This month’s open thread on climate topics. Please be respectful and substantive.

The post Unforced Variations: June 2026 first appeared on RealClimate.

The 4th UArctic congress on the Faroe islands finished with the message that the Arctic Council is still alive. It has overcome recent setbacks with difficulties concerning two of its member states, Russia and the US. The Arctic Council represents 8 nations together with indigenous peoples and has observers from around the world, and this wide-ranging diversity of course coloured the congress. It was allegedly the largest scientific conference on the Faroe islands ever, as it was combined with the Oceans Connectivity Conference.

I participated at the congress as a member of the Arctic Monitoring and Assessment Programme’s (AMAP) climate expert group to present the Arctic Climate Update reports for 2024 and 2026 (in progress).

Some of the topics that interested me included climate change, so-called “climate interventions”, and the Atlantic meridional overturning circulation (AMOC) which also has been discussed previously by Stefan. I noted that the question concerning changes to the AMOC is still being discussed, and scholars have different opinions on whether it will collapse or if it merely will weaken to a more modest degree.

While everybody at the congress agreed on climate change, the question concerning climate interventions, formally known as “geoengineering”, was highly controversial. I was invited to a side meeting organised by Ocean Visions on rescuing the sea ice. 

I expressed my doubts about the idea of geoengineering in the Arctic, as the region seems to have a semi-permanent cloud-cover that moderates the sea ice albedo during summer. The strongest Arctic warming by far, after all, takes place during the winter and the Polar Nights, in darkness when albedo cannot play a role. 

The sea ice does nevertheless play an important role for the warming, as retreating sea ice exposes a warmer ocean surface underneath. But sea ice is not necessarily the top amplifying feedback mechanism in the Arctic, according to AMAP’s Adaptive Actions in a Changing Arctic report for the Barents Sea region from 2017 (p.65). There are also some preliminary analyses suggesting that low-level cloudiness increases where the sea ice retreats. However, the cloud cover in different reanalyses and satellite observations are not entirely consistent, and more work is needed to address this question. 

Other reasons why spraying particles into the air is a bad idea is that it doesn’t deal with the oceanic acidification, which is particularly a problem for the cold oceans in the Arctic. It’s a big concern for marine ecosystems. Furthermore, it completely ignores the fact that the global hydrological cycle is entangled with the greenhouse effect and is also affected by our exploitation of fossile resources.

Finally, geoengineering opens a can of worms when it comes to conspiracy theories that may pave way for new unnecessary conflicts. We have plenty of such examples from the past: chemtrails, Haarp, Climate denial, 5G-COVID, and anti-vaxxers.

The post “The Arctic Council is not dead” first appeared on RealClimate.

The fantasy version of the normal updating of scenarios for a new round of CMIP simulations doing the rounds is bad faith BS.

As climate folk will know, the community is currently embarking on a new round of climate model simulations to support analyses and projections for the next IPCC report (due in 2028/9). This new effort has been dubbed CMIP7, because it is the sixth iteration of the CMIP effort (IYKYK), that started in the late 1990s. For each of these iterations, a new set of projections has been formulated for the modeling groups to use and the ones for this round were just published (van Vuuren et al., 2026). So far, so totally normal.

Why do scenarios need updating? Why can’t we use the three scenarios that Hansen et al. (1988) first came up with in the early 1980s? Three reasons. First the scenarios need to be continuous with the trajectories of the observed changes. The ‘join’ point was 1984 for Hansen’s original scenarios, then 2000 for the CMIP3, 2005 for CMIP5, 2014 for CMIP6, and it will be 2023 for CMIP7. As you can imagine, things have changed over the last 40 years (the Montreal Protocol, the Clean Air Acts, renewable energy price falls, fracking, the Paris Agreement, actual climate policies, reversal of climate policies, etc.). All of these things are the result of humans behaving in ways that humans behave and which are not easily predictable ahead of time. This is why future simulations have long been described as ‘projections’ and not predictions.

Second, the rationale for future scenarios has shifted in light of what we (as a society) are doing. At the beginning it made sense to think about a spread of baseline scenarios where no climate policy was enacted: “Business as usual” so to speak. But now? we have already done things and so ‘business’ is no longer ‘usual’. Now, ranges based on ‘current policies’, ‘current aspirations’ and ‘possible backsliding’ are perhaps more useful. Additionally, we are now much closer to 2100 than we used to be (also obvious, but often forgotten) and so scenarios need to be extended out further.

Third, what we are making scenarios for has expanded enormously. In 1984 there were only concentrations of well-mixed greenhouse gases, the solar cycle and the occasional volcano to project, but now, we have emissions of GHGs including CO₂, plus all of the halogenated gases, the short lived climate forcers (CH4, aerosols, NOx, SO₂/SO₄), land use change (deforestation, irrigation, agricultural shifts), possible anthropogenic impacts on dust and fire, and freshwater inputs from melting glaciers and ice sheets that are not otherwise represented in models. Did I mention nitrogen inputs, solar particle fluxes, and volcanic emissions of water vapor as well as sulfates? Keeping this all coherent and up-to-date is an enormous undertaking.

This all means that, duh, of course the scenarios would be updated for CMIP7.

People (hi Roger!) acting as if the publication of new CMIP7 scenarios is some huge policy shift or an admission that previous scenarios are no longer ‘official’ are just bull-shitting. This is something that was planned for and expected for literally years. Previous scenario sets were used in previous rounds, a new set will be used for the new round – that is all there is to it.

Oh noes!

The supposed focus of the ire are the high end scenarios of RCP85 (CMIP5) and SSP5-85 (in CMIP6). The reasons why these were set up in the first place (back in 2007!) was that IPCC wanted to span what had appeared in the literature before then – going beyond (in sophistication) what the (CMIP3) SRES projections had done. But the IAM folks involved decided (correctly) that they didn’t have the time to start from scratch, and so they decided to split up the task – come up with a spread that covered 99% of published scenarios with ‘representative’ concentration pathways (RCPs – gettit?) for the climate models to use quickly (in CMIP5), and back-fill plausible socio-economic pathways later (to be used in CMIP6). Thus the CMIP5 models (which were run in 2007-9 or so) used the RCPs (including RCP85 – which was so-called because it reached 8.5 W/m² of direct radiative forcing from GHGs in 2100).

For climate modellers, the reasons why the pathways are the way they are is a secondary concern – if they were only to be given CO₂ levels (and other GHGs etc.), the basic need is just for a low, middle and high scenario that encompass our most ambitious climate policy pathways, a worst case scenario (‘Burn it all!’), and something in the middle. When it came time for CMIP6 (2016-2019 or so), the SSPs (that had been promised a decade earlier) were ready, and so they were used. But for the purposes of the climate modellers, the drivers underlying the SSPs were not really that relevant. A climate model really doesn’t care how cooperative or antagonistic regional economic blocks are – it just responds to the resulting emissions. That there is a need for high end scenarios should be obvious – where are the tipping points in the system? what are the impacts of a 2ºC warmer world? what about 3ºC or 4ºC? Are these worth avoiding perhaps? Having seen these results, the answer (IMO) is definitely yes!

In the last few years, a number of people have pointed out that assumptions underlying the highest SSPs don’t look as plausible as they used to seem (this is also true of the lowest projections, but people seem less bothered by that). Note this is many years after all the models that were ever going to use them were run. But this is less of an issue that some people portray. Climate impacts rely on a chain of calculations – a specific set of emissions, a resulting concentration pathway, and a modelled sensitivity. Similar impacts can arise with lower emissions, but greater carbon cycle feedbacks and higher sensitivities, and given that each of these steps are quite uncertain, it is not really worthwhile for climate models (or modelers) to get too attached the specific storylines the IAM folks put together. Hence the collective shrug from climate modelers around the RCP85 ‘dialogue’ in the last couple of years.

Let me give two examples why high end scenarios are important: Impacts on ice sheets are a very important part of the climate change and have yet to be fully integrated into the standard climate models. So independent efforts with ice sheet models were set up using the output from the CMIP5 and CMIP6 models – they used two scenarios, RCP85 and RCP26 to bracket possibilities in ISMIP6 (Seroussi et al, 2020). Interestingly, they found that, particularly for Greenland, that none of the models had melt rates as high as observed even with RCP85 forcing. Thus for a situation where the (ice sheet) models are insufficiently sensitive, a higher than expected forcing might give you a more likely outcome.

A second example is the use of ‘global warming levels’ in the last IPCC assessment. These were averages of the models when they reached particular temperature levels (2ºC, 3ºC, 4ºC etc.), but for that to work, enough models had to reach those temperatures in order to make an average – and in practice for 3ºC and 4ºC, this was only possible with SSP5-85 scenarios. Other assessments used the higher signal-to-noise ratios in the high end scenarios to estimate sensitivities across many systems that would have been noisier and more uncertain if that was not available. How the models got there is basically irrelevant. The new high end runs will also be used for this (note that H gets to 8.5 W/m² only about 20 years after RCP85).

Note that even the harshest critics of RCP85 will admit (in academic circles at least) that these are legitimate uses. However, some of the more stupid commentaries equate the mere mention of RCP85/SSP585 with scientific misconduct, claiming that counting the number of times the ‘naughty’ words of RCP85 appear in publications or assessments is a damning indictment of the entire field’s integrity. This is so dumb and lazy that I find it hard to credit.

But wait!

The funny thing is that there are real issues with the way this whole endeavor has grown up. First, because CMIP is the only (serious) climate projection game in town, as climate change has become more salient, CMIP projections have been used to inform a far wider array of science than was imagined back in 2007. Not all of those uses are optimal. For instance, the ERA5 reanalysis still uses CMIP5 projections of solar forcing from 2008 which didn’t turn out to be so good at matching what actually happened. Consultants and banks have used CMIP6 projections as if they were real predictions, and the adaptation community have often assumed that specific CMIP6 pathways are the most likely outcomes.

All of these misuses are compounded by the fact that it appears to need a decade to update these pathways in the light of new science and societal decisions and changes. This is way too long – annual updates of the process should be achievable if funders prioritized it.

And finally, there are of course far more scenarios that would be interesting to explore in climate models (policy specific scenarios, delta scenarios (where only one thing changes at a time), in-between scenarios, annually updated scenarios etc.) than can possibly be performed given existing computational capacity. This is (right now) prohibitive, but at the rate that faster, more efficient machine-learning emulators are advancing, it might not be for much longer.

A serious critique of the climate modeling enterprise would be focused on these issues (for instance), rather than tilting at RCPs.

References

1. D. van Vuuren, B. O'Neill, C. Tebaldi, L. Chini, P. Friedlingstein, T. Hasegawa, K. Riahi, B. Sanderson, B. Govindasamy, N. Bauer, V. Eyring, C. Fall, K. Frieler, M. Gidden, L. Gohar, A. Jones, A. King, R. Knutti, E. Kriegler, P. Lawrence, C. Lennard, J. Lowe, C. Mathison, S. Mehmood, L. Prado, Q. Zhang, S. Rose, A. Ruane, C. Schleussner, R. Seferian, J. Sillmann, C. Smith, A. Sörensson, S. Panickal, K. Tachiiri, N. Vaughan, S. Vishwanathan, T. Yokohata, and T. Ziehn, "The Scenario Model Intercomparison Project for CMIP7 (ScenarioMIP-CMIP7)  ", 2025. http://dx.doi.org/10.5194/egusphere-2024-3765
2. J. Hansen, I. Fung, A. Lacis, D. Rind, S. Lebedeff, R. Ruedy, G. Russell, and P. Stone, "Global climate changes as forecast by Goddard Institute for Space Studies three‐dimensional model", Journal of Geophysical Research: Atmospheres, vol. 93, pp. 9341-9364, 1988. http://dx.doi.org/10.1029/JD093iD08p09341
3. H. Seroussi, S. Nowicki, A.J. Payne, H. Goelzer, W.H. Lipscomb, A. Abe-Ouchi, C. Agosta, T. Albrecht, X. Asay-Davis, A. Barthel, R. Calov, R. Cullather, C. Dumas, B.K. Galton-Fenzi, R. Gladstone, N.R. Golledge, J.M. Gregory, R. Greve, T. Hattermann, M.J. Hoffman, A. Humbert, P. Huybrechts, N.C. Jourdain, T. Kleiner, E. Larour, G.R. Leguy, D.P. Lowry, C.M. Little, M. Morlighem, F. Pattyn, T. Pelle, S.F. Price, A. Quiquet, R. Reese, N. Schlegel, A. Shepherd, E. Simon, R.S. Smith, F. Straneo, S. Sun, L.D. Trusel, J. Van Breedam, R.S.W. van de Wal, R. Winkelmann, C. Zhao, T. Zhang, and T. Zwinger, "ISMIP6 Antarctica: a multi-model ensemble of the Antarctic ice sheet evolution over the 21st century", The Cryosphere, vol. 14, pp. 3033-3070, 2020. http://dx.doi.org/10.5194/tc-14-3033-2020

The post Scenarios, schmenarios… first appeared on RealClimate.

This month’s open thread on climate topics. El Niño coming (shades of 2023), competitive predictions, ill-advised geo-engineering ideas, and massive shifts in renewable energy roll out. Surely something there to discuss…

The post Unforced Variations: May 2026 first appeared on RealClimate.

Somewhat belated open thread for this month! (Oops). Please stick to climate related topics, and remain respectful.

The post Unforced Variations: Apr 2026 first appeared on RealClimate.

Confirmation bias and a profound lack of curiosity mark the latest ABC (Anything But Carbon) contrapalooza in DC this week and a decade-old albedo error trips them up.

I occasionally dip into the contrarian-sphere to see if there is anything new that might be of actual interest. I am usually disappointed, and last week’s escapade was no different. The quality of the talks was pretty abysmal – bad slides, monotone reading of notes, and abundant errors, misunderstandings, fallacies and cherry picks but, if there was a theme, it was that everything was so complicated and uncertain that no-one can know anything. This is a notable contrast to previous outings where everything was definitely due to the sun or ‘natural’ variability (anything but carbon remains the organizing principle).

Multiple speakers (including Willie Soon, John Clauser) purported to be very irate that the CERES Earth’s energy imbalance (EEI) record is calibrated to the changes in the in situ heat content data (dominated by the ocean heat content changes). Quite why they were so exercised was a little mysterious because their sources of information on this topic were the papers that clearly explained why and how this was being done (i.e. Loeb et al. (2009) or Loeb et al. (2018)). [Basically, the satellite data for the EEI does not have a good enough absolute calibration to be an independent estimate, and so the CERES EBAF product is adjusted to match the (much better characterized) in situ heat gain (Jul 2005-Jun 2015) in a way that does not affect the trends]. Also the EEI based on in situ data is apparently wrong because the AI told them so. Ok then.

In both Soon’s and Clauser’s talk, a particular figure made an appearance – Fig. 11a from Stephens et al. (2015).

Unsurprisingly, this was used to claim that the CMIP5 models (and, by implication, all models) were terribly wrong, can’t be trusted etc. etc. Oddly, neither of them chose to show the comparison with the later CMIP6 models (Jian et al., 2020):

Or even the earlier CMIP3 models from Bender et al (2006):

Well, it’s not so odd, since these comparisons are much more favorable to the models. But lets look closer…

The CERES observations in the three plots do not agree at all! The 2015 figure has maximum albedo in March and October, while the other two have maxima in June and December – a 2 or 3 month phase shift. Something is wrong here. Fortunately, the CERES project has a very accessible website for downloading data, and it’s trivial to get the incoming solar flux and reflected solar flux for every month. The albedo is just the ratio, and we can average the months to create a climatology. The differences in the averaging periods makes no visible difference, and the differences in the EBAF version are likely to be minor (though that is harder to check). However, the bottom line is that the CERES data in the 2015 figure is wrong, while the 2006 and 2020 papers are correct.

We can speculate about what led to this (possibly related to the first month with data being March 2000 assumed to be January?), but there are two immediate consequences. First, the CMIP5 models (like the CMIP6 and CMIP3 models) turn out not to be so bad: phasing is ok, but the annual mean albedo can be a little variable. Second, it’s likely that the other panels in Fig 11, Figs. 5a-c, the discussion about them in section 6 etc. in the Stephens et al (2015) are also affected by this. Despite citing Bender et al (2006), and also Kato (2009) (see his figure 1a) who have it correct, the phase offset was not addressed. The Stephens et al paper has since been cited over 240 times, and it seems odd that no-one else had noticed this issue [Aside, if you know of a reference that does make this point, please let me know in the comments].

Why now?

Interest in the EEI is obviously growing, both as a function of the increasing length of the CERES timeseries and the fact that the EEI is growing. Even the WMO is elevating this metric in importance. So one might expect the contrarian-sphere to try and undermine it – that’s just what they do.

But here is the difference between doing real science and what is on show at the DC contrapalooza. Scientists are curious about what is actually going on. Given a discrepancy, they want to understand what’s happening. The changes in albedo over the CERES record are indeed interesting and a little challenging to explain (the CERESMIP project is looking into this in more detail), but the scientists’ goal is to dig deeper until it becomes clear. For Soon and Clauser, discrepancies are just weapons – they don’t care that something doesn’t look right – in fact they want it to look wrong regardless of whether it’s an error in an old paper, or an ambiguous statement that they can read uncharitably, or a genuine issue. Thus the chances of them checking into this themselves is zero – despite their frequent claims that they want to ‘follow the data’.

Do I expect everyone to check every figure in every paper they cite before using them in a presentation? No. But this example outlines how important open science is. When something comes up like this, people should be able to check quickly that the label and the contents match. It also highlights the danger of leaving issues uncorrected in the literature. I don’t know if this issue has been brought to the attention of the journal or the authors already, but even papers from a decade ago get cited and used (see here for another example). We owe it to everyone (yes, even the contrarians!) to make sure that the literature is as free of error as we can make it.

References

1. N.G. Loeb, B.A. Wielicki, D.R. Doelling, G.L. Smith, D.F. Keyes, S. Kato, N. Manalo-Smith, and T. Wong, "Toward Optimal Closure of the Earth's Top-of-Atmosphere Radiation Budget", Journal of Climate, vol. 22, pp. 748-766, 2009. http://dx.doi.org/10.1175/2008JCLI2637.1
2. N.G. Loeb, D.R. Doelling, H. Wang, W. Su, C. Nguyen, J.G. Corbett, L. Liang, C. Mitrescu, F.G. Rose, and S. Kato, "Clouds and the Earth’s Radiant Energy System (CERES) Energy Balanced and Filled (EBAF) Top-of-Atmosphere (TOA) Edition-4.0 Data Product", Journal of Climate, vol. 31, pp. 895-918, 2018. http://dx.doi.org/10.1175/JCLI-D-17-0208.1
3. G.L. Stephens, D. O'Brien, P.J. Webster, P. Pilewski, S. Kato, and J. Li, "The albedo of Earth", Reviews of Geophysics, vol. 53, pp. 141-163, 2015. http://dx.doi.org/10.1002/2014RG000449
4. B. Jian, J. Li, Y. Zhao, Y. He, J. Wang, and J. Huang, "Evaluation of the CMIP6 planetary albedo climatology using satellite observations", Climate Dynamics, vol. 54, pp. 5145-5161, 2020. http://dx.doi.org/10.1007/s00382-020-05277-4
5. F.A. Bender, H. Rodhe, R.J. Charlson, A.M.L. Ekman, and N. Loeb, "22 views of the global albedo—comparison between 20 GCMs and two satellites", Tellus A: Dynamic Meteorology and Oceanography, vol. 58, pp. 320, 2006. http://dx.doi.org/10.1111/j.1600-0870.2006.00181.x
6. S. Kato, "Interannual Variability of the Global Radiation Budget", Journal of Climate, vol. 22, pp. 4893-4907, 2009. http://dx.doi.org/10.1175/2009JCLI2795.1

The post A reflection on reflection first appeared on RealClimate.

We previously highlighted Roy Spencer’s poor practices in comparing models with observations, but we’ve now dug down a little deeper, and it’s not pretty.

A few weeks ago, the email exchanges of the CWG authors were published after a court ordered them to be made public. In them, there is an interesting exchange between Steve Koonin and Roy Spencer. Koonin wanted Spencer to address the (obvious) complaint that Roy’s comparison of ‘Corn Belt temperature trends’ figure was a cherry pick.

Roy agrees to look into it, but whether he ever got back to Koonin is unclear. In any case, no public statements or responses have been made. The conversation did however reveal where the data came from and Roy’s method for making the comparison, inspiring me to try and replicate the analysis more appropriately. So let’s see what we can find out.

NOAA Climate Divisions

NOAA has a great website with its Climate Division data (ClimDiv) which is an aggregated product from the individual station data, but averaged at the division, state, and regional levels. It has averages for 9 regions of CONUS, the big river basins, NWS areas, and multiple agricultural regions. The Corn Belt map for averages is below:

Spencer most likely used this precomputed index (across 11 states, though his figure says 12 – he might have tried making his own 12-state index but I don’t think it matters much, other than making it trickier to replicate). Anyway, the Corn Belt data (code 261 for the area weighted observational data) can be downloaded directly and the JJA trends computed for the 50 year period 1973-2022 (following Spencer’s graph). I’m sure it will come as a great shock to our readers that of all the regions computed in the NOAA ClimDiv dataset over this period, the Corn Belt index has almost the lowest summer trend (0.1ºC/dec though with a wide (95%) confidence interval [-0.07,0.27]). Quel surprise!

Note that the annual trends are much less noisy and more similar across the regions (but that wouldn’t be as useful, now, would it?).

This can be compared with the CMIP6 models data from a roughly analogous area (100-81ºW, 40-46ºN). This is relatively easy to extract from the ClimateExplorer website for 144 individual simulations (using the historical + SSP245 scenario). Spencer (I think) chose the option to download the ensemble mean for each model (or perhaps just a single run from each model). In either case, he discards very relevant information from the ensemble for each particular model.

So let’s plot the JJA trends for the Corn Belt region, including two elements that Spencer ignored. First, the uncertainty in the OLS trend in the observations (which is large) and, second, the spread across the individual model ensembles. As we did previously, we can plot the trends against the climate sensitivity (here I’ll use the Transient Climate Response) to see the difference that the ‘hot models’ make (IPCC AR6 assessed that the likely range of TCR was 1.4-2.2ºC, very likely 1.2-2.4ºC). Note that noisier the statistic (short periods, small regions, etc.) the less clear any difference related to climate sensitivity will be.

As in Spencer’s original plot, the bulk of the simulations have a larger trend than observed (all but 2 in fact). However, we are not done. Recall that the simulations are an ensemble of opportunity. Some of the models have sufficient ensemble members to reasonably estimate the standard deviation and the 95% spread for the trend in that model, but others have only one member and no spread can be calculated. If we assume that the average standard deviation (in this diagnostic it’s 0.11ºC/dec), is a good estimate in those cases, we can plot the ensemble means, along with an estimated 95% spread, for each model.

Now we have a slightly more regular statistical comparison, and we can see that this observation is within the 95% spread of about half a dozen models. Thus it’s a plausible (if not likely) match. Given the large number of comparison one could theoretically do, if one area (specifically chosen) is an outlier is not so surprising.

However, the point of the criticism of Spencer’s figure to begin with was not that these Corn Belt temperature trends have been well predicted, but rather that this index was cherry picked to provide the worst possible comparison. So what would a different index have given? I have not checked them all [The other crop regions are not so easily translated into lat/lon rectangles!], but I did look at the NorthWest region (124-111ºW, 42-49ºN) – this region has warmed more than the CONUS average by about the same degree that the Corn Belt region was below.

And… there is a much stronger coherence with the ensemble. But this would not have provided as useful a talking point of course. Does this mean the models are perfect? Of course not. A worthwhile analysis would have looked at a spread of such comparisons and made a statement about the utility of the models based on that collective analysis. Just looking at the one or two small areas or single seasons is not going to be informative.

Useful vs. Useless model-observation comparisons

My point here is not to discuss the utility of the CMIP6 projections for the corn, wheat, soy regions of the US etc. That is best left to people who have better domain knowledge about those applications. However, I do want to (again) stress a few points:

• Observations have uncertainty too. Whether it is in the linear trend and/or structural, it needs to be accounted for in any comparison.
• Observational data comes from a single realization. You need to think about what the irreducible uncertainty associated with internal variability means.
• The model ensembles are complex. You cannot substitute what is easily downloadable from ClimateExplorer for an analysis in and of itself.
• Whether the observed trend is or is not close to the model ensemble mean(s) is not a good test of the skill of the model(s) or multi-model ensemble.
• The appropriate test is whether the real world observation is exchangeable with an ensemble member from the model. The visual comparisons shown above address this, but it can also be done quantitatively in ways that take into account both the observational uncertainty and spread.
• Comparisons such as presented by Spencer in the DOE or Heritage Foundation reports are fundamentally flawed since they never deal seriously with any of this, and will continually be called out as cherry picking.

One last thought. Steve, take a moment and think about why Roy didn’t ever address the critiques. Just spitballing here, but the intersection between people generally thought of as ’eminent’ scientists and folks that engage in this kind of hackery is empty.

The post Spencer’s Shenanigans: Part II first appeared on RealClimate.

The mystery of why the last million or so years of glacial variability are so different to what came before just got more mysterious…

It’s easy to understand why the ice ages have such a hold on our imaginations. Putting aside the cavemen, woolly mammoths, and sabre-toothed tigers of popular culture, the scientific questions around the pacing of the glacial cycles, their magnitude, variability, and impacts are truly profound.

Despite huge strides in understanding the ice ages – from the ground-breaking work of Hayes, Imbrie and Shackleton (1974) that demonstrated the skill of the Milankovitch model in the 1970s, the paradigm-busting results from the Greenland Ice Cores in the 1990s, the discovery of the Heinrich events, etc., there remain plenty of real and abiding mysteries including:

• Why are the 100kyr cycles so strong?
• What are the details of the carbon feedbacks on glacial-interglacial cycles?
• What triggered the ice ages in the first place? (i.e. why did the impact of Milankovitch cycles get much larger over the last 2.5 million years?)
• Why didn’t humans develop agriculture in the last interglacial?
• What triggers the Dansgaard-Oeschgar oscillations?
• and… what caused the change from lower magnitude 40kyr cycles to 100kyr cycles across the Mid-Pleistocene Transition (MPT)?

We have good evidence from the deep Antarctic ice cores of the coupling between CO₂ and temperature over the last 800kyrs and from ocean sediment proxies, we have reasonable estimates of the coupling between CO₂ and temperature over the long cooling during the Cenozoic (the last 65 million years). But, until now, we haven’t been able to really examine that intervening period – the early Pleistocene.

Theories, of course, abound. The obvious one is that the long term declines in CO₂ crossed a threshold that allowed for larger ice volumes that had more resonance with the 100kyr cycles. Another is that the early ice advances (which were more spread out but less voluminous) scraped all the soils off the rocks and that subsequent ice sheets were less mobile. I think most folks expected the data (when it arrived) to basically confirm what people expected.

But sometimes the observations don’t confirm your preconceived notions. The nice thing about science is that scientists (ideally) tend to get excited at this point (instead of, say, trying to deny the new information). So what has just happened?

Two new papers, Marks-Peterson et al. (2025) (direct link) and Shackleton et al. (2025) (direct link) in Nature this week report on analyses of very old Antarctic ice. These samples come from the “blue ice” in the Allan Hills in Antarctica where multi-million year old ice surfaces after having been deposited and transported over large distances. This is quite distinct from deep drilling in places where you hope the ice has not moved much, and while it doesn’t have the nice stratigraphy of the cores, you can sample snapshots of the atmosphere over a much longer time – in this case, almost 3 million years – albeit with coarser dating.

There are two main measurements presented. The first are the GHG concentrations in the air bubbles trapped in the ice (Fig. 1), and the second is a record of mean ocean temperature inferred from the ratio of noble gases in the air bubbles (Fig. 2).

The first and most dramatic (or rather, non-dramatic) result, is that CO₂ levels appear to have barely changed (on average) over this key period – dropping only 20-30ppm over the onset period. That isn’t nothing, but it’s only about 0.45-0.7 W/m² in forcing, and would lead to around 1ºC in global surface cooling. The CH₄ levels might have been expected to fall too, but they seem to be static. [Note that this method is not sampling the glacial/interglacial variations which are apparent in the more recent records]. The second, and somewhat confounding, result is that the global ocean seems to have cooled by about 2ºC over the same time period (with the global surface temperature change would have been larger).

So we have a conundrum. The onset of NH glaciation did happen as the planet cooled (as might be expected), but the first guess for what caused that cooling (long term trends in CO₂ and/or CH₄) does not appear to work.

How might this be resolved?

There are always multiple potential ways out of a conundrum: subsequent analyses might find an issue with the observations, there might be a hyper-sensitivity to the small CO₂ changes at this time (but why?), there might be something else driving the change (volcanism? dust aerosols?), or… what? None of these possibilities are obvious winners, and of course, they are not mutually exclusive. Eric Wolff (direct link) in his commentary seems to think that the ocean is doing the driving, but I think that might be backwards.

The funny thing is that paleo-climatologists have been wanting these old ice analyses for a long time – with the anticipation that they would help answer these questions. But they seem to be posing many more questions than they have answered.

Broader issues

One thing this shows is that scientists can’t be complacent. As we’ve seen with surprising climate events even over the last few years (2023, Antarctic sea ice, the increases in the Earth’s Energy Imbalance), the more you look at the planet (or even the universe) the more surprising things you find. Science is an active search for deeper understanding – and we are not done yet.

Final thought

At face value, these results seem to suggest that CO₂ declines were not the dominant/only cause of the cooling at the onset of the ice ages, despite expectations. Some of the usual suspects are certainly going to claim (fallaciously) that this means that CO₂ can’t be the cause of anything. This is obviously a stupid argument so feel free to judge anyone that makes it.

Nonetheless,…

There are more things in heaven and earth, Horatio, than are dreamt of in your philosophy.

References

1. J.D. Hays, J. Imbrie, and N.J. Shackleton, "Variations in the Earth's Orbit: Pacemaker of the Ice Ages", Science, vol. 194, pp. 1121-1132, 1976. http://dx.doi.org/10.1126/science.194.4270.1121
2. H. Heinrich, "Origin and Consequences of Cyclic Ice Rafting in the Northeast Atlantic Ocean During the Past 130,000 Years", Quaternary Research, vol. 29, pp. 142-152, 1988. http://dx.doi.org/10.1016/0033-5894(88)90057-9
3. J. Marks-Peterson, S. Shackleton, J. Higgins, J. Severinghaus, Y. Yan, C. Buizert, M. Kalk, R. Beaudette, V. Hishamunda, D. Eves, A. Carter, A. Kurbatov, J. Epifanio, J. Morgan, I. Nesbitt, M. Bender, and E. Brook, "Broadly stable atmospheric CO2 and CH4 levels over the past 3 million years", Nature, vol. 651, pp. 647-652, 2026. http://dx.doi.org/10.1038/s41586-025-10032-y
4. S. Shackleton, V. Hishamunda, Y. Yan, A. Carter, J. Morgan, J. Severinghaus, S. Aarons, J. Marks-Peterson, J. Epifanio, C. Buizert, E. Brook, A.V. Kurbatov, M.L. Bender, and J. Higgins, "Global ocean heat content over the past 3 million years", Nature, vol. 651, pp. 653-657, 2026. http://dx.doi.org/10.1038/s41586-026-10116-3
5. E.W. Wolff, "Climate snapshots trapped in ancient ice tell a surprising story", Nature, vol. 651, pp. 592-593, 2026. http://dx.doi.org/10.1038/d41586-026-00636-3

The post The Puzzling Pleistocene first appeared on RealClimate.

Guest commentary from Nathan Lenssen (Colorado School of Mines)

A new analysis of historical temperatures suggests that things are getting warmer faster, but what does it mean for the future?

A study (Foster & Rahmstorf 2026) was published on Friday claiming evidence that “Global Warming Has Accelerated Significantly”. This study is an update by the authors of a similar study they published in 2011 where they found no statistical evidence for an acceleration in global warming. Both studies sought to determine if there is a detectable acceleration in warming, after statistically removing the effects of ENSO, volcanoes and changes in solar forcing from the observed global mean temperature (GMT) series (through to 2024).

As I’ll discuss further below, there was no detectable acceleration in the raw GMT series – this doesn’t mean there isn’t any, but that the noise (internal variability etc.) doesn’t allow us to see if there is clearly. Thus, the study has detected an acceleration in the rate of warming of inferred long-term trends – which we can pretty confidently attribute to anthropogenic effects. This study has understandably gotten substantial attention in the media. Here, I will outline what I think we have learned from this study, what this means for our understanding of the current state of the climate system, and what it means for projections of climate change (Hint: not much).

FR26 make three contributions in this recent work: (1) the production of an “adjusted” GMT series that removes statistically estimated impacts of a few short term changes in GMT, to hopefully leave just the warming associated with changes in anthropogenic forcings, (2) the detection of an acceleration in the rate of warming on this series using three different statistical methods, and (3) a forecast that 1.5ºC warming will be reached by ~2030. The methods used here are generally sound, particularly by engaging with the state of the art in changepoint detection methods as one of the methods for acceleration detection (Beaulieu et al. 2024). The figure below shows the three statistical methods for detecting changes in trend, all of which provide statistically significant evidence that the recent trend is faster than previous trends. 

 Given the assumptions made by the authors, this provides statistically robust evidence that acceleration has been detected. On first glance, this may be surprising or alarming as, to the zeroth order from our understanding of the Earth’s system’s response to CO₂, we expect a roughly generally linear warming in GMT given the exponential rise in CO₂ due to the log-scaling of GMT with CO2. Acceleration could be the result of the decrease of cooling anthropogenic forcings (as is hypothesized for some regional accelerations detected in Beaulieu et al. 2024) or substantial feedbacks/tipping points that are causing the Earth to warm faster than the simple CO₂ forcing physics dictates. Note though that the climate models that are used to inform our future projections also expect an acceleration around now (of course, given the assumptions that went into them).

However, as the authors point out, their method of ‘removing’ ENSO could be improved (for instance, Compo and Sardeshmukh (2010)), and there is still some imprint of natural climate variability in their adjusted time series. Note that an estimate of the “true” natural variability of the climate system, and correspondingly the “true” forced response, is one of the white whale problems in climate science! FR26 does an credible, but necessarily imperfect, job of isolating the forced response, but don’t account for this uncertainty in their statistical tests.

While we can’t know the true internal variability perfectly, we have climate models which provide an estimate of this variability. The figure below shows that the CMIP6 models (screened for a likely Transient Climate Response (TCE)) have a spread that fully contains the observed climate signal. Notably, the ensemble mean of these models demonstrates a slightly greater than linear warming (minus the effects 1991 eruption of Mount Pinatubo).

We can look at this more closely. If we look at the trends in individual model simulations for the last 13 years (2013-2025) and the 13 years before that (2000-2012), on average, the models show a slight acceleration over the same period highlighted by FR26 (0.18ºC to 0.30ºC). However, while there is a difference in the mean of these distributions, they are not clearly separate. This shows that, at least in model land, the acceleration in trend (given the internal variability and model uncertainty) is going to be difficult to detect. Note that comparisons between the models and the real world are complicated by any divergences in the forcings in the scenarios (designed more than 15 years ago) and what actually happened (Hunga Tonga, the IMO regulations, Chinese aerosol decreases etc.).

So where does this leave us? There is no detectable acceleration in the raw observed GMT, but there is an acceleration in GMT when removing the linear effects of ENSO, volcanoes, and solar variability, and there is slight acceleration in GMT when estimated using a multi-model ensemble of climate models. John Kennedy recently discussed some of these results in the context of FR26, expanding to a wider discussion of estimates of warming rate. He hits the nail on the head by pointing out two key open questions: “If there is an acceleration, what is physically driving it?” and “What will happen to the warming rate in the future?” The question about mechanism is key to trustworthy predictions of the future rate, and this is not addressed in the new paper.

The prediction of 1.5ºC warming by ~2030 made in FR26 is made in this context by estimating the rate of warming in this adjusted GMT. While made in the imperfect context discussed here, this estimate is reasonable when compared to a more comprehensive attempt to estimate this date . However, as John states, we already know the planet was warming, we have some evidence for acceleration, but we need a better path forward to predict how GMT and subsequent regional climate will change under continued CO₂ emission

References

1. G. Foster, and S. Rahmstorf, "Global Warming Has Accelerated Significantly", Geophysical Research Letters, vol. 53, 2026. http://dx.doi.org/10.1029/2025GL118804
2. G. Foster, and S. Rahmstorf, "Global temperature evolution 1979–2010", Environmental Research Letters, vol. 6, pp. 044022, 2011. http://dx.doi.org/10.1088/1748-9326/6/4/044022
3. C. Beaulieu, C. Gallagher, R. Killick, R. Lund, and X. Shi, "A recent surge in global warming is not detectable yet", Communications Earth & Environment, vol. 5, 2024. http://dx.doi.org/10.1038/s43247-024-01711-1

The post How robust is our accelerometer? first appeared on RealClimate.

This month’s open thread for climate related topics.

The post Unforced Variations: Mar 2026 first appeared on RealClimate.