You are here:
Estimating Accuracy of Land-Cover Composition From Two-Stage Clustering Sampling
Citation:
Stehman, S. V., J. D. WICKHAM, L. Fattorini, T. G. WADE, F. Baffetta, AND J. H. Smith. Estimating Accuracy of Land-Cover Composition From Two-Stage Clustering Sampling. REMOTE SENSING OF ENVIRONMENT. Elsevier Science Ltd, New York, NY, 113(12):1236-1249, (2010).
Impact/Purpose:
A common use of land-cover maps is to calculate the area or proportion of area of each land-cover class within each unit of a spatial partition of the region mapped. The area classified as forest, agriculture, and developed land within each 10 km by 10 km block forming a partition of the region mapped is an example of land-cover composition data that may be obtained from a land-cover map. Applications of land-cover composition data span a variety of spatial units, including watersheds (e.g., Jones et al. 2001a, Stanfield et al. 2002), subbasins or subwatersheds (Wimberly and Ohmann 2004, Jennings et al. 2004), buffer areas around a sampling station (Comeleo et al. 1996, Driscoll and Donovan 2004), townships (Glennon and Porter 1999), 7.5 km by 7.5 km blocks (Riitters et al. 2004), 400, 800, and 1600 m radii circular plots (Bakker et al. 2002), or 5 km by 5 km blocks (Jones et al. 2001b, Wickham et al. 2002).
Description:
Land-cover maps are often used to compute land-cover composition (i.e., the proportion or percent of area covered by each class), for each unit in a spatial partition of the region mapped. We derive design-based estimators of mean deviation (MD), mean absolute deviation (MAD), root mean square error (RMSE), and correlation (CORR) to quantify accuracy of land-cover composition for a general two-stage cluster sampling design, and for the special case of simple random sampling without replacement (SRSWOR) at each stage. The bias of the estimators for the two-stage SRSWOR design is evaluated via a simulation study. The estimators of RMSE and CORR have small bias except when sample size is small and the land-cover class is rare. The estimator of MAD is biased for both rare and common land-cover classes except when sample size is large. A general recommendation is that rare land-cover classes require large sample sizes to ensure that the accuracy estimators have small bias.