Uncertainties in SWAT extreme flow simulation under climate change

- Extreme flow is modeled by SWAT combined with a regional climate model.
- SWAT parameter uncertainty under climate change is investigated.
- Uncertainty from scenarios, parameters and probability models is substantial.

SummaryUncertainty in climate change impact analysis has been widely recognized. Analyzing it becomes an important task particularly when impact analysis results are used for adaptation purposes. A methodology aiming to investigate the impact of climate change and the separate and combined impacts of several uncertainty sources on future extreme flows in the Lanjiang catchment, East China is proposed. A regional climate model PRECIS (Providing REgional Climates for Impacts Studies) is applied to downscale the General Circulation Model (GCM) outputs and the extreme flows are simulated by the SWAT (Soil and Water Assessment Tool) model. Besides emission scenarios and extreme value models, the main uncertainty source, namely SWAT parameters is taken into account and the sequential uncertainty analysis method is employed to analyze the parameter uncertainties. The SWAT model calibration and validation results indicate that the model has a good performance in Lanjiang catchment. The projected future extreme flows show that the design discharges of small return periods are likely larger than those in the baseline period, while those of large return periods will be likely smaller than those in the baseline period at Misai, Quzhou and Lanxi stations. The uncertainty analysis results show that for small return periods such as 5 years, the uncertainty introduced from the SWAT model parameters is much larger than those from emission scenarios and extreme value models at Misai and Quzhou stations while for large return periods such as 50 years, the uncertainty introduced from all three sources are substantial. However, the already large uncertainty due to SWAT model parameters, emissions scenarios and extreme flow distributions might be dwarfed by GCM uncertainty, which is not concerned in this study.