Report on Corn and Soybean Field Yield Calculation Experiment
Introduction
To improve risk assessment, we at N.O.A.H. use a model that calculates the potential yield for each of the fields we insure. We set out to test how well our model works on a real example. We used agro-physical and agrometeorological soil parameters, as well as time-series historical satellite data.
The data owner, a farmer from West Hallock, Illinois, USA, kindly agreed to provide us with data from four of his fields. For these fields, with areas of 45.6 Ha, 30.88 Ha, 16.22 Ha, and 30.02 Ha, and planted with corn and soybeans, we obtained baseline data for 2023.
Data Collected
Agro-Physical Soil Data
- Soil moisture reserves
- Tillage depth
- Volumetric weight
- Organic content
- Nitrogen content
- Soil acidity (pH)
- Zinc and Boron content
Agrometeorological Data
- Solar radiation
- Precipitation amounts
- Temperature profile for the year
Satellite Data
- Optical data from Sentinel-2
- Radar data from Sentinel-1
Results
The yield predictions obtained were compared with the actual 2023 yields. We're managed to reach a prediction accuracy in the range of 94.95 - 97.18% for corn and 90,89% for soybeans.
Conclusion
The accuracy of our model's predictions in this real-world example demonstrates its effectiveness in calculating field yields, making it a valuable tool for farmers and opens a way for us to help crop insurance agents in their job.
