We identify 5-day precipitation events over the contiguous U.S. using Tempest Extremes (TE). TE uses object-based techniques to identify and track 3-dimensional space-time objects, which we analyze here as events based on 5-day precipitation. This event-based analysis complements analyses that yield maps and other statistics of changes in 5-day precipitation.

We analyze precipitation simulated by the regional climate models RegCM4 and WRF, which were driven by either the ERA-Interim reanalysis or GCM simulations. Simulations used 12-km grid spacing. To evaluate the quality of the simulations, we use NARR and PRISM observational data. Analysis time periods are chosen for contemporary (1981 - 2005 and 1991 - 2010) and RCP 8.5 scenario (2075 - 2099) time ranges. Daily precipitation is accumulated over 5-day time periods and input into TE to identify sequences of 5-day precipitation events. Extreme precipitation occurs in a 2-5 day precipitation event.

We use six analysis regions covering the contiguous U.S. that are chosen to evaluate model performance in contemporary climate and to analyze regional projected changes in sequences of 5-day events. Contemporary simulations tend to fall near the observational data for all regions. Scenario results indicate a future with increasing frequency, duration and intensity of sequences of 5-day precipitation events for most regions of the U.S. studied here.