In this study, we examine the systematic and random variations in recorded radiance in two of the LANDSAT Thematic Mapper bandpasses for images obtained over a uniform agricultural area before and after harvest. Large variations will affect the accuracy with which conclusions may be reached regarding ground cover from an analysis of recorded multiband radiance data. We found the variations to be sufficiently large to suggest a need for caution in the quantitative analysis of multichannel digital radiance data to discriminate terrestrial cover types.