Abstract
Climate‐driven changes in extreme precipitation and floods severely impact societies and ecosystems, yet regional spatial and temporal nuances remain poorly understood, including in the Hawaiian Islands. This study utilised stationary and non‐stationary Generalised Extreme Value (GEV) distributions, annual maximum series (AMS), and circular analysis to examine extreme hourly rainfall across 117 gauges on the five most populated Hawaiian Islands (Kauaʻi, Oʻahu, Maui, Molokaʻi, and the Big Island) from 1966 to 2020. Stationary GEV results show that Kauaʻi, Oʻahu, and the Big Island exhibited higher return levels (the event magnitude at a given return period) than the other islands, with windward sides consistently exceeding leeward sides. The 1990–2020 AMS revealed that most places showed no trend in annual maximum hourly rainfall, while some significant positive and negative trends were identified across the islands. Non‐stationary GEV results showed that extreme rainfall is spatially heterogeneous—Oʻahu, Molokaʻi, and Maui are dominated by decreasing return‐level trends, while the Big Island and Kaua‘i show predominantly increasing trends, particularly for rare 20‐ and 100‐year events. Notably, Maui exhibited consistent island‐wide decreases regardless of frequency or duration. Circular analysis identified two primary, persistent seasonal peaks for extreme events: October—December and February—March. These findings highlight a critical duration dependence: short‐duration infrequent extremes are intensifying across most islands, even where frequent events are decreasing, posing a significant threat to emergency responses. As the first study investigating hourly extreme rainfall in Hawai‘i, this work provides a vital understanding of non‐stationarity on tropical islands to mitigate future flooding threats.