6 Conclusions
This study presents an advanced hierarchical sensitivity analysis for a climate–influenced hydrological model system to quantify different sources of uncertainty in the hydrological impacts of future climate change. A multilayer hierarchical uncertainty quantification framework was developed to integrate with the variance–based sensitivity analysis method to estimate the relative importance of the uncertainty sources considered. The Latin hypercube sampling strategy was applied to calculate the sensitivity indices defined for this hierarchical sensitivity analysis. Variant uncertainty sources, including three different GGESs, thirteen plausible GCMs, two hydrological models and twenty sets of uncertain parameters, were quantified at the catchment scale. The spatio–temporal variability of the uncertainties considered in hydrological (annual discharge and annual peak discharge) predictions was comparatively analyzed using the four– (three–) layer hierarchical framework.
The sensitivity analysis results indicated that the GCMs and hydrological parameters are generally the main contributors of uncertainty in the discharge projections at the interannual scale. The uncertainty of GGESs is the smallest contributor of hydrological projections at the interannual scale, but the uncertainty of GGESs shows large variability over the projection periods. At the intra–annual scale, GCMs contribute the largest uncertainty of the discharge predictions particularly during summer season. In contrast, the uncertainty due to GGESs, hydrological model and parameters is generally limited, with the larger contributions in winter than in summer. At the spatial scale, a large spatial variation was identified in the sensitivity indices, suggesting that a single result for certain locations or one time point hardly capture the overall sensitivity information for a complex problem. Nevertheless, the spatial heterogeneity of the sensitivity indices does not affect the rank of the relative importance for uncertainty sources. The proposed framework is mathematically rigorous and general and can be applied to a wide range of climate–influenced models with more or different sources of uncertainty. The sensitivity results can provide key information for the knowledge of the spatio–temporal variations of various uncertainty sources in hydrological projections under future climate impacts.