Hoax or blunder?
David E. Wojick
dwojick at shentel.net
Thu Dec 18 17:03:29 EST 1997
For your interest and contemplation:
This letter from two prominent climate change modelers appeared in the
Washington Post (Tuesday, December 16, 1997; Page A 26).
"On a Butterfly's Wing
"In an otherwise informative article about climate change, John Casti
states, "The flapping of the proverbial butterfly's wings in Brazil today
can percolate into a tornado in Kansas". ["A Look at the Complexities of
Global Warming," Outlook, Nov. 30] This kind of statement might contribute
to public confusion and dilute the resolve needed to formulate a reasonable
response to the threat of climate change. Given the potential uncertainty
that Mr. Casti cites, a reasonable person would ask how we possibly could
know anything about climate change?
"In fact, the picture is not so bleak: In the early 1990s, climate
simulation models successfully predicted the impact of the Mount Pinatubo
eruption on global climate. To reconcile these two seemingly contradictory
assessments, we need to distinguish between the limited predictability of
tornadoes and volcanoes, and the much better predictability of climate. For
the former, the metaphor about a butterfly is useful because it suggests
that the ability to predict a weather event for a specific place and time
-- such as severe weather -- diminishes after a few days. However, the
prediction of climate refers to weather statistics over extended periods of
time and not to any particular weather event.
"Thus, we can predict with reasonable certainty that tornadoes will occur
with nearly constant annual frequency within the United States, just as we
can predict that average summer temperatures will be much warmer than
average winter temperatures. Therefore, we have much more confidence in the
prediction of climate statistics than in long-term weather forecasts.
"Scientists have identified certain influences (such as air pollution from
cars and industry) that can alter the climate -- i.e., change the weather
statistics over extended periods. Climate simulation models are the tools
we use to predict the magnitude of this climate change. Their use enables
us to anticipate that the buildup of greenhouse gases over the next century
will lead to global warming, punctuated by more frequent, intense heat
waves and droughts. Butterflies can flap their wings all they want, but it
will be harder for them to stay cool."
Mesoscale Atmospheric Process Branch
NASA -- Goddard Space Flight Center
Center for Climate Systems Research
The writers' lack of understanding of the butterfly effect, and its
implications for climate change science and policy, is profound.
1. Regarding the Butterfly -- it is not all powerful.
The "butterfly effect" is part of the new science of nonlinear dynamics or
"chaos theory". Basically this is the science of systems with lots of
feedback mechanisms and the climate system is certainly one of these. The
butterfly concept is the intrinsic unpredictability of system behavior due
to extreme sensitivity to initial conditions.
The point that they could model the short term effects of Pinatubo is well
taken. I call this the "scale of predictability". Some people call it
"momentum". A phenomenon of a given size is likely to be predictable for a
given time. Other things being equal, the bigger the event the longer the
time. Hurricanes don't dissipate in hours (although their track is often
unpredictable). Pinatubo was a monster, so we could predict its effects for
a year or so, but not for 3 years. Summer will indeed be warmer than winter
for quite a long time, unless something really drastic happens to the axis
of rotation of the earth. Let's hope not.
My benchmark example: If it rains 2" the ground will probably be wet after
24 hours. Whether it will be wet after a week is unknowable. If it only
rains 0.2" it is unknowable even for the 24 hour scale. In many cases the
duration of predictability is proportional to the scale of the cause. I
expect that determining scales of predictability will be a big part of the
new science once it really gets going. As it is we spend a lot of money
predicting the unpredictable.
2. Regarding climate statistics -- there are many Butterflies.
It is sort of true that the butterfly effect only applies to a "particular
event", whatever that means, but there is no end of Butterflies. That there
is a constant or "nearly constant" number of tornadoes each year has to be
nonsense. First, this must mean tornadoes of a certain size or larger, I
assume big ones, since I have little tornadoes in my fields whenever it is
windy and nobody counts them. How could we possibly know how many there
are, since many go unreported? The only possibility, and that remote, is
satellites. Even if we have such they must be quite new, so we have no
record. I do know there are good tornado years and bad, some legendary, at
the state level.
Of course it would be wonderful if their claim were true, because then we
could say "There goes the last tornado this year" or "Only 3 more tornadoes
to go this year". But universal constants are hard to come by and I am sure
this isn't one. Their "nearly constant" probably means they have five years
of satellite data and some variation. Of course the long term average may
not change much -- that's the stability of chaos, which is a point the
writers are tripping over.
Chaos is produced by feedback and all of the major climate mechanisms
provide strong feedback. In fact chaos is a powerful form of stability. The
temperature at my place can change 40 degrees in two days but has not
changed 200 degrees in a million years, so far as we can tell. While
average global temperature has varied chaotically for millions of years, it
appears never to have gone outside of a 15 degree range. That's stability!
The price of this stability is extreme unpredictability within the stable
range, due to the ongoing nonlinear struggle of feedback mechanisms,
so-called "drivers" and "limiters". The butterfly effect.
But chaotic systems, wherein dwells the Butterfly, are also called "far
from equilibrium systems". They exhibit something called "strange
statistics". In far from equilibrium systems the average value is a rare
event, so no good in prediction. (Unless you can find use in a prediction
of the form "it's not likely to be like this.")
My benchmark example: Average annual rainfall in Virginia is about 36
inches, or roughly 0.1"/day. How often does it rain about this much?
Rarely. It either doesn't rain, like most days a year, or it typically
rains from 0.25 to 1.0 inches, sometimes more. Show me the useful statistic
for prediction (hint: if it rains it pours.) Of course if you can get
someone to bet you that it will rain 0.1" on the tenth day of next month,
go for it.
Likewise the general statistics in climate change are that it is usually
getting either warmer or colder, CO2 levels are usually going up or down,
and so on. In the million year scale our data indicates it has usually been
either very warm or very cold. Intermediate values are rare, occurring
mostly in passing from one extreme to the other. That is why we call it a
far from equilibrium system.
This leaves the predictability of the seasons. What $10 billion in climate
change research will buy. Unfortunately it doesn't quite justify the
writers' claim that "Therefore, we have much more confidence in the
prediction of climate statistics than in long-term weather forecasts." We
3. The Fallacy of Equilibrium.
In fact these models only make stable predictions at all because they rely
on what I call the "fallacy of equilibrium", which means not including
those pesky Butterflies. Such predictions only tell us that if all natural
greenhouse gas sinks and sources stay in perfect equilibrium for a century
then we will make a difference. But these sinks and sources -- the oceans,
biospheres, weathering of rocks, atmospheric chemistry, and so on, not to
mention the heat of the sun that drives the whole system, have never been
stable, so why should they sit still for a century? The supposition is
preposterous, given the chaotic nature of the feedback mechanisms, but the
modelers have made it. Here much of the real climate change research, as
opposed to the modeling, has actually paid off by elucidating many of these
chaotic feedbacks in detail. Climate is chaotic at all scales.
As far as the writers' worry that "This kind of statement might contribute
to public confusion and dilute the resolve needed to formulate a reasonable
response to the threat of climate change." It should. We do not and cannot
control the climate.
This Post letter is Kuhn's Problem personified (scientists refusing to
learn a new fundamental theory). They don't understand the new science of
nonlinear dynamics and they don't care. Given that they are in the business
of modeling the dynamics of climate, this ignorance is appalling. Or is it
just that there is no money for modelers in admitting unpredictability?
Best regards to all,
Historical notes: Kuhn being of course the historian of science Thomas
Kuhn, whose pioneering book in the late 1960's: "The Structure of
Scientific Revolutions", first identified this recurring problem of
resistance to revolution. Nobody wants to consider, much less admit, that
their life's work is wrong. My dissertation dealt with this problem and my
first major paper: "The Structure of Technological Revolutions" (1973),
extended it to conceptual revolutions generally. An expanded version of
that paper is Chapter 14 of Bugliarello and Doner's "History and Philosophy
of Technology" (1979). I got 'retrained' in chaos theory in the late 1980's
by an accident of fate.
I cannot overestimate how profound this problem is. If the scientific
community does not understand, or refuses to accept, what is plainly before
its eyes, how can the public be expected to do so? The educational
requirements are staggering. In the meantime our public policy is absurd.
David E. Wojick, Ph.D., P.E.
dwojick at shentel.net
(Click on "PowervisioN" OR "NAAQS on the Web")
391 Flickertail Lane
Star Tannery, VA 22654
More information about the Coral-list-old