Global temperatures over the past decade

It is well known that the past decade or so has seen less global warming than might have been expected – but what is the cause? This is more of a discussion post, rather than any new analysis.

The most recent decade has seen observed global temperatures at the lower limit of the model projections. There seem to be 3 possibilities for this relative slowdown in the rate of warming:
1) Internal climate variability
2) The assumed radiative forcings are wrong
3) The climate simulators used are too sensitive to greenhouse gases

I think there is evidence that all 3 possibilities are playing some role. Firstly, climate simulators show a range of internal variability behaviours, and a decade with no global warming (or even a cooling) is not implausible – various analyses indicate that around 5% of decades should exhibit a cooling trend globally, perhaps because the warming is in the deeper ocean. In fact, we might have expected a cooling decade sometime in the next few decades anyway. If internal variability is the sole cause, then we might expect a more rapid warming over the next decade.

However, another option is that the radiative forcings used in the climate simulations are somehow incorrect. The most obvious culprit would be the emission of aerosol precursors, which help cool the planet. The scenarios used by the IPCC optimistically tend to show a rapid reduction in aerosol emissions from 2005 onwards, which may not have happened. So, perhaps some of the relative lack of warming is due to the fact that we are not using the observed forcings, but projected forcings after 2005? In addition, even if the projections have the correct forcings, the models may be too sensitive to aerosol reductions, and so the projections produce a more rapid warming than that observed. Other candidates for producing incorrect forcings are stratospheric effects or volcanic aerosols.

The final possibility is that the higher climate sensitivity models are too sensitive to greenhouse gases. Recent analyses in the ‘Detection & Attribution’ framework have suggested this, along with initial work on examining hindcasts from decadal predictions.

What would help answer this? More time to see what happens, of course. A better understanding of recent aerosol trends, and whether the models are responding correctly would be very beneficial. Also, more regular updates to observed emissions and radiative forcings to allow more concrete detection and attribution of trends.

My suspicion is that all 3 possibilties are playing some role. The next few years will be very interesting indeed!

About Ed Hawkins

Climate scientist in the National Centre for Atmospheric Science (NCAS) at the University of Reading. IPCC AR5 Contributing Author. Can be found on twitter too: @ed_hawkins

31 thoughts on “Global temperatures over the past decade

  1. Hi Ed,

    Apologies, that must have seemed a little terse. I hit post by mistake while attempting to do my own wobbly version of multi-tasking.

    At the modern end of the record we have to deal with new systems coming in to operation, old systems being phased out, reduced data availability in near-real time and a host of other things. Systematic errors and data homogeneities are not something that can easily be assessed in real time. For example, land station data homogenisation tends to rely on identifying breaks based on neighbour comparisons which require a reasonable length of data after the break has occurred in order to identify that it has happened and make appropriate adjustments.

    At the earlier end of the record it’s clear we’d gain some additional insight by extending observational archives back as far as we can.

  2. Hi,

    followed a tweet here. Why has it not warmed this last decade? Because there has always been natural variation. The climate warmed as much from 1910 to 1940 as it did from 1970 to 2000, but CO2 hardly rose in the first period. So, we know that natural variation can easily give us warming equivalent to 1910-1940, so the first and best assumption for any warming not explained directly by CO2 (about 0.3°C) is that it is natural variation.

    Solar activity is an obvious candidate. Solar modulation of cloud cover is pretty conclusion. The connection between cosmic rays and climate is all bit confirmed by laboratory tests at CERN. In contrast there is NO EXPERIMENTAL coroboration with the speculative scaling up of CO2 by hypothetical positive feedbacks.

    I hope that helps. If not get in touch.

    1. Hi Mike,

      You need to remember that CO2 is NOT the only thing that we believe affects the climate. You need to realise that aerosols (as discussed in the post) have masked much of the recent warming due to CO2, as they cool the climate. You need to realise that the sun may have played some role in the 1910-1940s warming, and that there were several large volcanic eruptions just before 1910 which also cooled the climate, from which the climate was recovering. We consider all of these factors.

      As for the cosmic ray experiments – CERN scientist Jasper Kirkby, said about his recent cosmic ray experiment:

      “At the moment, it actually says nothing about a possible cosmic-ray effect on clouds and climate, but it’s a very important first step”

      I hope that helps!

      1. I’ve read the solar papers that say the solar TSI from the last half of the 20th century are insufficient to cause the warming of the 1st half, which is what you are implying here. The observations don’t match that, however. If you look, you will see that the average TSI from the second half century is just about equal. The idea or bias that the sun can’t participate in late 20th century warming is quite wrong, in other words.

        1. Hi James,
          Solar TSI has an influence – but it is small. We know this because if it was a strong effect then there would be a clear 11 year solar cycle in global temperatures, which there isn’t. Also, the stratosphere would be warming, not cooling, if the sun was responsible.

  3. It’s worth mentioning the wok of Foster & Rahmstorf (2011), who adjust various temperature datasets to remove the estimated impact of ENSO, volcanic aerosol and solar variability.
    The resulting warming rate is found to be steady over the 1979–2010 period studied. Punchline figure here.

    1. But does ENSO capture the full possible extent of the impact of internal climate forcing on a temperature trend? When Forster and Ramsdorf write

      “The resultant adjusted data show clearly,
      both visually and when subjected to statistical analysis, that
      the rate of global warming due to other factors (most likely
      these are exclusively anthropogenic) has been remarkably
      steady during the 32 years from 1979 through 2010.”

      Then they obviously think it does. FR2011 is a better example of this type of study in that they name what internal variability is, in their case ENSO, and so we can argue about whether they are correct. In other attribution style analysis internal climate variability is often relegated to little more than the residual of an analysis or is expressed in the error bars of a particular estimate.

      Ed I was wondering what you thought of this aspect of attribution studies given you seem to specialize in internal variability? Am I being unfair in my description of these studies? Are there any attribution studies where you believe internal variability is well accounted for?

      1. Hi HR,
        ENSO is obviously not the only component of internal variability – plenty of other pages here consider other aspects – but it is a large component.

        In many standard ‘detection & attribution’ (D&A) analyses the internal variability is indeed treated as a residual, and checked for consistency with our understanding and with estimates of internal variability from climate simulations. In that sense it is well accounted for, but we are doing some more work on extending D&A type analyses to consider internal variability more explicitly.

        Other studies such as Thompson et al:
        consider more than just ENSO in determining the internal variability components of the global temperature trends.

        But, there is certainly scope for combining better understanding of internal variability with attribution studies.

  4. I don’t know that i completely agree with your premise, imo, climate/weather are terms we use when talking about the movement of energy from the tropics to the poles. Very few people expected the amount of change happening at the poles, in the last decade, or said so, and whilst temperatures away from the poles may not have shifted much, the extra energy in the system, going poleward, has manifested itself in a whole array of extreme weather, worldwide.

    1. Hi John – I agree that what is happening in the Arctic is faster than expected. However, the Arctic is quite a small part of the planet in surface area, and it is the global average which has shown less warming than might have been expected.

  5. Ed,

    I hope it’s not too late to post on this?

    I have recently been in conversation with a colleague who has done some fairly simple 4-parameter linear regression on global air temperature and his findings suggest another possibility to the slowing down of temperature increases over recent years – the AMO. His analysis,, seems to do pretty well in modelling Temp, especially over the last 35 years or so.

    I note that you wrote a post about the AMO in 2010 considering the possibility of its influence but I can’t find a follow-up.

    There remains the question about which is the independent or driving variable – AMO or temperature, but they seem to be significantly correlated. More likely they operate in a feedback relationship, although I admit to not knowing much about the system. I would be very interested to hear your views.


    1. Hi Rob,
      It’s never too late to discuss!

      The role of the AMO would come under the natural variability possibility I discussed, and is definitely a possibility for having an influence. I note that your colleague has not included anthropogenic aerosols in his regression model which seems to be an important omission.


      1. Ed,

        Thanks for the reply. Point taken about internal variability.

        However, I’m still left wondering why, if he’s managing to explain 88% of the variance of mean global temperature over the last 150 years with just two variables (ghg, AMO), the inclusion of AMO isn’t a specific variable in (any?) GCMs? After ghg, is there any other driver that can explain as much of the variance (18%). Anthropogenic aerosols seem to pale into insignificance beside it.

        What am I not getting here?


        1. Hi Rob,

          The point is that the AMO might itself be influenced by the anthropogenic aerosols and/or volcanoes (see recent papers by Ottera et al & Booth et al.), and so prescribing the observed changes to the AMO may actually be missing the driving factor. We must always remember correlation is not causation! And, we know that aerosols are a significant factor – their radiative forcing is not small.


  6. I am the ‘colleague’ mentioned by Rob Brown above.

    First of all I fully accept that correlation is not causation. However I think is valid to use MLR to:
    1. Quantify the influence of known drivers (like TSI for example).
    2. Identify other potential drivers (like AMO).
    I have sometimes queried whether or not the AMO is the ‘heartbeat’ (i.e. driver) or the ‘pulse’ (i.e. an indicator of whatever is the driver). At the moment it is best to consider it as a heartbeat until a mechanism is found.

    I aslo accept your point about anthropogenic aerosols. Though I would point out that Foster and Rahmstorf didn’t includ them either. I’ve not looked at their influence. They only change the conclusions, relative to my earlier posting, very slightly. The earlier 4-parameter model suggested the GHGs account for 49% of 1976 to 2005 temperature increase; the new model gives 59%.

    My results are similar to those in Zhou, J., and K. Tung, 2012 and I’ll posting on this separately.

    1. Hi Ron,
      I agree that MLR is a valid approach to consider some of these issues, as you have already noted from the literature. There are also a couple of papers by Lean & Rind which you may find interesting.

      I think you would need to include ENSO in your model – we know this has a large impact on global temperatures, and is definitely a heartbeat, rather than a pulse. The AMO is not so clear cut – it could be either or both.

      Also, Foster & Rahmstorf didn’t include aerosols because they were trying to remove the variability components to find the trends, which is different from what you are doing.


  7. Thanks for commenting.

    I’m aware of the Lean & Rind Papers.

    I did try a 5-parameter version: AMO, TSI,GHGs, Volcanoes & ENSO (but no Anthropogenic aerosols). For the period 1871 to 2011 it increased the explained variance from 88.6% to 91.5%. The reasons I did not pursue it were fistly, given the serial correlation in the variables, adding more parameters is as likely to be mopping up the noise as adding information and secondly its effects are short term (a few years) and I was more concerned with the longer term.

    That said, the error bands for ENSO coefficient in the 5-parameter model were low, relative to the coefficient, and there is no doubt that ENSO is a powerful global climate indicator.

  8. Hi Ed.
    I’d like to suggest a fifth possibility besides the three you mentioned plus the possible observation error mentioned by John.

    5) External variability.

    the Sun has been a lot less active since 2003 than in the previous five decades. Now, I know that various studies purport to show that solar variability is not a big factor in Earth’s surface temperature variation, and I realise that those studies have been successfully published in ‘the literature’. I’m an amateur climateer with an HNC in Mech Eng and a degree in Hist/Phil Sci with no publication history apart from on my own blog, but if you have the time, and the open mind, I’ll attempt to show you those studies are wrong. Are you up for it?

  9. Regarding that natural variability suggestion: funny how that works. At the start of the century, every heat wave and every fractional degree of increase was presented as part of the constant, ever-onward increases in temperatures we should expect. Those increases were not weather, they were ‘climate.’ When global temps plateaued, suddenly we were told that such a LACK of increase was ‘just weather.’

    While you may get a ten year period without warming during a general increase, that is not what was observed. What occurred was a lack of warming during the FIRST decade of the century. Which is fundamentally different. We didn’t increase for five decades and then see a ‘pause’ – the lack of increase occurred at the start of the century.

    Rather than deal with this obvious fact, the response has been to sleight of hand. Now, every hotter has disappeared, and we’re being told ‘this is the hottest decade,’ or ‘this was one of the five hottest years.’ Well, yeah – but there was no increase.

    There is a very well known phenomenon in science. When someone’s hypothesis is challenged with new evidence, rather than following the logic of the new evidence, he backs up. The advocate retrenches, points to different support, changes the subject. Anything but allow that he might have been wrong in the first place. Whether the consensus position is right or wrong will be demonstrated in the coming decades. But right now, they are behaving like they guy who refuses to deal honestly with new data. Such behavior doesn’t change the science, but it does support a rational skepticism of the strength of support for the hypothesis. If you’re confident of your data and your analysis, there’s no need to weasel.

    1. Hi Mark – I disagree. The ‘weather events’ (e.g. EU heatwave of 2003, Russian heatwave of 2010) that we have seen have been attributed to the fact that temperatures have increased, rather than they are increasing currently. Big difference. And, global temperatures are still increasing – the linear trend is still positive, but not significant. But, as the post says, there are several possible explanations for the most recent decade and as scientists we are trying to work out which ones are responsible.

      1. Hi Ed,

        Could you clarify what is meant by:

        “And, global temperatures are still increasing – the linear trend is still positive, but not significant.”

        Is it that the measured/estimated trend is not beyond what is plausible due to experimental/observational errors given uncertainties due to coverage, bias, choice of end points etc., or that it is not significantly unlikely given natural variance?

        I am never quite sure if people are commenting on whether the trend is real or whether it is surprising.

        This also applies to when I read the IPCC reports.

        WGI chapter 3 deals with observations yet Appendix 3.A details a whole lot of guff about natural variability not required to answer the question “Are we confident that we have observed some warming, and if so how mach?”. Given that the question of whether or not the warming is unlikely to be natural is dealt with in Chapter 9 where natural variability is correctly considered I would have thought that Chapter 3 might confine itself to whether the observations are sound. The clue being that it is a chapter on observations.

        I think there be numerous occassions where it may be important that the observed trend is sound irrespective of whether or not it is natural, e.g. species migration, and in this case extreme events.

        When people say that there has been no significant warming since xxxx, I never know if I should infer that we are unsure as to whather the observed trend is robust or merely that it is not beyond the bounds of happenstance. That clearly bothers me.


        1. Hi Alex,

          When I said ‘not significant’, I meant that it is not significantly different from zero, given the natural variability, i.e. it could have occurred by chance if the actual long term trend was zero. The observational errors would add to this uncertainty. I agree that this is a bit of a red herring. However, we would expect the trend to reemerge sooner rather than later…


Leave a Reply

Your email address will not be published. Required fields are marked *