‘Scalescape’: An R package for estimating distance-weighted landscape effects on an environmental response

Erin Lowe, Nelson Institute For Environmental Studies, University of Wisconsin - Madison, Madison, WI, Benjamin Iuliano, Integrative Biology, University of Wisconsin - Madison, Claudio Gratton, Department of Entomology, University of Wisconsin - Madison, Madison, WI and Tony Ives, Department of Integrative Biology, University of Wisconsin, Madison, Madison, WI

Background/Question/Methods
For any study on the effects of landscape context, a central challenge is determining the relevant landscape extent for a given biotic or abiotic response. This involves both determining the distance from discrete study sites at which elements of the landscape (e.g., the amount of natural area or elevation) influence the response of interest and determining how much landscape elements contribute up to that distance (their “weight”). We give a brief overview of existing methods for defining this spatial extent of a landscape effect, outline issues with these methods, and introduce a new R package called ‘scalescape’ that provides a user-friendly tool to estimate the extent and weight of landscape elements while solving many of the problems with existing methods. Conceptually, ‘scalescape’ follows the approach of other ‘spatial smoothing methods’ which weight landscape elements based on their distance from a focal site. Given a map of predictor variables it uses maximum likelihood to estimate both the spatial range over which the predictors affect the response and the magnitude of these responses.
Results/Conclusions
‘Scalescape’ builds on previous methods in several ways. First, it integrates well-used functions for performing regressions in R with landscape weightings of spatial predictor variables. This provides the first user-friendly interface to implement spatial smoothing with commonly used regression packages. It can also accommodate a wide range of model types and structures including gls() models which account for spatial autocorrelation in the residuals. Finally, ‘scalescape’ performs hypothesis tests using bootstrapping which gives correct type I errors and better statistical power than previous methods.