Key points are not available for this paper at this time.
Climate change can intensify multi-day extreme rainfall and increase damaging riverine floods. However, rare-event precipitation is highly uncertain and propagates nonlinearly to inundation and losses. This study develops a probabilistic flood-impact assessment for the Cagayan River Basin (northern Luzon, Philippines) by linking Bayesian extreme-value inference with uncertainty propagation along the rainfall–flood–damage chain. Basin-mean 1-day and 7-day annual maxima are modeled with Bayesian generalized extreme value inference. The 100-year return level (RL100) of design rainfall under SSP1–2.6 and SSP3–7.0 is estimated by combining three CMIP6 GCMs using continuous ranked probability score-based Bayesian model averaging. Flood response is represented with Rainfall–Runoff–Inundation (RRI) model simulations driven by a fixed storm template (Ondoy–Pepeng) scaled across rainfall totals. From these simulations, a rainfall–peak-stage relationship and quadratic rainfall–loss functions for buildings, rice, and corn are derived. Monte Carlo sampling of RL100 rainfall is propagated through these relationships to obtain posterior distributions of peak stage and event losses, summarized by credible-range widths for central (50%) and tail-inclusive (95%) intervals. SSP1–2.6 remains close to the present-climate reference with relatively narrow uncertainty, whereas SSP3–7.0 shifts RL100 upward and widens the distribution. Uncertainty further expands when mapped to peak stage and losses; the 95% range width nearly doubles for building damages (1944 to 3817 million PHP) and increases for rice (1336 to 2494 million PHP). These results show that deterministic RL100-based estimates can understate tail risk under stronger forcing, and that probabilistic reporting is essential for robust adaptation planning.
Kurihara et al. (Sat,) studied this question.