A State-Level Comparative Analysis of the GAP and NLCD Land-Cover Data Sets

Abstract

Two nationwide land-cover mapping efforts, the GAP Analysis Program (GAP) and the USGS National Land Cover Data (NLCD) program, address the need for intermediate-scale land-cover information to support a diverse user community. The data sets are comparable, but have different objectives, classification systems, and methodologies. A comparative analysis of the GAP and NLCD data sets is needed throughout the United States to determine their relative strengths and limitations and better inform the user community of their applicability for various applications. This study conducted comparative analyses of the GAP and NLCD data sets for Kansas. The data sets were generalized to a common set of land-cover classes, and pixellevel comparisons were made at the state and ecoregion levels. The GAP and NLCD had an overall classification agreement of 67 percent at the state level, with most of the classification disagreement occurring between cropland and grassland. The cropland area classified by GAP was comparable to cropland area estimates reported by the USDA, while NLCD appeared to underestimate cropland and overestimate grassland. The single-date/multiple-data source classification approach and the sub-optimal early-spring dates of Landsat TM data used to produce NLCD resulted in substantial confusion in croplandgrassland discrimination. The multiple-date classification approach used by Kansas GAP provided better discrimination of most land-cover classes. Some classification disagreement, however, was attributable to methodological differences between GAP and NLCD. Accuracy assessment found an overall accuracy of 87 percent for GAP and 81 percent for NLCD, and GAP had higher accuracies for most individual land-cover classes. The Kansas GAP and NLCD land-cover products were found to be comparable in terms of characterizing broad scale land-cover patterns, but the Kansas GAP land-cover map appears to be more appropriate for localized applications that require detailed and accurate land-cover information.