Measuring Cities from the Sky
camera + flight = awesome maps
Not long after humans successfully started lobbing objects into the air, the tremendous potential to gather data became apparent. Balloons were already being employed for reconnaissance as early as the Civil War (in much the same way spy satellites are used today). Pioneers like Sherman Fairchild, revolutionized the way we saw cities by photographing them from the air.
The aerial perspective allows us to measure things on the ground as they really are with far less effort than it would be on the ground itself. Put simply, it's the bird's eye view. As both aviation and measuring equipment became more sophisticated, so to did our ability to accurately measure land. Planes gave way to satellites and we augmented our photographs with other spectrums of measurement.
This is of particular interest to those concerned with cities as it affords a convenient way to measure the pattern, extent, and cohesiveness of urban development. To use a dirty word: sprawl.
Land cover data is a much more sophisticated version of these early urban photographs but the basic idea is the same. To put this data to work we start with the Mark I Eyeball, comparing patterns of development. We could accomplish this using widely available aerial images such as those included in Google Earth. Land cover data goes further by codifying the nation into types of land. The extent and boundaries of human settlement become readily measurable.
Land Cover Legend
National Land Cover Dataset
This data-set, produced by the Multi-Resolution Land Characteristics (MRLC) Consortium is the combined effort of pretty much all of the major federal agencies concerned with geography and land, from NASA to the Fish and Wildlife Service. I guess that makes it the Captain Planet of remote sensing data. At the heart of the data is Landsat satellite data.
The NLCD provides continuous coverage of the lower 48 in 30x30 meter grid cells. That works out to about a fifth of an acre which is granular enough to measure low-density residential development. To cover such a large area it takes 9 billion cells (and a good 16gb of hard drive space). Overall, its pretty accurate too. Their own testing put the accuracy around 80%. While its not as granular as say parcel data, it makes up for it with scale. Land cover data facilitates comparison at a regional and national scale. The goal is to crank these out every 10 years like the census so we can keep track of how land in the country is changing.
The best part about the NCLD? Also like census data, it is freely available for anyone to download and use. The expressed goal of the project is to provide fuel for academic research, monitoring, and plain old curiosity...
A Rorschach Test for Urban Land
The goal of these map galleries is to provide a back of the envelope way to compare the size and configuration of urban areas. Sort of a Rorschach test for urban land. I was partly interested in creating a more standardized and updated version of this popular comparison of Barcelona and Atlanta displayed below. While I like the simplicity of this graphic I ultimately decided to include a fuller picture of urban development including the relative intensity of development and terrain context. I included transit lines, but as I am comparing American cities to each other, I also included interstates. The population values represent what's visible at that scale rather than any official administrative boundaries.
Hopefully these graphics will spur some interesting comparisons and questions. What do 5 million people in Atlanta look like versus 5 million people in Boston? Do all southeastern Sun-Belt cities really look interchangeable? Why do cities take on the pattern that they do?
I was also interested in producing these images for their own sake. Each map is like an urban snapshot capturing the net result of history, terrain, water, and growth. They capture each places' decisions about and reactions to land and development. In a way, the shape of each metro almost becomes a unique, yet recognizable pattern, like a fingerprint.