Normalization of Landsat thermal imagery for the effects of solar heating and topography

Abstract

This paper is an investigation of three simple normalization procedures for suppressing the effects of solar heating and topography in daytime thermal data. The first method is the hyperspherical direction cosine (HSDC) transformation, which separates the pixel vector into an illumination/albedo component and a spectral component. The second method, a model correction, is based on the assumption that, once an elevation correction using the normal lapse rate has been applied, temperatures are proportional to the instantaneous solar heating as measured by the cosine of the solar illumination incidence angle. The third method is a statistic-empirical correction. These three normalization methods were applied to a test site in the Humboldt Range, Pershing County, Nevada, using Landsat Thematic Mapper data. It was found that geological patterns were much clearer in the normalized data than in the original temperature information. The HSDC correction brought out lithological differences, helped discriminate between gravels and spectrally similar sedimentary rocks and resulted in a significant increase in classification accuracy. The model correction appeared to inadequately compensate for the cool temperatures found at high elevations, and therefore underestimates the actual decline in temperature with elevation. Nevertheless, the rock contacts are relatively clear, and the classification produced the second highest overall accuracy. The statistic-empirical classification resulted in improved elevation correction, but it over-corrected north-facing slopes and produced only intermediate improvements in accuracy.