Walkability: What is it really?
The question is very simple: What is walkability? It is the answer that is the most difficult to answer. This has been a hot topic for the past 20 years or so, and we are yet to find a universal definition. Funny thing is it is not even in the major dictionaries. The difficulty in creating a “dictionary-appropriate” definition for walkability may lay in the fact it concerns a variety of disciplines, the studies vary in how they identify and explain walkability according to the properties they deem worthy, which both enriches and complicates the issue further.
The ambiguity of this non-definition was my starting point. There are a good deal of definitions specific to study, and because of this, the studies start to contradict each other at one point. Why don’t we investigate this problem in the scope of CaaDM, then? I had an ulterior move, as well, since walkability was the main theme of the studio I was assisting in the same semester I took CaaDM. I thought it would be hilarious to deal with the issue as a PhD student, and teach and guide architects-to-be about it at the same time. Being both the student and the instructor of the subject was as frustrating as I predicted it could be. The more I studied it, the more critical I was of the students’ works; and as the students delved deeper into their projects, they came up with a whole lot more walkability problems which made my data table larger. Good old rabbit hole.
So what did I do in order to find out ‘what makes a street walkable’? I created a very long attribute list which I would be simplifying and normalizing in the end to tone down the noise. There was also ‘personal rating’ working as the label which helped unfold the different relations between attributes and their effects on walkability. I used mutual information and correlation matrices to find out the attribute-attribute and attribute-walkability dependencies. The findings were not surprising, but they were at the same time eye-opening in the sense that they helped me see the relations I would have overlooked.
Mutual information matrix suggested that lane width, obstacles on the sidewalk, sidewalk width, amenity concentration and the number (and ages) of people using that street tell us the most about our personal rating of walkability. In truth, we could deduce all of this on our own. It is the smaller relations that mean a lot more to me. For example, sidewalk width is linked to obstacles and amenities; meaning the larger the sidewalk, the more justifiable shopkeepers find it to put all their store stuff out there. This is particularly a big problem in Turkey. Shopkeepers just love to occupy sidewalks. Another link could be found between amenities and lane width. This tells us that the larger streets may attract more amenities.
One interesting attribute I was insistent on keeping was ‘store owners’ and whether we could identify their presence on the sidewalk from Google Street View images (Spoiler alert: We could.). This might not make sense to most people, but shopkeepers especially in Turkey enjoy spending time socializing outside their shops. You can see them chatting, playing backgammon (needless to say while drinking tea), or just people watching. This attribute was mostly linked to the number of amenities and façade transparency. It is easy to see that as the number of amenities increases, shopkeepers find more ways to spend time together socializing. This didn’t really have major impact on the personal rating of walkability, but they surely contribute to the liveliness of a street.
Correlation matrix lets us see the positive and negative relationships between attributes. Looking at its results together with the results from mutual information matrix also revealed more about the dependencies of attributes in terms of personal rating of walkability. The positive correlations were sidewalk width, amenities and people with higher dependency; car-free streets, bikes and transparent façade with less dependency. The negatives were the lane width and obstacles. The interesting point here is that when we look at the dependency between the sub-attributes of amenities, the type of amenity doesn’t have significance in walkability. Only there is stronger dependence between retail and financial amenities, which may suggest more shopping equals more need for cash, or banks like to cluster around economically active areas.
This study was just a sample and well… Algorithms are hungry for data. It would be great to have a very large dataset for this problem, but still, I am happy with the results. It shows the potential of what data can tell us. It might be too utopian (but I am hopeful) to imagine a world where a globally collective database is used just for understanding walkability which, to be frank, as big-city-dwellers we all need to be more invested in. Jeff Speck says that living in walkable cities is greener than living in a sustainable gadget filled house. And a greener world we all dream of will get closer with more walking, less driving.
