Investigating Environmental Influences Behind Summertime Nocturnal Extreme Rainfall Events in the United States Northern Great Plains

Abstract

Extreme rainfall events (EREs) in the United States are known to cause significant and deadly flash flooding events. Despite advancements in extreme rainfall forecasting in recent decades, owing to the increase in numerical weather prediction model skill and resolution, extreme precipitation forecasting remains a significant challenge and motivates NOAA’s Precipitation Prediction Grand Challenge initiative. One particularly difficult forecasting problem involves summertime nocturnal convection, which has been shown to produce a high percentage of summertime (June-August) EREs. This nighttime occurrence poses additional complications when messaging warnings and potential impacts to at-risk communities. One of the primary forecast challenges is understanding how environmental variables control the predictability of nocturnal EREs and their severity, as the events are strongly influenced by interactions between the synoptic-scale and mesoscale processes that can be difficult to accurately represent and predict. To further investigate this, this study investigates the large-scale environmental conditions that favor summertime nocturnal EREs in the Northern Great Plains of the United States using self-organizing maps (SOMs). SOMs are used to cluster EREs of varying severity, i.e., extreme and less extreme, using large-scale synoptic meteorological variables to understand how environments vary across extreme rainfall events and inform the predictability of extreme rainfall. First, EREs are identified using a high-resolution database, that uses NCEP Stage IV 6-hour quantitative precipitation estimates (QPE) and average recurrence interval (ARI) thresholds from NOAA Atlas 14. Within this database, events are defined by 6-hour Stage IV QPE accumulations exceeding the 10-year and 100-year ARI thresholds. A SOM is then trained on 500-mb geopotential height anomalies associated with the resulting more-extreme (100-year ARI exceedance) and less-extreme (10-year ARI exceedance) events. Following the SOM training, events assigned to each SOM pattern (node) are then separated into more-extreme and less-extreme events and used to generate composites of additional atmospheric variables, such as total column water vapor (TCWV), 850-mb winds, 850-mb temperature advection, vertically integrated water vapor flux (VIWVF), and VIWVF convergence. By doing this, we can identify atmospheric patterns associated with the different SOM nodes and allow for the investigation of whether certain patterns are associated with more-extreme rainfall events compared to less-extreme events. This thesis will compare the clustered environmental conditions between more-extreme and less-extreme rainfall events and detail a few of the node regimes resulting from the trained SOM. Results from the SOM framework are also compared to full-domain bulk composites of all more-extreme and less-extreme event environments to evaluate the added value of the SOM approach in resolving distinct synoptic environments compared to traditional composite approaches. In addition, the large-scale environmental evolution preceding the events is analyzed. Results show that both the bulk and node regime composites reveal a common set of environmental features associated with summertime nocturnal extreme rainfall events in the Northern Great Plains that are known to be favorable for extreme rainfall. The bulk composites also identify some key environmental differences between more-extreme and less-extreme events that were consistently present across the evaluated SOM node regimes. However, the bulk composites do not capture the full range of environmental differences between more-extreme and less-extreme events. The implications of these new findings will be shared to advance our collective knowledge of extreme rainfall events and how SOMs can be useful tool in identifying common patterns.