Researchers release open global city boundaries dataset for consistent cross-country urban definition

A research team announced an open dataset of city boundaries covering more than 10,000 global cities, aiming to give cross-country urban research a definition of "city" whose meaning stays as consistent as possible across regions. It matters because many economic and social studies must aggregate geocoded indicators—pollution, greenery, building height, violent events—to the city level, and the standard datasets UCDB/GHSL often distort in low- and middle-income countries.

Method and calibration

Rather than mechanically applying a single global threshold, the team searched for rules that work across regions yet suit research purposes. Their approach combines three indicators—GHSL population density, nighttime lights, and GHSL built-up area—and treats a place as urban when at least two of the three hit. Thresholds are not globally uniform but set per country, defined by that country's largest city. The team manually inspected the 30 largest cities and countries, adjusting thresholds in a few clearly unreasonable cases to find a standard that works across regions.

Problems with existing datasets

According to @paulnovosad, in high-income countries the algorithm's boundaries are already close to the GHSL frontier, with disagreement concentrated at suburban edges because the economic boundary of a city is inherently fuzzy. In low-income countries, however, density-threshold-based UCDB/GHSL more easily mislabels dense rural areas as urban. In the Dhaka case, the GHSL-UCDB boundary reportedly extends far beyond common sense, sweeping in southern cities hours away by car or bus with large stretches of farmland in between; India's Hajipur is likewise misjudged as one giant "city" when it is closer to a rural area spanning scattered small towns across Bihar.

Background and significance

The team argues that whether low- and middle-income countries can form more efficient cities is closely tied to prosperity, yet the relevant research has long had a rural bias, with a core reason being the lack of good city boundary data. Unable to find an off-the-shelf dataset usable for more than 10,000 global cities, they developed one themselves and released it openly, aiming to lay groundwork for future comparative global city research.

2026-07-15 ~ 2026-07-15 · 9 related posts