As a big fan of walkable cities and car-free environments, I have often come across websites like walkscore, which assign a rating to each house based on how walkable its location is. While individual scores are interesting, I always insist on seeing the full picture, so I set out to calculate my own simple walkability score for a few select cities and then plot it on a map! But let us start by defining what we mean by “walkable”
What makes a place “walkable”?
In its simplest form, a walkable environment is one that invites and encourages people to move around on foot as opposed to using a motorized vehicle. Usually, this word is used to refer to mid or high-density neighbourhoods, with mixed-use development (both residential and commercial) which result in people being able to access their destinations on foot in a relatively short time. So a basic definition could be distance to the most frequent destinations. However, there are many other factors that can influence the attractiveness of walking in a certain environment. For example:
- Sidewalk/pedestrian infrastructure
- Intensity/speed of motorized traffic
- Orography (extreme slopes discourage walking)
- Weather conditions (extreme heat/cold)
- Security
A complete definition of walkability should take all these factors into account, and that is indeed how websites like Walkscore do it. However, for the purpose of this article I will focus on a simpler definition, using only the distance to typical destinations. While this certainly yields an incomplete picture, it still uncovers city design patterns that are of great interest. Moreover, the data needed to do it is readily available, which makes it relatively easy for anyone to create a map of their own city.
Plotting a City
Let us start simple: how far away is each household from their most common destinations? The first step is to obtain and plot data about the geographical location of the households themselves. I will use my hometown of Burgos, Spain to showcase these steps; using data from the Spanish cadaster, we can plot the shape of each building. The data also contains the primary use of each building, so we will filter out commercial and industrial properties to only keep the residential ones:
Ok, this shows us the layout of the city, but now we need to add the points of interest. What are some typical destinations? A typical example is groceries – pretty much everyone needs to do them regularly. Now, there are many different types of establishments where one could get groceries, but in the spirit of keeping things simple let us start by considering only supermarkets. Using the Google Places API, we can retrieve the coordinates of every business classed as a supermarket, and then plot them on top of the map above:
Now we can see roughly which areas are well covered by supermarkets, and which ones are somewhat lacking. A simple way to assign a score to each dwelling would be simply calculating the distance to the nearest supermarket. However, what we are ultimately trying to assess is the convenience of reaching a destination on foot; in this context, there is no practical difference between 5km and 20km, since a person is highly unlikely to walk either distance if motorized transport is an alternative. It would also make no sense that 10 meters are twice as bad as 5, since, again, there would be no practical difference between these distances. So let’s design a score that takes these into account:
- We will assign 0 to the lowest score (totally not walkable), and 1 to the highest (fully walkable)
- We’ll set a threshold of 150 meters below which the score will be maximum – we assume that any distance below 150m does not make a difference
- We will also set an upper threshold of 1500m. We assume everyone will consider this “impractical” and will always choose an alternative, so higher distances do not decrease the walkability. Anything above 1500m will have the lowest score
- Between 150 and 1500, we will have a linear decay function – so halfway through the distance, the score will be 0.5
Ideally I would like the thresholds to be a bit higher – I think people will definitely walk up to 2km. However, we are using the distance “as the crow flies” here, meaning a straight line between the dwelling and the supermarket. Ideally, the street travelling distance should be used. This is doable with the Google Maps API, but somewhat complex and expensive, so for this post I decided to use only the “straight-line” distance. In order to compensate, I will lower the thresholds a bit – so I am assuming that 2km through the streets is, on average, 1.5km in a straight line. However, feel free to change these when generating your own maps!
Visualizing a Walkability Score
Using this logic, we can assign a score to each living unit, and then plot it using a color code to visualize the scores throughout the whole city:
Interesting! Now we can better see some patterns. In general, the city is pretty well served, except for some neighborhoods in the West (Barriada San Juan Bautista) and a village in the South East (Barrio de Cortes). These are lower-density neighborhoods with no commercial surfaces available.
But in the rest of the city, the high population density makes it commercially viable to have several supermarkets in a relatively small area, which in turn makes it possible for people to walk to them. The average score per dwelling (not per building) is 0.87, which is very high!
We can calculate similar scores for other “typical” destinations. For example, this is a similar map but for pharmacies instead of supermarkets:
A little better coverage than for supermarkets – the average score per household is 0.94. Let’s do the same for parks:
Also a very good score – 0.89. The scores for parks are sometimes misleading, however – we are just using points of interest that Google Maps classifies as “parks”, but the wilderness or countryside can fulfill the same needs and will not always be picked up by Google Maps.
We could calculate similar scores for many other categories, but these three are a good summary of a household’s needs. We can therefore calculate an average of all categories, and give an overall score to each household. We can then plot this on the map to have an “overall walkability” score:
We could call this map the “walkability summary” of Burgos. The average overall score is 0.89, which, again, is very high! There are some isolated areas, but in general we could conclude that this is a very walkable city (although let’s remember that there are many factors we are not taking into account).
Different city typologies
The first example with Burgos is perhaps a bit of an extreme case… how do other cities fare?
The image below shows Boadilla del Monte, a high-income residential suburb located West of Madrid, composed mostly of single-family homes. Although there are some businesses and amenities, large areas are very far away from all of them, resulting in high car dependency for a significant percentage of the inhabitants. The average score is 0.67, which not terrible, but considerably worse than Burgos.
What about other countries? Below we can see the city of Haarlem, in the Netherlands. With an overall score of 0.78, in paper it is slightly worse than Burgos. However, this score does not take into account the outstanding bicycle infrastructure that Dutch cities have, which substitutes a large percentage of the motorized traffic and makes longer distances in the daily routine of its inhabitants more workable than if they only walk.
Finally, we have an extreme example of car-dependency. Centennial, CO, USA, is a suburb of Denver, and is the stereotypical North American suburb – extremely low density and single-use zoning:
The score for Centennial is a mere 0.33 (thank the parks for that, because the supermarket walkability is an appalling 0.10!).
Do you think these scores accurately reflect how easy it is to walk through each of these cities? Feel free to contact me for more details about the methodology, for the scripts, or for any suggestions or criticism!
