The EUROHARP project1 was a large scale comparison and evaluation of contrasting nutrient pollution models used as policy support tools to estimate river water quality. Accurate estimation of diffuse pollution, such as nitrate and phosphorus, from agriculture is a major challenge. Results need to be accurate and responsive to changes in land use and land management to assess compliance with EU policy instruments, to monitor trends in water quality, and to target mitigation measures in time and space.
Contrasting types of modelling tools that predict river flows and concentrations of pollutants were compared with measured data. Models were chosen as examples of different approaches (process based, semi-empirical, and conceptual) used as policy support tools in EU Member States. Representative river catchments were chosen from 17 European catchments covering a range of climates (from north to south), soils, hydrology, and land uses.
Results of the models EveNFlow and PSYCHIC showed that both were capable of acceptable performance in modelling flows, nutrient concentrations and loads in five out of six test catchment areas. In the sixth area, in Greece, results were less accurate, due to limitations in input data (rainfall, groundwater, point sources). Accurate predictions of daily river flows were the primary factor influencing the satisfactory prediction of nitrate and phosphorus concentrations and loads. However, both models showed adaptability and were capable of performing reasonably well when confronted with limited or unknown data in the other five test catchments. The results demonstrated that diffuse pollution tools can be applied to catchments where climate and agricultural practices are very different to those in the areas where such models were originally developed. This adaptability demonstrates that some diffuse pollution tools have the flexibility required for demanding and strategic policy support purposes. Important generic principles for diffuse pollution support which emerged from this work included:
Model selection should be governed by end-user requirements, taking into account the availability of input data (e.g. screening tools may be best suited to less complex approaches, whereas the system feedback implicit in scenario modelling would benefit from adopting relatively more complex approaches). Modellers need to actively engage with catchment data managers to ensure valid assumptions are made.