Articles | Volume 11, issue 2
Hydrol. Earth Syst. Sci., 11, 997–1011, 2007
https://doi.org/10.5194/hess-11-997-2007
Hydrol. Earth Syst. Sci., 11, 997–1011, 2007
https://doi.org/10.5194/hess-11-997-2007

  22 Mar 2007

22 Mar 2007

Assessing the model performance of an integrated hydrological and biogeochemical model for discharge and nitrate load predictions

T. Pohlert, L. Breuer, J. A. Huisman, and H.-G. Frede T. Pohlert et al.
  • Institute for Landscape Ecology and Resources Management, Justus-Liebig-University Gießen, Heinrich-Buff-Ring 26, 35392 Gießen, Germany

Abstract. In this study, we evaluate the performance of the SWAT-N model, a modified version of the widely used SWAT version, for discharge and nitrate predictions at the mesoscale Dill catchment (Germany) for a 5-year period. The underlying question is, whether the model efficiency is sufficient for scenario analysis of land-use changes on both water quantity and quality. The Shuffled Complex Evolution (SCE-UA) algorithm is used to calibrate the model for daily discharge at the catchments outlet. Model performance is assessed with a split-sampling as well as a proxy-basin test using recorded hydrographs of four additional gauges located within the catchment. The efficiency regarding nitrate load simulation is assessed without further calibration on a daily, log-daily, weekly, and monthly basis as compared to observations derived from an intensive sampling campaign conducted at the catchments outlet. A new approach is employed to test the spatial consistency of the model, where simulated longitudinal profiles of nitrate concentrations were compared with observed longitudinal profiles. It is concluded that the model efficiency of SWAT-N is sufficient for the assessment of scenarios for daily discharge predictions. SWAT-N can be employed without further calibration for nitrate load simulations on both a weekly and monthly basis with an acceptable degree of accuracy. However, the model efficiency for daily nitrate load is insufficient, which can be attributed to both data uncertainty (i.e. point-source effluents and actual farming practise) as well as structural errors. The simulated longitudinal profiles meet the observations reasonably well, which suggests that the model is spatially consistent.

Download