Dessler e-mail to Lindzen on 1/17/11

Dick-

First, I’d like to comment on your statement that, “The issue of
whether clouds can cause El Nino is a red herring.” I agree, and I
assume from this that you also have no physical evidence that clouds
are causing surface temperature changes. If you did, you would be
advancing it as a fatal flaw in my paper.

What’s puzzling is that I just re-read the comment you submitted to
Science about my paper. In it, your first and main point is: “These
results imply that [Dessler’s results are] ... not indicative of the
cloud feedback, but is largely a consequence of the temperature
changes induced by non-feedback cloud variations.” That sure sounds
like “clouds cause ENSO” to me.

So have you concluded that your comment to Science is now a “red
herring”? Are you going to withdraw it?

OK, now on to the meat of your e-mail. You make the argument that
there are certain time scales over which the feedback must be
calculated.

To be honest, I simply don’t follow the logic of your argument.
Luckily, it sounds like this is really point number 2 from your
Science comment, which is reasonably clear: “A second problem arises
from the use of regression over the whole record. The problem here
stems from the fact that feedbacks introduce temporary imbalances to
the radiative budget (over time scales of hours to months), but over
longer periods (years to decades depending on climate sensitivity),
the system equilibrates so as to eliminate these imbalances (6). Using
the whole record acts to distort the feedback estimates by including
equilibration as well as feedback. More accurate estimation of
feedback requires the isolation of the specific feedback signals (5,
6).”

I have two responses. First, as I wrote in my response to the Science
comment, “Their second criticism indicates confusion between forcing
and feedbacks. It is forcing (e.g., an increase in greenhouse gases,
a brightening Sun) that generates imbalances in the Earth’s radiative
budget. This radiative imbalance then causes the planet to warm,
restoring equilibrium. Feedbacks do not create a radiative imbalance
— they simply change the magnitude of the warming necessary to restore
radiative balance. When estimating the magnitude of a feedback, there
is no requirement that the Earth’s surface temperature be either in or
not in equilibrium.” Clearly, as articulated in your Science comment,
this argument is fundamentally wrong.

Second, I wonder what the source is of the claim that we must consider
equilibrium time scales in our analyses of feedbacks. It appears to
be reference 6, which is Lindzen and Choi 2009. But the claim is not
proven in that paper --- it is simply assumed (in paragraph 5 of that
paper). Can you provide a reference where some evidence is presented
in support of this claim?

Thanks!

Dessler e-mail to Spencer on 1/17/11

Roy-

Correlation does not prove causality. Period. I honestly cannot
believe you want to argue this point. The fact that you do lays bare
the intellectual bankruptcy of your “clouds cause climate change”
hypothesis. It’s now evident that there really is no actual physical
evidence to support it.

In science, correlations allow you to construct hypotheses, which then
must be tested with physical arguments. The history of science is
littered with the corpses of high r squared correlations that fooled
people into assuming causality. I’m afraid you’re another victim.

I would NEVER assume correlation proves causality. My cloud feedback
calculation is supported by a firm causal link: ENSO causes surface
temperature variations which causes cloud changes. This is supported
by the iron triangle of observations, theory, and climate models.

As to your comment:
“After all, the Southern Oscillation Index is an ATMOSPHERIC index,
and for you to claim that changes in the trade winds DON'T cause a
change in cloud cover, which can in turn affect ocean temperatures, is
treading on thin ice.”

Of course changes in trade winds can change clouds.

But what causes changes in the trade winds during ENSO?

It’s the changing SST distribution. Get it? Surface temps drive clouds. QED.

Thanks for this interesting discussion. Believe it or not, I think
we’ve actually reached closure. I have no more questions for you.

To the reporters/bloggers on this list: I would encourage you to
write/blog about this exchange. I think that this was an unusual and
frank exchange of views that people would be interested in. FYI, I’m
also archiving these e-mails on my previously dead blog,
http://sciencepoliticsclimatechange.blogspot.com/.

Thanks!

Lindzen e-mail to Dessler on 1/14/11

Dear Andy,

It is clear where you are going wrong. You are confusing changes over long periods during which equilibration can occur with fluctuations from which feedbacks can be determined. Note that in order to measure feedbacks from outgoing radiation, one must look at time scales that are short compared to equilibration times (years) but long compared to the action of feedback processes (days). That fluctuations in clouds, volcanos, etc. occur over the relevant time scales is obvious as is the fact that such fluctuations must, of necessity, cause changes in surface temperature. The issue of whether clouds can cause El Nino is a red herring. Incidentally, we have looked at your data. If you restrict yourself to relevant time segments, your r-square goes up greatly, and if you perform an analysis of leads and lags, you too get a negative feedback from fluctuations in outgoing radiation that lag SST (whether you choose segments or your entire record -- though again, r square is much larger for segments). The ambiguities in the choice of segments in Lindzen-Choi 2009 disappear when one uses 2 or 3 month smoothing.

Best wishes,

Dick

Spencer e-mail to Dessler on 1/14/11

Andy,

I cannot believe you keep saying "correlations do not prove causality", when you yourself have used a (near-zero!) correlation in your paper to support causality in only one direction!

But we (including Lindzen) are claiming that relatively LARGE correlations, with a clear lead-lag relationship, DO strongly suggest causation.

I do not know how you can, in all honesty, continue this mantra of yours...oversimplifying our position into something like "clouds cause El Nino", or some such thing.

After all, the Southern Oscillation Index is an ATMOSPHERIC index, and for you to claim that changes in the trade winds DON'T cause a change in cloud cover, which can in turn affect ocean temperatures, is treading on thin ice.

And if, perchance, Kevin Trenberth is coaching you on these talking points, I wish he would just come out and defend them himself.


-Roy

Dessler e-mail to Lindzen on 1/14/11

Dick-

Your question gets to the same issue that I am after: what is the
cause of the temperature variations over the last 10 years. If it
turns out that clouds are responsible, then that would indeed
challenge my feedback estimate.

However, I suggest that ENSO is responsible for the temperature
variations. I've repeatedly asked Roy for evidence that clouds are
responsible, and he cannot provide anything beyond some ambiguous
correlations. The problem with that of course is that correlations do
not tell us causality.

If you have evidence that the variations over the last decade are
caused by clouds, then please let me know what that evidence is.

Thanks!

Lindzen e-mail to Dessler on 1/11/11

Dear Andy,

I find this exchange a little peculiar in that you never seem to address what I understand Roy's main points to be.

First, Roy notes that outgoing radiation (especially in the visible) varies for many reasons of which feedback to surface temperature is but one. Other obvious examples range from volcanic aerosol production to cloud triggering by Kelvin-Helmholtz instabilities. I know of no one who really questions that surface temperature is not the only or even the main source of fluctuations in visible radiation.

Second, any changes in outgoing radiation will also cause changes in surface temperature. This is simply a matter of elementary thermodynamics. For non-feedback changes in radiation, the temperature changes will follow rather than lead temperature changes.

Third, because of the first two points, simple regressions of outgoing radiation on surface temperature will not be a useful measure of feedback. Spencer and Braswell illustrated this in their paper. There are, of course, other problems with the use of simple regressions as well.

It would be interesting to see your response to these points.

Best wishes,

Dick

Dessler e-mail to Spencer on 1/10/11

Hello, Roy.

If you don't mind, I'm going to keep after you on one important point.

In these e-mails, we have discussed two hypotheses for the observed
lead/lag relation and decorrelation. My hypothesis is that all of the
observations are due to canonical ENSO physics --- not clouds causing
temperature changes. The fact that climate models get these relations
supports this, and it is consistent with the surface energy budget.

Your hypothesis --- clouds cause temperature changes --- has no
supporting data (beyond just the correlation & decorrelation) and it
is not consistent with the surface energy budget.

Here's my question: Why would anyone (including you) accept your
hypothesis when the mainstream view fits the data better.

I'm interested in your thoughts.

Thanks!

Spencer e-mail to Dessler on 1/7/11

Happy New Year Andy,

I've been catching up this week after having last week off for the holidays.

The evidence I presented you was NOT just the decorrelation of the data, as you claim...as I mentioned in my previous e-mail , it was TWO-fold: It is also the lagged relationship, with radiative flux changes preceding the temperature response, then the temperature changes either simultaneous with (os as we will see with ENSO, preceding) the radiative feedback response. The phase space plots we published are one way of revealing the lagged relationship.

Without taking this time lag into account, you will get correlations -- and thus regression slopes-- close to zero...as you do, even in the climate model regression.....no matter what the feedbacks are. We demonstrated this in Spencer & Braswell (2010).

Now, let's look at the oceans, since you want to emphasize the signature of ENSO during the period of record....

By far the most precise measurements of global SST variations come from AMSR-E on NASA's Aqua satellite and, as you know, a CERES radiation budget instrument also flies on that satellite. I did a calculation with these data somewhat similar to the one you did.

Attached find a plot somewhat analogous to yours, but from Aqua SST versus radiative flux....As you can see, the radiative feedback response occurring AFTER the temperature changes suggest strongly negative feedback. Also shown in the same plot AR4 climate model results from their 20th Century runs on the same plot.

The peak correlation of the satellite data in the plot was 0.68, at 11 months lag. At zero lag, the correlation is only 0.27. (What was the correlation of the data you showed in Fig. 2 of your Science paper? I did not see one listed.)

Clearly, there are cause-effect things going on here that cannot be revealed in plots like your Fig. 2, unless these time lags are taken into account.

As a result of our discussion, I've decided that we should do another publication, focusing on the lag relationships seen in the Aqua data and what they might mean for feedback and climate model validation.

-Roy

Dessler e-mail to Spencer 1/3/11

Roy-

Happy new year!

Thanks for your response. I do think we are making progress now.

In your last message, you confirm that the only evidence supporting your hypothesis is the observed correlation between surface temperature and cloud radiative forcing as well as what you refer to as "decorrelation" of the data.

Because of the complexity of the Earth, there are always multiple hypotheses for every correlation. For the observed correlations you note, I have a competing hypothesis which I think explains the data better than yours.

Here is mine: the heating the surface is caused by energy stored in the ocean. This drives changes in atmospheric circulation, which change clouds. Clouds play a very small role in the surface energy budget during ENSO.

There are several strong pieces of evidence that support my point of view:
(1) Climate models successfully simulate the same lead/lag relationship -- see my attached figure. This suggests that correlation you identified is just normal ENSO physics.
(2) Energy budgets of the surface also suggest virtually no role for clouds during ENSO (e.g., see Trenberth publications).
(3) As far as the decorrelation of the data goes, I have not looked at this in the climate models. However, I think you sent a figure a few e-mails ago showing a climate model that reproduced that. Again, this suggests that it's normal ENSO physics. In addition, I note that the models generate about the same r^2 as the data.

Finally, you have no evidence supporting your hypothesis beyond the mere existence of the correlation. Because of that, your theory explains nothing (e.g., you cannot tell me what percentage of all of the variability is due to "forcing" versus "feedback") and makes no testable hypotheses.

So here are my questions for you:
(a) Do you have any evidence that my proposed hypothesis is wrong?
(b) Does your hypothesis make any testable predictions that would
allow it to be falsified? If so, what are they?
(c) Does your hypothesis explain anything that my theory does not?

Thanks!

[Figure caption: The slope of the regression of energy trapped by
clouds vs. surface temperature, as a function of the lag between the
time series. Negative values indicate that changes in clouds lead
changes in surface temperature. The black line is calculated from
observations cited in Dessler [2010] and the red lines are the climate
models from that same paper.]

The blog is back ... at least temporarily

Roy Spencer and I have been having an interesting and useful exchange of ideas on the cloud feedback. The goal is not just for us to hear contrasting views, but for the interested public to listen in on the conversation. My goal is to publish our correspondence as it occurs, rather than have a big dump of e-mails periodically. Roy published the first bunch on his blog, and you can read it here. Subsequent e-mails will be posted here as they arrive.

Thanks and welcome to the conversation.