Assessing the spatial, spectral, and temporal consistency of topographically corrected Landsat time series composites across the mountainous forests of Nepal

Abstract

In mountainous regions slope and aspect result in variations in the illumination condition and the same land cover type can therefore show differences in the reflectance depending on the orientation of the terrain slope towards the sun and the sensor of the satellite. Different topographic illumination correction methods exist and their performance varies depending on e.g. sun zenith and land cover type. Similarly a variety of evaluation criteria exist to assess the performance of the topographic correction methods and each evaluation criteria usually only considers certain aspects of the ability of the topographic correction method in reducing signal-to-noise. In this article we present a novel framework for the evaluation of topographically corrected image composites using cloud computing. We evaluated six topographic correction methods (Bin Tan, C-Correction, Minnaert with slope, Sun-Canopy-Sensor plus C-Correction, Statistical-Empirical, and Variable Empirical Coefficient Algorithm) in forest areas of four Landsat footprints in Nepal for a time series of image composites from 1988 to 2016. Our evaluation shows that the Statistical-Empirical topographic correction method provides the best overall performance, but in some years and footprints also other methods can show the best performance. We discuss these differences and related reasons in detail and give recommendations on the best use and evaluation of topographic correction methods.