Concepts

Every dataset in Urban Growth Center describes a city or a piece of one. This page explains what a city is, how cities are subdivided, and how to join datasets to them.

Cities#

A city is an urban agglomeration: a contiguous built-up region with one boundary and a stable id. It is not a municipality, district or metro authority. Administrative definitions of a city vary across countries, so we define cities by built-up extent to keep them comparable across borders.

The city boundary file includes every city with a population of at least 50,000, along with its name, population estimate, country code and area.

Administrative splits#

Some questions need more detail than a whole city. For these, cities are split along administrative lines: an ADM1 split is the part of a city inside one first-level administrative unit, such as a state or province, and an ADM2 split is the part inside one second-level unit, such as a district. Administrative boundaries come from GADM.

Splits are selective, not a complete administrative map:

• Not every city has splits, and a city may have ADM1 splits without ADM2 splits or the reverse. Check each level separately.
• Small slivers are dropped, so a city’s splits may not cover its full area. Totals across splits (population, area or other measures) may not add up to the city total. For whole-city figures, use the city-level observation.

Levels and identifiers#

Every published polygon, whether a city or a split, has its own unique id. Datasets record which kind of polygon each row describes in boundary_level:

To join a dataset to boundaries:

1. Filter the dataset to one boundary_level.
2. Join to the matching boundary file on id. Names are labels, not keys.
3. Keep the time columns (year, month or epoch). Time-series datasets have several rows per polygon.

Splits also carry parent_id, which links a split to its city, and adm_id, which identifies the source GADM unit. Use parent_id to bring in city attributes. Never join split-level measures on it: that attaches them to the whole city.

Choosing a level#

Use cities for comparisons across cities. Use splits when variation within a city matters and the measure is meaningful at that finer level. Copying a city value onto its splits does not create a split-level estimate.

Each dataset reference includes a short Python join example: population, PM2.5, GHSL, congestion, Doing Business, ADM1 splits and ADM2 splits.