Multispectral analysis of maritime clouds at night in the presence of atmospheric water vapor

Abstract

Multispectral analysis methods are exercised using AVHRR channels 3, 4, and 5 to improve upon single-wavelength thermal imagery at night. An algorithm was developed yielding cloud location and water vapor distribution from channel 3-4 and 4-5 differences, respectively. Water vapor effects on pixel registration for cloud were examined using two candidate subscenes, one cloudy and dry, the other, cloudy and moist. A positive water vapor/cloud-free correlation was found using statistical techniques on the candidate subscene scatter plots. TOVS water vapor channels verified the analyzed water vapor pattern in the 4 - 5 difference image. V-shaped cloud/clear thresholds were applied to various subscene scatter plots to account for the positive correlation of water vapor-to-cloud registration. Results showed that in regions of higher water vapor concentrations, pixels nearest cloud boundaries were likeliest to be misclassified as clear. Images containing significant water vapor gradients and cloud variations required threshold refinement for best results.