Watershed modeling with arcswat: calibration and validation for the prediction of flow, nitrate and phosphorus load

Abstract

The European Water Framework Directive requires that all surface waters and groundwater within defined river basin districts must reach at least ‘good’ status by 2015. Thus, the directive requires the development of management strategies to restore rivers and lakes to “good” status within a specified timeframe. Simulation models are essential tools to evaluate potential consequences of proposed strategies and to facilitate management decisions. One of the most commonly used river basin model is ArcSWAT, a combination of the simulation model SWAT with a GIS user interface. To test the application of ArcSWAT under German conditions, a relatively small watershed (52 km²) with sufficient data on soils, land use, climate, water flow and river water quality was selected. The watershed shows a wide variety of land use (intensive farm land, extensive pasture, and forest), soils (light sandy soils and heavy loams), surface slopes (flat to 30 % slope) and a few potential point sources of nutrients, like fish ponds, wastewater treatment plants and a public compost plant. The measured Phosphate concentration over a period of 20 years was relatively uniform and low (0.2 mg/l), whereas the Nitrate concentration varied considerably between 10 and 50 mg/l. The model was calibrated and validated for the prediction of flow, Nitrate and Phosphate concentration and load at the main basin outlet. Sensitive model parameters were determined and adjusted within feasible ranges to minimize model errors monthly flow and quality data. The calibration resulted in good model predictions of the monthly discharge, whereas the error in Nitrate and Phosphorus concentration and load was much larger. Possible reasons for the poor simulation results for the nutrients are incorrect amounts of fertilizers in the automatic fertilization option of ArcSWAT. This resulted in incorrect yield and biomass production in the model. For a further improvement of the simulation quality, different arable crops and adequate model fertilization values have to be considered. The model can be used for German conditions with data usually available in Germany without too much calibration work for simulating and predicting the monthly and annual discharge. To simulate nutrient concentration and load, much more calibration work is necessary.