You are here:
Designing a Multi-Objective Multi-Support Accuracy Assessment of the 2001 National Land Cover Data (NLCD 2001) of the Conterminous United States
Citation:
STEHMAN, S., J. D. WICKHAM, T. G. WADE, AND J. SMITH. Designing a Multi-Objective Multi-Support Accuracy Assessment of the 2001 National Land Cover Data (NLCD 2001) of the Conterminous United States. PHOTOGRAMMETRIC ENGINEERING AND REMOTE SENSING. American Society for Photogrammetry and Remote Sensing, Bethesda, MD, 74(12):1561-1571, (2008).
Impact/Purpose:
The Multi-Resolution Land Characteristics Consortium (MRLC) has completed the 2001 edition of the National Land Cover Data, NLCD 2001, for the conterminous United States (Homer et al. 2007), and is finishing NLCD 2001 for Alaska, Hawaii, and Puerto Rico over the coming year (www.mrlc.gov). NLCD 2001 is considerably different from NLCD 1992 (Vogelmann et al. 2001). Whereas NLCD 1992 was a land-cover dataset, NLCD 2001 is a land-cover database consisting of three main products mapped for every 30-meter cell in the U.S. (Homer et al. 2004). The products identify one of 16 classes of land-cover (Table 1) convolved to a one-acre minimum mapping unit, percent urban imperviousness, and percent tree canopy. The impervious surface and canopy cover datasets provide perpixel estimates from 0 to 100 percent. In this article, we describe the framework and development of the sampling design for assessing accuracy of NLCD 2001.
Description:
The database design and diverse application of NLCD 2001 pose significant challenges for accuracy assessment because numerous objectives are of interest, including accuracy of land cover, percent urban imperviousness, percent tree canopy, land-cover composition, and net change. A multisupport approach is needed because these objectives require spatial units of different sizes for reference data collection and analysis. Determining a sampling design that meets the full suite of desirable objectives for the NLCD 2001 accuracy assessment requires reconciling potentially conflicting design features that arise from targeting the different objectives. Multi-stage cluster sampling provides the general structure to achieve a multi-support assessment, and the flexibility to target different objectives at different stages of the design. We describe the implementation of two-stage cluster sampling for the initial phase of the NLCD 2001 assessment, and identify gaps in existing knowledge where research is needed to allow full implementation of a multi-objective, multi-support assessment.