A novel framework for uncertainty quantification of rainfall-runoff models based on a Bayesian approach focused on transboundary river basins

Study region: The transboundary Imjin River basin, Korea. Study focus: The primary aim is to propose and validate a novel framework for assessing the uncertainty in hydrological models, particularly rainfall-runoff models (RRMs), considering transboundary river basins with limited data accessibility. By utilizing an adaptive Markov chain Monte Carlo (MCMC) simulation method combined with three comprehensive uncertainty assessment measures, the developed framework focuses on evaluating the uncertainty inherent in RRMs. A key component of this framework is the delayed rejection adaptive Metropolis (DRAM) algorithm, which is employed to explore behavioral simulations defined by four likelihood functions (LFs). The proposed methodology was applied to the transboundary Imjin River basin using the Sejong University rainfall-runoff (SURR) model, a case study that involves a database of five-year extreme flood events. New hydrological insights for the region: The application of this framework in the transboundary Imjin basin demonstrated its effectiveness in quantifying and addressing the uncertainty in RRM predictions. The integration of the DRAM algorithm with uncertainty indices provided a robust mechanism for evaluating and improving the reliability of RRM outputs for transboundary basins. Effects of LFs in blending with the DRAM algorithm were confirmed by uncertainty measures and the behavior of the upper and lower uncertainty bounds. These insights could provide an approach to develop more accurate and reliable water resource management strategies in global transboundary contexts.