Detection and Visual Analysis of 3-D Atmospheric Fronts – from Individual Cases to Ensembles

Abstract

Atmospheric fronts are a widely used conceptual model in meteorology, most encountered as two-dimensional (2-D) front lines on surface analysis charts. The three-dimensional (3-D) dynamical structure of fronts has been studied in the literature by means of “standard” 2-D maps and cross-sections and is commonly sketched in 3-D illustrations of idealised weather systems in atmospheric science textbooks. However, only recently has the feasibility of the objective detection and visual analysis of 3-D frontal structures and their dynamics within numerical weather prediction (NWP) data been proposed, and such approaches are not yet widely known in the atmospheric science community. In this thesis, I investigate the benefit of objective 3-D front detection for case studies of extra-tropical cyclones and for comparison of frontal structures between different NWP models. I build on a recent gradient-based detection approach, combined with modern 3-D interactive visual analysis techniques, and adapt it to handle data from state-of-the-art NWP models including those run at convection-permitting kilometre-scale resolution. The parameters of the detection method (including data smoothing and threshold parameters) are evaluated to yield physically meaningful structures. I illustrate the benefit of the method by presenting two case studies of frontal dynamics within mid-latitude cyclones. Examples include joint interactive visual analysis of 3-D fronts and Warm Conveyor Belt (WCB) trajectories, as well as identification of the 3-D frontal structures characterising the different stages of a Shapiro–Keyser cyclogenesis event. The 3-D frontal structures show agreement with 2-D fronts from surface analysis charts and augment the surface charts by providing additional pertinent information in the vertical dimension. A second application illustrates the relation between convection and 3-D cold-front structure by comparing data from simulations with parametrised and explicit convection. Investigation into “secondary fronts” that are commonly shown in UK Met Office surface analysis charts shows that for this event the secondary front is not a temperature-dominated but a humidity-dominated feature. Building on this 3-D front detection approach, I extend the detection of individual 3-D fronts towards front-feature-based time series analysis and ensemble clustering. Ensemble simulations have become a standard in NWP. However, ensemble simulations generate large amounts of data and their analysis remains a challenge. In this thesis, I develop a manual and automated front-tracking algorithm based on geometric and physical characteristics to derive time series of frontal attributes from a selected cyclone system. These frontal attributes characterise the 3-D front by one-dimensional (1-D) physical properties, such as the average slope of the 3-D frontal structure. By tracking a selected front over successive time steps, time series of frontal attributes are derived to provide a compact view of the development of frontal attributes. To order and cluster ensemble simulations according to frontal attributes, a selected front automatically tracked across all ensemble members and front attribute time series are derived for each member. These feature time series are then ordered and clustered using time series distance measures in combination with k-means clustering, resulting in distinct clusters that represent different patterns of frontal evolution across ensemble members. Integrated into the 3-D interactive visual analysis framework Met.3D, my approach allows a comprehensive analysis of the spatio-temporal evolution of 3-D atmospheric fronts and thus contributes to the challenge of rapid analysis of large ensemble weather forecasts, as well as having great potential for operational weather forecasting.