4.3 Sensitivity Analysis for Annual Peak Discharge
The annual peak discharge, which represents the annual maximum value of annual discharge, was chosen as an output of interest for the sensitivity analysis. The variability of annual peak discharge by averaging hydrological projections from the GGESs, GCMs, hydrological models and parameters is shown in Figure 8. The dispersion of annual peak discharge from the four uncertainty sources is comparatively large. The sensitivity indices for the predictions of annual peak discharge at the Hengshi hydrological station were estimated and the results are displayed in Figure 9. It can be seen that GCMs and hydrological parameters account for the two largest portions of uncertainties in the discharge predictions, with the average contributions of 31.57% and 50.2%, respectively (Figure 9b and d). The sensitivity indices of hydrological parameters are in the range of 36.0%–57.1% during the projection period, which are higher than that of annual average discharge (33.34%–47.76% in Figure 5), suggesting larger influence of hydrological parameters on the extreme hydrological events. Compared with the hydrological parameters, the uncertainty from the hydrological model structures tends to be smaller and accounts for 10.8%–23.1% of the total uncertainty. The GGESs are still the least important source of uncertainty for the annual peak discharge predictions, accounting for 0.01%–3.78% of the total uncertainty, which are also smaller than the uncertainty in annual average projections. The variability of the total variance of annual peak discharge is shown in Figure 9e, and the dispersion of individual source is generally large.