You are here:
OVERVIEW OF US NATIONAL LAND-COVER MAPPING PROGRAM
Citation:
Wickham, J D. OVERVIEW OF US NATIONAL LAND-COVER MAPPING PROGRAM. Presented at Workshops on EON 2000, Montpelier, France, October 28-30, 2002.
Impact/Purpose:
The objective of this task is to produce land-cover and related products that are needed to meet Annual Performance Goals (APG) under GPRA Goals Clean Air, Clean Water, and Healthy and Safe Communities, and to meet the critical needs of EPA Regional Offices.
Description:
Because of escalating costs amid growing needs for large-scale, satellite-based landscape information, a group of US federal agencies agreed to pool resources and operate as a consortium to acquire the necessary data land-cover mapping of the nation . The consortium was initiated in the mid- I 990s and focused on acquiring Landsat TM and ancillary data for the conterminous United States. Three successful mapping programs were: 1) Coastal Change Analysis Program (C-CAP), whose objective was to map current land cover and back-cast for change assessment in coastal areas of the US; 2) GAP, whose objective was to create detailed maps of vegetation on a state-by-state basis for identification of habitats most in need of protection, and; 3) the National Land Cover Data (NLCD) mapping program, whose objectives were to map general land-cover classes for the conterminous United States and provide an accuracy assessment of that data.
NLCD land-cover mapping for the 1990s, led by the U.S. Geological Survey (USGS), was completed in 2001, and represents the first consistently classified, seamless land-cover map of the conterminous U.S. despite 20 years of Landsat TM availability. The U.S. Environmental Protection Agency has been one of the principal partners in the NLCD mapping project, and has led the accuracy assessment of the land- cover data. Accuracy assessments have been conducted by U.S. federal administrative regions, have
been completed for 9 out of IO of those regions, with the remaining region in progress. Reference data were collected for approximately 16,000 points to conduct the accuracy assessment using an innovative, two-stage cluster sampling design that maintained geographic spread among the sample points, collected 100 samples per class, and minimized cost. A primary finding of the assessment was that overall land- cover accuracy at the Anderson et al. (1976) Level I thematic detail (water, urban, barren, forest, agriculture, wetland) approached Anderson's nominal standard of 85% correct.
The success of the NLCD program has fostered its continuation, and in 2001 the Consortium re- convened to initiate acquisition of Landsat and ancillary data to continue the NLCD, GAP, and C-CAP mapping programs. NLCD has undertaken a new methodological approach to land-cover mapping that includes: 1) "smart" scene selection using seasonal (temporal) NDVI to select Landsat TM data; 2) regression-based land-cover "derivatives" (e.g., impervious surface) for input into the land-cover mapping process and database development; 3) rule-based (e.g., CART) land-cover mapping, and 4) mapping by ecoregion-like mapping zone to improve spectral separation of classes and improve edge- matching. The mapping is again being led by the USGS. As mapping zones are completed, EPA will be conducting change detection accuracy assessments. Pilot studies of change detection accuracy assessment of the Chesapeake Bay region are now in the planning stages and will be undertaken when the ca. 2000 land cover for that region are completed.