Can remote sensing-based yield proxies substitute crop cutting experiments in Indian agricultural insurance models?

Accurate and scalable yield estimation is fundamental to the effectiveness of agricultural insurance protecting smallholder farmers from climate-induced yield losses in farming systems where basis risk remains a critical challenge. This study examines the potential of remote sensing ( RS )-based yield proxies to supplement and progressively replace traditional crop-cutting experiments (CCEs) in India's area-yield crop insurance model presented as a proof-of-concept pilot for the Baghelkhand region of Madhya Pradesh. A hybrid approach combining crop simulation (DSSAT) and machine learning was applied to predict yield variability across rainfed paddy farming systems in the Baghelkhand region of Madhya Pradesh. Vegetation indices from Landsat and Sentinel satellites and weather variables (rainfall, temperature, evapotranspiration) from NRSC and CHIRP were analyzed via Google Earth Engine and ArcGIS. Using five-fold cross-validation and data from 372 farmer interactions, the model achieved strong performance (R 2 = 0.81) at the district level. Results show that stage-wise aggregation of predictor variables significantly improves accuracy over seasonal averages. A novel threshold yield adjustment framework enables early warning at pre sowing (ex-ante) and post-harvest stages (ex-post), reducing basis risk by approximately 7% compared to the existing model. The study further integrates the Farmer Satisfaction Index into loss assessment methodologies as a deliberate mechanism to embed farmer-centric subjectivity into an otherwise data-driven model, addressing moral hazard concerns and promoting more equitable and need-based insurance payouts advocating a transition from rigid historical TY benchmarks to real-time data-driven adjustments for more resilient and farmer-centric agricultural risk management systems, with validation across broader agro-climatic zones identified as the next research priority.