Advancing texture metrics to model landscape heterogeneity

Abstract

We advance methods to study landscape heterogeneity and assess the suitability of these methods to detect boundary regions between land cover types. We employ 2nd-order texture metrics on continuous and on discrete vegetation data in North Eurasia and on a smaller region therein. Furthermore, we advance 2nd-order texture metrics to incorporate spatial scales in novel ways. The metrics entropy, contrast, and homogeneity are found to detect boundary regions well on discrete data in the smaller study area, but create fuzzy results on continuous data and when models are run for the whole of northern Eurasia. The scales we consider affect the size of the boundary regions, but also the metric values. The results elucidate the need for aligned scale determination and data restriction procedures. Our approach offers new opportunities to model landscape heterogeneity across scales.