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.