COMPARISON OF LANDSAT THEMATIC MAPPER AND HIGH-RESOLUTION PHOTOGRAPHYTO IDENTIFY CHANGE IN COMPLEX COASTAL WETLANDS

Landsat Thematic Mapper (TM) images were used to generate pre- and pos t-hurricane classifications of a complex wetland environment in southe rn Louisiana. Accuracies were estimated as 77% and 81.5% for the pre- and post-classifications that included water, emergent vegetation, flo ating vegetation, and mud flats. From the two classifications, areas o f emergent vegetation loss were identified. The classifications and ch ange map were compared to similar output generated from high resolutio n color infrared photography. The comparison showed spatial scale of t he sensor was the most important factor in separation of classes in th is type of wetland environment. Classifications derived by using the T M images provided good class separation when one class dominated more extensive areas (>30 m), but not when mixtures of wetland types were o n the same order as the TM sensor spatial resolution Boundary pixel mi xtures were problematic, however problems also occurred in areas of fa irly continuous canopies containing small pockets of water and floatin g vegetation, and in areas of degrading marsh. Both areas were predomi nately misclassified as emergent vegetation. In the case of change det ection, loss of emergent vegetation occurring as small pockets was not identified, whereas loss of degraded marsh was identified but the spa tial continuity and extent overemphasized. In combination, these miscl assifications resulted in the TM change analysis overpredicting emerge nt vegetation loss by about 40%.