Climate models
"The models are always wrong" is one of the most common claims about climate change. It can be checked: scientists have published warming projections since 1970, and the years they were predicting have now happened. The first half of this page compares those predictions, and today's climate models, with what was measured. The second half shows what the models say comes next. That part cannot be checked yet, so it is marked as projection throughout.
How to read the charts: solid lines and bars are measurements; dashed lines and hatched bars come from models, projections or estimates.
Part 1. How well did past predictions do?
Predictions made decades ago, and the models used today, set against the temperatures that were actually measured.
17 projections from 1970 to 2007
In 2020, Zeke Hausfather, Henri Drake, Tristan Abbott and Gavin Schmidt collected every projection of global warming published between 1970 and 2007 that gave a time series of future temperatures, and compared each one with the warming over the years it was predicting, up to 2017 at the latest (Geophysical Research Letters 47, e2019GL085378; data on GitHub). The projected rates and the verdicts ("matched" means the difference was within the uncertainty of the two trends) are the study's. The measured rates here are calculated by this page from ERA5 air temperature over the same years; the study used several surface temperature records, so its measured rates differ slightly.
Hansen's 1988 scenarios
Scenarios from Hansen et al. (1988), Global climate changes as forecast by Goddard Institute for Space Studies three-dimensional model, Journal of Geophysical Research 93, 9341: temperatures 1958–2019, where the model runs end, as published by RealClimate; forcing 1958–2017 as digitized by Hausfather et al. (2020). The model runs start in 1958 with the measured rise in greenhouse gases; the scenarios differ from 1988 on. To compare like with like, the measured ERA5 temperatures are shifted so that their 1958–1987 average matches scenario B's. Forcing is the extra energy that the added greenhouse gases trap, in watts per square meter, counted from 1958. Hansen's scenario forcings cover greenhouse gases only, so the measured line does too: well-mixed greenhouse gases from the Indicators of Global Climate Change 2025 (Forster et al. 2026, data), which use newer estimates of each gas's effect than were available in 1988.
Hansen et al. 2025: a test running now
Hansen's recent predictions, scored against NASA's measurements
Hansen, Kharecha, Sato and 15 co-authors, many of them at NASA's Goddard Institute for Space Studies (GISS), 2025: Global warming has accelerated: are the United Nations and the public well-informed? Environment: Science and Policy for Sustainable Development 67(1), 6–44 (doi:10.1080/00139157.2025.2434494), published February 2025. Measured in the paper's own terms: NASA GISTEMP v4 (data.giss.nasa.gov/gistemp, built from NOAA's GHCN station and ERSST sea surface data), each month taken relative to that month's 1880–1920 average, as in the table Hansen's group keeps at columbia.edu/~jeh1 (this site's version matches it to within 0.004 °C), then averaged over 12 months. The grey line is ERA5 above 1850–1900, as used elsewhere on this site, for comparison. The hatched box is the projection in Hansen's monthly chart Global Surface Temperature (Base Period 1880–1920), version with data through September 2026, read off the chart image with its axes calibrated on the gridlines (to within 0.01 °C).
Today's climate models vs measurements
CMIP6 is the set of climate model runs behind the IPCC's 2021 report. Shown here: one run (r1i1p1f1) from each of the 16 models with complete global air temperature in both the historical experiment (to 2014, driven by measured greenhouse gases, air pollution, volcanoes and the sun) and its continuation with the middle scenario SSP2-4.5 (from 2015), as averaged by Neil Swart, CCCma (GitHub). Both models and measurements are global near-surface air temperature, relative to 1850–1900. Measured: ERA5, converted to 1850–1900 with Copernicus's offset of 0.88 °C for 1991–2020 (as on the air temperature page); complete years only.
Part 2. What the models say comes next
None of this has happened yet. These are projections and estimates from the main international assessments, shown as published, each with its range. What actually happens depends above all on how much more CO₂ is emitted.
Warming by scenario (IPCC)
Projection
IPCC Sixth Assessment Report, Working Group I, Summary for Policymakers, Table SPM.1 (ipcc.ch): best estimate and very likely range (90% chance the outcome falls inside) of the 20-year average global surface temperature, relative to 1850–1900. The lines to 2100 are the same assessment year by year, from IPCC AR6 Chapter 4 (Lee et al. 2021; files in Chris Smith's AR6 repository), each point a 20-year average plotted at the middle of its 20 years, shifted from 1995–2014 to 1850–1900 with the IPCC's 0.85 °C; the shaded ranges (shown for the low and high scenarios, as in the IPCC's own figure) can differ from the table by up to 0.1 °C. The measured line is the 20-year average of ERA5, plotted the same way. The IPCC combined the CMIP6 models with observed warming and other evidence, so these ranges are narrower and less hot than the raw model spread. Current-policies and pledges estimates: UNEP Emissions Gap Report 2025 (unep.org). "Long-term level now": the straight-line trend through the last 30 complete years of ERA5, read off at its last year (the same method as the air temperature page).
Emissions vs the path to the targets
Measured Required path
Measured: total greenhouse gas emissions by gas, in billions of metric tons of CO₂-equivalent (each gas weighted by how much it warms over 100 years), from the Indicators of Global Climate Change 2025 (Lamb et al., in Forster et al. 2026, data). Fluorinated gases are in the data from 1990 (about 1% of the total then). Required paths: IPCC AR6 Synthesis Report, Table SPM.1 (ipcc.ch): the median cut from 2019 levels in the scenarios that limit warming to 1.5 °C and to 2 °C. "At least an even chance" (1.5 °C) and "at least two in three" (2 °C) refer to how strongly the climate reacts, which is not known exactly: for each scenario, at least half (or two-thirds) of the plausible climate responses stay below the limit. "No or limited overshoot" means passing 1.5 °C by at most about 0.1 °C, for up to several decades, before coming back below it by 2100, for 2030, 2035, 2040 and 2050. The straight lines between those years are a simplification; the IPCC gives only those years, each with a range (in the tooltips).
What it would take
The cut needed for 2 °C, year by year
Year by year: the cut from the latest measured year to the IPCC's 2030 level for 2 °C (see the chart above), spread evenly over the years to 2030, starting from EDGAR's first estimate for 2025. Comparisons: greenhouse gas emissions of countries in 2025 from EDGAR (EU Joint Research Centre, 2026 report, excluding land use); CO₂ from burning coal, oil and natural gas in 2024 from the Global Carbon Project (as on the slow signals page); CO₂ from deforestation and other land use from the Indicators of Global Climate Change 2025; methane from oil, gas and coal: at least 120 million tonnes a year (IEA Global Methane Tracker 2025), counted as CO₂-equivalent with the IPCC's factor of 29.8. Options: IPCC AR6 Working Group III, Summary for Policymakers, Figure SPM.7 (ipcc.ch), the estimated potential of each option in 2030 and its net lifetime cost per tonne of CO₂-equivalent avoided. The IPCC publishes these only as a chart, so the values were read from the figure in its PDF (checked against the report's own totals; see the methods page).
The remaining carbon budget
Estimate
Remaining carbon budgets from 1 January 2026: Indicators of Global Climate Change 2025, Forster et al. 2026, Earth System Science Data 18, 3889 (doi:10.5194/essd-18-3889-2026), calculated by Robin Lamboll with the paper's central setup (data). The paper rounds the 1.5 °C budget to 130 billion metric tons. The years left use the same paper's emissions for 2024: 38.6 billion metric tons of CO₂ from fossil fuels and industry plus 4.6 from land-use change. Today's figure assumes emissions continue at that rate, which is a simplification.
The next five years (WMO)
Forecast: chances, not certainties
WMO Global Annual to Decadal Climate Update 2026–2035, published 28 May 2026 (wmo.int), prepared by the UK Met Office from forecasts by climate centers around the world. Temperatures relative to 1850–1900.