Modeling landscape structure response across a gradient of land cover intensity

Quantitative analyses in landscape ecology have traditionally been dominated by the patch-mosaic concept in which landscapes are modeled as a mosaic of discrete patches. This model is useful for analyzing categorical data but cannot sufficiently account for the spatial heterogeneity present in continuous landscapes. Sub-pixel remote sensing classifications offer a potential data source for capturing continuous spatial heterogeneity but lack discrete land cover classes and therefore cannot be analyzed using standard landscape metric tools. This research introduces the threshold gradient method to allow transformation of continuous sub-pixel classifications into a series of discrete maps based on land cover proportion (i. e., intensity) that can be analyzed using landscape metric tools. Sub-pixel data are reclassified at multiple thresholds along a land cover continuum and landscape metrics are computed for each map. Metrics are plotted in response to intensity and these 'scalograms' are mathematically modeled using curve fitting techniques to allow determination of critical land cover thresholds (e. g., inflection points) where considerable landscape changes are occurring. Results show that critical land cover intensities vary between metrics, and the approach can generate increased ecological information not available with other landscape characterization methods.