2025 Global Mean Surface Temperature

Most of the annual GMST temperature data is in for 2025, and as expected, it was essentially a statistical tie with 2023 for the 2nd and 3rd the warmest year on record. La Niña conditions developed in 2025 which prevented it from competing with 2024. 

Above, here's how 2025 looks in several major datasets, and I added the recently published DCENT-I surface temperature dataset. I put the 2-sigma uncertainties for HadCRUT5 in dotted lines. This graph is set to the 1951-1980 mean (left scale), and this allows you to see that most of the variability between the datasets occurs in the late 19th century. The right scale is offset to show the warming above the 1850-1900 baseline for HadCRUT5 with the 1.5°C target as a dotted horizontal line. Below I set the same data to the 1850-1900 baseline and plotted centered 30-year means from HadCRUT5; this has the effect of better showing warming for each dataset above this baseline (though variability between the datasets artificially shows up in recent decades). I then used the most recent 30-year trend to estimate the current anomaly for the 30-year mean from HadCRUT5. NASA, JRA-3Q and ERA5 don't go back to 1850, but I found the proper baseline adjustment to align them to the other three.

Dataset	1996-2025 Warming Rates		1970-2025 Warming Rates		2023 - 2025 Mean
BEST	0.24 ± 0.04°C/decade (2σ)		0.21 ± 0.02°C/decade (2σ)		1.53°C
NOAA	0.25 ± 0.04°C/decade (2σ)		0.20 ± 0.02°C/decade (2σ)		1.38°C
NASA	0.25 ± 0.04°C/decade (2σ)		0.20 ± 0.02°C/decade (2σ)		1.40°C
HadCRUT5	0.24 ± 0.04°C/decade (2σ)		0.20 ± 0.02°C/decade (2σ)		1.47°C
DCENT-I	0.24 ± 0.04°C/decade (2σ)		0.20 ± 0.02°C/decade (2σ)		1.56°C
JRA-3Q	0.25 ± 0.05°C/decade (2σ)		0.21 ± 0.02°C/decade (2σ)		1.50°C
ERA5	0.27 ± 0.05°C/decade (2σ)		0.21 ± 0.02°C/decade (2σ)		1.52°C
Average	0.25 ± 0.05°C/decade (2σ)		0.20 ± 0.02°C/decade (2σ)		1.48°C

Above I show the trends for the last 30 years as well as the trends since 1970. Set to a 1850-1900 baseline, most datasets came in under 1.5°C for 2025, but the target is for a long-term, which I take to mean at least a decade. The 2023-2025 mean averaged just slightly under 1.5°C. Extending the most recent 30-year trend to 2025 puts the current long-term anomaly at 1.38°C, and this suggests that we're on track to cross the 1.5°C target in the early 2030s and the 2.0°C target in the 2050s.

Leon Simmons on X prepared the gaussian distribution of daily GMST anomalies for every day since 1940. For 2025, the GMST for every day was grater than the 1991-2020 mean.  

The new GloSAT dataset goes back to 1785, so I set this to a 1790-1830 baseline and aligned it with HadCRUT5. According to this dataset, we hit the 1.5°C target this year, though I think it's better to continue to use a 1850-1900 baseline, since so many assessments of the consequences of warming use this baseline for estimates.  

There appears to be acceleration in 30-year trends as well. The trends I calculate are going to be affected by ENSO, which explains a good deal of the waviness in the overall upward trend in GMST warming trends. This shows that while many claim that there are "pauses" in global warming, that perception is really just an artifact of ENSO. Not only is global warming continuing, it's acceleration. That said, the acceleration may be due the fact that we're cleaning up aerosol pollution, de-masking the accumulated effects of GHGs.

This can also be seen by looking at the trends of El Niño years, La Niña years, and neutral years. All appear to be warming at about the same rate. I have not calculated the uncertainties for these trends, but I suspect they overlap, even though visually neutral years appear to be warming less rapidly than El Niño years, La Niña years. Since La Niña conditions developed in 2025, contrarians predictably pivot again back to the same kinds of fake arguments they used after 1998 and 2016. They started counting the months for which we've seen "no warming" while ignoring the fact that we should expect La Niña years to fall below the overall trendline and El Niño to land above it. I poke a little fun at this here.

Despite loud objections to the contrary, observations are well within the expectations of CMIP6 model projections, especially when filtered to show results from models calculating ECS to be near 3°C. We've had about 1.2°C warming in this model-observation comparison.

To illustrate the long-term effect of carbon dioxide on global mean surface temperature, I also updated my plot of the correlation between GMST and CO2, with the usual caveats that need to be acknowledged. Correlation isn't causation (evidence for causation is here), and there are other forcings at play here, particularly the forcings of other GHGs and aerosol pollution. These forcings historically have roughly canceled each other out, but if we continue to clean up air pollution while emitting methane and other GHGs, this will be less the case going forward. 

Above I show a similar plot of GMST anomalies with CO2 forcings, and here you can see that the residuals have a slope of ~0 C/W/m^2. All in all, 2023-2025 has been exceptional, and it will be interesting to see if the apparent step up in global warming continues. But clearly the CO2 continues to be a very good predictor of GMST anomalies, and there's no reason to think this will change moving forward.