The DOE Challenge to Sherwood's ECS Estimate

Accounting for Climate Forcings Puts ECS Near 3°C
From ClimateBrink

The so-called "Climate Working Group," hired by the Department of Energy to write what Roger Pielke Jr termed a "red team" response to climate science (my initial response is here) is predictably critical of the central scientific estimate for ECS. The first ECS estimate I know of was calculated by Arrhenius, who concluded that 2xCO2 would cause between 4-6°C warming. This value was revised downward by Gilbert Plass in the 1950s to ~3°C, and since the 1970s this has become the standard estimate. The IPCC currently says the likely range is 2.5-4.0°C, largely as a result of Sherwood et al 2020 (S22),[1] which is still to date the most comprehensive assessment of ECS (Sherwood's likely range was 2.6-3.9°C). There is a growing body of scientific literature arguing that recent observational evidence is more consistent with an ECS closer to 4°C, suggesting that the IPCC may be a bit conservative on the warming response to doubling CO2. But predictably, the Climate Working Group wants to push ECS in the opposite direction. Regarding Sherwood's paper, the DOE report says:

Lewis (2022) raised a number of concerns about this result, including methodological errors, outdated input values, and use of subjective Bayesian priors in the analysis. Lewis’ analysis found that climate sensitivity is estimated to be much lower and better constrained than in the Sherwood et al. analysis – median 2.2°C (1.8–2.7°C in the 17–83 percent likely range, and 1.6–3.2°C in the 5–95 percent very likely range).

Lewis (L22)[2] indeed had several criticisms of Sherwood's paper. Lewis identified three of these:

1. A "coding error" that led to "the standard deviation for ΔCO2PETM to be one tenth of the correct value," which they admit does not affect their main results.
2. A re-evaluation of the "priors" used as "input values" to calculate ECS.
3. A change in method from a "subjective" to an "objective" Bayesian statistical method.

These appear to correlate with what the DOE report summarizes as "methodological errors, outdated input values, and use of subjective Bayesian priors." Since the uncertainty calculation for ΔCO2 during the PETM doesn't affect Sherwood's results, I'm going to leave that alone. The other two categories of criticism I think warrant further evaluation.

Subjective vs Objective Statistical Method

The DOE report criticizes Sherwood et al 2020[1] (S20) for using "subjective Bayesian priors," as if this is somehow problematic. This is echoing L22, which claims, "This study estimates climate sensitivity using an Objective Bayesian method with computed, mathematical priors, since subjective Bayesian methods may produce uncertainty ranges that poorly match confidence intervals." We are given the impression from the DOE report (and to some extent L22) that the "objective" method is inherently superior to the "subjective" method, and Lewis implies this with the title of the paper, "Objectively combining climate sensitivity evidence." One might wonder why anyone would choose a subjective Bayesian method if the objective method is inherently better. 

And yet in a PubPeer discussion about his paper, Lewis admits that it's not quite this simple. There is subjectivity involved in both S20's "subjective" method and L22's "objective" method. Nic Lewis writes,

It is the method of combining the evidence from which climate sensitivity is estimated that is objective, not the choice of evidence used. The statistical method used is indeed objective, in the sense (as the paper states) that it only depends on observed data and model assumptions. As the paper goes on to say, In Objective Bayesianism (as used in Lewis 2022) as well as Subjective Bayesianism (as used in Sherwood et al 2020), however, subjective choices will still be made by the investigator in relation to the data and [statistical/physical] model used.

Gavin Schmidt's discussion of S20 (before L22 was published) provides a pretty good explanation for why S20 chose a "subjective" Bayesian approach.

This is fundamentally a Bayesian approach, and there is inevitably some subjectivity when it comes to assessing the initial uncertainty in the parameters that are being sampled. However, because the subjective priors are explicit, they can be adjusted based on anyone else’s judgement and the consequences worked out. Attempts to avoid subjectivity (so-called ‘objective’ Bayesian approaches) end up with unjustifiable priors (things that no-one would have suggested before the calculation) whose mathematical properties are more important than their realism.

I'm not a statistician, and I don't think it's necessary to argue one method is superior than the other. It seems clear to me that the choice of an "objective" Bayesian method doesn't privilege L22 over S20, nor is there a substantial criticism of S20 for following a more "subjective" approach. No Bayesian methodology forces a low or high ECS result, and far more significant to L22's results are the decisions regarding priors, which Lewis admits involves subjectivity. In effect, L22's argument is not that L22 used an objective Bayesian approach and therefore objectively lowered S20's ECS estimate (as is often implied). The argument is that if you use Lewis' priors with an objective Bayesian approach, you can come up with a result that is lower than S20's result. So we have to evaluate whether Lewis' choices are actually improvements over S20 to evaluate his revision to ECS.

Evaluating Priors

After the approach in S20, Lewis evaluated three types of evidence (process, historical, and paleoclimate) in order to revise and update the S20 estimate for ECS. I'm not going to engage in a wholesale evaluation and revision of L22 input values. Rather, I'm going to argue that, within the choices Lewis made, L22 made systematic revisions that point in the direction of lowering ECS, and several of these are not justified by the best available evidence. I'm not going to identify all of these unjustified revisions. Rather, I'm going identify just one or two in each section and show how correcting these values affect his estimates for ECS.

Evaluations of ECS from Process Evidence

Process evidence requires a quantification of feedbacks and the forcings associated with 2xCO2. You can see the effect of Lewis' revisions in the data table below.

Table 1 from L22

The most notable change for me is the change in cloud feedbacks, which Lewis reduced from 0.45 W/m^2/C to 0.27 W/m^2/C. While cloud feedbacks are extremely difficult to quantify, Ceppi et al 2021[23] was able "constrain global cloud feedback to 0.43 ± 0.35  W/m^2/C." Recent satellite evidence over the last 25 years has also shown a significant reduction in albedo, which may be explained either by cloud feedbacks or reductions in aerosol pollution, but it seems pretty clear here that reducing the total cloud feedback to 0.27 W/m^2/C is unwarranted. I'm going to revise it to 0.43 W/m^2/C. I agree with the revision of forcings for 2xCO2 to 3.93 W/m^2, so if I keep Lewis' Planck response at -3.25 W/m^2/C, that leaves the scaling factor. I've reviewed the paper on this, and I'm not convinced it's necessary after reducing F2×CO2. With Lewis' scaling factor, ECS increases to 2.47°C; without it, ECS becomes 2.87°C.

Evaluations of ECS from Historical Evidence

It would take too much time to go into a lot of detail here reconciling all the factors that lead to L22 cutting S20's estimate by more than half. Lewis wrote that the most significant revisions were to aerosol forcings and the historical pattern effect, but I'm going to look at just aerosols and leave everything else the same.

Table 2 from L22

Several studies in the wake of both the covid pandemic and the implementation of regulations governing aerosol pollution from ships have made Lewis' aerosol revisions pretty much untenable. As best I can tell, the L22 aerosol revision was largely based off of Glassmeier et al. 2021.[19] Several studies that L22 would have had access to (some he even cited) contradict his estimates [20][21][22]. The AR6 estimate for aerosol forcings is -1.3 W/m^2,[12] and most of his other adjustments to S20 were simply updated from AR6 values. I see no reason, given the evidence we now have, to lower aerosol forcings below -1.3 W/m^2. But if I make just that one correction and leave everything else from L22 the same, the corrected ECS estimate from historical data jumps up to 3.1°C. If I update S20's values to those published in AR6, S20's revised estimate for ECS is 4.93°C.

Evaluations of ECS from Paleoclimate Evidence

L22 evaluated much of the data used by S20 to estimate ECS from three sets of paleoclimate data, one from a cooler time period, like the Last Glacial Maximum (LGM) and two from warmer time periods, like the mid-Pliocene Warm Period (mPWP) and the Paleocene-Eocene Thermal Maximum (PETM). In many of the revisions from from historical data, L22 just used updated values from AR6, which to me is completely understandable. For paleoclimate evidence, though, he didn't do this, and after investigating a bit, I think I have figured out why. Here's L22's Table 3 which shows the paleoclimate data with both S20 and L22's revisions.  

Table 3 from L22

Let's evaluate these revisions to see if they are actually updates and corrections to S20, and what impact might come from evaluating L22's revisions.

LGM

The most striking change here is that Lewis lowered S20's ΔT for the LGM from 5°C to 4.5°C, but the IPCC AR6 argued for 6°C (5 - 7°C), a full degree increase above S20's estimate. By 2022, multiple studies had been published that suggest that 5°C is likely too low a value. I've counted six recent and respected studies[4][5][6][7][8][9] that arrived at values near ~6°C, including one published after the publication of L22. There are also at least two respected studies[10][11] that arrived at ~7°C, including one published after the publication of L22. It would seem that S20 was a bit conservative in his use of 5°C. While there are a few studies that arrive at values below 6°C,[28] it's clear that Lewis should have adjusted it upwards to ~6°C.

L22's revision reduces S20's estimate of ECS from the LGM from 2.63°C to 1.97°C, but if we correct L22's ΔT to 6°C and leave everything else the same, the improvement to his ECS estimate is significant. The following formulation comes from  L22. Using L22's revised values in Table 3 for the LGM while changing only the value for ΔT, we get:

S = 6/{(1+0.135)[-0.57 + (-6.67+0.5*0.1*6^2)/3.93]} = 2.92°C

If I chose to use 7°C, ECS would increase to 3.4°C. With just this one correction to one of L22's input values, we not only improved on L22 but showed that S20 likely underestimated the value of ECS from the LGM. This also brings L22's estimate much closer to the ECS values seen in the paleoclimate literature.

mPWP

For the mid-Pliocene Warm Period (mPWP), Lewis again lowered S20's ΔT estimate for the mPWP from 3°C to 2.48°C. But AR6 estimated the mPWP to be ΔT = 2.5°C to 4°C. Lewis cited Haywood et al 2020 for his ΔT value, but that paper actually concluded 3.2°C, consistent with AR6. L22 also cited Tierney et al 2019[14] and McClymont et al 2020,[16] but these papers only evaluated SSTs, which Tierney points out did not warm as much as land.[15] This explains L22's apparently lowballed estimate for ΔT. If we correct L22's estimate for ΔT with AR6 and Haywood ΔT (which were the global estimates), we again need to increase S20's estimate for ΔT from 3°C to 3.2°C. Our next correction to L22, changing only their ΔT estimate, gives us this:

S = 3.2/{[log(1+0.32)/log(2)](1+0.4)(1+0.135)(1+0.67)} = 3.01°C

So with just this one correction to one of L22's input values, we've improved L22's estimate from 2.33°C to 3.01°C, which is much closer to S20's estimate of 3.36°C and Haywood's estimate of 3.6°C (2.6–4.8°C).

PETM

L22 did not update many values from S20 with the exception of the "CO2 ERF relative to with log(concentration)." The ΔT value for both L22 and S20 is 5°C. AR6 estimates that CO2 increased from 900 ppm to 2000 ppm CO2 and GMST increased to 10-25°C above the 1850-1900 mean, with a median estimate of 15°C. The IPCC's ΔT value isn't helpful here because it's relative to a preindustrial baseline, not relative to temperatures before the onset of the PETM. Most estimates for the PETM ΔT fall between 5°C and 8°C, but Tierney et al 2022[17] arrived at 5.6°C with a 95% range of 5.4°C to 5.9°C. So let's stay conservative and update L22's ΔT to 5.6°C. Here we get:

S = [1/(1+0.135)]/{[(1+1.117)*log(1+1.667)/log(2)]*(1+0.4)/5.6 - 0/3.93} = 1.18°C

This is clearly wrong, since increasing ΔT should increase ECS, so it appears that L22 made a mistake in writing this formula. The formula as written is not the formula he used to calculate 1.99°C for ECS during the PETM. We can still estimate how much our improved ΔT would affect L22's calculated ECS if we understand ECS scales with ΔT, so our updated ECS would be 1.99*5.6/5 = 2.23°C, which puts it closer to S20's estimate of 2.38°C. But since I can't trust L22 here, I decided to dig up Sherwood's formulation, which is below:

I was suspicious that L22 made a mistake solving for λ in the above (in S20, ECS = S = -ΔF2xco2/λ). So I solved for S and this is what I found (leaving out the β term, since it's value is 0 in both L22 and S20):

S = -ΔT*ln(2)/[(ln(ΔCO2)(1+fch4)(1+ζ)]

This formula doesn't include Lewis' fco2nonLog term, but with this formula, I can reproduce S20's 2.38°C, if I use the data from S20. S20's value for ΔCO2 is 2400/900 = 2.667. Lewis says ΔCO2 for S20 is 1.667 and then writes his formula with a ln(1+ΔCO2) term. If I use Lewis' values in Table 3, I calculate his ECS to be S = 2.22°C without Lewis' extra term. If I add his extra term, I get this formula

S = -ΔT*ln(2)/[(1+fco2nonLog)(ln(ΔCO2)(1+fch4)(1+ζ)]

Here I get S = 1.05°C, and this allowed me to identify Lewis' mistake. Lewis' "1+fco2nonLog" should just be "fco2nonLog." If I remove that mistake, I can reproduce his S=1.99°C with ΔT=5°C and then calculate S=2.23°C with ΔT=5.6°C. According to Table 3, this extra term was omitted in S20, but he included it because of "Meinshausen et al. (2020); supplemental material S5.3.4." The supplemental material deals with the fact that the logarithmic assumption begins to break down above ~2000 ppm, so a "change from 900 to 2400 ppm" is "11.7% higher than that based on a logarithmic relationship.[27]. I haven't evaluated whether he's done this correctly, but either way the correction is rendered unnecessary if CO2 didn't increase significantly above 2000 ppm, and more recent evidence suggests that's actually case.

If I update S20's ECS calculation with Tierney's ΔT = 5.6°C, I get an ECS of 2.67°C. However, Tierney's analysis found that ΔCO2 for the PETM was smaller than that used in S20. Her value for ΔCO2 was 2020/1120 = 1.8, which is likely a much better estimate of the actual ΔCO2 for the PETM, given the literature I've seen. Since CO2 likely didn't increase significantly above 2000 ppm, I think we can ignore Meinshausen's correction factor. If we use Tierney's values for ΔT and ΔCO2, using all the other values from S20, we get an ECS of 4.16°C. This result is much lower than her calculated value of 6.5°C, so this is a conservative update to S20 and L22.

Considering all three paleoclimate estimates, our improvements increased L22's values for ECS, and 2 of the 3 updates also increased S20's values for ECS. Our summary results for paleo ECS estimates are:

• LGM: 2.92°C (from L22's 1.97°C)
• mPWP: 3.01°C (from L22's 2.33°C)
• PETM: 4.16°C (from L22's 1.99°C)
• Average: 3.36°C

These revisions are also consistent with recent evidence that IRF (and therefore ERF and ECS) scales with climatic base state,[18] meaning that sensitivity is likely larger when climate is warmer. 

Conclusion

If I take the average ECS from all three types of evidence, we get:

• Process: 2.87°C
• Historical: 3.11°C
• Paleoclimate: 3.36°C
  
• Average: 3.11°C

Obviously this simple average of these three methods doesn't take into consideration the uncertainties associated with each of these methods or the uncertainty of the mean of the three types of evidence. But I think it does show that, with small improvements to the data used in these calculations, Lewis' reductions to the best estimate for ECS essentially disappear.

It should be clear that the main reason why Lewis was able to lower Sherwood's ECS estimate is  because of the subjective choices made by Lewis, and many of these had a significant impact on his ECS estimates. Likewise, the few updates and corrections I've offered show that it's likely these choices were made in order to lower ECS, not simply to update or correct the evidence found in S20. The gold standard for evaluating these matters, though, is not updating Bayesian priors in studies; it's replication. Studies estimating ECS at 3°C or higher are clearly replicable in later analysis. Studies estimating ECS closer to 2.2°C are not. Here are a few examples published after 2020:

1. Ceppi et al 2021[23] quantified cloud feedbacks such that lower ECS estimates are implausible. “Considering changes in just these two factors, we are able to constrain global cloud feedback to 0.43 ± 0.35 W⋅m−2⋅K−1 (90% confidence), implying a robustly amplifying effect of clouds on global warming and only a 0.5% chance of ECS below 2 K.”
2. Myhre et al 2025[24] examined trends in Earth's energy imbalance and found that the models that perform best a simulating both SW and LW trends have an ECS at 4 C or larger. "The trends in net EEl and surface warming trend over the first two decades of this century provide little constraint on climate sensitivity. However, we present robust findings for trends in LW and SW EEI. These trends, and their relationship to climate sensitivity, are more physically based than the net EEI trend. ... All models, given as the 99.999% level of the distribution, with an ECS of 2.93 K or below, are outside the CERES range."
3. Zhu et al 2022[25] showed that while models with ECS near 5°C simulate an LGM that is too cool, models with an ECS near 4°C perform quite well. "PaleoCalibr has a lower ECS (∼4°C) and a 20% weaker aerosol-cloud interaction than CESM2. PaleoCalibr represents a physically more consistent treatment of cloud microphysics than CESM2 and is a valuable tool in climate change studies, especially when a large climate forcing is involved."
4. Brown et al 2022[26] evaluated ECS during the Miocene and came to an estimate of 3.9°C. "Having accounted for non-CO2 greenhouse gasses and slow climate feedbacks, we estimate global mean surface temperature change for a doubling of CO2—equilibrium climate sensitivity—to be 3.9 °C (1.8–6.7 °C at 95% confidence) on the basis of comparison of our record of radiative forcing from CO2 with a record of global mean surface temperature change."

For historical data, the graph at the top of this page shows that accounting for known forcings allows us to make good predictions for the increase in temperature since 1850 when ECS is near 3°C. There is simply no warrant to the DOE climate report's conclusions about ECS. There is no evidence of "methodological errors" that lower S20's conclusions for ECS when corrected. Lewis' revised "input values" were sometimes incorrect, and genuine updating increases the ECS estimates for L22 (and frequently S20). And the use of "subjective Bayesian priors in the analysis" is not a legitimate criticism of S20. An ECS of 3°C (or higher) remains a robust conclusion in climate science. I consider it irresponsible of the DOE report to suggest otherwise.

References:

[1] Sherwood, S. C., Webb, M. J., Annan, J. D., Armour, K. C., Forster, P. M., Hargreaves, J. C., et al. (2020). An assessment of Earth's climate sensitivity using multiple lines of evidence. Reviews of Geophysics, 58, e2019RG000678. https://doi.org/10.1029/2019RG000678

[2] Lewis, N. Objectively combining climate sensitivity evidence. Clim Dyn 60, 3139–3165 (2023). https://doi.org/10.1007/s00382-022-06468-x

[3] Sherwood, S. C. and Forest, C. E.: Opinion: Can uncertainty in climate sensitivity be narrowed further?, Atmos. Chem. Phys., 24, 2679–2686, https://doi.org/10.5194/acp-24-2679-2024, 2024.

[4] Tierney, J.E., Zhu, J., King, J. et al. Glacial cooling and climate sensitivity revisited. Nature 584, 569–573 (2020). https://doi.org/10.1038/s41586-020-2617-x

[5] Tobias Friedrich et al. Nonlinear climate sensitivity and its implications for future greenhouse warming. Sci. Adv.2,e1501923 (2016). DOI:10.1126/sciadv.1501923

[6] Snyder, C. W. Evolution of global temperature over the past two million years. Nature 538,

226–228 (2016).

[7] Bereiter, B., Shackleton, S., Baggenstos, D., Kawamura, K. & Severinghaus, J. Mean global

ocean temperatures during the last glacial transition. Nature 553, 39–44 (2018).

[8] Holden, P. B., Edwards, N., Oliver, K., Lenton, T. & Wilkinson, R. A probabilistic calibration
of climate sensitivity and terrestrial carbon change in GENIE-1. Clim. Dynam. 35, 785–806
(2010).

[9] Vincent T. Cooper et al. ,Last Glacial Maximum pattern effects reduce climate sensitivity estimates.Sci. Adv.10,eadk9461(2024).DOI:10.1126/sciadv.adk9461

[10] Osman, M.B., J.E. Tierney, J. Zhu, R. Tardif, G.J. Hakim, J. King and C.J. Poulson: Globally resolved surface temperatures since the Last Glacial Maximum, Nature 599, 239-244, 2021.

[11] James E Hansen, Makiko Sato, Leon Simons, Larissa S Nazarenko, Isabelle Sangha, Pushker Kharecha, James C Zachos, Karina von Schuckmann, Norman G Loeb, Matthew B Osman, Qinjian Jin, George Tselioudis, Eunbi Jeong, Andrew Lacis, Reto Ruedy, Gary Russell, Junji Cao, Jing Li, Global warming in the pipeline, Oxford Open Climate Change, Volume 3, Issue 1, 2023, kgad008, https://doi.org/10.1093/oxfclm/kgad008

[12] IPCC AR6. https://www.ipcc.ch/report/ar6/wg1/downloads/report/IPCC_AR6_WGI_TS.pdf

[13] Haywood, A. M., Tindall, J. C., Dowsett, H. J., Dolan, A. M., Foley, K. M., Hunter, S. J., Hill, D. J., Chan, W.-L., Abe-Ouchi, A., Stepanek, C., Lohmann, G., Chandan, D., Peltier, W. R., Tan, N., Contoux, C., Ramstein, G., Li, X., Zhang, Z., Guo, C., Nisancioglu, K. H., Zhang, Q., Li, Q., Kamae, Y., Chandler, M. A., Sohl, L. E., Otto-Bliesner, B. L., Feng, R., Brady, E. C., von der Heydt, A. S., Baatsen, M. L. J., and Lunt, D. J.: The Pliocene Model Intercomparison Project Phase 2: large-scale climate features and climate sensitivity, Clim. Past, 16, 2095–2123, https://doi.org/10.5194/cp-16-2095-2020, 2020.

[14] Tierney, J. E., Haywood, A. M., Feng, R., Bhattacharya, T., & Otto-Bliesner, B. L. (2019). Pliocene warmth consistent with greenhouse gas forcing. Geophysical Research Letters, 46, 9136–9144. https://doi.org/10.1029/2019GL083802

[15] Lunt, D. J., Haywood, A. M., Schmidt, G. A., Salzmann, U., Valdes, P. J., Dowsett, H. J., & Loptson, C. A. (2012). On the causes of mid-Pliocene warmth and polar amplification. Earth and Planetary Science Letters, 321, 128–138. https://pubs.usgs.gov/publication/70044190

[16] McClymont, E. L., Ford, H. L., Ho, S. L., Tindall, J. C., Haywood, A. M., Alonso-Garcia, M., Bailey, I., Berke, M. A., Littler, K., Patterson, M. O., Petrick, B., Peterse, F., Ravelo, A. C., Risebrobakken, B., De Schepper, S., Swann, G. E. A., Thirumalai, K., Tierney, J. E., van der Weijst, C., White, S., Abe-Ouchi, A., Baatsen, M. L. J., Brady, E. C., Chan, W.-L., Chandan, D., Feng, R., Guo, C., von der Heydt, A. S., Hunter, S., Li, X., Lohmann, G., Nisancioglu, K. H., Otto-Bliesner, B. L., Peltier, W. R., Stepanek, C., and Zhang, Z.: Lessons from a high-CO2 world: an ocean view from  ∼ 3 million years ago, Clim. Past, 16, 1599–1615, https://doi.org/10.5194/cp-16-1599-2020, 2020.

[17] J.E. Tierney, J. Zhu, M. Li, A. Ridgwell, G.J. Hakim, C.J. Poulsen, R.D.M. Whiteford, J.W.B. Rae, L.R. Kump, (2022) Spatial patterns of climate change across the Paleocene–Eocene Thermal Maximum, Proc. Natl. Acad. Sci. U.S.A. 119 (42) e2205326119,
https://doi.org/10.1073/pnas.2205326119.

[18] Haozhe He et al. ,State dependence of CO2 forcing and its implications for climate sensitivity.Science382,1051-1056(2023).DOI:10.1126/science.abq6872

[19] Franziska Glassmeier et al. ,Aerosol-cloud-climate cooling overestimated by ship-track data.Science371,485-489(2021).DOI:10.1126/science.abd3980

[20] Zhaoyi Shen et al. ,Using the fast impact of anthropogenic aerosols on regional land temperature to constrain aerosol forcing.Sci. Adv.6,eabb5297(2020).DOI:10.1126/sciadv.abb5297 https://doi.org/10.1126/sciadv.abb5297

[21] Zelinka, M. D., Klein, S. A., Qin, Y., & Myers, T. A. (2022). Evaluating climate models’ cloud feedbacks against expert judgment. Journal of Geophysical Research: Atmospheres, 127, e2021JD035198. https://doi.org/10.1029/2021JD035198

[22] Gettelman, A., Lamboll, R., Bardeen, C. G., Forster, P. M., & Watson-Parris, D. (2021). Climate impacts of COVID-19 induced emission changes. Geophysical Research Letters, 48, e2020GL091805. https://doi.org/10.1029/2020GL091805

[23] Paulo Ceppi, Peer Nowack. Observational evidence that cloud feedback amplifies global warming. Proceedings of the National Academy of Sciences Jul 2021, 118 (30) e2026290118; DOI: 10.1073/pnas.2026290118 https://www.pnas.org/content/118/30/e2026290118

[24] Gunnar Myhre et al. Observed trend in Earth energy imbalance may provide a constraint for low climate sensitivity models.Science388,1210-1213(2025).DOI:10.1126/science.adt0647

[25] Zhu, J., Otto-Bliesner, B. L., Brady, E. C., Gettelman, A., Bacmeister, J. T., Neale, R. B., et al. (2022). LGM paleoclimate constraints inform cloud parameterizations and equilibrium climate sensitivity in CESM2. Journal of Advances in Modeling Earth Systems, 14, e2021MS002776. https://doi.org/10.1029/2021MS002776

[26] Brown, R. M., Chalk, T. B., Crocker, A. J., Wilson, P. A., and Foster, G. L.: Late Miocene cooling coupled to carbon dioxide with Pleistocene-like climate sensitivity, Nat. Geosci., 15, 664+, https://doi.org/10.1038/s41561-022-00982-7, 2022.

[27] Meinshausen M, Nicholls ZRJ, Lewis J, Gidden MJ, Vogel E, Freund M et al (2020) The shared socioeconomic pathway (SSP) greenhouse gas concentrations and their extensions to 2500. Geosci Model Dev 13:3571–3605. https://doi.org/10.5194/gmd-13-3571-2020

[28] Annan, J. D., Hargreaves, J. C., and Mauritsen, T.: A new global surface temperature reconstruction for the Last Glacial Maximum, Clim. Past, 18, 1883–1896, https://doi.org/10.5194/cp-18-1883-2022, 2022.