Multi Scale Yield Prediction for Climate Change Scenarios.

Variance in crop production is highly influenced by weather and will become more uncertain as climate chance progresses. For this purpose it is important that yield modelling approaches are robust, flexible, and easily downscaled. This will allow agricultural practices to be tested at aggregated and highly localized scales simultaneously, while taking into account climate, weather, soil, and biological response in synthesis. A key issue to address is the accurate prediction of yield with the same model at both the national and field level.

For 5 years FarmLink has been compiling the single largest collection of quality yield data on research standard combines. The data captured covers over 6 million acres across 26 states and continues to grow each harvest. Precision yield data collection across multiple climatic and environmental conditions in multiple soil types has allowed the creation of an environmental and biological response database. From this research database a hybrid regression model has been used to characterize the complex interactions of climate, weather, soil, and field attributes. This has been the linchpin of a biological response model enabling the prediction of yield and estimate yield potential from any parcel of land in the continental U.S.

To detect variability of yield and land productivity this extensive field database can be leveraged to provide accurate yield prediction for multiple climate scenarios at the field, county, state, and national levels. This can be done using aggregated parameters for shifts in harvest season averages or with a modeled weather system of daily precipitation and temperature. Predicting changes in yield and productivity with increased accuracy will enable better management of risk in the face of climate change as well as facilitate agricultural development to adapt with changing weather patterns on a season by season basis.