As often happens, Eli Rabett has yet another fascinating post, this time about how Barry Bickmore has solved some of the mysteries created by Christopher Monckton. Monckton responded to the expose of his misdeeds the way he usually does: by threatening Bickmore. But you can read about that from the Rabett (honor among bloggers and all that).
First, Bickmore figures out where Monckton got his “data” for the IPCC projections of CO2 concentration; it seems that Monckton was less than totally forthright about this. Again I’ll let Bickmore tell the story. Second, when Bickmore asked the House of Lords about Monckton’s membership in that august body he received this reply:
“Christopher Monckton is not and has never been a Member of the House of Lords. There is no such thing as a ‘non-voting’ or ‘honorary’ member.”
I noticed something odd in Bickmore’s revelations about Monckton’s manipulation of IPCC CO2 scenarios. Monckton insists that actual CO2 concentration isn’t rising exponentially, that in fact the increase is linear, saying:
I am, of course, familiar with the fact that, over a sufficiently short period (such as a decade of monthly records), a curve that is exponential (such as the IPCC predicts the CO2 concentration curve to be) may appear linear. However, there are numerous standard statistical tests that can be applied monotonic or near-monotonic datasets, such as the CO2 concentration dataset, to establish whether exponentiality is being maintained in reality. The simplest and most direct of these is the one that I applied to the data before daring to draw the conclusion that CO2 concentration change over the past decade has degenerated towards mere linearity. One merely calculates the least-squares linear-regression trend over successively longer periods to see whether the slope of the trend progressively increases (as it must if the curve is genuinely exponential) or whether, instead, it progressively declines towards linearity (as it actually does). One can also calculate the trends over successive periods of, say, ten years, with startpoints separated by one year. On both these tests, the CO2 concentration change has been flattening out appreciably. Nor can this decay from exponentiality towards linearity be attributed solely to the recent worldwide recession: for it had become evident long before the recession began.
For some reason, I don’t feel inclined to take Monckton’s word for it. I guess I’m just a natural-born skeptic.
So let’s “calculate the trends over successive periods of, say, ten years, with startpoints separated by one year” to find out whether or not “the CO2 concentration change has been flattening out appreciably.”
We’ll use CO2 data from Mauna Loa with the seasonal signal removed; the de-seasonalized data look like this:
That certainly doesn’t look linear — and in fact it isn’t. There’s an increase in the growth rate (rather than constant growth rate as for a linear trend) that is definitely statistically significant.
But we’re really interested in whether or not CO2 is increasing fast enough to be called “exponential.” Despite Monckton’s claims about “standard statistical tests” for exponential growth (he talks better than he knows), the simplest way to tell is just to log-transform the data. If the data are growing exponentially, the log-transformed data will be growing linearly. Here’s the logarithm of the CO2 data:
Well well! Over time, the growth of CO2 has NOT been linear, but it also has NOT been exponential. It’s been faster than exponential (as the logarithm has grown faster-than-linear, i.e., it has accelerated). And yes, the acceleration of log(CO2) (the faster-than-exponential growth of CO2) is statistically significant. That settles it.
But hey, let’s give Monckton his due, and apply a test he himself recommended. We’ll compute the linear regression slope for 10-year intervals with starttimes spaced 1 year apart. Here’s what we get:
Note that the rate is increasing overall, it’s even increasing recently; the last 10-year interval has a higher growth rate than the 1-year-preceding interval.
We can even compute the linear regression slope for 10-year intervals of the logarithm of CO2 concentration, with starttimes spaced 1 year apart:
Once again, the rate is increasing overall, it’s even increasing recently; the last 10-year interval has a higher growth rate than the 1-year-preceding interval. So even recently, the growth of CO2 has been faster than exponential.
It’s true that recently, the growth rate isn’t increasing as fast as it used to be, so CO2 concentration isn’t increasing as much faster than exponential as it used to be. But it’s still faster than exponential. And of course, we expect some variations because, like all geophysical data, the CO2 concentration shows noise as well as signal.
But the fact remains that Monckton’s claim that CO2 has been increasing linearly rather than exponentially just doesn’t hold water. Most telling is the fact that Monckton claims this linear-rather-than-exponential growth for the time period 2002-2009. The noise just doesn’t allow such a determination. If Monckton knew what he was talking about, he’d surely know that.
One final note: in his regular publications through the SPPI (the “Science” and Public Policy Institute), Monckton often graphs global temperature, which is sometimes referenced as “Source: SPPI global temperature index,” and sometimes referenced as “Data source: SPPI index, compiled from HadCRUt3, NCDC, RSS, and UAH.” But I can’t seem to find the actual data. Anybody know where to get the “SPPI global temperature index” data?