Monday, June 09, 2014

A Hockey Stick Question

There is an interesting argument I have seen with regard to Mann's hockey stick reconstruction of past temperatures. I am not sure if it is correct. The hockey stick has been the subject of a lot of heated controversy pro and con and I am not in a mood at the moment to try to wade through it, so I thought I would sketch the argument and see if anyone here can give me a good reason to believe that it is either true or false. 

It goes as follows:

Most of the reconstruction is based on proxies for temperature such as tree rings, since nobody was producing reliable instrumental data in 1500 A.D. The recent part, on the other hand, is based on instrumental data. The claim I have seen is that Mann's graph shows proxies up to one date, proxies and instrumental up to another date, instrumental only starting sometime around 1960-80. That fits what sense I can make of the graph in his paper, the exact date at which the proxies disappear being hard to make out. The further claim, and the critical part, is that if you run the proxies up all the way to the present they do not show as sharp a rise as the instrumental data.

The instrumental data is a more reliable source of information than the proxies for the period when we have it, so if the only objective is reconstructing what has happened to global temperatures over the past six hundred years or so it makes sense to use it where available. But part of the point of Mann's paper is that the recent temperature rise is unprecedented. To show that he needs an apples to apples comparison, a comparison between the proxy record for the past and the proxy record for the recent period of rising temperature or a comparison between the instrumental record for the past and the instrumental record for the recent period. The latter is impossible, since there is no instrumental record for most of the period he is looking at, but the former is not. 

If the proxy record for the past century fails to show the rapid temperature rise in the instrumental record, that is evidence that the proxy record is an unreliable source of information, in particular unreliable as a way of detecting rapid rises. And it is probably more unreliable the farther back we go, since the farther back we go the less relevant data is available. If so, the fact that the proxy record shows no rapid rises in the past may not mean that they did not happen.

Can anyone tell me if this argument is correct? Does the proxy record for the second half of the 20th century show a substantially milder rise in global temperature than the instrumental record?


John said...

I've had the same question.

There are statistical techniques to account for measurement error (I find it easiest with a Bayesian approach, but whatever). For this, you might have to a Kalman filter where the different data is merely evidence of some underlying process that you can only really observe recently.

Daublin said...

My understanding is that tree rings are not managing to proxy temperature at all. There are too many other variables that impact a tree's growth, for example precipitation.

EH said...

Yes, this critique is entirely correct; although not specific to the hockey stick controversy per se. It is the same issue that motivated the 'hiding of the decline', for instance.

The underlying assumption of paleoclimatology is that temperature has a strong relationship to certain measurable 'proxies'. That is a nice idea, but when it hits the wall hard, as has happened with many tree ring chronologies, it will result in some fallout.

'hiding the decline' was a willful attempt to delay this realization from going public. The hockey stick arises from the same discrepancy in a more subtle manner. Here mere incompetence rather than bad intent may have been the original driver; although the unwillingness of those involved to own up to basic errors kindof makes it part of the same story.

jonabbey said...

My understanding is that the 'hockey stick' was based on a fairly broad collection of proxies, and that the tree ring proxy has a known problem in certain environmental circumstances, including that pertaining in recent years.

But you'd have to look to the primary literature for the whole story, I'm sure.

Glen said...

Your argument is correct. Tree ring proxies in particular "diverge" - the trend on the sorts of series that Mann used does the opposite of the instrumental record - the trend there goes DOWN in the latter half of the last century.

Craig Loehle explains some of the problems with tree rings as a temperature proxy in this CA blog post:

Realizing that all of the "shaft" in the hockey stick came from tree rings, Loehle therefore did a study that used JUST proxies that WEREN'T based on tree rings (and didn't tack on or average in the instrumental record). He got a curve that looks like those on this page:

(The most relevant paper of his is: Loehle, C. and Hu McCulloch. 2008. Correction to: A 2000 Year Global Temperature Reconstruction based on Non-Treering Proxy Data. Energy & Environment 19:93-100 )

In short, when using actually defensible reconstruction methods - comparing proxy data to proxy data and excluding known-unreliable proxies, current temperatures are not clearly "unprecedented" in human history - it could well have been warmer than today during the MWP.

(We can say for certain that it WAS warmer than today in recent memory in a great many specific locations just from looking at the movement of tree lines, but we can't say for certain that it was warmer overall in any particular year or year range because the error bands are too large)

Glen said...

Oh, and the best Loehle published paper explaining the divergence problem wrt tree rings is probably: Loehle, C. 2009. A Mathematical Analysis of the Divergence Problem in Dendroclimatology. Climatic Change 94:233-245

dWj said...

One of the basic properties of linear regression etc. is that the predicted value of the y variable has less variance than the actual y variable; the ratio is R^2. The less well you can measure y, the less variable will be the best estimate you can give to y based on your measurements. I wonder whether that might play into this as well.

Anonymous said...

i always remember scottish medieval vineyards and even romano british vine production when thinking about proxies. greenland colonization also comes to mind. anyone know how reliable it is as data?

jonabbey said...

There are plenty of anecdotes about the MWP, but they don't provide evidence for a higher _global_ average temperature.

David Friedman said...

This post was echoed to Google+ and picked up some interesting comments. One of the most interesting was a link to an interview with Richard Mullen about the NRC report on Mann's work.

That report is commonly asserted to have validated the Hockey Stick. According to Mullen, who was a referee for the report, that is not true. The conclusion of the report was that Mann's conclusions were valid for the past 300 years, but not new, and that his conclusions for the earlier period, which were new, were not supported.

The relevant part of the interview is in the first five minutes or so—interesting stuff.

Josh Sacks said...

There are multiple, independent lines of reasoning which invalidate the Mann's work. The argument you state is not one of them.

The original and most complicated argument (originated by Steve McIntyre of ClimateAudit fame) against the hockey stick has to do with the selection algorithm used to generate it.

The Hockey stick does NOT use all available data. As you mention, there are questions about how good some of the temperature "proxies" are. To fix this, Mann filtered the input set of proxies attempting to only capture data sets which accurately "proxy" temperature. How did he do this? We have accurate knowledge of temperature for the last 50 years. Mann "screened" the proxies by correlating them against known temperature over the recent past. Mann only used proxies which seemed to correlate well.

Seems reasonable, but if you think about it, it's a completely flawed process, almost guaranteed to generate a hockey stick.

Imagine that all the input proxies were in fact random walks with no actual data content. What does Mann's process do? Well, it selects a handful of proxies which randomly happen to correlate to modern temperature. The last 50 years shows increasing temperatures, and we know that the end result must approximately match the real temperature for this time period (since the selected proxies are correlated with real temperature over this period). Prior to the correlation period, we'd expect the summation of random walks to be essentially flat over time.

This is EXACTLY what Mann produced. A long flat line, followed by a sharp spike over the past 50 years.

Mann did NOT include actual thermometer data in his work. Rather he SELECTED proxies which correlated with actual thermometers- a process which fundamentally biased the results.

Josh Sacks said...

There are other arguments against Mann's work as well. Most of these are critiques of the underlying proxies such as "strip-bark" which has known issues as may be more of a CO2 or drought proxy than temperature.

Other critiques involve the statistical validity of the proxy data- ClimateAudit has shown how some of the tree proxies depend on as few as 2-3 trees for some time periods and that these trees have enormous influence on the proxy data and overall hockey stick graph. These proxies are also very sensitive to which sets of data you use- often including more data fundamentally changes the shape of the proxy curve, implying these are not particularly reliable proxies.

Other funny critiques are pointing out that sometimes the hockey-stick analysis uses the input series "upside-down". That is they interpret the proxy signal as being "warm" when the original proxy author interpreted it as "cold" and vice-versa.

Bishop Hill said...

The essence of your argument is correct, but there are a few details that are not. I will not bother with these here.

In Mann's dataset, the tree ring series which suggested unprecedented warmth were mainly the bristlecone pines plus a few others. In the literature at the time, it was generally agreed that the bristlecones were unreliable as temperature proxies because they were contaminated with a non-climatic signal.

Mann's algorithm was biased, because it picked up these unrepresentative proxies and overemphasised them in the final result.

I can recommend a good book on the subject. :-)

Commenters need to be careful about "hide the decline". This referred to a Keith Briffa proxy series which was not used in the Hockey Stick itself.

J.E.S said...

David: SOmewhere between the 10 and eleven minute mark in the video, Mullen admits to having been funded, at least in part, by the Koch brothers.

Economists love to say that "incentives matter." Well, how about the incentive for a cash-strapped scientist to tell the story that his rich, free-market doctrinaire funders want him to tell?

Anonymous said...

The interesting thing is that hockey sticks abound in proxies that have no tree rings whatsoever.

Makes the attacks on Mann and his work completely irrelevant.

Glen said...

Anonymous claims: "The interesting thing is that hockey sticks abound in proxies that have no tree rings whatsoever."

Yes, that would be interesting if it were true, but as far as I know it's not.

See, the thing the RealClimate gang and their ilk don't seem to grasp is that to skeptics, what makes Mann's hockey stick count as "a hockey stick" is not so much the steepness of the "blade" as how straight and long the "shaft" is. So it's not sufficient to merely find a chart that upticks at the end! Even if you find a proxy that has a strong uptick at the end WITHOUT (as is often done) lazily pasting in instrumental values or using a "validation" process to throw out or "flip" all samples that don't end in an uptick, the resulting chart STILL fails for our purposes to count as "a hockey stick" if:

(a) it goes back <400 years so all it's telling us is that things have gotten warmer since the Little Ice Age (something everybody already knew and agreed with before Mann came along).


(b) the "shaft" is wobbly in a way that shows lots of significant natural variation prior to the last century, often including an MWP that roughly matches recent according-to-proxy temperatures and a hefty LIA between the MWP and today.

What made Mann's work matter was that he claimed to show with high "skill" that temperature trends had been amazingly steadily flat for over a THOUSAND YEARS prior to the last 50 in which they leapt up to levels previously "unprecedented".

So if you're thinking of that RealClimate post (or the SkepticalScience page or Tamino posts that echoed it) claiming to show a half-dozen other "hockey stick" trends: none of those fit the bill as "confirming the hockey stick". Most fail because they don't go far enough back to tell us anything interesting, a few fail because they show too much variation to be a non-BROKEN hockey stick (eg: Moberg's or Loehle's sticks are broken in two places), and others fail because they combine two types of data inappropriately or continue to reuse the exact same data-snooped series we already knew the shape of before.

At least, that's my impression. I think when you said hockey sticks "abound" what you really meant was "proxy series that don't go back very far and have some sort of uptick at the end abound", which isn't quite the same thing. But please let me know if I'm wrong about this. (Also: please use a name other than "Anonymous", as doing so makes it hard to have conversations!)

Tibor said...

J.E.S.: Well, while this would potentially give him an axe to grind, one could say the same about pretty much any research that is not funded by the researcher himself (and there is not much of that). There is always someone "behind it". If a research is tax funded one could claim "they just show what the government wants them to show". If it is a private contributor with some known opinions, you can say it again. While it might be a reason for a bit more caution, it is not a "licence" to reject something a priori.

An example:

Someone I discussed the issue about (who diasgreed with me but was very reasonable with good arguments and fun to talk with) sent me an article about minimum wage from the "Centre for economic and policy research"


Since CEPR is to my knowledge basically a socialist/progressive analogue of the CATO institute (they, of course, have some minimum wage articles too), it is obvious that they have an axe to grind. But that does not mean (necessarily) they are trying to mislead or lie. They may focus on stuff that supports their political beliefs and miss some other whereas their critics will do the opposite and give more detailed account of that other stuff. If someone is deliberately dishonest, well, that is a whole different story, but you cannot derive that from just knowing who funds the guys.

Anonymous said...

Glen, what's the criticality of the many hockey sticks to our understanding of the climate system and our impacts upon it?

Foundational, or merely an interesting curiosity?

Glen said...

Anon: I'm not sure I can answer that question unless you give me a reference so I know what you mean by "the many hockey sticks". I've explained why some of those I'm *guessing* you might have been thinking of don't tell us much of interest, but you haven't confirmed that my guess was correct, so I don't know what your premises are.

Do you just mean the reconstructions that were based on tree rings? Do you mean some hypothetical others that weren't which you claimed "abound" but you haven't actually described, named, or given a link to? Are you including alleged "hockey sticks" that aren't even from temperature proxies?

And while you're answering that, perhaps you could add: whatever this group is that you are thinking of, what's its significance to YOU? If the attribute "hockey-stickness" doesn't matter at all to our understanding of the science why are you trying to rescue it? Why not just admit that Mann screwed up? Why squint and pretend to see "hockey sticks" elsewhere so you can defend his conclusions as "plausible" however flawed the underlying science was?

Anonymous said...

One of these days the pseudo-skeptics might abandon attacking MBH'98 and thinking they've gutted the edifice of AGW and it will come crashing down and so AGW ceases to exist and we can think about other things.

In the meantime, we don't need MBH'98 and its verifications and analogues to know that we are the cause of the observed changes in the climate system. Again, it far more an intellectual curiosity and a factoid than *the* foundation of all climate science.

Glen said...

Anon: So are you now admitting that truly HS-shaped proxies don't "abound"? Were you just making that up when you said it before?

Of course the hockey stick is not "*the* foundation of all climate science". Nobody said it was. Its significance is largely rhetorical.

Climate science doesn't rely on any of the MBH curves, but the claim of "unprecedentedness" does. If you were to believe that MBH99 is essentially a correct summary of what happened to the temperature trend, this might lead you to believe two important corollaries:

(1) short-term "natural variation" has been demonstrably insignificant in magnitude and couldn't possibly have been enough by itself to explain much of the positive trend seen in the last half of the 1900s.

(2) Recent temperatures are "unprecedented" during the era for which humanity has significant written records.

If you believe #1, you're not going to look very hard for "natural" trends that might help explain the data. If you believe #2, you're not going to look very hard at past historical records from warmer times to estimate how "bad" climate change might become in the near future.

If Loehle's or Moberg's curves are more correct than MBH99 was, then current temperatures are NOT "unprecedented" - we are currently within measurement error of where things were about a thousand years ago. We're not in "uncharted waters"; there's no reason to predict imminent horrible environmental effects never seen before that man can't adapt to because we adapted to it just fine last time. There's also less reason to predict "runaway"-type feedback effects - if they didn't show up in the past they are unlikely to do so now.

So that's why MBH matters to skeptics. An additional reason MBH matters to alarmists is that alarmists are trying to maintain an aura of infallibility. If they admit that MBH was based on bad statistics but kept around for its propaganda value, people have to wonder what other claims in climate science might have been similarly motivated.

Anonymous said...

I get it, Glen - the beef you have isn't really with hockey sticks at all. It's all about the potential policies to mitigate AGW that you feel are unnecessary because you think the MBHs and their verifications and replications rest on shaky ground. There's a site called climateaudit that specializes in going after proxy reconstructions. Not terribly successfully, I might add.

As is often the case, pseudo-skeptics are putting the policy cart before the science horse. Unfortunate.

Tibor said...


I noticed you discussing this on facebook with a couple of people on "climate change discussion" page. Is it worth following that page? From what I noticed, most of the people there were not very reasonable (regardless of their opinions on climate change) and most of their arguments were ad hominem accusations. Is that an exception or a rule there? And if it is a rule, are the occasional good exceptions good enough to offset all the nonsense? Thanks

Unknown said...

David Friedman said...


The critical question is whether what Mullen said was true. Googling around, I found the following:

We roughly agree with the substance of their findings,” says Gerald North, the committee’s chair and a climate scientist at Texas A&M University in College Station. In particular, he says, the committee has a “high level of confidence” that the second half of the twentieth century was warmer than any other period in the past four centuries. But, he adds, claims for the earlier period covered by the study, from AD 900 to 1600, are less certain. This earlier period is particularly important because global-warming sceptics claim that the current warming trend is a rebound from a ‘little ice age’ around 1600. Overall, the committee thought the temperature reconstructions from that era had only a two-to-one chance of being right.

That fits what Mullen said, except that he thought it was three hundred years rather than four hundred.

J.E.S. mentions funding from Koch, doesn't mention that Koch was in a list of funders that also included Bill Gates.