No significant warming for 17 years 4 months

Gail. No doubt all readers who think that CO2 drives temperature will acknowledge that the GISP graph you linked “proves” that C2 cools the earth and is an Ice House not a Greenhouse gas

Gail Combs says (June 17, 2013 at 5:54 am): “Again note the cherry picking of values as outlined by Mauna Loa Obs.”
Cherry picking or good observational science? Measuring the background level of something so easily affected by local sources is no easy task. CO2 is also measured at the South Pole (and elsewhere). Even in such a “pristine” location, precautions must be taken. On the linked page, watch the air sampling video, then compare the two graphs of air samples taken downwind and upwind of the station. I’ve read that background CO2 measurements from around the world are quite comparable, so either there’s a lot of cherry-picking going on, or CO2 measurement is one of the more reliable aspects of climate science.
“The whole hoax is based on the ASSumption that CO2 is uniformly mixed in the atmosphere…”
1) The assumption is actually that CO2 is “well-mixed”, which isn’t the same as “uniformly mixed”.
2) I don’t believe the concept of CAGW or even AGW requires that CO2 be “well-mixed”, though in that case the GCMs might need to take into account persistent geographic variations in so-called “greenhouse gases”.
3) If I understand the AGW concept correctly, the critical area is the upper atmosphere, i.e. the “effective radiating level” or ERL. Anybody know if CO2 is “weller-mixed” in the upper atmosphere than at ground level?
“The Japanese satellite (JAXA) shows CO2 is not ‘well-mixed’. map 1 and map 2”
That depends on what the meaning of is “well-mixed” is. 🙂 Even on these maps the CO2 levels vary less than 20 ppm (by eyeball) over the (limited) coverage areas, or roughly 5%. Considering the non-uniform CO2 sources and sinks, that’s “well-mixed” to me. Check out a larger set of maps. Note the variation of CO2 on time scales as short as a month. Again, it looks like the CO2 is getting stirred around pretty well.
BTW, two observations on these maps:
1) I understand the satellite measures the entire column of atmospheric CO2, so even if CO2 is more uniform in the upper atmosphere, the reading would be skewed by the levels closer to the surface; and vice-versa, although less so because the absolute amount of CO2 decreases with altitude.
2) Comparing the same month year over year, the maps get redder. Near Hawaii in April 2010 I see an orange square, consistent with the 392.52 ppm CO2 reading from Mauna Loa. In April 2013 there’s a reddish square over Hawaii, consistent with a Mauna Loa measurement of 398.40 ppm. The scale has to be exaggerated to make such a small change noticeable, but as a result a 1.5% change seems to set the maps on fire. 🙂
“The Trouble With C12 C13 Ratios”
I visited that link before and found Chiefio’s musings quite intriguing. Likewise the carbon isotope section of Murry Salby’s talk covered at WUWT. I’m not sure if the satellite data support or contradict Salby and Chiefio.

1) I understand the satellite measures the entire column of atmospheric CO2, so even if CO2 is more uniform in the upper atmosphere, the reading would be skewed by the levels closer to the surface; and vice-versa, although less so because the absolute amount of CO2 decreases with altitude.
Did you notice the other oddity? There is no possible way that the Japanese satellite measurement OF the entire air column supports global CO_2 on the edge of 400 ppm. Eyeballing the graphs the means should be what, 380 ppm, and as you note, it should be skewed high compared to Mauna Loa, not low.
I’m thus troubled by what appears to be a 5 or 6% error in normalization. That’s actually a rather lot. Mauna Loa is what it is, whether or not you approve of its data reduction methodology at least it has been consistently applied for a long time so its own value is apples-to-apples. But it has been used as a sort of gold-standard of global atmospheric CO_2, and I’d rather bet that its readings are used as primary input into the GCMs.
The satellite data suggests many things. First, that the Mauna Loa readings are a lousy proxy for global CO_2 levels. Second, that they are far too high — as you note, if CO_2 concentration on average increases with depth (which makes moderate sense and is consistent with it not being perfectly homogeneous over the globe as it is) then the mountaintop reading should be lower by some percentage than the mean of the entire air column, not higher. Given that surface readings are frequently in the 400 to 500 ppm (depending on where you look) Mauna Loa could be off by 10% or 20% on the high side compared to true top-of-the troposphere average CO_2 concentration. Since that is where the atmosphere nominally becomes transparent to LWIR emitted from the CO_2 because the atmosphere itself has thinned to the point where it is no longer optically opaque, this suggests that the emission temperature being used in the models is derived from a point too high in the DALR (and hence too cold) — thereby exaggerating warming.
If it “only” a 5% effect I’d worry but perhaps it isn’t “important” (bearing in mind that the entire post-LIA warming in degrees Kelvin is order of 0.5%, so small numbers cannot safely be ignored when trying to explain the 0.1-0.2% that might be attributable to anthropogenic causes). If Mauna Loa is off by 10% or more for any reason whatsoever, that cannot possibly be ignorable. For example, if top of the troposphere global mean CO_2 were really 370 ppm and not 400 ppm, that’s a huge difference.
One wonders if there are any reliable controls. The other possibility is of course that the satellite isn’t correctly normalized. One wonders if contemporaneous soundings support one or the other. That actually might be a large enough error to kick the GCM spaghetti back down into agreement with nature all by itself, although systematically correcting them back in time is going to be very difficult. One also wonders why Mauna Loa produces so very different a number (just as one wonders why LTT is diverging from land surface temperature records, or was until they stopped adjusting it because they had to after one final push of all the older temperatures down).
rgb

And rgb, spectrums of IR from the poles show CO2 emitting instead of absorbing. This goes w/Chiefo’s idea that the TOA (defined as no convection) at the poles during an inversion is actually at the surface, and CO2 acting as a coolant instead of an insulator. So the standard theory of ~1C warming from CO2 doubling might not be so solid, IMO.

rgbatduke says (June 18, 2013 at 7:51 am): “Did you notice the other oddity? There is no possible way that the Japanese satellite measurement OF the entire air column supports global CO_2 on the edge of 400 ppm. Eyeballing the graphs the means should be what, 380 ppm, and as you note, it should be skewed high compared to Mauna Loa, not low.”
Well, I’m not so sure. In the 2013/04 map, for example, Hawaii doesn’t seem out of line with the rest of the northern hemisphere, and in fact several places are even redder. The maps seem to cover most CO2 sources, but a lot of space is blank. Plus, as you mention later in your comment, calibration/sensitivity/reliability of this new satellite tool is unknown. And this says CO2 at the South Pole is within 6 ppm or less of Mauna Loa. Most or even all of the difference is explainable as a gradient from the major CO2 sources to the north.
BTW, the GOSAT site also has a neat animation of global CO2 distribution. It’s a simulation, so take with a grain of salt, but it smooths out the monthly maps nicely.
Last night I re-watched the “Seasonal Forests” episode of the BBC’s “Planet Earth”. It mentioned that the vast northern forests are a major source of atmospheric oxygen. In the GOSAT simulation, Canada and Siberia turn dark blue (low CO2) in late summer, indicating massive photosynthesis and oxygen production. Watching the simulation unfold, I felt the same thrill up my leg that Chris Matthews gets listening to Obama. 🙂

Some Moncktonesque statistical treatment here, what exactly does “bounded by the means of the trends on these published uncertainties” mean in relation to statistical significance? Using the same data I obtained a trend of 0.089±0.118 ºC/decade which indicates statistically significant warming at the 85% level. To say that there has been “no statistically-significant warming in 17 years 4 months” as Monckton does is meaningless.
Not wanting to speak for Mr. Monckton, of course, but I suspect he is referring to for the linear fit, usually interpreted as a measure of the signficance of the fit compared to the null hypothesis of no relationship. I also think he is using “accepted” values for the conclusion, which a lot of people have grudgingly been coming to accept in the climate community.
At the same time, I completely agree with you. There is nothing special or magical about 17 years, or 16 years, or 13 years. Or rather, there might well be, but we don’t know what it is because we do not actually know the relevant timescales for variation of the climate as opposed to the weather. The climate as reflected in some sort of average global temperature derived from either the thermometric record or proxies has never been constant, never been linear, never had anything like a single time constant or frequency that could be associated with anything but a Taylor series/power series fit to some sufficiently small chord or fourier transform ditto. Well, I take that back — there is a pretty clear fourier signal associated with Milankovitch processes over the last 3.5 million years, only the period changes without warning three times over that interval (most recently to roughly 100 ky) and we don’t know why.
The much better way to assert precisely the point Monckton makes above is by asserting no point at all — simply presenting the actual e.g. LTT and SST and LST records over the last 34 years where LTT, at least, is consistently and accurately measured, SST is increasingly precisely measured, and sadly, LST estimates are comparatively dubious. Over that entire interval, LTT (as arguably the best measure of actual warming for a variety of reasons) suggests a non-catastrophic warming rate on the order of 0.1 C/decade but with (obviously!) large error bars. SSTs lead to a similar conclusions. Measurements of SLR (any mix of tide gauge data and satellite) lead to a similar conclusion.
Deconstructing the causes of the warming, decomposing it into (say) a flat fit pre-1997 and a second flat fit post-1999 (which reveals that most of the warming occurred in a single discrete event associated with the super-El Nino/La Nina pair in between as far as we can tell from the data) or a linear fit, or an exponential fit, or throwing new fourier, linear, or otherwise functional components in coincidence with decadal oscillations, the solar cycle, the level of solar activity, CO_2 concentrations, stratospheric water vapor, stratospheric ozone, or the density of kitchen sinks (per household) in Zimbabwe is in some sense all complex numerology, climatological astrology. The temperature record is what it is, not what we would fit it to be, not the stories we make up to explain it because we damn sure cannot compute it!
That’s the real mistake Monckton made — he presented an aggregate view of the GCMs, because that is what the IPCC did in its notorious AR4 summary for policy makers, which contains an absolutely horrendous abuse of statistics by using the averages and standard deviations of model results over many completely different GCMs to make quantitative assertions of the probabilities of various warming scenarios as if “structure of a particular climate GCM” is an independent, identically distributed variable and reality is somehow bound to the mean behavior averaged over this variable by the central limit theorem which is sheer madness. This isn’t Monckton’s error, it is a standard error made by the IPCC, an error where incompetence in statistical analysis is beautifully married to political purpose to the detriment of scientific reasoning), but he perhaps should have avoided perpetuating it (as Nick Stokes rather overvehemently insisted above and just presented the spaghetti snarl of actual GCM model results themselves as they say precisely the same thing, only better, without the irrelevant and incorrectly computed probabilities.
Madness and incorrect computation that is, of course, perpetuated and reflected in your citing 0.089±0.118 ºC/decade which indicates statistically significant warming at the 85% level.. Let’s restate this in Bayesian language. If there is no bias in the process associated with generating the points being fit (an assumption we can frame as a prior probability of there being occult bias), if the underlying climate is a linear function (an assumption we can frame as the prior probability of the climate behaving linearly over some interval of time, an assumption we can actually make at least semi-quantitative if one can believe the proxy record over the Holocene and bewares the fact that much of that proxy is intrinsically coarse grain averaged over intervals longer than 33 years), if the error bar you obtain from a linear fit (presumably from the distribution of Pearson’s ) over a remarkably short interval where we do not know the timescales of the relevant noise compared to the linear trend is relevant (again, one can try to guestimate the probable timescales of noise compared to the fit interval, but the curve itself strongly suggests that the two are comparable as it decomposes into two distinct and quite good fits joined at the ENSO in the middle), and if there is nothing else going on that we, in our ignorance, should be correcting for, then your linear fit yields warming with a considerably wider variance than you are allowing for — none of the Bayesian probabilities above are optimal for the linear fit to be precisely meaningful, and the uncertainties they introduce all broaden the uncertainty of the result or worse, reflect the probability that the linear fit itself is nonsense and cannot be extrapolated with any real confidence at all.
I strongly suggest that you read the Koutsoyiannis paper on hydrology that has as it first graphic a function plus noise at a succession of timescales (in fact, Anthony, this graph should be a front-page feature on WUWT somewhere, IMO, as a permanent criticism of the plague of linearizing a function we know is nontrivially nonlinear). On a short enough timescale it appears linear. Then it appears exponential. Then it appears sinusoidal. But is that really its behavior? Or is the sinusoidal merely systematic noise on a longer term linear behavior? Note that no possible statistical analysis on the original fit interval can reveal the tru-er longer time behavior, and at no time can one separate out the unknown longer time behavior from the fit.
In the meantime, here is a statement that perhaps everybody — even Nick Stokes and in other venues Joel Shore — can agree with. The last 15 to 17 years of the climate record — depending on which side of the discrete “event” of the 1997/1998 ENSO you with to start on — are not strong evidence supporting the hypothesis of catastrophic anthropogenic whatever, global warming, climate change. So much so that the warmist community stopped using the phrase global warming over this interval, substituting the non-falsifiable term “climate change” instead because any weather event can safely be attributed to human activity and who can prove you wrong other than by asserting statistical arguments so abstruse that any member of the lay population will fall asleep long before you finish them, where the catastrophe itself is always immediate and exciting and real to them. Every experience is a peak experience if you are demented and forget past experiences, after all.
Note that I carefully avoid stating that the data “falsifies” the assertion of CAGW, AGW, or any particular “scenario” put forth in AR4’s summary for policy makers. To make such an assertion I would have to have prior knowledge that not only I lack, but everybody lacks. The issue of e.g. UHI bias and cherrypicking in the contemporary land surface temperature record is an open question at this point, with evidence supporting both sides. Who really knows if there is a bias, what the sign of the bias is, what the magnitude of the bias is? Even given the best possible intentions and totally honest scientists constructing it, totally honest scientists have biases and often cannot help incorporating them into their database — an assertion I make with considerable empirical evidence derived from meta-studies in e.g. medical science.
There is considerable unacknowledged uncertainty in climate science — the egregious treatment of various Bayesian priors as equal to unity or zero (in a way that reflects one’s prejudices on the issue) in order to avoid diluting the purity of one’s conclusions with the inconvenient truth of uncertainty. But quite independent of Bayes, it is entirely safe to say that an interval of essentially flat temperatures does not support the assertion of aggressive, catastrophic, global warming. Indeed, 13 years (starting MOST generously in the year 2000) is 1/8th of a century. If 1/8th of the twenty first century has been climate neutral in spite of a significant increase in CO_2 in that time and over an interval in which the GCMs unanimously call for aggressive warming, one would have to lack simple common sense to assert that this is evidence for the correctness of the GCMs and the likelihood of a catastrophic warming increasingly confined to the 7/8 of the century remaining.
rgb

rgbatduke says (June 19, 2013 at 8:07 am): “Let’s restate this in Bayesian language.”
I gather from what follows the above that the “Bayesian” language must be spoken in very long and very complex sentences. 🙂
No worries, though. I ran it through Google Translate and after about 15 minutes of processing it spit out “It’s not statistically significant, Phil.”. 🙂
One final thought: We must at all costs keep RGB away from the cryptic Steve Mosher, lest their mutual annihilation destroy the entire planet. 🙂

{ rgbatduke says:
June 19, 2013 at 8:07 am }
I’ll bet you’re a hoot at a party!
;<)