Testing the relevance of binary, mosaic and continuous landscape conceptualisations to reptiles in regenerating dryland landscapes

Abstract

Fauna distributions are assessed using discrete (binary and mosaic) or continuous conceptualisations of the landscape. The value of the information derived from these analyses depends on the relevance of the landscape representation (or model) used to the landscape and fauna of interest. Discrete representations dominate analyses of landscape context in disturbed and regenerating landscapes; however within-patch variation suggests that continuous representations may help explain the distribution of fauna in such landscapes. We tested the relevance of binary, mosaic, and continuous conceptualisations of landscape context to reptiles in regenerating dryland landscapes. For each of thirteen reptile groups, we compared the fit of models consisting of one landscape composition and one landscape heterogeneity variable for each of six landscape representations (2× binary, 2× mosaic, and 2× continuous), at three buffer distances. We used Akaike weights to assess the relative support for each model. Maps were created from Landsat satellite images. Reptiles varied in their response to landscape context; however, the binary landscape representation with classes ‘intact/disturbed’ was best supported overall. Species richness was best described by a binary landscape representation with classes ‘wooded/clear’, whereas reptile abundance was best described by a mosaic landscape representation with classes defined by vegetation type. Five out of ten reptile species responded strongly to a single landscape representation, with the most relevant representation and conceptualisation varying among species. Our findings support the use of multiple landscape conceptualisations and representations during analyses of landscape context for fauna in regenerating landscapes.