Agro‑Climatic Hazard Profiling of Rice and Wheat Under SSP245 Using a Spatial Multi‑model Framework in India
Аннотация
Climate change is reshaping agro‑ecological dynamics across India, exposing staple crops such as rice and wheat to compound climatic stressors. This study develops a multi‑hazard, spatially explicit framework for district‑level risk profiling under the SSP245 scenario, integrating machine learning classification, principal component analysis (PCA), and composite hazard indexing. Among seven tested algorithms, Random Forest achieved the highest predictive accuracy (0.994) and R² (0.976), effectively capturing non‑linear hazard interactions. PCA reduced dimensional complexity, with two principal components explaining over 80% of variance and enabling clustering of stress patterns across crops. A 12‑variable Synthetic_MAP index, derived from remote sensing indicators, quantified multi‑seasonal hazard convergence and identified high‑risk districts such as Surendranagar and Thanjavur (scores > 1700). To enhance operational relevance, the framework integrates agro‑biomass proxies from satellite vegetation indices with early warning indicators, stratifying districts into resilience tiers. These outputs directly support the Gramin Krishi Mausam Sewa (GKMS) by enabling crop‑stage‑specific advisories, hazard‑triggered alerts, and district‑level bulletin customization. By aligning climate thresholds with forecast parameters, the model strengthens GKMS’s capacity for anticipatory decision‑making, precision extension services, and adaptive agronomic planning. Overall, this integrated approach provides a scalable decision‑support tool for climate‑smart agriculture, early warning systems, and national food security preparedness, demonstrating the value of combining remote sensing datasets, GIS‑based spatial analytics, and machine learning for operational climate risk management.