Detecting statistically significant annual changes in cycling volumes based on crowdsourced GPS-data

It is essential for local authorities to conduct retrospective analyses of the impacts of infrastructure measures and barriers on cycling. Typically, such evaluations can only be performed at specific locations through permanent counting stations, or with the help of on site surveys. However, to gain a comprehensive overview across an entire network, extensive GPS track data proves invaluable. A bivariate correlation analysis is employed to examine the linear relationship between GPS tracks and data from permanent counting stations. To facilitate comparison, the annual GPS tracks are aggregated into hexagonal grids, and the annual changes are quantified using various approaches. A spatial correlation analysis is then conducted for each approach using Moran’s I, identifying clusters of significant changes. These results are compared and validated against known infrastructure measures and barriers, using a German City (Dresden) as a case study. The analysis reveals a moderate to strong linear correlation between GPS data and permanent counting station data. Infrastructure measures and barriers are identifiable across all methods of analyzing annual changes, and, in certain instances, shifts in cyclist routes to or from alternative nearby roads are also detected. Given that certain clusters of significant change cannot be directly attributed to specific infrastructure measures or barriers, it is crucial to incorporate multiple approaches to analyze annual change. This methodology helps mitigate the risk of false inferences, enhancing the robustness and reliability of the findings.