
A Scientific Reports study applies interpretable machine learning to a large yield monitor dataset of 134 US maize and soybean crop-site-years. The models predict yield with R² above 0.87 and use Shapley values to rank the effects of weather, soil and terrain.
The authors aim to inform breeding and precision management.
The study improves understanding of which environmental factors limit yields, which may help future supply forecasting. It has no immediate effect on prices.
After 2,467 past weather events, CORN moved a median +0.1% over 5 sessions (51% higher; middle half -1.5% to +1.7%), and +0.6% over 20.
After 1,964 past weather events, SOY moved a median +0.1% over 5 sessions (51% higher; middle half -1.7% to +1.8%), and +0.3% over 20.
Moves of each commodity's benchmark from the last close before the event. Historical reads pool past events of the same kind; they describe what happened, not what will. Not investment advice.