Visual Analytics of Volunteered Geographic Information: Detection and Investigation of Urban Heat Islands

Abstract

Urban heat islands are local areas where the temperature is much higher than in the vicinity and are a modern phenomenon that occurs mainly in highly developed areas, such as large cities. This effect has a negative impact on energy management in buildings and also has a direct impact on human health, especially for elderly people. With the advent of volunteered geographic information from private weather station networks, more high resolution data is now available within cities to better analyze this effect. However, such data sets are large and have heterogeneous characteristics requiring visualinteractive applications to support further analysis. We use machine learning methods to predict urban heat islands occurrences and utilize temporal and spatio-temporal visualizations to contextualize the emergence of urban heat islands to comprehend the influencing causes and their effects. Subsequently, we demonstrate the analysis capabilities of our application by presenting two use cases.

A progressive development of a visual analysis interface of climate-related VGI

“…In another study, Seebacher et al (2019) propose an adaptive workspace consisting of five views that facilitate prediction and visual analysis of urban heat islands based on volunteered data from a private weather station network. Tenney et al (2019) describe a crowd sensing system that captures geospatial social media topics and visualizes the results in multiple views.…”

Summer Nights in Berlin, Germany: Modeling Air Temperature Spatially With Remote Sensing, Crowdsourced Weather Data, and Machine Learning

“…Higher density T air monitoring has been achieved by high-resolution measurement networks [13] and vehicle-mounted temperature sensors (traverses) [14]- [16], but these approaches are costly to implement and maintain. Recently, crowdsourced atmospheric data has emerged as a source for low-cost, spatially dense, and long-term measurements of T air in urban environments [12], [17]- [20]. Crowdsourced T air data from "Netatmo" citizen weather stations (CWS), which belong to the realm of the "Internet of things" and volunteered geographic information, has proved to be reliable to study urban T air with its worldwide distribution, easy and free access, and high spatial density [17]- [19], [21]- [24].…”

“…Moreover, in contrast to parametric approaches, ML algorithms can handle a large set of available predictor variables and provide mechanisms for estimating variable importance [35]. However, very few studies so far have fused crowdsourced T air data and remote sensing data to map T air [12], [20]. Shandas et al [15] used Random Forest (RF) regression using Sentinel-2 and vehicle-mounted temperature sensors to model T air spatially.…”

“…Shandas et al [15] used Random Forest (RF) regression using Sentinel-2 and vehicle-mounted temperature sensors to model T air spatially. Seebacher et al [20] predicted UHI occurrences using private weather station network data, land-use and land-cover data from OpenStreetMap, meteorological data, and RF classification. They recommended the addition of other influencing variables in prediction models of UHIs, including the morphology of urban surfaces [20].…”

“…Seebacher et al [20] predicted UHI occurrences using private weather station network data, land-use and land-cover data from OpenStreetMap, meteorological data, and RF classification. They recommended the addition of other influencing variables in prediction models of UHIs, including the morphology of urban surfaces [20]. Notably, Venter et al [12] combined Sentinel, Landsat, and LiDAR data with crowdsourced Netatmo data to model annual mean, daily maximum and minimum T air using RF regression in Oslo, Norway.…”