Evaluation of the Impacts of Hurricane Hugo on the Land Cover of Francis Marion National Forest, South Carolina Using Remote Sensing

• Main Author: Kulkarni, Amit
• Other Authors: DeWitt Braud
• Format: Others
• Language: en
• Published: LSU 2004
• Subjects: Geography & Anthropology
• Online Access: http://etd.lsu.edu/docs/available/etd-08302004-153703/

Hurricane Hugo struck the South Carolina coast on the night of September 21, 1989 at Sullivans Island, where it was considered a Category 4 on the Saffir-Simpson scale when the hurricane made landfall (Hook et al. 1991). It is probably amongst the most studied and documented hurricanes in the United States (USDA Southern Research Station Publication 1996). There has been a Landsat TM based Hugo damage assessment study conducted by Cablk et al. (1994) in the Hobcaw barony forest. This study attempted to assess for a different and smaller study area near the Wambaw and Coffee creek swamp. The main objective of this study was to compare the results of the traditional post-classification method and the triangular prism fractal method (TPSA hereafter, a spatial method) for change detection using Landsat TM data for the Francis Marion National Forest (FMNF hereafter) before and after Hurricane Hugos landfall (in 1987 and 1989). Additional methods considered for comparison were the principal component analysis (PCA hereafter), and tasseled cap transform (TCT hereafter). <br><br> Classification accuracy was estimated at 81.44% and 85.71% for the hurricane images with 4 classes: water, woody wetland, forest and a combined cultivated row crops/transitional barren class. Post-classification was successful in identifying the Wambaw swamp, Coffee creek swamp, and the Little Wambaw wilderness as having a gain in homogeneity. It was the only method along with the local fractal method, which gave the percentage of changed land cover areas. Visual comparison of the PCA and TCT images show the dominant land cover changes in the study area with the TCT in general better able to identify the features in all their transformed three bands. The post-classification method, PCA, and the TCT brightness and greenness bands did not report increase in heterogeneity, but were successful in reporting gain in homogeneity. The local fractal TPSA method of a 17x17 moving window with five arithmetic steps was found to have the best visual representation of the textural patterns in the study area. The local fractal TPSA method was successful in identifying land cover areas as having the largest heterogeneity increase (a positive change in fractal dimension difference values) and largest homogeneity increase (a negative change in fractal dimension difference values). The woody wetland class was found to have the biggest increase in homogeneity and the forest class as having the biggest increase in heterogeneity, in addition to identifying the three swamp areas as having an overall increased homogeneity.