A novel survey design for modeling species distribution of beavers in Algonquin Park, Canada

Abstract

AbstractUnderstanding spatial and temporal variation in beaver abundance is a central goal for a wide range of management issues, ranging from species reintroductions to mitigation of environmental and economic impacts. Yet due to high costs associated with surveys, many studies are limited to a single regional estimate, or a complete census of a smaller study area extrapolated to the surrounding landscape. We present a survey design that allows for predicting beaver abundance across the broader landscape through interpolation. In October 2019, we conducted an aerial survey in a 15,000 km2 study area around Algonquin Provincial Park in Ontario, Canada. We counted 145 colonies on 73 plots that averaged 4.5 km2 (+/−3.27 SD). Our regional estimate for beaver abundance of 0.55 (95% CI = +/−0.18) colonies/km2 is comparable to historical surveys conducted in the region in the 1970s. We then predicted beaver abundance in unsampled plots using a Poisson generalized additive model (adjusted R2 = 0.81, deviance explained = 55.9%) that included non‐linear responses to elevation (P < 0.001), shoreline complexity (P = 0.003), and availability of shade‐tolerant hardwoods (P = 0.001). Our species distribution model predicted strong east‐west patterns in beaver abundance across the study region associated with spatial patterns in elevation and forest composition.