Optimising Species Distribution Models: Sample size, positional error, and sampling bias matter

Abstract

Species distribution models (SDMs) have proven valuable in filling gaps in our knowledge of species occurrences. However, despite their broad applicability, SDMs exhibit critical shortcomings due to limitations in species occurrence data. These limitations include, in particular, issues related to sample size, positional error, and sampling bias. In addition, it is widely recognized that the quality of SDMs as well as the approaches used to mitigate the impact of the aforementioned data limitations are dependent on species ecology. While numerous studies have experimentally evaluated the effects of these data limitations on SDM performance, a synthesis of their results is lacking. However, without a comprehensive understanding of their individual and combined effects, our ability to predict the influence of these issues on the quality of modelled species-environment associations remains largely uncertain, limiting the value of model outputs. In this paper, we review studies that have evaluated the effects of sample size, positional error, sampling bias, and species ecology on SDMs outputs. We integrate their findings into a step-by-step guide for critical assessment of spatial data intended for use in SDMs.

Rising Temperatures, Falling Leaves: Predicting the Fate of Cyprus’s Endemic Oak under Climate and Land Use Change

“…The Global Biodiversity Information Facility database documenting 92 occurrences [144]; The WOODIV database, which provided occurrence data in 10 × 10 km grids, albeit with significantly less precision [146]. Aware of the limitations imposed by discarding data marked with positional uncertainty-namely a reduced sample size and compromised model precision, which could hinder accurate species distribution assessment-we adopted the framework recommended by [147] and developed by [148], particularly the Nearest Environmental Point method as implemented in the enmSdmX 1.1.2 R package [148], to counteract these limitations. This approach enabled the inclusion of an additional 23 occurrence points from the WOODIV database into our analysis, culminating in a comprehensive dataset encompassing 298 occurrences.…”