The map has come to precede the territory.
Working with interactive mapping introduces challenges not typical to static mapping. In the process of creating the Cambridge Housing Authority’s interactive map of public housing developments, I ran across some unexpected accuracy issues. Specifically, this project aimed to provide basic information about available public housing developments for potential residents. The most basic fact about these developments concerns where they are located. Yet, beyond particular scales of the map, the point data used to represent location became skewed. Being familiar only with the location of the central office, I was only made aware of this inaccuracy thanks to some property managers who actually worked at the development sites.
This problem stems from the reliance on administrative data at the agency level. Because the agency isn’t concerned with the use of the exact geographic location of its public housing developments, it relies on a list of street addresses – something that’s pretty obvious for administrative purposes. However, when dealing with map-making, especially in an interactive environment where not every detail is intimately known to the map-maker, it’s possible for users to notice slight inaccuracies in the data that the map-maker would not otherwise see. Clearly, addresses will point someone to the right place for navigation, but in an interactive setting in which users can zoom in to the exact location of the street address, they are only shown a street corner and never an actual building.
In the image above, the data provided the same address for two very different properties. When the addresses were geo-coded, the two sites were given the same coordinates, effectively hiding the Washington Elms property behind Newtowne Court. Additionally, at this scale, the geo-coded point barely relates to the location of the actual property.
This problem led me to consider alternative ways of generating high-detail point data for interactive environments. To be sure, it is entirely possible to limit the extent to which the user can zoom, preventing the necessity of any additional research, but that’s no fun, anyway. Clearly, public housing developments appear as polygons at close zoom levels. Using City of Cambridge parcel data, I was able to generate a map of the actual lots in which the properties were built. Then, I provided this as a layer on my map:
Beyond a particular zoom level, the application will switch this layer on, so that the user has a stronger idea of where the properties are actually located. However, to avoid additional complexity, it’d probably be acceptable to generate centroids of the parcels, and match those points to the property data.
This underscores some of the challenges with using administrative data found in public agencies. Addresses will undoubtedly point someone in the right direction (they wouldn’t be addresses, otherwise!), but in terms of representing locations on a map at various scales, geo-coded addresses may not be accurate enough.