A patient receives emergency treatment at a Karachi hospital during a life-threatening heatwave on June 27, 2024.
Summary
- Mortality depends on exposure, vulnerability and adaptive capacity, making precise death tolls impossible to model.
- The four billion deaths at 3°C figure is not a forecast or modelled projection.
- The estimate of two billion deaths at 2°C is implausible.
- The timeframe for the 3°C / four billion deaths estimate is undefined.
- The four billion figure is an educated worst-case estimate, not a robust scientific projection.
The proposition that four billion people or more would be dead if — and more likely, when — global warming reaches 3 degrees Celsius (°C) has gained some currency, mainly due to the 4DB (four billion dead) website and associated activities.
Now, that is half the current global population, and on current warming trends of 0.3-0.35°C/decade, the world will hit 3°C around 50 years from now, perhaps earlier, in part because there is little prospect of a rapid decline in fossil fuel emissions. So that’s a mind-blowing average of 80 million people a year dying due to climate impacts every year from now to 2075? Is that realistic?
The number of greater than four billion dead at 3°C first appeared in a 2025 report on Planetary Insolvency: Finding our balance with nature, published by the UK Institute and Faculty of Actuaries (IFoA) and the University of Exeter.
The number appears in Figure 12: Planetary solvency risk impact and likelihood definitions (illustrative) on page 32 of the report. In this figure, the scale of impacts listed for 2°C of warming include mortality of “>2 billion deaths” and GDP losses of 25%, as well as “>4 billion deaths” and GDP of 50% for 3°C of warming (Figure 1).
Figure 1: Planetary solvency risk impact and likelihood definitions
The 2°C estimate is even more startling, because Earth has already hit 1.5°C for all practical purposes, with the average for 2023-25 above 1.5°C, and a strong El Nino likely on the way for later this year which could push annual warming towards 1.7°C for 2026-27. At the current, accelerated warming rate, the Earth will reach 2°C around 2040, just fifteen years from now.
So the figure of two billion dead at 2°C would mean an average of 130 million people a year dying every year from now till 2040 due to climate change? To my mind, that is simply not credible, and nothing I have read tells me that is a remotely-likely estimate.
Likewise, the risk matrix gives a figure of 1-5% dead (80 to 400 million deaths) for 1.5°C, a level of warming Earth has already reached. Estimates of actual mortality are difficult due to the direct and second- and third-order impacts, such as: hotter-climate > drought > food-shortage > displacement > conflict > mortality, and so on. The 2025 report of the Lancet Countdown on Health and Climate Change estimated heat-related mortality has increased to an average 546,000 deaths per year, though not all are specifically related to climate heating, and that is only one element of the story.
So it is important to understand what these numbers are, and are not, and where they came from.
There is no qualitative indication in the report as to their source. On page 27, the report acknowledges that “very limited research has been carried out on the potential for large-scale loss of life in relation to these interconnected risks on which to base an assessment”.
I asked a colleague in the UK who had worked with the authors of the report, and he was told that the figures came from the Climate Endgame paper. But its author, Luke Kemp, said this was not the case, and “he confirmed that he does not make any forecasts about mortality in the paper” (emphasis added).
As the caption to Figure 12 says, it is a risk assessment matrix that illustrates general levels of risk. I was told that the authors have now privately clarified that the mortality numbers “are absolutely NOT forecasts or predictions and we don’t use them as such” (emphasis added). But this is now the way that sites like 4BD are using it.
One interpretation would be that the risk matrix was simply indicating orders of magnitude rather than specific projections. So either it was too subtle in its distinctions, and/or it was a bit of a stuff-up in that it was assumed that the four billion figure came from somewhere, but it did not. So there is still a question of how the four billion figure was derived.
I understand there have been suggestions that a correction or clarification be issued, at least saying that the figures “are absolutely NOT forecasts or predictions”. And I had conversations with people working on the 4BD project outlining the story above and why some nuance was necessary in using such numbers, but the die had been cast.
So how can we think about this? One guess is the table was saying that at 3°C, an expert elicitation would find that mortality would likely be in the billions, rather than tens or hundreds of millions. Or alternatively, this was a plausible worst-case scenario, but not derived from models. As discussed below, that is a reasonable proposition based on other literature. As for two billion dead at 2°C, I can see almost no credible evidence that is even in the ballpark, just half a degree warmer than at present.
Another, likely interpretation is that Figure 12 had used some numbers for GDP loss (in column 1) and had simply applied the same number for mortality (in column 2), so 50% loss in GDP equals 50% mortality. This in itself is a brave assumption: in one case, during the Great Depression — when, for example, US GDP fell 30% between 1929 and 1933 — mortality rates did not increase, and in some cases improved, though there were later adverse outcomes for children born at that time.
So there is no mortality analysis in the table at all, and no epidemiology; it simply flows from GDP impacts.
So where did the 50% decrease in GDP at 3°C come from? An earlier IFoA report in 2023, The Emperor’s New Climate Scenarios, whilst recognising that “climate change is complex, nuanced and characterised by deep uncertainty”, provided a chart of damage functions relating temperature and GDP loss (Figure 9, page 25), and asks “at what point do we expect 50% GDP destruction – somewhere between 2070 and 2090 depending on how you parameterise the distribution”, and that’s around 3°C. Depending on assumptions about when GDP hits zero (called the ruin parameter), the figures could be higher (80%) or lower (30%) because these are abstract models of possible damage functions. So it looks like a 50% loss of GDP was a figure of choice.
So what is the basis of the damage functions? The Emperor’s New Climate Scenarios says that “Insurance leaders have unequivocally stated that if climate change raises average temperatures to 4˚C above pre-industrial levels most assets will be uninsurable”, and that “without insurance, investment, finance, business slow to a halt – we will no longer have an economy.” (page 27). I find this assumption highly problematic. As insurance premiums rise right now, in some cases dramatically due to extreme climate impacts, many households and small businesses are making decisions to continue to operate without asset insurance.
The only footnote to these statements is a view by one insurance CEO, Thomas Buberl of AXA at a Davos panel, reported by Bloomberg, that at 3-4°C “it’s not insurable anymore”, but what he said was much more specific: he was talking about basement retail premises in New York and Mumbai and that is what the headline said: “Climate change could make your basement uninsurable within a decade”. That is a very narrow and specific base on which to draw a global conclusion.
Now, how much of an economy would be left at 3, 4 or 5°C is a good question, but the method used here is far too narrow. I remember a long time ago James Lovelock saying at 4 or 5°C there may be 500 million people left eking out a miserable existence at the poles, or words to that effect.
To reiterate, it is almost impossible to put specific numbers on such future impacts, due to the radical uncertainty about the social impacts of a physical system that itself is non-linear in many important aspects, characterised by abrupt changes whose specific human consequences are somewhere between difficult and impossible to model, and where quantifying social impacts is of limited efficacy.
Vulnerability and adaptation
Even if future (non-linear) physical changes are well known, mapping their human impacts involves several more degrees of difficulty because the risk (potential damage) varies with exposure and vulnerability. There are three factors:
- Hazard: the physical changes in a climate system subject to abrupt change;
- Exposure: the presence of people, livelihoods and ecosystems in that physical space; and
- Vulnerability: the propensity of these human systems to be negatively impacted due to their sensitivity and/or limited adaptive capacity.
In the uber-rich, low-rainfall Gulf states, for example, whilst unlivable heat is becoming the norm, adaptation paid for by stupendous oil and gas revenue — desalination, 24/7 air conditioning, using flood-lit beaches at night rather than during the day, irrigating date palms, importing almost all the food and most of the labour — reduces vulnerability, even though the whole project seems a bit crazy.
So mortality in a hotter climate will be affected a great deal by adaptation capacities to reduce vulnerability, and that is largely a product of national income, and international assistance.
But there are also hard boundaries that cannot be easily adapted to, for example rice yields diminish once temperatures exceed 35°C at the time of flowering, and by 37°C the damage becomes critical. By 2050, between one-quarter and two-thirds of rice production capacity will be subject to high or extreme heat stress risk (Figure 2).
Smart adaptation would include moving to a plant-based diet to free a great deal of agricultural land now used to grow livestock and grow feed for livestock, and to reduce methane emissions. A well-planned and managed retreat from low-lying coastal land would also help, rather than just waiting for the inevitable to happen. And it depends on the nature of international politics. A coordinated global mobilisation to face the coming crisis would yield a very different result from states pretending that climate collapse was simply not on the agenda. Will states cooperate in the face of unprecedented adversity, or close borders and go to war?
The timeline
Another unresolved issue is the time frame. Many impacts do not manifest immediately once a certain temperature level is reached. For example, climate history teaches us that every one degree of warming will likely result in 10-to-20 metres of sea-level rise in the longer term, over many centuries. The fastest rises in the paleoclimate record are three-to-five metres in a century, so the full sea-level rise from 3°C of warming could take a thousand years to manifest, perhaps less and perhaps a good deal more.
So when the figure of “4 billion dead” is raised, is that when the thermometer ticks past the figure “3°C” later this century, or over a thousand years as coastlines, agriculturally-rich deltas and low-lying states drown? My hunch is that it would be the latter.
A similar issue is the collapse of the Atlantic Meridional Overturning Circulation. Whilst a number of recent research papers conclude this tipping point is close at hand, models suggest the process is likely to take 100 years, so once again the exposure increases over that time, and so does the opportunity to reduce vulnerability with adaptive measures, though only up to a point.
Many other system-level changes have similar characteristics.
The world at 3°C
In a 3°C hotter world, large parts of the tropics will suffer “near-unlivable” extreme heat conditions, there will be less rainfall over significant parts of the dry subtropics, and this combined with increased evaporation rates will lead to drying out and desertification across the dry subtropics. New extremes — of rainfall and heat, flooding and drought — beyond human experience and beyond model expectations will occur. And a committed sea-level rise of many metres will be in the slow process of inundating coastal cities and deltas.
Some research, which is contested, suggests 3°C could cut global GDP by half. Last year, Australian researchers concluded that at 3°C, by the end of the century, the estimated harm to the global economy would be 40%, which “could devastate livelihoods in large parts of the world”. Coral reef systems would be gone, and that alone means coastal ecosystems would only be able to provide 20–50% of the fish protein that they do today for half a billion people around the world.
In 2021, the Australian Academy of Science published a report co-authored by David Karoly on Risks to Australia of a 3-degree warmer world. Amongst other things, it said that at 3°C, heatwaves would happen as often as seven times a year, with events lasting 16 days on average, “fire risk (driven by record heat, dryness and fuel) will increase by 30 per cent or more in south-eastern Australia”, and yields of key crops would reduce “by between 5 and 50%, depending on crop and location”. Other research estimates that “beyond 2°C warming, the declines in suitable areas for the 30 crops [analysed] become more pronounced – in some cases approaching and passing 50%”. That in itself would cause global chaos.
A 2020 study on extreme heat found that at 2.7°C, up to 3.5 billion people will be exposed to temperatures outside the “human niche”, that is, the climate conditions that have served humanity well over the past 6000 years. Further research published in 2023 described a zone of near unlivable heat, “a situation found in the present climate only in 0.8% of the global land surface, mostly concentrated in the Sahara, but in 2070 projected to cover 19% of the global land” (Figure 3). Prof. Marten Scheffer said those pushed outside the climate niche might consider migrating to cooler places: “Not just migration of tens of millions of people but it might be a billion or so.”
Figure 3: Projected zone of heat of “near-unliveable conditions” at 2.7°C global average warming (“Quantifying the human cost of global warming”)
And at 2.7°C, scientists say that the Arctic would be “transformed beyond contemporary recognition: the Arctic Ocean would be essentially ice free for several months in summer, the area of Greenland that reaches melting temperatures for at least a month would roughly quadruple, and the area of permafrost would be roughly half of what it was in preindustrial times.”
Potsdam Institute Director Prof. Johan Rockstrom says that such a level of warming is “something that humanity has absolutely no evidence that we can cope with… Push ourselves to 2.5°C – we’re in unknown terrain. It would lead to a complete melting of the big ice sheets, which would be a 10-metre sea level rise… There would be a collapse of all the big biomes on planet Earth – the rainforest, many of the temperate forests – abrupt thawing of permafrost, we will have complete collapse of marine biology, we will have a shift of large parts of the habitability on Earth.”
Twenty years ago, a group of US security analysts constructed a 3°C scenario:
“Massive nonlinear events in the global environment give rise to massive nonlinear societal events. In this scenario, nations around the world will be overwhelmed by the scale of change and pernicious challenges, such as pandemic disease. The internal cohesion of nations will be under great stress, including in the United States, both as a result of a dramatic rise in migration and changes in agricultural patterns and water availability. The flooding of coastal communities around the world, especially in the Netherlands, the United States, South Asia, and China, has the potential to challenge regional and even national identities. Armed conflict between nations over resources, such as the Nile and its tributaries, is likely and nuclear war is possible. The social consequences range from increased religious fervour to outright chaos. In this scenario, climate change provokes a permanent shift in the relationship of humankind to nature’ (emphasis added).”
So how many would die, and how many would survive in this 3°C world?
Displacement and mortality
The “human niche” papers on near-unliveable heat found that these extremes are projected to envelop 1.2 billion people in India, 485 million in Nigeria and more than 100 million in each of Pakistan, Indonesia and Sudan. Many would be forced to move to a more liveable climate. Another study from the same year of 2020 concluded that warming of 2°C could provide more than 500 million people additional incentive to emigrate, whilst warming of 3°C could provide additional incentive-to-emigrate to well over a billion people.
The idea that a billion people may be displaced may seem fanciful, but the UN also agrees with this figure: “Unless we change the way we manage our land, in the next 30 years we may leave a billion or more vulnerable poor people with little choice but to fight or flee.”
The figures may be much, much higher than this, but there are so many variables, so many social unknowns, that any figure can only be a guesstimate by social scientists and security analysts, based on a scientifically-credible scenario.
Climate disruption kills people in many ways, including by direct physical impacts (heat stress, floods, cyclones and other extreme events), by severe food insecurity, through displacement and conflict, by increased disease and poorer health outcomes, and so on.
The 2025 report of the Lancet Countdown on health and climate change reported a 63% increase in heat-related deaths since the 1990s, reaching an estimated 546,000 yearly deaths on average in 2012–21. The higher number of heatwave days and drought months in 2023 compared with 1981–2010 was associated with 123.7 million more people experiencing moderate or severe food insecurity in 124 countries analysed. As hotter and drier weather increases the risk of wildfires, 2024 had a record-high 154,000 deaths from wildfire smoke-derived small particulate matter (PM2·5) air pollution.
What about future mortality rates? Joshua Pearce of the University of Western Ontario says if warming reaches or exceeds 2°C, it is likely that mainly richer humans will be responsible for the death of roughly one billion mainly poorer humans over the next century. This is based on a review of the literature by Pearce and Parncutt, which found mortality costs of carbon emissions converged on the “1,000-ton rule”: an estimate that one future premature death is caused every time approximately 1,000 tons of fossil carbon are burned. Their best- and worst-case figures were 300 million and three billion for 2°C.
By this logic, the mortality range at 3°C would be in the range of 450 million to 4.5 billion. Of course, this assumes a linear relationship between emissions and mortality, but that assumption is based on observed conditions that are far different from those humans will face in the second half of the century.
It should also be noted that in 2019, Rockström told The Guardian that in a 4°C-warmer world: “It’s difficult to see how we could accommodate eight billion people or maybe even half of that. There will be a rich minority of people who survive with modern lifestyles, no doubt, but it will be a turbulent, conflict-ridden world.”
So four billion dead at 3°C? Immediately or when the full physical and social impacts are realised over many centuries? Depending on deeply uncertain adaptation and other political responses? Between expert estimates of degrees of magnitude and worse-case conjecture, it’s anybody’s educated guess.