An Uncertainty Quantification Framework for Hydrologic Prediction under Climate Change
Friday, 19 December 2014: 4:43 PM
In recent years, there has been an increasing concern about the implications of climate change on water resources. Climate change has posed significant challenges to watershed management by changing the frequency and severity of extreme weather events, such as heat/cold waves, storms, floods and droughts. In this study, a regional climate model (PRECIS) is used to generate high resolution climate change scenarios over the province of Ontario, Canada. An advanced step-wise cluster approach, as a surrogate of the hydrologic model, is then developed to establish the precipitation-runoff relationship and address uncertainty in hydrologic cycles, providing a robust support for hydrologic prediction. To explore the potential interactions among hydrologic parameters, an ANOVA-based probabilistic collocation method is also proposed for uncertainty propagation in a reduced dimensional space. Multivariate inference is conducted to reveal statistical significance of model terms in the five-dimensional second-order and third-order polynomial chaos expansions (PCEs), respectively. A set of truncated PCEs can thus be generated through discarding insignificant terms, leading to a remarkable reduction in computational efforts. Results reveal that the truncated PCEs are good functional representations of the precipitation-runoff processes in terms of efficiency and accuracy. The performance of the third-order PCEs is better than the second-order PCEs through a comparison with Monte Carlo simulation. Finally, the impacts of climate change on watersheds in Ontario are evaluated and analyzed by using the future climate projection as the inputs of the surrogate models. This study meets the need for watershed adaptation to the changing climate. The outputs from this study will significantly improve our understanding of the challenges that Canadian watersheds are facing under various climate change scenarios, and provide scientific basis for watersheds management and adaptation.