The proliferation of affordable web and mobile technologies has made geographic knowledge production more accessible to many. This has been allied with greater use of open source tools among the general population. Geographers work alongside others to produce high quality cartographical material used daily and across many digital platforms. At the same time, the production of these volunteered geographic resources remains highly uneven. Mirroring the uneven geographies of economic and social spaces and practices, the unevenness of geographic knowledge production is also now clearly evident. Places that are left out of our understanding of the world are reflective of and contribute to the ongoing and uneven power relationships that produce that cartographic knowledge. In turn, uneven geographic production through such projects reflect and reproduce broader social and spatial inequalities. In seeking to understand the process of the uneven geographies of knowledge production, we can examine how the public contributes to map making. This is often called volunteered geographic information or VGI. These practices sit within broader Public Participation Geographic Information systems (PPGIS) and an extensive body of work has been compiled around these. VGI spans small-scale community art projects and weekend hobbyist activity to anti-capitalist practices and disaster relief mappings, such as that seen in the aftermath of recent natural disasters. VGI is supported by an infrastructure and produced as electronic resources for all of those connected to use and re-use.
One of the most significant VGI projects currently in use is Open Street Map (OSM), a free digital map accessible to all and where reviewed contributions are aggregated using a single database. (It is so much more than this but for brevity this to me represents the easiest way to define OSM at this stage.) It is used widely across a range of web interfaces through Application Programming Interfaces (APIs) and connected tile-mapping services. Since its inception in 2004, OSM has grown to provide a rich and useful database of places and map features. It is an ongoing and iterative project to represent the world cartographically such that volunteers can add to, access and reproduce maps of any part of the world. It stands apart from proprietary mapping systems provided by private companies like ESRI, Google and Microsoft and their partners. Because it is an iterative contributory database, OSM remains a partial and informal arrangement of space, organised by contributors who are not full-time cartographers but volunteers. There are few enough cartographers left in universities, let along outside of them. This partiality of contribution is an indicator of uneven geographic knowledge production. These partial contributions have been examined across entire countries but many have not connected these with structural social processes. Instead, the uneven production of geographic knowledge is often framed as a problem of inadequate access to a reliable web connection. Given that volunteers can map anywhere they wish and the near ubiquity of high quality web access in Europe, unreliable web connection seems an implausible factor in uneven mapping efforts. OSM mappers can select an area or a set of objects like swimming pools, or forests) and concentrate massed and individual efforts at placing new objects within the database.In fact, this provides the basis for a social network of mappers across dozens of countries who contribute to the OSM for a plurality of reasons.
In this two part blog post, I want to address this implausibility by examining the partiality of contributions in Dublin, Ireland. I want to examine the ways in which cartographic poverty is embedded within broader structures of social inequality. Not just a result of inadequate access, cartographical poverty is an outcome of class bias in the representation of urban areas. Cartographic poverty affects not just our ability to find a location on a map but who gets represented and in what ways. If volunteered geographic information projects like OSM are committed to mapping that which is not currently mapped, why are specific places left unmapped? in this first part, I talk a little bit about OSM in Dublin.
Maryland, Dublin, not Maryland the state.
This is Maryland, south inner city Dublin. It is a neighbourhood which consists of several hundred small houses, tucked in behind St James’s hospital.
A walk around the neighbourhood would show a couple of rows of housing, connected by narrow streets, dominated by the parked cars of residents. In OSM, the housing on Ave Maria road for example, is represented by one single polygon shape. The housing is ex-municipal housing, now largely privately owned as a result of failed policy of the 1980s. Each polygon shape on the OSM on Lourdes or Ave Maria roads represents eight to ten individual housing units. Judging from the chimneys evident on satellite imagery, the number of housing units per polygon varies by a small amount. The Maryland neighbourhood is an area of Dublin that has a relatively poor (albeit gentrifying) population. By the state’s index of relative affluence and deprivation, the district where Maryland is located has a score of -1.5 where the lowest (most deprived) score is -22 and the highest (most affluent) is +22.
South and east of Maryland is Westbrook road, near Dundrum. This neighbourhood has lower density housing, a large golf course nearby and is an area of greater affluence. Its located in an area where the corresponding score in the state index of affluence and deprivation is +15.4. As can be seen from the embedded map below, each housing unit is represented by a single polygon.
These mapping polygons, lines and points are placed within the OSM database by volunteer mappers, myself included. I am very occasional OSmapper, inspired mostly by cycles within the Dublin, Meath and Wicklow region at weekends. It is not that OSmappers like me are individually biased when placing features on the map on this or that occasion: this partiality of representation is a structural issue. What I mean by this is that mappers as a whole are mapping more features (buildings, houses, libraries, schools, green spaces etc) within the wealthier areas of the Dublin region than in the poorer areas. In the next section, I will discuss this in more detail.
We tend to give more attention to things that are nearby to us than things that are further away. This is the reason why 50 deaths in a landslide in Brazil commands less of our attention than 3 deaths in Wexford. The analysis of OSM contributions points toward this tendency but for volunteer mapping. I call this cartographical poverty: the tendency for smaller numbers of volunteer map contributions in areas of relative deprivation when compared with areas of greater affluence. I will expand upon this in my next post but this is more than merely a question of the availability of reliable internet connection. It is more a question of who gets mapped and who does not. Remember that in OSM, all features fall into a set of nested categories (using classes and tags) so houses can be tagged by contributors as buildings and houses.
In Dublin’s 40 poorest districts, those with the lowest scores using a 5-class natural break in the data, there were 5,063 features tagged as a house in the OSM database last summer. In the wealthiest 76 areas, there were 29,990 features marked as ‘house’. Not only that, the residential landuse tag was used for 272 features within the bottom 40 poorest areas but used 1,229 times within the 76 wealthiest. The area of these landuse features also reveals an interesting set of data. The residential landuse features in the richest 76 districts have a total area of 53,916,247 sqm or just under 44,000 sqm per district. In the poorest 40 districts of the region, the total area of the 272 features mapped is 23,076,252 sqm or a little under 85,000 sqm per district. If residential landuse is tagged in poorer areas, they are likely to larger and less detailed. This differential extends to other tagged features in the poorest and wealthiest districts: tagged green spaces, leisure parks and swimming pools.In the image below, the darker the red, the larger the area. I am using it is here as a crude indicator of feature mapping granularity.
If OSmappers like me can map anywhere they want, using well-known base maps from the usual providers, why is there a difference in the number of areas mapped and tagged as residential landuse and in features marked as housing?
Cartographic poverty then, and I will explore this in more detail in the next post, might be understood as a differential in the representation of mapped features across space based on the structural inequalities evident in other ways.