.comment-link {margin-left:.6em;}

Wednesday, January 07, 2009


The Australians War on Science Keeps on Rolling.

Tim Lambert has been keeping a tab on the Australian's war on Science, counting up the number of erroneous or misleading articles on science (mostly about global warming). He's up to Number XXX, but he will have to change it to XXXI with an article from January 6 by Jon Jenkins, "The Warmaholics Fantasy".

Now, Jenkins is an Adjunct Professor of Virology, specialising in computer modelling. Why is it that the Australian never gets actual climate scientists to comment?

Heck, I'm a Senior Lecturer in Pharmacology, and I do a lot of computer modelling (okay, so its molecular modelling), why do they never invite me to comment? (insert pout here). Professor Jenkins opens with a standard litany of denialist taking points, satellites more reliable than ground based temperature recording (it's not), and that the "hockey stick" graph is fraudulent (it isn't). Professor Jenkins should be a little more careful with his accusation of fraud though. In his article he says "They [the satellite measurements] show minuscule warming, all in the northern hemisphere, which not only stopped in 2000 but had completely reversed by 2008 (see graph)."

Now the graph is not in the online version so I have taken the liberty of reproducing the graph under fair comment rules (its the graph at the top, click on the image to enlarge it, the Australian seems to have pinched it anyway). The data is from the UAH MSU global monthly lower troposphere temperature data set. You can down load it yourself and graph it. Just like I did, the result is in the bottom image (again, click on the image to enlarge it). You can compare my graph with the image accompanying Professor Jenkins article.

One of the first things you can see is that the Australian graph finishes just a wee bit early. Now I don't know if Professor Jenkins prepared the graph (it appears to be redrawn from an illustration in a US newspaper article published in October, whoever did the redrawing was being a bit naughty in producing misleading attributions), but he references it, and should have checked it for accuracy and completeness. Because the data set he uses finished in August, missing out the September, October and November data. If you include that data, the temperature anomaly bounces right back, regaining the supposed "wiped out" warming. Now the November data were there way back in December, so there is no excuse for showing a truncated graph. Professor Jenkins may not have produced this misleading graph, but he should have checked it (finding the original data wasn't hard).

The "trendline" shown in the Australians graph isn't strictly speaking a trendline, it's just a high order polynomial fit, which will generate a nice looking smooth line but will not actually reflect the trend of the underlying data. It's rather naughty to use polynomial fits in these sorts of situations. Again, Professor Jenkins appears not to have produced the fit, but he should have known immediately that this sort of polynomial fit is inappropriate. In my graph I show both a linear fit (black line) and a rolling 3 year average fit (red line). Both more appropriate models for this sort of noisy data. As you can see, global warming has not gone away. As well the UAH data agrees with the other satellite data and the two land based data systems in showing that warming hasn't stopped. Indeed the UAH data shows a warming of 0.14 degrees C per decade, comparable to the 0.16 degrees per decade found by surface stations (the RSS satellite data finds warming of 0.18 degrees C).

The dip in 2008 was caused by a La Nina event, and is not surprising at all. See also here and here for more discussion of 2008 temperature trends.

Professor Jenkins goes on at length about a whole range of other issues in climate change, but it won't surprise you that he is wrong about them too (eg Volcanoes as CO2 generators). Ironically, this piece was published the day before the Quadrant Hoax hit the Australian headlines. You think the editors might have done a little fact checking too.

But, Professor Jenkins the bottom line is you shouldn't be too quick to accuse scientists of fraud when you can't be bothered to check out the datasets and draw your own graphs (and do your own modelling).

UPDATE: Tim Lambert blogs the Jenkins article here. Also, Jenkins has in his article the statement "Science is only about certainty and facts." It is not, it is about inference to the best explanation. But again, the irony of bringing up certainty when Jenkins hasn't done his fact checking is palpable.


It's rather naughty to use polynomial fits in these sorts of situations.

It's not "rather naughty", it's completely preposterous and dishonest. OBVIOUSLY the end of the polynomial will take a sharp turn, and if he is a science professor of any kind he knows that perfectly well. Who the hell gave this moron a doctorate?
We could fit a sine wave to these data if we wanted; it wouldn't be a very good fit though ;-) I wonder if Prof Jenkins has a value for the goodness of his fit? Does he say why he thinks a low-order polynomial should be a good fit? Also, I wonder what he makes of the sudden steepening in the gradient of his fit; it would have things changing now significantly more than at any point in the past 28 years.

Given the obvious correlated noise in the data, we really shouldn't draw conclusions one way or another based on time-scales close to (or less than) four years.
"why he thinks a low-order polynomial should be a good fit?"

Because Excel cannot do polynomial fits beyond 6th order.

If he did a 30 order fit - you know 1 order per year (degree of freedom) - he'd end up with a smoothed zigzag. But that would give the game away.

'Cause then he would have to explain what particular 30th order physical process he believes underlies his model (hint: there are no 30th order physical processes, and I can't off the top of my head think of any 6th order ones either).

This trope has been around for a couple of months now and someone needs to stomp on it.

Thanks for your effort.
It's better to just not read 'The Australian'. You'll feel much better for it. And I hope you haven't been buying it with your own money.
I have a feeling that I would be highly amused by seeing where the polynomial ends up in, say, 2050, or where it was in, say in 1900. Unfortunately, my graphing skills are non-existent (I still can't figure out how to import GISS data into octave, though I guess I could copy-paste it..)

has anyone tried this?
Y'all might be interested, I did get the GISS data into Octave, and figured out how to do some pretty graphs.

I ran a 6th degree best fit over GISTEMP data from 1880-2008. Funnily enough, I didn't see a downward curve at the end, but a distinctly upward one.

Oh, and I extended the line to 1700 -> 2100, and I notice that the planet will be 7 degrees warmer in 2100, and that there was a global ice-age before about 1700. Nifty graphs here (down the bottom, last two images).
Awesome post/thread. Just a quick one (I'm just a lowly telco engineer with an interest in this stuff through my snowboarding company) -- as I have forgotten my uni statistics theory; but why are higher order polynomials naughty/egregious/dishonest etc?

Tim M
why are higher order polynomials naughty/egregious/dishonest etc?

This SHOULD be obvious: Because they are just coincidental fits to the raw data -- they aren't in any way predictive.
Post a Comment

Links to this post:

Create a Link

<< Home

This page is powered by Blogger. Isn't yours?