This is like having three different weather apps for your farm that try to forecast how much crop you’ll harvest, then checking which one is most accurate. One app uses plant science rules (AquaCrop), one uses simplified physics of how crops grow, and one uses a learning robot (a neural network) that learns from past seasons’ data to predict future yields.
Reduces uncertainty in how much farmers will harvest in Illinois by comparing and validating different modeling approaches (process-based crop models, semi-physical models, and neural networks), enabling better decisions on planting, irrigation, fertilizer use, contracts, and risk management.
Access to historical yield and weather/soil datasets for Illinois and the modeling expertise to calibrate both process-based crop models and data-driven neural networks to local conditions.
Hybrid
Time-Series DB
High (Custom Models/Infra)
Data availability and quality for each region (weather, soil, management practices) and the effort required to calibrate models like AquaCrop and train neural networks for different crops and geographies.
Early Majority
Combines and directly evaluates three different modeling paradigms—AquaCrop (a process-based FAO crop model), semi-physical models, and artificial neural networks—on the same Illinois dataset, giving a clearer picture of trade-offs between interpretability and accuracy for regional yield prediction.