Estimation of long-duration maximum precipitation during a winter season for large basins dominated by atmospheric rivers using a numerical weather model

Abstract The Probable Maximum Precipitation (PMP) estimation for long durations during winter and spring seasons is important to develop the Probable Maximum Flood for snowmelt-driven regions since extreme floods are often characterized by snow-accumulation and snowmelt processes rather than by a single rainstorm event. Although several studies have estimated the PMP for a single storm duration, little attention has been given to the PMP estimation for long durations on the order of several months. This study proposes a new framework using a numerical weather model (NWM) to estimate the long-duration Maximum Precipitation (MP) during the winter season, which is the first part of a two-part effort to develop the PMP during the winter and spring seasons. As a demonstrative case, we estimate the MP for the 6-month winter period (October to March) for the drainage areas of Bonneville Dam (621,600 km2) and Libby Dam (23,270 km2) in the Columbia River Basin dominated by atmospheric rivers (ARs). In the proposed framework, the historical AR events are identified based on the integrated water vapor transport thresholds used in the AR category scale. The precipitation depths during the identified AR events are then maximized by simultaneously optimizing the AR position and its atmospheric moisture. Finally, the design precipitation sequence is formed by substituting each historical AR event with the corresponding maximized AR event, acting as the basis of long-duration MP. As a result, the maximum 6-month winter period accumulated basin-average precipitation depths: long-duration MP, for the drainage areas of Bonneville Dam and Libby Dam, are estimated to be 961.0 mm and 1,101.7 mm, respectively. To the authors’ knowledge, this is the first study estimating the MP for long durations on the order of several months and for very large basins (above 100,000 km2) by using the NWM-based approach.