Climate highly sensitive to C02 changes

Courtesy of Watts Up With That, a nice new article that examines carbon content of fossil soils to estimate the typical sensitivity of climate to changes in C02.

Paleo tagging past climate sensitivity Watts Up With That?

The article in its original format is here, though full text is behind a paywall (so I supposed we should thank Watts for stealing it for us):

Fossil soils constrain ancient climate sensitivity ? PNAS

The key finding is:

"Although further work is needed, the geologic evidence (2) (Fig. 1) is most consistent with long-term, future climate change being more severe than presently anticipated (5)."

 

Ah, golly, another one like TimBikes and Uforya.

If you really want to have a discussion, it would be helpful if you could post about a single point, and we could stay on the discussion of a single point.

Otherwise, crap, we have to ping back and forth across an entire spectrum of stuff and there's never any conclusion about anything. With TimBikes and Uforya, as soon as you try to get something pinned down on one issue, bing, they post a laundry list of other stuff.

OK, just looking at the length post above:

GENERAL COMMENTS ON METHODS

In general, studies that look at a single variable, in isolation, are going to be wrong, or at least unreliable, per se. The net climate change is the product of a number of relatively small forcings, as shown here:



Second, studies of correlation, as opposed to studies that actually model the physics of the atmosphere, tend to be weak. Particularly if they are single-variable studies (literally correlation) instead of, at least, "regression analyses" where you put multiple factors on the table at once.

Beyond those two general comments, and certainly not wanting to try to get into the detail of every item of the laundry list you posted, I note the following, just scanning it.

LOOKING AT C02 IN ISOLATION IS NOT VERY INFORMATIVE.

"The" driver is supposed to be atmospheric C02, and the effect is logarithmic. The first three graphs don't show atmospheric C02, they show annual hydrocarbon use. Thats not the same. It's not even close to being the same (e.g., atmospheric C02 was not near zero at the start of the graph).

So, if you're going to rely on a single-variable analysis (which, as I said above, is not the way to go). You at least have to put the right variable in there. A single-variable analysis that looks at the wrong thing is just doubly wrong.

So, if "Its got to be C02", then, at a minimum, you've actually got to put C02 on your graph.

THE EVIDENCE IS VASTLY DEEPER THAN ONE NUMBER (GLOBAL TEMPERATURE PREDICTED BY A MODEL)

Next, Lindzen's argument? Again, that's a tiny slice of the actual evidence. Even within the narrow band of evidence focused on models, it ignores most of the evidence. And it's poorly stated as is: The correct statement is, the models reasonable replicate temperature changes in the past. But as we come to the present, if we drop manmade forcings out of the models, they diverge enough from the actual temperatures that we can tell that the man-made forcings matter.

But all of that pretends that there's only one output from the model, or one piece of evidence to look at: global average temperature. That's foolish.

For example, we know that the top of the atmosphere (statosphere) is cooling, that warming is greatest at night (as opposed to during the day), and that warming in general is greatest in the Arctic and less in the tropics. (Warming, or the lack thereof, in the Antarctic is a special case and is well predicted by the general circulation models used to model global temperature).

Why does that detail matter? Because all of that detail is exactly what is predicted with greenhouse gases as the forcing, AND, to a first approximation, that detail rules out a lot of alternative explanations. That, by itself, doesn't "prove" that greenhouse gases are the culprit. But that, by itself, can rule out some explanations.

DETAILS ON TEMPERATURE CHANGE PRETTY CLEARLY RULE OUT THE SUN OR URBAN HEAT ISLANDS AS EXPLANATIONS OF WHAT IS BEING OBSERVED.

For example, let me discuss the hypothesis that the current warming is due to the sun. That hypothesis is based on studies of correlation between sunspot activity and temperature. As noted above, single-variable correlation studies are inherently weak. (And, for the last 30 years, we have direct, satellite-based measurements of solar output, so for the last 30 years, we know that the sun has, in fact, gotten slightly dimmer, on average, over the last three solar cycles.) But more to the point, if the current warming were caused by increased solar activity, then we would see the most warming where there was the most solar input: at the top of the atmosphere, during the day, and in the tropics. (I'm not just winging it on that, based on "common sense", that's what general circulation models show when you turn up the value for insolation.) As stated above, none of those things are true.


Similarly, if, as some claim, the measured temperature increase is due to the urban heat island effect, that could not possibly generate a decrease in stratospheric temperatures. At least, not by any means I've ever seen discussed.

So, to discuss this as if the only bit of data is mean global temperature as predicted by a model, that's just not right.

From that point forward, you're just all over the map. Let me point out a few obvious inconsistencies.

WARMING IN ONE AREA, USING ANCIENT EQUIPMENT, DOES NOT DISPROVE GLOBAL WARMING.

See your chart labeled "The oldest thermometer record in existence ...". You're not seriously saying that readings, in a single location, negates any possible inferences about global warming, are you? Because that's what that particular chart compares: Temperature in one location, going back to 1660, and contrasts two periods of warming, from that one location, and ... uses that to show that global warming is wrong?

So, later, you're absolutely convinced that the modern land-based temperature record are wrong, using modern instrumentation. But you'll take the data from a 1660 thermometer -- I can't even imagine what that is -- from one location, as evidence that global warming can't be correct?

Because there's no link to the original there, I've had to track down the source. It's the British Hadley center.

Oddly, I'm pretty sure that the device that we recognize today as a thermometer was invented in the early 1700s by Farenheit. I can't even imagine what the recording instruments were in 1660.

Ah, but me being me, I had to track down the original source materials.
http://www.rmets.org/pdf/qj74manley.pdf

What does the fellow who put that together say about the early part of the data.
"For the first six decades to 1720, the figures are printed in italics as an indication that they must be considered less reliable, based as they are on extrapolation from the results of readings of highly imperfect instruments in uncertain exposures at a considerable distance, generally in southeast England; or on estimates based on interpretations of daily observations of wind and weather."

So, not only is it one location, but your source actually managed to emphasize the very portion of the timeseries that the original author bothered to round and print in italics as a clear warning about the reduced reliability of the data. Interpretations of daily observations of wind and weather, indeed.

THE ACTUAL COMPUTER CODE USED TO GENERATE THE NUMBERS IS FREELY AVAILABLE.

Ah, golly, the computer code. I write computer code for a living, OK? I understand putting tests into the code for sensitivity analysis and to check how how the outputs respond to changes in inputs. Research computer code is always ugly. Mine sure as hell is.

But the code actually used to generate (e.g.) the GISS temperature series, along with all the data used to generate the temperature series, is freely available on-line. Unlike you, I've actually downloaded it and looked at it (although I program exclusively in SAS these days and can only "follow along" with what they do). None of the stuff that you allege is true about the code actually used to generate the series. You should check that out for yourself, as I did, rather than rely on the opinions of someone who has an axe to grind. Be warned: It's not easy. So? So if somebody gives you an easy explanation of why that code is wrong, biased etc. -- chances are very good that they haven't actually bothered to read and understand the actual code in use either. Because it's a damned hard thing to do, as I know from personal experience.

Data @ NASA GISS: GISS Surface Temperature Analysis: Sources

Ah, then we get into the usual stuff about how all the temperature data are wrong.

MEASURED WARMING AND URBAN HEAT ISLANDS AND SUCH: WHAT WATTS DOESN'T TELL YOU

Let me just point out the obvious logical inconsistency first: If the temperature data are wrong, then evidence that refers to the temperature data, including that cited by you, is wrong. Logically, you can't have it both ways.

In terms of the actual temperature measurements in the modern era, as far as I can tell, Watts truly does not understand what either NASA or NOAA do with the data, and Watts fails to acknowledge even the most basic facts.

What do I mean by basic.

Well, for starters, three-quarters of the globe's surface is water. Average water temperatures have warmed over the past century. Not as much as land temperatures, but that is exactly what you would expect based on the underlying science.

Similarly, world temperature numbers are geographically weighted. You don't just take the N data points, add them up, and divide by N (though you will see a lot of simple-minded amateur analyses that do that). You're trying to get at the average temperature per unit of area. You weight each station by the area it represents. That means rural areas count more than urban areas, because there's more rural area than urban area. OK, then, what percent of US land area is urban? Highest I've seen, under any reasonable definition, is 6%. For the globe as a whole, I have not seen an estimate over 5%.

Upshot: Water is 75% of the globe, maybe 5% of the remaining 25% is urban. Urban areas account for a little over 1% of the earth's surface, and the global temperature numbers are calculated to reflect that.

So, that's so basic, and so obvious, why doesn't (e.g.) Watts start his discussion from that, and then get into the details and the photos and whatnot? My guess is, he's not too fond of the accurate perspective it gives on what matters and what doesn't in terms of generating the global temperature estimates.

(I should note that I have left out a lot of the details of the calculation. The data are not even just a simple weighted average, because they've had to deal with changes in population of temperature stations over time. The entire calculation approach was constructed, from the start, to try to generate a consistent time series for changes in temperature from a shifting and sometimes inconsistent base of temperature monitoring stations. What I'm saying is, the folks who generate the global temperature numbers have dealt with these issues right from the very start, and designed what they do with that in mind.)

Second, NOAA actually looked at the specific issue raised by Watts regarding station siting. They took only the stations Watts approved of, and they took all stations, and after they've done what they do to remove the crap from the data, the two sets of stations gave the same trends. My takaway from that is that they are very good at removing the crap from the data. I've posted that on Priuschat about a dozen times, so, after wasting an hour on this response, I'm not going to look it up again.

Third, even for the narrow issue of US land temperatures (about 2% of the earth's surface), Watts and similar completely ignore the climate reference network. Back around 2000, it was clear that you'd like to have a better set of optimally-sited stations with high-quality instruments. There's now about 5 years of data from those stations, and it can be compared to the historical climate network data. Guess what? No difference in the measured US land temperature anomaly between the new, high-quality, remote rural climate reference network, and the existing historical climate network. If you were of an open mind, you could read about it here:

http://www1.ncdc.noaa.gov/pub/data/uscrn/publications/annual_reports/FY09_USCRN_Annual_Report.pdf

In particular, you'd want to look at figure 13, page 26.

So, did Watts ever clue you in on that? Wonder why not? (OK, I'm being flip, but I'll apologize if Watts somewhere referenced this finding.)

Fourth, GISS doesn't use urban area trends, in doing the trend calculation. Well before Watts got into this, the folks at GISS were sufficiently worried about about heat island effects that they force temperature trends in urban areas to match the surrounding rural areas.

Fifth, the entire "urban heat island" effect has been thoroughly studied, by meterologists and climate scientists. You can find a decent summary of some new results here:

Does Urban Heat Island effect add to the global warming trend?
RealClimate: No man is an (Urban Heat) Island
RealClimate: The Surface Temperature Record and the Urban Heat Island
If you don't like the fact that I've liked to summaries of the evidence on realclimate, then you can follow the links from there back to the original research if you wish.

And, just as a commonsense test, where is global warming occurring fastest: the arctic. Not a lot of cities up there.

Sixth, if you want another test, please note that the satellite data, over the past 30 years, also show substantial warming (though how much depends on whose reconstruction of temperatures you look at.) No cities in space, either.

So, when I look at this issue of Watts perseveration on the US historical climate network, I see the following:
1) Three-quarters of the earth's surface is water. The oceans are warming. The highest rate of warming is in the arctic. No cities in either place. Urban areas are a small fraction of the earth's surface and the calculations reflect that. Not perfectly, but to the extent possible.
2) Watts made his point, NOAA tested it systematically, and it made no difference. His "good or best" stations gave the same trend as the entire climate network. He can continue to publish pictures of the worst examples, that's his prerogative. But the systematic analysis has been done.
3) The new climate reference network (all in remote areas, all with state-of-the-art instrumentation) gives the same average US temperature anomaly as the historical climate network (the stations Watts looked at). I suspect that Watts et al. will continue to ignore that, but the data's out there if you care to look.
4) Each of the major US global temperature products addresses urban heat island and other data defects in its own way. GISS, for example, simply forces urban trends to match rural trends. If urban = city, there is no urban heat island effect in GISS, by definition. NOAA has an explicit data-based adjustment.

As an afterthought, my favorite pet peeve: You can either bash them for poor incoming data, or you can bash them for what they do to clean up the poor incoming data and make it consistent, but it's just friggin' illogical to bash them indiscriminantly for both. Pick one. If the incoming data have problems, then they need to make adjustments, and the adjusted data WILL differ from the incoming. Because, if the incoming data have problems, the adjusted data HAVE TO differ from the raw data -- otherwise you wouldn't have dealt with the problems. The mere fact that the adjustments make a difference, or make a difference in one direction on average, is not evidence of anything other than that the incoming data have problems and require adjustment. You need to know a whole lot more, and do a whole lot more, to show that an unadjusted series is preferable to the adjusted series. You can't logically maintain that the unadjusted data are better but the unadjusted data are crap. Pick one.

---
NO MORE LAUNDRY LISTS, PLEASE

For the rest of it -- that a scientist somewhere would say something, that a graph from some source shows a medieval warm period, and so on, that's not worth much. You really need to put all the evidence on the table and make a judgment. I mean, for that one scientist who says it's a crock, if I quote another one who says it's not, does that make us even? What kind of logic is that?

Which gets me back to my original point.

I've wasted yet another hour responding to yet another one of these laundry list posts. I've stopped doing that for TimBikes and Uforya, because that's clearly a rhetorical technique for them -- it avoids having to focus, logically and clearly, on any one issue. This is the first time I've seen you do this, so, OK, I got suckered in one more time. I'm not going to do it again.

If you want to talk about any one point, and discuss that in a logical, fact-based manner, putting all the evidence on the table, I'm willing to play ball. But if you're planning to be another TimBikes or Uforya, where you want to jump all over the place, all the time, with laundry-list posts, nah, forget it. I've put in my due diligence on this one post, I'm not doing that any more.

 

Ah, well, I can't help myself. There's a whole thread on the Jones interview distortion here. You really ought to read what he actually said, in the published BBC interview, as opposed that got published as a newspaper propaganda piece. They aren't even close to the same.

For example, for the past 15 years, what Jones said is that the warming trend barely misses passing traditional (5% confidence level) tests of statistical significance. He didn't say "no trend", he clearly took pains to give a full answer.

In my thread on the Jones interview, posted on Priuschat, I actually downloaded and looked at the data. What a concept, considering that the entire discussion is about the data.

I'll summarize, based on my analysis of statistical significance. (For you tech-heads out there, just straight OLS with no correction for possible serial orrelation of the error term. In other words, the simplest possible calculation.)

If they'd asked Jones the "since 1995" question about NASA GISS, the answer would have been: Yes, statistically significant warming.

If they'd said "1994" instead of "1995", about the Hadley data, the answer would have been : Yes, statistically significant warming.

So that's clearly a fairly fragile result. I mean, next year, when the results (likely) are "statistically significant", are you going to change your mind about global warming. No, of course not, you'll forget about this and find yet some other objection.

In fact, the 15-year-trend test is useless. It doesn't tell you anything.

Why? Just ask whether or not it reliably tells you whether or not global warming is occurring. A test that gives you the wrong answer most of the time is a lousy test. That's not rocket science.

So, we all agree that the NASA and Hadley data show global warming. By that, I mean that if you accept those data, it shows global warming. Right? No argument on that conditional statement?

OK, then, if you go to that dataset that (if taken at face value) clearly shows global warming, pick 15-year intervals at random, how often would you NOT find a statistically significant warming trend? Answer, more than 60% of the time, you would NOT find such a trend in the data.

The upshot is that 15 years is typically too short an interval to show anything reliably out of the annual temperature data. The "noise" of year-to-year fluctuation is stronger than the roughly 0.02 degrees centigrade per year "signal" of long-term warming. You need more data to separate the signal from the noise.

So, to make this clear:

• We have a dataset that clearly shows global warming (when taken at face value).
• We have a test, supposedly for warming, that gives us a NO answer more than 60% of the time.
• Is that a good test for global warming? No.

 

Your references are typically orth listening to, IMHO, even if I don't look at all. But nothing from that A23A reference, or the truncated URL for that, will play for me.

 

I don't normally do this, but here's the entire article, quoted, at the end. If you can figure out what was postulated, what was retracted, and what this all means, you're a better man than I.

Al Gore's wall of seawater? Never heard that one before.

BTW, 7 CM = like 3 inches. So, whatever the reason, they ain't withdrawing it 'cause it's a "wall of seawater".

Oh, wait, I must be doing that "actually reading and understanding the links thing" that gets me into trouble elsewhere. Or was it declasse' of me to translate CM to inches or something?

But all kidding aside, I can't make head or tail of this. I suspect that this was withdrawn for issues related to the technical details of the analysis, rather than the conclusion.

Looks like they got a plausible answer, but not for a defensible reason. Is that so?


SECOND EDIT: Finally figured it out. The material mentioned at the top of the article has nothing to do with the withdrawn paper. The withdrawn paper is discussed in the middle. They bring in another, separate paper that was *not* withdrawn, toward the end. They have two different references to the withdrawn paper, with two different text labels. It really is every bit as much of as mess as it seems to be.

I clicked the link for the withdrawn article and read the abstract. They definitely predicted at least 3" of sea level rise in the next century. Here's literally what the abstract of the article in question says:

"In response to the minimum (1.1 °C) and maximum (6.4 °C) warming projected for AD 2100 by the IPCC models, our model predicts 7 and 82 cm of sea-level rise by the end of the twenty-first century, respectively. The range of sea-level rise is slightly larger than the estimates from the IPCC models of 18–76 cm, but is sufficiently similar to increase confidence in the projections."

OK, I THINK I FINALLY UNDERSTAND WHAT THEY SAID:

The article in question, Siddall et al, was published late last year in Nature Geoscience. It predicted a minimum 3" of sea level increase. This paper has been withdrawn.

There was a different paper, published at almost the same time, Proceedings of the National Academy of Sciences, by RS. Rahmstorf and M. Vermeer, that, by contrast, predicted a minimum of more than 2 feet of sea level increase.

That PNAS paper (2 feet) has *not* been retracted, the Nature Geoscience (3 inches) *has* been retracted.

If you bother to read the retraction (there I go again, actually looking at the content of the references), the scientists who pointed out the problems with the paper that was withdrawn were -- S. Rahmstorf and M. Vermeer. The guys withdrawing the paper agreed that Rahmstorf and Vermeer were correct (in the errors that they pointed out).

Upshot: This actual content of this posting is 180 degrees contrary to the spin with which it was posted. The actual content of the article is this:

Two projections of future sea level increase, based on analysis of the paleological data, were recently published in peer-reviewed journals. One of them (Vermeer/Rahmstorf, PNAS) predicted a minimum increase over the next century of more than 2 feet, the other (Siddall et al., Nature Geoscience) predicted a minimum of more than 3 inches. Clearly, those papers disagree. The paper predicting minimum 3 inches was withdrawn, because they agree that they made a mistake in their calculation. That mistake was pointed out by the authors of the first paper. The paper predicting minimum of more than 2 feet stands.

The upshot is that, of these two, the one that says the IPCC under-predicted likely sea level increase stands.

As a final note, I can't help but point out that Rahmstorf is, in fact, a regular contributor on realclimate. Conspiracy theorists, please keep your shirts on -- note that the authors of the second paper agreed with the Vermeer/Rahmstorf critique of their work.


Here's the original referenced article in full:

"
Scientists have been forced to withdraw a study on projected sea level rise due to global warming after finding mistakes that undermined the findings.
The study, published in 2009 in Nature Geoscience, one of the top journals in its field, confirmed the conclusions of the 2007 report from the Intergovernmental Panel on Climate Change (IPCC). It used data over the last 22,000 years to predict that sea level would rise by between 7cm and 82cm by the end of the century.
At the time, Mark Siddall, from the Earth Sciences Department at the University of Bristol, said the study "strengthens the confidence with which one may interpret the IPCC results". The IPCC said that sea level would probably rise by 18cm-59cm by 2100, though stressed this was based on incomplete information about ice sheet melting and that the true rise could be higher.
Many scientists criticised the IPCC approach as too conservative, and several papers since have suggested that sea level could rise more. Martin Vermeer of the Helsinki University of Technology, Finland and Stefan Rahmstorf of the Potsdam Institute for Climate Impact Research in Germany published a study in December that projected a rise of 0.75m to 1.9m by 2100.
Siddall said that he did not know whether the retracted paper's estimate of sea level rise was an overestimate or an underestimate.
Announcing the formal retraction of the paper from the journal, Siddall said: "It's one of those things that happens. People make mistakes and mistakes happen in science." He said there were two separate technical mistakes in the paper, which were pointed out by other scientists after it was published. A formal retraction was required, rather than a correction, because the errors undermined the study's conclusion.
"Retraction is a regular part of the publication process," he said. "Science is a complicated game and there are set procedures in place that act as checks and balances."
Nature Publishing Group, which publishes Nature Geoscience, said this was the first paper retracted from the journal since it was launched in 2007.
The paper – entitled "Constraints on future sea-level rise from past sea-level change" – used fossil coral data and temperature records derived from ice-core measurements to reconstruct how sea level has fluctuated with temperature since the peak of the last ice age, and to project how it would rise with warming over the next few decades.
In a statement the authors of the paper said: "Since publication of our paper we have become aware of two mistakes which impact the detailed estimation of future sea level rise. This means that we can no longer draw firm conclusions regarding 21st century sea level rise from this study without further work.
"One mistake was a miscalculation; the other was not to allow fully for temperature change over the past 2,000 years. Because of these issues we have retracted the paper and will now invest in the further work needed to correct these mistakes."
In the Nature Geoscience retraction, in which Siddall and his colleagues explain their errors, Vermeer and Rahmstorf are thanked for "bringing these issues to our attention".