Relative performance of ecological niche and occupancy models for predicting invasions by patchily-distributed species

Abstract

Ecological niche models (ENM) have been used with mixed success for predicting the geographic extent of non-native species to aid management and conservation. This approach is problematic for predicting invasions of patchily-distributed species (e.g., pond-breeding amphibians), whose occurrence is often determined by local habitat conditions. Here, we tested the performance of bioclimatic ENM for predicting occurrence (from repeated surveys) of two non-native pond-breeding anurans at 71 wetlands in British Columbia, Canada: permanent pond specialist American bullfrog (Lithobates catesbeianus), and generalist green frog (Lithobates clamitans). For L. catesbeianus, we assessed the risk of invasion beyond the invasion front. We found higher correlation between ENM and occupancy predictions for L. clamitans (r s = 0.58), than for L. catesbeianus (r s = −0.26). L. clamitans occurrence was highest at low elevations and high annual precipitation; in contrast, L. catesbeianus occupancy was predicted by wetland connectivity and distance from a historic introduction site [low at isolated ponds >50 km from the introduction site, and high (>0.8) at all ponds with >10 % water within 500 m]. Conditional on successful dispersal, four sites beyond the L. catesbeianus invasion front surveyed in this study were at high risk of invasion due to high habitat suitability (proportion of area occupied = 0.33; 0.04–0.83, 95 % CI). In conclusion, ENMs may be useful for informing invasion management for climate driven wetland species, but repeated sampling is necessary to predict invasions for habitat-driven wetland species.