The Influence of Geographic Sampling Methods on Vegetation Map Accuracy Evaluation in a Swampy Environment

摘要: The effectiveness of five sampling methods was evaluated empirically to determine the thematic accuracy six-class vegetation maps two study areas in Okefenokee Swamp produced by computer-assisted supervised classification Landsat TM data. An exhaustive survey permitted estimates overall, user's, and producer's accuracies be made each method. spatial pattern errors also examined with aid autocorrelation analysis. By comparing Kappa coefficients agreement among error matrices generated different applying Chi-square tests, stratified random method found perform best for evaluation a swampy environment where exhibited mixture simple complex patterns. higher resolution image data employed produce map tended cause stronger positive resultant pattern.