Application of Reanalysis Datasets in Hydrologic Modeling for Dam Safety: A Novel Approach to Storm Transpositioning

Hydrologic hazard studies range in complexity from a preliminary assessment based on the probable maximum precipitation to an in-depth, risk-based analysis using precipitation-frequency estimates, storm spatial and temporal patterns, and stochastic hydrologic modeling. One of the limiting factors for performing the complex studies is data availability, thus there is great value in the use of reanalysis products when studying facilities in data-sparse regions. Additionally, precipitation-frequency estimates from the Hydrometeorological Design Studies Center (e.g., NOAA Atlas 14) are not available in all locations, may contain mixed distributions of storm types, and do not provide estimates of imporance to dam safety studies (i.e., > 1000-year return period).

Storm typing is perhaps one of the greatest contributions to hydrologic science in the realm of dam safety. Frequently, storm typing is done via manual classification of individual storm events, though more recently automated algorithms have been developed to identify synoptic-, meso-, and local-scale events. Precipitation-frequency estimates can then be derived for specific storm types using historical analyses of precipitation. Here, we compare the results from several reanalysis products: (i) Livneh et al (2015) daily precipitation; (ii) CHIRPS satellite-estimated precipitation; (iii) DAYMET daily precipitation; (iv) PRISM daily precipitation; and (v) Climate Forecast System reanalysis. We provide context for the application of the different storm types through case studies in the southeast United States and intermountain West. Additional hydrometeorological variables that are important to stochastic modeling are briefly discussed, followed by current and potential future applications of these data.