community3.earth
Where people can still live as the world heats
community3.earth helps towns and small communities get ready before the world is 3 °C warmer than pre-industrial times.
- Heat past the limit, a month a yeara week a year
- Food and water failure
- Drowned coast
- Desert and dry steppe today
Fly to
What each level of warming leads to
Held at 3.0 °C for a very long time
- of the world’s land newly fails one of the three tests
- 4.6%
- people live there today
- 1.6 billion
A long-run picture, not a forecast. This is where each level of warming leads if it is held for a very long time. We are not saying these temperatures will be reached.
Above 4 °C the heat shading is our own extrapolation from published results at 1 to 4 °C.
1.5 °C, the lowest setting, is about where single years are now (Berkeley Earth).
What the colours mean
The first three are the tests. Each line gives the share of the world’s land and the people who live there today.
Lethal heat without reliable powered cooling
0.3% of land220 million people
A month or more of hours each year beyond the measured limit at which a healthy young adult, in shade and at light activity, can keep a steady body temperature. Past it the body goes on heating for as long as the exposure lasts, so everyone depends on powered cooling.
Lighter red: a week or more of such hours each year. Both reds together: 2.8% of land, 1.6 billion people.
Food and water failure
1.0% of land55 million people
Too dry or too hot to grow staple crops or supply water locally.
Dark red: fails the heat test as well.
Drowned coast
3.4% of land1.4 billion people
Below the sea level that this much warming eventually commits us to. At 3.0 °C: about 18 m (range 10 to 24 m).
Deserts and dry steppe today
25% of land560 million people
Land that already fails the food and water test today, or does by 1.5 °C: deserts, and some dry steppe that is grazed and farmed now. Not counted in the headline figures.
People are counted where they live today (GHS-POP, 2020).Heat: the limit measured by Vecellio and colleagues (2022), mapped as hours a year by Vecellio and colleagues (2023) up to 4 °C; our extrapolation above.Food and water: WorldClim 2.1 climate, warmed with CMIP6 model patterns.Sea level follows a straight line through the middle of the ranges the IPCC assessed in 2021 for about ten thousand years after the warming. The range at 1.5 °C is widened by us (the IPCC gives 6 to 7 m), and above 5 °C it is our own extension.Earth imagery: Solar System Scope, CC BY 4.0, based on NASA imagery; modified.
What it means
What “uninhabitable” means here
On this map, a place is uninhabitable when people could not go on living there in the long run, with the global temperature held at that level for a very long time.
A place fails in one of two ways.
- It can no longer supply food or water once supply chains break.
- Heat is lethal without powered cooling, and the power cannot be relied on.
Coast is shaded separately, where melting ice would in time put it under the sea. Each of the three is a test we can measure from published data.
The three tests
Test 1
Lethal heat without reliable powered cooling
Wet-bulb temperature measures heat and humidity together. Past a certain point sweat can no longer shed the body’s heat, even in the shade, and its temperature keeps rising. Air conditioning protects people only while the power holds.
30.6°C wet-bulbthe measured limit in humid heat. It is lower in dry heat.
The rule we use
Shaded where at least 240 hours a year, a month of eight-hour days, go beyond the limit measured for healthy young adults at light activity in shade: the point where the body can no longer hold its temperature steady.
Sources: Limit: Vecellio and colleagues, 2022. Map of hours: Vecellio and colleagues, 2023, PNAS
Test 2
Food and water failure
The land is too dry, or its growing season too hot, to feed and water the people on it. They would depend entirely on supply from elsewhere.
0.20aridity indexa year of rain divided by what the air could evaporate.
The rule we use
Shaded where the aridity index is below 0.20, or no three-month growing window has rain of at least half of evaporative demand, or every growing window averages daily highs of 36 °C or more.
Source: WorldClim 2.1 climate, warmed with CMIP6 model patterns
Test 3
Drowned by sea-level rise
Ice sheets and the deep ocean take thousands of years to finish responding to warming. This is the coast that ends up under the sea once they have.
18m at 3 °Cthe central long-run figure. The assessed range is 10 to 24 m.
The rule we use
Shaded where land lies below the sea level that the warming level commits us to over roughly 10,000 years, following the range assessed by the IPCC.
Land that fails both the heat test and the food and water test is drawn in dark red. A lighter red marks places with a week or more of such heat each year. The published heat data stop at 4 °C. Above that, the heat layer is our own extrapolation.
What this map is not
A map like this is easy to over-read. These are the limits we ask you to keep in mind while you use it.
Methods and sourcesNot a forecast for a date. The map carries no years. It shows where each level of warming leads if it is held for a very long time: heat and drying as climate models have them when the level is first reached, sea level after about ten thousand years.
Not a claim that these temperatures will be reached. It shows what published data say could follow if they are.
Not a prediction that people will leave. People live today in places that fail these tests, on imported food, river water and air conditioning. The map shows where that dependence becomes total.
Not precise at the scale of a town. Each cell of climate data covers tens of kilometres. A single town needs its own assessment.
Not all caused by warming. Deserts were already deserts. By 1.5 °C, 25% of the land outside Antarctica fails the food and water test, almost all of it already in the 1970–2000 climate record we start from (WorldClim 2.1). It is drawn separately, as deserts and dry steppe today. Some of it is semi-arid steppe that is grazed and farmed now, not desert.
Not a list of safe places. The map tests three things. Unshaded land still faces fire, river floods, storms, drought years and heat below these limits.
Evidence
The pace of warming has picked up
The planet is warming faster than it was. Over the last fifteen years it has also warmed faster than the central projections said it would. Scientists do not agree on how long that will last.
1Measured
What the data show
Berkeley Earth combines temperature readings from weather stations with sea-surface measurements from ships and buoys into one global record that starts in 1850. Temperatures here are the rise above the 1850 to 1900 average, a period often used as the pre-industrial baseline.
Berkeley Earth’s published figures
- September 2026. The second warmest September on record, behind September 2023.
- 1.65 ± 0.11°C
- The 12 months from October 2025 to September 2026. The second warmest such period on record.
- 1.458 ± 0.074°C
- The chance that 2026 becomes the warmest year on record.
- About 96%
Part of this heat is temporary. El Niño is a natural warming of the tropical Pacific that lifts the global average while it lasts. La Niña is its cool opposite. El Niño has now reached record strength on the conventional Niño 3.4 index, a measure of sea temperature in the central tropical Pacific. Berkeley Earth expects record and near-record warmth to continue through the end of the year “due to the short-term influence of a record-breaking El Niño.”
Source: Berkeley Earth, “September 2026 Temperature Update”, 22 October 2026 (opens in a new tab). Ranges are 95% confidence intervals: the span the true value very likely lies within.
Figure 1
The record
Global average temperature, 1850 to October 2026, in °C above the 1850–1900 average
- Yearly average
- 2026: January to October only
- 10-year average
- Latest 12 months
- 3.0 °C, the level we prepare by
Bars are calendar-year averages. The line is the 10-year running average, plotted at the middle of each ten years, so it ends in late 2021 at 1.32 °C. The ring is the average of the 12 months from November 2025 to October 2026: 1.53 °C.
2026 is a partial year: January to October only, shown as a paler bar. The latest month is preliminary. Dashed lines mark 1.5, 2.0 and 3.0 °C.
Our values run a little above Berkeley Earth’s printed figures. We re-baselined the data ourselves, and the result sits 0.02 to 0.04 °C above the figures Berkeley Earth prints in its reports: for the 12 months to September 2026 it prints 1.458 °C, and the same months come to 1.50 °C here. We have not established why.
Bar chart of the global average temperature for each year from 1850 to 2026, in degrees Celsius above the 1850 to 1900 average, with a line for the 10-year running average. 1900: 0.17. 1950: 0.10. 1980: 0.57. 2000: 0.69. 2010: 1.02. 2020: 1.28. The three warmest complete years are 2024 at 1.56, 2023 at 1.49, 2025 at 1.46. January to October 2026 averages 1.55. The latest 12 months average 1.53. The latest 10-year average, centred on late 2021, is 1.32. Dashed lines mark 1.5, 2.0 and 3.0 degrees; 3.0 is the level we want communities to be ready before.
Show the numbers: every year
| Year | Yearly average | 10-year average |
|---|---|---|
| 2026 (10 months) | 1.55 | – |
| 2025 | 1.46 | – |
| 2024 | 1.56 | – |
| 2023 | 1.49 | – |
| 2022 | 1.18 | – |
| 2021 | 1.13 | 1.30 |
| 2020 | 1.28 | 1.29 |
| 2019 | 1.25 | 1.24 |
| 2018 | 1.13 | 1.18 |
| 2017 | 1.20 | 1.15 |
| 2016 | 1.31 | 1.12 |
| 2015 | 1.17 | 1.10 |
| 2014 | 1.01 | 1.07 |
| 2013 | 0.95 | 1.04 |
| 2012 | 0.93 | 1.01 |
| 2011 | 0.90 | 0.98 |
| 2010 | 1.02 | 0.94 |
| 2009 | 0.94 | 0.93 |
| 2008 | 0.80 | 0.92 |
| 2007 | 0.93 | 0.91 |
| 2006 | 0.93 | 0.91 |
| 2005 | 0.98 | 0.89 |
| 2004 | 0.82 | 0.86 |
| 2003 | 0.90 | 0.85 |
| 2002 | 0.92 | 0.86 |
| 2001 | 0.84 | 0.83 |
| 2000 | 0.69 | 0.81 |
| 1999 | 0.69 | 0.79 |
| 1998 | 0.93 | 0.75 |
| 1997 | 0.78 | 0.72 |
| 1996 | 0.65 | 0.69 |
| 1995 | 0.77 | 0.69 |
| 1994 | 0.60 | 0.69 |
| 1993 | 0.55 | 0.66 |
| 1992 | 0.51 | 0.63 |
| 1991 | 0.70 | 0.61 |
| 1990 | 0.72 | 0.59 |
| 1989 | 0.52 | 0.56 |
| 1988 | 0.64 | 0.55 |
| 1987 | 0.60 | 0.56 |
| 1986 | 0.49 | 0.54 |
| 1985 | 0.39 | 0.53 |
| 1984 | 0.41 | 0.52 |
| 1983 | 0.58 | 0.50 |
| 1982 | 0.40 | 0.47 |
| 1981 | 0.61 | 0.46 |
| 1980 | 0.57 | 0.43 |
| 1979 | 0.44 | 0.41 |
| 1978 | 0.35 | 0.39 |
| 1977 | 0.47 | 0.38 |
| 1976 | 0.15 | 0.35 |
| 1975 | 0.23 | 0.31 |
| 1974 | 0.17 | 0.28 |
| 1973 | 0.41 | 0.27 |
| 1972 | 0.27 | 0.25 |
| 1971 | 0.14 | 0.23 |
| 1970 | 0.25 | 0.24 |
| 1969 | 0.30 | 0.22 |
| 1968 | 0.16 | 0.22 |
| 1967 | 0.22 | 0.20 |
| 1966 | 0.21 | 0.21 |
| 1965 | 0.14 | 0.23 |
| 1964 | 0.04 | 0.21 |
| 1963 | 0.31 | 0.22 |
| 1962 | 0.26 | 0.24 |
| 1961 | 0.32 | 0.23 |
| 1960 | 0.23 | 0.23 |
| 1959 | 0.29 | 0.24 |
| 1958 | 0.32 | 0.24 |
| 1957 | 0.30 | 0.26 |
| 1956 | 0.06 | 0.27 |
| 1955 | 0.13 | 0.25 |
| 1954 | 0.21 | 0.24 |
| 1953 | 0.41 | 0.22 |
| 1952 | 0.34 | 0.21 |
| 1951 | 0.25 | 0.22 |
| 1950 | 0.10 | 0.24 |
| 1949 | 0.17 | 0.26 |
| 1948 | 0.20 | 0.27 |
| 1947 | 0.26 | 0.25 |
| 1946 | 0.21 | 0.25 |
| 1945 | 0.33 | 0.28 |
| 1944 | 0.43 | 0.31 |
| 1943 | 0.31 | 0.33 |
| 1942 | 0.28 | 0.34 |
| 1941 | 0.35 | 0.35 |
| 1940 | 0.46 | 0.33 |
| 1939 | 0.35 | 0.32 |
| 1938 | 0.34 | 0.29 |
| 1937 | 0.36 | 0.27 |
| 1936 | 0.21 | 0.26 |
| 1935 | 0.16 | 0.25 |
| 1934 | 0.20 | 0.21 |
| 1933 | 0.06 | 0.18 |
| 1932 | 0.18 | 0.16 |
| 1931 | 0.23 | 0.14 |
| 1930 | 0.19 | 0.15 |
| 1929 | −0.03 | 0.12 |
| 1928 | 0.13 | 0.12 |
| 1927 | 0.13 | 0.11 |
| 1926 | 0.21 | 0.09 |
| 1925 | 0.07 | 0.07 |
| 1924 | 0.03 | 0.07 |
| 1923 | 0.04 | 0.07 |
| 1922 | −0.01 | 0.03 |
| 1921 | 0.09 | 0.00 |
| 1920 | 0.04 | −0.02 |
| 1919 | −0.02 | 0.00 |
| 1918 | −0.09 | −0.01 |
| 1917 | −0.22 | −0.03 |
| 1916 | −0.09 | −0.04 |
| 1915 | 0.16 | −0.06 |
| 1914 | 0.08 | −0.08 |
| 1913 | −0.09 | −0.10 |
| 1912 | −0.09 | −0.10 |
| 1911 | −0.19 | −0.09 |
| 1910 | −0.18 | −0.09 |
| 1909 | −0.20 | −0.12 |
| 1908 | −0.19 | −0.14 |
| 1907 | −0.15 | −0.14 |
| 1906 | 0.04 | −0.13 |
| 1905 | −0.06 | −0.09 |
| 1904 | −0.21 | −0.06 |
| 1903 | −0.17 | −0.04 |
| 1902 | −0.05 | −0.02 |
| 1901 | 0.09 | 0.00 |
| 1900 | 0.17 | 0.00 |
| 1899 | 0.07 | 0.01 |
| 1898 | −0.06 | 0.03 |
| 1897 | 0.13 | 0.01 |
| 1896 | 0.11 | 0.01 |
| 1895 | −0.02 | −0.02 |
| 1894 | −0.11 | −0.04 |
| 1893 | −0.12 | −0.02 |
| 1892 | −0.13 | −0.03 |
| 1891 | −0.04 | −0.07 |
| 1890 | −0.15 | −0.08 |
| 1889 | 0.13 | −0.10 |
| 1888 | 0.00 | −0.09 |
| 1887 | −0.20 | −0.07 |
| 1886 | −0.14 | −0.07 |
| 1885 | −0.14 | −0.04 |
| 1884 | −0.15 | −0.04 |
| 1883 | −0.02 | −0.04 |
| 1882 | 0.02 | 0.02 |
| 1881 | 0.09 | 0.05 |
| 1880 | 0.01 | 0.06 |
| 1879 | 0.08 | 0.07 |
| 1878 | 0.36 | 0.08 |
| 1877 | 0.30 | 0.08 |
| 1876 | −0.03 | 0.08 |
| 1875 | 0.00 | 0.07 |
| 1874 | −0.09 | 0.07 |
| 1873 | 0.01 | 0.07 |
| 1872 | 0.02 | 0.03 |
| 1871 | 0.01 | 0.03 |
| 1870 | 0.03 | 0.04 |
| 1869 | 0.09 | 0.05 |
| 1868 | 0.12 | 0.04 |
| 1867 | 0.07 | 0.02 |
| 1866 | 0.08 | 0.02 |
| 1865 | 0.09 | 0.00 |
| 1864 | −0.07 | 0.00 |
| 1863 | −0.10 | −0.02 |
| 1862 | −0.15 | −0.04 |
| 1861 | −0.10 | −0.07 |
| 1860 | −0.02 | −0.07 |
| 1859 | −0.01 | −0.07 |
| 1858 | −0.10 | −0.06 |
| 1857 | −0.17 | −0.03 |
| 1856 | −0.07 | −0.02 |
| 1855 | 0.08 | −0.01 |
| 1854 | 0.03 | – |
| 1853 | 0.06 | – |
| 1852 | 0.05 | – |
| 1851 | 0.04 | – |
| 1850 | −0.07 | – |
Data: Berkeley Earth (www.berkeleyearth.org (opens in a new tab)), high-resolution global temperature record (beta release), retrieved 16 November 2026. CC BY-NC 4.0 (opens in a new tab). Chart and re-baselining to 1850–1900 by community3.earth.
Our analysis of their data
- The warming rate from 2010 to October 2026, against the rate from 1970 to 2009.
- 0.37 against 0.19°C per decade
- From 1970 to 2009 the world warmed at 0.19 ± 0.04 °C per decade. From 2010 to October 2026 the rate was 0.37 ± 0.15 °C per decade, about twice as fast.
- The difference is unlikely to be chance: it passes a 95% significance test with the data to October 2026. Cut the record at the end of 2024 and it falls just short. Cut it at the end of 2022 and the test finds nothing.
- The record ends during a record El Niño, and the recent windows begin in the cooler La Niña years of 2011 and 2012. That flatters the latest rates.
- We did not remove El Niño, volcanic or solar effects from the data. Foster and Rahmstorf did. Their result is in part 2.
Computed by us from Berkeley Earth’s monthly data file. Berkeley Earth is not responsible for these numbers. Each ± is two standard errors, a measure of how uncertain a fitted rate is. We widened them to allow for autocorrelation: one month’s temperature resembles the last, so months are not independent measurements.
Figure 2
The pace
Warming rate over the preceding 15 years, in °C per decade
- Rate over the preceding 15 years
- ± two standard errors
- 1970–2009 rate
Each point is the slope of a straight line fitted to 180 months of temperatures, plotted at the month the window ends. The band is ± two standard errors, widened to allow for autocorrelation. The dashed line is the rate from 1970 to 2009, 0.19 °C per decade.
The latest window, November 2011 to October 2026, gives 0.40 ± 0.19 °C per decade, the highest of the 163 windows. Before 2015 the highest was 0.28.
Caveat. The record ends in a record El Niño, and the latest windows start in the cooler La Niña years of 2011 and 2012. That flatters them. A 15-year window is short: look at how far the line has swung before.
Line chart of the warming rate over rolling 15-year windows, in degrees Celsius per decade, with a band of plus or minus two standard errors. The first window ends in December 1864 and the last in October 2026. For windows ending before 2015 the rate ranges from −0.23 to 0.28, the high point being the window that ends in December 2006. The latest window, November 2011 to October 2026, gives 0.40 plus or minus 0.19, the highest of all 163 windows. The rate from 1970 to 2009, 0.19, is drawn as a dashed reference line. The record ends in a record El Niño, which flatters the latest windows.
Show the numbers: every 15-year window
| 15 years to | Rate | ± |
|---|---|---|
| October 2026 | 0.40 | 0.19 |
| December 2025 | 0.39 | 0.19 |
| December 2024 | 0.36 | 0.19 |
| December 2023 | 0.31 | 0.16 |
| December 2022 | 0.28 | 0.13 |
| December 2021 | 0.29 | 0.13 |
| December 2020 | 0.30 | 0.13 |
| December 2019 | 0.26 | 0.14 |
| December 2018 | 0.24 | 0.15 |
| December 2017 | 0.23 | 0.15 |
| December 2016 | 0.19 | 0.17 |
| December 2015 | 0.13 | 0.11 |
| December 2014 | 0.12 | 0.11 |
| December 2013 | 0.14 | 0.14 |
| December 2012 | 0.11 | 0.16 |
| December 2011 | 0.12 | 0.17 |
| December 2010 | 0.17 | 0.19 |
| December 2009 | 0.14 | 0.17 |
| December 2008 | 0.17 | 0.15 |
| December 2007 | 0.24 | 0.20 |
| December 2006 | 0.28 | 0.20 |
| December 2005 | 0.25 | 0.21 |
| December 2004 | 0.19 | 0.20 |
| December 2003 | 0.22 | 0.19 |
| December 2002 | 0.19 | 0.22 |
| December 2001 | 0.16 | 0.17 |
| December 2000 | 0.16 | 0.16 |
| December 1999 | 0.22 | 0.16 |
| December 1998 | 0.26 | 0.16 |
| December 1997 | 0.18 | 0.13 |
| December 1996 | 0.17 | 0.14 |
| December 1995 | 0.14 | 0.16 |
| December 1994 | 0.08 | 0.16 |
| December 1993 | 0.09 | 0.18 |
| December 1992 | 0.14 | 0.25 |
| December 1991 | 0.15 | 0.27 |
| December 1990 | 0.20 | 0.22 |
| December 1989 | 0.20 | 0.30 |
| December 1988 | 0.25 | 0.20 |
| December 1987 | 0.20 | 0.22 |
| December 1986 | 0.18 | 0.23 |
| December 1985 | 0.22 | 0.24 |
| December 1984 | 0.25 | 0.18 |
| December 1983 | 0.25 | 0.19 |
| December 1982 | 0.24 | 0.20 |
| December 1981 | 0.24 | 0.20 |
| December 1980 | 0.19 | 0.19 |
| December 1979 | 0.15 | 0.15 |
| December 1978 | 0.15 | 0.14 |
| December 1977 | 0.10 | 0.14 |
| December 1976 | 0.02 | 0.13 |
| December 1975 | 0.02 | 0.13 |
| December 1974 | 0.02 | 0.14 |
| December 1973 | 0.02 | 0.15 |
| December 1972 | −0.05 | 0.12 |
| December 1971 | −0.08 | 0.11 |
| December 1970 | −0.01 | 0.14 |
| December 1969 | 0.01 | 0.14 |
| December 1968 | −0.01 | 0.15 |
| December 1967 | −0.05 | 0.18 |
| December 1966 | −0.07 | 0.17 |
| December 1965 | −0.07 | 0.16 |
| December 1964 | 0.00 | 0.19 |
| December 1963 | 0.07 | 0.21 |
| December 1962 | 0.07 | 0.21 |
| December 1961 | 0.05 | 0.22 |
| December 1960 | 0.04 | 0.20 |
| December 1959 | 0.01 | 0.20 |
| December 1958 | −0.05 | 0.21 |
| December 1957 | −0.08 | 0.23 |
| December 1956 | −0.11 | 0.22 |
| December 1955 | −0.08 | 0.20 |
| December 1954 | −0.09 | 0.30 |
| December 1953 | −0.08 | 0.23 |
| December 1952 | −0.13 | 0.15 |
| December 1951 | −0.16 | 0.13 |
| December 1950 | −0.13 | 0.13 |
| December 1949 | −0.04 | 0.26 |
| December 1948 | 0.02 | 0.14 |
| December 1947 | 0.11 | 0.15 |
| December 1946 | 0.14 | 0.14 |
| December 1945 | 0.18 | 0.11 |
| December 1944 | 0.19 | 0.11 |
| December 1943 | 0.22 | 0.13 |
| December 1942 | 0.23 | 0.10 |
| December 1941 | 0.24 | 0.10 |
| December 1940 | 0.21 | 0.11 |
| December 1939 | 0.17 | 0.11 |
| December 1938 | 0.16 | 0.11 |
| December 1937 | 0.14 | 0.11 |
| December 1936 | 0.12 | 0.10 |
| December 1935 | 0.10 | 0.09 |
| December 1934 | 0.11 | 0.09 |
| December 1933 | 0.11 | 0.09 |
| December 1932 | 0.17 | 0.09 |
| December 1931 | 0.21 | 0.11 |
| December 1930 | 0.20 | 0.12 |
| December 1929 | 0.12 | 0.16 |
| December 1928 | 0.12 | 0.16 |
| December 1927 | 0.13 | 0.16 |
| December 1926 | 0.12 | 0.17 |
| December 1925 | 0.12 | 0.16 |
| December 1924 | 0.13 | 0.17 |
| December 1923 | 0.15 | 0.18 |
| December 1922 | 0.16 | 0.18 |
| December 1921 | 0.16 | 0.19 |
| December 1920 | 0.09 | 0.22 |
| December 1919 | 0.05 | 0.23 |
| December 1918 | 0.06 | 0.21 |
| December 1917 | 0.07 | 0.21 |
| December 1916 | 0.09 | 0.18 |
| December 1915 | 0.05 | 0.23 |
| December 1914 | −0.08 | 0.22 |
| December 1913 | −0.17 | 0.17 |
| December 1912 | −0.17 | 0.18 |
| December 1911 | −0.22 | 0.15 |
| December 1910 | −0.22 | 0.15 |
| December 1909 | −0.19 | 0.16 |
| December 1908 | −0.12 | 0.18 |
| December 1907 | −0.05 | 0.18 |
| December 1906 | 0.01 | 0.23 |
| December 1905 | −0.01 | 0.21 |
| December 1904 | 0.03 | 0.24 |
| December 1903 | 0.04 | 0.19 |
| December 1902 | 0.08 | 0.19 |
| December 1901 | 0.15 | 0.17 |
| December 1900 | 0.16 | 0.17 |
| December 1899 | 0.13 | 0.17 |
| December 1898 | 0.12 | 0.16 |
| December 1897 | 0.11 | 0.18 |
| December 1896 | 0.04 | 0.18 |
| December 1895 | −0.05 | 0.17 |
| December 1894 | −0.08 | 0.16 |
| December 1893 | −0.10 | 0.17 |
| December 1892 | −0.18 | 0.20 |
| December 1891 | −0.23 | 0.24 |
| December 1890 | −0.21 | 0.23 |
| December 1889 | −0.16 | 0.24 |
| December 1888 | −0.16 | 0.21 |
| December 1887 | −0.17 | 0.22 |
| December 1886 | −0.11 | 0.23 |
| December 1885 | −0.06 | 0.22 |
| December 1884 | −0.01 | 0.20 |
| December 1883 | 0.03 | 0.17 |
| December 1882 | 0.04 | 0.17 |
| December 1881 | 0.06 | 0.17 |
| December 1880 | 0.05 | 0.15 |
| December 1879 | 0.06 | 0.14 |
| December 1878 | 0.10 | 0.14 |
| December 1877 | 0.06 | 0.10 |
| December 1876 | 0.03 | 0.09 |
| December 1875 | 0.06 | 0.09 |
| December 1874 | 0.07 | 0.09 |
| December 1873 | 0.10 | 0.08 |
| December 1872 | 0.12 | 0.07 |
| December 1871 | 0.16 | 0.07 |
| December 1870 | 0.16 | 0.08 |
| December 1869 | 0.12 | 0.10 |
| December 1868 | 0.08 | 0.11 |
| December 1867 | 0.02 | 0.11 |
| December 1866 | −0.02 | 0.11 |
| December 1865 | −0.07 | 0.10 |
| December 1864 | −0.09 | 0.09 |
Data: Berkeley Earth (www.berkeleyearth.org (opens in a new tab)), high-resolution global temperature record (beta release), retrieved 16 November 2026. CC BY-NC 4.0 (opens in a new tab). Trend fits and chart by community3.earth. Berkeley Earth is not responsible for them.
2Published
What scientists are saying
Whether the speed-up is real, and whether it will last, is argued in the published literature. These are the main positions, including the evidence against.
Berkeley Earth, in its report on 2025
“The warming spike observed in 2023 to 2025 has been extreme and suggests an acceleration in the rate of Earth’s warming.”
“The warming spike in 2023 to 2025 suggests that the past warming rate is no longer a reliable predictor of the future …”
The sentence goes on to say that additional factors have created conditions for faster warming, “at least in the short-term.”
“However, whether that greater warming rate persists over the long-term or is only present briefly is currently hard to predict.”
These lines are from Berkeley Earth’s annual report. Its September 2026 update reports the month’s temperatures and the El Niño, and makes no statement about the warming rate.
Other published work
The human-caused rate
The annual Indicators of Global Climate Change assessment puts human-caused warming at 1.37 °C in 2025, rising at 0.27 °C per decade (range 0.2 to 0.4) over 2016 to 2025.
Significant once natural swings are removed
Foster and Rahmstorf removed the estimated effects of El Niño, volcanic eruptions and the solar cycle from five global temperature records. The speed-up then passes a 98% confidence test in all five, with rates of 0.34 to 0.42 °C per decade after a change in 2013 or 2014. In the raw data the test does not reach 95%, and the authors note that El Niño’s effect on 2023 and 2024 may not be completely removed.
An earlier 2 °C
James Hansen and colleagues argue that the climate is more sensitive to carbon dioxide than the best estimate of the Intergovernmental Panel on Climate Change (IPCC), and that 2 °C will be reached in the 2030s, not at mid-century. Their peer-reviewed paper of 2025 gave 2045. The higher sensitivity is a minority position.
Evidence against: not yet detectable
Using data to the end of 2023, Beaulieu and colleagues detected no change in the warming rate beyond the 1970s in most temperature records. They calculated that the rate would have to rise by at least 55% for a surge to be detectable. The study predates the data for 2024 to 2026.
How this compares with climate models
Warming over the last fifteen years has been faster than the central projections, and faster than nearly every model from the two previous generations. It sits in the upper part of the range of today’s models. It has not outrun the hottest of today’s models, which warm faster than the real world has.
What has risen faster than most models simulate is Earth’s energy imbalance: the gap between the energy the planet takes in from the Sun and the energy it sends back to space. It has about doubled in twenty years, and scientists do not yet agree on why.
Sources: Hausfather, “Comparing climate models with observations”, The Climate Brink, December 2025 (opens in a new tab); Foster and Rahmstorf, “Global Warming Has Accelerated Significantly”, Geophysical Research Letters, 2026 (opens in a new tab); Loeb and colleagues, Surveys in Geophysics, 2024 (opens in a new tab); Carbon Brief, “State of the climate”, 3 November 2026 (opens in a new tab).
3Ours
Our planning estimate
We plan for about 2.0 °C by around 2030 and 3.0 °C between 2045 and 2050.
That is at the pessimistic edge. It is roughly what continuing the pace of the last six years would give, and that window runs from a La Niña into a record El Niño. So most scientists expect the long-term average to rise more slowly. Published estimates put 2 °C in the 2030s (Hansen and colleagues) to the 2040s or early 2050s (central IPCC estimates), and current policies at about 2.8 °C by 2100 (UNEP). The IPCC figures are for medium to very high emissions. With deep cuts in emissions, the IPCC’s central estimate stays below 2 °C.
This estimate is ours. It is not Berkeley Earth’s, and it is not a forecast from any published source. Nobody knows what the warming rate will be over the next two decades.
We may be too pessimistic. We plan for the early dates because preparing too late costs far more than preparing too early.
Sources: Hansen and colleagues, “Another El Nino Already? What Can We Learn from It?”, 6 February 2026 (opens in a new tab); IPCC Sixth Assessment Report, Working Group I (2021), as analysed by Carbon Brief (opens in a new tab); IPCC Sixth Assessment Report, Working Group I (2021), Summary for Policymakers, Table SPM.1 (opens in a new tab); UN Environment Programme (UNEP), Emissions Gap Report 2026, as reported by Carbon Brief, 21 October 2026 (opens in a new tab).
The gap, in numbers
Measured from the long-term level in Berkeley Earth’s record, 1.42 °C in October 2026:
- needed from now to reach 2.0 °C by the middle of 2030. That is 3.9 times the most recent 15-year rate, the fastest in the record.
- About 1.6°C per decade
- needed from now to reach 3.0 °C between the start of 2045 and the start of 2050. That is 1.7 to 2.2 times the most recent 15-year rate.
- 0.68 to 0.87°C per decade
- when the long-term average would reach 2.0 °C and 3.0 °C if the most recent 15-year rate (0.40 °C per decade) simply continued. At the 1970 to 2009 rate: 2057 and 2110.
- 2041 and 2066
Arithmetic on Berkeley Earth’s data, by us. We read “around 2030” as the middle of 2030, and count the range for 3.0 °C from the start of each year. Single months and single El Niño years run ahead of the long-term average. How our dates compare with published rates: Methods and sources.
Figure 3
Where the pace leads
Today’s long-term level, extended to 2060, in °C above the 1850–1900 average
Measured
- 12-month average
- Trend line through the last 30 years
- Long-term level today: 1.42 °C
Simple extrapolations, not forecasts
- 1970–2009 rate, continued: 0.19 °C per decade
- Most recent 15-year rate, continued: 0.40 °C per decade
- 95% ranges
Published estimates of when 2 °C is reached
- Hansen and colleagues: 2030s
- IPCC central estimates: 2040s to early 2050s
Ours
- Our planning estimate (pessimistic edge): 2.0 °C by about 2030, 3.0 °C between 2045 and 2050
The long-term level today, 1.42 °C, is where a straight line through the last 30 years of data stands in October 2026. Single years sit above or below it: the latest 12 months average 1.53 °C.
The grey and violet lines extend that level at the 1970 to 2009 rate and at the most recent 15-year rate. They are simple extrapolations, not forecasts: a fitted line knows nothing about emissions, aerosols, clouds or the oceans. Shading shows their 95% ranges, from the uncertainty in today’s level and in each rate.
The dashed band is our planning estimate (pessimistic edge). It reaches 2.0 °C about 6 years before the early edge of the faster line’s range (2036), and 3.0 °C 4 to 9 years before it (2054).
The outlined bars on the 2.0 °C line are published estimates of when 2 °C is reached. The IPCC bar spans its central estimates for the three higher-emission scenarios only. Their very likely ranges run from 2032 to 2084. In the IPCC’s two lowest-emission scenarios the central estimate stays below 2 °C all century, so they have no place on the bar.
Line chart from 2010 to 2060. Measured: the 12-month average temperature, and the long-term level today of 1.42 degrees Celsius above the 1850 to 1900 average. From that level two straight lines are drawn. At the 1970 to 2009 rate of 0.19 degrees per decade the line reaches 2.0 degrees around 2057 and 3.0 degrees around 2110. At the most recent 15-year rate of 0.40 degrees per decade it reaches 2.0 degrees around 2041 and 3.0 degrees around 2066. Both are simple extrapolations, not forecasts. Published estimates of when 2 degrees is reached are marked on the 2.0 degree line: Hansen and colleagues, 2030s; IPCC central estimates, 2040s to early 2050s. Our planning estimate, at the pessimistic edge, is drawn as a dashed band: 2.0 degrees by about 2030 and 3.0 degrees between 2045 and 2050. It rises far more steeply than either line.
The same comparison as a table
| Case | Reaches 2.0 °C | Reaches 3.0 °C |
|---|---|---|
| Simple extrapolations, not forecasts | ||
| 1970–2009 rate, continued 0.19 °C per decade | 2057 (2049 to 2069) | 2110 (2094 to 2134) |
| Most recent 15-year rate, continued 0.40 °C per decade | 2041 (2036 to 2055) | 2066 (2054 to 2101) |
| Published estimates | ||
| IPCC central estimates the three higher-emission scenarios; in the two lowest the central estimate stays below 2 °C | 2040s to early 2050s | Not by mid-century: 2.0 to 2.4 °C for 2041–2060 in these three scenarios |
| Hansen and colleagues a minority position | 2030s | no date given |
| Ours | ||
| Our planning estimate pessimistic edge; needs 0.68 to 0.87 °C per decade | about 2030 | 2045 to 2050 |
Data: Berkeley Earth (www.berkeleyearth.org (opens in a new tab)), high-resolution global temperature record (beta release), retrieved 16 November 2026. CC BY-NC 4.0 (opens in a new tab). Trend fits, extrapolations and chart by community3.earth. Berkeley Earth is not responsible for them. Published estimates: Hansen and colleagues, 6 February 2026 (opens in a new tab), who measure from 1880 to 1920; IPCC Sixth Assessment Report (2021) central estimates for the three higher-emission scenarios, about 2041, 2046 and 2052, as analysed by Carbon Brief (opens in a new tab); mid-century best estimates from Table SPM.1 of the same report (opens in a new tab), as reproduced by Carbon Brief.
Mission
Towns that can keep their people alive
community3.earth is being set up to help existing towns and small communities get ready, before the world is 3 °C warmer than before industry, to keep their people alive and their society working.
The 3 in our name is that level of warming, and it is a deadline, not a destination. In our view a town that starts preparing at 3 °C has started too late: the work takes many years (our own plan, further down this page, runs for twenty), and it needs supply chains, credit and spare labour that extreme heat, grid failure, drought, fire and flood will already be wearing down. So the aim is to be ready before 3 °C, not to begin at it.
We plan to make standard modules for power, cooling, water, food storage, a clinic and communications, and install them in buildings a town already has. Then we extend outward to the whole town. Later we build new districts beside the towns best placed to receive people.
Status, November 2026 We are at the start. No town has signed and nothing has been built yet.
What we are building
- First
The Survival Core
Existing schools, halls and clinics converted into cooled refuges with their own power, water, food reserve, clinic and communications. Designed to run for two weeks with no grid.
- Then
Town systems
Once a town has a core, it extends outward in stages: a second water source, local power that runs without the grid, one cool room in every home, fire breaks and protected food growing.
- Later
Receiving districts
New factory-built neighbourhoods beside towns that can hold and have agreed to grow. The land sits in a community land trust so homes stay affordable. We build one only after delivering a core there.
- Throughout
Open designs
Everything a community needs to build and repair the system will be published under an open licence. Anyone may use it, with us or without us.
- Throughout
Local ownership
Whatever we install ends up owned by a body in the town: the municipality, a utility, a school district, a tribal government or a co-operative. The town can run it without us.
- Throughout
The training corps
A paid programme of 12 to 24 months that trains local residents and career-changers as technicians while they build in their own region. We expect skilled people, not money, to be the limit.
The figures in these cards are design targets from our plan. They are not results: nothing has been built or tested yet.
What we will not do
We will not build private bunkers.
No single-household or closed-membership survival facilities. Wealth pays for more places for more people, not a better place for one.
We tell towns the truth, including when the answer is to leave.
If our assessment is that a town cannot be held, we say so and we do not sell it permanent infrastructure. We publish the method so it can be challenged.
The town owns what is built.
Everything ends up owned by a body in the town. We will not hold a town’s water or power hostage to a subscription.
Adaptation has limits. None of this replaces cutting emissions.
The plan
The plan, stage by stage
Six stages. Choose a stage to watch one town change.
These are targets we plan against. They are not forecasts or promises. We are at the start. No town has signed and nothing has been built yet.
- Blue glow cooled and safe
- Yellow line local power
- Grey line the wider grid
- Mint built under this plan
- Dashed planned, not built
November 2026 to February 2027
Find out if it is real
In the scene
We build nothing in the first 90 days. We test three things cheaply: whether a real town will sign up for an assessment, whether someone with means will back a named town, and whether experienced people will join.
Targets for day 90
- 40conversations: ten each with towns, people with means, builders, and experts and funders.
- Version 0.1of the survival standard, four pages long, published with the town assessment method.
- 15towns shortlisted, and every one visited.
- 3town assessments signed, each paid for by the town or by a sponsor.
- 1pilot building named, with its owner’s written agreement, and two or three founding hires or committed part-timers.
November 2026 to the end of 2028
Prove it in one town, then five
In the scene
One question matters here. Can we deliver a working core to a real town, at a cost the town can pay, that the town then runs on its own? We start with one building and publish the results, failures included.
Decision point. We go on to the next stage only if three things are true. A town has run a drill without us. A government body has paid for a core. The cost of each refuge place is heading below half the pilot’s.
Targets for the end of year 2
- 200 to 500places in a pilot refuge: one building, commissioned by the end of 2027.
- Summer 2028The pilot runs a full drill with no grid. We publish the results, including what failed.
- 30assessments delivered, and 5 towns with a core delivered or under construction.
- 8,000 to 10,000people with a refuge place, in 5 to 8 buildings.
- 20 and 50The first 20 trainees in the Corps and the first 50 volunteers in the Reserve.
2029 to 2031
Regional scale before the surge
In the scene
Our planning estimate puts 2.0 °C at about 2030. We expect the first event that changes public opinion in our regions during these years. The aim is to have a factory, trained people and town financing ready before it.
Targets for the end of year 5
- 75 to 100towns with a core, and 150,000 to 250,000 people with a refuge place.
- 50 a yearOne module factory able to ship 50 cores’ worth of modules a year, with work under way in three regions.
- 10towns, at least, with whole-town systems under way.
- 200 to 500factory-built homes as the first receiving district breaks ground, beside a town that can hold and has agreed to grow.
- 25towns, at least, sign a mutual-aid compact. 2 independent builders are licensed, and the Corps trains 500 to 1,000 people a year.
2032 to 2036
Federation
In the scene
We assume demand will by now exceed what any single organisation can meet. Success means the work no longer depends on us. Regional companies do the building. The centre keeps three jobs: the standard, the bank and the school.
Targets for the end of year 10
- 500 to 1,000towns certified to core level, covering three to five million people.
- 50 or moretowns with full town systems.
- 10 to 20receiving districts, giving 50,000 to 100,000 places.
- 6 to 10regions with a module plant, owned by regional companies or co-operatives.
- 1 yearof staple food in reserve for the whole network, alongside a polytechnic, small-batch production of essential medicines, a bond bank and a mutual insurer.
2037 to 2046
Continuity
In the scene
Our planning estimate has the world approaching 3.0 °C. The aim shifts from hardening single towns to keeping a working society across a network of them: food, power, water, cooling, medicine, technology and the ability to invent.
What success looks like in 2046
- 1 to 2 millionplaces to receive people, across receiving districts and the towns that can hold.
- 2 weeksEvery dwelling in a network town stays survivable through a two-week outage in peak heat.
- Every regionmakes essential medicines, and the network can rebuild or make its own inverters, battery packs, compressors, pumps and membranes.
- Thousandsof communities beyond us use the standard, including people who simply downloaded the designs.
- A small centreThe centre is a small foundation. If it vanished, the network would carry on.
No fixed date. Prepared from the first year.
When it gets much worse
In the scene
We expect demand to come in jumps, after disasters. A deadly heat wave with grid failure, or failed harvests, could multiply the need overnight. The response is decided in advance. Every stage must leave us able to triple output within a year.
Triggers we plan for. A mass-casualty heat event with grid failure. Harvest failure in several regions at once. Insurers or lenders leaving a region. A government asking for hundreds of towns in two years. Large-scale migration into regions that can hold. A break in the supply of key components.
What must be ready beforehand
- 10mobile cores held in reserve for emergency use.
- 50, 2,000, 20,000Reserve volunteers by year 2, year 5 and year 10, each trained to deploy for two to six weeks.
- Open designscomplete enough to build from without us, and licensed builders who have each already built at least one core.
- WeeksTraining that turns a tradesperson into a technician for this work in weeks.
- DaysStandby funds that can be drawn in days, and a stock of parts that take long to order: transformers, switchgear, chillers and membranes.
Get involved
Help us get this right
We need people who know more than we do. Tell us who you are and how you would like to help. The form asks a few extra questions that depend on your answer, so your message reaches the right person. We read every message.
We are at the start. No town has signed and nothing has been built yet.
Most wanted
Climate and earth-system scientists
We need better timelines than our own, and advice on which infrastructure to build first. If we are wrong, we want to hear it from you.
The questions are optional. We need only your name, an email address and your agreement.
- Timelines.Tell us what you consider plausible, and where our planning estimate is off.
- The map.Check the method and the thresholds behind it.
- What to build first.Say what a town should do before anything else.