Combining Ecohydrological Catchment Modeling and Water Quality Monitoring Data to Assess Nitrogen Pollution in the Swist River Basin, Germany

Abstract

Legislation like the EU Water Framework Directive demands the development as well as implementation of catchment-scale monitoring and management plans to control pollution of surface waters and ensure their good status. Water pollution with excess amounts of reactive nitrogen (N) - for example nitrate - can cause environmental problems like eutrophication while also posing potential health risks to humans. Knowledge of pollution sources and their relative importance in a given catchment is a prerequisite for developing targeted monitoring strategies and efficient management plans. The process of quantifying the contribution of each pollution source to overall pollution in a catchment is called source apportionment. Due to the abundance of pollution sources as well as their heterogeneity in time and space, source apportionment solely based on water quality monitoring is not feasible in large catchments, especially with regard to non-point source (diffuse) pollution. Ecohydrological computer models can be used to complement water quality monitoring data in space and time, identify critical source areas and carry out source apportionment studies. The overall research aim of the present thesis is to assess N pollution of surface waters in the Swist river basin, Germany, from 2012 to 2018 using the ecohydrological catchment model SWAT (Soil and Water Assessment Tool).<br /> Calibration of SWAT using Monte Carlo analysis generated a SWAT model ensemble allowing for analysis of modeling uncertainty. Furthermore, two independent reference data sets on diffuse N pollution in the Swist basin were available for comparison with the SWAT simulation results: (1) An emission inventory from upscaled edge-of-field monitoring data and (2) raster data generated by the AGRUM modeling system. The calibrated SWAT models predicted a mean diffuse N emission load of 8.1 kg/(ha a) for the entire Swist catchment area in the study period (median: 6.9 kg/(ha a)). Agricultural tile drainage was identified as the most important emission pathway in the catchment, shaping much of the spatial variability in diffuse N pollution. In contrast, due to missing groundwater contact in large parts of the catchment (a consequence of lignite mining north of the study area), surface waters received only minimal N from groundwater, possibly indicating a dominance of faster (tile drains) over slower (groundwater) emission pathways in the catchment. The highest diffuse N loads were predicted near the headwaters of the Schiessbach tributary by all SWAT models, making this region a likely critical source area for N in the catchment. On average, wastewater treatment plants contributed approximately one fourth (26.8 %) of overall N pollution in the simulations. As all these findings were basically consistent between the SWAT simulations and the two reference data sets, they can be used with relative confidence in the future to focus monitoring efforts, devise emission control strategies and implement effective mitigation measures.<br /> However, comparison of the available data also revealed some meaningful discrepancies, highlighting the remaining uncertainties in the results. Diffuse N emission loads simulated by the SWAT ensemble ranged from 4.7 to 11.4 kg/(ha a) (excluding one outlier at 15.9 kg/(ha a)). Since the two reference data sets mostly agree with SWAT in the order of magnitude of these values, the results are considered a realistic appraisal of diffuse N pollution in the Swist catchment. Still, with a factor of more than two (or three when including the outlier) between the upper and lower end of the uncertainty interval, the exact amount of N released to the Swist and its tributaries remains difficult to quantify. Although the general ranking of the individual pathways is mostly stable between the SWAT models and likewise confirmed by the reference data, the relative uncertainty intervals associated with their N contributions are even wider than for the overall emission loads. Apart from the Schiessbach headwaters, no other obvious critical source area candidate emerged among the SWAT simulation results and the reference data, with regions of high N emission loads mostly fluctuating between the different model results. The simulated contributions from wastewater treatment plants ranged from 14.7 to 36.7 % of overall N emissions in the catchment.<br /> For a future continuation of SWAT modeling in the Swist catchment to assess N pollution, there are several possibilities to potentially reduce the uncertainty in the simulation results. First, including additional data and/or objective functions in model calibration probably helps to better constrain the SWAT parameters. Second, the correction of some deficiencies in the model setup (i.e. input data and model structure) - for example in the amounts of applied N fertilizer or the omission of combined sewer overflows - makes model calibration presumably more efficient and may eliminate potential bias in the simulation results. In summary, ecohydrological catchment modeling with SWAT in the present thesis was successful in generating novel insights regarding N pollution in the Swist catchment. Here, model-based source apportionment benefited immensely from the extraordinary wealth of monitoring data available for the Swist catchment and the possibility to compare the SWAT modeling results to two independent reference data sets for the study area. This underscores the importance of comprehensive monitoring and knowledge of relevant local conditions to reach a sound understanding of surface water pollution on the catchment scale.