Comparative study of nitrate leaching models on a regional scale.

In Europe and North America the application of high levels of manure and fertilisers on agricultural land has led to high levels of nitrate concentrations in groundwater, in particular on sandy soils. For the evaluation of the development of the quality of groundwater a sound quantitative basis is needed. In this paper a comparison has been made between observations of nitrate concentrations in the upper groundwater and predictions of nitrate leaching models. Observations of nitrate concentrations in the upper groundwater at three different locations in regions with mainly sandy soils in the eastern and northern part of the Netherlands were used to test the performance of the simulation models to predict nitrate leaching to the upper groundwater. Four different types of simulation models of different levels of complexity and input data requirement were tested. These models are ANIMO (dynamic complex process oriented model), MM-WSV (meta-model), WOG (simple process oriented model) and NURP (semi-empiric model). The performance of the different simulation models was evaluated using statistical criteria. The dynamic complex process oriented ANIMO model showed the best model performance. The MM-WSV meta-model was the second best model, whilst the simple process oriented WOG model produced the worst model performance. The best model performance showed the dynamic complex process oriented ANIMO model in predicting the nitrate concentrations in the upper groundwater of the Klooster catchment. The good performance of the ANIMO model for this catchment can be explained by the additional information about the use of manure and fertilisers at farm level in this study area. The ANIMO model may be a good tool to predict nitrate concentrations in the upper groundwater on a regional scale. However, the use of a detailed process oriented simulation model requires a comprehensive set of input data. If such a comprehensive data-set is not available the MM-WSV model (meta-model) proves to be a good alternative. The WOG and NURP models are suitable for long term (>8 years) predictions of average nitrate concentrations in the upper groundwater on a regional scale.