The deficiency of the ‘economic optimum’ application for evaluating models which predict crop yield response to nitrogen fertiliser

Abstract

Much effort is devoted to testing N-fertiliser models against measurements of yield in N-fertiliser experiments. We show that the ‘economic optimum’ application of N-fertiliser can often only be determined very imprecisely from such measurements, even when they are accurate. Hence any attempt to evaluate a simulation model by comparing simulated with experimentally determined optima, or to relate these experimental optima to other factors thought to influence yield, will be uninformative. Big differences between simulated and measured optima will occur even with good models, which may explain the apparent poor performance of some simulation models in the literature.

We suggest, instead, the systematic inspection of the differences between simulated and observed yields at the levels of N-fertiliser applied in an experiment. A study of the relationship of these differences to the applied level of N-fertiliser and to other factors of interest will be a more informative method of evaluating a model, and of suggesting other factors for inclusion in it. An example of the use of the method is given.