Visual Analytics of Spatio-Temporal Event Predictions : Investigating Causes for Urban Heat Islands

Abstract

Due to ongoing urbanization and industrialization processes, city planners are facing an increasing number of challenges. One such challenge is the phenomenon of locally increased temperature { the urban heat island effect. This effect is a well-known phenomenon in both the domain of city planning and meteorology. The appearance of urban heat islands is not only influenced by many apparent variables, such as weather, vegetation, and surface characteristics, but also by rather inconspicuous parameters, like industry, transportation infrastructure, population density, energy management, and air pollution. Because of this large number of influencing factors, analyzing the causes for this effect is a complex task. To improve urban climate and energy management, innovative applications that provide support for decision-making tasks are needed. We propose a visual analytics system which enables expert users to explore temperature conditions of a city area. At its core, the system uses a random forest classification model, trained on heterogeneous, nationwide collected data, which facilitates interactive parameter steering of spatial and meteorological features. Users can explore the influence of several variables on urban heat island forecast events. The system provides spatio-temporal event predictions that offer insights about future conditions and the effect of various variables on the formation of urban heat islands. To present connections between diverse urban heat island properties and forecast events, we compose different mature, interactive visualizations. Through several use cases, we demonstrate that our system allows users to focus on relevant features while getting a solid overview of the urban heat island situation in a specific area of interest. The integration of a combination of selected reasonable visualizations with a prediction model based on ensemble learning offers a viable solution for an adequate analysis of the urban heat island effect that can be enhanced by further functionality in the future.