From satellite drought indices to migration flows: tracing climate signals across the Sahel

Linking satellite-derived environmental indicators to human mobility outcomes requires bridging remote sensing, climate science, and migration research. In the Central Sahel, where drought increasingly threatens livelihoods, understanding how climate stress translates into population movement remains complicated by a fundamental scale-of-analysis problem: patterns visible at national levels may obscure, or even reverse, at regional scales. This study traces the climate-migration signal across Burkina Faso, Chad, Mali, and Niger, quantifying drought's contribution to internal migration while examining how spatial heterogeneity shapes the relationship.

This study analyzed 77,783 internal migration flows (2005–2010) from a derived gravity model, linking them to drought severity measured via the Soil Moisture Agricultural Drought Index (SMADI) which is a composite satellite indicator integrating soil moisture, temperature, and vegetation health. A symmetric push-pull framework treated origin and destination conditions identically, addressing methodological critiques of traditional asymmetric gravity models. Machine learning algorithms (Random Forest, XGBoost) captured non-linear relationships, with climate attribution quantified through five complementary methods including SHAP value decomposition.

The results reveal that scale of analysis fundamentally shapes conclusions about climate-migration relationships. In three countries, drought contributed modestly but consistently to migration prediction: Chad (5–9% of model explanatory power), Burkina Faso (6–18%), and Niger (4–38% depending on attribution method). Mali, however, showed negative climate attribution (−23%), i.e., adding drought variables degraded predictive accuracy. This counterintuitive finding traces to within-country heterogeneity: the Mopti region exhibited an inverse drought-migration relationship (r = −0.22), likely reflecting the Inner Niger Delta's flood-pulse ecology where drought improves rather than undermines local livelihoods. Aggregating across regions with opposing signals cancels the climate effect and introduces prediction error.

Despite this heterogeneity, robust patterns emerged across all four countries. Push factors at origin dominated predictions (>99% of importance), while destination pull factors contributed negligibly, suggesting Sahelian migration functions primarily as stress response rather than opportunity-seeking behaviour. Rural-origin corridors showed 2–2.5 times higher climate sensitivity than urban-origin flows. Critically, partial dependence analysis revealed non-linear drought-migration relationships with plateaus at extreme drought severity, consistent with the immobility hypothesis wherein severe stress erodes the resources necessary for movement, potentially trapping vulnerable populations in place.

These findings carry two implications for interdisciplinary climate-mobility research. First, national-level analyses risk masking or misrepresenting climate signals when subnational regions exhibit opposing relationships, regional stratification is not merely preferable but essential for valid inference. Second, the transition from mobility to immobility at extreme drought levels suggests that climate adaptation policy must address both displaced populations and those trapped by insufficient resources to move. Bridging satellite drought monitoring with migration outcomes is methodologically feasible, but the bridge must be built at appropriate spatial scales.