Population density — how this layer is made

What you are looking at

The colour of the map shows how many people live on each square kilometre of ground, for the year 2020. There are no points to click. This map is a value at every cell rather than a set of places.

Where the data comes from

The figures come from GHS-POP, the Global Human Settlement population grid published by the European Commission's Joint Research Centre. It is modelled and not measured. GHS-POP does not count people. It takes each country's census and spreads the count for a census area across the map in proportion to how much housing satellite imagery sees there. The census is the only real headcount, and the model decides only where within a census area those people sit, never how many of them there are. We do not model or estimate anything ourselves, and moving the published figures onto the grid a web map uses keeps each country's total intact.

Colour

Colour represents how many people live on a square kilometre. Most of the inhabited world holds a few hundred people or fewer to the square kilometre, while a few city centres hold tens of thousands, so the steps rise steeply. Each step up the key is roughly three times the one below it, running 10, 30, 100, 300, 1,000 and upward, so a quiet rural district and a crowded city each show their own detail. The steps are fixed numbers and the same everywhere, so a colour always stands for the same density wherever you look.

Where we have no figure, over open sea or over ground the model never covered, you see the basemap through the layer rather than a colour. No data means we have nothing there, never a zero standing in for a low value.

You can choose between warm, magma and cool schemes in Settings, and the density each step stands for does not move.

Why the detail is not as fine as it looks

The grid is drawn finely, and most of that fineness comes from the buildings rather than from the population figures. Across the world the average census area feeding the model covers more than 30 square kilometres, so over much of the map a whole town has one measured figure spread out to look detailed. A country publishing one national total is drawn with the same crisp texture as one publishing fine local counts. We build the layer only as fine as the figures support, at roughly one kilometre to a cell, and the map smooths it as you zoom in, which adds no detail the census never had.

Which year

GHS-POP runs from 1975 to 2030 in five-year steps. The years up to 2020 are estimates, and 2025 and 2030 are projections rather than measurements. We show 2020, the most recent estimate, so that an undated layer gives you a figure worked out from a census rather than a forecast. A control to move between the years may come later.

Known weaknesses

• Thinly populated country is close to uninformative. A test against Sweden's population register found the model close to reality where people are dense and almost unrelated to it where they are sparse. It found only about a quarter of the rural cells that had people in them.
• Cities come out wrong more often than not. Another study found roughly three quarters of urban cells misjudged, with central areas given too many people and dense residential areas too few.
• The error grows the closer you look. Accuracy that holds for a whole country falls away at the scale of a local authority and below, and nobody has published a figure for the grid's own resolution.
• A country's census can be a decade or more old. Later years are then that pattern scaled by modelled growth rather than a fresh survey. We use GHS-POP everywhere rather than a newer national source where one exists, so the whole map is built the same way.

Source and licence

GHS-POP R2023A, European Commission, Joint Research Centre — Schiavina, M.; Freire, S.; Carioli, A.; MacManus, K. (2023). Used under CC BY 4.0.