Spatial analysis of cycling risk in Helsinki using crowdsourced risk exposure data

Increasing the modal share of cycling in urban transport is a way to improve public health and for cities to meet their sustainability objectives and urbanization challenges. However, safety concerns remain a significant barrier to the widespread adoption of cycling. Identifying areas where cyclists are more prone to accidents is crucial for improving cycling safety, but traditional cycling accident data often lacks normalization for risk exposure (i.e. volume of cycle traffic), which can obscure high-risk areas. In this thesis, cycling risk index values were calculated in the City of Helsinki by combining official and surveyed accident data and Strava Metro risk exposure data on a 500 m x 500 m grid. Spatial heterogeneity of the risk data was assessed using global Moran’s I statistics, followed by local indicators of spatial association statistics to locate significant high-risk clusters. Subsequently, both global and local regression models were constructed to identify variables affecting cycling risk, using infrastructure data and traffic safety survey responses. The results reveal significant spatial clustering of cycling risk in Helsinki and show that areas with the highest absolute risk differ from those with the highest exposure-normalized risk. When normalized with exposure data, risk levels decreased in downtown Helsinki but increased in northern and eastern parts of the city. At a local level, difficult or unsafe crossings and traffic exposure were the most significant explanatory variables for overall risk before exposure normalisation, while combined length of shared roads and cycleway length were the most significant variables after normalization. The explanatory variables effectively modelled absolute cycling risk but explained exposure-normalized risk only moderately. The results suggest that improving bikeability in downtown Helsinki can affect the most cyclists, and that improvements in the northern and eastern parts of the city may yield outsized positive results for overall cycling safety. To better model exposure-normalized risk in the future, a broader set of explanatory variables should be explored, while verifying the local correlation of Strava cycling trip count data with real cyclist count data could further refine risk data normalization.

Grantor dc:publisher

Helsingin yliopisto

Year dc:date.issued

2024

Author dc:creator

• Alkiomaa, Jonne Jukka Aleksi

• pyöräily
• riskit
• liikenneturvallisuus
• joukkoistettu data
• Strava

Statement dc:rights

• In Copyright 1.0

Language dc:language.iso

eng

Handle dc:identifier.uri

http://hdl.handle.net/10138/587928

OAI identifier oai:identifier

oai:helda.helsinki.fi:10138/587928