Extreme rainfall events trigger natural hazards, including floods and debris flows, posing serious threats to society and the economy. Accurately quantifying extreme rainfall return levels in ungauged locations is crucial for improving flood protection infrastructure and mitigating water-related risks. This paper quantifies the added value of combining rainfall observations with Convection Permitting Model (CPM) simulations to estimate sub-daily extreme rainfall return levels in ungauged locations. We assess the performance of CPM-informed estimates of extreme return level against a traditional interpolation techniques. We find that kriging methods with external drift outperform inverse distance weighting for both traditional and CPM-informed approaches. We then assess the effectiveness of the two methods under different scenarios of station density. At the highest station density (1/196 km²), traditional interpolation methods outperform the CPM-informed method for durations under 6 h. The performance becomes comparable between 6 and 24 h. For lower station densities (1/400 and 1/800 km²), the CPM-informed method outperforms the traditional method, with average reductions in fractional standard error of 24 %, 13 %, and 8 % for return periods of 2, 10 and 50 years, respectively for a rain gauge density of 1/800 km², and 16 %, 8 %, and 3 % for density of 1/400 km². Information from CPM simulations can thus be useful for estimating sub-daily extreme rainfall events in ungauged sites, particularly in data-scarce areas in which the density of rain gauges is low.

中文翻译：

---

低于日的降水量返回未测量位置的水平：将观测与对流相结合的额外价值允许模拟


极端降雨事件会引发自然灾害，包括洪水和泥石流，对社会和经济构成严重威胁。准确量化未测量位置的极端降雨回降水平对于改善防洪基础设施和降低与水相关的风险至关重要。本文量化了将降雨观测与对流许可模型 （CPM） 模拟相结合的附加值，以估计未测量位置的低于每日的极端降雨返回水平。我们评估了 CPM 对极端回报水平的估计与传统插值技术的性能。我们发现，对于传统方法和 CPM 知情方法，具有外部漂移的克里金方法的性能优于反距离加权。然后，我们评估了两种方法在不同站点密度情景下的有效性。在最高的站点密度 （1/196 km²） 下，传统的插值方法在 6 小时以下的持续时间优于 CPM 知情方法。性能在 6 到 24 小时之间变得相当。对于较低的站点密度（1/400 和 1/800 km²），CPM 信息方法优于传统方法，对于 1/800 km² 的雨量计密度，在 2 年、10 年和 50 年的重现期内，分数标准误差平均分别减少了 24 %、13 % 和 8 %，对于 1/400 km² 的密度，分数标准误差平均减少了 16 %、8 % 和 3 %。因此，来自 CPM 模拟的信息可用于估计未测量站点的每日以下极端降雨事件，特别是在雨量计密度较低的数据稀缺地区。