| Description |
The synoptic and mesoscale connections between Low-Level Jets (LLJ) and precipitation in the United States Great Plains region (GPR) are explored to better understand their relationship, especially as it relates to precipitation extremes. The accurate representation of these features and their associations through high-resolution modeling and advanced data assimilation methods are also examined. This is accomplished through three distinct but related topics, including 1.) The variations in the climatology of the LLJ structure during wet and dry periods, 2.) The three-way interactions between the LLJ, Mesoscale Convective Systems (MCS), and Cold Pools (CP), and 3.) The impacts of non-Gaussian ensemble distributions on data assimilation. The first part of the study explores the factors leading to drought and flood events in the GPR. It was found that the LLJ provides a consistent source of moisture advection and ascent in its exit region, however, it is the synoptic environment that ultimately dictates whether an LLJ is productive at precipitation formation. Accordingly, the LLJ provides regular convergence and lift, which is either realized or squandered according to the environment already in place as a result of the synoptic anomalies present. The second part of this study uses WRF simulations and ensemble sensitivity analysis to analyze interactions between the LLJ and MCSs, and CPs for a case on the 23-24 of May 2011. These components are sensitive to each other, which leads to divergences in simulated solutions according to how well the various interactions are handled in terms of timing, location, and strength. It is demonstrated that the LLJ is stronger as a result of moist convective processes. The third part of the dissertation looks at the effects of non-Gaussian data assimilation for two cases. Non-Gaussian ensemble distributions are widespread and regular. Non-Gaussianity is associated explicitly with convection and modestly associated with other features such as dry lines. Filters that are designed for non-Gaussian data assimilation provide value, particularly at higher resolutions and near non-Gaussian features. For filters that assume a non-Gaussian prior distribution, the corresponding posterior distribution is degraded. However, such filters can perform well when the posterior spread is minimalized. |
