…are flawed. That’s putting it mildly:
Professor Curry said: “It’s not just the fact that climate simulations are tuned that is problematic. It may well be that it is impossible to make long-term predictions about the climate – it’s a chaotic system after all. If that’s the case, then we are probably trying to redesign the global economy for nothing”.
I’ve been saying that’s likely the case for years. I’ll look forward to reading her paper.
I am very experienced with computer modeling or in my profession computer simulation. That is how I earn my living and have for “a very long time”.
Computer models are fine, when their problem domain is well specified, including inputs and outputs and behavior is deterministic even under pseudo-random stimulus[1].
You get into the weeds when some or none of the above is true. Also there is a fundamental difference in how models are used when attempting science vs. engineering. In engineering the outcomes are fixed and knowable/predictable in advance. Science is pulling back the curtain on the previously unknown. The extreme success in computer modeling in the field of engineering has over time been eyed with envy in the scientific community. It is rather amusing to know that this was not always the case. In fact there was a time when over-reliance on a “tool” to do difficult science was considered something “less than desirable” [2]. Somewhere in the 1980’s that changed. I’m convinced some of it had to do with arms control. SALT and START that went into effect in the 70’s and 80’s had the real consequence of prohibiting actual detonation of nuclear devices even below ground. To replace that testing those responsible for maintenance of the nuclear stockpile in the US resorted to computer modeling of same. I can’t imagine the hours that had to go into verifying those computer models. The work is classified but I imagine required a lot of affirmative experimentation to go along with it. But again even in the extremes of a nuclear detonation you are working in areas of known constraints, with predictable outcomes. Modelling has since branched out into other sciences to varying degrees of success. Weather models have become quite good within their 72 hour horizon. What isn’t as well known by the public is that those models also require precise inputs in terms of atmospheric pressures, temperatures (and etc. I’m not a meteorologist) within a 3d grid in order to have any predictive power. Take away those precise data inputs and the models become useless. And those data measurement techniques have improved greatly over the years. If these same computer weather models were somehow to be made available to weather folk of 60 years ago, its unclear they would have provided useful forecasts, lacking the corresponding advancements in atmospheric measurement. As the saying goes Garbage In Garbage Out.
The domain of climate models have few of the attributes I have outlined above (para. 2). And I agree that the fact that they require tuning, yield trends that diverge from observation and have yet to demonstrate that they can be run backwards in time to show conformance to the historical record should all be issues of concern.
[1] True “random” stimulus using a digital computer is impossible without specialized hardware and in fact is rather undesirable. Pseudo-random stimulus is based around modulo arithmetic (in most cases) and can mimic nearly random stimulus in most situations to a high enough degree to accomplish what is needed, with the added benefit that it can be made repetitive to allow further problem examination of a given stimulus sequence.
[2] Who Got Einstein’s Office? Eccentricity and Genius at the Institute for Advanced Study Regis – See the chapter on John von Neumann.
Oh come on. You write some code that produces stable predictions of Earth’s temperature, and then you add a factor for mankind’s increasing evil, gluttony, and greed.
for (year=1975; year < 2100; year = year + 1)
temp(year)=16+random(1.0) ;Earth's average temperature in Celsius
'yep. Model looks right
for (year=1975; year<2100; year=year + 1)
begin
evil = 0.02 * (year – 1975)
gluttony = 0.02 * (year – 1975)
greed = 0.01 * (year – 1975)
temp(year) = 16.0 + random(1) + evil + gluttony + greed
end
'Yep. Model correctly predicts a 6.25 C increase in temperatures
That took me all of five minutes.
I could do the same thing in a far more roundabout way using a shallow water model or full Navier-Stokes equations, but it would just be a waste of processing time.
Hi George you should forward your CV to NOAA 🙂
Your “stable” model has a warming bias. I’ll leave that as an exercise for the student.
Some would also probably dispute your monotonically decreasing values for evil, gluttony and greed prior to 1975. If I were counting down to zero for an arbitrary definition of EG&G, 1971 might be a better date, that was when Nixon first imposed wage and price controls and the US economy was supposedly in for a few years of a “utopian paradise” or at least no inflation until after the ’72 election…. And soon thereafter you couldn’t buy gasoline either…. Life was so great back then….
I’ve long wanted to run a number of simulations and analyses with interval arithmetic, to see how robust their results actually are.
If that’s the case, then we are probably trying to redesign the global economy for nothing
Oh, it’s not for nothing. It suits the global socialist elites just fine.
Never let a crisis go to waste. …and if there isn’t one, just make it up!
Well, first you model the butterfly, then you model what put the butterfly in that exact spot and made it flap its wings exactly when it did — which is almost certainly because of another butterfly at a different time and place.
Easy-peasy.
The whole idea of modeling the earth’s climate with differential equations is ridiculous. The Navier-Stokes equations, like all results of continuum mechanics, are derived by setting up the constitutive equations in integral form (based on the unshakable assumption of the fluid as a continuum, rather than a set of discrete particles), and differentiating them to make them more tractable. As an astute professor of mine once observed, “Nature doesn’t differentiate, nature integrates.” That’s why the flap of a butterfly wing makes absolutely no difference in the real world, but does in the world of differential equations that are being solved by discrete approximation. The NS equations have no hope, in their native form, of predicting the dominant feature in the earth’s atmosphere, turbulence. In a discrete approximation numerical solution, they appear to do so, but don’t really. Climate models have nothing to do with weather, climate, or reality.
Navier-Stokes also falls apart if there’s condensation or evaporation, and the work-arounds required to use NS for a steam turbine are quite elaborate.
And yet they’re using NS simulations to study a system that is to a large degree driven by condensation and evaporation.
Go figure
I don’t think it is the intractability of the governing equations so much as it is not having the right ones. NS is only applicable to some aspects of the entire problem of radiative-convective equilibrium encompassing land, ocean, and atmospheric systems. And, NS solutions are not always chaotic, or at least, there can be dominant, deterministic modes of response.
There are dominant natural modes apparent in the data which do not appear to be chaotic. In fact, the past century+ of data show a long term mean global temperature trend (which itself is probably just a segment of a very long term cycle) plus an approximately 60 year periodic component riding on top of it.
That pattern was laid in well before atmospheric CO2 rose precipitously in the mid-20th century, and causality demands that it had little to nothing to do with it. AGW rode the upswing of that ~60 year cycle in the late 20th century up until it reversed in the mid-2000’s, producing the “pause” or “hiatus” that killed the hypothesis, though it hasn’t stopped thrashing in its death throes yet.
But, natural cycles are given short shrift, because the goal is to convict humans, via their combustion of fossil fuels, of a crime against nature.
The comment about climate being a chaotic system seems strange. WEATHER is chaotic; climate is the statistics of weather. Even chaotic systems have well-defined statistical behavior (this is made precise in ergodic theory.)