Still, I’ve never submitted a paper knowing I was not going to have funds to cover page charges.

Obviously, before commenting on a comment, I would wait for Steven Schwartz to respond. After all, he may well simply say what I would have said, in which case, my comments would be redundant.

Meanwhile, I am trying to see what the effect of uncertainty is when I use “my” method. I’m getting values near Scafetta’s right now. But the uncertainty in estimating the autocorrelation is necessary.

There really is no need to write a comment on the comment. As you say earlier, Steve Schwartz will be given the chance to reply to the comment and the two will be published together. Steve Schwartz is a smart guy and I am sure he will have no difficulty responding.

More generally, on the subject of page charges, most journals will waive them if you plead that you have no source of funding, provided that your paper is not too long and you do not require color printing. But whether a journal will publish your paper or not is a different question. If you do write a paper, don’t forget to end it with the Credo: “I believe in global warming almighty, …”

1) Scaffetta assume the data can be described as the sum of two series, each with it’s own time scale two time scales, and finds
A) very fast one with a time scale of about 1/2 a year (note: very close to zero, as data are annual averaged, so you can’t get less than 1 year!) and

B) a slow one with a time scale of 8 or so years ±2 years!

2) I assume two processes with two time scales also! I know it looked like one, but What are my time scales?
A) I assume one has a time scale of ZERO right away. This is because I identify it as “white noise” due to measurement uncertainty. (Given how some of the agencies calculation this, there could acutally be a time scale associated with measurement uncertainty. However, I just assumed zero.
B) About seven or eight years ±2 years.

So, to make our two results “the same” all we need is for Scafetta to say: “Oh. I bet that really speedy process is measurement noise, which retains some correlation due to the way GISS smooths spatially and temporally to do whatever it is they do to determine “correct” the temperature at stations.)

What else did Scaffetta do:

He subtracted out the mean computed by GISS and then redid the computation. This makes sense in context of these simple models. The reason is: The “forcing” files in the climate models don’t include white “TSI” noise or “earth albedo noise” or any other “noise” factor that might cause the forcing term describing in the simplified equations to be random. So, to the extent that the system responds to the random componenet of the extrenal forcing, that means the climate models can’t get the precisely right autocorrelation for residuals, especially not anything at the smaller scales.

(This isn’t even a slam on models. The modelers tell us they are only interested in the average signal. That’s why they don’t use realistic values for the random component of the external forcing. The idea that you can use the smoothed component of the forcing may be fine when predicting the future climate, but many aspects of the random “weather noise” would be off in some amount.))

And both the Schwartz and the Scaffetta method need the residuals to be there.

I haven’t done that. What I did instead was create “Lumpy”, where I fit the simple lumped parameter model to the forcings given by GISS. I don’t remember the time scale for that– but I get fairly low sensitivities, no matter what I try to do.

Hmmm… I should be doing these more formally and submitting.
Also, maybe I should show Scaffetta the “Lumpy” way!

Lucia, since you have had a few words to say about other’s comments on Schwartz’s paper, I was wondering if you could comment on this comment by Nicola Scafetta:
http://www.fel.duke.edu/~scafe.....357713.pdf

Tamino was outed as Grant Foster on CA a while back

As Lucia points out, a GCM is not earth, even if it was correctly implemented. However I have never seen a verification effort of a GCM; hence we cannot even assume that the GCM does what the modeller intended it to. Using an unverified and unvalidated hypothesis to disprove another hypothesis is disingenuous. That’s not science and the modelling community should better learn the difference between reality and hypothesis (model). It is not the first time in climate science the two are confused.

That’s why I also haven’t bother to write up my Taylor Series analysis, which Roger Pielke Sr. suggested might be nice for a particular issue of a journal that was dicussing the sort of non-GHG issues for modeling.

When responding to Steve, I didn’t mean to imply the whole paper is wrong– but that part is.

However, in response to you, later, I’m adding that I’m pretty sure problems with all four prongs of the argument in the paper. The question is are they big problems or small ones. I know two of the four counter arguments in GASM are dead in the water.

I know the GCM ‘proof’ are meaningless. That’s the “figure 1″ bit in GASM. The reason is that if your look at the derivation of “Schwartz” method to work, it requires that the extrenal forcing contain a sizeable amount of white noise. The earth may experience sizeable “white noise” forcing, but the GCM runs most certainly do not. So, based the derivation of the Schwartz method, it might work for earth, but it can’t work for GCM runs — unless the modelers include a sizable amount of white noise in the forcing files.

So proving the Schwartz method doesn’t work for GCM data is utterly meaningless. I need to figure out how to write this up clearly with examples. But the GCM “proof” is pointless (and not because GCM’s may be right or wrong. It’s pointless because they make a certain type of approximation, which may well work for climate but will result in some differences in the autocorrelation for GMST– at least if the Schwartz method has any validity.)

So, that means two of the criticisms are flat out wrong. That leaves two possible criticisms in GASM alive.

* The issue about the bias in the correlation (equation 4 in GASM) is definitely valid with regard to Schwartz. Though… I’m actually not sure how much difference it ultimately makes on the answer we wish to learn: the time constant. (I’m going to be adding that correction to “the lucia method”.)

* The issue about the uncertainty intervals is important and that is definitely a problem for Schwartz 07. HI need to look at that for “the lucia method” also. (And I need to look at bias at the same time.)

Very good questions. I knew I was being obscure, but of my readers know what I was talking about because I’ve discussed it before in the context of what Schwartz07 did wrong, but how the criticism at Real Climate was off the mark. (That is: Tamino’s criticism at RC was more wrong than Schwartz.)

Grant, Annan, Shmidt & Mann has a 4 pronged complaint of what is wrong with Schwartz paper. The measurement uncertainty affects two of the prongs a lot, a third prong at least a little, and the fourth one not so much. (The fourth prong is the GCMs.)

The first prong in the “criticism” is that the shape of figure 7 in Schwartz is inconsistent with the idea the climate could be described approximately using a simple “lumped” climate with a single time constant. This conclusion is wrong. The shape is perfectly consistent with the claim climate has one constant!

The reason the Schwartz figure is perfectly consistent has to do with measurement noise! In laboratory experiments, the shape of the autocorrelation is well know to be affected by noise. This is particulary true of the behavior near lags of zero. (I can pull quotes from Hinze 1957. I need to go to my husbands office and find Bendat and Piersal. )

In their papers, neither Grant, Annan, Shmidt & Mann 08 nor Schwartz07 recognize that the measurements are the sum of two processes: a) The climate system and b) the uncertainties in measuring GMST.

Those uncertainties are admitted by both Hadley and GISS. Also, we know they exist, because if you compare Hadley and GISS, they aren’t identical. (More about this later. )

The discussion of the “lumped” climate would suggest the GMST of the planet itself should look a certain way (that would be a horizontal line) in Schwartz. However, if you figure that the measurement uncertainty exists, and is white, you discover that the data should looke precisely l ike as they do in Schwartz’s figure:

I discuss this here.

http://rankexploits.com/musing.....-suggests/

If you read that article, you’ll see that ig turns out that if you do consider the measurement noise, and solve for the amount of the time constsnat and the noise, you get back an amount of noise that is near the amount admitted by the agencies!

I”m going to be discusing the AFSM08 more later, but there are other problems with that paper.

Since you brought up the GCM’s, I’ll make this comment. Using the method Schwartz recommened and to get time constants for GCM’s does not represent a valid test of the method’s utility when used to obtain the time constant of the earth . However, this one prong of their criticism is unaffected by “white noise”.

GASM08 seem to assume that Schwartz’ paper amounts to a “claim” that any “lumped parameter” that conserves energy should act as an AR(1) process, simply because Schwartz does that to the earth. However, Schwartz07 doesn’t make that claim. His claim is his method can be used if the system conserves energy, is a subject to an external forcing that consists of deterministic linear trend trend plus white noise. (Arguments can be made for externally applied white noise forcing on the real earth.)

GCM’s are manifestly not driven by white noise plus a trend. The forcings applied are highly smoothed removing all high frequency fluctuations (such is variability in the TSI, any temporal changes in albedo due to famers plowing fields different things growing, period smaller volcanos, what not). This is justified as unimportant in the long term (and that may well be correct.)

However, it has the consequences that it would affect the autocorrelation of the residuals of the “weather noise”. Specifically, it affects the higher frequency components. And…. it’s those components that matter the most when assessing the first few lag components of the autocorrelation!

For this reason, it is easy to show that even if Schwartz’s method worked for the earth, it could never work for GCMs!

So, this is a problem with the GASM paper– but it’s unrealted to the measurement noise issue.

Of course, there are problems with Schwartz paper as he too should have considered the measurement noise. I’m trying to see what happens if we consider the remaining criticisms in
GASM08 paper.

The things I need to do are:

Add the uncertainty term Annan describes and concluded that if incorporated into my analysis. (Actually, I did this: it makes the time constant tiny bit lower 7 years ±2 year for Met land only data. (I always to the inappropriate data first. Then I surprise myself with the “real” answer. )

But, I need to check whether doing the Annan adjustment makes sense. I need to read the reference Annan cited and figure out if “correcting” with that makes sense. (Maybe I need to adjust for different lags etc? I don’t know.)

I need to check the montecarlo stuff to find uncertainty intervals. GASM seem very keen on checking if 30 year lags are eliminated. So, I really should re-run the Monte Carlo method with the white noise added, because for all I know, the error bars are that large.

Anthony has stated several times that he really doesn’t approve of Tamino attacking people from behind a wall anonymity.

Who cares what Anthony thinks? It’s supposed to be about the argument. But when Watts’ arguments get pummeled he’s got to fall back on something, so we have the anonymity bashing. A rush to “out” somebody’s real life identity becomes a surrogate for scoring an intellectual point.

And I agree with Arthur that Schwartz’s method’s inability to come up with a known lag time is the coup de grace for his paper.