A new “analysis” craze has been sweeping the climate contrarian world of late. Not content with cherrypicking one relatively cool La Nina year (even though 2008 was still in the top 10 in the instrumental record), some of the usual suspects are now using higher-order polynomial fitting in an attempt to illustrate a severe downward “trend” in global temperature.
Chief among them (what a surprise) is Alberta’s very own “fact chucker”, National Post columnist Lorne Gunter (you read that right – LG will not rest until all facts are distorted or removed as seen in recent posts here and here). Roy Spencer (of UAH satellite-derived tropospheric temperature fame) has also jumped on the polynomial bandwagon, and now superimposes a polynomial curve on his monthly update of the UAH record. [Update May 15: Spencer no longer uses the fourth-order polynomial curve in his monthly update]. Can a “scientific” conference on the subject, perhaps jointly sponsored by the Heartland and Fraser institutes, be far behind?
Now it turns out that Gunter, who is mathematically challenged to say the least, has most likely been relying on the wisdom and Microsoft Excel skills of his fellow Albertan, oil industry insider and engineer Allan MacRae (or Allan M.R. MacRae, as he most often styles himself). MacRae is a minor but fast rising star in the contrarian firmament, as we shall see.
I first stumbled across the Gunter curve in a small commentary by Chris Horner at GlobalWarming.org that claimed the Earth has “cooled since George Bush took office.” A barely discernable National Post credit on the accompanying illustration led me to Gunter’s October 20, 2008 column entitled “Thirty years of warmer temperatures go poof”. And there it was in all its awesome majestic stupidity:
The only data set presented was the University of Alabama (UAH) satellite-based lower troposphere estimate, which not so coincidentally, shows less warming than the other commonly cited global temperature data sets. (See some general discussion in the first part of this recent post).
I quickly realized that the “trend line” was a higher-order polynomial, and recalled that Excel had just such a graphing functionality built in (I had often wondered why anyone would use it). Within a few minutes I had reproduced a reasonable facsimile of the graph, using Excel’s sixth-order polynomial trendline, and then just for fun I created the animation that graces the top of this post. It cycles from the usual linear trend line through the various polynomial degrees, each pointing more severely downward at the end point.
And there I left it until a few weeks ago. I had been intending to post on this, but quickly realized that the whole matter had been discussed at length elsewhere. It turned out that Gunter’s column had been posted on WattsUpWithThat.com and had received a thorough going over at Greenfyre’s (which means I need not take on that thankless task again – been there, done that).
Far be it from me, then, to dwell too long on the obvious, but the following points do need to be made:
- By any reasonable long-term measure, global warming has not “stopped” or “disappeared” or “gone poof”, as explained in Tamino’s excellent “What if …” post at Open Mind.
- There is no conceivable justification for fitting a polynomial trend to a temperature time series; it can only be construed as a blatant attempt to mislead by purporting to show a bogus downward trend that has “erased” global warming.
Of course such a curve will change greatly from month to month, at least as the end point is approached. The graph below illustrates this phenomenon.
Here we can see that the end portion has changed upward considerably since MacRae first published his polynomial curve – already a considerable amount of warming has reappeared! As well, as recently as early 2007, the curve was at its peak at the end point. Presumably that’s why this brilliant idea didn’t come up back then.
But where did the curve come from? Greenfyre was led astray by the cryptic credit to Andrew Barr, a National Post illustrator, and mistakenly thought the curve had been drawn by hand. But other sharp-eyed readers, had already concluded, as I did, that the curve had been done in Excel. Somehow, I can’t imagine Gunter (or Barr) downloading data, reformatting it and adding trend lines to it.
But Gunter explicitly stated that “the chart was not produced by Douglass and Christy” (mistakenly attributing the UAH data to David Douglass instead of Roy Spencer). So it was a mystery, until a combination of Googling and luck led me to a blog post by Allan MacRae at icecap.us, which featured this graph as its centrepiece.
This version showed the use of Excel more clearly (dark blue line graph with diamond points being the Excel default). More to the point, the exact same version of data (up to August, 2008) was used, even though Gunter had one more month at his disposal when he published. And MacRae’s comment that “ALL the global warming over the past three decades has disappeared!” eerily presages Gunter’s conclusion that “all of the rise in global temperatures since 1979 has disappeared.”
It turns out MacRae is not new at the climate disinformation game. In 2002, he wrote an anti-AGW piece along with Friends of Science “scientific advisors” Tim Patterson and Sallie Baliunas. “Climate Change is Natural” ran in the Canadian Globe and Mail newspaper on November 19, 2002 (posted at AlbertaResidentsLeague.com). The piece also ran, together with an ardent anti-Kyoto “rebuttal”, in the PEGG, a publication of the Association of Professional Engineers, Geolgists and Geophysicists of Alberta (APEGGA).
MacRae was identified by the Globe as “a professional engineer, investment banker and environmentalist”. As Frank Magazine sardonically observed at the time he “sounds like a real ecologist, eh?”
From ZoomInfo and other sources, I’ve pieced together a slightly more elaborate resume of MacRae’s “scientific research” in the Alberta oil patch going back twenty-five years:
- 1984-1996: Canadian Occidental Petroleum (including involvement “in the management of the Syncrude, OSLO and PCEJ oilsands projects” from 1984 to 1991, according to his 2007 Alberta royalty review submission).
- 1996-200?: Consultant, working “with a major US-based multinational on international oil acquisitions”
- 200?-200?: President and CEO of Odyssey Petroleum (see ZoomInfo cache)
- 200?-????: CEO of Aterra Energy Corporation (see FindArticles cache)
In the intervening years, MacRae wrote a few letters to the editor, listed here along with other writings. But in 2008, MacRae stepped back into the contrarian big leagues. A correspondence with Roy Spencer led to MacRae being acknowledged as a major inspiration for Spencer’s, um, unorthodox ocean-driven CO2 hypothesis, as outlined at WattsUpWithThat:
Even though I have been playing with the CO2 and global temperature data for about a year, it was the persistent queries from a Canadian engineer, Allan MacRae, who made me recently revisit this issue in more detail.
MacRae’s own blog post hypothesizing CO2 rise as an effect of rising temperature (rather than as a cause) appeared at icecap.us in February, 2008. That led eventually to full induction into Marc Morano’s infamous list of “international scientists” who “dissent over man-made global warming” (2009 edition):
Professional Engineer Allan M.R. MacRae of Alberta, Canada, authored a scientific analysis critical of man-made global warming in 2008. “The IPCC’s position that increased CO2 is the primary cause of global warming is not supported by the temperature data,” MacRae wrote on February 5, 2008. Variations in atmospheric CO2 concentration lag (occur after) variations in Earth’s Surface Temperature by ~9 months. The IPCC states that increasing atmospheric CO2 is the primary cause of global warming – in effect, the IPCC states that the future is causing the past. The IPCC’s core scientific conclusion is illogical and false,” MacRae explained.
Since MacRae has also identified such a recent steep downward trend in temperature, I suppose we’ll see a similarly steep decrease in atmospheric CO2 any month now. Or not.
Be that is it may, it does seem that Spencer’s new-found fondness for polynomial fitting the UAH temperature series was very likely inspired by MacRae. Spencer’s October 11, 2008 update at Climate Research Newsdid not have the polynomial curve, but in November, there it was. However, Spencer’s polynomial fit is only fourth-order, not sixth. Presumably this is considered somehow more sober and “scientific”, albeit at the cost of making less warming “disappear”. Spencer’s explanation of the curve (now at his own drroyspencer.com website) is interesting:
The smooth curve in the graph is a fourth-order polynomial fit to the data, which smooths out the large amount of monthly variability in the data and helps reveal the underlying ‘trends’. (There is no claim that this curve has any predictive power for the coming months or years.)
Of course, in the official UAH data archive the only trend published is the conventional linear least squares fit. It still stands at +0.13 deg C per decade, somewhat less than RSS and the surface data sets, but significantly positive. Surely its omission from Spencer’s monthly updated chart is an oversight that will be rectified shortly. Or perhaps not.
Spencer’s caveat about “predictive power” is apparently not shared by MacRae, who back in October threw all caution to the winds:
Furthermore, the best fit polynomial shows a strong declining trend. Are we seeing the beginning of a natural cooling cycle? YES.
Perhaps, then, this would be a good time to extend MacRae’s polynomial curve and compare its awesome predictive power to the simple linear trend that has been made to disappear. Poof, indeed!
Oops. Recent UAH observations are close to or even above the linear trend, while the polynomial curve appears to project another ice age within a few years.
Anyway, now that the true source of the polynomial curve has been revealed, perhaps the National Post would like to update the Gunter column with an appropriate credit to Allan MacRae. Credit where credit is due and all that. If journalistic professionalism or ethics are not motivation enough, Post editors should spare a thought to poor Andrew Barr, who has unfairly taken much of the blame for the statistical travesty.
After all, all the hapless illustrator did was pretty up an ugly graph. The phrase “putting lipstick on a pig” comes to mind, somehow.
It is pretty funny what he did. I could have a lot of fun with posts like that, Tamino would blow a gasket.
Extending a polynomial fit always results in bs, immediately after the last data point so this last graph is not surprising at all and doesn’t discount the otherwise humorous math in my opinion.
[DC: “He”? Do you mean MacRae? Or Gunter or Spencer? The polynomial fit was “bs” before I extended it, although extending shows how ridiculous it is. It seems you’re pretending it was all a joke on their part. If only that were true … sadly and pathetically, they are all quite serious.]
There is something reasonable to be said for non-linear fits though. I don’t find the argument that only a linear trend can be used particularly compelling. There is shape there, although I’ve heard the AGW argument to the contrary.
[Jeff, have you actually read any of the IPCC reports? Or looked at the Hadley CRU website? It appears you don’t realize that binomial smoothing is standard for the longer surface data sets. Linear trends are applied to the satellite data only because of the shortness of the record. You should also check out Tamino’s posts on “lowess” smoothing
e.g. this one. Perhaps a properly tuned lowess fit would be an improvement on the linear trend – I might try that some time.
But not all “non-linear” fits are created equal. Of course, a higher order polynomial is bogus and misleading in this case; the undue influence of the end point makes that clear. And the intellectual dishonesty is compounded by Spencer’s apparently deliberate omission of the linear trend. But if you want to defend that sort of thing, go right ahead.]
DC, thanks for this.
I first encountered this graph or a variant at the tail end of last year. I went so far as to create a webpage so that people could properly see the graph, but then took it down again. The page is now back up, (but only for a short while).
What got up my nose about this incarnation was that I found it on Spencers blog. Spencer has a PhD in a maths based subject. He has no excuse for this sort of misdirection.
“Jeff, have you actually read any of the IPCC reports? Or looked at the Hadley CRU website? It appears you don’t realize that binomial smoothing is standard for the longer surface data sets.”
Of course I have seen the binomial smoothing including some of the highly manipulable versions which straighten endpoints based on user input parameters. I’m not an idiot . . . honest. 😉
These useful tools are also abused by people who want to make the data look different than it is.
As far as the joke in this case we are in agreement. I was just appreciating the humor, it’s a little heart warming to see an AGW supporter get equally upset about exaggerated presentation. You should see how hot I get about the endless bad math in temperature reconstructions. How about the hideous CPS method Mann08 used. It probably took a year off my life.
Extending any high order polynomial fit is mathematically meaningless though and not useful for extrapolation. So the last graph proves nothing but you are still correct.
The video you did pretty well makes your case.
[DC: I do hope you are not accusing the IPCC or Hadley CRU of “abuse” of “useful tools”.
If extending a polynomial fit is meaningless, then so is the trend (first derivative) at the end point. It’s obvious that contrarians would not use a polynomial fit if it didn’t show a downward trend at the end. Extending the graph simply makes the dishonesty clearer for those less mathematically inclined than you or me.]
Hey DC, thanks for the mention … it alerted me to your existence! Nice work, hope to see much more.
Some reason for the snip?
[DC: I moved your last comment to unthreaded per my comment policy].
Pingback: Maintaining the Consensus « the Air Vent
Pingback: “Suppressed” Carlin report based on Pat Michaels attack on EPA « Deep Climate
To sum up:
Creating ‘ground-up’ models of climate using the known behaviors and effects and the most accurate current estimates of their values and coefficients: bad use of computers, phony baloney models, can predict anything/can’t even predict current climate.
Modeling temperature as simple sixth power of time: brilliantly incisive mathematical proof of cyclic nature of temperature.
So glad that’s all cleared up now.
Pingback: More heavy lifting from the “suppressed” Alan Carlin « Deep Climate
Thanks for this. Amazingly well done!
Wondered why your statistical analysis did not include the years 1940 to 1979, for which data is readily available? What would the the straight line fit for the expanded period be?
Do you have an explanation why there are periods of temperature decline when the CO2 curve shows relentless increase?
[DC: This article discusses contarian “analysis” of the satellite-derived tropospheric temperature record. That record exists only from 1979. There is no satrellite data before that time.
Twentieth century global temperature, including the slight decline between roughly 1940 and 1970, can only be explained by a combination of increasing anthropogenic influence and various natural factors. Please see IPCC Chapter 9 Executive Summary or the IPCC FAQ 9.2. Of course, there have been no decade-over-decade declines in the last forty years.]
Thanks for the response. Still curious as to the mechanism that allows the temperature to depart from increasing while the catalyst concentration is monotonically increasing. Natural inputs to the system (volcanic eruptions), cause instantaneous changes.
Your last statement “Of course, there have been no decade-over-decade declines in the last forty years” seems at variance with your assertion that you have only valid data since 1979. One, that is only 30 years, and two we have experienced static or “slightly decreasing temperatures since 1998. Can you offer an explanation, including natural forcings that have offset the anthropogenic CO2 forcing since 1998?
[DC: The satellite-based tropospheric record goes back to 1979, but the surface records all show decadal increases in the 1980s relative to the 1970s. So my statement stands.
Comparison to the single year 1998 is misleading. When one compares decades or longer intervals there is absolutely no evidence of “static” or “decreasing” temperatures. None whatsoever.
Within the 2000s, we do see natural variation from year to year, much of it from ENSO. Nevertheless, 2008 was the tenth warmest year in the instrumental record, but was relatively cool compared to immediately preceding years, due to confluence of a pronounced La Nina and a longer than usual inter-cycle solar trough.]
Pingback: Anatomy of a lie: How Marc Morano and Lorne Gunter spun Mojab Latif’s remarks out of control « Deep Climate
…the interaction of science, advocacy and politics in both the global warming and eugenics cases share a number of characteristics:
• Powerful advocacy groups claiming to represent both science and the public in the name of morality and superior wisdom.
• Simplistic depictions of the underlying science so as to facilitate widespread ‘understanding.’
• ‘Events’, real or contrived, interpreted in such a manner as to promote a sense of urgency in the public at large.
• Scientists flattered by public attention and deferent to ‘political will’ and popular assessment of virtue.
• Significant numbers of scientists eager to produce the science demanded by the ‘public.’
Prof. R. S. Lindzen, MIT
[DC: Can this possibly be a serious comment? Alas, it appears the answer is yes. It’s off topic, though – you have been warned]
Pingback: Do you subscribe to James Lovelock's conclusion that it's too late to stop the effects of GW? - Page 5 - PriusChat Forums
I wonder how that higher order fit is doing now.