Decoding the architecture of drought: SHAP-enhanced insights into the climate forces reshaping the Sahel

Abstract. The Sahel region faces increasing drought variability, driven by complex interactions between climatic indices and hydrological extremes. This study explores the correlation between the Standardized Precipitation Evapotranspiration Index (SPEI) and multiple climatic indices – including the Global Mean Temperature (GMT), Indo-Pacific Warm Pool (IPWP), Atlantic Multidecadal Oscillation (AMO), and North Tropical Atlantic Index (NTA) – using trend analysis, cross-correlation, and an innovative SHAP-driven (SHapley Additive exPlanations) clustering approach. The Seasonal Kendall (SK) test identified statistically significant decreasing SPEI-12 trends in 57.5 % of the grid cells, especially in the western (Senegal, Mauritania) and southeastern Sahel (South Sudan), while 19.3 %, mainly in central-western areas (Burkina Faso, Niger), showed significant increases. Correlation analysis revealed strong negative relationships between SPEI-12 and GMT (up to −0.76) and IPWP (−0.71), underscoring their role in drought intensification. Conversely, AMO (0.40) showed a positive correlation, meaning that during its warm phase rainfall tends to increase, alleviating drought severity, while its cold phase intensifies drought. This reflects a spatially heterogeneous influence distinct from the consistently negative effects of GMT and IPWP. Using the SHAP-driven clustering, AMO and NTA emerged as key discriminators of regional drought regimes. Thus, correlation analysis and RF/SHAP highlight complementary perspectives: parameters such as GMT and IPWP drive overall drought intensification, while parameters such as AMO and NTA govern the regional differentiation of drought patterns. This study introduces a novel framework that integrates explainable Artificial Intelligence (AI) into drought assessment, offering actionable insights for climate adaptation and water resource management in the Sahel.