Visual analytics and microclimate analysis: A use case for a visualization tool developed for mobile measurements

Mobile transects are conducted in urban climate research to study spatial variations of atmospheric variables. However, the analysis of the resulting microclimate data sets is a non-trivial and therefore time-consuming task. The data are usually multivariate and afflicted with uncertainties, e.g., due to sensor lags. Furthermore, the spatial context of the measurements has to be taken into account by including surrounding urban features that influence the measurements into the analysis. If several transects have been conducted on a similar route, the individual measurement points also need to be identified between all runs to make the values comparable. To tackle these challenges, a visualization framework tailored to transect data sets has been developed. Its core functionalities include data preprocessing, such as sensor lag correction or time-detrending, a flexible visualization of all observed variables within their spatial context in a three-dimensional environment, and analysis capabilities including clustering and source area computation.

In this presentation, the functionalities of the visualization tool are demonstrated using transect measurements collected at Power Ranch, Gilbert, Arizona. This data set was acquired using a golf cart equipped with sensors measuring surface temperature as well as air temperature and relative humidity at two different heights along a certain route over the course of one year. We show how the framework can be used to gain insight into possible impacts of local urban design on the surrounding microclimate. In particular, transect data and data from surrounding weather stations were integrated to approximate the sensor-specific source area at each measurement point along the transect route. The land use fractions contained within each resulting source area are then used to quantify the relationship between the measured signal and the surrounding urban form. These correlations are analyzed in terms of their temporal variability over the course of a day and a year. Furthermore, data from all transect runs are aggregated and clustered to highlight route sectors of similar temporal value distribution. The challenges during this analysis are described, as well as the solutions provided by the visualization framework.