AN EVALUATION OF THE COASTWATCH CHANGE DETECTION PROTOCOL IN SOUTH-CAROLINA

The NOAA sponsored CoastWatch Change Analysis Project (C-CAP) will uti lize remote sensing technology to monitor changes in coastal wetland h abitats and adjacent uplands on a cycle of 1 to 5 years. Two study are as in South Carolina were selected to test various C-CAP change detect ion protocols using near-anniversary Landsat Thematic Mapper data obta ined in 1982 and 1988. Fort Moultrie (dominated by salt and brackish m arsh) and Kittredge (40 river miles inland and dominated by bottomland hardwoods and riverine aquatic beds) study areas were used to evaluat e a modified C-CAP classification scheme, image classification procedu res, change detection algorithm alternatives, and the impact of tidal stage on coastal change detection. The modified CoastWatch Classificat ion Scheme worked well and can be adapted for South Carolina with mino r adjustments. Unsupervised ''cluster-busting'' techniques coupled wit h ''threshold 3 majority filtering'' yielded the most accurate individ ual date classification maps (86.7 to 92.3 percent overall accuracy; K appa coefficients of 0.85 to 0.90). The best change detection accuracy was obtained when individual classification maps were majority filter ed and subjected to ''post-classification comparison'' change detectio n (85.2 percent overall accuracy, kappa coefficient of 0.82). Suggesti ons are made concerning appropriate change detection matrix logic and the format of change detection legends. The multiple date images selec ted for coastal change detection should meet stringent tidal stage gui delines which have yet to be fully documented.