Illumination modelling for topographic correction of Landsat 8 and Sentinel-2A imageries

Automated classification algorithms for satellite imageries require spectral correction from terrain effects due to shading. Such terrain effects can produce reflectance bias of pixels in the same category. This study was aimed at exploring robust algorithms for correcting satellite imageries from terrain effects, applicable for either Landsat 8 or Sentinel-2A imageries. Mount Halimun-Salak and Mount Gede-Pangrango, Bogor, West Java, Indonesia, were selected as the window areas to evaluate the algorithm. We developed algorithm, which combined solar position modelling, illumination modelling, and simple statistical model to remove the terrain effects. The algorithm was proven to be able to solve overcorrection problems and result relatively consistent sensitivities in SWIR, NIR and blue bands of either Landsat 8 or Sentinel-2A imageries in both window areas from different acquisition dates and times.