Accuracy and uncertainty quantification of using a large climate ensemble dataset for process-based flood quantile estimation

Eight major river basins on Shikoku Island (21,000 km²), Japan. Large climate ensembles enhance the robustness of regional flood quantile estimation but impose substantial computational demands. This study evaluates two event-based simplification approaches (the aggregating grid event (AGE) method and the precipitation stretching (PrecStre) method) using the high-resolution (hourly, 150 m) Rainfall–Runoff–Inundation model. Reference 100-year discharges for all river grids are derived from all valid precipitation events of a 720-year climate ensemble. The AGE method achieves the highest accuracy and lowest uncertainty (BIAS −0.9%, MAPE 6.48%). Although the PrecStre method is more computationally efficient and lower interpretation cost, it shows inferior accuracy and robustness; under its optimal configuration, it yields a BIAS of −4.2% and a MAPE of 26.18%. Precipitation variability increases the internal uncertainty of PrecStre, but routing processes attenuate this effect and reduce downstream MAPE to 14.9%. Expanding climate ensemble members decreases uncertainty in flood quantile estimation and improves density of candidate events near the target return period, which can further improve the accuracy. The AGE method is expected to provides a benchmark for regional flood risk management, while the PrecStre method needs improvements for fulfilling explanatory needs of practical river planning. This study provides a theoretical foundation for developing regional adaptation planning based on large climate ensemble. • Aggregating Grid Event yields best discharge quantile estimates at low cost. • Routing lessens precipitation variability impact in precipitation stretching. • Using events of similar magnitude improves precipitation stretching performance.