Atmospheric effects in multispectral remote sensor data

Abstract

The problem of radiometric variations in multispectral remote sensing data which occur as a result of a change in geometric and environmental factors is studied. The case of spatially varying atmospheres is considered and the effect of atmospheric scattering is analyzed for realistic conditions. Emphasis is placed upon a simulation of LANDSAT spectral data for agricultural investigations over the United States. The effect of the target-background interaction is thoroughly analyzed in terms of various atmospheric states, geometric parameters, and target-background materials. Results clearly demonstrate that variable atmospheres can alter the classification accuracy and that the presence of various backgrounds can change the effective target radiance by a significant amount. A failure to include these effects in multispectral data analysis will result in a decrease in the classification accuracy.