AGW question for Citizen Runner

Citizen Runner wrote:(1) Let me first emphasis that we were postulating recent warming due to an unknown forcing.

(2)Without attribution to some physical phenomena we can't say with confidence what that unknown forcing is going to do in the future. There have been statistically significant changes in the surface temperature instrument linear trend.

... Why then would one conclude that the current trend will continue indefinitely?.

(3) As an aside I've read a bit of "Anti-Fragile" the other night and have some issues with it. Perhaps I'll comment when I have more time.

(1) Perhaps you were. In my original very crude naive analysis I was simply looking at trends in data, with no thought to causation. Now sometimes (often perhaps) trends in data have little or no connection with real behaviour and if that is the case we can treat any derived forward prediction as rubbish. In this case my crap prediction so far does a better job than the deterministic models of IPCC so I won’t throw it out yet.

(2) No of course we can’t treat my “prediction” or any other with confidence, at least not initially. But with the benefit of new data we can test the “hypothesis” and have it gain confidence, or not, through Bayesian updating.

(3) OK, but don’t make the debate about the book, I was simply using an interesting quote that was timely and aligned with my thinking. I doubt Taleb shares my views on AGW, although I could only find very little written by him on the topic.

You may have seen in the news that Wally Broecker, the scientist who popularized the term "global warming" died the other day. I've linked his 1975 paper "Climatic Change: Are We on the Brink of a Pronounced Global Warming?" a number of times as it addresses a number of popular climate contrarian talking points.

a) "They changed from 'global warming' to 'climate change'" - See title of paper that popularized term 'global warming'.

b) "In the mid-70s they were predicting cooling" - Acknowledges recent (at that time) cooling, attributes it principally to "dust" (aerosols in current vernacular) and a cyclical ocean current contribution (much like the "shill model" but with ~80 year cycle), and projects imminent warming trend.

c) "The models are wrong" (All models are wrong, but some are useful. --- George Boxx) - Projected 1.1 C temperature rise at 403 ppm CO2 which is about 10% higher than observed GISS-loti trend over the relevant period of about 40 years. The CO2 contribution to warming in this "Broecker model" is based on earlier modeling done by Manabe etal circa 1970. I'd add that the Hansen 1988 has performed pretty well with regards to net trend over 30 years.

In your initial "naive prediction" you explained the derivation of your model, but if you provided some explanation of your attribution estimate I've missed it. It's not clear how you can test a hypothesis about attribution without a model of causality.

Also there exists published literature on global warming attribution using Bayesian techniques.

Relative to your quote, I think Taleb uses "theory" in places where he means "hypothesis" is the standard scientific usage. He makes a clear distinction between theory and empirical approaches, but a scientific theory is a broad explanation for a wide range of phenomena with predictive usefulness, supported by consilience of the aggregate empirical evidence, usually integrating and generalizing many hypotheses. That is theory isn't distinct from empiricism in this context.

In so much as I've read Taleb, the potential "black swan" event he addresses is the potential for impacts much worse than projected rather than the potential that we've got the physics fundamentally wrong as you've postulated.

Thanks for the reply CR, lots to think about there. I only have time for a quick reply so I’ll just make a philosophical observation.

Your feedback underscores a key difference between how we seem to think about the world. Reading between the lines, I suspect you may work in a field involving relatively well behaved phenomena that can be modelled and analysed deterministically based on confident physical models. I do not; in the earth sciences, a great many phenomena defy deterministic representation and must be modelled by other means, often probabilistically. And more importantly, through the use of subjective judgement when available data do not support frequence-based probability estimates, which is virtually always the case for defining frequency of rare events (defining the tails of the pdf).

In Thinking Fast and Slow (fantastic book), Kahneman diminishes the value of subjective judgement, although he admits to coming around to seeing its validity in select situations after a long collaboration with one of his opponents, Klein, whose research revolves around situations where subjective judgement is the key to success. Gladwell’s Blink touches on this also.

In Superforecasting, Tetlock distinguishes between two different types of thinkers, foxes and hedgehogs. I feel like maybe you have the hedgehog view of AGW and I that of the fox.

Citizen Runner wrote:...In so much as I've read Taleb, the potential "black swan" event he addresses is the potential for impacts much worse than projected rather than the potential that we've got the physics fundamentally wrong as you've postulated.

CR,

I only drew Taleb in with reference to the line I quoted, to the effect that we should remain robust against changes in theory (since theories about phenomena are bound to change, whereas our observations of those phenomena don't necessarily change along with the evolving theories). You're taking a bit of a dive into his writings to follow a line of reasoning I didn't intend or rely on. That said, we can segue outside the constraints of the AGW discussion to talk a bit about "black swans" and rare events if you like.

The idea of a "black swan" isn't necessarily one of unexpectedly large negative effects, rather it relates to an occurrence, positive or negative, that was unexpected because our models (mental or otherwise) didn't allow for it. Or, put another way, our assumptions about the probability distribution underestimate the likelihood of events in the outer ranges (tails) of the distribution. This can happen in various ways; for example, we can assume normality (or some other well-behaved, low variability distribution) with exponentially decreasing probability with distance from the mean/median/mode (where often the decay is a power law, much slower than exponential decay), or, alternatively, we can underestimate the true variability for well-behaved distributions (a 10% error in estimated standard deviation results in a much, much higher error in the estimated 5-sigma event probability, as an example).

It is usually easiest to estimate a measure of central tendency (mean, median, mode, geometric mean, etc), and very much harder (or even impossible) to get a good estimate of the frequency of extreme events. When the tail(s) of the probability distribution are well represented by a power law with steep enough slope / exponent, the overall behaviour of the phenomenon is governed by rare events, and therefore deriving an understanding of the future from average conditions (or average plus some measure of variability) may be entirely inappropriate and inadequate.