Wednesday, April 22, 2015

Not so fast, with that so-called hiatus...

Sou | 5:57 PM Go to the first of 19 comments. Add a comment

Update: Patrick Brown, the lead author, has just written an article about his paper, at realclimate.org. He is answering some questions in the comments.

Sou 14 May 2015

Anthony Watts (archived here) has posted a press release about a new paper in Nature's open access journal, Scientific Reports.  The research was led by Patrick T. Brown, a PhD student Duke University. What the scientists did was compare recent and projected climate change with that which has occurred over the past 1,000 years - mostly in the northern hemisphere (which is where there is most data). They looked at the recent global temperatures and compared them with projections of climate models - and drew what I see as overly-confident conclusions.

I feel a bit stretched on this subject, and am happy to be shown if I'm wrong about this.

A puzzle - RCPs haven't diverged yet

The paper is puzzling me, because it indicated that the current warming is more in line with RCP6.0 than RCP8.5. However in regard to atmospheric greenhouse gas concentrations, the various pathways would only now be just starting to diverge. So how can the scientists say that the recent values and trends are different for RCP8.5 and RCP6.0? I'd have thought they should be the same.

Here are the concentrations for CO2 under the main RCPs, from van Vuuren et al (2011) :

And here they are for CH4:

What Patrick Brown and co. did

I figured I was missing something so I looked into the paper a bit more. It seemed to me that what Patrick and his colleagues were comparing wasn't the recent temperatures and trends with the modeled temperatures and trends for the same period. For the recent temperatures, they were looking at actuals from 1993 to 2013. For the modeled data they were analysing over the period 1993 to 2050. And yet they drew conclusions about a recent short period - from 2002 to 2013 inclusive.

The authors did various statistical analysis on modeled data and observations. In the summary the authors state:
In this work we created a very large ensemble of stochastic realizations of unforced GMT [global mean (surface) temperature] noise that were based empirically on the instrumental record and reconstructions of surface temperature over the past millennium. We used this ensemble to create an empirical estimate of the EUN [Envelope of Unforced Noise] and used it in the comparison between observations and the forced signal produced by CGCMs [Coupled Atmosphere-Ocean General Circulation Models] over the 20th and 21st centuries. We find that:
  • the interdecadal variability in the rate of global warming over the 20th century (i.e., acceleration from ~1910–1940, deceleration until ~1975, acceleration until ~2000) is within the 2.5–97.5% EUN, even if the forced signal is represented as a linear trend, indicating that this observed interdecadal variability in the rate of warming does not necessarily require interdecadal variability in the rate-of-increase of the forced signal.
  • ...recently observed GMT values, as well as trends, are near the lower bounds of the EUN for a forced signal corresponding to the RCP 8.5 emissions scenario but that observations are not inconsistent with a forced signal corresponding to the RCP 6.0 emissions scenario.

The "recently observed GMT values" are, I think, the values from 1993 to 2013. It's a bit hard to tell from reading the paper. The researchers did do an analysis, which was described as (my paras):
 ...linear trend distributions of the ESRUN superimposed on the three RCP forced signals, as a function of trend length. These distributions account for all overlapping trends of a given length between the years 1993–2050.
1993 was chosen as the start year in order to extend the analysis as far back as possible while still avoiding any major volcanic eruptions and 2050 was chosen as the end year to avoid the post-2050 reduction in the rate of increase of the RCP 4.5 forced signal56. These trend distributions are compared to the recently observed trends calculated on the GISTEMP dataset from 1993–2013, 1994–2013, … , 2010–2013. The goal of this comparison is to see if observed trends over the 1993–2013 period are consistent with the mean rate-of-increase of the CGCM-produced forced signals between 1993–2050.

Remember Pat'n Chip's AGU Poster?

You might recall the work of Pat'n Chip, with their running comparisons of global surface temperatures in a poster session at AGU14. This comes across as being something similar. If so, it suffers a similar flaw. The flaw is that every run of observations ends in the same year. Well, it couldn't go beyond 2013 because at the time of preparation, the data for 2014 wasn't in.

What I think they've done is the same. Looked at the trend from 1993 to 2013, from 1994 to 2013, 1995 to 2013 etc etc. What this would mean is that the trend is dominated by the actual mean surface temperature for 2013. Add 2014 and you will get a different trend for each of the runs.

Grant McDermott wrote a nice critique of this technique back when Pat'n Chip used it. You can read it here on his blog.

Comparing different periods?

The other problem as I see it is that the scientists are comparing a trend projected out to 2050 with actuals only going to 2013. It says nothing of what will happen between now and 2050. As I showed above, there is not a lot of difference between the greenhouse levels under the different scenarios at this time. Starting soon, the RCPs will diverge, but they haven't really done so yet. So to make the claims the authors have, in regard to the different pathways, don't appear to me to have any sort of solid foundation.

In the press release (not the paper), Patrick Brown is quoted as saying:
"Based on our analysis, a middle-of-the-road warming scenario is more likely, at least for now," said Patrick T. Brown, a doctoral student in climatology at Duke University's Nicholas School of the Environment. "But this could change."

How a "negative" becomes positive - just add the latest year

In the paper, the authors didn't go quite that far. Here is an indication of how far they did go:
Over the years 2002–2013, a linear trend fit to GISTEMP reveals a slightly negative slope (Figure 3a, 3b, 3c). Because special attention is often given to cooling episodes embedded in the long-term global warming signal, it is valuable to quantify the likelihood of such events occurring.

It struck me that this was a very short period of time, so I put together my own charts, using the same data source as they did, GISTemp. I've animated the periods 2002-2013 and 2002-2014. As you can see, once you add in 2014, the trend again becomes positive.

Data Source: GISS NASA

In any case, for the scientists to say it was a "slightly negative slope" from 2002-2013 is a bit cheeky, given the R2 value for the linear trendline. If they'd waited a couple of months, they'd have missed the opportunity!

And if they'd looked at the slope from 2003 onwards, they would have seen it was a "slightly positive slope" from 2003 to 2013 - and an even more "slightly positive slope" from 2003 to 2014!

Data Source: GISS NASA

The length of time between temperature hikes

The scientists played with the numbers some more, and defined a Cancellation Timescale as "the longest period of time where there is at least a small chance (5%, 2.5% or 0.5%) that the positive trending forced signal could be completely ‘canceled out’ by a negative unforced fluctuation, resulting in a flat trend line". Here is one of their charts:

Figure 3 (e) probability of observing at least 1 negative linear trend of the given length over the period 1993–2050 (see Methods). The red dashed line in both (d) and (e) represents the negative 11-year trend that was observed in GISTEMP from 2002 to 2013. All trends were calculated with an ordinary least squares procedure. Source: Brown15, and with lines to probability added by me.

They found that the longest period of no trend in surface temperature, before it would be considered outside the different percentiles of model projections was:


RCP6.0: 13  years
RCP8.5: 8 years


RCP6.0: 17  years
RCP8.5: 11 years

The authors report at 97.5% confidence:
However, a negative trend of 11 years matches the 2.5% Cancellation Timescale and exceeds the 5% Cancellation Timescale (8 years) associated with the RCP 8.5 forced signal, indicating that we may be able to reject this rate of forced warming at the 97.5% confidence level.

And at the 99% level they write a proviso:
For the RCP 8.5, 6.0, and 4.5 forced signals, the 0.5% Cancelation Timescales are 19 years, 27 years and 23 years respectively. This would indicate that the global warming hiatus would need to continue for 8–16 years beyond 2013 before it could be said with over 99% confidence that the true forced signal is not increasing as quickly as these CGCM-produced forced signals.
They did some other different analysis to check their "Cancellation Timescale".

A couple of problems with the analysis, as I see it

Now it's quite likely that I'm missing something here, but it seems to me that that Patrick Brown in the press release is claiming a lot more certainty than is warranted. Here again is what he is quoted as saying:
"Based on our analysis, a middle-of-the-road warming scenario is more likely, at least for now," said Patrick T. Brown, a doctoral student in climatology at Duke University's Nicholas School of the Environment. "But this could change."

I'd say Patrick is drawing that conclusion based on a flawed analysis of the global mean surface temperatures from 1993 to 2013. Flawed because all the successive trends he computed are overly influenced by the mean temperature in 2013. Add in 2014 and he'd get a different result. Flawed also because up until around now, the difference between greenhouse gas concentrations for each RCP is minimal. They haven't started diverging yet to any great extent.

So have at it in the comments, if you want to. In any case, adding another year does show that their "flat" or "negative" curves turned positive in 2014. And this year is shaping up as another warm one, so far.

From the WUWT comments

As you can imagine, the comments from the deniers at WUWT are the usual mix of "it's not warming", "scientists don't know nuffin'", "it's all a hoax" and "it won't be bad". These are interspersed with various strawmen claiming the science says what it doesn't.

MarkW doesn't understand the difference between short term natural variability, such as from ENSO events, and the swamping of all other factors over the longer term. What he was told was correct. Over the medium to longer term, greenhouse warming is swamping all other factors.
April 21, 2015 at 4:28 pm
It was only a few years ago we were being told that CO2 was so strong that it completely swamped all other factors.

Pat Ch wrongly thinks that global surface temperatures haven't risen in 18 years. He's been reading too much WUWT and no actual climate science.
April 21, 2015 at 7:18 pm
The hiatus was not modeled in either middle of the road models or any other. To assign it a value now as if it was is absurd. To pretend it is 11 years instead of 18 is childish. Do you think climate is aware of the common calendar? This is babble. An attempt to portray scientific disaster as a mere statistical misunderstanding .
Below is a chart showing how much the temperature has risen in the past 18 years. I've taken 18 years ago as 1996 and drawn a line across for the moving average for that year.

Data Source: GISS NASA

Doug S probably triple or quadruple locks all his doors and windows at night, guarding against an invasion of the USA by hordes of middle east militants. He's nuts!
April 21, 2015 at 8:55 pm
People should consider the threat of slow, decade change due to sea level rise against the threat of Islamists storming into your town on a Sunday afternoon and decapitating and raping your daughter, granddaughter, son and wife. Perhaps you’ll be left alive to contemplate the danger of climate change and 2 degree Celsius rise in a hundred years.
It’s a matter of common sense and perspective. Religious believers become fixated on a subject and loose the ability to think and act rationally.

joelobryan mutters something irrelevant about Hollywood films:
April 21, 2015 at 9:10 pm
This paper champions the fact that the Null Hypothesis of naturally changing climate cannot be rejected when real-world data is considered. The Null hypothesis is only rejected by those who accept a programmed CG simulation similar to a HollyWood superhuman movie of CG animations.

David Cage is a conspiracy theorist of the "it's all a hoax" type, claiming fraud and fakery:
April 21, 2015 at 11:23 pm
They do not mention the possibility that the climate scientists fiddled the results to match the previous variability when the supposed warming existed. Could it also be that there is not forcing caused by CO2 resulting in temperature changes but that CO2 is the result of temperature changes so instead of forcing there is negative feedback and a fundamentally stable system as the signal analysis would suggest?

References and further reading

Patrick T. Brown, Wenhong Li, Eugene C. Cordero and Steven A. Mauget. "Comparing the Model-Simulated Global Warming Signal to Observations Using Empirical Estimates of Unforced Noise." Scientific Reports, April 21, 2015 DOI: 10.1038/srep09957 (open access)

Van Vuuren, Detlef P., Jae Edmonds, Mikiko Kainuma, Keywan Riahi, Allison Thomson, Kathy Hibbard, George C. Hurtt et al. "The representative concentration pathways: an overview." Climatic Change 109 (2011): 5-31. DOI 10.1007/s10584-011-0148-z (open access)

Climate capers at Cato - by Grant McDermott

HotWhopper: A silly poster by Pat'n Chip at AGU14 provides a learning experience


  1. Hot Whopper regulars will no doubt recall there was a spate of papers that came out in the 2013 - 2014 timeframe that used Bayesian analysis to argue for a very low transient climate sensitivity (TCS). As Sou alludes to above, this is because surface temps have been suppressed lately due to the natural variability inherent in the system, and predominately due to the largely prevailing La Niña conditions over the past decade or so.

    Well, all that excess heat that's been trapped in the oceans over the past few decades due to the anthropocene greenhouse effect has got to emerge sooner or later, and 2014 was probably just the start of it. I dare... no I *double dare* folks like Nic Lewis to repeat their analyses in 2016 once the figures for 2015 are in. How much you wanna bet that all we'll get is [*crickets*]? :-)

  2. I will have a look at the paper.

    From memory the different emission scenarios are not expected to diverge very quickly initially due to the logarithmic respond of CO2 forcing to CO2 concentrations. So the uncertainty in the CO2 ECS dominates which warming pathway you end up on initially.

    But by the time they do diverge it's too late as the recovery time after reducing emissions varies a lot depending on the emission scenario. So waiting to see what the actual CO2 ECS is instead of picking a lower emission pathway now is dangerous.

    1. In the abstract they say the hiatus is still consistent with a middle range ECS, but probably rules out a high range ECS. So the results don't, say, rule out a middle range ECS therefore leaving a low range ECS as most likely

      "The empirical EUN also indicates that the reduced GMT warming over the past decade or so is still consistent with a middle emission scenario’s forced signal, but is likely inconsistent with the steepest emission scenario’s forced signal."

    2. Sou,

      I don't think the paper is flawed, I am happy to accept their conclusion stated in the abstract.

      Maybe them using RCP to categorise forced signals is a bit misleading as the RCPs are based on rates of CO2 concentration which are well-known? I interpret their comment about RCP 8.5 as meaning the ECS is less that which emerges from the CMIP5 RCP 8.5 model projections.

      What I suspect they did was look at a lot of different time intervals depending on what looked interesting. I am impressed they covered so much in one paper!

    3. Harry, can you explain to me how they concluded that the last few years are likely inconsistent with RCP8.5 but likely not inconsistent with RCP6.0, when there is virtually no difference between the atmospheric concentration of GHGs at present between the two RCPs.

      Seems to me if it's consistent with one, it's got to be consistent with the other.

    4. I can't. I did try. But they cover a lot of stuff.

      They appear to be talking about the model ensembles when they refer to the RCPs, not the CO2 concentration scenarios. Eg when they say RCP 6.0 they are talking about the CMIP5 model ensemble for the 6.0 W/m2 forcing by 2100 which is called the AIM model elsewhere.

      Maybe the Real Climate people will look at it, they seem able to get to the heart of complex studies.

    5. Harry, yes they are talking about model runs, but the runs are tied to the representative concentration pathways. The way I read it is that the researchers are not arguing that some models are better than others. What they are claiming is that we are on the RCP6.0 pathway.

      You've got me thinking - it raises the question of emissions and atmospheric concentration of GHGs, which the paper didn't discuss.

      You could be on the right track in a way. They might not have been talking about the path we are taking in terms of emissions and the impact. Rather they might have been arguing that the models don't match the pathway and are "running hot" without actually saying so. I don't think the evidence they came up with shows that - either way. (There are too many inconsistencies in the paper IMO.)

    6. I looked up some of the references and what the lead author said in news articles - he does seem to say he considers RCP 6.o likely and RCP 8.5 not likely based on current observations. But how is that possible given the limited amount of time to allow the RCPs to diverge?

      I did a quick read and yes, the whole point of the models is they are standardised except for the parameters they want to vary - this makes sense. This implies they should not have diverge much yet.

      The IPCC AR5 reports say the different RCPs do not diverge until later in the century.

      I am going to watch his videos on youtube. If I still can't work it out I am going to email him and see what he says.

    7. Harry, I'm in the group that has recently made observations at Eli's, that thinks that increasing climate instability will feed back to cause wobbly economic outcomes wherein a representative concentration pathway of 60 W/m^2 in 2100 is more likely than the 8.5 alternative. I certainly hope so, because fulfillment of 8.5 would certainly spell the end for > 25% of biodiversity, and for anything but pre-industrial localised human communities.

      Having said that it's possible that fracturing economic conditions may simply remove restrictions on pollution emission and that rogue governments and industry may still have sufficient wherewithal to burn baby burn in an essentially business-as-usual scenario. If that does eventuate we really are stuffed.

      RCP6.0 is no picnic though, and with a central ECS value and by the time of peak emissions in future centuries this scenario is entirely able to bring about pretty much most of the destruction of civilisation and the biosphere that is certain with RPC8.5.

    8. Greg Laden has posted about the study on his blog. He has also made some comments on his facebook timeline


    9. This comment has been removed by the author.

    10. Sorry Greg Laden made the comments on a facebook group, not his timeline.

  3. "Based on our analysis, a middle-of-the-road warming scenario is more likely, at least for now ... But this could change" is just pap. This whole exercise smacks of grabbing a final few hairs from the tail of the hiatus as it disappears back into the undergrowth.

    I don't expect to see many charts displayed at WUWT that extend beyond 2013. Maybe Antarctic sea-ice, but even that seems to be bumping up against a ceiling.

  4. In any case, you gotta love the term "Envelope of Unforced Noise". It's right up there with photography's "Circle of Confusion" as a contender for Best Scientific Name for Band.

  5. What is interesting is where it is now, just ~4 months later. The current 8-month anomaly on GISS is .755C. Right now it's possible the yearly rate of warming for 2015 could hit .15C to .2C - one year.

    Why is the surface warming so rapidly? Wasn't the cool phase of the PDO supposed to last for decades?

    1. JCH - I was having a conversation about PDO on another blog earlier today. Most likely the cool PDO epoch is not over, as there is much variability within each. The current cold phase has been in effect since approx 1999, so if it were to end now it would be extraordinarily short. Typically the phases last 20+ to 30 years, so it is more likely we are seeing a shorter +PDO spell in the midst of a -PDO epoch. This link is a great resource for historical PDO values, with interactive graph.

  6. Sou

    I'm inclined to agree with your view - this may be one of those studies which is overly sensitive to the recent slowdown in the rate of surface warming. Since the rate of surface warming appears to be modulated by variability in the rate of ocean heat uptake it's arguable that the recent short-term slowing will not have much (perhaps any) effect on TCR (as formally defined), never mind ECS.

    Since the surface / troposphere is *not* the climate system as a whole, only a small part of it (the majority being ocean), the strength of conclusions based only on short term variability in surface warming is debatable.

  7. Patrick Brown, the lead author, has just written about this paper at realclimate.org. Here is the link.

    Global warming and unforced variability: Clarifications on recent Duke study


Instead of commenting as "Anonymous", please comment using "Name/URL" and your name, initials or pseudonym or whatever. You can leave the "URL" box blank. This isn't mandatory. You can also sign in using your Google ID, Wordpress ID etc as indicated. NOTE: Some Wordpress users are having trouble signing in. If that's you, try signing in using Name/URL. Details here.

Click here to read the HotWhopper comment policy.