Incorporating climate model similarities and hydrologic error models to quantify climate change impacts on future riverine flood risk

Abstract

Quantifying uncertainty in projecting climate change impact on flood frequency analysis is particularly relevant for long-term water resources planning and management. This study examines uncertainties arising from (1) a climate model, with and without accounting of intermodel similarities, (2) a hydrological model with two different error model definitions, which have previously received less attention in studies propagating uncertainty through climate change impacts on flood response. Through a Bayesian modeling framework, a proposed statistical framework is utilized to explore various definitions of sources of uncertainty and to develop a series of nested formulations that can evaluate the leverage of specific uncertainty sources in quantiles of future streamflow. To be specific, climate model similarity, hydrologic prediction error, and frequency analysis are formally modeled with an appropriate likelihood function. The quantiles inferred by each formulation are compared in a case study of the Yongdam Basin in Korea. Results indicate that variance in hydrologic response is underestimated if climate model similarities are ignored, and in many cases, the inferred quantiles of streamflow projection are biased accordingly. Furthermore, a simple error model used in defining hydrologic uncertainty may create incorrect information in determining the quantiles in streamflow projection. The approach presented here of quantifying uncertainty has the potential to better depict overall multi-source uncertainties in projections of climate change impacts on hydrologic response. Finally, our results also indicate that careful flood planning management may be required for the Yongdam Basin in future summers.