Where the numbers are heading

Solid: measured. Dashed: forecast. Darker band: 80% likely range. Lighter band: 95%.

Expected Sep 2027

428.8 ppm (95%: 427.6–430.5)

Expected Sep 2028

431.2 ppm (95%: 430.0–433.3)

Tested: we hid the last 24 months (Oct 2024–Sep 2026) and forecast them from the data before. The actual value was inside the 95% range 23 of 24 times and inside the 80% range 16 of 24. Average miss: 0.8 ppm, against 4.0 ppm for “same as the same month last year”. It beat the simple guess and the ranges held about as often as they should.

Monthly average CO₂ measured at Mauna Loa, Hawaiʻi, since 1958. The yearly wiggle is plants in the northern hemisphere breathing in and out.

Extends the recent rise and seasonal pattern; it cannot know about future emissions changes or volcanic eruptions.

Source: NOAA Global Monitoring Laboratory, Mauna Loa

Solid: measured. Dashed: forecast. Darker band: 80% likely range. Lighter band: 95%.

Expected Jun 2027

1,941 ppb (95%: 1,934–1,949)

Expected Jun 2028

1,948 ppb (95%: 1,934–1,963)

Tested: we hid the last 24 months (Jul 2024–Jun 2026) and forecast them from the data before. The actual value was inside the 95% range 24 of 24 times and inside the 80% range 19 of 24. Average miss: 4 ppb, against 11 ppb for “same as the same month last year”. It beat the simple guess and the ranges held about as often as they should.

Monthly global average methane from NOAA’s marine surface sites. NOAA publishes these with a few months’ delay.

Methane growth has sped up and slowed down before for reasons scientists still debate; a trend model will miss those turns.

Source: NOAA Global Monitoring Laboratory, global average

Solid: measured. Dashed: forecast. Darker band: 80% likely range. Lighter band: 95%.

Expected 2030

1.20 °C vs 1951–1980 (95%: 1.07–1.93)

Expected 2035

1.31 °C vs 1951–1980 (95%: 1.17–2.09)

Tested: we hid the last 10 years (2016–2025) and forecast them from the data before. The actual value was inside the 95% range 6 of 10 times and inside the 80% range 4 of 10. Average miss: 0.15 °C vs 1951–1980, against 0.14 °C vs 1951–1980 for “no change from the last year”. Here the method did no better than the simple guess and its ranges were too narrow: read this forecast as a rough sketch.

Yearly global average surface temperature, land and ocean, compared with the 1951–1980 average. The current year is left out until it is complete.

A statistical extension of the warming since 1970, not a climate model: El Niño years and volcanoes push single years well above or below it.

Source: NASA GISS Surface Temperature Analysis (GISTEMP v4)

Solid: measured. Dashed: forecast. Darker band: 80% likely range. Lighter band: 95%.

Expected 2031

4.42 million km² (95%: 2.78–5.13)

Expected 2036

4.43 million km² (95%: 2.25–5.39)

Tested: we hid the last 10 years (2017–2026) and forecast them from the data before. The actual value was inside the 95% range 10 of 10 times and inside the 80% range 10 of 10. Average miss: 0.62 million km², against 0.29 million km² for “no change from the last year”. Here the method did no better than the simple guess, so read the forecast as a rough guide only.

Average Arctic sea-ice extent each September, when the ice is at its yearly low. Satellite records start in 1979.

Single summers swing a lot with weather, so the range is wide; the ice could also decline faster or slower than the past trend.

Source: NSIDC Sea Ice Index (G02135 v4)

Solid: measured. Dashed: forecast. Darker band: 80% likely range. Lighter band: 95%.

Expected 2030

8.63 billion (95%: 8.56 billion–8.72 billion)

Expected 2035

9.03 billion (95%: 8.96 billion–9.21 billion)

Tested: we hid the last 10 years (2016–2025) and forecast them from the data before. The actual value was inside the 95% range 3 of 10 times and inside the 80% range 3 of 10. Average miss: 33.7 million, against 438.6 million for “no change from the last year”. The point forecast beat the simple guess, but the ranges were too narrow: the real uncertainty is larger than the bands show.

Total world population each year. The UN’s own projections use births, deaths and migration by age; this only extends the recent trend.

Population growth is slowing; a trend model notices that only as it shows up in the data.

Source: World Bank (UN Population Division estimates)