Correlations between spatial sampling biases and environmental niches affect species distribution models

Aim Spatial sampling biases in biodiversity data arise because of complex interactions between geography, species characteristics and human behaviour, including preferences for or against particular habitats; are therefore not necessarily independent the environmental niches species. We evaluate when correlations spatial likely to affect distribution models (SDMs) developed both with without attempts correct these biases. Innovation A virtual ecologist framework was used simulate either no bias that were correlated (positively negatively) one variables define The niche simulated autocorrelation operating at multiple scales. Virtual samples then model distributions, evaluated based on their ability rank suitability sites correctly. Main conclusions Correlations frequently reduced correlation predictions, but relative importance effects varied type (greater decline as broadens) (models built using detection/non-detection less affected than those detection-only data). Bias-correction effectiveness depending structure also highly variable across methods dependent type. implications results is a greater concern SDMs where: (1) effort non-random respect an gradient thought be species’ distribution; (2) being modelled has broad niche; (3) modelling contain only information detections (i.e., presence only).