There are several options to choose from when looking for maps of urban areas; however they are often out of date, especially on a global scale. With the rapid urbanization of the earth’s cities, timeliness is an important component of monitoring urban growth.
The nighttime lights dataset is great for visualizing cities. (Download either from VIIRS of DMCP.) I’m using the DMCP dataset which comes in annual composites on a scale from 0 – 100, with 100 representing the brightest, most consistent lights. You need some threshold to discriminate between the core urban areas and the less intense suburbs/sprawl in the surrounding areas. A new function at 10.2.1, Binary Threshold, is one option for this type of analysis.
The algorithm behind Binary Threshold is based on the Otsu method, an old school image classification technique developed to discriminate between foreground and background in an image and minimize intra-class variance between the two classes.
Here’s what the raw nighttime lights data looks like over Europe:
And after binary thresholding:
Here’s a close-up of Central Russia:
And central Russia after binary thresholding:
You can add another step and mask out all of the core urban areas as identified by binary thresholding, then call everything that still has a positive value “suburban,” and areas with a value of 0 are rural. If you are working in a small study area, you probably want to incorporate ancillary data, but on a large or global scale, this is a reasonable approach.