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.