How good is our crop model at predicting yield and nitrogen needs?
Estimating yield might be the most difficult task for any model to perform because of the highly variable and highly complex interactions that occur within a field, a decision zone and at even smaller scales. The goal of our service is not to accurately predict yield. The key component to making a nitrogen recommendation is to get the yield goal set correctly, since that is the biggest driver in the process.
The model yield prediction does somewhat (not linearly) impact plant nitrogen uptake. The plant uses nitrogen for growth of stalks, leaves, roots, etc. and stores and remobilizes nitrogen throughout the season, so there is quite a bit of buffering capacity in the system. We want to be reasonably close, however we believe being within 20-30 bushel of actual is adequate. In general, the model tends to predict yields slightly higher than the actual measured yield.
We can directly compare the new model to the previous Ceres Maize model using the 2014-2016 PRNT (Pioneer Regional Nitrogen Trials) dataset. These experiments were conducted by university researchers at locations across the corn belt and were sponsored by Pioneer Agronomy Sciences. An analysis of that data shows that the new model reduces the mean average error for the economic optimum nitrogen rate by about 16% over the Ceres Maize model.
Granular Nitrogen is a full service offering to our customers that includes the tremendous value provided by a certified services agent. We believe their extensive local agronomic knowledge and experience along with the peace of mind and convenience they provide for our growers is a key part of the value proposition to a grower.