Nitrogen Response Modeling with a Multidegree Spline Plateau Approach

Nitrogen response models are often estimated as plateau type models. Existing plateau models are nonlinear in the parameters. The model introduced here is a multi-degree spline plateau model, which is linear in parameters. The linearity allows using approximation methods so that estimating a spatially varying coefficient model with Bayesian Kriging could be a commercially feasible approach to guide precision nitrogen applications. With proper knot selection, the spline plateau model can fit as well as the linear and quadratic plateau models. The computational efficiency of the spline plateau model was demonstrated using R-INLA and it provided fast and accurate estimation.