Combining geostatistical and biotic interaction model to predict amphibian refuges under crayfish invasion across dendritic stream networks

Aim Biological invasions are pervasive in freshwater ecosystems, often causing native species to contract into areas that remain largely free from invasive species impacts. Predicting the location of such ecological refuges is challenging, because they are shaped by the habitat requirements of native and invasive species, their biotic interactions, and the spatial and temporal invasion patterns. Here, we investigated the spatial distribution and environmental drivers of refuges from invasion in river systems, by considering biotic interactions in geostatistical models accounting for stream network topology. We focused on Mediterranean amphibians negatively impacted by the invasive crayfishes Procambarus clarkii and Pacifastacus leniusculus . Location River Sabor, NE Portugal. Methods We surveyed amphibians at 168 200‐m stream stretches in 2015. Geostatistical models were used to relate the probabilities of occurrence of each species to environmental and biotic variables, while controlling for linear (Euclidean) and hydrologic spatial dependencies. Biotic interactions were specified using crayfish probabilities of occurrence extracted from previously developed geostatistical models. Models were used to map the distribution of potential refuges for the most common amphibian species, under current conditions and future scenarios of crayfish expansion. Results Geostatistical models were produced for eight out of 10 species detected, of which five species were associated with lower stream orders and only one species with higher stream orders. Six species showed negative responses to one or both crayfish species, even after accounting for environmental effects and spatial dependencies. Most amphibian species were found to retain large expanses of potential habitat in stream headwaters, but current refuges will likely contract under plausible scenarios of crayfish expansion. Main conclusions Incorporating biotic interactions in geostatistical modelling provides a practical and relatively simple approach to predict present and future distributions of refuges from biological invasion in stream networks. Using this approach, our study shows that stream headwaters are key amphibian refuges under invasion by alien crayfish.