On the spatial statistical analysis of urban user-generated geodata
- Tuesday 29 March 2022
- 15:00 - 16:30 (BST)
- Online (via Zoom) GET DIRECTIONS
Dr René Westerholt, Juniorprofessor for Spatial Modelling at the Department of Spatial Planning, TU Dortmund University (Germany), will present at this UBDC research seminar.
Following Dr Westerholt's presentation, there will be the opportunity for attendees to pose questions.
User-generated geodata reflecting parts of daily life have grown rapidly over the last decade and a half. With the entry of smartphones into the market and in combination with the widespread use of mobile networks and positioning technologies, large geospatial data sets have emerged in a relatively short time. Among the best-known examples are social media feeds, but we also routinely leave traces about aspects of our daily geographic lifestyles when using loyalty cards, through mobile tracking apps such as when exercising, and when using dockless bikes or e-scooters. Many of these non-scientific datasets are not collected according to standardised protocols. On Twitter and Instagram, for example, users are (more or less) free to decide what content to share and what images to upload. These decisions also have a strong geographical dimension. Where we spend time, what topics we tweet about in the immediate vicinity of our whereabouts, and how the urban environment leads us to use social media in certain ways contributes to a high degree of spatiotemporal heterogeneity in corresponding datasets. This is a challenge for spatial-statistical methods that are often tailored to more traditional scientific datasets. This talk addresses the interface of partial representations of everyday life in data and spatial analysis methodology.
In this talk Dr Westerholt will address two aspects of the analysis of user-generated geographic data from urban environments. The first aspect is to examine how urban environments influence the production of user-generated data. An empirical study conducted in the city of Heidelberg (Germany) is presented (Westerholt et al. 2020). This study utilised gamification, spatial statistics, and OpenStreetMap data collection to investigate how people moving around a familiar urban environment change their geographic behaviour in response to spatially structured incentives that are unknown to the participants beforehand. The results show that participants who are confronted with a spatially structured incentive system are more likely to stay in high-scoring areas, have a longer play time, walk longer distances, and are more willing to take detours. The results can be applied to the case of spatially distributed incentives in cities and thus represent an important insight for a better understanding of many user-generated geospatial data. The second aspect of the talk will be methodological. He will report on a novel method for dealing with complex spatial heterogeneities in social media and related datasets (Westerholt 2021). The method is based on Moran eigenvector filtering. However, instead of controlling for spatial effects as in the original method, the presented approach is based on accentuating certain types of spatial structures while attenuating others that would act as confounders. The results show that this pre-filtering of data allows for better interpretability of social media datasets. The two perspectives presented thus look at the overall topic from both the data and the methodological side.
Dr René Westerholt is Juniorprofessor for Spatial Modelling at the Department of Spatial Planning, TU Dortmund University (Germany). Prior to his current post, René has held a position as Assistant Professor for Urban Analytics at the University of Warwick (UK). He completed his PhD in Geography at Heidelberg University (Germany) in 2018, after graduating from Osnabrück University (Germany) with both a Bachelor's and Master's degree in Geoinformatics. His primary research interests include spatial analysis methodology, user-generated data, and place-based geographical information.
Registration for this online event is available via Eventbrite and full details and instructions for joining will be circulated post-registration.