You are here:
ASSESSING THE ACCURACY OF NATIONAL LAND COVER DATASET AREA ESTIMATES AT MULTIPLE SPATIAL EXTENTS
Citation:
Hollister, J W., M. L. Gonzalez, J F. Paul, P. V. August, AND J L. Copeland. ASSESSING THE ACCURACY OF NATIONAL LAND COVER DATASET AREA ESTIMATES AT MULTIPLE SPATIAL EXTENTS. PHOTOGRAMMETRIC ENGINEERING AND REMOTE SENSING 70(4):405-414, (2004).
Description:
Site specific accuracy assessments provide fine-scale evaluation of the thematic accuracy of land use/land cover (LULC) datasets; however, they provide little insight into LULC accuracy across varying spatial extents. Additionally, LULC data are typically used to describe landscape composition and structure in watershed and regional ecological studies. The accuracy of the statistics defining landscape structure can not be assessed solely from site specific measures of thematic accuracy. To address these shortcomings, we used photo-interpreted LULC data from Rhode Island and Massachusetts to determine the accuracy of land cover area and edge (patch perimeter) measurements from the National Land Cover Dataset (NLCD) across multiple spatial extents ranging from 2 km2 to 210 km2. The area of forest, water, developed and agricultural lands were positively correlated (p < 0.0005) with the reference data at all extents. Area measurements for these same classes were accurately depicted (rs 0.92) at most extents. The amount of edge of water, developed, and agricultural lands was also positively correlated (p < 0.0005) with the reference data at all extents. Forest edge was positively correlated (p < 0.001)with reference data at all extents except 5.3 km2. Accurate estimates of the total amount of edge (rs 0.92) were found at most extents for only the developed class. The amount of area and edge of barren lands, wetlands, orchards, and rangeland were poorly depicted at all extents.