Climate Variability and Rice Production: Extreme Temperature Index Modeling: Seasonal Yield Response Assessment

It creates a novel dataset that explicitly separates weather conditions (heatwaves, rainfall, humidity, sunshine) in the flowering and grain filling phases simultaneously for all three rice seasons (Aus, Aman, Boro) in Bangladesh, filling a critical gap in weather stress studies.

Strong Methodological Framework: It presents a custom-made Heat Stress Index (HSI) as well as a consensus feature-selection scheme (Pearson-MI-RFE) and a Non-Negative Least Squares (NNLS) stacking ensemble. The ensemble (chronological splitting (leakage free) and statistical tests (Friedman, Wilcoxon, DM) provides a highly accurate prediction,

In terms of Actionable Climate Risk Quantification, it presents for the first time a SHAP-driven seasonal vulnerability ranking (Aus > Aman > Boro) and quantifies specific yields loss projections under IPCC warming scenarios (+3°C: 43.1% loss for Aus). It provides a concrete, data-driven basis for targeted climate-adaptation strategies in rice production.