We present RainyDay, an open-source stochastic storm transposition (SST) framework for generating large numbers of realistic rainfall “scenarios.” SST “lengthens” the rainfall record by temporal resampling and geospatial transposition of observed storms within a user-defined region to extract space-time information from regional rainfall data. Relatively short (10-15 year) records of ground-based weather radar are sufficient to estimate rainfall and thus to simulate flood events with much longer recurrence intervals including 100-year and 500-year events. We describe the SST methodology as implemented in RainyDay, and compare rainfall IDF results from RainyDay to conventional estimates from NOAA Atlas 14. Then, we demonstrate some of the flood frequency analysis features that are possible when RainyDay is integrated with a hydrologic model, including robust estimation of flood hazards in a changing watershed. The U.S. Bureau of Reclamation (USBR) is supporting the development of a web-based version of Rainyday, a “beta” version of which is available at http://her.cee.wisc.edu/projects/rainyday/.
Challenges remain for applying SST in flood frequency efforts across the Western US, the primary domain of interest to USBR. These include questions surrounding storm transposition in mountainous terrain, estimation of very extreme rainfall and flood frequencies (>1,000 return periods), and how to properly initialize hydrologic models for flood frequency analyses in snowmelt-dominated watersheds and watersheds with extreme variations in elevation, topography, and land cover. These issues will be explored as the research project continues.