Managing the Scalability of Visual Exploration Using Game Engines to Analyse UHI Scenarios

Large data sets have proven valuable to reveal behaviour phenomena in multidimensional problem spaces. As two-dimensional scientific visualizations are often limited in such scenarios, 4D spatial-temporal visualizations are increasingly raising interest. The huge data size often hinder loading of larger simulation scenes or longer animation sequences. This leads often to performance degrades and thus hinder the broad acceptance of such visualization techniques. We propose a visual pre-processing approach to overcome this challenge. In our research we study how domain knowledge can help to reduce data, guide exploration and the keep scene reloading effort to a minimum. We conduct a feasibility study for the visual pre-processing on heat emission data from cars using cellular automata for simulation and the Unity3D game engine for visualization. The results show a data reduction rate of more than 90% in the model/file size that leads without loss of the visual quality to significantly smoother scene change during urban heat island exploration.