Incorporation of latitude-adjusted bioclimatic variables increases accuracy in species distribution models

Abstract

Latitudinal variations in species niche due to factors such as phenotypic plasticity and local adaptation pose fundamental problems for ecological niche modelling at large geographic scales. Species adaptation to varying climates across latitudes thus necessitates the latitudinal adjustment of climatic variables for incorporation into more accurate models of habitat suitability and species distribution. Previous studies have identified changes in species’ elevational distributions across latitudes, however few studies incorporate latitude-adjusted variables into suitability or distribution models. In this study, we utilized generalized additive models (GAMs) to adjust bioclimatic variables by latitude in order to remove latitudinal variation in ecological niche models. We incorporated these latitude-adjusted variables into maximum entropy (MaxEnt) models of two species of toad, Bufo sachalinensis (previously Bufo gargarizans) and Bufo stejnegeri, in northeast Asia. We additionally identified thresholds in bioclimatic and terrain variables that delineate boundaries between the two Bufo species where their ranges overlap. We found that GAM adjustment of bioclimatic variables effectively removed latitudinal variation in said variables. MaxEnt models incorporating these latitude-adjusted variables had significantly improved model fit (AUC and TSS) for both Bufo species. We identified a potential sympatric area shared by the two species and delineated a geographic boundary thresholding the probable presence of one species over the other. This study shows a clear improvement for distribution models with the inclusion of latitude-adjusted bioclimatic variables for both narrow and wide-ranging species. This has implications for future distribution modelling practices, especially for geographically widespread species, and has potential for modelling species distributions in climate change scenarios. Further, the methodology for delimiting species thresholds has future use in modelling contact zones of related species.