Authors: Levi Wolf*, University of Bristol, Sean Fox, University of Bristol
Topics: Urban Geography, Remote Sensing, Geographic Information Science and Systems
Keywords: data science, remote sensing, urban geography, machine learning
Session Type: Paper
Start / End Time: 12:40 PM / 2:20 PM
Room: Harding, Marriott, Mezzanine Level
Presentation File: No File Uploaded
Urban data science is stuck with a difficult representational problem. Formal urban boundaries drawn by governments and administrations often do not reflect the effective territory of the city. In other words, there is often a spatial mismatch between de facto urban boundaries and de jure administrative boundaries. This mismatch creates a critical governance challenge by contributing to fragmented planning, investment and service delivery. Such fragmentation leads to “sub-optimal socio-economic results” in cities and metropolitan regions in wealthy nations. Unfortunately, this problem—urban political fractionalisation—is difficult to study empirically: the typical forms of urban data studied by demographers and economists are bound by these jurisdictions as well. Census data is often gathered from subdivisions of existing administrative units, so analysis cannot peek underneath or between these lines to accurately estimate the true level of fractionalisation. Further, census data is usually infrequently gathered (meaning that contemporary changes in urban structure are invisible) and path-dependent (since enumeration units are usually only adjusted, not redrawn wholesale). Thus, we develop & apply a new method for determining contiguous urban areas from satellite imagery. We discuss preliminary results on the use of the apparent urban area to measure & model political fractionalisation, as well as the structural, social, & political factors involved in jurisdictional gerrymandering.