Outlook · Now City

Cities outlook

When does each large city stop being a place to build for? Under the max-max scenario (SSP5-8.5 warming, 70 m of sea-level rise), every city above one million people gets an estimated expiry period and the trigger that sets it: heat pushing it out of the human climate niche, the sea reaching its core, or neither within the 10,000-year long now.

Loading…

Which comes first, heat or the sea?

Every city sits on two clocks. The heat clock runs on the assessed SSP5-8.5 warming path and stops when half the city's people live where mean annual temperature has passed the 29 °C edge of the human climate niche. The sea clock runs on the fastest plausible ice-melt schedule and stops when half of them sit below the water. The diagonal is where the two clocks agree; above it the sea comes later than the heat, below it the sea comes first.

Every city above one million people. Dot area is 2025 population. Horizontal position is the year half the city crosses the niche edge; vertical position is the year the sea reaches half of it. A city in the "no date" band of either axis never reaches that mark on that clock. Hover for the numbers; click to jump to the row.

When the cities expire

All cities above one million people, counted by expiry period and colored by trigger. Heat dates use one climate model's spatial pattern and mean annual temperature only; a wet-bulb index would pull the humid coastal cities earlier.

Every city

Sorted earliest first. Click a row for the sentence and the numbers behind it. Heat year and sea year are the year half the city's people are affected on each clock; the expiry period runs from that point to the year nine-tenths are.

City	Country	Pop (M)	Expiry	Trigger	Heat year	Sea year	Half under (m)	MAT 2100 (°C)	Water	Incoming (M)

How the dates are made

The city list is the JRC Global Human Settlement Layer's Urban Centre Database (R2024A): every centre with a 2025 population of one million or more, sampled over the centre's footprint. Both clocks use the same three marks: the point at which a tenth, half, and nine-tenths of the city's 2020 population is affected. The expiry period runs from the half mark to the nine-tenths mark of whichever clock reaches the half mark first.

Heat. For each 1 km cell, its local warming per degree of global warming comes from the CHELSA CMIP6 field (GFDL-ESM4, 2071 to 2100 minus 1981 to 2010) divided by that model's own global warming between the periods, the standard pattern-scaling assumption. The global warming level at which the cell's mean annual temperature reaches 29 °C, the edge of the human climate niche in Xu and colleagues (2020), is then read off the IPCC AR6 assessed warming path for SSP5-8.5, extended past 2100 with the low end of AR6's 2300 range so later dates stay conservative. Because the model's pattern is driven by the assessed path rather than the model's own cooler run, the dates land earlier than the raw field would suggest. The SSP3-7.0 date is shown in each row for comparison.

Sea. Drowning is a still-water bathtub on GEBCO 2024 sub-ice elevation at 1 km, counting only cells the world ocean can reach (a basin floods only once the sea has a path to it, so Baku stays dry until the Caspian connects at about 26 m). The height at which half the city's people are under water becomes a year with the schedule in the Max SLR chapter: 2 m by 2100, 12 m by 2500, 22 m by 3500, 70 m by 7000. That is the fastest plausible schedule; beside it each row says how the same height reads in the AR6 envelope, which quantifies nothing beyond 2300.

Water stress is WRI Aqueduct 4.0 for 2080, shown as a flag. Incoming migrants come from the same Now Cities gravity model the data explorer maps, summed over the footprint and a 5 km ring. What the model omits: defences, subsidence and storm surge; humidity in the heat clock; more than one climate model's pattern; and any adaptation term at all. Full derivation and schema in the methodology.