Cycling risks communication using Web-GIS:

different visualisation techniques and their effectiveness

Alessandro Cristofori

 

Introduction

The level of cycling in the UK is one of the lowest among the most developed countries (Peucher,2008). Despite this, cycling has always been encouraged by UK government institutions; for example, the vision of the Cycling Action Plan for Scotland (CAPS) is to sensibly increase the number of journeys taken by bicycle by 2020. Cycling reduces motor vehicles traffic, transport emissions and greenhouse gases (de Hartog et al., 2010), because of environmental benefits cycling can be considered a sustainable transport. However, people who cycle are sensibly more likelier to incur accidents (de Hartog et al., 2010) and risk is a prominent deterrent to cycling (Parkin et al., 2007; Winters et al., 2011), hence a hamper to the diffusion of this transport as an alternative to motor-vehicles.

With this dissertation we wanted to evaluate the effectiveness of GIS in dealing with collisions data, both to communicate accidents risks and to generate awareness in actual and potential cyclists; the assessment considered different visualisations techniques and a route planner based on a safe path choice model, the study area is the City of Edinburgh.

 

Methodology

 

Database

Cycling accidents and road network were loaded into a PostGIS database. Accidents data were sourced from Department from Transport, road network from OpenStreetMap.

 

Route planner

Information contained in both accidents and roads dataset were explored to determine for each accident and road segment its degree of risk. Risk elements considered were type of road, type of intersection, number of accidents per road segment, number of casualties, casualties severity, risk related to cycling manoeuvre and risk related to colliding vehicle. Risk elements were summarised in a comprehensive risk score for each road segment, spatial analyses and geoprocessing were performed using ArcMap and QGIS. Weighted risk scores and segments lenghts were used to identify edges costs with Multicriteria evaluation. PgRouting with Dijkstra's algorithm was used as routing engine to find the safest path given a start and an end point.

Casualty severity for manouvres and intersection types Distribution of cycling accidents risk Edinburgh City centre Riskiest road segments for cyclists in Edinburgh City centre Safest path example with PgRouting plugin for QGIS
 

 

Web-GIS

The user can explore the map and choose among diffrent types of visualisation, accidents points, dangerous streets and a customisable heatmap, picking two points on the map the planner returns the safest and quietest route. The interface and interaction are provided by JavaScript and OpenLayers, maps are managed by Geoserver and queries to DB are made in SQL and PL/Pgsql.

Web-GIS system diagram Cycling accidents represented as points Representation of roads cycling risk Heatmap showing concentration of buses/bicycles collisions Safest (green) and quietest (red) paths
 

 

Test

Both the model and the Web-GIS were tested. 50 random paths were generated and the risk optimised algogrithm was compared to its non-optimised counterpart, for each attempt risk reduction and length were measured with the aim to assess not only the effectiveness in reducing cycling risks but also how the lenght/time criterion was taken into account. For the Web service test, 20 people between regular and non-regular cyclists were asked to complete a survey after having used the Web-GIS and having compared it with CycleStreets.net, questions wanted to assess visualisation techniques, model effectiveness and site usability.

 

Results

If we consider the first test and both criteria together the accident aware algorithm performs better in 26 out of 50 attempts, in we consider one criterion at the time we can actually assess the tendency of the accidents aware algorithm to prefere shorter and less safe routes instead of longer but safer ones, as the other algorithm does. We can see that both safety and length criteria were successfully included.

 

 

 

In the second test the majority of people in both groups declared that visualisation techniques used in our Web-GIS are effective at communicating cycling risk. Almost half respondents in both groups said to be ready to consider the safe route suggested by our planner, with a prevalence in the non-regular cyclists group, and roughly the same number of people declared to prefer our service compared to CycleStreets. In Open-ended questions we asked respondents to indicate which improvements they would apply to our service, most answers indicated the inclusion of route distance/time, a user defined measure for the degree of risk and general improvements to the safest paths sometimes judged too long, yet capable to avoid dangerous areas.

 

 

Discussion

The exploration of new techniques filled existing gaps and revealed that the message of road risk based on complex criteria is perceived differently from experienced and non-experienced cyclists. As for the effectiveness of representation techniques, users from both groups declared to be more aware of risks and their location after seeing the map. Our representations can be considered successful; the most appreciated is the combination heat map / safest route.

The second test showed some imprecisions in the optimisation of the algorithm, sometimes safest routes are perceived as too long and sometimes found routes are not actually the safest. Other factors to consider in unsuccessful outcomes are risk factors choice and their modelling. Our research was focused on safety and distance, but as extensively witnessed by previous studies on this regard (Aultman-Hall et al., 1997; Howard and Bruns, 2001; Winters et al., 2011) criteria playing a role in the choice of cycle routes are highly subjective, depending enormously on cyclists characteristics.

 

Conclusions

The first test made evident that safety has been effectively included, the second showed some weaknesses. Experienced cyclists do not always take into account safety and some unexperienced cyclists consider travel distance/time as more important.

We can suggest the use of our system not as a stand-alone web-GIS for safest route choice but as a valid alternative to other suggested routes; for example, the fastest and the quietest. In addition, as demonstrated by our survey this function could be highly considered by non-regular cyclists and might prompt these road users to cycle.

 

References

Aultmann-Hall, L., Hall, F. L., Baetz, B. B., 1997. Analysis of bicycle commuter routes using geographic information systems: implications for bicycle planning. Transportation Research Record, 1578, pp. 102-110.

De Hartog, J. J., Boogaard, H., Nijland, H., Hoek, G., 2011. Do the Health Benefits of Cycling Outweigh the Risks?, Ciencia & Saude Coletiva, 16(12), pp.4731-4744.

Howard, C., Burns, E. K., 2001. Cycling to Work in Phoenix: route choice, travel behaviour and commuter characteristics. Transportation Research Record, 1578, pp. 39-46.

Parkin, J., Wardman, M., Page, M., 2007. Models of perceived cycling risk and route acceptability, Accident Analysis and Prevention, 39, pp. 364-371.

Pucher, J. and Buehler, R., 2008. Making Cycling Irresistible: Lessons from The Netherlands, Denmark and Germany, Transport Reviews, 28(4), pp. 495-528.

Transport Scotland, 2013. Cycling action plan for Scotland 2013. [pdf] Available through: http://www.transportscotland.gov.uk/report/j0002-00.htm http://www.transportscotland.gov.uk/report/j0002-00.htm[accessed 19 July 2014].

Winters, M., Davidson, G., Kao, D., Teschke, K., 2011. Motivators and deterrents of bicycling: comparing influences on decisions to ride, Transportation, 38(2011), pp. 153-168.