Steve January 11, 2021

Six scientists analysed the article and estimate its overall scientific credibility to be ‘low’. more about the credibility rating

A majority of reviewers tagged the article as: Cherry-picking, Exaggerating, Misleading.


The article in The Guardian misleadingly claims that the latest generation of climate models are more sensitive to carbon emissions than scientists previously thought. Instead, reviewers of this article state that the best estimate of climate sensitivity, or the amount that global temperature rises in response to increased CO2, remains within the range presented in the IPCC 5th Assessment Report[1]. Some updated climate models have exhibited higher climate sensitivities, but research so far suggests that is unlikely to be accurate.

Several reviewers stated that the article is based on one model that showed high climate sensitivity (described in a news commentary in Nature[2,3]), while overlooking the model’s limitations. As the reviewers explain below, the model performed well in short-term forecasts, but did not accurately simulate the global warming trend over the last century (see figure below)[4].

Figure—The new model (blue and red) shows a cooling trend in global mean land temperature from 1900-1980 and a much higher warming trend since 1980 than observed temperatures (black)[4].

Although the article in the Guardian mentions that scientists are still evaluating this model, the headline and introduction mislead readers by stating that “climate worst-case scenarios may not go far enough,” and that projections of the future “may need to be revised.” These claims overstate scientific confidence in the new models. More broadly, the article overlooks other climate models and scientific research on climate sensitivity and how it is influenced by clouds[5-7].

See all the scientists’ annotations in context. You can also install the Hypothesis browser extension to read the scientists’ annotations in context.


These comments are the overall assessment of scientists on the article, they are substantiated by their knowledge in the field and by the content of the analysis in the annotations on the article.

Richard Betts, Professor, Met Office Hadley Centre & University of Exeter:

Equilibrium climate sensitivity (ECS) is a very important topic of great interest, but also complex and still subject to considerable uncertainty and ongoing research. The article does recognise the uncertainty in the latest models and quotes a leading scientist in this regard, but the headline and first two sentences overstate the scientific confidence and give the impression of a more substantial update to scientific understanding than is actually the case. The caveat in the second sentence reads as if the high ECS in the new models is provisionally accepted and merely needs further checking, rather than being something that the modellers are not yet sure about either way. The article is not clear on the uncertainty in the previous estimates of ECS and gives the impression that there was previously a fairly precise estimate of “around 3°C”, but in fact that was already fairly uncertain, as explained by the IPCC[1]. The latest research would be more accurately characterised as still not yet ruling out high climate sensitivity that was previously thought possible but relatively unlikely, rather than suggesting that high climate sensitivity is now more likely than previously believed.

Mark Richardson, Research Assistant, UCLA/NASA JPL:

This article’s main theme is broadly accurate. Many new climate models, which better fit satellite cloud measurements, have a high value of “climate sensitivity.” It includes important doubts about how “such a high figure does not fit with historical records”.

Technical errors like confusing the national origin of a model and the description of a provided percentage don’t affect these main conclusions.

Readers should be aware that this article is slanted to the “shock” and “surprise” of these results and misses context from past work, such as the uncertainty ranges provided in past UN climate reports. However, the article is clear about these being mostly model results, reports how only around a quarter of these new models show very high sensitivity, and mentions the counterbalance of evidence from the historical record. Taken together I mark this article as “accurate”, with elements that are “exaggerating” and “imprecise/unclear”.

Reto Knutti, Professor, ETH Zürich:

The main problem with this Guardian article is not incorrect statements, but that it is cherry picking one single result, misinterpreting it, surround it by strong quotes from people who have no expertise in that area, thereby painting a doomsday scenario (“modelling suggests climate is considerably more sensitive to carbon emissions than thought”) that is highly misleading and completely unsupported by the evidence.

The oversimplification is in two steps, the first in a Nature comment on the original paper[3]. Here Tim Palmer is incorrect in saying that the results of the original paper “support the estimates [for climate sensitivity]”[2,3]. The fact that the new high climate sensitivity model does well on six hour forecasts does not imply a) that the model is fine in general, and does not imply b) that it is the only model that is capable of doing that. The relationship between short-term forecasts and climate feedbacks is not demonstrated, and the evidence from hundreds of other papers on the topic is ignored.

On a) the agreement on short-term forecasts simply means that this model is doing this particular thing well. However, this particular model is one of the two worst in simulating the warming over the past 40 years. It shows basically no warming globally until 1980 or 1990, even cooling over land, and a massive surge after about 1990[4]. The details are likely complicated, but are probably related to too-strong feedbacks (high climate sensitivity), compensated by too strong aerosol cooling until 1990, a hypothesis put forward ages ago in energy balance models. Indeed the Guardian piece mentions that, but only briefly at the end.

On b) as pointed out by others, nobody has demonstrated that the high sensitivity is related to short-term forecast skill in these newer models. Everything is different in the new model, so the improvement may have come from some other change. One would have to demonstrate that many versions of that model with low climatology are doing significantly worse on short-term forecasts, and would have to demonstrate that this is also true for models from other centers.

The fact that one particular test cannot rule out a high climate sensitivity does not make it likely. It simply means that we do not know. But there are other lines of evidence that point to the canonical 2-4.5°C range. One is the recent warming since 1950 or so, the other is paleoclimate estimates, the third is process understanding and feedback estimates from cloud data and surface observations

We and others have shown recently that almost all of the high climate sensitivity models in fact tend to overestimate recent warming[5], the Met Office model being one of the two worst, and taking that into account suggests that many of the new CMIP6 models are biased high, and that future warming is similar to what it was in earlier models. While it is correct that we are seeing models with high climate sensitivity, the evidence is growing that there are issues with at least some of these models.

In summary, the Guardian article is cherry-picking a single technical paper and over-interpreting it as being relevant for the prediction of long term warming, without sufficient context on the vast amount of literature that does not support such a conclusion.

Joeri Rogelj, Research Scholar, International Institute for Applied Systems Analysis (IIASA):

The article correctly reports that the most recent versions of some climate models estimate more warming for a given increase in CO2 concentrations. It is also correct in highlighting that how clouds are represented in these models is the likely reason for these higher estimates. However, it does not report all the science available on this topic and its claims are thus misleading. The article suggests that the only information currently available from scientific studies is that things look much worse, and that otherwise more research is needed to understand whether this is really the case. However, available studies that have looked at what these new model projections mean have found that models with higher warming are worse in capturing global warming trends over the past decades, making their projections of very high warming less probable[4]. This is important context that was omitted from the article.

Paulo Ceppi, Lecturer, Imperial College London:

The article suggests that new modelling results have led scientists to revise their understanding of the role of clouds for climate change. This is misleading. Current scientific consensus is that clouds will amplify warming, but most evidence does not point towards an extreme amplifying effect. So while extreme high warming scenarios cannot be entirely ruled out, there is no convincing evidence that the amplifying effect of clouds is stronger than we previously thought.

While the new Nature paper cited in the article is thought-provoking[2], its significance is overstated, and the paper should have been discussed in the context of other recent evidence that does not support the high-sensitivity modelling results.

Piers Forster, Professor, University of Leeds:

In this article there is some science on the latest climate models and their cloud feedbacks that have been reported before. However, quotes by Rockstrom and the writer [Watts] are unsupported opinions that make misleading interpretations of the new science that warming will be worse than we thought. These claims are not supported by other lines of evidence showing our estimates of warming rates have been stable over time.

Mark Zelinka, Research Scientist, Lawrence Livermore National Laboratory:

I would characterize the article as overstating the importance of a single result from a single model without sufficient context. A highly sensitive model might have a great-looking climatological state and perform really well on short-term weather forecasts, but even a broken clock is right twice a day. This is hyperbole of course: This is an excellent model and by no means “broken”, but it remains to be rigorously shown that this model is giving a more accurate picture of our future than what we previously thought. More generally, the consequences of continued greenhouse gas emissions are dire enough if climate sensitivity is in line with the existing scientific consensus; whether this new more sensitive model is more accurate is mostly immaterial to that fact.


The statements quoted below are from the article; comments are from the reviewers (and are lightly edited for clarity).

“Worst-case global heating scenarios may need to be revised upwards in light of a better understanding of the role of clouds, scientists have said.”

Richard Betts, Professor, Met Office Hadley Centre & University of Exeter:

This appears to be based on the view of a scientist who was not involved in developing the new models. A quote later in the article shows that a scientist with a leading role in the latest models is much more cautious and said it needs more research before conclusions can be drawn.

Piers Forster, Professor, University of Leeds:

There is no consensus in the community what these would be. Different scientists have different views, so there is no standard worst-case scenario to revise, see our CCC blog.

“Recent modelling data suggests the climate is considerably more sensitive to carbon emissions than previously believed”

Richard Betts, Professor, Met Office Hadley Centre & University of Exeter:

This statement is not correct. Equilibrium Climate Sensitivity (ECS) has always been quite uncertain, and the high values in the new models are within the range previously thought possible, although relatively unlikely. A value of around 5°C in some of the CMIP6 models is outside the “likely range” of ECS assessed in the IPCC 5th Assessment Report (AR5), but it is still not in the AR5 “very unlikely” range of above 6°C[1]. The recent studies cited in this report conclude that the high ECS values cannot be ruled out by the methods used in those studies, but they do not conclude that high ECS is more likely than previously believed.

Piers Forster, Professor, University of Leeds:

The latest models might also be over sensitive[6].

“25% of them show a sharp upward shift from 3C to 5C in climate sensitivity”

Richard Betts, Professor, Met Office Hadley Centre & University of Exeter:

It’s not clear what this statement is based on, as no source is given. Although several CMIP6 models produce an ECS above the upper end of the range of the CMIP5 models, this range was 1.2°C – 4.7°C so it is not clear where the description of a shift from 3°C to 5°C comes from.

Mark Richardson, Research Assistant, UCLA/NASA JPL:

I don’t see where this statement comes from either. The top 25% of models out now have climate sensitivity of about 4.8°C or above, so one part of this statement gives a broadly accurate summary. But several of those are new versions of models that previously had climate sensitivity quite a bit higher than 3°C, so there wasn’t a “shift” of those models.

“For 40 years, it has been around 3C”

Richard Betts, Professor, Met Office Hadley Centre & University of Exeter:

Actually the estimates of Equilibrium Climate Sensitivity (ECS) have long been around 1.5°C – 4.5°C, with higher values still thought possible. The IPCC 5th Assessment Report judged that there was up to a 10% chance of ECS being greater than 6°C, and the likelihood of an ECS of 5°C was assessed as less than 33% but more than 10%[1].

Piers Forster, Professor, University of Leeds:

I completely agree with Richard Betts, and in fact the latest models agree very well with this range of ECS in general[6].

“He said climate sensitivity above 5C would reduce the scope for human action to reduce the worst impacts of global heating. “We would have no more space for a soft landing of 1.5C [above preindustrial levels]. The best we could aim for is 2C,” he said.”

Mark Richardson, Research Assistant, UCLA/NASA JPL:

This is a fair interpretation of the consequences: climate sensitivity of 5°C would make it almost impossible to achieve the stated temperature targets.

But there is still plenty of evidence for climate sensitivity being in the historically likely 1.5 – 4.5°C range, so this should be kept in mind.

“the EU’s Community Earth System Model”

Mark Richardson, Research Assistant, UCLA/NASA JPL:

The Community Earth System Model is from US institutions, but it’s accurate to say that it’s from “leading research bodies” and has a high climate sensitivity.

“Previous IPCC reports tended to assume that clouds would have a neutral impact because the warming and cooling feedbacks would cancel each other out.”

Mark Richardson, Research Assistant, UCLA/NASA JPL:

“Assume” is ungenerous—this was based on calculations using physics and the available observations. For example, in 2007 they looked at how climate model clouds affected their climate sensitivity and talked about a “very large inter-model spread”, i.e., you could get net warming or cooling, and we didn’t have the information to rule out either[8].

Reto Knutti, Professor, ETH Zürich:

IPCC does not “assume”, they quantify based on observations and models. And while cloud feedbacks are uncertain, in 2013 the conclusion was already that the overall effect was likely positive (i.e., contributing to warming), specifically only 17% probability for it being negative (cooling).[1]

IPCC AR5 WG1 page 592: “Based on the preceding synthesis of cloud behaviour, the net radia­tive feedback due to all cloud types is judged likely to be positive. This is reasoned probabilistically as follows. First, because evidence from observations and process models is mixed as to whether GCM cloud feedback is too strong or too weak overall, and because the positive feedback found in GCMs comes mostly from mechanisms now sup­ported by other lines of evidence, the central (most likely) estimate of the total cloud feedback is taken as the mean from GCMs (+0.6 W m–2 °C–1). Second, because there is no accepted basis to discredit individual GCMs a priori, the probability distribution of the true feedback cannot be any narrower than the distribution of GCM results. Third, since feed­back mechanisms are probably missing from GCMs and some CRMs suggest feedbacks outside the range in GCMs, the probable range of the feedback must be broader than its spread in GCMs. We estimate a probability distribution for this feedback by doubling the spread about the mean of all model values in Figure 7.10 (in effect assuming an additional uncertainty about 1.7 times as large as that encapsu­lated in the GCM range, added to it in quadrature). This yields a 90% (very likely) range of −0.2 to +2.0 W m–2 °C–1, with a 17% probability of a negative feedback.”[1]

“But in the past year and a half, a body of evidence has been growing showing that the net effect will be warming. This is based on finer resolution computer models and advanced cloud microphysics.”

Mark Richardson, Research Assistant, UCLA/NASA JPL:

I think this is generally fair: the available satellite data in particular have shown that clouds seem to act in a way that coincides with higher climate sensitivity.

The fusion of satellite data with climate models has been vital for this.

“Scientists caution that this is a work in progress and that doubts remain because such a high figure does not fit with historical records.”

Richard Betts, Professor, Met Office Hadley Centre & University of Exeter:

It’s correct to have included this, but the message here seems less confident than that conveyed by the headline and opening sentence.

“Catherine Senior, head of understanding climate change at the Met Office Hadley Centre, said more studies and more data were needed to fully understand the role of clouds and aerosols. “This figure has the potential to be incredibly alarming if it is right,” she said. “But as a scientist, my first response is: why has the model done that? We are still in the stage of evaluating the processes driving the different response.”

Richard Betts, Professor, Met Office Hadley Centre & University of Exeter:

It’s correct to have included this statement from a leading authority on the new models, but the headline and opening sentence of the article do not reflect the caution given here.

“The more we learn, the more fragile the Earth system seems to be and the faster we need to move,” he said. “It gives even stronger argument to step out of this Covid-19 crisis and move full speed towards decarbonising the economy.”

Piers Forster, Professor, University of Leeds:

In fact the more we learn, quite often we find that nature is quite robust, even when trying our best to destroy it, see our CCC blog.



  • 22 June 2020: This post was updated to include a comment by Mark Zelinka

Read More