Location: Hydrology and Remote Sensing Laboratory
Title: Improved forest change detection with terrain illumination corrected landsat images Authors
|Tan, Bin -|
|Masek, Jeffrey -|
|Wolfe, Robert -|
|Huang, Chengquan -|
|Vermote, Eric -|
|Sexton, Joseph -|
|Ederer, Greg -|
Submitted to: Remote Sensing of Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: May 17, 2013
Publication Date: June 22, 2013
Repository URL: http://handle.nal.usda.gov/10113/59908
Citation: Tan, B., Masek, J., Wolfe, R., Gao, F.N., Huang, C., Vermote, E., Sexton, J., Ederer, G. 2013. Improved forest change detection with terrain illumination corrected landsat images. Remote Sensing of Environment. 136:469-483. Interpretive Summary: Land cover and land use change affect regional and global carbon and water cycles. Land cover changes detected from remote sensing images are affected by varying illumination conditions due to the effect of topography. Illumination correction is an important step in pre-processing high-resolution remote sensing data for forest change detection studies. This paper presents an operational illumination correction algorithm to minimize the variability of observed reflectance for similar surface types due to topography and Bidirectional Reflectance Distribution Function (BRDF) effects. Results show that the illumination correction process greatly improves forest change detection accuracy. The overestimation of forest loss and regrowth produced from original Landsat data has been reduced. The accurate forest change information is critical for the modeling carbon and water cycles, and for the forest management required by USDA Forest Service.
Technical Abstract: An illumination correction algorithm has been developed to improve the accuracy of forest change detection from Landsat reflectance data. This algorithm is based on an empirical rotation model and was tested on the Landsat imagery pair over Cherokee National Forest, Tennessee, Uinta-Wasatch-Cache National Forest, Utah, San Juan National Forest, Colorado and Sinkyone Wildness State Park, California. The analyses showed that the illumination correction process successfully eliminated the correlation between Landsat reflectance and illumination condition. A set of forest change maps was retrieved from the corrected Landsat imagery pairs and compared with the change maps from the original Landsat imagery pairs. This comparison shows significant disagreement between two sets of change detection results. Both sets of results were validated through visually comparing with high-resolution (1 m or less) time series images. The validation results demonstrate that the illumination correction process greatly improves the forest change detection accuracy. In particular, a decreased overestimation of forest loss and regrowth was seen and major forest changes were successfully captured. It was found that the disagreement rate between change maps from the original and corrected Landsat images increases with increasing inclination angle. The relationship between illumination condition and the disagreement rate is a V shape curve. The lowest disagreement rate occurs when illumination condition is slightly smaller than horizontal field’s illumination condition value. The shape of the curve is site-specific. The correction for topographic illumination should be considered as a standard pre-processing step for land cover classification and land use change detection, especially for mountainous areas.