Submitted to: Photogrammetric Engineering and Remote Sensing
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: January 6, 2012
Publication Date: September 1, 2012
Citation: Goslee, S.C. 2012. Topographic corrections of satellite data for large-scale monitoring. Photogrammetric Engineering and Remote Sensing. 78:973-981. Interpretive Summary: As part of the joint USDA NRCS and ARS Conservation Effects Assessment Program, the possibility of using satellite imagery for regional monitoring of grazinglands is being explored. Several difficulties must be overcome when using Landsat or other satellite images for such monitoring, including the shading effects caused by changing Sun angles in mountainous areas. Six topographic correction methods were tested for Landsat imagery of central Pennsylvania to identify the methods that were most effective for regional programs. Two methods, the c-correction and the Minnaert correction, were the most effective at removing topographic shading effects, although neither was entirely successful. This research contributes to the development of a regional monitoring program for grazingland extent and productivity that will improve understanding of the environmental effects of pasture management practices.
Technical Abstract: Satellite imagery has become widely available, creating the potential for long-term regional monitoring programs. To enable comparison of images taken at different times, topographic effects on reflectance caused by the interaction of sun position and terrain slope and aspect must be corrected. Previous studies of topographic correction methods have often used only single scenes taken under good conditions, and even then have found mixed effects. Six correction methods were tested on subsections of two Landsat scenes in central Pennsylvania. Both TM and ETM+ platforms were included, and images were taken at low, medium and high solar elevations. The effects of the corrections on the distribution of reflectance values and two ratio-based vegetation indices, on the correlation with illumination, and on the difference between north- and south-facing slopes were assessed. Lambertian methods should not be used for large-scale monitoring: these corrections are not band-specific, and thus do not remove topographic effects on vegetation indices. Instead, band-specific non-Lambertian methods such as the c-correction or Minnaert methods should be used. Incorporating slope categories and land cover in calculations of the band-specific correction coefficients may improve the results under certain circumstances. The recommended methods are still not ideal: neither provides effective correction under conditions of low or negative illumination, though such conditions are often found at low solar elevations. Topographic corrections are one component of an suite of atmospheric and platform-specific corrections; such an integrated approach may offer a way to reduce topographic effects even at low illumination.