Essay by Eric Worrall
When the observations disagree with the model, claim the model is right anyway?
World may warm by 25 per cent more than current projections suggest
Official projections for how much the world will warm are already catastrophic, and now it appears they significantly underestimate future warming
By Michael Le Page
19 August 2026Climate models projecting higher warming for a given level of carbon dioxide have been discounted as unrealistic, but a new way of assessing models suggests they are accurate.
…
How much the world will warm depends on how much more CO2 we put into the atmosphere and how the planet responds to that CO2. Different climate models vary in how much warming they project for a given CO2 level. For instance, some models suggest that a doubling of CO2 will result in the average global surface temperature increasing by around 1.6°C above pre-industrial levels in the short term. Others suggest it could be as high as 3°C.
…
Their results suggest that natural variability limited temperature rises between 1981 and 2014 – a period that includes the so-called global warming hiatus in the first decade of the 21st century. “It’s really exactly this period 1981 to 2014, which was used in the previous papers, where you see the strongest bias down,” she says. “It was really bad luck.”
…
“In my opinion, the filtering out of natural variability is a valuable step forward,” says Drew Shindell at Duke University in North Carolina, although he isn’t yet convinced that the short-wave and long-wave trends are a good way to assess models.
“I think it’s a plausible conclusion and a reasonable approach,” says Steven Sherwood at the University of New South Wales in Sydney.
…
Read more: https://www.newscientist.com/article/2585248-world-may-warm-by-25-per-cent-more-than-current-projections-suggest/
The abstract of the study;
Recent Temperature and Energy Imbalance Trends Point to Higher Estimates of Future Warming
G. Gyuleva, E. Fischer, R. Knutti, S. Sippel
First published: 11 August 2026
https://doi.org/10.1029/2026EF008356Digital Object Identifier (DOI) VIEW METRICS
Abstract
Climate models simulate a wide range of 21st century warming for a given forcing scenario. Constraining this uncertainty is a central challenge in climate science because of its implications for climate policy and adaptation. The transient climate response (TCR) is a key idealized metric used to quantify future warming in response to an exponentially increasing CO2 concentration. Climate models span a range of 1.3–3 K for TCR. In attempts to constrain this range, emergent constraints on TCR based on historical temperature trends consistently pointed toward TCR values at the lower end of the range of models. However, recent evidence from trends in the short-wave and long-wave components of Earth’s energy imbalance (EEI) at the top-of-atmosphere suggests that models with higher TCR lie closer to the observed EEI trends. Here, we reconcile this discrepancy and provide a revised range for TCR of 1.9–2.6 K. Using a statistical variability-filtering approach, we show that previous temperature-based constraints were biased low due to internal variability, according to our method. We then provide an EEI-based constraint and demonstrate that short- and long-wave EEI trends have a strong potential to constrain future warming, due to the much larger inter-model spread in EEI compared to surface temperature. When considering the recent 2001–2025 period, our results show that both surface temperature and EEI trends support higher TCR values than previously estimated. This result implies that it is increasingly difficult to exclude high climate sensitivity models from the plausible range of future warming.
Read more: https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2026EF008356
Essentially what they appear to be trying to do is work out what warming should have been using measurements of global energy imbalance, apply known mostly ocean cycle forcings such as El Nino and La Nina, then tag whatever discrepancy is left as random “natural variability” which can be discarded for the purpose or predicting future global temperature. Unsurprisingly their conclusion is natural variability has suppressed global warming, and we can expect more warming in the future.
The scientists admit that their approach assumes they have identified all significant forcings, and their results would be wrong if they accidentally discard a true climate forcing by assuming it is random natural variability.
… We acknowledge that if a forced mechanism not represented by models is responsible for parts of the observed SST pattern, our statistical variability-filtering approach might alias it into variability, rather than a forced change. We therefore interpret our variability-filtering approach as an estimate of the variability contribution of recent warming patterns, conditional on forced pattern effects not aliasing into our model. As discussed in Section 1, literature remains inconclusive about whether the effect is caused by internal variability, a forced response not simulated by models, or even an artifact of observational products. …
The scientists admit the massive discrepancy between surface temperature based models and global energy imbalance based models, and that this might be due to observational error:
Emergent constraints based on raw historical GSAT trends for 1981–2025, and raw historical EEIPC1 trends for 2001–2025, produce very different TCR ranges (Figures 3a and 3b): 1.2–2.2 K for the GSAT-based constraint and 1.9–3.1 K for the EEI-based constraint. … This mismatch could have several reasons: (a) observational error, (b) internal variability …
They admit variability probably isn’t the explanation for this gigantic discrepancy between EEI (Earth’s Energy Imbalance) and surface temperature models;
Our results provide some evidence that variability is most likely not the main cause of the discrepancy between GSAT- and EEI-based constraints. If it is owed to a systematic bias in the GSAT-to-EEI coupling, then we should expect this discrepancy to persist regardless of the period in consideration. The fact that the discrepancy partly becomes resolved by aligning the two periods suggests that processes in the real world might have changed between 1981–2025 and 2001–2025 in such a way as to make higher future warming more likely. Even so, aligning the periods of the constraint to 2001–2025 does not fully resolve the discrepancy, so a part of it may still be owed to model biases in the coupling between surface temperature and energy imbalance, such as identified by Chen et al. (2026) and Olonscheck and Rugenstein (2024).
Let’s just say in my opinion this approach is unconvincing. Plugging in known forcings and being left with a huge unbridgeable gap between energy imbalance calculations and temperature models to me screams missing variable or bad observations. Going on to suggest the energy imbalance models are more reliable predictors of future temperature, despite them doing a poor job of explaining observations, in my opinion is just speculation.
The authors of the paper above have no idea what is happening to the alleged missing heat, any more than Trenberth did when he described the missing heat as a “travesty” in Climategate.
Even if the alleged missing heat is real and not just a long standing measurement error, there is no reason to assume we’ll ever see that heat again. A few warm years is not robust evidence global warming has suddenly reverted to some alarmist trend. If that alleged missing heat is being swallowed by the ocean depths, on any meaningful human timescale that heat is gone. The ocean depths have barely warmed since the last ice age, they have the thermal capacity to absorb thousands of years of excess surface heat with no noticeable impact on the ocean surface or near surface regions humans care about.
This is a whopping load of climate alarmist gibberish.
“How much the world will warm depends on how much more CO2 we put into the atmosphere and how the planet responds to that CO2.”
The statement above says it all, in their view CO2 is the control knob for our climate, no need to check out anything else.
More proof of expensive educations being wasted on lemmings.
>> but a new way of assessing models suggests they are accurate.
Uhm.. No, Mr Le Page, this does not mean what you think it means. .
I guess you have the right to come up with a new way to look at data and models, but unless you can show how the other groups are wrong in their analysis, any valid additional analysis providing a new result increases the uncertainty for all results!
This is similar for the different “temperature products” using the same satellite data, but since R. Spencer makes some very valid points about the other products ignoring trouble with some MSU-channels (https://www.drroyspencer.com/2019/04/uah-rss-noaa-uw-which-satellite-dataset-should-we-believe/), which allows him to ignore the other temperature products and incorrect.
Actually, there isn’t a lot of difference between UAH and NOAA Star. (only have data to 2021)
They both show obvious near zero trends from 1980-1997, and from 2001-2015, with warming spike+step changes at the 1998 and 2016 El Ninos.
Trend-wise, over the whole record NOAA Star has a slightly more positive trend, but since 2000, UAH has a slightly more positive trend….
There are slight difference, but the story they tell is still the same..
…. Warming at El Nino events only.
To claim “real-world proof” requires a level of empirical isolation, historical precedence, and direct cause-and-effect alignment that real-world climate data does not offer. The paleoclimate record shows temperature driving CO2, not vice versa. The geological record shows high-CO2 ice ages and that the current ice age was caused by tectonic activity ~34 million years ago. The modern record shows warming trends that began before major emissions. Cherry picking periods of correlation is not proof. The open planetary system makes isolating CO2 from water vapor, clouds, and solar cycles empirically impossible outside of theoretical frameworks or climate models.
NONE of the models take in account the variable sun energy output in either the short or long terms. None of the models have been verified on past warm and cold periods since the end of the last ice age. Other studies do show the sun output started upward in the last half of the 1800’s, hence the world warmed and is warming. Current levels are less than the 1000 AD, 600 AD and the entire Holocene warm periods, periods without high CO2 levels which leaves either the sun warming the world or somebody with a very large electric heater hooked up to a magic power source.
In my opinion, it’s not just speculation, it’s poor science.
When a model doesn’t fit, ascribe it to random outliers which don’t require explanation because they’re “random.” That’s not objective science, that’s nonsense. Go back to the pseudo-humanities where you belong. (I’m a longtime member of AGU. The linked article is an embarrassment.)
In my opinion, it’s not just poor science, it is propaganda dressed up to appear as science. It is Trans-Reality Alarmism with a very thin science look-alike veneer.
Isn’t that what fitting a trend line does? If a statistically significant trend can’t be demonstrated for a time series that is longer than the definition of the phenomenon then that strongly suggests that the phenomenon does not have a trend. It sounds like the whiners are unhappy that reality doesn’t support their wishful thinking.
As I said when the retired RCP8.5 and derivatives…they will just recalibrate the models. Problem solvered.
Yes, that’s what they are trying to do here.
In CLINTEL’s book Frozen Views of the IPCC, which I read with great interest in its French translation, there was a section devoted to a so-called “pattern effect,” mentioned in the AR6. The idea was indeed that natural variability had “masked” the “true warming” and that, to put it briefly, “we haven’t yet seen anything of what is going to hit us.” This poses a major epistemological problem, because it makes the theory absolutely unfalsifiable. And this epistemological problem becomes downright comical when one remembers that the IPCC and its affiliates, who have been “right from the start,” as we are constantly being told, have thoroughly berated anyone who emphasized natural variability as a factor at least as important as anthropogenic factors in climate change. Whatever one may think about the issue, one has to admit that mainstream climate scientists are fond of the “heads I win, tails you lose” approach.
Lol.
We need a new way of undressing those models.
This is a very long winded and wordy way of saying, don’t believe what you see in the ‘real’ world.
The authors are clearly anxious to provide a reason why actual observations do not align with computer models. Their claim the difference is due to natural variation is about as unconvincing for an excuse as it gets.
We have spend decades listening to so called climate experts telling us their computer models have shown dire warming and here are the graphs to prove it.
The graphs each and every one since the start of this computer model based crisis was introduced all show a pathetic corelation with recorded data.
The models do not accurately predict future climate conditions and are even incapable of hind casting which tells us the models lack important input components.
The modelers refuse to come to terms with their impossible task of incorporating cloud cover and cloud evolution into the models because they can never know those key drivers of weather/climate input.
The team presenting this feeble excuse for failure, should have said. “our models are very accurate in the model world we have created but bear no relation to events in the real world.”
The term ‘real’ is recognised and used by them in their study. That infers they know they are living in a make believe space, of their own construction.
Maybe they should invoke the Apollo mission scientists. ‘In data we trust, for anything else, bring us a computer model’…..
I’m taking note of “may”, “projections” and “suggest” in that one-sentence word salad.
Separately, specific to this the thrust of this article:
“If it disagrees with experiment (observation), it’s wrong. In that simple statement is the key to science. It doesn’t make any difference how beautiful your guess is, it doesn’t matter how smart you are, who made the guess, or what his name is — if it disagrees with experiment (observation), it’s wrong.”
Um…shouldn’t the models be taking “natural variability” into account? If not, what, exactly, are the models modeling?