Biased correlated random walk and foray loop: which movement hypothesis drives a butterfly metapopulation?

Animals in fragmented landscapes have a major challenge to move between high-quality habitat patches through lower-quality matrix. Two current mechanistic hypotheses that describe the movement used by animals outside of their preferred patches (e.g., high-quality habitat or home range) are the biased, correlated random walk (BCRW) and the foray loop (FL). There is also a variant of FL with directed movement (FLdm). While these have been most extensively tested on butterflies, they have never been tested simultaneously with data across a whole metapopulation and over multiple generations, two key scales for population dynamics. Using the pattern-oriented approach, we compare support for these competing hypotheses with a spatially explicit individual-based simulation model on an 11-year dataset that follows 12 patches of the federally endangered Fender's blue butterfly (Plebejus icarioides fenderi) in Oregon's Willamette Valley. BCRW and medium-scale FL and FLdm scenarios predicted the annual total metapopulation size for ≥ 9 of 12 patches as well as patch extinctions. The key difference, however, was that the FL scenarios predicted patch colonizations and persistence poorly, failing to adequately capture movement dynamics; BCRW and one FLdm scenario predicted the observed patch colonization and persistence with reasonable probabilities. This one FLdm scenario, however, had larger prediction intervals. BCRW, the biologically simplest and thus most parsimonious movement hypothesis, performed consistently well across all nine different tests, resulting in the highest quality metapopulation predictions for butterfly conservation.