Implications of New Observing and Data Assimilation Strategies for Moist Convection in the Atmosphere

Abstract

Obtaining a faithful probabilistic depiction of moist convection is challenging because of intrinsic predictability limits of geophysical systems at small scales as well as errors present within models and observations. Understanding each of these challenges is further complicated by assumptions made in data assimilation (DA) techniques. In order to untangle sources of uncertainty in convective weather regimes and ultimately improve state estimates for severe convective storms, we evaluate novel data assimilation and observing strategies using a WRF modeling system adapted and expanded from the National Severe Storm Laboratories Warn on Forecast System. As a core component of our vision for moist-convective state estimation, we evaluate particle filtering (PF) methods as an alternative to DA methods based on linear estimation theory. From a Bayesian perspective, PF’s represent prior and posterior error distributions non-parametrically rather than assuming a Gaussian distribution and can accept any type of likelihood function. This approach is known to reduce bias introduced by data assimilation methods that rely on only the first two statistical moments to perform state estimation. The form of PF used in this research implements a localization strategy that makes it affordable for large geophysical applications, such as convection-allowing weather models. Beyond the methods used for DA, our work also explores fundamental aspects of how mesoscale convective systems (MCS’s) are constrained by observations in analyses and forecasts, with implications for the configuration of observing systems and choices made in parameterization schemes. The first part of this dissertation examines posterior ensembles and their forecasts for selected severe weather events between 2019 — 2020, comparing results from the PF with those from an Ensemble Kalman Filter (EnKF). We find that PF-based DA with 64-member ensembles produces posterior quantities for microphysical variables that are more consistent with model climatology than comparable quantities from an EnKF, which we attribute to a reduction in DA-induced bias. These differences are significant enough to impact the dynamic evolution of convective systems via cold pool strength and propagation. The second phase of this dissertation leverages the same mesoscale analysis and forecasting system to examine observation collection strategies for the evolving Maryland Mesonet. This portion of research introduces a framework for conducting observing system simulation experiments (OSSEs) and establishes limitations on the marginal utility of surface observing system density for EnKF assimilation in moist-convective contexts. The final portion of this dissertation expands upon earlier findings with the PF by evaluating a set of techniques intended to improve state estimation with the PF by addressing deficiencies caused by model error and pre-processing constraints placed on observations, as well as relaxing assumptions of regime-invariance and Gaussianity made for likelihood functions. We find that modifications to key parameters within the cloud microphysics scheme adopted for this research benefit precipitation and hydrometeor estimates produced by the PF—in a manner that is difficult to discern from similarly configured EnKF experiments. Additional benefits are found in observing configurations for reflectivity that remove the typical distinction between reflectivity returns above and below a cut-off threshold, which we again hypothesize to be a design choice that has a great impact on PF-based DA. Lastly, we find that the removal of Gaussian assumptions does not improve the resulting state estimates, but note that potential improvements may be achieved with observation uncertainty quantification strategies that better account for analysis error when forming non-parametric representations of observation errors within the DA framework.