Visual Analysis of Multi-run Spatio-temporal Simulation Data
Alexey Fofonov · 2016
Multi-run simulations are widely used to investigate how the simulated processes evolve depending on varying initial conditions in different fields of sciences, such as astrophysics, ballistics, and oceanography. This kind of experiments is typically executed using high-performance computers and generates a large amount of data to analyze. The main goals of this work are an efficient visualization of multi-run time-varying spatial data, pairwise and ensemble-wise comparison of simulations, as well as simulation feature detection and investigation, and exploration and classification of simulation results. Also, this work aims generalization of developing approaches to be applicable for data of any spatial configuration and of a multi-field nature. This work introduces a number of approaches and tools, which can significantly improve an analysis of multi-run time-varying spatial data on all data aggregation levels from a field distribution of an single time frame up to overview of the whole simulations ensemble. To investigate a single-field spatial data a new isocontour based similarity measure is introduced. For an interactive visualization and for detailed comparative analysis of multi-field spatial data algorithms for a splat-based progressive rendering of continuous scatterplots are developed. For a comparative visualization of multi-run and multi-field spatio-temporal simulation data a new visualization technique, called multi-run similarity plots, is introduced. For an overview of a multi-run and multi-field spatio-temporal simulation data on a highest aggregation levels two histogram based visualizations are proposed. Finally, a scheme of a top-down analysis of a multi-run spatio-temporal simulation data is introduced. In this work an application on the real world data sets, results of a user study and evaluation by the domain experts, as well as detailed discussions about scalability of the methods, their efficiency and restrictions, are presented.