New Data for Old Storms: Can New, Convection-Allowing Ensemble Simulations of Historic Storms Help Minimize Present-Day Flood Risk?

Accurate estimation of the potential “upper limit” for extreme precipitation is critical for dam safety and water resources management. While dams are necessary to managing water, dam failures present significant risks to life and property: for example, 2209 fatalities occurred in the Johnstown, PA 1889 dam failure, and the 2017 Oroville Dam spillway damage has resulted in costs upwards of $870 million. In order to estimate the “upper limit” of precipitation for a given location, the concept of “probable maximum precipitation” (PMP) is often used. It is well-known that existing PMP methods could benefit from updates to both storm data and storm analysis methods used. However, the reality that extreme events are by definition rare means that old storms with limited observational data are often used to define the historical upper bound of precipitation for a given region. Observations of old storms are generally notably limited in spatial and temporal coverage, and often incomplete or of dubious quality. This directly affects the quality of the resulting values used in dam safety evaluations and design.

As computational power and numerical weather forecasting models have improved over the past decade, it has become possible to simulate “PMP-type storms” explicitly. Convection-allowing ensembles in particular offer considerable promise to the weather forecasting community, and such an approach is tested here to evaluate whether ensemble simulations of four historical extreme precipitation events important for PMP development (Ward District, CO May/June, 1894; Rattlesnake, Idaho November, 1909; Savageton, WY September, 1924; Penrose, CO June, 1921) could benefit the existing PMP estimation process. Specifically, we use multiple members of the 20th Century Reanalysis dataset (version 2c) to initialize 4-km grid-spacing WRF simulations which are then considered for their value in acting as a potential surrogate dataset to replace the current storm analysis (i.e., augmenting observations and providing high-resolution spatial and temporal storm patterns). In addition, model results may be able to expose erroneous historical observations used in existing PMP estimates. Examples of each type of application will be provided.