The Holocene conundrum

I’m trying, and failing, to work from home while looking after the kids (the ice cream van has just arrived outside). I thought I might just discuss a paper that I became aware of yesterday (forget how, but it might have been a tweet from Oliver Bothe). The paper is The Holocene temperature conundrum by Liu et al. and which is essentially a comparison between Holocene temperature reconstructions and climate models.

The basic issue seems to be that the temperature reconstructions (Shakun et al. 2012; Marcott et al. 2013) suggest that after the Holocene maximum (8000 – 9000 years ago) we cooled by about 0.5oC. Climate models, however, suggest that we should have gradually warmed (by about 0.5oC to 1oC) over the last 11000 years. The basic result is shown in the figure below. The blues lines are the two reconstructions (Shakun and Marcott), the yellow and black are from two different ways of stacking the climate models runs, and the red is from the climate models, but biased in a way that tries to compensate for potential biases in the sea surface temperature reconstructions.

Comparions of Holocene temperature reconstructions and climate models (Liu et al. 2014).

Comparions of Holocene temperature reconstructions and climate models (Liu et al. 2014).

In the paper they try to compensate for various biases that may be (I assume) present in the reconstructions and do (as the red line in the above figure shows) get a better comparison between the models and the reconstructions. They show more detailed tests in the Supplementary Information. However, they seem to argue that although this may help to improve the global comparison, it doesn’t help to resolve inconsistencies are individual sites.

The other thing the paper discusses is the sensitivity of climate models to different forcings. The forcings that are relevant here are those due to greenhouse gases (GHGs), ice sheet changes, variations in insolation, and meltwater flux. What they ignore is variations in volcanic activity or solar variability. So, they argue that it is possible that the difference between the reconstructions and climate models is to do with the sensitivity of the climate models to one or more of these forcings. The paper says

Whatever the biases, the model biases have to exhibit a common warming bias across all of the current models with a total magnitude of at least ∼1 °C, such that removal of this model bias can generate a global cooling of ∼1 °C, which overcomes the 0.5 °C warming by GHGs and ice sheets to leave a net cooling of 0.5 °C as in the M13 reconstruction.

So, the one thing that struck me was that this difference can’t really be an indication of the models being over-sensitive to greenhouse gas (GHG) or ice sheet forcings. The CO2 concentration increases by about 20ppm over the Holocene, and the ice sheets retreat. Both of these should cause warming. Even if the climate models are over-sensitive, they can’t be so over-sensitive as to turn a warming trend into a cooling trend. That’s physically implausible. It doesn’t mean that climate models aren’t over-sensitive to GHG and/or ice sheet forcing; it just means that this discrepancy is too big to say either way.

Solar insolation, on the other hand, remains reasonably constant over the Holocene, but the distribution changes. What struck me about this is that the Milankovitch cycles are thought to be associated with orbital variations in which the total insolation doesn’t change much, but the distribution does. So, it is possible – I guess – that climate models are insufficiently sensitive to how the solar flux is distributed across the planet. This seems to be what the paper concludes by saying

The biases in current models, if they exist, are more likely to be related to their sensitivity to the orbital forcing and additional feedbacks in climate models,

where by “additional feedbacks” it means feedbacks not currently included in most climate models and which may be triggered by variations in solar insolation (I think).

Of course, the other possibility is that there are biases in the reconstructions that amplify – or produce – a cooling trend. I saw a few tweets, with links that I didn’t follow, today that seemed to suggest that there are disagreements about details of these reconstructions.

To be honest, one reason I thought I might write about this was that I was a little concerned that some might use this paper to argue that it is further evidence that climate models are too sensitive, but I think it did a fairly good job of covering that. I also think that comparisons between changes in our past climate and climate models is an interesting topic. I did, however, get somewhat distracted in the middle of writing it, then made dinner, and almost decided to bin it. So, I don’t know if I’ve given this as much thought as I should have, but I thought I’d post it anyway. Some may find it interesting and maybe others will have more insight into this than I currently have.

This entry was posted in Climate change, Climate sensitivity, IPCC, Science and tagged , , , , , , . Bookmark the permalink.

32 Responses to The Holocene conundrum

  1. BBD says:

    Yes, it looks as though the models are missing or insensitive to orbital (Milankovitch) forcing which is thought to have shaped the Holocene climate as reconstructed by Marcott et al.

  2. Rattus Norvegicus says:

    The other possibility of course is that forcing estimates are incorrect.

  3. Rattus,
    Which ones? I think the CO2 increase over the pre-industrial Holocene is reasonably well accepted (about 20 ppm). The ice sheets retreated, so that should lead to warming. I guess what the models didn’t include was any Solar variability and volcanoes, which – I think – they conclude can’t explain the effect. They also conclude the the Meltwater Flux can’t either (I presume this simply means cold water from the retreating ice sheets, so this would make sense). So, the only one that seems to make sense – to me at least – is that the model sensitivity to variations in insolation (in terms of how it is distributed across the globe) and the resuting feedbacks is what is missing or wrong. That’s assuming the reconstructions themselves don’t have issues.

  4. John Mashey says:

    I’m in middle of chip conference, so haven’t looked closely, but which interval does the 20ppm difference cover? See Fig 2B of Ruddiman, et al(2011). Of course, the depths of previous ice age was ~200ppm.

  5. It’s meant to be from 11kyr, or so, till just before pre-industrial times. If I understand Ruddiman (Figure 6) he’s arguing that it’s mostly anthropogenic and started around 6kyr ago. Interesting. I’m not sure what CO2 trend was used in the climate models in the paper.

  6. John Mashey says:

    Yes, he’s arguing that from a peak around 10Kya of ~265, without humans, Earth would have been expected to have gotten below 250ppm pre-Industrial, but humans added ~30ppm over what would have been expected, all +/-5ppm, I think. Of course, similar effect for CH4,which contributes as well, i.e., top half of Fig 6.

  7. BBD says:

    I notice that Liu14 assumes retreating ice sheets post-HTM. I’m not clear that ice sheets did continue to retreat post-HTM, or at least not to the extent that albedo reduction would be a significant feedback. Sea level reconstructions suggest only very slight melt over the last ~5ka.

  8. Reading the paper a strong impression is that the observed cooling over Holocene might be mostly or totally spurious and due to seasonal biases in the interpretation of the data. That could turn the clear cooling to an essentially constant temperature over most of Holocene or even to a weak warming trend. Such a behavior would result from subtracting the difference between the biased red line and the unbiased black model results from the empirical data in Figure 1. As the proposed bias is in the data, it would be in a sense more correct to adjust the data than to add a bias to the model prediction. (Models are even in that case needed for the estimation of the size of the bias.)

    As the models indicate rather strong warming there would still be some inconsistency, but much less than in the case that the bias in data is this strong.

  9. BBD says:

    Interestingly, Kutzbach et al. (2011) used CCSM3 to compare with-and-without anthropogenic forcings during the Holocene and found that the without-anthro runs showed much greater cooling than suggested by the paleo reconstructions. Ruddiman is a co-author:

    The results support more strongly inferences by Ruddiman concerning indirect effects of ocean solubility/sea-ice/deep ocean ventilation feedbacks that may have contributed to a further increase in late-Holocene atmospheric CO2 beyond that caused by early anthropogenic emissions alone.

  10. anoilman says:

    There are so many things wrong with this that its not more use than a curiosity or entertainment. Models aren’t designed to be used this way. Garbage in, Garbage out.

    There are many many longer term cycles not modeled at all, and small numbers there would add up over longer time frames. Again, Garbage in Garbage out.

    I suppose it kind rankles the modeler’s usual quips of hind casting to know accuracy.

  11. BBD says:


    There are so many things wrong with this that its not more use than a curiosity or entertainment.

    Liu14 or Kutzbach11 or both?

    I agree that current GCMs aren’t designed for and will struggle with hindcasting the entire Holocene.

  12. Pekka,
    That was my impression too.

    Interesting, thanks.

    Quite strong views there 🙂 . I did wonder about whether or not these are quite small effects and so, as you say, ignoring some other small effects could end up appearing to have a significant effect.

  13. DocMartyn says:

    The way that aerosols collapsed,indicated by particulate levels in the Greenland cores, just before the warming is a bit of a clue as to what allowed the Earth to warm. The way Greenland ice-core Nitrate levels tracks temperature is also nice. NOx increases as oceanic dust deposition rates collapse.

  14. Perhaps this is the place to ask a basic science question I’ve been wondering about for some time. Lets say you have a planet with half the surface at 40 C and the other half at -40 C. The average temperature is considered to be 0 C, right? Now, consider another planet with all its surface temperature at 0 C. It has the same average temperature. However, due to the non-linearity of blackbody radiation, the first planet will emit more radiation. Now, if you add atmospheres to these planets, everything will become even more complicated, because surface temperatures are only a proxy for top-of-atmosphere radiation.

    How are these effects dealt with?

  15. Following up on BBD comment on ICE extent – Lui et al use the ICE-5G reconstruction from Peltier 2004 (this information is given in the precursor paper He et al (2013)). An animation of the ICE-5G is given at

    Looking at the animation, rather than processing the netCDF files, I cannot see any change in ice cover since 8ka BP. So ice decline cannot be giving a warming signal after this time – unless the climate system has a slow response to the removal of this forcing (eg through vegetation responses). In CCSM3, the ICE only model (Fig 2) flatlines after 8kaBP, but LoveClim and Famous both show some increase.

    Note that summer snow cover probably increases in the late Holocene, the neoglacial, but the model should simulate this rather than having it as a forcing – it would be interesting to see how well the model does this.

  16. Richard,
    Thanks, that’s an interesting observation.

    You may need to explain the motivation behind your question, but I have a few comments.

    Firstly, when we talk about equilibrium non-greenhouse planetary temperatures that is typically defined as being the temperature at which the planet will radiate as much energy back into space as it receives from the Sun (or star, if we’re being general). It doesn’t really tell you how that temperature is distributed across the surface of the planet.

    So, two different planets that receive the same amount of energy from their parent stars would – to a first approximation – have the same non-greenhouse equilibrium temperature.

    I guess if you consider the moon, then it has – via this definition and ignoring albedo – the same non-greenhouse equilibrium temperature as the Earth, but the temperature on the Moon is very different to the Earth since it is basically set by how much radiation hits each part of the surface. So, very hot at the Equator at the meridian, cools towards the poles and towards the night side.

    On the other hand, if you mean how do we deal with this when we discuss surface temperatures on the Earth, then in some sense we don’t. What we’re interested in is how it is changing and – typically – we determine anomalies. This is really just telling us – on average – how much warmer we are relative to some baseline temperature. As long as we determine that baseline consistently, it doesn’t really matter too much. In fact, anomalies are actually relative to the baseline for each grid cell. In other words determine an average temperature for each grid region (which, I think, is a 5 degree square region on the surface) and for each time period (i.e., particular month or day) and then the anomaly is the difference the measured temperature for the same time period (i.e., the same month or day) and the baseline value for that grid region.

    I don’t know if that helps, but if you clarify why you’re asking, maybe someone can elaborate further.

  17. Andrew Dodds says:

    Interesting from other deglaciations:

    (Yes, not exactly prime source material..)

    Most of the time, it looks like the deglaciation results in a temperature ‘overshoot’ – the last 3 especially. This can’t be literal – there’s no momentum as far as I can tell or imagine. That at least suggests that the Holocene climatic optimum is a real thing and would perhaps be more pronounced if humans had not been present.

  18. ATTP, thanks for answering. I’m sorry if I’m not very clear, climate change is a communications minefield and I’m still trying to come to grips with that.

    As for why I’m asking: I’m wondering if a small amount of temperature anomaly may be explained by a different distribution of temperatures, without any change in radiative forcing. Lets say the tropics cool down 2 C. That gives us a large amount of energy in our energy budget. Lets get rid of this energy by heating the poles. Now, in order to do that, the poles need to heat up more than 2 C. So, we have a positive temperature anomaly but not a change in radiative forcing.

    I guess I would need to do some calculations to see how big this effect might be, but first I wanted to know I’m not making some conceptual mistake somewhere.

    Basically I’m saying that a 0.1C anomaly from a 30C baseline isn’t the same thing as a 0.1C anomaly from a -20C baseline. By averaging anomalies we are treating them as if they are. By doing this, we’re approximating things. I’m fine with that, I’m just wondering how much or how little error this approximation introduces.

  19. @Andrew Dodds There can be an overshoot if there is stored heat in the ocean that is suddenly released. The climate suddenly warms but the warmth cannot be maintained. I wrote about part of this recently on my blog.

  20. Rob Painting says:

    Long-term cooling from the early Holocene until the Industrial Revolution is also consistent with paleo sea level markers through this interval. The equatorial ocean saw a fall in ‘relative’ sea level over the last 4-5000 years as the Laurentide ice sheet had completely gone by this time, and the Antarctic and Greenland ice sheets were effectively stable.

    Ocean volume was therefore probably unchanging and, as water was siphoned away to fill collapsing glacial forebulges and the sagging of the ocean floor adjacent to the continents (continental levering), the equatorial ocean (the far field) experienced a drop in sea level. Sea level there fell about 3 metres over the last 4-5 thousand years – leaving ‘3 metre beaches’ littered throughout the equatorial ocean. Indeed habitable coral atolls were formed by this fall in relative sea level, with the cemented reef structure proving resistant to wave attack and thus suitable for colonization by the seafarers such as the Lapita people.

    As Richard Telford’s link shows, Peltier’s GIA model doesn’t support the idea of ice sheet loss contributing to warming throughout the Holocene.

  21. Bouke,
    Yes, I think you’re right that it would be possible to have a non-zero temperature anomaly without changing the outgoing flux. However, I don’t believe that this is what’s happening since we observe warming everywhere. Yes, there is polar amplification, so the Arctic is warming faster than the rest of the globe, but that might imply an even more concerning issue – if most of the warming is at the poles, then the change will be greater than if most were at the equator.

    Something to bear in mind is that our understanding of anthropogenic global warming comes from basic radiative physics. If we add greenhouse gases we will trap outgoing radiation, produce an energy imbalance, and will increase surface temperatures. The temperature anomaly measurements are simply an indicator of this process. We have much more evidence for this. Increasing ocean heat content. Measurements of the change in the outgoing spectrum. We see reductions in polar ice (both sea ice and ice sheets).

    So, although I think you’re right that it is technically possible to have a positive temperature anomaly without changing the outgoing flux, I don’t think it’s all that relevant an issue (there’s also the issue of such a change being physically plausible – why would the planet suddenly change how it emits radiation to space).

  22. Rob and Richard,
    So, I guess what I’m uncertain of now is what this implies about this paper. If the climate models had the correct models for the ice sheets, then the only warming effect should really have been from the small increase in CO2 (20ppm in the last 6kyr). An ECS of 3 degrees, would then suggest that this would be about 0.3 degrees of warming. So, is part of the problem here that they assumed continued ice sheet retreat and didn’t model the potential increase in snow cover probably, or is there more to this? I guess, it also seems that the models didn’t allow for a possible overshoot through the release of stored energy from the oceans.

    I guess I’m asking if – as AoM seems to have suggested – that really the models used in this paper weren’t really suitable for making this kind of comparison.

  23. Don’t forget the methane increase – but I’m not sure what the relative importance of CH4 and CO2 increases in the Holocene without checking.

    Snow should be modelled by the GCM – it is part of the climate rather than an external forcing. Also vegetations changes (which cause albedo changes) should be modelled by the coupled vegetation models

    I don’t see why AoM is suggesting that the models are not appropriate. OK they are at the simple (ie fast) end of the spectrum. Other model results exist for the Holocene (snapshots rather than continuious), I can check if they show the same Holocene trends.

    I’m not surprised that the seasonality bias of proxies is important – a recent paper in COPD (which I discussed here and was then asked to review, had great problems with a PMIP3 mid Holocene model proxy comparison, with obvious implications for Marcott et al. There are very clear (and predictable) discrepancies between proxies in the Norwegian Seas (surface proxies follow summer insolation, subsurface proxies follow winter insolation as subsurface temperature set during winter ventilation)

    If I think of anything interesting to say about Lui et al, I’ll try to post it tonight.

  24. Richard,

    There are very clear (and predictable) discrepancies between proxies in the Norwegian Seas (surface proxies follow summer insolation, subsurface proxies follow winter insolation as subsurface temperature set during winter ventilation)

    That’s very useful. I hadn’t quite established why they were biased, but that’s clear now.

    If I think of anything interesting to say about Lui et al, I’ll try to post it tonight.


  25. izen says:

    @-Bouke van der Spoel
    ” I’m wondering if a small amount of temperature anomaly may be explained by a different distribution of temperatures, without any change in radiative forcing. Lets say the tropics cool down 2 C. That gives us a large amount of energy in our energy budget. Lets get rid of this energy by heating the poles. Now, in order to do that, the poles need to heat up more than 2 C. So, we have a positive temperature anomaly but not a change in radiative forcing.”

    You are correct that the distribution of temperature across the globe and its variation can affect average temperature and total emissions in non-intuitive ways given the fourth power relationship. The Science of Doom website went into this in some detail a few years back, this is probably as good a place as any to start.

  26. izen,
    Thanks. Have only had a quick read, but it looks like a really good post by SoD. Very useful.

  27. anoilman says:

    richard telford, Anders, its not that I think models can’t be implemented or work, but the current models were designed for a very specific set of requirements. I don’t think Paleo analysis requirements were on the list. I’m suggesting that we may not know all the 1000 year etc longer time frame cycles. (But… I defer to those know know better.)

    I remember Anders had a post a while back about scientific discovery. “Eurika it worked! Hmm… that’s odd.”

  28. The question I would pose first about the applicability of the models for this case is:

    How well can the models tell the seasonality effect, and perhaps other biases in the empirical data?

  29. AoM, some specifics would make your argument a little more credible. LOVECLIM and FAMOUS are Earth System Models, designed to be fast enough to be useful for palaeo analyses. CCSM3 is a climate model, there is no obvious reason why it cannot be applied to past climate as well as future climate. Now some of the previous generation of models that needed flux corrections would be problematic to apply to long runs with gradually changing forcing, but I don’t think this problem applies to these.

    I’m not sure I thought of anything interesting to write about Lui et al, but I wrote it anyway

  30. BBD says:

    And then there’s science. Paleoecologists like Richard Telford refine the context in which proxies should be interpreted by paleoclimatologists. Geologists and paleo oceanographers reconstruct sea level variability over the Holocene and this helps to reduce uncertainty over the interpretation of eg. marine sediment cores. It’s all good.

  31. anoilman says:

    richard telford: You mistake my opinions for credible. 🙂

  32. jai mitchell says:

    A few thoughts,
    indicates that absorbed shortwave radiation effects are a much larger factor than modeled.
    shows that far infrared emissivity of open water is much less than ice, leading to increased ice sheet dynamics.

    I think that this discussion should include Ruddiman’s land-use and mid Holocene agricultural argument for the Anthropocene (really the Deleocene!) It seems that earlier analogs to the Holocene in the ice core record would have the world returning to ice age conditions between 2 and 6 thousand B.C.E.

    your image:
    indicates that the models severely underestimate ice sheet and GHG warming dynamics as well as (possibly) underestimate paleo aerosol effects (due to the LARGE underestimation of the former).

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.