Estimating error in an analysis of forest fragmentation change using NorthAmerican Landscape Characterization (NALC) data

We describe an approach for estimating measurement error in an analysis of forest fragmentation dynamics. We classified North American Landscape Chara cterization (NALC) images in four path-row: locations in the Upper Midwest to characterize changing patterns of forest cover. To estimate error, we ca lculated the differences in values of forest fragmentation metrics for over lapping scene pairs from the same time frame (or epoch). The overlapping im age areas were subdivided into landscape partitions. We tested the effects of amount of forest co:er, landscape phenology, atmospheric variability (e. g., haze and clouds), and alternative processing approaches on the consiste ncy of metric values calculated for the same place and approximate time but from different images. Two of the metrics tested (average patch size and n umber of patches) were more sensitive to image characteristics and containe d more measurement error in a change detection analysis than the others (pe rcent forest cover and edge density). Increasing the landscape partition si ze moderately reduced the amount of an-or in landscape change analysis, but at the cost of reduced spatial resolution. Processes used to generalize th e forest map, such as small-polygon sieving and majority filtering, were no t found to consistently decrease measurement error in metric values and in some cases increased error. Predictive models of error in a forest fragment ation change analysis were developed and significantly explained up to 50% of the variation in error. We demonstrate how, in a change analysis, predic ted error can be used to identify locations that exhibit change substantial ly greater than the error in value estimation. (C) Elsevier Science Inc., 2 000.