Applying Spatial Autocorrelation Analysis to Evaluate Error in New England Forest-Cover-Type Maps Derived from Landsat Thematic Mapper Data

Abstract

A spatial autocorrelation of error analysis was performed to compare the patterns of error in a classified Landsat Thematic Mapper (TM] forest-cover-type image for two forested areasone public and one private. TM data were classified to generate a detailed forest type map, and intensive ground reference data covering approximately 3600 ha were collected for both study areas. lbo difference images were produced by comparing the reference inventory with the classified data, pixel by pixel. The subsequent spatial au tocorrelation analysis indicated that concentrated blocks of error were more pronounced in the public lands study area than in the private lands study area, where error was more evenly distributed. The results indicated that systematic sampling is not always suitable for assessing error in TM data. lntroductlon