A Review of Machine Learning in Species Distribution Modeling under Climate Change in West Africa

Abstract

Species distribution modelling (SDM) has become a central instrument for anticipating how biodiversity is redistributed under climate change, and machine learning now supplies most of the algorithmic capacity behind these predictions. West Africa combines high climatic exposure, dense reliance on wild biological resources, and thin biodiversity records, so the region is a demanding setting in which the strengths and weaknesses of machine learning SDMs are exposed. Scope. Peer-reviewed studies published between 2020 and 2026 were synthesised to describe how correlative SDMs, tree-based ensembles, and deep learning have been applied to West African flora, fauna, disease vectors, and agricultural pests, and to assess the treatment of data quality, model transferability, and reporting standards. Key findings. Maximum entropy modelling remains the dominant approach, yet tree-based ensembles and, more recently, deep neural networks are being adopted where occurrence volume permits. Reported discrimination is high across algorithm families, with area under the curve values frequently above 0.90, though internal validation was shown to overstate reliability when sampling bias and extrapolation into novel climates were left uncorrected. Plant studies concentrated in Benin, Togo, Nigeria, and Burkina Faso projected habitat contraction for a majority of exploited woody species, with northward and upslope displacement of suitable conditions, while vector and pest models projected expansion of climatically suitable areas for several malaria vectors and invasive insects across the region. Persistent occurrence-data scarcity, spatial sampling bias, and inconsistent reporting were identified as the leading constraints on inference. Implications. Bias-aware calibration, spatially independent validation, standardised reporting, and investment in regional occurrence data are needed before machine learning SDMs are used to guide conservation prioritisation, protected-area design, and public-health planning in West Africa. This review positions the regional evidence within methodological developments and defines data-supported priorities for the next phase of work.