"Our greatest responsibility is to be good ancestors."

-Jonas Salk

Saturday, November 13, 2010

The Empiricist Fallacy

The mid-20th century scientist H H Lamb seems above criticism. Defenders of the CRU avoid criticizing their founder, while their opponents refuse to let go of the "Medieval Warm Period" sketch that he made, that made it into the first IPCC report (without attribution) on the basis of authority. There's little doubt that Lamb was one of the founders of climatology, but to put that in perspective, my wife assures me that serious psychologists don't take Freud seriously anymore either. I actually read a fair chunk of Lamb's magnum opus, and here's some commentary I made on it in 1995 in a "scientists used to predict global cooling" thread on usenet.
It is important to note exactly who made those predictions, (or more
properly, who expressed those worries) about an imminent ice age, and
who is now predicting rapid global warming. By and large these are not
the same people. The first group was essentially the observational
paleoclimatologists. Bryson still claims that "the proper tool of the
climatologist is the shovel". The compendium by Lamb which Tom Moore
takes as his primary reference was essentially the pinnacle of achievement
in that field.

With all due respect (I mean this quite seriously - the erudition and
breadth of knowledge of these people, Lamb in particular - is enormously
impressive) to that group, their grasp of mathematics and statistics
was weak, and of physics weaker still.

For instance, Lamb's prediction in particular of imminent and rapid
cooling was based on, essentially, a crude Fourier analysis (best fits
of sinusoidal curves to his record). Since one of the dominant features
was a rapid rise over the last century, the *presumption* of a cyclical
nature of the record forced a prediction of a rapid cooling *precisely
because there had been a recent rapid warming*. And although the niceties
of periodograms had all been worked out by that time, Lamb seemed blissfully
ignorant of the need to take particular care when fitting sinusoids to
a record with significant information at its termination.

In the 1970s, a separate discipline of physical climatology was just
emerging from an infancy at the peripheries of mathematics and astrophysics.

Since the 1890s, physical climatologists or their precursors have always
asserted that the anthropogenic cooling of the human volcano was
counterbalanced and probably outweighed by anthropogenic warming of
the human greenhouse.

The groups making the assertions were essentially distinct, the group
asserting warming was making far more specific and testable predictions,
and the reasoning behind the assertions was far more clearly based in
established and demonstrated results in physical science.
While Bryson's warnings about global cooling were intuitive ("the human volcano"), Lamb's were dressed up in harmonic theory. Unfortunately, harmonic theory clearly doesn't apply even were the system largely unforced. In a forced system where you look only at the system behavior and not at the inputs, and apply harmonic analysis blindly (I doubt Lamb was aware of Tukey) you'll basically end up predicting that a rapid cooling is imminent for two reasons: first, the assumption that there is no trend, and second, the emphasis of the FFT on the edges of the record (hence "windowing").

Note: in the linked usenet article I express doubts that Schneider ever expected cooling, but that is incorrect; he was coauthor of a paper with Rasool in the 70s that compared the forcings and expected aerosol cooling to dominate. He didn't hold that position long. I have not seen a proper statement of what's wrong with that paper.

The basic idea of empiricism, which I see implicit in everything Curry is saying, by the way, is that the data "speak for themselves" and no context is necessary. This is silly, since it amounts to a presumption that sightings of cousin Albrecht wearing a baseball cap and carrying a feather duster wading in a pond are as likely as a duck (regular readers need not follow the link; they know where it goes).

The empiricist view has never entirely faded from climatology, as, I think, we see from Curry. But it's essentially useless in examining climate change. Under its precepts, the only thing that is predictable is stasis. Once things start changing, empirical science closes the books and goes home. At that point you need to bring some physics into your reasoning.

More to follow.


William M. Connolley said...

Hey, I remember reading that post (belated thanks for it). Ah the good old days, when sci.env was worth reading (and writing).

Anyway, you say "I have not seen a proper statement of what's wrong with that paper." You should read http://www.wmconnolley.org.uk/sci/iceage/, viz: "the Schneider paper uses incorrect (a fact of 3 too low) values for CO2 sensitivity". In fact my recollection is that footnote ?3? in the paper pretty well says this or something similar. The other "error" is that he assumed a vast increase in aerosol that didn't happen. There was some follow-up talk about that (on sci.env I think; but I ref it on my page) about subsequent crit in the letters page about his assumptions.

Anonymous said...

I'm currently reading Schneider's "Science as a contact sport", which talks about the global cooling thing a bit.

His main points were that their model had not included the stratosphere, which significantly reduced the sensitivity to CO2, and had assumed that the aerosols were uniformly distributed, which increased the sensitivity to aerosols.

Rattus Norvegicus said...

I actually disagree with your title. Empiricism is important in the early days of a science because it is necessary to describe the phenomenon being investigated. However, w/o an understanding of the forces driving the phenomenon being investigate it is not sufficient to make progress. After all where would we be if not for the early work of Agassiz on ice ages? After all, his work established one of the early questions which climatology attempted to answer.

Even once a theoretical basis is established, empiricism remains important in developing new historical data which poses questions for the theoretical portion of the science to answer.

Michael Tobis said...

Empiricism is, of course, necessary to science. The fallacy that I wish to address is the idea is that it is sufficient.

Essentially much of the legitimate-sounding naysayer discourse amounts to a demand to "prove" the case for a fairly high sensitivity without any evidence other than the empirical.

There are several weapons designed to prevent reasoning from evidence. The idea that "the whole business depends on the hockey stick" os one. It presumes that the system is so mysterious that one can know nothing about it.

Unknown said...

On Lamb and Tukey:
It is sure that both Lamb and Tukey read Julius von Hann, probably his Handbook of Climatology (English edition 1903).

Tukey called one of spectral windows "Hanning", partly because von Hann used something like that, and partly because Hamming (Tukey's colleague) made a spectral window which is related to that.

It seems that von Hann did not explicitly weight data in the frequency domain, but just smoothed time series with "1,2,1" weights. (I may have missed something.)

On Schneider:
Schneider (1975) paper has comments on Rasool and Schneider (1971) paper.

Anonymous said...

Hansen commented a bit on Schneider in his interview with Spencer Weart for the AIP.

From the second link:
[...] [Rasool and Schneider] wrote a paper on the effect of aerosols on climate, and ended up saying that the Earth may be headed towards the next Ice Age. I actually helped Steve with his calculations because he was not really a radiation transfer person, so I gave him equations for the scattering the solar part of the problem. He did the thermal part of the problem, and actually, he did a kind of slightly sticky CO2 calculation. He understated the greenhouse effect by a factor of two or so. We later figured out looking at his computer program that it was his mistake.
I’m curious. What was it?
Just his calculation of the infrared opacity of the greenhouse gas greenhouse gas, CO2, was wrong.