Radiometric correction techniques and accuracy assessment for Landsat TM data in remote forested regions

Subtle change detection analysis in remote sensing relies on some form of radiometric consistency. Radiometric correction techniques developed in previous studies often require ancillary information such as climate data, illumination geometry, ground reference data of pseudo-invariant features (PIFs), and satellite calibration data. Most studies do not have the luxury of all these data. A relative radiometric correction technique of consistent quality applicable to study areas that lack urban development has not been generally accepted by the remote sensing community. A series of Landsat-5 thematic mapper (TM) and Landsat-7 enhanced thematic mapper plus (ETM+) images spanning 18 years was obtained for a primarily forested area in central British Columbia, Canada. Different techniques of radiometric correction that do not rely on ground reference data, climate data, or the subjective selection of PIFs were assessed for these images. They included an atmospheric transfer model that requires no ancillary climate data, a simple scaling function, and two scatterplot-based regression functions. Assessment of radiometric consistency was performed qualitatively by using edge detection and quantitatively using analysis of old-growth forests in equilibrium and measures of biomass accumulation in clearcuts. For these three methods of assessment, the two scatterplot-based regression functions yielded the best radiometric fidelity. These two techniques can be completely automated and are equally applicable in any Landsat TM- or ETM-based change detection studies.