Accuracy assessment and validation of classified satellite imagery of Texas coastal wetlands

Two methods of determining wetlands and landcover classification accuracy for Landsat thematic mapper (TM) imagery are presented and evaluated. A partial Landsat TM scene of the upper Texas coast (Galveston Bay) was classified according to the NOAA CoastWatch-Change Analysis Program landcover classification scheme. Thirteen major wetland, upland and open water landcover types were identified using a series of supervised and unsupervised classification procedures. The classification accuracy was estimated using two accuracy assessment methods: ground truthing 407 sites located with Global Positioning System (GPS) and direct overlay of wetland categories with comparable United States Fish and Wildlife Service (USFWS) National Wetland Inventory (NWI) data. Comparisons of the accuracy assessment estimates found that the ground truthing method consistently produced higher accuracy estimates than that of the direct overlay method. It was concluded that ancillary data as a means of accuracy assessment for classified imagery as an alternative to ground truthing should be used with caution