Sensitivity Analysis of Boundary Detection on Spatial Features of Heterogeneous Landscape

Relevant management decisions are directly subjected to our ability to characterize landscape pattern: the amount and arrangement of spatial variability. The primary objectives of our project were (1) to develop new exploratory spatial data analysis (ESDA) methods to evaluate and quantify landscape spatial heterogeneity across spatial scales, and (2) to assess the sensitivity of widely used landscape measures. Findings of this research assist decision-making in spatial analysis by providing more coherent boundary detection techniques and more reliable measures of landscape spatial heterogeneity. Specifically, using local statistics to examine the spatial properties of spatial subsets of the landscape, subsets within global data sets, we developed several exploratory spatial data analysis methods to detect boundaries and to quantify spatial pattern. Among them we are presenting here the local boundary cohesiveness index for evaluating boundary strength across scales. By attaching significance values to boundaries, one could perform landscape level analysis using only sharp, or only transitional boundaries, or refine the analysis as a function of their importance as a function of scale (neighborhood). This newly developed method is applied to forest data to assist sustainable forest management. The combination of statistical foundation to attach significance to individual boundaries or even segments of boundaries and the analysis across scale of the persistency of these boundaries, increase the ability of the analysis to identify, characterize and compare different types of boundaries and their possible role within the landscape. Keyword: landscape spatial heterogeneity, spatial autocorrelation, boundaries, randomization test, spatial overlap