Multi‐Objective Calibration of an Ecohydrological Model for Discharge, Phosphorus and Sediment Dynamics in a German Lowland Catchment

Phosphorus losses from agriculture remain key drivers of eutrophication in lowland catchments, where flat topography and artificial drainage complicate the representation of transport processes in water‐quality models. The SWAT+ model was implemented for the Kielstau catchment in Northern Germany, a small rural lowland system with intensive agriculture, diffuse inputs from runoff, erosion, and point inputs from six wastewater treatment plants. Effects of simultaneous calibration for daily discharge, sediment and total phosphorus (TP) data on model performance and the trade‐offs inherent to synchronous multi‐objective calibration were assessed. A joint multi‐metric framework was implemented using 10 000 model runs generated by Latin hypercube sampling of 25 sensitive SWAT+ parameters. Performance was evaluated using univariate performance criteria, duration‐curve metrics, weighted objective functions, and Euclidean distance. Discharge performance was consistently high, whereas sediment and TP showed larger uncertainty, especially during high‐flow events. Parameter interaction analysis highlighted CN2 (curve number controlling runoff generation), soil‐moisture availability, and HRU slope length and steepness (controlling flow energy and erosion potential) as dominant controls on event‐driven erosion and phosphorus transport. No single run achieved equally strong performance across all variables. Duration‐curve and hydrograph diagnostics revealed systematic biases, including underestimated TP during low flows and sediment peaks during events. Weighted and multi‐objective selection methods yielded the most balanced solutions, demonstrating the value of extending evaluations beyond aggregated goodness‐of‐fit metrics using temporal and segment‐specific diagnostics to support robust calibration strategies in lowland systems. Nevertheless, model applications require explicit acknowledgement of structural limitations, parameter sensitivities and trade‐offs.