You are here:
Estimating Landscape Pattern Metrics from a Sample of Land Cover
Citation:
Hassett, E. M., S. V. Stehman, AND J. D. WICKHAM. Estimating Landscape Pattern Metrics from a Sample of Land Cover. LANDSCAPE ECOLOGY. Springer, New York, NY, 27:133-149, (2012).
Impact/Purpose:
Development of measurements from maps (i.e., landscape metrics) is a defining characteristic of landscape ecological research (Wu and Hobbs 2002; Turner 2005). Landscape metrics were first introduced as a new area of ecological research (Krummel et al 1987; O’Neill et al 1988) to further explore linkages between spatial pattern and process. Following their introduction, software was developed to streamline calculation of landscape metrics (McGarigal and Marks 1995), which in turn helped to further advance their development and use. Landscape metric development has included mathematical refinement (Li and Reynolds 1993; Riitters et al 1996), analysis of statistical correlations (Riitters et al 1995; Cain et al 1997), and sensitivity to pixel size (Wickham and Riitters 1995), map extent (O’Neill et al 1996), map format (raster versus vector) (Wickham et al 1996), map classification accuracy (Hess 1994; Hess and Bay 1997; Wickham et al 1997; Langford et al 2006; Shao and Wu 2008), map thematic resolution (Buyantuyev and Wu 2007), fuzzy representation of map classes (Arnot et al 2004), and gradients of land-cover composition (Neel et al 2004).
Description:
Although landscape pattern metrics can be computed directly from wall-to-wall land-cover maps, statistical sampling offers a practical alternative when complete coverage land-cover information is unavailable. Partitioning a region into spatial units (“blocks”) to create a sampling frame introduces artificial boundaries and patch truncation effects that potentially bias sample-based estimators of landscape pattern metrics. The bias and variance of estimators of nine landscape pattern metrics were computed for each of four 120-km x 120-km regions at two dates using the 1992 and 2001 National Land-Cover Data (NLCD) of the United States. Sample-based estimators of change in these metrics between NLCD 1992 and NLCD 2001 were also evaluated. The biases of the sample-based estimators were generally small for the nine pattern metrics for both the estimators of status and estimators of change. However, biases for estimating certain landscape pattern metrics in some regions were large enough to merit concern. The smaller sample block size (10-km x 10-km) we investigated generally yielded larger biases but smaller variances of the estimators relative to a 20-km x 20-km sample unit. Stratified random sampling improved precision of the estimators relative to simple random sampling. The methodology we developed to investigate properties of sample-based estimators can be readily extended to evaluate other landscape pattern metrics and sample block sizes.