This study develops a crop yield emulator as a computationally efficient alternative to the DSSAT process-based model, enabling fast, large-scale assessments of how temperature, rainfall, CO2, and nitrogen influence crop yields. Implemented using fixed-effects polynomial regression in R, the emulator was trained on DSSAT outputs and validated across seven globally important crops, maize, wheat (spring and winter), rice (indica and japonica), soybeans, sorghum, groundnuts, and potatoes. It showed high predictive accuracy (R2 > 0.90) for most crops, especially spring wheat, groundnut, and potato, and performed reliably under climate scenarios such as RCP7.0. The emulator captured key crop-specific responses, such as maize’s sensitivity to heat and nonlinear yield responses to nitrogen inputs. While its performance is slightly lower for maize and Indica rice compared to other crops, it remains a scalable and cost-effective tool for climate scenario analysis. This tool offers valuable support for agricultural decision-making and climate adaptation planning by enabling rapid, data-driven insights into yield outcomes under future conditions.
This study develops a crop yield emulator as a computationally efficient alternative to the DSSAT process-based model, enabling fast, large-scale assessments of how temperature, rainfall, CO2, and nitrogen influence crop yields. Implemented using fixed-effects polynomial regression in R, the emul...