A data-driven approach to sub- seasonal drought forecasting in Bihar, India

Sub-seasonal forecasting, which provides climate outlooks spanning from two weeks to two months, has emerged as a critical tool for enhancing agricultural resilience and planning in monsoon-dependent regions, notably India. This study seeks (1) to develop a point-based forecasting approach to predict dry weeks at selected sites within Bihar, utilizing ECMWF variables and Indian Meteorological Department (IMD) grilled product data to capture temporal variability and site-specific characteristics of dry spells, (2) To predict regional dry conditions across Bihar using a spatial model that incorporates the Standardized Precipitation Evapotranspiration Index (e.g., SPEI-2 or SPEI-3) for a comprehensive, spatially-informed assessment of drought risk, and (3) to evaluate the performance of forecasting models across various lead times using classification metrics, such as Area Under the Curve (AUC) and Kolmogorov- Smirnov (KS) statistics, and others, to accurately assess their effectiveness in distinguishing drought conditions. This research aims to improve advisory systems, support proactive nursery and transplanting interventions, and ultimately enhance resilience in rice farming practices under increasingly variable climatic conditions.