Can we make better graphs of global temperature history?

MEA08 EIV, Ljungqvist10

1880-2013 CE



1850-2013 CE


HadCRUT4 Global annual average, Cowtan&Way (infilled)

1850-2013 CE

Berkeley Earth

Land+Ocean annual mean

Combining this data is certainly a challenge, and there are multiple approaches that could be used that range from the very simple to the very complex. More subtly the uncertainties need to be properly combined also. Issues range from temporal and spatial coverage, time-dependent corrections in d18O for long term geologic processes or ice volume corrections, dating uncertainty etc.

Finally, rendering the graphics calls for additional skills – not least so that the different sources of data are clear, that the views over different timescales are coherent, and that the graphics are in the Wiki-standard SVG format (this site can be used for conversion from pdf or postscript).

Suggestions for other data sets to consider, issues of calibration and uncertainty and trial efforts are all welcome in the comments. If we make some collective progress, I’ll put up a new post describing the finished product(s). Who knows, you folks might even write a paper…

This post was inspired by a twitter conversation for Sou from Bundunga and some of the initial data links came via Robert Rohde (of Global Warming Art and now Berkeley Earth) and Dana Royer.

Page 3 of 3 | Previous page

123 comments on this post.
  1. Eli Rabett:

    Well, of course Horace Mitchell and his scientific visualization studio at Goddard MD should be the best person in the world for something like this:), but Eli has a couple of ideas. Now if he only could draw. . .

  2. Keith Pickering:

    FYI, proxy temp anomaly data for 1730-1995 derived from Anderson et al. 2013 can be found at:

  3. Dan:

    Hi Gavin and Sou,

    This is a good idea in essence, but I am not sure it will bring anything new to the discussion. Having a better historical reconstruction is a very good idea.

    However, the problem is that the graphs are produced by scientists, and are interpreted by scientists who have an understanding of the science and explanations behind their construction and most importantly the results. The general public can see historical trends, but without understanding the science, can easily jump to the wrong judgment or conclusions as to what is presented and why it looks the way it does.

    Would a new reconstruction demonstrate that past temperatures were no higher or CO2 concentrations were no higher than current? No. Monckton et al would still use this information to hoodwink people, as they have been doing.

    It is therefore easy to see why somebody may jump to the conclusion that temperature has been greater in the past, or CO2 concentrations have been greater in the past, as they clearly have been. The reasons why this is the case, and whether this could be presented graphically, needs to be captured.

    As Keith #46 rightly states, forcing is a critical issue. But then the question is how to capture and present such data?

    One option could be to take the current solar energy flux and backcast global temperatures based on the historical CO2 concentrations, and planet albedo if this can be estimated, to show how warm it would have actually been historically, if the Sun output had been the same as is present. Therefore showing a greater correlation between CO2 and solar influence over the lifetime of the Earth, possibly, although there will be many other factors.

    Personally, I would think a graph showing the timeframe of modern man (say the last 250k years) would be also beneficial to focus the mind, as this is something that people can grasp far easier than millions of years ago. But that is not saying that far longer back in history should also be captured.

  4. wili:

    You might add a line in the graph giving the estimated number of species or families (though the latter may need more explanation for most viewers). This would go some way to countering the meme I have often seen in denialist posts that a warmer world would automatically have more ‘life.’

  5. Nick Stokes:

    I have tried a somewhat novel approach – a user-adjustable graph. Mouse clicking stretches the time scale from hundreds to millions of years. The graph can be dragged, and the temp scale varied. It’s a bit rough at the moment, but I’m hoping to improve. I did a post here.

    [Response: Thanks! – It’s a nice concept, but obviously needs some work. Note that there are scalings and baselines that need to be thought about (for instance, you are plotting LR04 upside down since temperature is inversely proportional to d18O). – gavin]

  6. Name:

    The proposal is very good and has the potential to be built upon to provide greater clarity on other matters of urgency.

    The graph which is an attempt by some editors at Wikipedia to produce a complete record for the Phanerozoic (as shown in Gavin’s post) sets the scene for a vertical line on the right-hand-side of the graph, if by 2100 the rise in temperature is 4-6 degree C above pre-industrial. That then shows clearly that the rate of temperature change in the coming decades has never previously been experienced, if you compare it to what has gone before. If the derivative is then determined and, for example, extinction data (eg as indicated at ) are plotted against it, the starting point which is the temperature graph can evolve to illustrate impacts. In the case of extinction intensity versus temperature change, there is a strong correlation between extinctions and change in temperatures. Life does not like sharp changes.

    In essence, therefore, the temperature graph could be the first step in an endeavour to create a open-source system that allows people to explore potential impacts. It is the potential for ordinary people to be able to explore impacts for themselves that could emerge as the value-add of the endeavour.

  7. Vince:

    The link for GISTEMP is dead. I think it should be

    [Response: Yes – thanks. – gavin]

  8. Steve Fish:

    Re- Comment by Eli Rabett — 14 Mar 2014 @ 3:32 PM, ~#51

    Hi Eli. I don’t have any ideas but am pretty competent with Adobe Illustrator. I have made thousands of teaching illustrations, but in my biological specialty. Steve

  9. Sou:

    Thanks, Gavin for letting me off the hook:) Now I can take it off the back-burner and will monitor progress here.

    I’m not sure if this has been mentioned already – Kevin Anchukaitis @thirstygecko pointed me to this paper by James Hansen, Makiko Sato, Gary Russell and Pushker Kharecha, which has a data supplement with this txt file:

    It may be of value in this exercise.

    [Response: It’s linked above. – gavin]

  10. wili:

    Even many people who follow climate issues quite closely will have no idea what “d18O” means. I would advise converting it to the estimated temperatures with a note that you are using this measure as your proxy here.

  11. Pete Dunkelberg:

    Different graphs and or pictures are good at different times. Didn’t a senator recently tell us that things were fine with high CO2 in dinosaur times, without mentioning that his state was under water at the time?

    Perhaps there could be a graph (along with some pictures) showing the area of the continental US on the same chart as CO2.

  12. perwis:

    Good initiative! For the context I also think it is important to include projections into the future in the graph. This strengthens the staggering perspective of what we are doing to the planet.

    My favorite climate graph of all time does exactly this. It is from the U.S. 2009 National Climate Assessment, and shows CO2 concentration from 800,000 years ago into 2100. The axes are linear and without any cutoffs. Simply amazing in its simplicity and so very powerful. It can be found here and is free to use:

    Otherwise I like Hansen & Sato’s graphs, e.g. in the paper referenced by Sou above.

  13. ardavenport:

    One alarming aspect of Climate Change is the rate of change of CO2, sea level rise, global temperature. Is there any way to extract from the data the rate of change of CO2 vs. time and rate of change of temperature, like ppm CO2/20 yrs or degrees /20 years? CO2 and temperature might have been high long, long ago, but the rate of change was very different back than from what it is today. And a high rate of change leads to more unpredictable weather — ask any farmed how important that is.

    A gray range showing upper and lower limits with the average line in the middle is always more readable than a graph full of error bars, but still captures the uncertainties in the measurements.

    Shouldn’t there be graphs of CO2 vs. time (left axis) and sea level vs. time (right axis), with a dashed line showing where sea level is in 1900 and 2000? Sea level rise is the most immediate attention-getting issue for a lot of people living near a coastline.

    NASA’s global temperature map is an excellent graphic of where climate change is going in time while showing the variations:

    Some places are hotter and colder but you can see that the average of the whole map is obviously getting hotter with time. With one look you can see both the average and how much variation there is, though it is only really effective as a video.

    Maps like this should be kept as up-to-date as possible.

    Climate data is inherently noisy. Just because there are a lot of ups an downs does NOT mean that we can’t see where things are going.

  14. Vendicar Decarian:

    Many Americans are incapable of understanding simple graphs, and unable to do
    basic arithmetic, so things have to be made simple, simple, simple.

    1: Follow the Wikipedia graph model
    Temperature is the most important scale, and the time scale must be compressed.

    2: Expand the last 100 years

    3: Expand the Pleistocene segment so that the transition into
    regular cycles of glaciation can be seen.

    2: Place a dot at time = 0 so that the current temp can be seen
    Label it as such with an arrow.

    3: Place a dot at time + 100 showing where the models place global temps
    in 100 years. Another where the final equilibrium temperature falls.

    Label as such with an arrow.

    4: Left hand side show dt ‘F inset dt ‘C

    5: Show error bars.

    6: Draw several lines downward at times of major biological events both
    evolutionary, and cultural labeled by number.

    7: Include some text for each numbered event available as a
    text caption.

    8: Put the image in the public domain.

  15. Dallas Dunlap:

    A real simple suggestion: In any graph showing global temperature over time,
    insert a horizonal line at the current average global temperature, so that the public can clearly see how,say, the temperature during the MCA compares to today’s temperatures.
    Also, something that I found useful in a debate I found myself in: When someone makes the claim that Mann et al removed the Medieval Warm Period, draw a horizontal line from the peak temperature of the MWP time period. People can look right at a graph and not understand what it shows.

  16. Villabolo:

    My apologies if this has already been brought up. provides an excellent means of generating temperature charts.

  17. wili:

    It would be great to include ‘consequences’ and ‘feedback’ graphs. One of the most dramatic of these is the loss of Arctic sea ice, especially volume, and especially contrasted with expected/modeled values as recently as 2007.

  18. Mike Roddy:

    Coming up with more precise temperature charts won’t solve the basic problem, which is media corruption. Six media companies (soon to be five) control 90% of content. They smirk at the public interest, and answer only to advertisers. We will continue to see inaccurate charts, or decent ones buried on page 17.

    These companies get away with it because they are only called out on lightly read blogs or stuffy academic analyses. We can’t have a democracy if the people remain uninformed about matters of life and death.

    The solution is to establish a robust media monitoring organization that is focused on climate. Bad actors would be linked and humiliated, in detail. It’s an idea I’ve floated for a couple of years, without success. If anyone knows of possible supporters of this notion, I can be reached at

  19. Mike Roddy:

    Producing accurate charts won’t help, because six (soon to be five) media companies control 90% of content. They answer only to advertisers, smirk at the public interest, and would only publish an accurate chart if it appears on page 17. The truth has become a harmless byproduct of geeks like us.

    Media critics dwell on lightly read blogs, or in stuffy academic analyses. The American public is far less informed on these issues than the people of, say, Uruguay or Latvia.

    We urgently need a climate change media monitoring organization that is funny and aggressive. Americans don’t see the truth because nobody is calling out those who are concealing it from them. I floated a plan to do something about this situation, but it went nowhere (“It’s already being done”), etc. A recent survey showed that climate change ranks #14 out of 15 concerns for the American public. Whatever we are doing is failing. If anyone knows of a person or organization willing to support this effort, I can be reached at

  20. mgardner:

    There’s a basic rule in education that every teacher learns the hard way, no matter how well educated one may be:

    You have to know where your student is at, and you have to recognize that it isn’t where you are at. (The more educated and specialized you are, the more that is true, especially in trying to connect with ‘the public’.)

    Scientists are human, and they can exist in their own bubble, and be parochial, and have a kind of tunnel vision. What impresses one’s peers may not really be useful in changing the mind of a hypothetical unbiased reader.

    The problem with ‘better graphs’ to counter ‘phony graphs’ is that the point of the phony graphs is not to convey information but to ‘look sciency’; to elevate the phony arguments and create a false equivalence ‘controversy’. (Remember “teach the controversy” for Evolution?)

    I wonder if rather than calling in the graphing experts, sympathetic marketing types would be more useful in structuring an explanation. Kudos to Edward Greisch @13 for mentioning Altemeyer’s decades of research, depressing though it may be.

    My experience with ‘not the best’ students is that they can access and internalize a narrative about physics with good quantitative understanding, but only if they can relate to that narrative.

    How about beginning at the beginning, with the *measurements* and what they mean rather than ‘data’– in other words, a series of slides illustrating *the science* that culminate in a nice graph. We need to face the fact that most people don’t have a clue what the global average temperature represents in the first place; statistically sophisticated arguments, however prettily presented, are not going to have much resonance.

  21. FurryCatHerder:

    I read most of the comments. I swear!

    I would strongly suggest that any chart which extends back more than 10KY very clearly note that civilization didn’t exist. For that matter, list the dominant life forms.

  22. GlenFergus:

    Ok, a first try at tidying that old WP graph:

    1. Hansen version of Zachos doesn’t fit, at either end (that shown is arbitrarily halved).
    2. What is the preferred estimate of global temps from L&R?

  23. Hank Roberts:

    > dominant life forms
    Those that change (or maintain) the climate system, and
    those that could be lost under current anthropogenic mismanagement,
    for example:
    Caldeira and Wickett, 2003, cited by 1500 subsequent papers

  24. Dan H.:

    Mike et. al.,
    The reason that climate change ranks so low among Americans has little to do with educating the people. Rather, people care less about those issues which interest or affect them less. It is also too long term for them to care. Many people can’t see past next week, let alone next year, decade, or century. Regarding news, the missing Malaysian jetliner has the monopoly on that at the moment. Even the Crimean election was relegated to the back pages.

  25. wili:

    Speaking of graphs, the Mauna Loa CO2 graph has risen dramatically recently. CC has a story on the first days above 400 this year we’ve just had turning soon into the first month ever above that mark.

    My calculations (which please do re-check for me, someone) are that we have just had the first 7-day period this year (11th-17th) during which the official current daily numbers average above 400.

  26. Meow:

    @70: We need to improve our arguments. In particular, we need to give up the idea that the dispassionate presentation of data and logical argument will persuade the general public.

    Data and logical argument address the intellect, but (for most people) leave the emotions untouched. Not one person in ten thousand looks at, say, Euler’s Identity and feels amazed.

    Rather, emotions rule most peoples’ lives, with the intellect following behind and assembling the debris into arguments supporting what the emotions have already decided.

    We must, therefore, address the emotions.

    Which means we must tell a story.

    It should be a story about a few people whom the viewer gets to know, not about multitudes who lose their faces in the scrum.

    It must be a story of peril and descent into darkness, but also one of hope and recovery.

    And it must be in the form that best reaches the emotions of the most people.

    Which means it must be a video.

    Got screenwriters?

  27. GlenFergus:


    1. That would be, doesn’t fit at one end. Hanson & Zachos ≠ Royer et al
    2. I used method from Hanson, Sato, Russell and Kharechka (2013), for consistency with 1.

  28. Vendicar Decarian:


    You can’t teach the willfully ignorant. In this instance, willfully ignorant American Republicans, and Conservatives in general.

    You can only convince those who are honest and willing to learn.

    Chipms don’t understand tables very well, and Average American Chimps are only marginally capable of comprehending the most simplistic graphs. Their minds are so polluted and thoughts so discordant that logical argument is mostly lost on them.

    A simple graph is the way to go.

    Wood for trees is a useful tool but it’s failure is in not providing error bars, and statistical confidence limits.

    As a result, it is therefore often hijacked by cherry pickers.

    The York plotter…

    Is good, but it doesn’t allow you to provide a simple link to any given plot. You have to rely on the apes to plug in some settings that you give to them second hand.

    A graphic is a good compromise, and has other utility like being able to be placed as a graphic in a web page, or used as a reference in a real publication.

  29. Vendicar Decarian:

    Re: 68,69 Bad actors

    An organization of 1000 people to monitor and respond to articles linked by the major “bad actors” – Dr*u*ge Report, etc., would be sufficient to take the wind out of the sails of the American Conservative Lie Machine.

    But given a few more years, that wind will be lost due to the change in climate itself.

    Have your lists ready when the revolution comes.

  30. GlenFergus:


    Royer et al (2004) and Hansen/Zachos don’t agree; not even remotely:

    My     Royer 2004    Hansen/Zachos
    60          2.28                  10.39
    50          1.40                  12.98
    40          0.33                    8.6
    30         -0.05                    5.0
    20         -0.29                    5.34
    10         -0.08                    3.66
    0            0.00                    0-ish
    Re-reading Royer, their adjusted Veizer T is meant to be a low-latitude shallow sea estimate. Is there grounds to factor it towards something more “global”, say x 2? (It would still be far away, but less so.)

  31. Dan Wang:

    It would be very nice to see a graph of global temperature versus time reaching back over at least a thousand years with all methods for data collection and correction and data sources spelled out in detail. It may be necessary to have different time scales also since the number of data points per unit time was lower in the past, and higher in recent times.

  32. BillS:

    It has now been been 6 days since Edward Greisch posted his comment at #13. It appears that his comment has been ignored while most have immersed themselves in the minutiae of graph creation.

    I think Greisch point is, “How do you preach NOT to the choir but to the non-choir?” This is a much more difficult problem than preaching (or graphing) to the true believers, i.e., the choir. (See Eric Hoffer on the notion of true believers.)

    The recent AAAS initiative, “What do we know” may be a good step in the right direction without any graphs at all “assuming that most people are actually smarter than chimpanzees”. One can hope.

  33. jyyh:

    dominant life form is a pretty hard thing to determine in paleontology, but how about something like this? (rights to alter the image afterwards reserved)

  34. thomas marvell:

    I seems like you are trying to display trend information to non-climate scientists. You should get input from the audience, and I am one of them. But I have looked at lots of these graphs.

    The graphs should be simple, intuitively obvious, and a bit entertaining.

    They should not be misleading.

    You might want to work with commercial graphic designers, who are experts at getting ideas across to the public.

    Some rules I would suggest are:

    1) There should be one line for each concept – temp. CO2, sea level, etc. Lots of lines for different estimates looked cluttered.

    2) To get a single line, one should use a commonly accepted notions of averaging. I would suggest the median estimate for each observation. Also, possible is the average, dropping highest and lower figures.

    3) The graphs should not include measures of uncertainty, such standard deviations. If the stand deviations are very large, the data series should not be used. Otherwise, text describing the graph can give rough descriptions of the uncertainty.

    4) There should be separate graphs for long and short time periods. Log graphs are not intuitively obvious.

    5) The graphs should not include technical words – especially the geological periods.

    6) The graphs should supply internet links to more detailed graphs and data descriptions. The major reason for this is so someone intending to mislead cannot use simplifying aspects of the original graphs to do so.

  35. Jim:

    Logarithmic scales in particular are a potential problem. I have a demonstration of some of the main issues here:

  36. Richard Badalamente:

    The rate of change of atmospheric CO2 since industrialization is unprecedented over that period for which we have reliable data. We know theoretically and can demonstrate experimentally that CO2 has a positive RF; other things being equal, more CO2 means more warming. Statistics have nothing to do with the demonstrable RF effect of CO2. Statistics come into play because of the “other things being equal” qualification with regard to the attempt to measure CO2-caused warming.

    My point is that perhaps one graph or chart is insufficient for the purpose of relating CO2 to a reconstructed graph of temperature. Perhaps a set of graphs showing CO2, TSI, aerosols, clouds, etc., might be useful as overlays — a sort of pictorial PCA?

  37. Dieter:

    Looks like my earlier comment was lost in the ether, but not to worry. I’ve been playing around with the idea of presenting multiple linear timescales on a single interactive graph here (simple proof-of-concept only). It could help with avoiding log scale, but only over a couple of orders of magnitude I guess. An expandable time-axis slider might be a better approach. Making graphics interactive also lets you keep them clean and simple but embed additional detail (references, links to datasets etc.) directly into the figure.

  38. GlenFergus:

    The Wikipedia graph has been updated. (The graphic in the OP is now the new version. The old one is here.)

  39. jyyh:

    fixed a couple of types in the wikipedia image explanation, thank you for updating.

  40. PaulW:

    Glen Fergus, on the updated Wiki graph, you’ve expanded the Royer 2004 temperatures by 2 times. While that makes the numbers closer to what temperatures were more likely to have been, the explanation that they were meant to represent shallow tropical oceans is not correct. They were meant to represent global temps. The issue with Royer 2004 is that they use a 50 million year smoothing routine (following on what Veizer did) which downplays the changes that occurred in the Cretaceous, the Permian, the Ordovician etc. I think the values produced by Rohdes directly from the Veizer isotopes is more representative. Just use Rohdes’ values as they are smoothed over 3 million years which then starts to capture more realistic temperature variations.

    In addition, the Hansen 2013 temperatures from 5.3M to 65M years ago are meant to be polar oceans rather than global temperatures. They have to be scaled back by 50% to be more representative of global temperatures. The Eocene should not be warmer than the Cretaceous, Antarctica did not glaciate over 33.6 Mya with global temps at +5.0C from today. The values should also be detrended at the same rate that Veizer and Royer and Rohdes have done over time because the isotopes suffer from diagenesis over time.

  41. MARodger:

    On the new-version Wikipedia graph, I think the ‘blobs’ to mark the temperature rise of 20th/21st century are a big improvement for putting geological temperature record in the context of recent changes. I was toying with a slightly different idea to do the same thing – a shaded band across the whole graph showing the temperature increase since 1900 and another shaded band showing the projected RCP8.5 range for 2100 – so the second of these would be something like this (although a red shading may be more appropriate).

  42. GlenFergus:

    Thank Pauls, I really would like to get it right, and definitely need help with that…

    Royer et al 2004: The rock record of glacial deposits can only be qualitatively compared to other records of climate, such as CO2. It is within this context that the low-latitude paleotemperature data of Veizer et al. (2000) and Shaviv and Veizer (2003), based on the shallow-marine δ18O carbonate record of Veizer et al. (1999; Fig. 3A), is so appealing.

    That says tropical (or at least subtropical) seas; do they change that by their re-interpretation? The objection to Veizer et al (plotted by Rohde), as I read it, is that they didn’t correct for CO2, pH and Ca++, so get it badly wrong (??)

    Hansen et al 2013: “Our first estimate of global temperature for the remainder of the Cenozoic assumes that Ts = Tdo prior to 5.33 Myr BP, i.e. prior to the Plio-Pleistocene, which yields a peak Ts of approximately 28◦C at 50MyrBP (figure 4). This is at the low end of the range of current multi-proxy measures of sea surface temperature for the Early Eocene Climatic Optimum (EECO) [79–81].

    That says global. How do you tell it is meant to be polar SST before 5.3 My? I agree it sure looks high. Maybe you could email me: Gmail gergyl.

  43. GlenFergus:

    MARoger: Nice. Doesn’t need to be red; can be a mistake to overstate IMO.

  44. Nick Stokes:

    In #55 I linked to an experiment with user-adjustable graphs which could adapt to widely varying time scales. I’ve now embedded this in a general climate plotter. It adds a few paleo datasets and allows mixing of different types, with an adjustable third axis. There’s still lots to be done on the actual data.

  45. PaulW:

    GlenFergus, Veizer and Zachos’ isotopes come from all over the world. They are really a collation of every study from every drilling program that they could get their hands on. And there are literally 18,000 datapoints in Veizer’s database and 14,000 in Zachos’.

    And then the question is how does one turn the isotope database into a global temperature estimate. There are formulae available to do that. The isotopes do vary with temperature based on location, proximity to the ocean, and especially with latitude.

    How does one take 18,000 isotopes from all over the world which have also moved over time with continental drift and Ph factors and diagenisis. We calibrate the whole database to known temperatures from other periods. Ph is not actually an issue because the whole database needs to be consistent with what we know.

    We know the last glacial maximum was around -4.5C, it was about +2.0C in the Eemian interglacial for example.

    We know that tropical oceans today can get to 32.0C or 17.0C warmer than the global average temperature. Did Eocene tropical oceans really get 16.0C higher than today’s 32.0C. That would be far too hot for complex life-forms. There are almost no animals that can live in a 48.0C ocean. Eocene temperatures seem to be about +6.0C compared to today based on the best estimates.

    We also know we had ice ages during the Carboniferous and the Ordovician as parts of Gondwana transited the south pole. Global temperatures had to be similar to today for large glaciers to survive for tens of millions of years at a time. etc.

  46. Pete.W:

    Here is just one example of a deliberate misuse of charts. (Its so easy to do!)

    For a while WUWT was choosing to link to this chart on their side bar as an example of their being “fair and balanced” by sharing real Arctic sea-ice-extent information with their visitors;

    But here is another chart that most of us would agree is far more useful;


  47. Tapio Schneider:

    Here is a graph for the past 3 Ma that I had put together for my own purposes (teaching and a book I am working on):

    The graph would be easy to extend to 5.3 Ma with the Lisiecki and Raymo (2005) benthic d18O stack, and to 65 Ma with the Zachos et al. (2001) data, with similar rescalings to estimated global temperatures as in this graph.

    Figure caption: Estimated global- and annual-mean surface temperatures over the past 3 million years. The temperatures are expressed as anomalies relative to the mean for 1961–1990. (a) Temperatures of the past 11.3 kyr, inferred from a variety of globally distributed temperature proxies. The orange line on the far right shows the smoothed direct temperature measurements from [NASA GISS]. (b) Temperatures of the past 420 kyr, inferred from the relative abundance of deuterium 2H to hydrogen 1H in an Antarctic ice core. (c) Temperatures of the past 3 Myr, inferred from the abundance of the oxygen isotope 18O relative to 16O in the calcium carbonate shells of benthic (bottom dwelling) foraminifera, recovered from ocean sediments. Shading in all panels indicates rough estimates of 95% confidence bands.

    Details: The temperature reconstruction in panel (a) is from Marcott et al. (2013). The temperature reconstruction in panel (b) is based on the European Project for Ice Coring in Antarctica (EPICA) Dome C ice core (Jouzel et al., 2007). The temperature anomaly of snow formation that Jouzel et al. estimated from the relative deuterium abundance in the ice is divided by a factor 2 to obtain an approximate global-mean surface temperature anomaly. The factor 2 accounts for the polar amplification of global climate change and is estimated from climate model simulations (Masson-Delmotte et al., 2010). The temperature reconstruction in panel (c) is based on the relative 18O abundance in the shells of benthic foraminifera, recovered from ocean sediments at sites distributed around the globe (Lisiecki and Raymo, 2005). Bintanja and van de Wal (2008) used a land ice model to deconvolve the effects of ice volume and temperature on the relative abundance of 18O in foraminifera shells. The resulting deep-sea temperature anomaly inferred by Bintanja and van de Wal (2008) was multiplied by a factor 1.59 to obtain an approximate global-mean surface temperature anomaly that matches the temperature anomaly inferred from the EPICA Dome C record in the period of overlap in a least-squares sense (cf. Masson-Delmotte et al., 2010). The confidence bands are approximations based on published uncertainty estimates: In panel (a), they are from Marcott et al. (2013); in panels (b) and (c), they are 2 K wide, roughly taking into account measurement uncertainties and the uncertainties in obtaining the estimated global-mean surface temperatures from the inferred Antarctic and deep-sea temperatures (Masson-Delmotte et al., 2010).

    The confidence bounds are rough, basically educated guesses based on various published sources of uncertainty. I would be interested in improving them.

  48. GlenFergus:

    Paul, pH is the key confounding effect investigated by Royer et al 2004. Have you read the paper? The section headed “The effect of seawater pH on the δ18O of marine carbonate“?

    I see Zachos et al 2006 (ref’d by Hansen et al 2013) has early Eocene mid latitude SSTs of ~26-33°C (excluding the error bars). Hansen et al’s global estimate for then is ~28°C (which I plot as +14°C vs a 1960-1990 mean of 14°C). That strikes me as potentially a little high, but unlikely way out.

    Are there similar estimates from the mid Cretaceous? I mean from plaktonic forams, not brachiopod shells? Maybe Bice et al 2006? Anyone know this literature?

  49. Bill Bunting:

    In case iy has not been proposed I would put Hans Rosling’s Mind the Gap software to yhe task. This is orecisely the kind of challenge that it was designed for.

  50. Vendicar Decarian:


    I think the change showing the bar representing projected temps is a good one.

    People still need some events on the chart to give them a feel for how long ago those 500 million years represents.

    Extend the box vertically and add some date sign posts.

    1: Emergence of land plants.
    2: Emergence of land animals.
    3: First mammals.
    4: First Birds.
    5: First Primates.
    6: Dinosaur extinction
    7: Emergence of modern humans
    8: Industrialization

    In addition the left should be augmented with an additional box that has a scale of 200 years.