The Geography of Human Vulnerability Through 2100

The Shared Socioeconomic Pathways describe five futures for the world economy and society. Their population projections, from the Wittgenstein Centre's demographers, differ mainly in fertility, mortality and education. Under SSP1, a sustainable path with rapid gains in schooling, the world population on this grid peaks at about 8.5 billion around 2050 and falls to 6.9 billion by 2100. Under SSP2, the middle of the road, it peaks at 9.5 billion around 2070 and ends the century at 9.0 billion. Under SSP3, a fragmented world of regional rivalry and slow development, it keeps growing to 12.7 billion. SSP5, fossil-fuelled rapid development, has a population path close to SSP1's, peaking at 8.6 billion. The latest United Nations projections, published in 2024, lie between SSP2 and SSP3, with a peak of about 10.3 billion in the mid-2080s.

The redistribution is more certain than the totals. Under SSP2, sub-Saharan Africa's population more than doubles, from 1.1 billion in 2020 to 2.4 billion in 2100, overtaking South Asia's 2.3 billion; East Asia and the Pacific shrinks from 2.3 to 1.6 billion. Under SSP3, sub-Saharan Africa and South Asia each approach 3.6–3.7 billion. The share of people living between the tropics rises from about 41% in 2020 to 48% under SSP1, 51% under SSP2 and 55% under SSP3; only in SSP5, where fertility falls fastest in poor countries and high-income countries keep growing, does it stay near 43%.

A simple way to see the shift is the population's centre of gravity — the point on the Earth's surface closest to the average position of every person. In 2010 it lay in Central Asia, near Dushanbe. By 2100 it has moved about 1,700 kilometres south-west in every scenario: to western Iran under SSP1 and SSP2, and to the shores of the Persian Gulf under SSP3. The direction is the signature of Africa's growth; the distance is the signature of how much growth there is. Under SSP5 the centre moves mostly west, because the Americas and Europe grow while Asia shrinks.

Danger-level heat is defined here as at least 30 days a year on which the daily maximum heat index — the "feels-like" temperature combining heat and humidity — reaches 41 °C (106 °F). That is inside the "danger" band of the US National Weather Service's heat-index scale (103–124 °F), where heat exhaustion is likely and heat stroke possible with prolonged exposure. In the 1995–2014 baseline, about 90 million people lived with a month of such days — 77 million of them in Pakistan's Indus plain, the rest mainly in northern India, Yemen and the Gulf. With the 2020–2039 climate and the population of 2030, it is already about 0.3 billion.

By the 2090s the number is 0.9 billion under SSP1-2.6, 2.2 billion under SSP2-4.5, 3.8 billion under SSP5-8.5 and 5.0 billion under SSP3-7.0. Chronic drying — a 20-year average twelve-month Standardised Precipitation-Evapotranspiration Index of −1 or lower, meaning that the average year is as dry as a moderate drought in the reference climate — reaches 0.4, 1.4, 2.6 and 2.8 billion people. Markedly heavier downpours — a 20% or greater increase in the wettest five-day rainfall of the year — reach 0.2, 0.8, 3.6 and 4.1 billion. People exposed to at least two of the three number 0.1 billion under SSP1-2.6 and 3.1 billion under SSP3-7.0.

The ensemble spread matters. Taking the 10th and 90th percentiles of the 26 models in each grid cell, heat exposure in the 2090s under SSP2-4.5 ranges from 1.1 to 3.6 billion people, and under SSP3-7.0 from 2.5 to 7.1 billion. For drying and heavy rain the range is far wider — zero to 7.7 billion for rain under SSP2-4.5 — because models disagree on where precipitation will change. These cell-by-cell percentiles exaggerate the range, since no single model is at its extreme everywhere, but they show that heat is the most robust of the three projections, and water the least.

Choose a scenario, a period and a hazard. Each cell of about 200 by 200 kilometres is shaded by the number of people in it who live beyond the threshold, on a logarithmic scale; cells where people live but the threshold is not crossed are drawn in grey, so the map shows both where people are and where they are exposed. The "population change" layer shows instead where the population grows or shrinks between 2020 and the chosen period.

Under SSP2-4.5, the danger-heat zone of the 2090s is anchored on the Indo-Gangetic Plain from Punjab to Bangladesh and extends along the Indus, across the Gulf and Mesopotamia, into the Sahel and coastal West Africa, and over mainland Southeast Asia and the Yangtze and Pearl River lowlands. Under SSP3-7.0 it covers nearly all of Bangladesh, Pakistan and Cambodia, three-quarters of South Asia's people and much of West Africa from Senegal to Nigeria. Drying is concentrated in the Middle East and around the Mediterranean — Egypt, Iraq, Türkiye, Iran — and in Mexico and Afghanistan, spreading under SSP3-7.0 to southern Africa and Brazil; heavier downpours fall mainly across South Asia, the Sahel and West Africa.

Hover over or tap a cell to see its country card: population today and in the 2050s and 2090s under every pathway, the people exposed to each hazard, and the average years of adult schooling, the capacity proxy used in section 7. The totals in the corner are exact sums from the 0.25° grid; the map itself rounds each cell's exposed share to tenths.

Growth in exposure can be split into three parts, following the method Jones, O'Neill and colleagues used for heat in the United States. The climate effect is the baseline population multiplied by the change in where the hazard occurs: people already living somewhere that becomes dangerous. The population effect is the change in population multiplied by the baseline hazard: more people in places that are already dangerous. The interaction is the change in population multiplied by the change in hazard: new people in newly dangerous places. Because danger-level heat, drying and heavier rain are almost absent in the baseline, the pure population effect is small; the action is in the other two terms.

For heat under SSP2-4.5, the climate effect adds 1.6 billion people by the 2090s, the population effect 0.1 billion and the interaction 0.4 billion. Under SSP3-7.0 the interaction grows to 2.3 billion — 46% of the total of 5.0 billion, nearly as large as the climate effect of 2.4 billion. In sub-Saharan Africa it accounts for 78% of the 1.1 billion people exposed; in South Asia for 47% of 2.7 billion. In East Asia it is close to zero or negative, because population falls in the places that become hazardous. Under SSP5-8.5, where the climate is hotter but the population smaller, the interaction is only 0.2 billion.

The interaction term is the arithmetic form of the report's central point. Much of the exposure of the late century will be borne by people who do not yet exist, in towns and cities that do not yet exist, in places that are not yet dangerous. That makes it the most preventable part of the risk: the buildings, cities and infrastructure those people will live in have not been built, and the decisions that shape them — where cities grow, how they are designed, how fast fertility falls as schooling rises — are being made now.

The SSP scenarios usually pair a population with a matching emissions path. But the two can be separated: each of the four population grids can be combined with each of the four climates, giving sixteen worlds. The matrix shows people exposed in the 2090s in each. Reading across a row holds population fixed and changes the climate; reading down a column holds the climate fixed and changes the population.

For heat, the climate lever is the stronger. With SSP2's population, moving from the SSP5-8.5 climate to SSP1-2.6 cuts exposure by three-quarters, from 4.9 to 1.2 billion. But the population lever is large too: under any climate, SSP3's population puts about 1.8 to 1.9 times as many people in the heat as SSP1's. Starting from the SSP3-7.0 world of 5.0 billion, the SSP1-2.6 climate alone brings it to 1.8 billion, the SSP1 population alone to 2.7 billion, and both together to 0.9 billion.

The population lever is not about numbers alone. SSP1 and SSP5 have small populations because they assume rapid gains in education, especially for girls, which lowers fertility and also raises the capacity to cope with heat — the subject of section 7. Nor is it a lever that works against anyone: the low-population paths are those in which people are healthier, better educated and richer. For compound exposure to two or more hazards, the climate lever dominates: with SSP3's population it ranges from 0.3 billion under SSP1-2.6 to 5.7 billion under SSP5-8.5.

Humanity is already concentrated: in the 2090s under SSP2, half of all people live on the densest 3.0% of inhabited land and 80% on 11.5%. Exposure to danger-level heat is concentrated far more. Half of the 2.2 billion people exposed under SSP2-4.5 live on 0.5% of inhabited land, 80% on 1.4%, and all of them on 6.1%. Under SSP3-7.0, half live on 1.0% of the land. Compound exposure to two or more hazards is the most concentrated of all: under SSP2-4.5, half of it is on 0.2% of inhabited land.

The concentration is regional as well. Under SSP2-4.5, South Asia holds 72% of the people exposed to danger-level heat in the 2090s, East Asia and the Pacific 17% and sub-Saharan Africa 6%. India alone accounts for 1.1 billion, followed by Pakistan (0.29 billion), China (0.24 billion) and Bangladesh (0.16 billion). Under SSP3-7.0 sub-Saharan Africa's share rises to 22%, with Nigeria's 0.49 billion exposed people overtaking Pakistan's; for two or more hazards, South Asia holds 54%, sub-Saharan Africa 27% and the Middle East and North Africa 11%.

Concentration cuts both ways. It means that a small number of places — the Indo-Gangetic Plain, the Indus valley, the Sahel and northern Nigeria, the Gulf and Mesopotamia, the river deltas of Southeast Asia — will carry a disproportionate share of the world's climate risk, and that losses there could overwhelm local institutions. But it also means that adaptation can be targeted: heat-health systems, cooling, water security and resilient infrastructure focused on a few hundred districts would reach most of the people at risk.

Exposure is not destiny. Whether heat, drought or floods kill people, impoverish them or merely inconvenience them depends on housing, health care, early warning, social protection, cooling and income. A companion report found that within countries each extra year of adult schooling has been associated with about a third fewer disaster deaths per person affected. This section uses the average years of schooling of adults aged 25 and over, projected for each country under each SSP by the Wittgenstein Centre, as a simple proxy for the capacity to cope.

Today, 3.4 billion people live in countries where adults average fewer than eight years of schooling, and these countries hold more than nine-tenths of the 90 million people exposed to danger-level heat in the baseline. In every pathway schooling rises, but at very different speeds. Under SSP1, SSP2 and SSP5 almost all the people exposed in the 2090s live in countries where adults average ten or more years of schooling — about the level of upper-middle-income countries today. Under SSP3, 2.0 billion of the 5.0 billion exposed to danger-level heat live in countries below ten years, and 0.9 billion below eight.

That is the deepest difference between the pathways. SSP3 is not only the world with the most people in the heat; it is the world in which the people in the heat are least equipped to cope with it, and in which the exposed population grows fastest where schooling stalls: under SSP3 the countries with heat-exposed populations and adult schooling below eight years in the 2090s are Pakistan, Niger, Burkina Faso, Sudan, Mali, Senegal and Chad. By mid-century the gap is already visible: in the 2050s, 1.2 billion heat-exposed people live in countries below ten years of schooling under SSP3, against 0.2–0.4 billion in the other pathways.

The projections point to five priorities. The first is to protect the hotspots that are already forming. The Indo-Gangetic Plain, the Indus valley and the Gulf will cross the danger-heat threshold for hundreds of millions of people within two decades in every scenario. Heat action plans like Ahmedabad's, which combine early warnings, cooling centres, outreach to vulnerable groups and changes to working hours, are cheap and proven; so are cool roofs, shade and green space, efficient cooling and labour rules that keep outdoor workers out of the worst hours.

The second is to build the cities of the late century for the climate they will face. Much of the growth in exposure will happen in places that are still small towns or fields. Urban plans, building codes and infrastructure standards set now — for ventilation and shade, for drainage designed for heavier rain, for water supply that does not depend on a single drying source — will determine whether these cities are livable. Retrofitting after they are built costs far more. The third is to treat education and health as climate adaptation. The development path that shrinks exposure most is the one in which schooling, especially for girls, and health care spread fastest; it lowers fertility and raises the capacity to cope at the same time.

The fourth is to cut emissions, which remains the largest lever for heat, downpours and compound hazards, and the only one that shrinks the hazard zone itself. The fifth is to plan for movement. The World Bank's Groundswell reports project that up to 216 million people could move within their countries because of slow-onset climate impacts by 2050, and that early action could cut this by as much as 80%. Well-managed migration to safer, more productive places is a form of adaptation, and the population projections here do not yet include it.

Population: decadal 0.25° grids for 2010–2100 under SSP1, SSP2, SSP3 and SSP5 from the World Bank Climate Change Knowledge Portal (CCKP), regridded from the one-eighth-degree SSP projections of Jones and O'Neill, which downscale the national projections of KC and Lutz using the Gridded Population of the World v4 base map and include urbanisation but not climate-driven migration. Climate: CCKP CMIP6 0.25° bias-corrected climatologies for 1995–2014 and for 2020–39, 2040–59, 2060–79 and 2080–99 under SSP1-2.6, SSP2-4.5, SSP3-7.0 and SSP5-8.5, using the median of 26 models (and the 10th and 90th percentiles for the ranges). Each 20-year period is paired with the population of its central decade (2030, 2050, 2070, 2090); the baseline uses 2010.

Hazards: danger heat = 30 or more days a year with a daily maximum heat index of at least 41 °C (CCKP variable hi41); chronic drying = 20-year mean SPEI-12 of −1 or less (spei12); heavier downpours = annual maximum five-day precipitation at least 20% above the 1995–2014 value in the same cell (rx5day). Exposure is the population of cells beyond each threshold. Decomposition follows Jones et al. (2015). Concentration curves rank cells by exposed people per km². Countries are assigned by rasterising Natural Earth 50m boundaries, with unassigned coastal cells given the nearest country; regions follow the World Bank. Capacity: mean years of schooling of adults aged 25+ from the Wittgenstein Centre Human Capital Data Explorer v3.3 for SSP1, 2, 3 and 5, population-weighted across age groups.

Caveats: the SSP population grids total 7.6–7.7 billion in 2020 and differ from the UN's 2024 projections; climate-driven migration, sea-level rise, coastal and river flooding and tropical cyclones are excluded; heat-index thresholds ignore acclimatisation, shade and air conditioning; SPEI and Rx5day thresholds are relative to the reference climate, so their baseline exposure is zero by construction; cell-wise ensemble percentiles overstate the spread; schooling is a coarse, national proxy for capacity. Related reports: Human Exposure to the Global Polycrisis, 2025–2100, Human Exposure Across the Polycrisis Century and Human Survivability Under the Global Polycrisis.