When big data misleads: limitations of large-scale online biodiversity databases for species distribution modelling

Abstract

Abstract Large-scale online biodiversity databases, particularly the Global Biodiversity Information Facility (GBIF), have become indispensable resources for macroecology, biogeography, and conservation biology. Their accessibility and extensive coverage have promoted widespread application in species distribution modelling, often without adequate assessment of data quality or ecological reliability. We evaluated the limitations of opportunistically collected occurrence records for ecological niche modelling using the West and Central African viper Bitis nasicornis as a case study. We compared MaxEnt models generated from GBIF records with models based on our own field observations collected between 1996 and 2025 across Liberia, Côte d’Ivoire, Ghana, Togo, and Nigeria, and with models integrating both datasets. Models built exclusively from GBIF data produced ecologically unrealistic predictions, identifying suitable habitats across rainforest, moist savanna, and dry savanna ecosystems despite the species’ well-documented restriction to lowland tropical rainforests, typically near permanent water bodies. By contrast, models based on field-verified records closely matched the species’ known ecological requirements and the current IUCN Red List distribution. Predicted suitable areas overlapped by only 8.6% between GBIF- and field-based models, revealing striking disagreement. Our analyses further suggest that this problem extends beyond B. nasicornis, with a substantial proportion of erroneous West African forest records originating from Benin, potentially affecting numerous rainforest taxa. We discuss how taxonomic uncertainty, spatial sampling bias, temporal mismatch, and the loss of ecological context compromise the reliability of large online datasets, and argue that robust field validation remains essential for ecological niche modelling and conservation decision-making.