This post is a reflection of Miller's (2010) article in the Journal of Regional Science “The data avalanche is here. Shouldn’t we be digging?“. In this sense, beacons in the title symbolize emergency locators, which are used in case of avalanche danger in the mountains. The utilization of various electronic/digital data sources such as point‐of‐sale data, location‐aware technologies, sensor networks, and the web apps is still (after 11 years) in a way untouched opportunity for geographers and other urban scientists. We are in the situation when we have different sorts of data, as well as appropriate methods to work with them, but the process that Miller calls Knowledge Discovery has so far been only partially explored and appropriated by urban scholars and planners. In this context, we ask several questions.
Can ‘data avalanche’ really help us to understand the timespace of the city? How to access individual sequences and why they should be called rather spatiotemporal proxies?
Just to illustrate a somewhat ambiguous and multi-layered relationship between data and knowledge, let us take a closer look at one particular set of "hard data". We will argue that the contribution of this data to an understanding of the more general spatiotemporal structure of a city is limited, unless linked to other fragments of urban knowledge.
We demonstrate here data of pick-up points from Brno Metropolitan Area (BMA), which were harvested in the period 1st December 2020 to 31st January 2021 from the public API. The following diagram shows attributes of all pick-up points in BMA shared by the package delivery company in the time frame of a single day.
Full - The pickup point was full, or near full. COVID - The pickup point was closed due to restrictive government measures. Vacation - The pickup point was closed due to reported vacation.. Technical - The pickup point was closed due to other technical issues.
At first glance, two different periods can be identified in the graph. While in the case of pre-Christmas madness and the New Year's period, we can talk about turbulent rhythms of hectic shopping (and delivering), the following period seems more stable in terms of package deliveries. The question is which of the two periods we consider to reflect ‘usual’ daily urban routine. In other words, which sequence of rhythms conveys the image of a normal city to us? We usually understand the pre-Christmas period as a kind of singularity. Compared to it, the rhythmicity of January corresponds maybe better to the expected (semi)regularity and ordinariness of everyday urban life. Ok, so far so good. Accepting this, we can go further in our exploration of the dataset.
We put peak (Wednesdays) and off-peak (Sundays) days in January on a map to examine the interconnectedness between temporal regime of pick-up points’ occupancy and qualities of urban space.
left map - peak days and occupancy; right map - off-peak days and occupancy The larger the map sign, the higher occupancy.
There is no easily interpretable pattern. Occupancy figures simply did not match with the spatial distribution of important transport hubs, city squares or other places of higher centrality. Overall spatial distribution of occupancy seems to be rather random. Is it disappointing? Not necessarily. It is a reminder that there is a difference between what spatiotemporal data can represent (or even what we think it should represent) and what it actually represents.
Back to our dataset. It turns out, that without knowledge of other attributes, such as the size of storage space and full knowledge of the supply process of pick-up points it is impossible to consider these data as true “hard data“ that can be employed without any contextual information in modelling urban timespace. In this case, basically, we do not know in what situations the storage space was filled.
This leads us to the conclusion that these types of data should be considered as ’social proxy data‘ or better said as “spatiotemporal social proxies’“, which reveal just small pieces of spatiotemporal knowledge. They are worthless when analysed and interpreted in isolation, useless for drawing conclusions when disconnected from their own context. However, they can still be useful. The key is to connect and find associations with other pieces - other proxies. Similar to avalanche beacons, there is need of more than one beacon to find someone in an avalanche...
References: Miller, H. J. (2010). The data avalanche is here. Shouldn’t we be digging?. Journal of Regional Science, 50(1), 181-201. Packeta (2021). Pick up points. Packeta s.r.o. https://widget.packeta.com/point/all-own.json