Evaluating the Importance of Non-Unique Behavioural Parameter Sets on Surface Water Quality Variables under Climate Change Conditions in a Mesoscale Agricultural Watershed

The parameter uncertainty in the eco-hydrological model Soil and Water Assessment Tool (SWAT) was estimated using non-unique parameter sets for the Altmühl watershed (Bavaria, Germany). The Sequential Uncertainty Fitting Algorithm (SUFI-2) was used to calibrate SWAT. The non-unique parameter sets found were subsequently applied to SWAT concurrently with climate change simulations to determine the variables of streamflow, nitrate nitrogen (NO₃ ⁻-N) and total phosphorus (TP). A suite of seven bias corrected climate change simulations provided reference (1970–2000) and future (2041–2070) climate data. The non-unique behavioural parameter sets that met an objective function of NSE >0.6 during calibration were applied to SWAT with the reference climate and with the future climate simulations. The best parameter set was also propagated through SWAT with each reference and future climate simulation in turn. Combining the non-unique behavioural parameter sets for estimating uncertainty bounds with an ensemble of climate change simulations led to a wider mean monthly spread (difference between maximum and minimum) of simulated NO₃ ⁻-N and TP than using the best run with the future climate simulations. More monthly data was considered using the non-unique approach, resulting in statistical significances for more months of the year and overall lower interquartile ranges. The study quantifies the non-unique behavioural parameter set contributions to the modelling prediction, which assists in making more informed decisions based on available knowledge, with its limitations, of the future simulations. We outline a simple approach that can easily be replicated for similar hydrological modelling studies.