Ecological niche modelling of tree and wood pipits in southern Africa and adjacent countries may help to delimit distributions based on citizen science data

Abstract

Distribution maps are generally based on documented records rather than true occurrence patterns. This may be problematic for cryptic, under-reported species that occur in areas poorly covered by observers. Species distribution models may help overcome this challenge. Here, all available records of the migratory Anthus trivialis (tree pipit) and resident Anthus nyassae (wood pipit) for southern Africa and adjacent areas were assembled to train generalised linear models, random forest and gradient boosting machine species distribution models. Sampling pseudo-absences from a common species’ similarly biased records helped to account for the spatial sampling bias present in the data. The model outputs suggest that A. trivialis and A. nyassae display a latitudinal habitat suitability gradient in the area of interest, opposing a latitudinal reporting gradient. The migratory behaviour of A. trivialis may blur its ecological niche. More and more reliable field observations are needed to confirm these findings. This study provides a clear framework to assist distribution delimitations from citizen science data by counteracting observer and sampling biases.