III. METODOLOGÍA
3.2. METODOLOGÍA EXPERIMENTAL
3.2.2. DESCRIPCIÓN DE LAS FASES DE LA IMPLEMENTACIÓN
spheric data sets
Interactive visual exploration techniques are not commonly used so far by at- mospheric scientists, although sophisticated methods and toolkits have been developed by the visualization community [92, 111, 153]. In this context, Nocke et al. [111], Tominski et al. [153] as well as Ladstädter et al. [92] state that the visualization techniques used by atmopheric researchers are frequently restricted to standard techniques, including 2D diagrams such as time series graphs or scatterplots, or colored 2D maps. These (static or animated) vi- sualizations are typically created using statistical toolkits such as MS Excel, R, or Matlab, or general purpose geographic information systems such as Ar- cGIS [112, 111], an observation I also made during my collaboration with urban microclimate researchers at the Arizona State University. While the resulting plots are easily understandable, they are frequently restricted to summarizing time series or scatterplots without including the spatial context in the case of 2D diagrams, or showing only one variable per image in case of spatial plots [111, 92]. This aggravates the combined visual exploration of the tem- poral development of investigated atmospheric parameters, their multivariate relationships, and their relation to the spatial context [111]. Thus, interesting features or patterns in the data might remain undetected [111, 92].
On the other hand, a large variety of sophisticated interactive visualization techniques for atmospheric data sets have been developed over the years, es- pecially for global and regional scale data sets stemming from observations, remote sensing, and simulations. Application areas range from multivariate exploration of time-varying uni-modal data sets [92, 149] to the integration of observation and simulation data for verification purposes, either qualita- tively [71] or quantitatively [163], to uncertainty visualization in climate and weather models [130]. Ladstädter et al. [92] demonstrate that visual explo- ration techniques can help to provide holistic views of the investigated climate data sets, which facilitates building hypotheses about the data. In their highly interactive tool SimVis, they incorporated techniques including interactive fea- ture selection, brushing and linking, focus and context, and flexible (algebraic) combination of variables. Enabling scientists to find unexpected features or patterns, such techniques can complement more quantitative statistical anal- yses, for which, in many cases, a hypothesis is needed beforehand [92]. Helbig et al. [71] use a variety of mapping techniques to combine several data fields in one view, through which users can navigate either on a desktop computer or in a virtual reality environment. Integrating data sets from both simulation and observation data, they also support the qualitative validation of simula- tion results. Nocke et al. [111] created a library for a suite of visualization techniques that can be used for several analysis purposes related to climate simulation data, using both visualization techniques that are typical for the atmospheric sciences, but also innovative approaches. Tominski et al. [153] visualize global and regional climate networks within their spatial reference frame, encoding important network information using the vertices of the re- sulting graph and enabling data filtering to reduce visual clutter.
While there has been advancement in visualizing larger scale weather and climate data sets, examples for the visualization and analysis of urban (mi- cro)climate data sets are rare, as also stated by Röber et al. [125]. However, several examples can be found for the visualization of urban microclimate sim- ulations. In this context, Röber et al. [125] use Avizo Green to visualize inner city ventilation, and interactions between an individual building and its en- vironment. Heuveline et al. [74] explore the utilization of augmented reality techniques to visualize airflow around building structures. A more integrated
approach was chosen by Wang et al. [162]. Not only limited to urban areas, the authors combine a microscale meteorological model with Google Maps / Google Earth to facilitate the manual initialization of the model in terms of the three-dimensional objects within the domain and the final visualization of the simulation results, using standard mapping techniques. The authors also demonstrate how Google Maps / Earth can be used to map observation data sets.
Even more rare are visualizations of observation data sets in an urban cli- mate context. There are some examples for the visualization of air quality data sets, which are obviously not necessarily bound to an urban context. For example, Qu et al. [120] developed a visual analytics system for station- ary air pollution measurements from Hong Kong. In their tool, they combine several visualization techniques, including a circular pixel bar chart, a par- allel coordinates plot with an s-shaped axis to encode wind direction, and a complete weighted graph showing pairwise correlations between measured at- tributes. Their system, however, does not include the spatial context of the measurement stations. Also dedicated to air quality data, Liao et al. [97] implemented a web-based air quality analysis system for stationary measure- ments that comprises a map view with pie-charts encoding the distribution of certain air quality indices over a selected time interval, a parallel coordinates plot, and a time-series plot.
Recently, with the advent of mobile sensing and data crowdsourcing, a va- riety of web-based visualization tools have been developed that either attempt to inform the general public about the collected data sets in their city (see, e.g., [168]), or that are rather designed as a data storage, management, and display systems [87]. While providing interactive navigation of the data sets, these systems are usually limited to standard mapping techniques and do not provide multivariate exploration facilities.