Yesterday, I commented on Gavin’s defense of AOGCM models that ‘saved’ their virtue by resorting to some novel notions. Today, I will show that even if we accepted one of the sillier arguments (which, the royal “we” does not), my analysis would still indicate the model-mean trend does not match the observed trend.
Recall, one of the novel analysis described was to: Gavin appeared to make included:
- Test models using made up data for 2009, and conclude that we should ignore the analysis based on 100% real data if the made up cases made the models look not so bad.
Today, I thought I’d show what the normalized error (d*) graphs look like if I use made up data . Like Gavin, I’ll made up temperatures for 2009 using data from (i) 2008 and (i) 2007. These are the same years Gavin chose in his analysis.
Because the earth surface temperatures for Jan and Feb 2009 are already available, I did not guess those; instead, I replaced only months for which data are not available. That is, I replaced March 2009 with March 2007 (or 2008), April 2009 with April 2007 (or 2008) &etc.
(Recall: My analyses use the AOGCM runs available at The Climate Explorer which include some runs not used in the AR4.)
Here’s the graphs using 2007 data to guess the missing months from 2009
Hmm…, 2007 was the “warm” possible guess. So, this is the comparison that should make the models look better.
Yet, somehow the normalized difference between models and trends make the models look quite bad. They look very bad at predicting trends beginning in 2001 and after. Recall, 2001 is worth discussing because that’s when the IPCC published a document describing the scenarios used to force the models. For years prior to that, the ‘prediction process’ as a whole was informed by knowledge of the measured temperatures.
The models don’t even look very good when tested against observed temperatures that were known when the models developed and the SRES were published.
Interestingly, if you either read earlier posts or scan down to the end where I reproduce the results based on 100% real data, you will see this comparison between models and observations looks worse than the one based on the real data alone.
Why do the models look worse than compared to real data? Especially for recent years? Because a) even guessing 2009 ends as 2007, the temperatures will not rise enough to catch up to the trend projected by the models and b) Particularly for recent years, the magnitude of the denominator in d*, is dominated by the estimate of the effect “earth weather noise” on the best estimate for the true underlying trend. So, the denominator tends to increase roughly as the square-root of the number of months in the trend. Adding 10 months to 98 months makes a big difference to the magnitude of d* starting in 2001!
Here’s the graphs using 2008 to guess the missing months from 2009
Recall, the temperatures in 2008 were cooler than 2007. So, of the two ‘guesses’ for 2009 temperatures, this is the “worse” case for the models. Of course, it a little bit worse than the test where we made up 2009 data by filling in with 2008 data.
Here’s what the graph looks like with real data
For your convenience, I’ve included the graphs showing the current results using no made-up data for March-Dec 2009.
As I mentioned: This result makes the models look less bad than the other two results.
So… why does the effect of guessing 2009 will warm to the level of 2007 not make the models look good?
Those who read Gavin’s post might wonder: Why did 1 year make such a difference to that analysis but not the d* one?
The answer is simple: Given the same quantity of data, the d* test has greater statistical power than the Michaels test (which is similar to the test Gavin suggested in a post.) That test seeks to determine whether the earth weather falls inside the range of “all weather in all models”. Due to model biases caused by uncertain understanding of the physics, range of “all weather in all models” is larger than the “earth weather noise” for the earth, whose physics are deterministic.
In contrast, the d* test compares the best estimate of the underlying climate trend to the range of climate trends in models.
It standard practice use statistical tests with more greater statistical power when several are available. Guess why? Because if you use a test with higher power, the total error rate, accounting for both Type I and Type II is smaller. Therefor, analysits whose goal is to make errors as infrequently as possible prefer test with higher statistical power.)
So, what will the analysis look like at the end of 2009?
Beats me.
Weather is pretty variable. It’s entirely possible a Super El Nino will form and temperatures will warm up and Mar-Dec 2009 will be warmer than Mar-Dec 2007. In which case, the earth’s temperature will catch up to the model trend.
Or maybe, temperatures will stay where they are now. Or maybe. . .
Obviously, it’s rather weird to test a hypothesis against predicted observations! But, if we happen to fill in with 2007 or 2008, as Gavin did when criticizing Michaels, the d* test will still say the models don’t look so hot!
When all is said and done: If the IPCC modeling process, which includes projecting forcings, running models and deciding how to weight model projections to obtain a trend did produce an accurate projection in the AR4, then the earth temperatures will eventually get back on track. If and when they do, we will know that the current deviation was a statistical outlier.
If, however, those making projections erred in any of those things, the earth temperature probably won’t get back on track.
That said: currently, statistical tests indicate the earth’s climate trend falls below the range projected by models.
Note: about comments
Due to my travel schedule comments will either be moderated or close on Monday morning.
Lucia,
I think it would be useful for anyone interested in climate models, and in particular the integrity of the data and the parametrization assumptions they use, and thus by extension their credibility as predictive tools, to have a look at Jim Hansen’s presentation at the Climate Change Congress in Copenhagen on March 11. As you will see from the quotes that follow, this goes most in particular for Gavin Schmidt.
On pages 6-8 of his presentation [the Pdf file of the full presentation is available on the web] Hansen makes the following points:
* The IPCC’s aerosol estimates are “pretty much pulled out of a hat” [sic].
* “We do not have measurements of aerosols going back to the 1800s.
We don’t even have global measurements today”.
* “Any measurements that exist incorporate both forcing and feedback”.
* ” Aerosol effects on clouds are very uncertain”.
I don’t know about you, but given their implications I find not just the statements themselves but the fact that they were not reported at all simply mind boggling. Imagine, the values for one of the key variables in the IPCC models “pretty much pulled out of a hat”. From “Death Train” Hansen himself. It beggars belief…
Maybe for this reason alone, we will a continuing and growing discrepancy between the models’ projections and the earth’s temperature trend. 🙂
PS: I hope your dad gets better. Cheers
tetris,
Yes, the current discrepancy could be due to aerosols. If aerosol loading are currently varying in a way that goes from warming toward cooling, that would cause a trend. But… if the modelers expected the opposite… th
The uncertainty in aerosols and the range of treatments in models can do all sorts of things to projection themselves and tests of the projections.
The fact that, for all practical purposes, the uncertainty in the range of aerosol loadings gave the modelers a fairly big tuning knob makes agreement in hindcasts less convincing than some might think. Maybe the model climate sensitivities are too high, or too low with aerosol loadings selected to give the correct temperature range. Or.. maybe something else.
By the same token: If the current cooling trend is caused by increased loadings of the sorts of aerosols that cool, things will get really bad should people decide to stop spewing those out.
Given the winter we just had, I would kind of hope 2009 warms to the level of 2007… except 2007 was a pretty cool summer here in Wisconsin.
By the way, I caught your update on twitter.
Yes, you have a greater chance of getting an incorrect falsification with the d* test, as Santer noted. I’m not sure why you think it’s the best test in this case.
Boris–
Both tests give the same chance of an incorrect rejection. If you pick p=5%. The d* test gives a lower chance of an incorrect ‘fail to reject’.
Unless your objective is to blind yourself to the reality that models are wrong, there is no reason not to pick the model with a higher statistical power. Unless blinded by the desire for a particular outcome, few would fail to see the rational basis for picking the test with the higher statistical power.
Boris–
BTW: The issue Santer discussed is a tiny over-strictness. As a practical matter, things he did not consider (i.e. non linerities in the underlying trend) overwhelm the small effect he discussed at length. Also, even that tiny effect could be corrected for easily by adding an adjustment factor computed based on Monte Carlo tests with synthetic noise. (This is how he showed the issue.) In contrast, the “Type II error” problem is huge.
Boris
It suppose that my flagging Hansen’s outlandish admission that the IPCC’s climate model aerosol data is “pretty much pulled out of a hat” is just more evidence of my being a “conspiracy theorist”.
Hansen’s statement that any aerosol measurements that do exist incorporate both forcing and feedback tells us that if we treat those “measurements” as “facts”, they must be covered in grease.
Just like other “facts” such Steig’s, Mann, et. al. finding a 0.1C per decade increase in Antarctic temperatures with a margin of error +/- 0.3C.
And there are some who persists in defending this as “science”. I just call it junk. GIGO.
Lucia,
Why don’t you comment directly at Real Climate?
Do you think it’s unreasonable to assume that this year will fall between 2007 and 2008? Recall that 2008 was the coolest year this decade (Century) and 2007 was the 2nd(?) warmest.
Nathan-how do you know she would even be allowed to comment? Anyone at CA could tell you their horror stories of having posts deleted, strawman-ified, censored, and/or blocked. The moderation there is very heavy handed. As for your second point, What? Statistically speaking, isn’t it ~always~ more likely that the next point will fall in-between the highest an lowest values? I thought everybody knew that!
Andrew_FL
Well, there you go.. .Giant conspiracy.
Yes, Andrew, that was my point that it wasn’t unreasonable to assume 2009 would fall between 2007 and 2008 – for some reason Lucia seems to be calling it silly:
“Today, I will show that even if we accepted one of the sillier arguments….”
This is off-topic, but it is sooooo Awesome that it can’t be ignored:
http://wattsupwiththat.com/2009/03/26/galactic-cosmic-rays-may-be-responsible-for-the-antarctic-ozone-hole/
Wow… It was cosmic rays all along. Who woulda gussed it?
Nathan,
I rarely comment directly at RC for several reasons. The main ones are:
1) I don’t like to fill captchas, be moderated, go back to check to see if my comment reappeared much later etc. I accept the right of all bloggers to moderate as they wish, but I find this policy pesky. So, generally, I’d rather just post here and people who want to discuss what I’ve said can discuss it here. Blog all over the net have operated on this “I can post at my own blog” principle for many years now. That’s one of the things I’ve always enjoyed about blogs.
2) Most my observations are too long to post in comments anywhere. In particular, I can’t post images to illustrate my points in a comment at RC.
3) Like most bloggers, I have set my blog to ping. If the authors at RC happen to be interested, they are automatically informed by this auto-pings sent out by nearly all blogs, and they can comment. Blog readers who read RC can easily discover the “blog reactions” and visit my blog too. So, posting makes my comments sufficiently visible to satisfy me.
That said, I did post a comment at RC about an hour ago. I probably won’t go back to see if it passes moderation until I return from my Dad’s in Florida.
Of course the temperature falling between 2007 and 2008 is a reasonable guess.
That said: The notion that we can ignore a current statistical rejection obtained with real, honest too goodness data based on guessed data is peculiarly odd.
When you suggest:
You mis-identify what I consider silly. The silly thing is to suggest or imply that the results based on data that have actually arrived should be ignored based on an analysis using made up data.
I describe the specific thing that was “silly” using this idea:
FWIW: If you examine the range of results at RC, Michael’s graph will look more or less the same if the temperatures during 2009 fall in between 2007 or 2008. If the temperatures are as low as in 2008, the models will look a little worse than they do with current data; if the temperature rise, they will look a little better. Uhm… Duh!
FWIW: The main difficulty with the Michaels graph at his blog is he left off GISSTemp, not how the results might change if we add the fake but plausible 2009 data. Had Gavin limited his focus to discussing Michaels leaving of GISSTemp he could have avoided this foolishness.
Andrew_FL
I’ve never had a comment rejected at RC. That said, the fact that they do filter and reject comments at RC bothers me and contributes to my preference not to comment there. But 1-3 above are my main reasons.
I guess you can watch to see if one of my comments shows up tonight, tomorrow …. whenever! 🙂
Tetris:
On pages 6-8 of his presentation [the Pdf file of the full presentation is available on the web] Hansen makes the following points:
I’m a little lazy and pressed for time. Can you provide a link? that would be ever so neat and peachy keen. The thought of James Hansen saying ANYTHING I’d agree with is … well, mind-boggling. THANX
Lucia I disagree.
You say this is “silly”
“…that we should ignore the analysis based on 100% real data if the made up cases made the models look not so bad”.
No he was not ignoring the analysis that’s based on the 100% genuine real data, but rather investigating if it’s sensible to draw the conclusions that Micheals did. And to do this he took a reasonable assumption on what the future data would be. It’s a perfectly reasonable thing to do. It’s not “silly”. Remembering of course that it was his claim that Michaels anaylsis would not work in years earlier than 2008, so he was attempting to show that it wouldn’t work as well in future years either.
You used your test and found that you agree with Michaels, this is not “silly” either. This is why you need to actually discuss it with him.
Nathan–
It is sensible to draw conclusions based on what Michaels did. Why would an analysis based on fake guessed data mean we can’t draw conclusions? In particular, if we make an intermediate guess for 2009, we get more or less the same answer as just stopping with current data.
What I posted in the current blog post doesn’t address Gavin’s analysis by ignoring data. But the previous one discussing type II error does so indirectly.
The claim that we ignore a rejection because we fail to reject with less data (as achieved by ignoring inconvenient data) is silly. I don’t need to discuss this with Gavin in comments to state that it’s silly here at my blog, in a bar, at the dinner table with my husband or anywhere.
He may have been attempting to do this, but he failed. Chip Knapperger who works with Pat Michaels discussed how he created his uncertainty intervals in comments here. Gavin did not replicate that. (Mind you, Gavin may not agree with Chip’s methods–but that would be a different criticism..)
However, with Gavin’s uncertainty intervals, Michaels’s current analysis would fall between the “fake data” analysis using 2007 vs 2008 to guess 2009. Look at the 2SD curves rather than the full range curves, squint and see. Gavin did not show that the results fall apart if you use the fake data. But even if he had, we would not ignore a current ‘rejection’. We are in a region where typeII error is large.
Lucia
What I am saying is that what he did wasn’t “Silly” – you may disagree with his results, but it certainly wasn’t silly.
” I don’t need to discuss this with Gavin in comments to state that it’s silly here at my blog, in a bar, at the dinner table with my husband or anywhere.”
Well, ok. Yet again all you do is reduce your blog to rhetoric. You could actually do something here, and actually engage in a discussion about a serious statistical point.
Nathan–
I know what you are saying: You think two specific notions gavin introduced are not silly.
I think they are silly. I think both notions are silly separately no matter what we discover about the accept/reject conclusion. I’ve explained why I think they are silly; in yesterday’s post, I put that in context of the typeII error.
That said: As I know some, like you, thinks what Gavin suggested is not silly, I showed that the models still look bad even if we were to accept the notion of needing to analysis using fake data was not silly.
I’m not entirely sure why you think posting my criticism of Gavin’s ideas at my blog is not engaging in a discussion of a serious statistical points. It’s true that I may not be discussing the notion that what Gavin did was silly directly with Gavin, but …. so?
Gavin has blog comments organized in a way that makes commenting uninviting. The moderation, delay in posting of comments, and his reply by inline comment and decision to moderate comments as he pleases puts visitors on an uneven footing. Based on comments of RC authors, it’s clear they don’t really welcome negative comments on their posts in comments.
In my opinion, the structure of the RC comments policy inhibits free discussion of differences of opinion and would make it nearly impossible to have a halfway decent discussion of any statistical point.
So, no, I’m rarely going to discuss much of anything in thatforum. Those who have different opinions of that forum will discuss there, and that’s fine with me.
Oh– btw. I was thinking of you as I drove home from the Opera on Saturday. We had another driving rain storm and I was wishing I could send it to Australia for you! (Luckily it didn’t turn to snow until 3 hours after I was safe tucked away in bed.)
rephelan [12660]
Me too, I’m, pressed for time….
Try: www. columbia.edu/jeh1/2009/Copenhagen_20090311.pdf
Once you read through the entire presentation, including Hansen using his grandchildren in a rather convoluted way [not dissimilar to the “death Trains analogy] the whole thing becomes even more grotesque.
I repeat: not one single media outlet reported his comments or dwelt on the implications, in terms of the IPCC’s climate models’ predictive credibility, of the aerosols variable being “pretty much pulled out of a hat”. If Wiki ever needed an appropriate example of GIGO, here it is.
[Boris, oh, Boris…., where is apology or straw man explanation for this piece of grease covered junk science?]
tetris,
WHat is your problem. Yes, aerosols are quite uncertain. You seem to think that the uncertainty could only possibly help the case of the denialists. Keep dreaming.
Lucia,
You’ve made no effort to quantify how often you will get a false rejection with the d* method, yet you are convinced that “the models are wrong”–so much so that you call it “reality.” Yet your analysis is incomplete. And then you have the temerity to suggest other have blinders on because they don’t believe your half-done analysis is convincing. Good luck with improving on this and getting it published. If it helps improve models, I’m all for it. But this Steve McIntyre type argumentation is only good for creating a following of deniers and isn’t very convincing to anyone who isn’t already convinced.
Boris–
It is untrue that I have made no such effort. I have not discussed the issue in any detail in recent blog posts. However, I have discussed the issue many times in the past. I have specifically discussed the fact that, owing to the volcanic eruptions, the Santer test makes too few false rejections for test of trends beginning in 1970. (See here. I have done quite a few tests of the Nychka correction as discussed in Lee &Lund and found that, for instances where the underlying trend is linear, that correction tends to make the test have too few false rejections.
The reality is: On the balance, for periods large numbers of volcanic eruptions, the Santer method creates a slight number of excess rejections. For the more recent trends, the Santer method may result in a slight number of excess rejections. That amount is no where near enough to substantively change the reject/accept results in this post (but when I get back from Dad’s I’ll be glad to super impose the Lee&Lund/Nychka correction so you can see that. As you recall, I used to routinely post using the Lee&Lund/Nychka correction.)
Oh? I think you have it backwards. I think it’s posts like Gavin’s silly one where, ignoring the typeII issue entirely, he makes bizarre claims that we set aside or fail to discuss rejections based on current real data because we get “fail to reject” if we reduce the sample size or guess that next year warms up. (While still getting rejections if next year stays the same temperature or cools.)
That type of argument is such dramatic special pleading only those who really wish to protect the virtue and sanctity of “the models” would possibly advance it. People who have not made up their minds read it and become suspicious of those who are making the silly unsupportable claims.
“the denialists”
Another unexpected use of the divide-and-conquer ad hom?
Andrew
1) Citizen lucia fails to grasp that nothing is “fake” if it is what the models expect it to be. Only data (measured, inferred or guesstimated) that fails to confirm The Models (Blessed be the Planet) is suspect or “fake”.
2) The future will always confirm The Models by definition.
3) Sounds to me like someone is due for a reprogramming visit at her nearest Gaia Adjustment Center.
@Andrew and Nathan:
So far, I’ve still had no comments censored at RC:
http://www.realclimate.org/index.php/archives/2009/03/michaels-new-graph/langswitch_lang/cz#comment-116607
So:
a) I sometimes comment there,
b) I hate the process having to wait and see whether the comment appeared hours later, knowing there is moderation hurdle and knowing that, should someone wish to make an inline comment, the “rebuttal” will *always* be directly next to my comment, but my response to the “rebuttals” will always be separated by numerous other comments on somewhat different subject but
c) I haven’t had any of my own comments not appear and
d) I obviously won’t be checking regularly to see if my comment spawned any further discussion.
In short, I don’t think RC comments are structured for anything resembling “discussion”. They are structured the way I would structure a blog written by a professor for students. In that case, moderation and inline comments make sense. But for bloggers… not so much.
Do their methods resemble those used by the Catholic Nuns here? If so, I suspect I will be able to resist. . .
Lucia and George,
Does the Church of Gaia do this-?
96% of Woodlands graduates took four years or more of Math, and over 68% took four or more years of Science
100% of our students donated 6,500 hours to community service
100% of the Class of 2007 earned admission to 200 colleges and universities.
Woodlands Academy annually enjoys 100% college acceptance.
Or does the Church of Gaia just produce “The Warmers”? 😉
Andrew
Andrew_KY–
I’m going to have to ask my sister about those numbers. We both graduated from Woodlands. She’s on the alumni board. But, based on numbers I know those statistics are accurate… but….
For example: take this one
I suspect the class of 2007 consisted of less than 100 girls. My graduating class consisted of 67 girls. Obviously, each girl was not admitted to 200 universities. If you examine the charming graduation photo here you will see the total number of students at graduation was fairly small. Suppose the class of 2007 consisted of 50 girls. Do they mean, each girl applied to a zillion billion colleges. Each girl was admitted to some and rejected by others. If we sum all admissions, the number is 200 (or greater?) I suspect that’s actually what they mean.
The claim that every single girl was admitted to some college somewhere and attended is almost certainly true. That situation would be typical of Woodlands and I would not be surprised if they could make the claim for every year since 1960!
Still… I’m going to have to give my sister some flak about the 200 universities statement next time I see her.
FWIW: The demographic at Woodlands ranges from uber-ultra conservative Catholic to tree-hugging-send revolutionaries to central-amearica and protest the exclusion of women from the priesthood catholic.
I have no idea where they collective stand on global warming. It wasn’t a big issue when I was in highschool. (I still remember the Illinois state mandated week long section on birth control wedge into some health class. Hilarious!)
1. Irrelevant. Resistance is futile.
2. Andrew_KY must also be absorbed.
Boris-In case you weren’t aware, aerosol forcing is critical to the argument that models confirm anthropogenic causation of twentieth century warming and a high sensitivity is justified. The fact that the time history and magnitude of this forcing is essentially produced “out of a hat” doesn’t bother you at all?
George,
When The Gaia Collective comes to absorb me, tell ’em they’ll find me on the Holodeck having drinks with JLH. lol
Lucia,
I know it is your nature to scrutinize numbers… did not Woodlands help you develop the very skill you are using to scrutinize theirs? 😉
Andrew
Andrew_KY.
Maybe. When I went, the nuns enrolling into classes refused to believe that girls could have really learned algebra in the eight grade, and so put all freshman into algebra. On the first day of class, Miss Curry gave us all a test on the first day of class. Afterwards, two of us sat in the back of the room teaching ourselves geometry while the other girls followed the class lectures on algebra. We were expected to take the tests on the same day as the sophomores and that worked ok.
There may be something about sitting in the back of the room and teaching yourself math. . .
Andrew_FL, tetris, Boris:
It is my understanding that aerosol values were not pulled out of hat (or out of anywhere else) but derived from the following detailed, carefully crafted mathematical model:
T1=Actual temp
T2=Predicted temp
A= forcing effect of aerosols (and other mysterious s**t)
T2 – T1 = A
Again, by definition the models are always right. Just relax and slip the pod under your bed….
George Tobin-ah yes, the inverse method!
Anderson, T.L., Charlson, R.J., Schwartz, S.E., Knutti, R., Boucher, O., Rodhe, H. and Heintzenberg, J. 2003. Climate forcing by aerosols – a hazy picture. Science 300: 1103-1104.
It is worth remembering that we had a red lunar eclipse last year. This is concrete evidence that stratopheric aerosols are very low compared to early years.
gavin made an interesting comment on one of the last threads at RC.
“For the 8 runs I have easily available (GISS-ER and GISS-EH), 2008 temperatures range from 0.12 to 0.59 deg C above the 1980-1999 mean. GISTEMP was 0.19 (0.17 for HadCRUT3v) on the same baseline. ”
So, first comment, he is changing the baseline again to 1980-1999 versus the normal baseline GISS uses of 1951-1980 (so that it will be more difficult for people to check I presume).
Using the more commonly used GISS baseline, the 1980-1999 mean was +0.248C. The GISS 2008 anomaly was +0.435C. GISS Model E ensemble mean temp for 2003 was +0.572C.
So, the 8 runs have a range of +0.368C to +0.838C anomaly for 2008 (quite a range in my opinion).
The low run for 2008 is -0.2C below the 2003 Model E estimate and the high run is +0.266C above the 2003 number (what forcing made temps drop by 0.2C in the last five years? – it certainly wasn’t GHGs which continued increasing at the same rate as previously).
And, the low run is -0.067C below the 2008 anomaly and high run is +0.403C higher than 2008 (average is +0.168C over 2008).
So, I conclude the newest GISS Model Es are adjusting the other forcings again (in the low run, Aerosols and solar would have to decline from -0.2C in 2003 to -0.5C in 2008 for the numbers to work).
– and/or the Models are just way off again.
Discussing aerosols and model fallacies at RC is a waste of time. Most have their head in the sand. I know so because I had the same discussion as Lucia over a year ago at RC (with much less statistics!). Same conclusions, though. And, same tired response from RC. One day they will realize that they are defending the undefensible.
Raven-there are lots of other observations showing that the air has been getting significantly clearer.
http://www.co2science.org/articles/V10/N24/Mischenkoetal2007.gif
Hi Lucia you have been quoted in support of (falsification of IPCC predictions) in support of letter sent to Obama by 120+ eminent climate persons: (congratulatiosn)
* “Surface temperature changes over the past century have been episodic and modest and there has been no net global warming for over a decade now.â€
This is correct using the global average surface temperature. An effective analysis of this issue has been presented at the weblog http://rankexploits.com/musings/category/climate-sensitivity/. However, using the global average upper ocean heat content changes, the warming in the 1990s and early 2000s ended in 2003, so the more rigourous metric for global warming indicated “no net global warming†for 6 years.
Vincent-that is not directly from the Cato Letter-that is from Roger Pielke Senior’s commentary on the Cato Letter. See:
http://climatesci.org/2009/03/31/open-letter-by-the-cato-institute-on-climate-science/
Vincent,
“Hi Lucia you have been quoted in support of (falsification of IPCC predictions) in support of letter sent to Obama by 120+ eminent climate persons: (congratulatiosn)”
Are you being ironic?
Lucia,
well I waited and saw your comment on RC… You didn’t exactly set the world on fire. I was expecting you to actually rebut Gavin’s claims. Do you think the difference between your conclusions and Gavin’s arise because you are using monthly data, rather than yearly?
Nathan-there is no such thing as bad publicity! But see my comment above-lucia is being quoted by RP Sr, not the Cato letter.
Nathan–
Setting the world on fire was not my goal was not my goal when commenting at RC. I was responding to someone who commented on a detail about my analyses. I have already explained to you that I have no intention of posting blog length dissertations commentaries in comments fields at RC for reasons stated above.
I doubt the differences between Gavin and my results arise becuase of the use of monthly vs. annual average data. However, if they do, then use of monthly data is preferred because it has higher statistical power and so has lower overall error rates. (That is: the sum of type II and type I errors is lower.) Both methods have equal type I errors.
Selecting statistical methods to maximize type II error is more subtle than simply throwing away recent data, but when either is done to promote the result one prefers to “prove” both tactics amount to cherry picking. One is cherry picking data; the other is cherry picking a statistical method..
Does anyone have a link to the Cato letter? (I’ll google, but I need to get used to this keyboard in the game room at the clubhouse at Dad’s condo!)
Lucia,
Upon a quick glance your criticism seems directed at the filling in/guessing of what 2009 temps may look like, and how that would affect Michaels’ graph.
If so, then the main point, that such a graph is very sensitive to the endpoint chosen and the dataset used, still stands. Or am I missing something here?
Bart
I found the Cato letter itself:
http://www.cato.org/special/climatechange/alternate_version.html
I see no basis for Nathan’s claim that I was “quoted in support of (falsification of IPCC predictions) in support of letter sent to Obama by 120+ eminent climate persons: (congratulatiosn)”.
1) The Cato letter does not include my name in anyway, shape or form.
2) Cato’s web page does not mention me. As far as I can tell, Michaels does not mention me.
3) The only thing I can find is Roger Pielke Sr. discussing what I have shown here at the blog: The earth’s surface temperature has not warmed recently.
This is a true observation, and I don’t see any reason why people aren’t permitted to notice the recent lack of warming. I don’t see anything wrong with Roger Pielke Sr. pointing out that I have plotted out recent temperature anomalies and compared them to the model predictions.
I’m a bit taken puzzled by Nathan’s at least seeming to insinuate that I should be upset that I happen to have noticed the flat temperatures, posted, and then been quoted. (I say “seem” because I think he is trying to be ironic in his previous comment.)
That some would suggest accurate information may not be uttered, posted or noted is the sort of rhetoric that plants the seeds of skepticism. How? By making it appear that those who believe the underlying trend is for warming require censorship to make their case.
If someone wants to argue about what the recent flat temperatures mean, that’s seems perfectly reasonable to me. If you want to argue the data are inaccurate, that is acceptable — just explain why you think they are wrong.
But if you want to suggest that we are not permitted to notice that the event has occurred…. well, don’t make any claims that your arguments are based on science. One of the very important features about science is that accurate observations trump models predictions.
In science, people get to report what the observations are. Other people get to quote those reports. No one is required to lie about the data simply because a commenter at a blog finds it inconvenient that politicians may learn the temperatures actually have not risen recently. I’m certaintly not going to lie about the data for fear that accurate statements might be quoted by someone somewhere. Even if I did lie beause I feared the truth would come out, my lieing would be pointless because plenty of people would notice the earth’s surface temperatures have not risen recently.
Bart-
Two comments:
1) The result is not “very sensitive” to the end point. Of course one will get somewhat different results depending on what happens in 2009. That always happens in any statistical analysis. With my method (following Santer), “guessing” 2009 as Gavin did will make the models look worse whether I end by replacing 2007 or 2008. With Michaels, the contrary to Gavin’s suggestions,the models will still look poor with Michaels method. (One of the reaons Gavin’s graphs make the models look “less bad” is he added extra curves outside the +/- 95% bands to indicate the full model range. That’s wider than the +/- 95% range.)
2) Because Gavin fails to consider typeII error, he is misintepretting “failure to reject” as “supporting” models. This is a typical undergraduate error made by those who don’t understand what the results of a statistical test even means. So, he seems to be suggesting that something falling inside the +/- 95% bounds somehow means supports the models and is therefor contradicts the current “reject”. Assuming this is what he means (and it seems to be what his readers think he means) then he is making a utterly foolish, ignorant claim. In reality, if we consider the power of this statistical test, we would anticipate failing to reject to occur even if the models are wll off track.
The fact that, using real honest to goodness data we are observing “rejects” so shortly after the 2007 publication of the AR4 when the projections were actually dissemimated to the public is very meaningful. So what if the predictions weren’t already wrong when the AR4 was published?
Lucia,
I thought RPS was putting words in your mouth when he said you justified the “Surface temperature changes over the past century have been episodic and modest” part of the sentence, as I’m not sure any analysis you have done touches on that statement (judging the “modesty” of 20th century warming requires fun paleo analysis, and I’m not really sure what they mean by “episodic”).
I hope your father is doing well.
Zeke–
Dad is doing GREAT! (It was a thyroid problem. We got that diagnosed, and he was better within 4 hours of taking his thyroid medication. It’s amazing.)
I guess I didn’t read what he said the way you seem to have done. What I think is that Roger Sr. is giving his own interpretation but also pointing to what the graphs actually show about what surface temperature did.
It’s true that I avoid saying whether 0.7C in one century is “modest” vs “catastrophic” or calling temperature changes “episodic” vs. “smooth”. Certainly, temperature went up more rapidly during some decades and less rapidly in others. They dropped during some periods (particularly after volcanic eruptions)– I’d have to have better access to my spread sheets to say what the longest sustained negative trend was! (I generally don’t memorize stuff like that.)
I would say that the Cato letter was written with the intention of making a political statement. At the same time, Pat Michaels did manage to make rather modest statements of fact but chose words to maximize the political impact. Words like “episodic” to tend to suggest that natural variations are very high. The potential that natural variability is large to very large is something skeptcis and stone-cold coolers have been suggesting; so, naturally, they are going to interpret “episodes” as indicating large natural variability. Are they correct? I dobn’t know.
By the same token… isn’t every one currently agreeing that that warming , as measured by global mean surface temperature, is what many would describe as “episodic”? It appears that while I’ve been down here, there has been a JGR letter suggesting that natural variability is very lage and so, the recent episode of non-positive and/or negative trends in GMST (often colled “cooling”) is not-consistent with rapid warming? (I haven’t read it yet. But, I did some blog skimming last night and just saw “spins” in both directions! )
You haven’t missed much in the GRL paper. Pretty much all it does is take some individual runs of a climate model and finds that (surprise surprise) about 5% of them are significantly different from the naive model at the 95% level. I would say that I’m looking forward to seeing you shred it, except you’ll probably think it’s beneath you, and anyway Pielke has pretty much done that already.
Lucia,
Welcome back!
Mike
Jonathan–
By “naive model” do you mean the mean of all models? Do you mean it shows 5% are different from the mean of all models at the 95% level? Really? Uhmm…. using any special test? Because for a correct test 5% of the population are supposed to be different from the mean at the 95% confidence level! That’s what it means.
Lucia, your description is what I think they’re saying (it’s not the only thing they say, but it’s one of the “key results”). But take a look for yourself; the paper is available online at http://www.cdc.noaa.gov/csi/images/GRL2009_ClimateWarming.pdf
Johnathan, Lucia, I have the suspicion that you are speaking of totally different papers. Lucia said JGR, Johnathan said GRL-Johnathan is clearly talking about this:
http://dotearth.blogs.nytimes.com/2009/04/03/cool-spells-in-a-warming-world/
But I’m not sure to what Lucia is referring. I think (but I could be wrong) it may be:
Swanson, K.L., and A.A. Tsonis, 2009. Has the climate shifted? Geophysical Research Letters, 36, L06711, doi:10.1029/2008GL037022.
Although that is also in GRL…
I have seen the phenomenon of one side spinning a result one way and the other side the opposite way. Interestingly, from what I know about Swanson and Tsonis, Swanson probably insisted on inserting his own spin into the article I just mentioned:
“Finally, it is vital to note that there is no comfort to be gained by having a climate with a significant degree of internal variability, even if it results in a near-term cessation of global warming. It is straightforward to argue that a climate with significant internal variability is a climate that is very sensitive to applied anthropogenic radiative anomalies. If the role of internal variability in the climate system is as large as this analysis would seem to suggest, warming over the 21st century may well be larger than that predicted by the current generation of models, given the propensity of those models to underestimate climate internal variability.”
I doubt this is a statement which Tsonis would agree with. Compare these to quotes from the authors in the media:
““When the climate kicks back out of this state, we’ll have explosive warming,†Swanson said. “Thirty years of greenhouse gas radiative forcing will still be there and then bang, the warming will return and be very aggressive.—
http://dsc.discovery.com/news/2009/03/02/global-warming-pause-print.html
““We realized a lot of changes in the past century from warmer to cooler and then back to warmer were all natural,” Tsonis said. “I don’t think we can say much about what the humans are doing,†he added.”
http://www.wisn.com/weather/18935841/detail.html
So here we have a case of one author spinning results one way, and the other another! Coauthors disagreeing about the implications of their research! Ain’t that something!
Lucia it was Vincent who claimed you were quoted in the Cato letter, not I.
“I doubt the differences between Gavin and my results arise becuase of the use of monthly vs. annual average data. However, if they do, then use of monthly data is preferred because it has higher statistical power and so has lower overall error rates. (That is: the sum of type II and type I errors is lower.) Both methods have equal type I errors.”
So how do you explain the difference in your conclusions?
Nathan:
What difference?
a) Gavin decrees results using fake 2009 data are completely different from those using 2008 data. His decree is incorrect. He is forgetting that fail to reject is not “completely different” from reject.
b) Gavin compares the trends to “all weather in all models”. He does this because he’d discussing Michael’s work- which is fine. But, that method has higher type II error.
There is no big difference in results there is a difference in interpretation. Gavin’s interpretation is dubious– for reasons I explained in my article and my point (b).
The difference in your conclusions.
He says it means one thing. You say it means another.
If you think his results are dubious, why not discuss it with him?
It is quite comic that you refuse to discuss your results with the people you are contradicting. Are you uncertain of your conclusions that you can’t actually confront him with it?
Nathan– What do you mean refuse? I explained the reasons I don’t post 3 page long comments in moderated comments at other blogs. I’m happy to discuss things in other forums.
Nathan — why do you always ask Lucia the same question?
Lucia
If you have discussed it with him, what did he say?
Andrew Kennett,
What do you mean? I ask her different questions.
Mostly I am trying to prompt her to engage in this disussion more formally – she posts a lot on this blog but that’s about it – why does she never go further?
Nathan you kep asking “why not discuss it with him?” (as per 12833 and Lucia answers and then you ask “why not discuss it with him?” OK you don’t like her answer but if it bothers you so much why not facilitate rather than repeat? It sounds like “are we there yet, are we there yet …”
(BTW that was a lighthearted comment not intended to be insulting)
Nathan,
What is it about the phrase “I don’t waste time posting comments on blogs with heavy handed moderation policies” that you don’t understand?
Andrew Kennett
I kept asking the same question because she didn’t really answer it the first time. The first time, she told me that she had commented on his blog – but that wasn’t actually asking Gavin, she was commenting on what some other person had said.
And if she’s unhappy with the moderation or speed of replies at RC, why not ask Gavin directly?
It’s just weird, why wouldn’t you want to discuss a point with someone that is in disagreement with you. If your motivation is just to find out what is going on surely asking the person who disagrees with you directly is the one of the best ways to learn.
Raven
So the only method that she can discuss this with Gavin is via RC?
Nathan — by posting here on a lightly moderated freely available blog that is viewed by the GISS and RC crews (some have commented here) and cheerleaders such as yourself the question is being asked in a free and open and public way. Others (such as yourself) can comment/respond as can anybody whose work / public announcements / blogs are being commented on. Other bloggers also link to comments and posts from here. If the question is posted off-line then we end up in a I-said-you-said situation which is useless to us all. Of course people are free to ignore blog posts and this often happens but just repeating “why not discuss it with him?†when the question has been answered is very uninformative.
Andrew Kennet
You misunderstand what I am asking her to do. I was hoping if she engaged in a discussion about why they came to different conclusions she would actually report back. See that way we all learn something. And then perhaps Gavin and Lucia might also learn something.
The point is that Lucia and Gavin used much the same data (when I asked her if she thought the difference was monthly vs annual she said no), but have come to very different conclusions. So I ask why is that? Lucia couldn’t answer that and so I asked her why she hadn’t asked Gavin. Surely that is the next logical step? Why are people content that there is such a discrepancy? Why don’t people want to find out any further?
It does seem typical of her posts that she reaches the point of mystery, then stops. Surely a scientific analysis would demand that she explore the mystery. Isn’t that the point of science?
“Of course people are free to ignore blog posts and this often happens but just repeating “why not discuss it with him?†when the question has been answered is very uninformative.”
Now this is not very fair. Lucia explained why she wouldn’t post on RC, she never explained why she wouldn’t just ask Gavin. I could say to you that just responding to my question with the same line “Lucia doesn’t respond on heavily moderated blogs” – it’s not really an answer is it?
Nathan I’m not so sure I do misunderstand and while I prehaps shouldn’t answer for Lucia (especially while it is night in the USA) maybe I’ll try. This blog is a place for Lucia to run analysis of climate data and models and to comment on other people doing similar things. Making these analyses and comments public allows others to comment, to comment on the methods and the calcs. If an analysis is done in response to another public analysis, either blog or published paper, there is a link and, understandabley a concluding comment. I can’t see a problem here. If the other analysist wishes to respond they may. If Lucia was to ask questions off-line and then post we would only see hearsay and the comments would be full critism of the hearsay not the analysis and that would be boring and uninformative. Lucia is also careful about attributing motive, the mysteries being investigated here are not why somebody did a particular analysis but whther the analysis itself is useful, so the question you keep asking is rather missing the point, and repeating the question and repeating the question still doesn’t make it a useful question especially as it has all ready been answered.
Gavin says if we put rose coloured glasses on then the sky doesn’t look blue. Lucia says that’s like saying if we put on yellow glasses then the sky isn’t blue. Nathan says Lucia why don’t you ask Gavin why he put on rose coloured glasses. Lucia says Gavin won’t answer in public and besides the colour of the glasses doesn’t change the sky just the way it looks. Nathan says ask Gavin. Lucia says Gavin won’t answer in public lets analyse the sky some more. Nathan says ask Gavin. (rinse repeat) yawn yawn
Lucia — sorry I got so involved with Nathan you can remove the lot without diminishing the world one little bit — I’m just trying to avoid writing up a ratehr pointless report
Andrew Kennett-
No problem. I’ve answered Nathan’s rephrased question about why I don’t post long discussions at RC. The first time I answered (here): I said this:
That’s my reason.
I also mentioned this:
But which Nathan seems to have taken to mean I posted the question he wanted me to ask. All I meant was that I posted a comment in response to something someone said. So, I might sometimes post brief comments. However, as it happens, I have no intention of getting involved in long conversations in comments at RC.
If Nathan doesn’t like it, and he thinks my discussing this with Gavin is a priority, as far as I’m concerned, he can go explain the issue to Gavin and try to convince at RC to have comments policies with objectionable features. (I suspect Gavin would tell him to pound sand. RC likes their policies. I even understand why they have them as I’ve seen what can happen in comments there. Moreover, if Gavin felt discussing the typeII error was a major priority, he could do it here. He isn’t– which is fine. )
If Nathan wants to learn what Gavin thinks about my dicussion of type II error, he can also ask Gavin.
For some reason, Nathan wants me to discuss this with Gavin. But the reasons he gives don’t correspond well with his dis-inclination to explore the options available for him to discover the answer he claims to seek.
FWIW: I have many theories why Nathan wants me to comment at RC. The theory at the top of my list is this: Nathan wants me to waste my time burying my opinions deep down in a heavily moderated comments field where no-one will read them, and where comments by those who agree with me are likely to be blocked. He wants me to waste my days returning to discover whether or not my comment appeared etc. In short: He thinks the time-wasting factor involved in my commenting at RC would suck up time and slow down my posting here.
If he only wanted to learn what Gavin thinks, he would ask Gavin.
Of course, I could be mistaken. There are other alternative.
After this, I will delete all future questions where Nathan asks why I don’t try to engage Gavin at his own blog. If he persists in asking, I will put him on auto-moderation.
Nathan:
Yes. I’ve had email conversations with Gavin. When there are substantive disagreements, they can be unpleasant and unproductive. Neither he nor I enjoy the experiences. So, I’m not going to get into email conversations simply to please Nathan. Posting at my blog suits me better.
If others enjoy emailing Gavin, they should feel free to do so.
lucia
” I have many theories why Nathan wants me to comment at RC.”
I really don’t care if you comment there at all.
I rarely read RC.
I asked you to disucss it with him, becuase I didn’t understand how you could come to different conclusions based on the same data.
“Yes. I’ve had email conversations with Gavin. When there are substantive disagreements, they can be unpleasant and unproductive. Neither he nor I enjoy the experiences.”
Well that’s a shame…
“After this, I will delete all future questions where Nathan asks why I don’t try to engage Gavin at his own blog. If he persists in asking, I will put him on auto-moderation.”
Goodness me. I aksed you exactly once if you were going to discuss this at RC, then I aksed you once if you were going to discuss it with Gavin.
Nathan,
You continually repeated essentially the same request. I am not required to publicly explain why I don’t post at RC, or publicly discus why I prefer not to exchange personal email with Gavin, or debate whether or not I should comply with your desires for me to behave as you prefer.
I answered your questions about why Gavin and I came to different conclusions: He diagnosis “fail to reject” as somehow contradictory with “reject”. His is a common mistake made by many people. However, it’s still a mistake.
If you want Gavin’s point of view, you can ask him and could have done so more than 21 days ago when I posted my article.
Nathan — I’ve just done a count on this thread. 1 call to post at RC that is true but 5 calls to ask Gavin directly. And this isn’t the first time. Now I’m not here to defend Lucia (and I often dislike her conclusions but can’t pierce her arguments) but please don’t play the victim — you’re not.
Lucia
Good grief…
Ok I’ll leave you alone. I don’t seem to be able to convince you to discuss this with others. Take care.