AI maps crop diversity hotspots to shield global food supply from climate change
Researchers have developed a machine-learning tool that identifies where to protect wild crop relatives most vulnerable to climate shifts, revealing that current conservation strategies miss critical populations. For food security and agricultural companies, this means more precise targeting of breeding programs that could help crops adapt to future environmental stress.
Originaltitel: Future-proofing agrobiodiversity: climate and niche-aware conservation planning using reinforcement learning
Abstract Despite substantial global commitments to expand protected-area networks, the strategic allocation of limited resources remains challenging. Spatial conservation planning helps identify priority regions that maximise conservation benefits per unit area. Yet, they also tend to neglect two fundamental aspects of conservation: climate-driven range shifts and the representation of environmentally distinct populations within species. Here, we propose a continental-scale conservation planning framework that explicitly accounts for both processes through novel routines implemented in the conservation planning software CAPTAIN. We apply this framework to European crop wild relatives (CWR), for which niche coverage is a focal priority, as it underpins their potential to support agricultural adaptation to future environmental stressors through breeding programs. Comparative analyses on a subset of 186 CWR associated with five focal crops show that accounting for range shifts and niche coverage leads to substantially different conservation priorities from those obtained with a baseline model based on current distributions only. These additions reduced the number of non-protected species by 64%, increased the average protected distribution range by 43%, increased mean niche coverage from 75.8% to 84.5% and reduced the number of species with less than half of their niche protected from 35 to 10. Applied to a more comprehensive checklist of 1,140 European CWRs, the final framework identifies continental-scale priority areas representing 93.5% of these taxa and includes 94.4% of its critically endangered species. Our results highlight the importance of incorporating both temporal dynamics and within-species environmental representation when designing conservation strategies under climate change. Repository The repository will be made publicly accessible after publication at doi: https://10.5281/zenodo.19855597