Understanding short-duration intense rainfall is crucial for mitigating flash floods, landslides, soil erosion, and pollution incidents. Yet, most observations from rain gauges are only available at the daily resolution. We use the new Global Sub Daily Rainfall dataset to explore extreme rainfall at both daily and sub-daily durations worldwide. Employing Single Gauge Analysis (SGA) and pioneering global-scale Regional Frequency Analysis (RFA), we reveal for the first time how Generalized Extreme Value Distribution (GEV) parameters change across climates and data durations (1h, 3h, 6h, 24h, and daily). This marks the first-ever near-global-scale RFA, made possible by the development of an algorithm that automates RFA on observed rainfall datasets. We compare our results with GEV applied to a gridded rainfall reanalysis (ERA5). Our key findings are that: 1) using ERA5, return levels are significantly underestimated across all climates for 1h rainfall and across all data durations for gauges in the tropical climate region. Even when accounting for differences between point and areal estimates, the median 1h return level estimates are approximately 40% lower compared to RFA. We therefore strongly advise against the use of reanalysis gridded rainfall for studying these extremes. 2) While most gauges show similar return levels with RFA or SGA, some differ significantly, and either method may yield the highest values. Thus, we strongly recommend using both SGA and RFA simultaneously to estimate return levels for a robust risk assessment in flood infrastructure design. 3) The interaction between daily and sub-daily GEV shape parameters varies across climate regions, rendering a universal method for inferring sub-daily rainfall extremes from daily extremes (e.g., using Intensity-Duration-Frequency curves) impractical. Our research provides innovative methodological insights that warrant consideration in future studies on rainfall extremes. Our results not only benefit local stakeholders globally but also serve as a crucial validation tool for the rising number of convection-permitting climate model experiments conducted worldwide.

中文翻译：

---

揭示不同气候下日和次日极端降雨之间的复杂相互作用


了解短时强降雨对于减轻山洪暴发、山体滑坡、土壤侵蚀和污染事件至关重要。然而，雨量计的大多数观测结果只能在每日分辨率下获得。我们使用新的全球次日降雨量数据集来探索全球日内和日内极端降雨。采用单量表分析 （SGA） 和开创性的全球尺度区域频率分析 （RFA），我们首次揭示了广义极值分布 （GEV） 参数如何随气候和数据持续时间（1 小时、3 小时、6 小时、24 小时和每天）而变化。这标志着有史以来第一个近全球规模的 RFA，通过开发一种算法来实现，该算法可以在观测到的降雨数据集上自动执行 RFA。我们将我们的结果与应用于网格降雨再分析 （ERA5） 的 GEV 进行了比较。我们的主要发现是：1） 使用 ERA5，热带气候地区 1 小时降雨量的所有气候和所有数据持续时间的回波水平都被明显低估了。即使考虑到点和面估计之间的差异，与 RFA 相比，1 小时回报水平估计的中位数也低约 40%。因此，我们强烈建议不要使用再分析网格降雨来研究这些极端值。2） 虽然大多数指标显示与 RFA 或 SGA 相似的回报水平，但有些指标差异很大，并且任何一种方法都可能产生最高值。因此，我们强烈建议同时使用 SGA 和 RFA 来估计回报水平，以便在洪水基础设施设计中进行稳健的风险评估。3） 日和次日平均 GEV 形状参数之间的交互作用因气候区域而异，因此是一种从日极端值推断次日降雨量极端值的通用方法（例如，使用 Intensity-Duration-Frequency 曲线）不切实际。我们的研究提供了创新的方法见解，值得在未来的极端降雨研究中考虑。我们的结果不仅使全球当地利益相关者受益，而且还为全球范围内进行的越来越多的允许对流的气候模型实验提供了重要的验证工具。