Unravelling the relationship between extreme precipitation duration and wet-day frequency at global scale.

Understanding the spatiotemporal interplay between the duration of extreme precipitation events and wet-day frequency (i.e., the relative number of days that exceed a minimal daily precipitation threshold) is critical for predicting hydrological risks in a changing climate. Here, we analyze ERA5 reanalysis and employ percentile thresholds to define extreme precipitation events in daily scale, investigating global patterns of the duration of extreme precipitation events across areas classified by wet-day frequency. Our analysis reveals that increased wet-day frequencies are generally associated with a steeper decay of duration distribution and lower mean durations, except in areas with a frequency exceeding 0.8, where the mean duration slightly increases. This overall negative correlation mainly stems from limitations in atmospheric moisture availability, while the anomalous increase in high wet-day frequency areas is attributed to their concentration in low-latitude regions, where extreme precipitation events have longer durations. Further analysis reveals a distinct latitudinal dichotomy under this phenomenon: mid-latitude regions (30°-60°) show a strong linear negative correlation (r = - 0.77), governed by synoptic-scale dynamics and wind-modulated precipitation systems. In contrast, low-latitude regions (0°-30°) demonstrate a nonlinear (initially increasing then decreasing) relationship with a significant correlation (r = 0.63) between these durations and amplified annual precipitation cycles, typically linked to seasonal rainfall clustering mechanisms. Threshold sensitivity tests confirm the robustness of these results, except in low latitudes and at the 99th percentile threshold, due to limited data sample. The MSWEP and PERSIANN datasets are also tested with similar results found. These finding can enhance prediction accuracy for extreme precipitation events and inform regional flood risk management.