Do CMIP5 models skillfully match actual warming?

by Nic Lewis

Why matching of CMIP5 model-simulated to observed warming does not indicate model skill

A well-known Dutch journalist, Maarten Keulemans of De Volkskrant, recently tweeted an open letter to the Nobel-prizewinning physicist Professor Clauser in response to his signing of the Clintel World Climate Declaration that “There is no climate emergency”, asking for his response to various questions. One of these was:

The CLINTEL Declaration states that the world has warmed “significantly less than predicted by (the) IPCC”. Yet, a simple check of the models versus observed warming demonstrates that “climate models published since 1973 have generally been quite skillful predicting future warming”, as Zeke Hausfather’s team at Berkeley Earth recently analysed.

The most recent such analysis appears to be that shown for CMIP5 models in a tweet by Zeke Hausfather, reproduced in Figure 1. While the agreement between modeled and observed global mean surface temperature (GMST) warming over 1970–2020 shown in the Figure 1 looks impressive, it is perhaps unsurprising given that modelers knew when developing and tuning their models what the observed warming had been over most of this period.

Figure 1. Zeke Hausfather’s comparison of global surface temperature warming in CMIP5 climate models with observational records. Simulations based on the intermediate mitigation RCP4.5 scenario of global human influence on ERF through emissions of greenhouse gases, etc. were used to extend the CMIP5 Historical simulations beyond 2005.

It is well-known that climate models have a higher climate sensitivity than observations indicate. Figure 2 compares equilibrium climate sensitivity (ECS) diagnosed in CMIP5 models and in the latest generation, CMIP6, models with the corresponding observational estimate on the same basis in Lewis (2022) of 2.16°C and (likely range 1.75–2.7°C). Only one model has an ECS below the estimate in Lewis (2022), and most models have ECS values exceeding the upper bound of its likely range. CMIP6 models are generally even more sensitive than CMIP5 models, with half of them having ECS values above the top of the 2.5–4°C likely range given in the IPCC’s 2021 Sixth Assessment Report: The Physical Science Basis (AR6 WG1).

Figure 2.  Red bars: equilibrium climate sensitivity in CMIP5 and CMIP6 models per Zelinka et al. (2020) Tables S1 & S2 estimated by the standard method (ordinary least squares regression over years 1–150 of abrupt4xCO2 simulations). Blue line and blue shaded band: best estimate and likely (17%-83% probability) range for ECS in Lewis (2022), derived from observational evidence over the ~150 year historical period but adjusted to correspond to that estimated using the aforementioned standard method  for models.

So, how is it possible that Hausfather gets an apparently good match between models and observations in the period 1970-2020? Does it imply that the models correctly represent the effects of changes in “climate forcers”, such as the atmospheric concentration of greenhouse gases and aerosols, on GMST, and accordingly that their climate sensitivities are correct?

The key question is this. Matching by CMIP5 climate models, in aggregate, with observed GMST changes would only be evidence that models correctly represent the effects of changes in “climate forcers”, such as the atmospheric concentration of greenhouse gases and aerosols, on GMST if resulting changes in their combined strength in models matched best estimates of the actual changes in those forcers. The standard measure of strength of changes in climate forcers, in terms of their effect on GMST, is their “effective radiative forcing” (ERF), which measures the effect on global radiative flux at the top of the Earth’s atmosphere once it and the land surface have adjusted to the changes in climate forcers (see IPCC AR6 WG1 Chapter 7, section 7.3)

It is therefore important to compare changes in total ERF as diagnosed in CMIP5 models during their Historical and RCP4.5 scenario simulations over 1970–2020 with the current best estimates of their actual changes, which I will take to be those per IPCC AR6 WG1 Annex III, extended from 2019 to 2020 using the almost identical Climate Indicator Project ERF time series.

Historical and RCP4.5 ERF (referred to as “adjusted forcing”) in CMIP5 models was diagnosed in Forster at al. (2013), for the 20 models with the necessary data. I take the mean ERF for that ensemble of models[1] as representing the ERF in the CMIP5 models used in Figure 1.

Figure 3 compares the foregoing estimates of mean ERF in CMIP5 models with the best estimates given in IPCC AR6. Between the early 1980s and the late 2000s CMIP5 and AR6 ERF estimates agreed quite closely, but they diverged both before and (particularly) after that period. The main reason for their divergence since 2007 appears to be that aerosol ERF, which is negative, is now estimated to have become much smaller over that period than was projected under the RCP4.5 scenario. Updated estimates of aerosol ERF also appears likely to account for about half of their lesser divergence prior to 1983, with the remainder mainly attributable to differences in ERF changes for land use and various other forcing agents.

Figure 3. Effective radiative forcing (ERF) over 1970–2020 as estimated in CMIP5 models (mean across 19 models) and the best estimate given in the IPCC Sixth Assessment Scientific Report (AR6 WG1). The ERF values are relative to their 1860–79 means.

The IPCC AR6 best estimate of the actual  ERF change between 1970 and 2020 is 2.53 Wm−2. The linear trend change over 1970–2020 given by ordinary least squares regression is 2.66 Wm−2, while the change between the means of the first and last decades in the period, scaled to the full 50 year period, is 2.59 Wm−2.

By comparison, the mean ERF change for CMIP5 models between 1970 and 2020 is 1.67 Wm−2. The linear trend change over 1970–2020 is 1.92 Wm−2, and the scaled change between the first to last decades’ means is 1.76 Wm−2.

It is evident that the AR6 estimate of the actual 1970–2020 ERF change is far greater than that in CMIP5 models. Based on the single years 1970 and 2020, the AR6-to-CMIP5 model ERF change ratio is 1.51. Based on linear trends that ratio is 1.39, while based on first and last decades’ means it is 1.46. The last of these measures is arguably the most reliable, since single year ERF estimates may be somewhat unrepresentative, and due to intermittent volcanism the ERF has large deviations from a linear relationship to time. As there is some uncertainty I will take the ratio as being in the range 1.4 to 1.5.

So, CMIP5 models matched the observed 1970–2020 warming trend, but the estimated actual change in ERF was 1.4 to 1.5 times greater than that in CMIP5 models. On the assumption that both the CMIP5 model ERF estimates and the IPCC AR6 best estimates of ERFs are accurate, it follows that:

  • CMIP5 models are on average 1.4 to 1.5 times as sensitive as the real climate system was to greenhouse gas and other forcings over 1970–2020[2]; and
  • CMIP5 models would have over-warmed by 40–50% if their ERF change over that period had  been in line with reality.

It seems clear that the ERF change in CMIP5 models over 1970–2020 was substantially less than the IPCC AR6 best estimate, and that CMIP5 models substantially overestimated the sensitivity of the climate system during that period to changes in ERF. Moreover, the divergence is increasing: the ratio of AR6 to CMIP5 model ERF changes is slightly higher if the comparison is extended to 2022.

In conclusion, Maarten Keulemans’ claim that “a simple check of the models versus observed warming demonstrates that “climate models published since 1973 have generally been quite skillful predicting future warming” is false.

Contrary to the impression given by Zeke Hausfather’s rather misleading graph, CMIP5 models have not been at all skillful in predicting future warming; they have matched the illustrated 1970–2020 observed warming (which was past rather than future warming until the late 2000s, when CMIP5 models were still being tuned) due to their over-sensitivity being cancelled out by their use of ERF that increased much less than the IPCC’s latest best estimates of the actual ERF increase.

Nic Lewis               5 September 2023

[1] ex FGOALS-s2, the Historical and RCP simulations of which were subsequently withdrawn from the CMIP5 archive.

[2] There are some caveats to the conclusion that CMIP5 models were oversensitive by a factor of 1.4 to 1.5 times:

  • the ensemble of CMIP5 models used in Forster et al. (2013) might not have been a representative subset of the entire set of CMIP5 model. However, there appears to be little or no evidence suggesting that is the case;
  • despite their careful compilation, the AR6 best estimates of the evolution of ERF might be inaccurate;
  • the CMIP5 model forcings derived by Forster et al. (2013) might be inaccurate. There are reasons to suspect that their method might produce ERF estimates that are up to about 10% lower than the methods used for IPCC AR6. However, Forster et al. present some evidence in favour of the accuracy of their method. Moreover, the agreement in Figure 2 between the CMIP5 and AR6 ERF time series between 1983 and 2007 (with divergences before and after then largely attributed to differences in particular forcing agents) is further evidence suggesting that the Forster et al. (2013) CMIP5 ERF estimates are fairly accurate; and
  • due to the heat capacity of the ocean mixed layer, GMST is more closely related to average ERF exponentially-decayed over a few years rather than to ERF in the same year. Using exponentially-decayed ERFs would somewhat reduce the 1.4 low end estimate given above for the ratio of AR6 to CMIP5 model ERF 1970–2020 increase estimates, perhaps by ~10%.

73 responses to “Do CMIP5 models skillfully match actual warming?

  1. I actually reckon that if the ocean modes had not shifted warmer from 1995 in response to weaker solar wind states, via the Northern Annular Mode, there would have been no global warming and probably cooling from 1995.

  2. So, improving on Michael Mann’s hockey stick graph showing a 5° increase in average temperatures since the year 1000?

  3. Single parameters for the entire globe. Land observations with urban influences of various magnitudes. Hidden internal manipulations in extremely complex models. Natural influences, both known and unknown, with unclear impacts,
    Strikes me as pretty mushy models to justify spending trillions of dollars, unless the money is actually used for payoffs of various kinds. In that case, just politics as usual. Nothing to see here, just move along and leave your wallet.

  4. Why are there no error bars presented for any of the effective radiative forcing (ERF) estimates?

    • The comparison being made is with Hausfather’s chart (Fig.1), which doesn’t show any error bars. In each case, the point is to compare a multimodel mean estimates with best estimates from non-model sources. In any case, the source for CMIP5 forcings doesn’t provide uncertainty ranges.

  5. Wei Zhang (MN)

    When did the CMIP models start on out-of-sample data? I’m guessing about 2006. Anybody can model the past with backtests. Just turn the nobs until it fits. It is going forward that counts.

    • Yes. If they were serious they would show their predictions for years (decades) into the future and serious evidence that they made them years (decades) ago.

      It is crazy that anybody gets more than zero points out of ten for predicting data they already have.

      • The climate models have NEVER made ANY correct predictions regarding global climate. Therefore, any claim that they are “accurate” in ANY sense of the word is nothing less than pure fraud. It’s amazing that there are so many fraudsters and then so many “useful idiots” who go along with the ruse.

        If anyone here, believes that the climate models have EVER made a SINGLE correct FUTURE prediction (the only kinds of predictions), please show it to me/us and I will show you why it and you are incorrect.

        If nobody here can show a SINGLE example of ANY climate model that has EVER made a CORRECT FUTURE PREDICTION, then can we please stop playing pretend and posting about this utter nonsense, and begin to tell the truth about climate.

  6. Clearly getting a rough match to the training data when you have something like 22 poorly constrained “parameters” to fiddle with does not indicate any “skill” in understanding climate, simply skill in regression fillting your parameters.

    It is significant that Hausfather chooses not to back before the tuning period ( hindcast ) back to 1900 and allow us to see how well the models reproduce the early 20th c. warming. We all know why he chooses not to do that obvious test of model skill.

    Another valuable test would be to look at TLS. We only have 1978 onwards but it would be worth examining. I’ve been told they get it roughly right , probably like the surface, but I’ll bet they do not reproduce the clear cooling steps after Mt P and El Chicon but instead produce a steady downward trend.

    I’ve not been able to find the corresponding CMIP5 field which corresponds to TLS. Any one know what that is?

  7. ” GMST is more closely related to average ERF exponentially-decayed over a few years”

    Yes, I used that concept to match volcanic forcing after Mt Pinatubo to the ERBE top of atmosphere energy budget:

  8. Valid physics is supported by evidence in Earth and other planets.
    None of you can prove it wrong: nobody has in 10 years now:

  9. Joe - the non climate scientist

    How hard has it been to project the future warming?

    The planet has been on a general warming trend since the late 1800’s . The models since the 1980’s have all predicted warming at roughly a similar rate.

    The overall trend has been reasonably consistent, thus the model predictions dont reflect any significant scientific insight.

    My apologies if I over simplifying the critique of the models.

    • I have demonstrated on WUWT that a simple linear extrapolation of historical temperatures, since the 1960s, is superior to Hansen’s infamous 1988 projections.

      • Joe - the non climate scientist

        Concur – That’s my point – whether it’s a super fined tuned climate model or a simple linear extrapolation of historical trend sense the 1960’s or 1880, they are all pretty much the same.

        Give me a climate model that replicates the cooling of the little ice age and the shift to warming – then I will be impressed!

      • It’s worse than that. climate models, despite their massive complexity are tuned to be basically CO2 plus noise. They “know” that the main warming is coming from CO2 , so that is the starting point. After that they add volcanics so that they get a dip in 1990-1994 and tune that so that it looks like they have the detail nailed but that is all they have. Any other discrepancies are just explained as “internal fluctuation” and given lots of silly names all ending in “oscillation” to pretend they are long term neutral so it does not matter that they cannot reproduce it.

    • “The models since the 1980’s have all predicted warming at roughly a similar rate. ”

      Wrong. There are a wide range of sensitivities to CO2 and some models produce much more warming. No work has been done to eliminate the crazies because the narrative always needs the : maybe as much as … line with stupid scary numbers.

    • Joe,

      That’s an important and often ignored point. Predicting that a long trend will continue isn’t a meaningful prediction.

      “Confirmations should count only if they are the result of risky predictions; that is to say, if, unenlightened by the theory in question, we should have expected an event which was incompatible with the theory — an event which would have refuted the theory.”
      — Karl Popper in “Conjectures and Refutations: The Growth of Scientific Knowledge” (1963).

  10. The models are all wrong for one reason and one reason only: they are based on false physics in which stupid climatologists with no correct understanding of the relevant physics, assume they can add to solar flux about twice as much radiation from the IR-active gases in the atmosphere and use the sum in Stefan-Boltzmann calculations to quantify surface temperatures. They can’t. That law only ever works for a single source of radiation.

    But that is what their energy diagrams clearly imply they do. Such diagrams even show more energy out of the base of the atmosphere than enters at the top. They also show molecules of “greenhouse” gases supposedly knowing they have to radiate more downwards and upwards. Clever molecules that can sense where the surface is I suppose! (/sarc)

    The surface temperature is warmer than that at the radiating altitude because of the non-radiative heat process explained in the above-linked paper and nowhere else in world literature. You can’t prove it wrong.

  11. >> Do CMIP5 models..
    well you are about 5 years late. Ever since the CMIP6 models have shown that the better cloud parametrization yields an about 25% different CO2 sensitivity, CMIP5 and all older models are just uncertain junk.
    Or do you claim that clouds are not important for the climate?

    • The article is in response to a recent claim by a journalist about a model-observation warming comparison, running up to 2020, that involves CMIP5 models.
      The CMIP6 models may have better cloud parameterization, but that doesn’t necessarily mean that they generally better represent the response of the whole climate system to changes in forcings. Indeed, the standard deviation of the ECS and, particularly, the effective climate sensitivity applicable to periods of 50 to 100 years is substantially greater in CMIP6 models than in CMIP5 models.

      • >> The CMIP6 models may have better cloud
        >> parameterization, but that doesn’t necessarily
        >> mean that they generally better

        Well, that might be debatable, but this also was not my point!
        I tried to say:
        Clouds are important for climate predictions, CMIP5 models do not represent clouds correctly (and we know that without doubt from CMIP6 model results), they therefore cannot be used for trustworthy climate predictions.

  12. Geoff Sherrington

    Accuracy of CMIP modelling can be seen by comparisons. In 2007, David Douglass et al published in Int. J. Climatol. (2007), “A comparison of tropical temperature trends with model prediction”.
    Table 2 from it:
    From the Abstract,
    “We examine tropospheric temperature trends of 67 runs from 22 ‘Climate of the 20th Century’ model simulations and try to reconcile them with the best available updated observations (in the tropics during the satellite era). Model results and observed temperature trends are in disagreement in most of the tropical troposphere, being separated by more than twice the uncertainty of the model mean. In layers near 5 km, the modelled trend is 100 to 300% higher than observed, and, above 8 km, modelled and observed trends have opposite signs.”
    Here is the modelled trend of temperatures at various altitudes above the Earth surface, expressed in milli⁰C per decade.
    Pressure Results #15 Average of all 67
    hPa milli⁰C/decade milli⁰C/decade
    1000 163 156
    925 213 198
    850 174 166
    700 181 177
    600 199 191
    500 204 203
    400 226 227
    300 271 272
    250 307 314
    200 299 320

    Model # 15 is Australia’s CSIRO MK3.0.

    Attention is drawn to the exceptionally good model results at mid-altitude, 300 mto 600 hPa. They agree with the average of the other runs/models to one millidegree C per decade. Are these models truly accurate estimates to one thousandth of a degree per decade?
    CSIRO might have set a performance standard that is hard to better. Can they explain how it was done?
    Geoff S

  13. Pingback: Do CMIP5 models skillfully match actual warming? • Watts Up With That?

  14. Europe is whining about Biden’s IRA. They can’t get enough subsidies. For “green” power, no subsidies means no power.

    Sept 6 (Reuters) – A fleet of U.S. offshore wind projects central to President Joe Biden’s climate change agenda may not move forward unless his administration eases requirements for subsidies in the year-old Inflation Reduction Act, according to project developers.

    Norway’s Equinor, France’s Engie (ENGIE.PA), Portugal’s EDP Renewables (EDPR.LS), and trade groups representing other developers pursuing U.S. offshore wind projects told Reuters they are pressing officials to rewrite the requirements, and warning of lost jobs and investments otherwise.
    Advertisement · Scroll to continue

    “The components needed for our projects to progress simply do not exist in the U.S. at this time, and we see no signs that the supply chain will be ready in time to meet our procurement schedule,” said David Marks, a spokesperson for the U.S. renewables division of Equinor (EQNR.OL).

    Denmarks’ Orsted (ORSTED.CO), a top offshore wind developer, warned last week that barriers to securing U.S. subsidies under the IRA, combined with soaring interest rates and supply chain delays, could lead to $2.3 billion in impairments for three projects, sending its stock plummeting.

  15. Pingback: Do CMIP5 models skillfully match actual warming? • Watts Up With That? - Lead Right News

  16. >…it is perhaps unsurprising given that modelers knew when developing and tuning their models what the observed warming had been over most of this period.

    Snide insinuations under a thin veil of plausible deniability?

    How uncharacteristic!

    • When a model is being built it is easy to make it appear accurate for past events. If a model is released in 2004 it makes no sense to claim it was accurate in it predictions from1980 to 2000. Only evaluate its accuracy from 2005 and on.

      Doesn’t that make sense???

      • Dependent data vs independent data. When verification data takes too long to emerge such as climate applications, you can at least jackknife the data honestly. That is where the data is sliced in …example… 4 parts. Then model is tuned to three parts and tested on the 4th which was not included in the development. You can repeat tuning the model on the other parts and end up with four different “independent” verifications. But there are ways to “cheat” that must be avoided to ensure success going forward.

    • Many things climate scientists have said have turned out to be false or at best arguable. Only now have they started to admit that climate models have very serious problems after 30 years of stonewalling and misrepresenting them. But based on CMIP6 it appears things are getting worse and not better. Nic’s suggestion is quite justified by this track record.

      What is really shameful is that most of the modelers knew all along how large the numerical errors and subgrid model errors were but mostly remained silent. Even Stevens and Palmer cover their recent admissions with the required narrative caveat that of course none of this means the climate crisis isn’t real. Talk about “insinuation under a thin veil of plausible deniability.”

      And you Joshua have repeatedly taken them out of context to confuse and obfuscate their core scientific findings.

    • Joshua of the snide remark claim.
      Please reflect on my comment above about exquisite accuracy claimed by one of the modelling groups. Then, contemplate the possibility that the exquisite result was derived from non-customary methodology of which some might not approve. “Snide” is not the right word if methods that disrespect the Scientific Method were used by the modellers. I do not know if it was.
      Consider also that this Nic Lewis analysis becomes irrelevant if ithe questioned methodology is widespread in the modelling community. Geoff S

  17. Richard Greene

    They are climate computer games, not climate models.

    The knowledge to construct a real climate model does not exist.

    The ability to predict the future climate may never exist, even with that great knowledge of every climate change variable.

    The 1970s models when using RCP 4.5 appeared to be fairly accurate until 2016. They averaged a global warming rise rate about half the rise rate when using RCP 8.5 over 400 years with the same “models”.

    So what?

    They are not real models of the climate of our planet, so an appearance of accuracy is just a lucky guess.

  18. Pingback: Do CMIP5 models skillfully match actual warming? - Climate-

  19. It’s wonderful how climate science operates in a do-it-yourself mode, independent about processes and protocols proven in other fields.

    Look at chapter 4 in AR6 WG1, discussing evaluation of models. Little mention of the large literature in other fields about validation of quantitative models. No mention even of simple methods, like out-of-sample testing or even a review by a multidisciplinary team of independent experts. It’s a weak basis for models to save the world, justifying drastic changes to the world’s eco y and societies.

    For out-of-sample testing, previous generations of CMIP models could be run with actual data (not scenarios) from after their creation through the present – and the predictions (not projections) compared with observations.

    One example of protocols that can provide a starting point is work of the Verification and Validation Committee of the American Society of Mechanical Engineers (ASME). See their ‘Guide for Verification and Validation in Computational Solid Mechanics’, their ‘Standard for Verification and Validation in Computational Fluid Dynamics and Heat Transfer,’ and ‘An Illustration of the Concepts of Verification and Validation in Computational Solid Mechanics.’

    • Insightful comments on model verification. Obviously, it’s easier to just make bogus claims of model accuracy than face a valid verification.

      If Fidelity or Vanguard marketed their investment funds by using backtests and changing the observations after-the-fact, they would be sued and fined and banned and/or jailed.

      But that is the MO for climate model verification

  20. The so-called climate models are completely inaccurate. There has never been a model yet that has generated even a single moderately accurate graph going forward. The millennials involved in their development think that by tweaking 20 arbitrary “tuning” parameters they have an accurate model because they can make it match a short period of time in the PAST. How absurd, ridiculous, and utterly foolish. You can match essentially any limited curve with 20 parameters. How stupid are these people? Forget noting that the models use an ECS that is at least 3 times higher than what we observe to be true to generate these hindcasts. And don’t mention that they can’t handle the equator. Or that the notion of model mean is in itself beyon absurdity. Do these people no nothing of basic math and statistics. Let’s see we have a bunch of models which don’t match anything and are full of OBVIOUS and LARGE errors … so let’s average them together and that will make them better. Yeah that’s the ticket! I would wager a bet that the vast majority of the model writing millennial coders could not derive the quadratic formula, even given the Google search page showing them the solution!

  21. Tuning is a key aspect of building climate models. “Tuning consists of adjusting the values of these parameters to bring the solution as a whole into line with aspects of the observed climate.”
    — “The Art and Science of Climate Model Tuning”

    But it makes validation by conventional impossible to rely on. Hence the need for testing vs out-of-sample data, as discussed here. And the use of other validation protocols.

    This should be a major project of climate science institutions. That it is not tells us much.

    • A better name for “The Art and Science of Climate Model Tuning” would be “Curve Fitting 101.” It’s just basic mathematics. Oops, I forgot the climate modelers never took or passed any actual math classes…

  22. Nic, it can be easily shown that the CMIP 5 and 6 models well follow their differing temperature sensitivities in the future period with various forcing scenarios such that, for example, the temperature trends for the 2014-2100 period for RCP 4.5 forced model trends have a very high correlation with the same models RCP 8.5 forced model trends. These high correlations extend to all CMIP 5 and 6 scenario combinations. Therefore, the question arises as to why many of the high sensitivity models no longer show those same correlations with temperature trends in the historical period. This is a question that has not received much attention from climate scientists until recently, and now including your recent post here.

    In the historical period it is interesting to look at Tdiff, which is the difference between a calculated temperature trend for that period based on model sensitivity and the actual model trend for that period. The Tdiff varies over a wide range for the CMIP5 and 6 models whereas smaller Tdiff values can be found for models with lower sensitivities and lower historical temperature trends and models with larger sensitivities and larger historical trends. Larger positive Tdiff values are found mainly for models with the highest sensitivities, although some models have negative Tdiff values (where the actual trend is larger than that calculated from its sensitivity.

    The range of differences in Tdiff values and to which models the differences accrue makes the concept of a model ensemble mean rather precarious. The differences indicate that modelers knowing full well that for their models to be taken seriously must attempt to get the historical and observed period reasonably correct. The models with the lowest Tdiff also most closely approach the observed trends for the historical period and have the lowest sensitivities. That result should put extra scrutiny on the high sensitivity models that approach the observed temperature trend.

    Below I have listed some papers that studied the high sensitivity and lower historical temperature trends for CMIP5 and 6 models and some analysis that I have done concerning the matter.

    • Ken, thanks for your comment. The main things that I would expect to affect historical warming trends in models, apart from their estimated ECS, are a) the trend in model ERF over the period involved relative to the model ERF for 4x CO2 (150 year abrupt4xCO2 simulation data is normally used to estimate their ECS); b) model ocean heat uptake (mainly the mixed layer to deeper ocean transfer coefficient); and c) the difference between model effective climate sensitivity on a 30 to 50 year timescale and their estimated ECS.

      • Nic, I agree with your factors that can affect the temperature trends in the historical period and of which there can be many combinations. I am of the mind that those factors should reflect the best observed estimates of them during the historical period and that as a consequence ideally the model temperature trends during the historical period should correlate with the model sensitivity as measured best by the model TCR, the model temperature trends in the future period with given scenarios and to a lesser degree with the model ECS.

        Evaluating model ERF values based on temperature changes and using those ERF values to predict temperaturen changes is too circular for me. An independent estimate of ERF works for me.

  23. Stephen Segrest

    Tired of feeling hopeless about climate change? Take a look at these charts. Free Read from my Washington Post subscription:

  24. It’s difficult to believe some people want to triple wind and solar electricity fast. Texas is an infamous leader in wind energy. Where has that gotten them?

    Texas declared its first power emergency since a deadly winter storm two years ago and came close to rolling blackouts as soaring temperatures roasted the second-largest US state.

    The declaration of a so-called Level 2 emergency late Wednesday came in response to shrinking supplies of available power and meant the Electric Reliability Council of Texas, the state’s grid operator, had to draw on reserves while pushing consumers to curb usage.

    Of course, members of the Church of Climate Doomers would sacrifice their children to reduce CO2 emissions. My children are happy I’m of a different religion.

  25. Here’s the ERCOT energy mix. Texas needs more coal, nuclear, and natural gas power plants.

    • We need more crypto miners too.

      ““The bitcoin miner said on Wednesday that it earned $31.7 million in energy credits last month from Texas power grid operator ERCOT. The company generated the credits by voluntarily curtailing its energy consumption during a record-breaking heatwave.

      The total value of the credits dwarfed the 333 bitcoin the company mined in August, worth about $8.9 million dollars as of the end of the month.”

  26. In my blog post here last year on Computational Fluid Dynamics, I give an explanation of why models can match GMST while getting all the details wrong. It has to do with conservation of energy. Models are tuned to match the top of atmosphere radiative balance. If their ocean heat uptake is roughly right as it seems to be then the GMST can be roughly right assuming that kinetic energy doesn’t change much in the atmosphere.

  27. Thanks, Nic. As always, great job.

  28. I don’t understand all the blather here about how the models “match” when they don’t actually match ANYTHING other than a short portion of a historical graph for which 20 customized “tuning” parameters have specifically been adjusted so as to “match”. And yet they still don’t match. Look at the hugely wide variation around the model mean (a totally meaningless concept in and of itself) of the so-called models. The variation of the models around the model mean is MUCH greater than the variation of the temperature as measured by thermometers, satellites, radiosondes or any other ACTUAL measurement around its mean. In other words, the models are MASSIVELY incorrect at EVERY point on these graphs, including the short segment of historical time that they are tuned for. And yet so many of you here blather on about how they match. What are you even talking about? The fact is the models have NEVER made even one single FUTURE prediction that was even mildly accurate. They don’t match anything.

    • About 15 years ago I first saw an average taken from dozens of various time series graphs of modelled temperatures like we see for CMIP comparisons with measured. (“Gobsmacked” comes to mind.) The separate responses covered a large span that IMO made them useless for claims that models were generating agreement. The averaged result is a dubious metric to compare with measured results.
      Why has there been so little published objection to this averaging?
      Geoff S

    • Thanks for the comments, Jonathan.

      They’re very instructive.

  29. The members of the Church of Climate Doomers have big plans for your money.

    More than 10 gigawatts of offshore wind projects along the US East Coast — the equivalent of roughly 10 nuclear power reactors — are at serious risk as higher costs force developers to re-crunch the numbers for proposals originally modeled years ago, before a runup in interest rates and material costs. Orsted A/S, the Danish wind giant, said this week it’s prepared to walk away from projects unless it gets even more government aid. Other developers are already paying tens of millions in penalties to exit contracts they say no longer make financial sense.

  30. Ireneusz Palmowski

    As water from melting sea ice in the south feeds the Humboldt Current, El Nino will weaken further in November.
    Another hurricane west of Mexico will effectively lower surface temperatures in the Nino 1.2 region.

  31. When will the “climate scientists” understand the basic concept that models are NOT and NEVER HAVE BEEN science? Models are NOT evidence. Models cannot be used as evidence. Models never will be evidence. Evidence is ONLY empirical. Belief systems and dogma are NOT evidence.

    Unfortunately, we now have Marxism/Communism masquerading as “science” in this so-called field of “climate science” which seems to be a field that only math and physics drop-outs go into nowadays.

    If you want to understand what real science is and what models are used for, you need to (1) understand math, (2) understand physics, (3) understand computer science, and then go study with a few Nobel Prize winning — especially retired or emeritus ones who are no longer subject to the tyranny and oppression of all the Marxist / Communist outfits masquerading as educational institutions — scientists. The real ones, by the way, NEVER call themselves “climate scientists”, because NO SUCH SCIENCE exists.

    Saying that “climate science” is A science is kind of like saying that athletics is A sport, only much more dramatically different.

    “Climate science” today is quite literally an enclave of Communist / Marxist propagandists with little math or physics education or understanding attempting to foist its communist dogma on the world by infesting ALL of the institutions and publications with indoctrinated communists and then BLOCKING all of the real scientists who actually understand math and physics.

  32. When will you all get it?

    The energy diagrams and the models all assume that surface temperatures (and increases therein) can be quantified using the Stefan-Boltzmann Law with an input equal to solar radiation (about 168w/m^2) added to back radiation (about 324w/m^2) less non-radiative surface cooling (about 102w^2)”- that is, a net of about 390w/m^2 which, if it were uniform 24/7 all over the globe and from a single source such as a sun delivering nearly 500w/m^2 (rather than only 168w/m^2) would achieve a temperature of about 288K – far less if the flux were variable because of the T^4 in Stefan-Boltzmann calculations.

    The Stefan-Boltzmann Law only ever works for a single source, as is easily demonstrated with a simple, cheap experiment that you could do in your backyard at night with two or more electric bar radiators and a thermometer.

    Because their physics is wrong, the models are wrong.

    Radiation to the surface of Venus, for example, would have to be well over 16,500w/m^2 of which the solar radiation is less than 20w/m^2. So how are you going to get that much radiation out of the less-hot atmosphere supposedly raising the surface temperature which is observed to rise between about 732K and 737K over the course of about 4 months that a location on the equator spends on the sunlit side.

    The main input of thermal energy that is what is raising the Venus surface temperature is not radiation of any kind, but rather the heat process explained in the 2013 paper “Planetary Core and Surface Temperatures.” You cannot correctly refute the physics therein. Nobody has in over 10 years.

  33. Hi Nic. I wonder if you have seen this paper.

    • Hi, Terry, thanks for the link. I had seen that paper, although I haven’t studied it in depth. It does point to significant internal problems remaining in the latest, CMIP6, models.
      For instance, its Fig.1(a) shows that over the 500 year preindustrial control simulation, which should be in equilibrium, time integrated heat flux is small into the ocean (as it should be) but very large into space. In fact, the net top-of-atmosphere energy flux is about 6 zettajoules per year, which is about what that flux has been estimated as averaging over the last half century or so of forced warming.

      • Thanks for the reply! I am interested in better understanding Figure 4e and its implications, especially for the Western US. It seems the models have a serious mass balance problem (even the authors make a big deal of it in the text), so what is the value of their “projections” for changes in precip and evap? I would like to see Figure 4e dissected in more detail. Hoping someone can do this.

  34. While warming is overstated and the dangers of warming are fabricated to a great degree–there are significant climate shifts occurring in front of us for whatever reasons that are displacing populations. AGW is false, but some sort of climate change is haapeningl how do we perform proper attribution?

  35. @Jonathan Cohler…Marxism and communism? That seems extreme. Dropouts? They have the proper degrees, but some of them are lying, some are believing other’s results since no one can test all claims while working on their own research.

    Also all empircal science uses models no matter what methods or type of research study being employed: Experimental, quasi-experimental, observational, RCT, systematic analysis, meta analysis, cohort, cross sectional, case study, case control study, or causal inference based.

    For example Michael Mann has all the proper degrees and lab courses and experience. I think Mann is lying. I think he knows better, but to say or imply he is not smart, not educated, not well trained is false.I think some of other AGW proponents assume the data and analyses from others are of higher quality than they are. Also, while climate scientists and meteorologists/atmospheric physicists may disagree on what models to use/trust or what the uncertainties are, both those who are AGW proponents and opponents use models and statistical analyses; this includes Roy Spencer, Judith Curry, Gavin Schmidt, etc.

    Models are not the begin and end all of evidence, however, and the results must be replicated outside of the original teams that made a finding and the methods and data must be transparent when it is affeciting so many lives.

    • Models are not and never will be evidence. Roy Spencer has NEVER used a model as EVIDENCE of anything. Until you understand the concept of what evidence is, you are not a scientist.

  36. Jonathan Cohler you are simply mistaken. Perhaps you mean he bever used the GCM’s that others used etc, but models are necessary in science. Here are some of Spencer’s publications:



    Asia-Pacific Journal of Atmospheric Sciences Article
    Published: 07 November 2013
    The role of ENSO in global ocean temperature changes during 1955–2011 simulated with a 1D climate model
    Roy W. Spencer & William D. Braswell
    Asia-Pacific Journal of Atmospheric Sciences volume 50, pages229–237 (2014)Cite this article

  37. Question for Nic – Something that confuses me a lot.

    The ability of a parametric model to predict/forecast future is typically tested by a) calibrating the model using data for one period and, b) using the calibrated model predict the future quantities for the next period. In a particular case of a climate model we should calibrate the mode, say, using data for 1990-2000 and see how it works for the next 10-20 years using the calibrated parameters and the known supposedly controllable inputs (greenhouse gases emissions). We can repeat this procedure in a rolling fashion (calibrate using 1995-2005, test for 2006-2016, etc). This is how we do it everywhere in time series analysis, no?

    How do we do it in climate science? Do we calibrate it using data for 1990-2020 and then show that the model fits the data for 1990-2020? What’s the predictive value of such model?

    • This isn’t an area I have much knowledge of. That said, my understanding is that in principle the physical parameterizations of GCMs are calibrated to reflect observed (past) relationships in variables relevant to those parameters, but in practice they are also calibrated by reference to past changes in variables such as global and regional temperature, radiative imbalance and cloud cover.
      So far as I am aware subsequent but relatively short term future performance is not normally used to validate GCMs. I’m not defending this approach, just stating my understanding of it.

      • e g., more like the casting of chicken bones to foretell the future…

      • Understood. Modern “data science” paradigm requires that the “test” dataset is separate from the “calibration” data set – an approach broadly used by quantitative finance models. Even stricter approach involves a “validation” data set where you are allowed to optimize hyperparameters (e.g. something related to the structure of the model, not to the underlying processes).

        If what you say is indeed the case — it surprises me a lot that climate science seems to exist in complete isolation from these proven approaches to timeseries modeling.

  38. @ jim2 | September 11, 2023 at 8:44 am in suspense

  39. Pingback: Attain CMIP5 models skillfully match steady warming? – TOP HACKER™

  40. Jonathan Baxter

    “The main reason for their divergence since 2007 appears to be that aerosol ERF, which is negative, is now estimated to have become much smaller over that period than was projected under the RCP4.5 scenario.”

    Short version:

    Is your model too sensitive? Crank up the aerosol forcing.

    Plus ça change, plus c’est la même chose.

  41. lol reading Zeke’s post you’d think CMIP5 was published in 1970 and has been successfully predicting temperatures for 50 years now

    “wow! what an amazing job of prediction! put them in charge of the global economy immediately!”

    of course Spencer showed long ago that even this bit of legerdemain is a bit of a reach