Atmospheric rivers (ARs) are long and narrow river-like features in the lower troposphere that carry most of the atmospheric water vapor fluxes from the tropics to the midlatitudes. Because of the large amount of moisture transported by these storms, they can cause heavy rainfall and major flooding events, such as the Midwest floods of 1993 and 2008. The overarching theme of this thesis is to understand the impacts of ARs on extreme precipitation and floods over the central United States.
First, to improve the understanding of the mechanisms leading to the development of these storms, three ARs that happened during the summer of 2013 are studied in detail. The work provided insight into the synoptic conditions associated with these storms. Moreover, I found that the source of moisture for ARs over the central United States can be located both in the tropics and subtropics, and evaporation over land can also add water vapor along the AR trajectory.
To understand the characteristics of precipitation during these storms, I focused on a 12-year period and used different high spatial and temporal resolution remote sensing-based precipitation products. These analyses showed that most of the AR-related precipitation is located in a narrow region (approximately 150km) within the area where the strongest moisture transport occurs.
The analysis of multiple long-term atmospheric reanalysis products has led to the development of the climatology of ARs over the central United States. This climatology is used to understand the AR characteristics, their long-term impacts on annual precipitation, precipitation extremes, and flooding over the central United States. AR characteristics (e.g., frequency, duration) are generally robust across the different reanalysis products. These storms exhibit a marked seasonality, with the largest activity in winter (more than ten ARs per season on average), and the lowest in summer (less than two ARs per season on average). Overall, ARs generally last less than three days, but exceptionally persistent ARs (more than six days) are also observed. In terms of their impacts on precipitation, AR-related precipitation is able to explain a large portion of the year-to-year variations in the total annual precipitation over the central United States. Moreover, 40% of the top 1% daily precipitation extremes are associated with ARs, and more than 70% of the annual instantaneous peak discharges and peaks-over-threshold floods are associated with these storms, in particular during winter and spring.
The relationship between the frequency of ARs and three prominent large-scale atmospheric modes [Pacific-North American (PNA) teleconnection, Artic Oscillation (AO), and North Atlantic Oscillation (NAO)] is investigated, and the results are used to statistically model the frequency of ARs at the seasonal scale. PNA and AO indices play a significant role in the winter season, when the AR frequency is the highest. Building on these insights, different spatio-temporal Bayesian hierarchical models are developed to describe the frequency of winter heavy precipitation events based on ARs and the large-scale atmospheric modes. The results suggest that over much of the central United States, PNA and AO can be helpful in describing the frequency of ARs in winter, which in turn can be useful to characterize the frequency of heavy rainfall events over the central United States.
Because of the large impacts that these storms have, their short-term predictability is examined by using outputs from five numerical weather prediction (NWP) models with a lead-time up to 15 days. While there are differences among the five NWP models, the results show that the skill in forecasting the occurrence and location of ARs over the central United States decreases with increasing lead time, and the models have positive skills up to the seven-day lead time.