2-DIMENSIONAL ESTIMATES OF LANDSAT-DERIVED CLOUD FIELD ALBEDO

A theoretical study has been conducted on the effects of cloud horizon tal inhomogeneity on cloud albedo bias. A two-dimensional (2D) version of the Spherical Harmonic Discrete Ordinate Method (SHDOM) is used to estimate the albedo bias of the plane-parallel (PP-IPA) and independe nt pixel (IPA-2D) approximations for a wide range of 2D cloud fields o btained from Landsat. They include single-layer trade cumulus, open an d closed cell broken stratocumulus, and solid stratocumulus boundary l ayer cloud fields over ocean. Findings are presented on a variety of a veraging scales and are summarized as a function of cloud fraction, me an cloud optical depth, cloud aspect ratio, standard deviation of opti cal depth, and the gamma function parameter v (a measure of the width of the optical depth distribution). Biases are found to be small for s mall cloud fraction or mean optical depth, where the cloud fields unde r study behave linearly. They are large (up to 0.20 for PP-IPA bias, - 0.12 for IPA-2D bias) for large v. On a scene-average basis, PP-IPA bi as can reach 0.30, while IPA-2D bias reaches its largest magnitude at -0.07. Biases due to horizontal transport (IPA-2D) are much smaller th an PP-IPA biases but account for 20% rms of the bias overall. Limitati ons of this work include the particular cloud field set used, assumpti ons of conservative scattering, constant cloud droplet size, no gas ab sorption or surface reflectance, and restriction to 2D radiative trans port. The Landsat data used may also be affected by radiative smoothin g.