Date of Award
Master of Arts (MA)
Geography and Geology
Dr. Michael P. Bishop
The utilization of satellite imagery acquired over rugged terrain is problematic because of anisotropic reflectance. A variety of environmental factors, such as the atmosphere and the topography, cause significant variation in the irradiant and radiant flux. Consequently, satellite imagery must be radiometric ally corrected to account for these variations which influence the at-satellite spectral response. In mountain terrain, the topographic effect is very pronounced, and satellite imagery must be normalized utilizing various techniques. The purpose of this research was to evaluate various implementations of the Minnaert correction technique which were designed to account for the influence of topography and land cover on sensor response. Specifically, a local and land cover stratification procedure was developed and tested to compute multiple Minnaert constants. SPOT HRV satellite imagery of the Nanga Parbat Himalaya was used because the topography is extreme, and the study area represents an excellent location for testing anisotropic-reflectance correction procedures. The SPOT-3 NIR image was used in the analysis as the atmospheric influence was minimal. A digital elevation model (DEM) was generated using SPOT-3 panchromatic stereopairs. The DEM was used to produce estimates of slope and slope aspect for accounting for variations in the local direct irradiant flux. Four subimages were selected to evaluate the spectral variance for two homogeneous areas and two heterogeneous areas. Effective anisotropic-reflectance correction should result in a decrease in spectral variation over homogeneous land cover and increased spectral variation among heterogeneous land cover areas. Descriptive statistics and semivariogram analysis was used for evaluation of the original and normalized images. Results indicated that the Cosine-correction method produced high radiance values throughout the image regardless of land cover. This over-correction was found wherever there were steep slopes, and is the result of not accounting for the correct magnitude of the surface irradiance. The Minnaert-correction procedure, using a global Minnaert constant, appeared to produce better results, however, global-regression analysis revealed results that did not accurately characterize the degree of anistropy for various land cover classes. M innaert correction based upon land-cover stratification produced statistically valid results, such that both topography and land cover effects on the radiant flux were generally accounted for. The window-based approach produced images that appeared to reduce the topographic effect, although overcorrection still occurred, and statistical results were invalid for small window sizes. These results indicated that the Minnaert-Correction procedure has the potential to be used to reduce the topographic effect in rugged terrain, if scale-dependent variation in the topography and land-cover characteristics can be locally evaluated to compute Minnaert constants. Furthermore, the diffuse-irradiant and adjacent-terrain irradiant flux need to be considered as overcorrection is problematic in complex terrain. More research on dynamic spatial-partitioning of the topography and spectral variation is therefore warrented.
Cacioppo, Stephen B., "Anisotropic-Reflectance Correction of Multispectral Satellite Imagery in Complex Mountain Terrain" (2002). Student Work. 586.
Files over 3MB may be slow to open. For best results, right-click and select "save as..."