Sensitivity of Selected Landscape Pattern Metrics to Land-Cover Misclassification and Differences in Land-Cover Composition

investigation of the sensitivity of these measurements to clasCalcuiation of landscape metrics from land-cover data is besification error (Hess, 1994). he objectives of this paper are coming increasingly common. Some studies have shown that two-fold: (1) to determine the sensitivity of landscape these measurements are sensitive to differences in land-cover metrics to land-cover misclassification, and (2) to determine composition, but none are known to have tested also their the sensitivity of landscape metrics to differences in landto land-cover misc~ass~f icat~on~ error sirnulascape condition. These objectives are necessarily connected. tion model was written to test the sensitivity of selected landIdeally, landscape pattern metrics would be insensitive to scape pattern metrics to misclassification, and regression misclassification but sensitive to differences in land cover. analysis was used to determine i f these metrics were signifiLand-cover data, mapped from Landsat TM for the Chescantly related to differences in land-cover composition. Cornapeake Bay Regional Watershed, were used for this study. parison of sensitivity and regression results suggests that The data were divided into 57 eight-digit U.S. Geological differences in land-cover composition need to be about 5 Survey (us~s) hydrologic units or watersheds. Sensitivity to percent greater than the misclassification rate to be confident misclassification is tested using a simulation model based on that differences in landscape metrjcs are not due to mjsclasa published land-cover error matrix (Green et a]., 1993). Sensification. sitivity to differences in landscape condition is tested by comparison to the amount of human land use in the watershed. Landscape condition is measured as the ratio of anIntroduction thropogenic land use to total area (U) (O'Neill et al., 1988). Analysis of landscape Pattern makes use of measurements of Low values reflect that a watershed is primarily forested, the connectedness (e.g., contagion, percolation), diversity while high values reflect dominance by human land uses. (e.g., ~hannon diversity, dominance), shape complexity (e.g., use of u as a measure of landscape condition is based fractal dimension), and size of land-cover patches to study on observations in ecology and biogeography that the Chessecological condition at local to regional scales (Turner and peake B~~ ~ ~ ~ i ~ ~ ~ l watershed was almost entirely forested Gardner, 1991). These metrics (O'Neill et al., 1988) have prior to conversion to human land use (Whittaker, 1975). been used to assess landscape condition (Krummel et al., 1987; Graham et al., 1991; Wickham and Norton, 1994), infer Landscape Pattern Metrics ecological process from pattern (Turner, 1989; Milne, 19921, Three landscape pattern metrics were chosen for analysis: and show how landscape configuration can impose conaverage patch compaction (APC), contagion (C), and fractal straints on biological populations (Browder et al., 1989; Hoodimension (F). APC, C, and F were selected because they ver and Parker, 1991; Flather et al., 1992; Pearson, 1993). were found to represent orthogonal axes among 55 landscape a regi0na1 perspective, land-cover Patterns may be conpattern metrics tested in a factor analysis (Riitters et al., sidered as either forcing or constraint functions for sub-re1995). Therefore, these metrics represent independent inforgional dynamics, or as integral parts of strictly regional mation about landscape pattern. The formulas for calculating models (Allen and Starr, 1982; O'Neill et al., 1994). Informathese metrics are listed in ~ ~ ~ ~ ~ d i ~ A. tion about land-cover patterns has proven useful for both loAverage patch compaction, APC, is the ratio of patch cal and regional assessments of ecological condition (Vos area to the size (area) of the smallest square that will contain and Opdam, 1993; Meyer and Turner, 1994). that patch. The ratios are averaged over all patches in a landMeasurements landscape Pattern are commonly made scape. APC has a value of 0 for linear patches and a value of from land cover (e.g., Krummel et al., 1987; Turner, 1987; 1 for perfectly square patches (Riitters et al., 1995). Turner, 1990a; Turner, 1990b; O'Neill et al., 1988; Graham et Contagion, C, measures the degree to which the landal., 1991; Olsen et al., 1993; Wickham and Norton, 1994; scape is composed of a few large or several small patches. wickham and Riitters, 1995; Wickham et a]., in press; RiitContagion ranges between 0 and 1. High values of contagion ters et al., 1995). However, measurement of landscape patindicate that the landscape is clumped into a few, large tern from land-cover maps has been undertaken without patches. Fractal dimension, F, is commonly calculated as twice J.D. Wickham and K.H. Riitters are with the Tennessee Valley Authority, Norris, TN 37828. R.V. O'Neill is with the Oak Ridge National Laboratory, Oak Photogrammetric Engineering & Remote Sensing, Ridge, TN 37831. Vol. 63, No. 4, April 1997, pp. 397-402. T.G. Wade is with the Desert Research Institute, Reno, NV 89506. 0099-1112/97/6304-397$3.00/0 K.B. Jones is with the Environmental Protection Agency, Las O 1997 American Society for Photogrammetry Vegas, NV 89193. and Remote Sensing