Upside Down Proxies in the IPCC AR5

A recent tweet:

Led me to yet another one of those articles which likes to pretend critics of the (in)famous Hockey Stick are just obsessed with one study published twenty years ago, hoping if they can prove it was fraudulent, something, something, something despite all the later work confirming the Hockey Stick was right.

That, of course, is complete drivel. Critics of the Hockey Stick have long criticized many temperature reconstructions and argued the later ones do not validate the original Hockey Stick. In fact, a common point many people have made is that the later work which supposedly confirms the original often uses data from the original. In fact, many later studies took output from the original one and used it as input for their analysis.

(It turns out if you use the results of a study in your calculations, your calculations may well match those of the original study. Who knew?)

Anyway, that's not what I wanted to look at today. That bugs me, but the article made me look at something which I hadn't thought about in some time. When I did a little digging into it, I discovered there was information I had never seen before. It's somewhat interesting.

To make a long story short, in 2008 and 2009, Michael Mann published temperature reconstructions using four proxies taken from Tiljander et al. (2003). He claimed these reconstructions rebutted critics who said his original Hockey Stick was dependent entirely upon a small amount of tree ring data as his new reconstructions were robust to the removal of all tree ring data. It turns out that claim was only true if he used the Tiljander proxies. If not, his conclusions were unsupportable.

This is important as Mann should never have used those four Tiljander proxies. To start, two of the Tiljander series are just derivatives of the other two created by applying some simple arithmetic to the other two series. You can start with two series, do a little multiplication and subtraction and wind up with two new series, but that doesn't mean the two new series have any new information in them.

Double-counting aside, Mann's reconstructions used two different methodologies (EIV and CPS) which both required proxies be calibrated against the modern record. The Tiljander series were taken from a lake where human activity disrupted the area a couple centuries ago. As a result, the people who created the series warned people the modern portion could not be used to represent temperatures. That means the Tiljander proxies could not be calibrated against the modern temperature record. Mann went ahead and did it anyway.

Partially as a result of this, and partially because Mann's methodologies were terrible, some of the Tiljander proxies got used "upside down" in some cases. Whether they were used right side up or upside down depended on the proxy, the reconstruction and the time period in question. Yes, Mann's methodology allowed him to do things like use proxies rightside up from 1000-1500AD then upside down after that.

There's a lot more which could be said about that, but it's a huge mess and I've discussed it all before. For now, hopefully that introduction will help you understand why I chuckled when I saw the article link to this figure taken from the Intergovernmental Panel on Climate Change's latest assessment report:

Versions of Mann's reconstructions show up about ten times in this one figure. I was curious what the report had to say about the Tiljander proxies being misused so I did some digging, and that's when I discovered the IPCC has posted explanations of how figures were created and the data used for them. I didn't know that before, and I have to say, I think it's a really good thing for the IPCC to do. They should be praised for it.

There's a lot of information/data to look through, but a couple things jumped out at me which I thought were worth highlighting. First, from the explanation for the figure shown above:

Replaced Mann et al. (2008) EIV-based reconstruction of NH land+ocean temperatures based on their full
network of proxy data, with one from Figure S8b of their supplementary information that used the full
network minus 7 potentially problematic proxy records (labelled "Ma08min7eivf" on the figure; labels
defined in the appendix Table 5.A.6). The 7 dropped records include the 4 from Tiljander et al. (2003)
which have non-climatic disturbance apparent in the most recent centuries

I think it is cheeky to refer to series as "potentially problematic" when it is known the Tiljander series cannot be calibrated to the modern temperature record, but still, this is... somewhat good? I guess? The IPCC used multiple reconstructions from Mann's 2008 and 2009 studies, all of which used the Tiljander proxies. That makes it hard to praise them for deciding to replace one of the reconstructions with a version that didn't use the Tiljander proxies.

More importantly, look at what the explanation goes on to say:

For the remaining Mann et al. (2008; 2009) series, the situation concerning the Tiljander et al. (2003)
records is as follows: (a) NH Ma08cpsl: Mann et al. (2008) SI Fig. S8a shows effect is negligible. (b) NH
Ma09regm: Mann et al. (2009) SI Fig. S8 shows effect is small after 900 CE, and before 900 CE the
version with these series removed is mostly cooler than the version used here. (c) SH Ma08cpls: should
be no effect because these records are in the NH. (d) Global Ma08eivf: graph and data on the following
webpage of Mann's shows that the effect is small, but I have not used this data as a replacement because
this page seems to be an unofficial supplementary information page related to these publications but not
formally published: (e)
NH Ma08eivl, SH Ma08eivl, SH Ma08eivf, Global Ma08eivl: Mann et al. have not estimated the effect of
the problematic records on these series.

There are a number of issues with this, including errors, but discussing that would take some time. What I find more interesting is the IPCC chose to use all these reconstructions knowing they used data which is known to be inappropriate. What's even more interesting is this entry in the explanation of another figure:

Mann et al. (2008)EIV “full” (=land+sst) reconstruction with the exclusion of 7 potentially problematic
proxies (4 of which are from Tiljander et al. (2003) sediment cores) could not be used because
uncertainty range data were not available, so the version that included these 7 proxies was used.

That logic is fascinating. The IPCC used the version of a reconstruction which includes data known to be corrupted because they didn't have uncertainty levels on the version of the reconstruction with the corrupted data excluded.

Personally, I think if you don't know how uncertain results are when corrupted data is excluded, you probably won't be able to figure out how uncertain they are by adding that bad data in. Apparently the IPCC disagrees.

(As a final note, the Tiljander proxies were not the only inappropriate series used in these reconstructions. They weren't even the only series used upside down. A number of others exist which were not excluded in any version of these reconstructions.)


  1. I'm guessing you mean of the chapter in the IPCC report (Working Group 1, Chapter 5) that figure is from? The heads of that chapter were two people I'm not particularly familiar with, Valérie Masson-Delmotte and Michael Schulz. If I remember correctly, they both work on paleoclimate timespans much larger than the hockey stick (~100,000 years). I could be wrong. They're not anyone I'm familiar with.

    If you're wanting to know about the lead author of something else, I'm afraid I don't know what you're referring to. I don't think the recipes say who they're written by. Which now that I think about it, is weird. How do you write a document using the pronoun "I" if you don't have your name attached to it?

  2. Such gyrations give an appearance of diligence and legitimacy. Except that it turns out to be a laughable exercise in cobbling together whatever flimsy scraps of information that can be found to support a predetermined conclusion. Nice find. It smells like the efforts of an unskilled undergraduate writing a paper on a topic of which he has no understanding, hoping the professor (or teaching assistant grading it) will be bamboozled by the number of words and not their meaning.

  3. MikeN, yup. At the time I wrote this post, I considered discussing that but couldn't remember the details that well. I planned on commenting on it in a future post as one of those "recipes" says issues with the Tiljanders proxies shouldn't affect the southern hemispheric reconstructions. That statement is true. However, I think what matters is not what should happen but what actually does happen. As nonsensical as it may be, the Mann's reconstruction for the southern hemisphere is dominated by proxies from the northern hemisphere.

    But hey, why would researchers at the IPCC know this? It's not like they have thousands of people reviewing their reports who could be even slightly familiar with what the papers they use actually do.

  4. Hi Brandon, I really appreciate the “digging” that you do, the knowledge you pass on, and look forward to your posts. Tip Jar time again! Thanks GFO

  5. Thanks! I apppreciate it!

    I still have a couple more posts to write about correlation and what it means, but these last couple posts have refreshed my interest in discussing paleoclimate issues. I'll probably start writing about them again as there is a lot of content I've never gotten through, and the IPCC has started getting things set up for its Sixth Assessment Report. Re-visiting how multiproxy reconstructions were handled in the past may be interesting for considering how the new report handles them.

  6. By the way, I should point out part of why I want to start writing about paleoclimate again is in the last week, I've re-visited a lot of old issues raised/covered by Steve McIntyre over the years, and I think it's interesting how so many of them have never been resolved. People like Kevin Anchukaitis may scoff at how people still discuss Michael Mann's original hockey stick, but the reality is that study is still highly relevant.

    For instance, one of the central issues in paleoclimatology is figuring out what data is a proxy for what variables (temperature, moisture, etc.). A significant amount of data used 20 years ago is still being used now, so understanding what that data was used instead of other data 20 years ago helps us understand why it is still used today. Which is part of why I can't feel sympathy for scientists when people try to obtain their e-mails. The field of paleoclimatology has consistently failed to offer any meaningful, much less forthright, explanation for the choices of what data to use/not use, or even to archive/not archive. E-mails can tell us a lot about those decisions.

    I happen to have access to far more e-mails than the general population. Reading them does shed light on that issue. I suspect the more e-mails people gain access to, the more they will be able to understand why multiproxy temperature reconstructions rely on the data they rely upon.

Leave a Reply

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