SkepticalScience Needs to Update their Escalator

Gary Pearse, the staircase went down between 1945 and 1975. Don’t panic.

There have been El Ninos in the past. If El Ninos mean stepwise warming, then we should have had such stepwise warming in the past. But, no, we do not see such. Something different is happening today (that is, the last forty years or so). El Ninos do not explain the circa 1977-97 warming trend, despite the apparent coupling of step-warming with ENSO events.
However, insolation can explain it. The question is whence the increase of insolation? Answer this and we will know the cause of the late warming trend and whether this will resume in the future. CO2 can be eliminated from consideration as a factor in climate.

so, what’s the Atlantic doing?

Illustrating why escalator’s are embarrassing to climate science, especially if, like Tim Yeo & John Gummer, you think the last decade having record temperatures means anything at all
Well, it means that the last decade had a number of record temperatures (record over the thermometric era, probably not records over the Holocene and certainly not record temperatures outside of the Holocene).
However, another point that is being missed in the discussion is this. If one takes a peek at Figure 9.8a from AR5:
and try to discern all of the little pastel-colored lines that make up the results presented for each of the CMIP5 models, how many of these exhibit a stair-step behavior? Would that be zero? How many of them exhibit the same range of temperature fluctuation around their “smoothed mean” behavior as is observed in the real climate? Would that also be zero? That’s a real question, as one cannot see most of the behavior of individual models in this obfuscating figure, but it is quite certain that MOST of the models have a variance that is between 2 and 3 times the variance observed in the actual climate (with absolutely zero individual models that exhibit “fifteen year intervals without temperature increase”). Would this obvious problem with the variance be a small fraction of the actual problem, because these pastel curves are themselves averages over many perturbed parameter runs for each model and hence the variance is already reduced by a statistical factor from the average variance for the individual runs? It would.
The models presented in figure 9.8a of AR5:
* Spend close some 8 or 9x as much time above the HADCRUT4 GAST estimate as they do equal to or below that estimate (outside of the reference interval — in fact let’s exclude the reference interval from all of the remarks in this list).
* In the rare intervals that the MultiModel Ensemble Mean (and don’t get me started on this statistical abomination) does go below HADCRUT4, it goes barely below. When it is above, it goes way above. To the extent that this mean reflects the general behavior of the already internally averaged underlying models, we can conclude on the basis of these two observations that:
* The p-value for the null-hypothesis “The [number of] models of CMIP5 [that] correctly compute the physics and solve the problem they are attempting to solve is absurdly small. That is, the probability of correct models producing this behavior by random chance is very small (easily less than 0.01), basically zero. There is no reasonable possibility that the climate models presented in figure 9.8a of AR5 or their MME mean are correct and that the deviation between the models and the actual climate outside of the reference interval is due to random chance.
* The individual, already averaged models contributing to the MME mean have variances 2-3 times too large. Since they are already averaged, it is probable that individual model runs exhibit global temperature fluctuations as much as an order of magnitude too large around whatever one imagines constitutes their “mean behavior”. IMO this is strong evidence that they accomplish their forced fit across the reference interval by incorrectly balancing competing dynamics — taking a large warming effect and partially cancelling it with a large cooling effect to produce a much smaller warming effect. Instead of damping temperature fluctuations with dynamic feedback, this effectively amplifies fluctuations by causing much weaker dynamics feedback damping mechanisms to be heavily overdriven and lagged before they can respond.
* And then, yeah, they no not exhibit anything like Hurst-Kolmogorov behavior, a.k.a. “stair stepping”, or the “escalator”. That is, they have a dynamic signature that is completely inconsistent with the actual data. They aren’t even qualititatively correct.
Sadly, I could go on. I sometimes wonder if anybody actually looks at anything but the SPM in the Assessment Reports. I don’t understand how any scientist who looked at this figure would not go “whoa, that’s not right” long before they got around to looking at the right hand end of it, where the model predictions are actively diverging from the observed temperatures, precisely as one would expect if the fit in the reference interval were being extrapolated out of its range of partial validity where the large cancelling terms no longer cancelled.
rgb

OK, I’ve tried to post the graphic from AR5 twice inline twice and it hasn’t worked. Here is a URL:
http://www.phy.duke.edu/~rgb/Figure9.8.jpg
Enjoy.
REPLY: I’ve converted to image for you. – Anthony

I was looking at this: http://www.skepticalscience.com/climate-of-doubt-escalator-updates.html
I wonder why they use a line for the temperature data, and a curve for the ice data. I’d like to see a curve applied to the temperature data.

How is it possible for anyone passing beyond Groundlevel at University anywhere not to have learn what it takes to present valid arguments that can lead up to a sound conclusion?
Please read Fallacies in argumentation (English text)

Or, play Logical Fallacy Bingo. It’s a lot more fun than reading a dry textbook… (the information included in each card will tell you what they are, as will a mouse hover over the space). I play all the time with postings. Since pointing out the fallacies never seems to affect the probability that the same person will commit them a second time, or third, or fourth, playing bingo is also every bit as useful…:-)