A radiative analysis of angular signatures and oblique radiance retrievals over the polar regions from the multi-angle imaging spectroradiometer.

摘要

This dissertation studies clouds over the polar regions using the Multi-angle Imaging SpectroRadiometer (MISR) on-board EOS-Terra. Historically, low thin clouds have been problematic for satellite detection, because these clouds have similar brightness and temperature properties to the surface they overlay. However, the oblique angles of MISR observe great contrast between these clouds and any underlying surface. This work demonstrates how the MISR instrument can contribute to our understanding of clouds in the polar regions.This dissertation makes the following key contributions to science: (1) With the aid of the MODerate Resolution atmospheric TRANsmission (MODTRAN) radiative transfer model, this dissertation provides the first explanation of the conditions that lead to the brighter or darker appearance of clouds relative to the surface at nadir by introducing the transition albedo. The transition albedo shows that clouds darken the surface under three conditions: (1) high surface albedos, (2) high solar zenith angles (low sun), and (3) thin clouds. Thin clouds over the polar regions fit all three criteria, and are therefore more likely to be missed by visible satellite retrievals. (2) This dissertation uses the MODTRAN model to explain the angular distribution of relative brightness between clouds and the snow surface through differences in absorption and sphericity of particles that make up clouds and snow surfaces. The effects of sphericity and absorption suggest that low level clouds appear bright at oblique forward-scattered angles because of the presence of small, spherical droplets. (3) This dissertation provides the first evaluation of several popular ice surface reflectance models in MODTRAN using MISR observations as model restraints. The Lambertian model is shown to underpredict BRF values in the blue and green bands unless au unrealistically thick aerosol layer is included. The Hapke 3-Parameter Model contained several programming errors which could not be completely resolved for bright surfaces. The Mishchenko Radiative Transfer Solution for Flat Layers outperformed both the Lambertian and Hapke models, but overpredicted forward scattering and underpredicted backscatter. The use of a Koch Fractal model decreased forward scattering, but still failed to catch the increase in backscatter noted in MISR retrievals. (4) This dissertation develops the Angular Signature Cloud Mask (ASCM) into a standard MISR product and evaluates the performance through global and regional analyses. The Band-Differenced Angular Signature (BDAS) uses the oblique cameras from MISR in the direction of forward-scattered radiation, for which low clouds over ice are a strength. The Angular Signature Cloud Mask (ASCM) applies thresholds to the BDAS to determine whether clouds are present for any given pixel. The Support Vector Machine (SVM) is used as a truth dataset for comparison to the ASCM. Agreement rates between the SVM and ASCM were 80.2% over water, 76.0% over land, and 90.0% over ice-covered surfaces.