A comparison of the performance of N simulation models in the prediction of Nmin on farmers' fields in the spring

The performance of three different models, which simulate changes in the inorganic N content of the soil, was evaluated in respect of their ability to predict N_min content in the spring under cereal crops. The models of British, Dutch and German origin, were tested using data from farmers' fields supplied by 70 farmers over two growing seasons in FRG. The models were run between harvest of the previous crop and spring of the following year, and predictions of N_min in the spring compared to soil measurements. The performance of the models was assessed by counting the number of cases in which predictions agreed within 10 or 20 kg (N) ha^–1 of the measurements. Predictions were less than ± 10 kg (N) ha^–1 of measured values in only 30–44% and 28–55% of cases in 1988 and 1989, respectively. Predictions were less than ± 20 kg (N) ha^–1 of measured values in 62–70% and 68–82% of cases in 1988 and 1989, respectively. Predictions in 1989 were better because the initial N_min content in the autumn was included in the model input. None of the models tested had been designed to use input data of the type available to farmers. It is concluded that, at present, the results are too variable for any of the models to be used with confidence as tools to aid in N fertilizer recommendations.