Comparison of NLCD with NWI Classifications of Baldwin and Mobile Counties, Alabama

Abstract

An assessment of the accuracy of National Land Cover Data 2001 as compared to National Wetlands Inventory mapping of Mobile and Baldwin Counties conducted by the U.S. Geological Survey's (USGS) National Wetlands Research Center (NWRC). Both classifications were checked against stratified randomly selected ground-based sites and with each other to compare the accuracy of the NLCD with NWI classification methods. For this accuracy comparison, numerous visits were made by photointerpreters to the Mobile Bay area to establish signatures for the modified NWI mapping. After all of the quadrangles in Baldwin and Mobile Counties were mapped, an accuracy assessment of those maps was conducted by field-checking the mapped classes with ground-based check sites. These same sites were used to check the accuracy of the NLCD and also as the basis for cross-walking the NLCD and NWI classification systems for direct comparison of the two methods. This accuracy assessment is dependent on a methodology and analysis developed for the unique characteristics and properties of NWI classification, which is mapped to polygon, as compared to NLCD, which uses raster classification and mapping. This cartographic presentation results in fundamentally different classification boundaries. Overlaying these map types cannot prevent boundary differences resulting in a tremendous proliferation of sliver polygons. The intersection of both maps resulted in three-quarters of a million polygons, nearly half of which were less than the 900-m2 pixel size used in NLCD mapping (the smaller of the minimum mapping units of NWI and NLCD versus NWI mapping was about 0.5 ha). It is the authors' opinion that this cartographic difference needs to be more fully explored before accepting the accuracy difference between NWI and NLCD presented here at face value.