Skip to content

High-Resolution Crop Yield Mapping Reveals the Spatial Variability of the 2012 Drought

We have successfully deployed our KGML-DA model to map the 30-m pixel-level corn and soybean yields across Champaign County, Illinois, spanning from 2010 to 2014.

This visualization demonstrates the effectiveness of our proposed KGML-DA framework for subplot-level simulation compared to traditional methods.

Explore the Data

Below is the interactive map visualizing these crop yields.

The 2012 Drought

The 2012 Midwest drought provided a perfect stress test for our models. Severe drought conditions in the pre-season and throughout July decreased initial soil moisture and severely affected corn pollination. As a result, the reported corn yield for 2012 dropped to 108.9 Bu/Acre.

  • The KGML-DA solution: By assimilating 30-m SLOPE GPP data, our framework captured the actual in-season crop conditions. This reduced the absolute error of county-level corn yield estimates by 9.5 to 21.5 Bu/Acre.

  • The result: The KGML-DA model successfully pulled up underestimated corn yields in 2010, 2011, and 2013, while suppressing the overestimated yield in 2012. Interestingly, soybean yield estimations showed less benefit from assimilating the 30m GPP data.

  • Drought resistance: The reported soybean yield was not significantly impacted by the 2012 drought. Soybeans are generally less sensitive to atmospheric conditions, and their relatively late planting allowed them to bypass the severest drought period from May to July.