Drought extremes have cascading impacts on agriculture, ecosystems, water resources, and human health. Forecasting droughts remains challenging without resolving tightly coupled soil–plant–atmosphere processes, including soil moisture (surface and root zone), groundwater deficits, atmospheric aridity, evapotranspiration (ET) partitioning, and vegetation responses across scales. Land surface models (LSMs) provide a mechanistic framework, but structural and parameter uncertainties limit large-scale reliability. Recent advances in artificial intelligence and machine learning (AI/ML), including physics-guided and explainable approaches, offer new opportunities to improve process representation, attribution, and prediction of drought across spatiotemporal scales. However, challenges persist in generalization to novel extremes and robustness of interpretability.

This session invites contributions on mechanistic and/or AI/ML approaches to hydrological extremes, including: (1) flash and compound drought mechanisms and prediction; (2) attribution and process understanding; (3) comparisons of mechanistic vs AI/ML methods focusing on predictability, uncertainty, and interpretability; and (4) impacts on ecosystems, agriculture, and hydrology. Early-career contributions are encouraged.

• B - Biogeosciences
• GC - Global Environmental Change