Cycling in bike-friendly Amsterdam
Where does the infrastructure fall short?
Introduction
The Netherlands is a country of cyclists. It has the most bikes per capita and the bike is used by many people to move around. This is also the case in the capital of the Netherlands: Amsterdam. The city has more than 500 km of bicycle lanes and 50% of the total commuting is done on bicycles (Gaadi.com, 2015). But unfortunately, since so many people use bikes, cycling accidents do happen in Amsterdam. Between 2008 and 2017, there were 5038 (rijkswaterstaat, 2018) reported accidents in Amsterdam where a bicyclist was involved. We wanted to know why there were so many accidents in the city. Therefore we wanted to investigate if there was a relation between bicycle accidents and infrastructure of the roads in Amsterdam. We formulated the following research question:
What is the relation between bicycle accidents and infrastructural attributes of the cycling paths in the city of Amsterdam?
First, we wanted to focus on where most of the accidents took place in Amsterdam by making a hotspot analysis of bicycling accidents. After this analysis we wanted to analyze those hotspots. Therefore we formulated the following subquestions:
Where did the bicycle accidents take place in Amsterdam?
What infrastructural characteristics are mainly present within hotspots where cycling accidents take place?
Literary importance
We first looked into some literature of bicycle accidents. In order to analyse where the accidents took place, we used a method that was also used by. Pulugurtha et. al (2007). They identified a GIS method which identifies predrestian crash zones. By using hotspots and the kernel density method, the authors were able to identify crash zones in Las Vegas. A more detailed explanation of this method will be provided in the method section of our research. Another research about this topic was done in Vancouver by Schuurman et al (2009), 31 of the 32 hotspots in Vancouver were located on downtown streets and two third (66%) of the hotspots were at intersections. Their main finding about the relationship between the built environment and the hotspots was the presence of bars and other places where you can buy alcohol.
Bike infrastructure
As mentioned before, the Dutch people are used to cycling often as a reliable and safe way of getting from point A to B. But accidents happen and they can have lethal consequences. A safe journey depends on several indicator’s a cyclist has to face. Beside a safe bike infrastructure on a human scale there are policies and measures to cultivate social relationships (Vassi & Vlastos, 2014). In this research we will keep our focus on the infrastructure. The International Transport Forum has four benchmarks to provide a safe cycling infrastructure, these are as followed out of Wardlaw (2014):
1. Speed management is a critical and effective tool to reduce the severity of bicycle/motor vehicle crashes.
Speed management protects cyclists as a sort of hidden infrastructure. The impact of a higher speed level results in an unequal mass speed relation which results in a higher percentage of fatal crashes. The speed limit on a safe mixed bicycle-motor way shouldn’t be higher than 30km/h. Speed-control devices such as humps, bollards and signage improves the speed limit compliance of motor vehicles and thus also the safety of the cyclists (OECD, 2013).
2. Where speeds cannot be lowered, or where traffic densities are high, authorities should seek to separate bicycle and motor traffic.
Most scholars agree about the effectiveness of this measure (Hoglund, 2017; Wardlaw, 2014; Ohlin, Alguren & Lie, 2019; Hull & O’Holleran, 2014; Vassi & Vlastos, 2014; OECD, 2013). A cyclist is the most vulnerable driver on the road, because of the mass and speed disadvantage. It is imported to have separated bicycle infrastructure at places where cars are accelerating (OECD, 2013). The Danish Cycling Embassy suggested to have a separated cycle track or lane at a medium traffic intendancy of a motorized road at 35 km/h. This is because the fatality risk of a crash with a motorized vehicle is significantly higher (OECD, 2013).
3. Separated bicycle tracks are an attractive option as they generally produce fewer and less severe crashes in their linear sections however, safety may be compromised at junctions, where crashes may increase unless specific counter-measures are undertaken.
Junctions, intersections and roundabouts are the main hazard to cyclists on cycle lanes. The most collisions (74%) happened at junctions (Hull & O’Holleran, 2014). Dutch roundabouts are safer because they are mostly designed with an orbital ring where motorized vehicles must give priority when entering and leaving the roundabout (Hull & O’Holleran, 2014).
4. Crash risk at bicycle track/road interfaces is exacerbated by poor sight lines, and confusion regarding expectations of cyclists vis-a-vis motorists and vice-versa. Proper design of junctions contributes to lower crash risk.
It’s important to lower the risk of accidents by implementing a proper design which has fewer barriers, clear signals, physically demarcated and/or seperate cycling lanes and gives priority to cyclists at junctions (OECD, 2013).
Data
The data is mostly composed of datasets and maplayers provided by the municipality of Amsterdam. These can easily be downloaded through several websites from the municipality and the government. The first dataset containing information about traffic accidents is a national dataset collected by Rijkswaterstaat. This data can be found through the municipality of Amsterdam which has selected only the information which is relevant to the city. From their selection we have made our own array of attributes and selected only the accidents which directly involved bikes. Here we knowingly made the choice not to include e-bikes and scooters since they could skew the results because of their higher speed and thus higher risk of accidents.
Other datasets we used are also provided by the municipality. They show the network of recognized bike paths. This is not a map of all the routes that can be biked, but rather a map showing the bike paths which, according to the municipality, are part of the city’s biking network. The attributes which are included in this dataset can be used in our analysis since some of these are clear infrastructural characteristics. These needed to be linked to the information on the accidents before they could be useful.
Furthermore we made use of built-in basemaps provided by ArcGIS, these clearly show the context in which our bike network can be placed. The basemap is vital for viewers unfamiliar with the urban layout of the city.
Methods
Map 1: 'optimized hotspot analysis'
The different datasets by the municipality of Amsterdam offer a variety of possibilities on how to visualize them. As a first step we wanted to see the density of bicycle accidents. For this we needed to prepare the database for usage. We had to delete some suspiciously high values from places which were not very logical. We suspect that data aggregation in an earlier stage may have created some odd values. After deleting these exceptionally high values and some unnecessary attributes from the attribute table, the database was ready for use. We tried to apply an ‘optimized hotspot analysis’ on the database containing all the accidents involving bikes in Amsterdam (Map 1). This yielded some interesting results which clearly bring out two major hotspots of bicycle accidents around the Dam square and Leidseplein. More information will be provided in the results. However this Hotspot Analysis created a cluttered representation. Therefore we tried an easier method provided by ArcGIS: ‘Kernel Density’, this method creates a heatmap of accidents in Amsterdam. This density map shows nearly the same image as the hot spot analysis and is more clear on where the hotspots are located. This method was also used by Schuurman et al.(2009) and Pulugurtha et al.(2007), where they investigated pedestrian injuries. This heatmap combined with a simple layer, again provided by the municipality, showing the biking network as the municipality intends it, forms an informative map on where bicycle accidents take place (map 1). The density map has been divided in five classes to create a visual distinction between higher densities and lower to no density. In this way smaller ‘hotspots’ with lower, but still above average scores also become visible and the same classes were used by Pulugurthan et al. (2007).With more classes the map will get more cluttered and the results will become less visible. With the usage of a heatmap we lose the ability to visualise coldspot, however for us this is not that important since we are only interested in the higher values.
To check if this outcome does not only show the population density (of the bike paths) throughout the city, we used an indication by the municipality which bike paths are the busiest, which are shown in map 2. Here the busier bike paths are yellow and red, while the ‘normal’ quiet bike paths are green. This way it becomes evident that the hotspots are not in the same place as the busiest biking routes in Amsterdam.
The next step in our analysis brought us to a more detailed look at the Dam square and Leidseplein. We wanted to know which characteristic infrastructural attributes are involved in accidents involving bikes. Through the data, provided by the municipality of Amsterdam, we found some interesting attributes to analyse. Most importantly these are: separation, width and road surface of bike paths. These came from the dataset which showed the bike network and had to be linked to the information on accidents since we wanted to know which attributes could have been involved in which accident. We did this via a buffer around the pieces of road with certain attributes and selected all the accidents within this buffer. This way all the accidents around the characteristics we study could be analysed together. We have made these three attributes into three different maps per hotspot to accentuate that different attributes have different locations but they can also overlap. If we were to combine these three maps per hotspot into one, this map would be unclear and would not show the intended result. To create these maps we used attributes from the dataset by the municipality. This dataset was already geocoded into a shapefile. The only preparations that had to be done was cleaning up the big attribute table and delete all the unnecessary attributes and project them into the same coordinate system. We ended up using the WGS 1984 projection because this way we were able to utilize the built-in basemaps from ArcGIS.
Where did the accidents take place?
In map 2 you can see the results of our density map of cycling accidents in Amsterdam between 2008 and 2017. More bicycling accidents took place in the red areas than in the yellow areas, almost all of these density points are in downtown Amsterdam. This was expected, because this was also the case in the research of pedestrian injuries in Vancouver and in Las Vegas (Schuurman et al., 2009) (Pulugurtha et al., 2007). However, in Vancouver and in Las Vegas you could see that these density areas were more spread in the downtown area, but in Amsterdam there is one big density spot for almost the whole city centre. This may be explained by the fact that the surface area of the city centre is smaller compared to those of Vancouver and Las Vegas.
Map 2: Kernel density
There are two main areas in Amsterdam with a very high density of cycling accidents. The first area is around the Damsquare. The second high density spot is the area surrounding Leidseplein and Max Eeuwen square. A lot of bars, pubs and clubs can be found in those two areas. This also correspondents with other scientific research, where they pointed out that the presence of bars and pubs play a role in traffic accidents (Schuurman et al., 2009).
What infrastructural characteristics are mainly present within hotspots where cycling accidents take place?
In order to say more about why these accidents take place around Leidseplein and the Dam square we made some maps about the cycling infrastructure in Amsterdam. Most of the accidents took place on the busiest cycling paths as you can see in the following map. The cycling paths at and around the Dam and Leidseplein are busy as well. So the amount of usage of a cycling path seems to be a characteristic of bicycling accidents. However some of the most crowded cycling paths are in Amsterdam-East. Another very crowded cycling path is the Overtoom, but all these streets are not located in the high density areas of bicycling accidents.
Map 2 (Kernel density) and map 3 (network usage)
So the amount a cycling path is used by cyclists is definitely an important factor in the occurrence of cycling accidents, but it is not the only one. Speed management, mentioned by OECD (2013) is not applicable on the high density places, because the maximum speed in the city centre of Amsterdam is 30 km per hour for cars, motors, bicyclists. In the literature it was found that there are more accidents at intersections. However, in our data it was often a little bit vague if the accidents actually took place at an intersection or at a road close to an intersection. Therefore we decided that we wouldn’t use intersections as a characteristic.
Conclusion
What is the relation between bicycle accidents and infrastructural attributes of the cycling paths in the city of Amsterdam?
Referring to our main question, we first started to analyse the places where a lot of bicycling accidents took place through hotspot and density analyses. The areas that stood out were the Dam square and Leidseplein.
Our main conclusion for the relationship between bicycle accidents and infrastructural attributes of the cycling paths in the city of Amsterdam is as follows. Both the Dam square and Leidseplein are one of the busiest areas in Amsterdam. Not only concerning mobility but also a lot of leisure entertainment, such as bars and clubs, are located in these particular areas. However the amount of usage and traffic of those areas can’t be seen as the only factor for all the bicycle accidents since the busiest cycling paths are not located in the areas with high accident density. Furthermore both places are located in an area of a maximum speed of 30 km per hour. For this reason most recommended safe biking infrastructure of the OECD (2013) measures are not necessary. This doesn’t mean there is no space for improvement. The lack of separated cycling paths could make areas with high accident density significantly safer. In the hotspot around the Dam square there were a lot of accidents on rough surfaces like clinkers and narrow cycling paths. These infrastructural characteristics could be improved within Amsterdam to ensure that there will be less accidents in the future.
Reflection
Constructing this storymap has made our understanding of conveying a spatial image to the neutral viewer a lot better. We turned the usual perspective around for this project. Most of the time we start with an interesting topic, conduct research on it and try to visualise this through a map, but this time we started creating maps next to conducting our research, which developed a more visually oriented result.
Unfortunately the corona crisis has had a big influence on the way we were able to work together. We experienced that working together physically was restricted by government rules. This made working with ArcGIS rather difficult since most of the map-making had to be done on one computer. Through video calling and meeting in pairs instead of with all four we were able to make ends meet.
The results could have been more precise if we had used analytic methods provided by, for example, SPSS. However since the limited amount of time and our main focus on a visual representation through GIS, we decided that this would have been impossible. We are happy with the results of our small research and the visuals we created with it, although they could have been researched further to get a more detailed understanding of how infrastructural attributes play a role in bike accidents.
References
Brezina, T., Ibesich, N., Niegl, M. & Lemmere, H., 2012, Requirements for high quality cycling infrastructure design, 2nd International Conference on Road and Rail Infrastructure (CETRA), (953-959)
Gaadi.com, 2015 List of top countries with most bicycles per capita. Available at: http://www.gaadi.com/cycles/news/list-of-top-countries-with-most-bicycles-per-capita (Accessed on: 23 november 2020).
Gemeente Amsterdam, 2020, Fietsnetwerk Amsterdam. Available at: https://data.overheid.nl/dataset/196aeef9-eeba-4442-b2df-1ca3e9f9f9f1 (Accessed on: 24 november 2020).
Hobma F.A.M & Jong P., 2016 ,Planning and Development Law in The Netherlands, Chapter 5, Planning Law Private Instruments, (71-91)
Hoglund, M., 2017, Safety-oriented bicycling and traffic accident involvement, International Association of Traffic and Safety Sciences, (152-162)
Hull, A. & O’Holleran, G., 2014, Bicycle infrastructure: can good design encourage cycling?, Urban, Planning and Transport Research: An Open Access Journal, Vol. 2, No. 1, (369–406)
OECD/International Transport Forum, Cycling, Health and Safety, (1-245)
Ohlin, M., Alguren, B. & Lie, A., 2019, Analysis of bicycle crashes in Sweden involving injuries with high risk of health loss, Traffic Injury Prevention, Vol. 20, No. 6, (613-618)
Pulugurtha SS, Krishnakumar VK, Nambisan SS: New methods to identify and rank high pedestrian crash zones: An illustration. Accid Anal Prev. 2007, 39 (4): 800-811
Rijkswaterstaat, 2018, Verkeersongevallen (BRON, landelijk). Available at: https://data.amsterdam.nl/datasets/G6xF0loDvppzog/verkeersongevallen-bron-landelijk/ (Accessed on: 16 november 2020)
Schuurman, N., Cinnamon, J., Crooks, V. A., & Hameed, S. M. (2009). Pedestrian injury and the built environment: An environmental scan of hotspots. BMC Public Health.
Vassie, A. & Vlastos, T., 2014, A review and critical assessment of cycling infrastructures across Europe, WIT Transactions on Ecology and The Environment, Vol. 191, (757-768)
Wardlaw, M., 2014, History, risk, infrastructure: perspectives on bicycling in the Netherlands and the UK, Journal of Tansport & Health, (243-250)
