Demographic Growth Inside the Global Risk Envelope

The world's urban centres held 1.63 billion people in 1975, 2.70 billion in 2000 and 3.69 billion in 2025, and the database projects 3.85 billion in 2030. The United Nations' 2025 World Urbanization Prospects, now using the same degree-of-urbanisation definition, counts 45% of the world's 8.2 billion people in cities and 33 megacities of 10 million or more, up from 8 in 1975. With world population projected to peak at about 10.3 billion in the 2080s, two-thirds of the growth to 2050 is expected to be urban.

The chart indexes each exposed population to 1975. The population in 100-year river flood zones grew fastest: 170 million in 1975, 338 million in 2000, 507 million in 2025 and 532 million in 2030, a factor of 3.0 to 2025. Population in 100-year coastal flood zones grew 2.5 times, to 106 million, and in the low-elevation coastal zone below 10 metres 2.4 times, to 349 million. Population in cyclone-wind zones and in zones of strong earthquake shaking grew more slowly than the urban total — by 2.0 and 1.9 times — to 169 and 227 million.

The hottest cities grew fastest of all. Cities with a mean annual temperature of 27 °C or more held 136 million people in 1975 and 413 million in 2025, three times as many. Warming moves that line: on the database's projected 2030 climate under a middle-of-the-road scenario, the population of cities above it roughly doubles again, to about 0.8 billion, though part of that jump reflects the switch from observed to modelled temperatures.

A simple test asks whether new urban residents settled in hazard zones more or less often than existing ones. The growth bias divides the share of population growth that occurred inside a zone by the zone's share of population at the start. A value of 1 means growth was neutral; above 1, it leaned towards the hazard. For river flood zones the value since 1975 is 1.56: 16.4% of the 2.06 billion people added to urban centres settled in a 100-year flood zone, against a starting share of 10.5%. Since 2000 the lean has eased slightly, to 1.37, but it persists.

The lean is strongest where growth was fastest and land was flattest. In East and South-East Asia, 24.6% of urban growth since 1975 settled in flood zones, against a starting share of 13.7% — a bias of 1.8. In Central and Southern Asia the bias is 1.48, in Northern Africa and Western Asia 1.34 and in sub-Saharan Africa 1.30; in Latin America (0.96) and North America (0.80) growth moved away from the floodplain. By income, the bias is 2.0 in upper-middle-income countries, led by China's coastal and river-delta cities, 1.36 in lower-middle-income and about 1.0 in low-income countries, and 0.77 in high-income ones.

These results echo the global study by Rentschler and colleagues, which found that settlements in the most hazardous flood zones expanded by 122% between 1985 and 2015, against 85% for all settlements, with the fastest growth in East Asia and 81% of new high-risk settlements in middle-income countries. Among the 439 cities of over a million people that added at least 200,000 residents since 2000, 44 put the majority of that growth in flood zones — including Shanghai, Bangkok, Suzhou and Baghdad. Remember that these are undefended zones: Shanghai's 25 million exposed residents sit behind extensive defences, while many smaller cities have none.

Low-lying coasts attracted growth roughly in proportion to their starting share. The urban population of the low-elevation coastal zone below 10 metres rose from 147 to 349 million between 1975 and 2025 — a growth bias of 1.08, or 0.96 since 2000. Coastal flood zones leaned slightly more (1.18), most sharply in Northern Africa and Western Asia, where the bias is 2.7, and in Europe (2.5) and Central and Southern Asia (1.6). Sea-level rise will enlarge these zones over the century, so the population already settled there is exposed to a moving boundary.

Cyclone wind zones and zones of strong earthquake shaking grew more slowly than the urban total, with biases of 0.79 and 0.75. That reflects geography more than caution: much of the century's urban growth has been in South Asia and Africa, away from the cyclone tracks of the western Pacific and the seismic belts of East Asia and the Americas. In Central and Southern Asia, where growth did reach a cyclone-exposed coast — Karachi above all — the cyclone bias is 1.8.

Still, 169 million urban residents live where cyclone winds are expected at least once a century, and 227 million where shaking strong enough to damage ordinary buildings is expected. These hazards are rare in any one year and catastrophic when they strike, which makes them the ones most easily forgotten in everyday settlement decisions — and the ones where building codes and enforcement do the most good.

Heat is the one hazard that every resident of a city shares. The database records each city's mean annual temperature from reanalysis data and a projection for 2030. Using 27 °C as a threshold — warmer than all but the hottest places in South Asia, the Sahel, the Gulf and mainland South-East Asia — the population of such cities tripled between 1975 and 2025, from 136 to 413 million, while urban population as a whole rose 2.3 times.

Low-income countries carry the largest share: 23% of their urban residents live in cities above the line, against 16% in lower-middle-income, 7% in upper-middle-income and 5% in high-income countries. Among cities born since 2000, 23% of residents live above it, against 20% in older cities. On the database's projected 2030 climate, the population of cities above 27 °C roughly doubles to about 0.8 billion, and the share of days that are unusually hot, by the standard of 1961–1990, rises from a median of 19% to 28%.

Heat compounds other hazards. Of the 413 million people in the hottest cities, 160 million also live in cities with large flood-zone populations and 79 million in cities with high levels of organised violence. Companion reports in this series follow heat stress in detail; here the point is spatial: the cities adding people fastest are, on average, among the hottest, so heat-resilient urban design — shade, ventilation, cool materials, green and blue space, and cooling for the most vulnerable — is a settlement decision too.

Counting the six flags for each city shows how hazards stack. In 2025, 1.67 billion urban residents (45%) lived in cities with none of them, 1.27 billion with one, 591 million with two, 142 million with three and 13 million with four or more. The share in cities with at least one rose from 47% in 1975 to 55% in 2025, and the share with two or more from 17% to 20%; on the projected 2030 climate, it reaches 23%, or about 0.9 billion people. Of the 2.06 billion people added to urban centres between 1975 and 2025, 61% settled in cities with at least one major hazard.

The commonest pairing is river flooding with coastal exposure — 322 million people live in cities with both, in the great deltas from the Nile to the Mekong and the Pearl. Floods combine with heat for 160 million, with organised violence for 133 million, and with cyclones for 79 million. Cyclones and violence coincide for 67 million, mostly in Manila and Karachi. The most multi-exposed large cities include Yangon (flood, coast, heat and violence), Tokyo and Nagoya (flood, coast and earthquake), Manila (flood, cyclone and violence), Basra, N'Djamena and Mandalay (flood, heat and violence) and Port-au-Prince (cyclone, earthquake and violence).

Compounding matters because the hazards interact and because the same city budget, the same drainage and the same hospitals must cope with all of them. A flood in a city already under heat stress and insecurity is harder to respond to and slower to recover from. And violence is a settlement driver in its own right: displacement by conflict pushes people into the cheapest, most exposed land at the urban edge, where planning is weakest.

The database records the year each urban centre first reached city density and size. Of the 11,422 centres in 2025, 3,911 crossed the threshold after 2000 — 1,076 in Central and Southern Asia, 861 in sub-Saharan Africa, 689 in East and South-East Asia, 497 in Northern Africa and Western Asia and 413 in Latin America — and they are home to 340 million people. Many are former towns and large villages in the densely settled plains of South Asia and the Nile, and fast-growing market towns across Africa.

These new cities are slightly more exposed to river flooding and heat than older ones — 15.1% of their residents live in flood zones against 13.6% in older cities, and 22.6% live above the 27 °C line against 19.7% — but less exposed to coasts, cyclones and earthquakes, and markedly less likely to stack two or more hazards (14% against 21%), because they are mostly inland and in regions with fewer coastal and seismic hazards.

New cities are where settlement decisions are cheapest to change. Their street grids, drainage and land markets are still forming; their flood plains are often not yet built on; their municipal institutions are being created. Planning that reserves floodplains for parks and agriculture, guides growth to safer, higher ground with trunk infrastructure, and sets building and drainage standards before the buildings go up costs a fraction of retrofitting or relocating later.

The atlas shows every urban centre with at least 300,000 people in 2025, plus smaller ones exposed to two or more hazards. Choose the year to size the points by population in 1975, 2000, 2025 or 2030 and watch the constellations of South Asia, eastern China, the Nile and West Africa brighten.

The default colour is the number of major hazards; the alternatives show the share of residents in river flood zones, how much of each city's growth since 2000 went into the flood zone, the decade in which it became a city, and its mean annual temperature in 2030.

Hover over or tap a city to see its card: population over time, the people in its flood zone, its exposure to coastal flooding, low elevation, cyclone wind and earthquake shaking, its temperature and whether it is flagged for violence.

The first lever is where growth goes. Risk-informed land-use plans and zoning that keep the most hazardous land — active floodplains, low coastal land, unstable slopes — for parks, agriculture and wetlands, and that steer growth to safer ground by building roads, water and transit there first, change settlement patterns at almost no cost. The flip side is making safe land affordable: much hazardous settlement is informal, driven by people who cannot afford anywhere else, so serviced land and housing finance on safer ground are risk-reduction tools.

The second is how cities are built where they already are. Most of the 507 million urban residents in flood zones will not move. For them, drainage designed for heavier rain, flood defences sized to growing cities, building codes and elevated ground floors, early warning, and room for rivers to spread safely reduce risk. Building codes enforced on new construction are the main defence against earthquake shaking and cyclone wind, and shade, ventilation, cool materials and green space against heat.

The third is information and incentives. Flood and heat maps published at the scale of neighbourhoods, disclosure of hazard to buyers and renters, insurance priced to risk where markets exist, and public investment that does not subsidise new building in hazard zones all shift private decisions. Where risk is already intolerable and rising, voluntary, well-supported relocation is an option of last resort. And because most of the urban growth still to come will be in South Asia and sub-Saharan Africa, where cities are growing fastest and planning capacity is thinnest, finance and technical support for municipal planning are among the highest-return investments in adaptation.

Data: GHSL Urban Centre Database R2024A (European Commission Joint Research Centre), thematic tables for general characteristics, GHSL population (GHS-POP R2023A, 1975–2030 in five-year epochs), exposure, climate, hazard-risk and socioeconomic indicators, for 11,422 urban centres defined by the degree of urbanisation within their 2025 boundaries. Exposure: population in the 10- and 100-year river flood zones (JRC/CEMS global flood hazard maps, undefended), in 20- and 100-year coastal flood zones (Vousdoukas et al.), below 5 and 10 m in the low-elevation coastal zone (CIESIN LECZ v3), in 100-year tropical-cyclone wind zones (STORM, categories 1–2), and in zones of MMI VIII or more shaking (GEM global seismic hazard map 2023.1). Climate: mean annual temperature (reanalysis, 2010 decade; SSP2-4.5 projection, 2030). Violence: battles, explosions and violence against civilians recorded within each centre in 2015 and 2020.

Growth bias = (Δ population in zone ÷ Δ total population) ÷ (zone population ÷ total population at the start). City flags as defined under section 4; hazard counts use each epoch's population shares, observed climate to 2025 and projected climate for 2030. New cities: year of birth (first epoch meeting the urban-centre criteria) of 2000 or later. Regions are UN SDG regions and income groups the World Bank's, as assigned in the database. City points: centres of ≥ 300,000 people in 2025, plus those of ≥ 100,000 with two or more hazards.

Caveats: flood zones ignore defences; all populations are counted within 2025 boundaries; the 2030 epoch is a projection; the 2030 temperatures come from a different source from the observed baseline, so part of the change in the heat line is methodological; the violence measure includes criminal and state violence and depends on reporting; and flags based on thresholds simplify continuous risks. Related reports: The Geography of Human Vulnerability Through 2100, Human Exposure to the Global Polycrisis, 2025–2100 and The Global Water Polycrisis.