Erhmm… it’s the IPCC AR4 projections that are falsified!
:)
:)
There seems to be some confusing about just which IPCC projections were falsified. In particular Tamino is misinformed. This misunderstanding is likely due to lack of clarity on my part. So, I’d like to clarify: It’s the AR4 predictions that are falsified by recent data — not the TAR.
So, criticism that, somehow, I have failed to falsify the TAR is….what is the word well, odd.
It is certainly is true that I did not falsify the TAR. I haven’t even run a hypothesis test on it. (Moreover, I’ve told some eager readers repeatedly that, no we can’t just pick 2001 or 1998 and falsify the TAR. We need to have a basis for a date. Moreover, I don’t see why I would bother to even try to falsify the TAR. Common! I’ve shown over and over that temperature rose during that period!)
It is true that some readers are asking me questions about the TAR, they are asking me about Rahmstorf 2002, and I am beginning to read the TAR, ask questions of the TAR and familiarize myself with analyses of the TAR.
However, the falsification I posted applies to the projections in the AR4. 
So, for the bullet points:
- My hypothesis test is of the AR4 not the TAR.
The projections are shown on a figure from the AR4, as illustrated on the figure included in my previous blog post:
Readers likely didn’t guess this, but of all the possible figures in the AR4, illustrating the same short term prediction, I selected that particular one because it shows observations in addition to predictions. (This is actually rather important in context of some published “verification” which rely on the popular “slide then eyeball the data” method required by some prestigious journals.)
Recall that prior to posting the figure showing the falsification, I explained my reasons for selecting specific dates to use when performing a verification in What Are The IPCC Projections? And How Not to Cherry Pick. Selecting a figure with data on it was a “short cut” for emphasizing the idea that it’s best to test projections against later data.
- The AR4 does, indeed, predict 2C/century for the beginning of this century.
It is worth nothing that IPCC does not expect warming rates to remain constant forever: they are expected to increase with time.
In addition, the IPCC modify their predictions as the “not - frozen- simple- predictive- models” are first improved and then tuned. So, different warming rates and their short term projections apply to different time frames and different warming rates are predicted in different reports.
So, in that context of the “blog-climate-wars” it is useful to compare predictions in the TAR to those in AR4.
The TAR predicted a warming rate of 1.5 C/century for 1990-2005. This is discussed in the AR4, under in Understanding Near Term Climate. TS 5.1 page 68:
Projections for 1990-2005 carried out for the FAR and the SAR suggested global mean temperature increases of about 0.3C and O.15 C per decade, respectively. [....] Projections for the TAR were similar to those of the SAR. [...] The results are comparable to the observed value of 0.2C per decade shown in Figure TS 2.6, providing broad confindence in such short-term projections.
So, indeed, we see that the TAR predicted 1.5 C/per decade from1990-2005.
Several paragraphs later, discussing this century, the AR4 states:
“Committed climate change due to atmospheric composition in the year 2000 correspondds to a warming trend of about 0.1 C/decade over the next two decades. About twice as much warming (0.2C per decade) would be expected if emissions were to fall within the range of the SRES marker scenarios. This result is insensitive to the choice among SRES initiatives. “
(Ref: Understanding Near Term Climate from the Technical Summary. TS 5.1 page 68: )
So, we see AR4 projects 2C/century of warming and applies for first few decades of this century not 1990-2005. In graphs, we see that the AR4 expect the warming rate to accelerate over time.
Clearly, it makes no sense to test the AR4 projection of 2C/century after 2000 against data from 1990-2000. TAR hind-cast/projections values that apply to 1990-2005, an earlier time frame, when less warming would be expected based on the theory of AGW itself.
So… 2C/century and after 2000 is the correct choice for the AR4
So, we see that the 2C/century short-term projection, was made by the climate scientist writing the AR4, and it applies to after 2000, not 1990-2000. This is one of the reasons testing against data for this century is appropriate.
In contrast, a prediction of 1.5C/century was made by the authors writing the TAR, and it applies to 1990-2005.
I have not run a hypothesis test to see if the TAR applied from 1990-2005. Notwitstanding Rahmstorf et al 2007. I don’t plan to.
The IPCC already thinks the predictions for the TAR were consistent with 1990-2005 data, and say so in the AR4. So, while some may agree with “The Rahmstorf Seven”, it appears that group did not sway the writers of the AR4 to announce that the TAR under-predicted warming.
So, it appears that, despite testy sounding language, Tamino and I agree:
- I did not falsify the TAR. (I didn’t even try.)
- Longer time frames do reduce uncertainty intervals.
- One can use statistics to test hypotheses. (This is actually a break-through based on some other criticism which suggests that we can’t.!) and
- It is reasonable to elevate “The IPCC is right” to null hypothesis making it difficult to dislodge. (This is important. Where it not so, we wouldn’t even be arguing over these α errors!) and (though this may come as a shock to Tamino)
- I did not account for uncertainty in the correlation coefficient in residuals.
I said this myself when I first performed the falsification:
It is also worth nothing that even though Cochrane-Orcutt widens the uncertainty intervals significantly, they may still be to small. The reason is that the uncertainty intervals I posted do not account for the uncertainty in the estimate for the serial autocorrelation on the uncertainty in the trend, “m”.
Why didn’t I correct? Simple: I don’t know how. I’ve told my readers. I’m teaching myself some of these methods because, for some reason, those wishing to convince skeptics the current period is not unusual simply refuse to quantify.
They want to wave their hands at graphs showing dips after volcanic eruptions. They want to talk about 1998 being a ‘high’– the result remains if we start after 1998. They want to talk about La Nina — sure it’s a factor. But do you have to make up a super-powerful 12 year long La Nina?
The only way to get this done is to read up on a standard method, do it, and post.
Evidently, Tamino has run the correction for uncertainty for the auto-correlation and my hats off to him for that. (And, Tamino, thanks for the reference. Now I can look it up.
)
Points of Agreement and Disagreement with Tamino
Though we have not corresponded personally, based on what’s written at his blog, I suspect we disagree on these issues:
- I think it’s fine to test the AR4. If more sophisticated analyses are required to get the uncertainty intervals large enough, but that’s not an insurmountable hurdle. Just do the analysis. Just them or stop telling people they have to believe you.
- I think one should test on merged data. I will be posting the results for the averages from the five major measurement groups.
I think showing the result for the merged group is required to reduce the instrument noise — which is not inconsiderable. Keeping the noise high without justification is sneaky way to avoid falsifying, and many, many people know this.
Moreover, posting the results for the average data must be done to prevent doubters from believing data are cherry picked for a certain result. (And bear in mind: there are doubters on both sides of the fence!)
- I think the idea that predictive models were “frozen” in 1990 is silly. Stop saying that. And cut out the mumbo-jumbo about physics. No one thinks the models are based on astrology, or the outcome of a darts competition between various programmers at GISS. Tuning is done. The IPCC documents admit the models used to “project/predict” are tuned. Ultimately, data is involved– and that includes temperature data collected after 1990!
So… right now… I still think we have a falsification
Yes. My hats off to Tamino on adjusting for autocorrelation.
But until he does this with merged data (or I read the reference and learn how) I’ll still think we have a falsification of the AR4 projection of 2C/century for this decade on our hands!
While I know many wish to insist that everyone must not only grant the IPCC the null hypothesis, and grant the IPCC an α no greater than 5%, and perform only two-tailed t-tests rather than a one-tailed t-tests, we must give even give the IPCC the benefit of the doubt when dancing around rejection with α=5%.
This is not so. No one is required to give the IPCC the benefit of the doubt on thsi. In fact, there is no rule that those who think the IPCC has never shown these simple models project to within 1C/century need grant the IPCC &alpha<= 5%. So, be fair with the 5%, or tough out the cold spell and wait for El Nino.
I'm in Chicago, and let me assure you, I don't like cold-ish spells. But I still know, that, right now, it’s cold relative to the last few years. But, just as my daffodils will survive today’s late spring snow, so will the theory of global warming. The only question is: How many blooms? And how much warming?
Comments
lucia (Comment#1410) March 27th, 2008 at 5:57 pm
Thanks Geoff.
It is preferable when people who “rebut” an argument or claim actually rebut an argument or claim that was made by someone. Don’t you agree?
Geoff Larsen (Comment#1413) March 27th, 2008 at 6:51 pm
Yes I agree Lucia. Setting up straw-man arguments is common but often I think it is not done intentionally. Often it is just a difference in perspective with ego sometimes getting in the way.
I think you & Tamino have much to contribute to the debate and you are both extremely good teachers. Sometimes “beliefs” get in the way of rational statistical debate (witness Tamino’s PCA analysis when MBH was brought into the picture; IMO a disaster to Tamino).
lucia (Comment#1414) March 27th, 2008 at 7:12 pm
Geoff– I also agree those arguing strawmen often do so unintentionally. Doubtless, Tamino didn’t intend a post rebutting … well… nothing.
I’m afraid I couldn’t even bring myself to read the PCA analysis. Very few people are interested in statistics for statistics sake. I’m not. Honestly, unless someone has a new result, interesting using the analysis, correctly, at this point, why should I care?
Ian Castles (Comment#1417) March 27th, 2008 at 8:41 pm
Lucia, Your analysis is convincing, and I don’t agree that the misunderstanding on the TAR/AR4 issue can be attributed to any lack of clarity on your part. Your chart posted above, which has been reproduced many times including on ‘Prometheus’ on 15 March, is clearly referenced to the ‘AR4′ with the Chapter, Figure number and page cited. If only IPCC citations were equally specific.
Perhaps the misunderstanding arose because of the appearance online of Rahmstorf7 in the same week as the SPM of AR4 WGI, in which the IPCC claimed observations from 1990 to 2005 enhanced confidence in their models. And Rahmstorf7 quoted TAR projections in support of the proposition that ‘Previous projections as summarized by IPCC, have not exaggerated but may in some respects even have underestimated the [climate] change.’ By comparing the latest IPCC projections (from AR4) with observations since 2001, you have shown that THESE projections have probably over-estimated the warming.
Roger W. Cohen (Comment#1422) March 28th, 2008 at 6:01 am
Lucia
I continue to be fascinated by your work and the interchanges it has stimulated.
Please help me out with this bit of thinking that goes to the point of what exactly has or has not been falsified. It seems to me that the main thing we should be interested in is how well the models predict/project/forecast SUBSEQUENT temperature behavior, not how well they have done in reconciling past behavior. The models can be tuned, adjusted, whatever. At the end of the day their credibility is based on their ability to predict. After all it is those predictions that are the basis for proposals for the various steps to curb emissions. So it makes sense to me that the main thing to look at is how well models have done in predicting global average temperature after they were frozen, not before. It’s like any good business. You make projections in a business plan, then steward against how well you have done compared to how well you thought you would do. So by this reasoning, and if one assumes the models were frozen one year prior to each IPCC report (a somewhat adjustable parameter), one should be concerned about four specific projections:
IPCC 1990 from 1989 to present
IPCC 1995 from 1994 to present
IPCC 2001 from 2000 to present
IPCC 2007 from 2006 to present
Then by this reasoning, can we say anything about:
– which of the four are falsifible
– the performance of models as a whole based on four sample projections
– is there any evidence of improvement
lucia (Comment#1423) March 28th, 2008 at 6:38 am
Roger–
I initially agreed on your dating scheme. However, it turns out that, since the IPCC FAR, SAR, TAR and AR4, rely on previously published documents, it’s necessary to find the document where the prediction was first published. This makes picking the “start” date a bit more difficult.
So, initially, I thought AR4 verification should “start” in 2006 or 2007, I changed my mind when Ian brought to my attention that models to be used were published in some preliminary report. So, from that point, they are “frozen”. (I’ll admit to not actually dredging up the preliminary report that Ian cited, That’s sloppy of me… but I’m a blogger. Hey!)
I’m willing to go with “once the specific models used are frozen we can start verification of predictive ability Though, I also accept your point of view which is: Only when the actual “predictions” are released to the public”.
But I do think that, to start the verification clock before publication of results, one must provide a convincing evidence that the specific models were frozen.
I think, (and I said this before) it’s even fair to have something of a double standard with respect to falsification:
To “verify”, models must both hindcast and forecast. So, one should be permitted to falsify using all relevant past data. However, one should not be able to verify predictive ability using data available prior to the time when models are frozen.
So, with respect to the TAR, if it could be shown to hindcast poorly against data from 1990, that could falsify. But, the TAR 1.5 hindcast the 1990-2001 period certainly doesn’t falsify if we accept the relatively large uncertainty bands associated with 95% uncertainty intervals.
That said, one can’t “verify” that the TAR forecasts well using data from 1990. Or at least, you can’t prove it that way to an audience with critical thinking skills. This is not an obscure statistical matter.
If limited to data after 2001, the TAR forecast looks no better than the denialist forcast of 0C/century. Those are sort of the two relevant hypotheses. (Maybe I’ll run them both on graphs and tables! )
Roger W. Cohen (Comment#1424) March 28th, 2008 at 7:23 am
Lucia // Thanks for the note. I accept that attempting to date from a time models were “frozen” is not practical, and you can even argue it is meaningless or at least undiscoverable. Then it is fair to start “only when the actual ‘predictions’ are released to the public;” i.e, publication date. This is fair to the models for the purpose of falsifying predictive capability because it gives the minimum time for assessment.
The idea one gets to from this is there can be dual criteria for verification:
1. Verification/falsification of Predictive Capability uses only forecasts, and the interval available may be truncated because of ambiguty of start time discussed above.
2. Verification/falsification using all relevant past data.
Clearly for FAR, the two criteria are identical or nearly so, while for AR4, they may be very different.
Now you can go through the various possible outcomes. Your case is for AR4: 2 is falsified; 1 is ?. Logically AR4 is falsified regardless of whether 1 is falsified. But what would we say if 2 is verified but 1 is falsified? The model is still falsified for the purpose of prediction and therefore can’t be trusted as a basis for policy. One can go through the other combinations and draw conclusions for eaach case.
By now my head is spinning so I’ll stop.
lucia (Comment#1425) March 28th, 2008 at 7:48 am
Roger–
In some sense, every thinking person, voter, policy maker, gets to assess amd make up their own mind about sub-questions and broader questions.
I think that with your point #2: Showing a method hindcasts is a necessary condition to have confidence in the model. But, it’s insufficient to give confidence. So, if it doesn’t hind cast, it fails even if it forecasts. You don’t even need to check if it forecasts.
But, if it hindcasts, we can move on to testing #1:
With regard to point #1: To give us confidence, a model must be shown to forecast. There are several possible states now:
If a model is untested, it is not proven to forecast. It is an untested, unproven model. Some can rely on it if they wish, using whatever arguments they prefer. People can make “the model is accurate” their null hypothesis. (This is actually unusual to make “the untested model predictions are accurate” a null hypothesis in science, but it’s the situation we have with regard to consensus forecasts.)
Still, no matter what: the model’s predictive ability is untested and unproven. People who point that out are simply correct.
If a model is later compared to data, and fails to forecast, it is a failed model. Period. It doesn’t matter if it can hindcast.
If a model is later tested and is does not fail, then it might be a good model. However, to decree the model “good”, the β error, tested against some alternate hypothesis of interest, must be low. After all, if we have only 1 data point, the way these things are set up, a good model has α chance of ‘failing’ through pure random chance. This rate of failure is constant no matter how much data we collect. (We have been picking α=0.5)
However, if a model is wrong, but some amount, the chance of showing it’s wrong is β. Obviously, we shouldn’t think “not failing” when there is only data point “confirming”. Not failing is only “not failing”.
What makes sense, in terms of AGW is to always state the β relative to the vigorously advanced alternate hypothesis. The only vigorously advanced alternate hypothesis I know of, that can be stated quantitatively, is “no warming”. (If there are others, let me know!)
Roger W. Cohen (Comment#1426) March 28th, 2008 at 8:47 am
Lucia
I accept the necessary but not sufficient argument for criteria #2. It allows AR4 to be falsified without worrying about the short forecasting period. Now it would be interesting to see if TAR fails criteria #1. It would be strange if TAR does not fail #1 but AR4 fails #2, for that would suggest that models have become worse.
The only ‘vigorously advanced alternate hypothesis’ might be the climate sensitivity of about 1 deg C for CO2 doubling advanced by Stephen Schwartz at Brookhaven based on empirical — not modeling — considerations. (Heat capacity, time constant, and sensitivity of Earth’s climate system. Schwartz S. E. J. Geophys. Res. , D24S05 (2007). doi:10.1029/2007JD008746). Presumably this predicts about 0.5 deg C 21st century warming, which is going to be hard to distinguish from zero. It is perhaps worth noting because it admits some but not much GHG warming. Other work suggests a substantially lower climate sensitivity than the IPCC central tendency by implication of a variety of findings that specific natural effects are larger than IPCC allows (solar variability, AMO, PDO, state-to-state transitions of natural ocean cycles, etc.). At this point it seems reasonable to continue to use zero as the alternative.
steven mosher (Comment#1427) March 28th, 2008 at 9:30 am
Lucia, I apologized to Tamino for MY misrepresenting his “you bet” posts. He also apologized for misunderstanding
that you were falsifying the AR4 as opposed to the TAR.
You should link to his apology which is in response to a post of mine on his site.
And then I would thank him publically on his site and thank him here. and links to all.
Frank O’Dwyer (Comment#1435) March 28th, 2008 at 2:47 pm
Lucia,
What is the material difference between TAR and AR4 projections, for your purposes, exactly? IIRC both of them project about 0.2C/decade.
The side point about 1990 vs 2001 is also a red herring, because Tamino’s post has in any case shown that the data from 2001 onwards don’t falsify a rate of 0.2C
Tamino has also made the point that ‘about 0.2C/decade’ is only 1 significant digit, which does not mean 0.2 bang on the nose, but may mean anything from 0.15 to 0.25. Using the actual data from the model runs he finds the real projection to be 0.0174.
Tamino’s post is here.
(By the way, this ‘It is preferable when people who “rebut” an argument or claim actually rebut an argument or claim that was made by someone. Don’t you agree?’ is rich in irony given your earlier rebuttals of claims not made by William Connelly, Boris, Tamino, and myself - but never mind!)
lucia (Comment#1436) March 28th, 2008 at 3:06 pm
Frank–
What do you mean by “what is the material difference?” I believe I discussed them above.
As for your suggesting that somehow the predictions are the same. authors of the AR4 disagree with you on that. They think AR4’s central prediction is 2C/century and the TARS is 1.5C/century. I have quoted the text written by the authors of the AR4, above, and provided a link to the document.
With regard to Tamino’s post: it has several shortcomings. With respect to his discussion of 2.0C/century. My results are posted here. They were written before Tamino wrote his post and there is nothing he has said that has led me to believe I should treat the numbers otherwise. (I have read the citation for the pumped up uncertainty intervals for the OLS and will comment further, later. But, I’m afraid you will need to be patient.)
Which claims not made by Connelly, Boris, Tamino or yourself did I rebutt and where? As far as I am aware, I commented on specific things each of you stated, generally here in comments but sometimes elsewhere.
It is true that I seem to have misunderstood one of Boris’s claims. He has a tendency to try to phrase claims in the form of rhetorical questions. We did discuss this after I misinterpreted his claim and made a statement in my post. However, my post was not specifically a rebuttal to him.
Otherwise, if I rebutted an argument that was not advanced, let me know. We can discover what the problem was. As you can see, we all agreed in comments that, quite often, these things happen accidentally.
Frank O’Dwyer (Comment#1437) March 28th, 2008 at 4:14 pm
Lucia,
“What do you mean by “what is the material difference?””
I mean what is the material difference between the projected trends. See below.
“As for your suggesting that somehow the predictions are the same. authors of the AR4 disagree with you on that. They think AR4’s central prediction is 2C/century and the TARS is 1.5C/century. I have quoted the text written by the authors of the AR4, above, and provided a link to the document.”
The text you quote actually says that the TAR projection was “similar” to the SAR which in turn is “about 1.5C”. That doesn’t mean it equals 1.5C.
The average TAR projection per the models turns out to be 1.7 which is indeed similar to about 1.5C. It’s also “about 2C”.
lucia (Comment#1438) March 28th, 2008 at 4:29 pm
Frank…
And your point is?
Frank O’Dwyer (Comment#1439) March 28th, 2008 at 6:59 pm
Well for a start, it is that Tamino’s argument is not affected by the difference between TAR and FAR projections at all (since both are ‘about 2C’, and in any case ‘about 2C’ is the figure he starts from).
Nor is it affected by whether the start point is 1990 or 2001 (he does the analysis for both).
What then is left of your rebuttal to Tamino above? Not much. Just the disagreement re ‘merged data’. And I don’t think you have addressed his point regarding that, which is that the measures do not all measure the same thing.
And you have still not addressed his contention that ‘about 2C’ is not the same as ‘exactly 2C’.
Phil. (Comment#1441) March 28th, 2008 at 7:42 pm
Frank what the AR4 says is:
AR).
The multi-model mean SAT warming and associated uncertainty ranges
for 2090 to 2099 relative to 1980 to 1999 are B1: +1.8°C (1.1°C
to 2.9°C), B2: +2.4°C (1.4°C to 3.8°C), A1B: +2.8°C (1.7°C
to 4.4°C), A1T: 2.4°C (1.4°C to 3.8°C), A2: +3.4°C (2.0°C to
5.4°C) and A1FI: +4.0°C (2.4°C to 6.4°C). It is not appropriate
to compare the lowest and highest values across these ranges
against the single range given in the TAR, because the TAR
range resulted only from projections using an SCM and covered
all SRES scenarios, whereas here a number of different and
independent modelling approaches are combined to estimate
ranges for the six illustrative scenarios separately. Additionally,
in contrast to the TAR, carbon cycle uncertainties are now
included in these ranges. These uncertainty ranges include only
anthropogenically forced changes.
Most of those scenarios have means significantly above 2ºC (only one slightly below).
They predict very much less variability over the different scenarios over the couple of decades following 2000:
There is close agreement of globally averaged
SAT multi-model mean warming for the early 21st century for
concentrations derived from the three non-mitigated IPCC
Special Report on Emission Scenarios (SRES: B1, A1B and
A2) scenarios (including only anthropogenic forcing) run by
the AOGCMs (warming averaged for 2011 to 2030 compared
to 1980 to 1999 is between +0.64°C and +0.69°C, with a range
of only 0.05°C).
lucia (Comment#1442) March 28th, 2008 at 8:18 pm
Frank,
What rebuttal to Tamino? I said I have other issues with Tamino’s post, but I will be addressing them later. Tamino posted yesterday. He didn’t link me. I don’t normally read his blog.
From a reader, I learned of his post, which mentions my name, but does not specifically state he is responding to me.
In comments some of my readers asked me to clarify what my analysis pertains to, and some of the difference in my choices, so I did. As that question was motivated by Tamino’s analysis, which at least seemed to be in reference to mine, but discussed the TAR numbers and time frame. I referred to it and linked. I will likely comment on other analytical choices at a later time. That’s what I meant to communicate when I previously said:
(I have read the citation for the pumped up uncertainty intervals for the OLS and will comment further, later. But, I’m afraid you will need to be patient.)
As for statements like
And you have still not addressed his contention that ‘about 2C’ is not the same as ‘exactly 2C’.
I also have not addressed Bill Clinton’s philosophical observation about the meaning of “is”.
I’m sure you wish to make some coherent point with “1.5 is about 1.7″ and “1.7 is about 2.0″ so 1.5 is about 2.0″. But when I read that, I simply want to continue with: “2.0 is about 2.3″, and “2.3 is about 2.5″. And soon, we can prove that 1.5 is “about” 1,000,000, though the two are, quite obviously, not exactly equal.
But surely that is not what you wish to say. I’m sure Tamino could not have meant to say any such thing. Since nobody is saying that, no rebuttal is required.
Avfuktare vind (Comment#1443) March 29th, 2008 at 4:04 am
The idea that “…predictive models were “frozen” in 1990…” requires that no model decisions were done since 1990, as the choice of one set of parameters and equations over another obviously would be influenced by the models fit to real numbers. The IPCC says the models have been developed since then, hence they were not frozen.
I think the real falsification of the IPCC models can best be done with the accumulation of heat, or lack thereof. Neither the atmospere nor oceans gained heat from 2004 as measured down to -700 metres, nor did we see a significant increase in latent heat. (Probably there has been no accumulation of heat in the climate system since at least 2001.) This lack of heat accumulation flies in the face of claims of a current radiative imbalance of 0,85 +/- 0,15. From that we know that the AOGCMs the IPCC prefer is falsified, with absolute certainty as they do not allow for a radiative imbalance at or below zero. (E.g. see Smith and Reynolds, 2005, Willis, et al, 2008, Lyman et al 2006 and of course AMSU data from RSS and UAH.)
Unfortunately it doesn’t tell us what the actual rate of increase due to AGW will be, nor if the natural factors obviously in play will dominate the next century or not.
steven mosher (Comment#1444) March 29th, 2008 at 4:52 am
for reference AR4 estimates that the LOWEST warming expected for 2011-2041 will be .64C or roughly .21 C per decade
and the text indicates about .2C for two decades following 2000. makes of it what will
Frank O’Dwyer (Comment#1445) March 29th, 2008 at 6:49 am
Lucia,
“What rebuttal to Tamino?”
Indeed. If you admit nothing up there rebuts his argument I certainly agree that it doesn’t.
Yet you still claim a ‘falsification’ when his argument is that you don’t have one.
“I’m sure you wish to make some coherent point with “1.5 is about 1.7″ and “1.7 is about 2.0″ so 1.5 is about 2.0″.”
Strawman.
I didn’t even write the words you attribute to me there, yet you use quotation marks to imply that I did. You have simply butchered what I did write, completely altered the meaning, and added an implication (’so’) that appears nowhere except in your mind. This suggests reading comprehension issues on your part, and that is being charitable.
It is preferable when people who “rebut” an argument or claim actually rebut an argument or claim that was made by someone. Don’t you agree?
Roger W. Cohen (Comment#1446) March 29th, 2008 at 10:05 am
Lucia // I would like to return to statistical issues for a moment. We had corresponded earlier about the possibility that multi-year weather issues in such a short timeframe might corrupt the assignment of a trend to an underlying climate change. In particular the ENSO cyle is a candidate for such a contamination. Visually there is evidence of a cyclic contribution to the data, and the ENSO index had been trending downward during this period. This results in high month-to-month correlations, thereby increasing the uncertainty of the OLS derived slope. In effect the Cochrane-Orcutt method allows for such additional factors to affect climate by providing a substantially larger uncertainty in the regressed slope. It is a tribute to the magnitude of the discrepancy between the IPCC models and the actual trend that the difference still allows rejection of the models.
An alternative approach is to try to model the ENSO cycle and use it as a control variable in a multivariate regression scheme. To see how this works, I used the published NOAA MEI (Multivariate ENSO Index) as the control variable in the model:
T = a + b t + c MEI,
where T is the recorded global temperature anomaly (I used the average of RSS and HadCRUT3), t is time, and MEI is the index. During the period Jan 01 through Feb-08, the index varied from about -1.3 to + 1.3 units and exhibited a downward trend of about 0.45 unit over the period. The cross correlation of T with MEI is highest for a temperature lag of 2 months, so I used that in the model. The following summarizes the results, comparing the base OLS results (without MEI) with the model including MEI as a control variable:
Base (No MEI) With MEI
Slope “b” degree/century 0.83 0.65
Slope “c” degree/MEI-unit N/A 0.094
r^2 0.03 0.33
Mean square residual 0.0093 0.0064
Estimate of error in “b” 0.51 0.42
(1 standard deviation)
You see that using MEI as a control variable reduces the underlying climate trend, as expected because MEI is trending downd over this period. However the effect is not large, a little over 20%. Also the r^2 is substantially higher, and the residuals and error in the slope estimate are smaller when the MEI control variable is included. I don’t know how to put up a graph on this site, but I can say that the residuals appear to be far less regular with the control variable. Correlations are still present, though significantly reduced, so it is still worth doing C-O on the result.
In summary, while the effect of ENSO in this time period is to slightly exaggerate the perceived underlying climate trend, including it as a control variable also allows its uncertainty to be reduced. It is very likely that the falsification result is strengthened by virtue of the reduced slope uncertainty and the knowledge that a known multi-year weather cycle has been accounted for
.Note: I modify the comment to insert an html table Wordpress stripped it so I pasted the original back in and then formatted the hard way.– lucia
lucia (Comment#1447) March 29th, 2008 at 11:49 am
Frank D:
Yes. I still claim falsification of the AR4. I believe there are shortcomings with Tamino’s analysis method. I have also said twice previously that I have not discussed some of those shortcomings and plan to do so later.
I think most readers will accept that I am not required to post and comment on every aspect of Tamino’s post instantly. My falsification of the IPCC projections on March 10. Tamino posted on March 26.
I feel not the slightest unease permitting myself a few days, or even weeks until I have time to write a decent post discussing the shortcomings in a way that my readers can understand. If you chose to believe this means I have no rebuttal, that’s your prerogative.
With regard to the merged data: I think that is important. I’ve said so. If you think otherwise, you can express your opinion. That’s fine. We disagree.
With regard to your accusation that I am arguing a strawman in response to your “about 2″ issue: When you first brought up this “about” issue, I asked you to clarify your point. You came back with a very brief statement I find similarly opaque: I have no idea what point you are trying to make. In an attempt to get you to clarify, I have explained what your comment seems to claim when I read — and I have read it quite a few times.
When I explained words seem to say, I said that I know you can’t be saying that. Therefor: no strawman.
My point is: if you wish me to know my thoughts on some particular idea in your head, you will need to explain your idea in a way that doesn’t sound like gooble-ty-gook.
Otherwise, if you keep repeating gooble-ty-goook, I’m just going to ignore you. Roger Cohen has has said something both interesting and coherent.
lucia (Comment#1448) March 29th, 2008 at 12:00 pm
Roger–
That’s really cool! I obviously do need to learn some of these other techniques. I plan to update the analysis spreadsheet as I learn an understand these — but it is always nice to have the back-of-the envelop order of magnitude analysis confirmed using a formal, standard method.
(I’m going to edit your comment to make the table appear as a table.
lucia (Comment#1449) March 29th, 2008 at 12:09 pm
Roger: There are two ways to insert images.
1) Host them on some site available to you (flicker etc.) Then, insert the image inside images tags. If you don’t know html, just paste the link and I’ll modify for you.
2) send me the image by email. You can click “Contact Lucia”, and send it.
Bob Tisdale (Comment#1454) March 31st, 2008 at 8:31 am
Roger and Lucia: The MEI by its name implies that vaiables other than temperature are included in its computation. In addition to SST, it includes air temperature, cloudiness, sea-level pressure,and two surface wind components. These may impose a trend that does not exist in the temperature only based ENSO indices such as NINO3.4 or ONI. You may wish to check.
lucia (Comment#1455) March 31st, 2008 at 8:50 am
Bob– Yes. The MEI doesn’t directly account for temperature. I think Roger is trying to do exploratory statistics to see if, given a metric that is supposed to tell us how strong ENSO is, and how many times it reversed from El Nino to La Nina during the period, how does that affect the computation of the trend.
I think the only important is this a good way to isolate the El Nino /La Nina effect on the estimated trend. It should at least be a good way to get an idea if it’s large or small, given the time frame, and the strength of the index.
Erhmm “some” are commenting on the TWO month change in CO2: How not to rebut a blog post. | The Blackboard (Pingback#1564) April 8th, 2008 at 10:25 am
[...] previously noted that a blogger who has taken on the form of a Bulldog “rebutted” my falsification of the IPCC AR4 prediction of 2 C/century warming for the early decades of 2000. How? by performing a entirely [...]
Geoff Larsen (Comment#1408) March 27th, 2008 at 5:48 pm
“I wander’d lonely as a cloud
That floats on high o’er vales and hills,
When all at once I saw a crowd,
A host of golden daffodils;”
William Wordsworth
I’m personally not sure the theory of global warming will survive, at least in its current form.
Nice comeback to Tamino.