Stochastic climate change projections by Bayesian estimation for uncertainty assessment of potential flood damage

This study applies a Bayesian estimation to quantify the future probabilistic rainfall and propagate uncertainty to future flood risk in the Cagayan Valley Region, Philippines, under climate change. Downscaled and bias-corrected CMIP6 precipitation data were analysed using Bayesian inference. Future probable rainfall is predicted by the MCMC method using the parameters of the generalised extreme value (GEV) distribution to estimate the probable extreme rainfall. The resulting probability density functions and 95% credible intervals clearly depict the uncertainty in the return level, showing that the credible intervals widen as the return period increases, especially for higher-emission scenarios. These probabilistic rainfall estimates were propagated through the Rainfall–Runoff–Inundation (RRI) model to generate stochastic flood simulations, representing flood hazards as probability distributions rather than deterministic outcomes. Higher-emission scenarios (SSP 2–4.5, SSP3-7.0 and SSP5-8.5) exhibited significant increases in extreme rainfall intensity. The simulated river water levels and flood risk in urban and agricultural areas are quantitatively assessed in the form of credible intervals and probability density functions. By visualising and propagating uncertainty from rainfall estimation to flood modelling, the framework provides a transparent and policy-relevant basis for evidence-based climate adaptation and risk-informed flood management.