GCMs-misuse? part II

rjw9y rjw9y at envsci.evsc.virginia.edu.
Thu Dec 9 16:15:26 EST 1993


I am the author of GCMs-misuse.  Thanks to Rick Adair for coming to my
rescue.  I responded to Len's grammar flame by e-mail to him.  My
feeling is that some jargon like '2xCO2' and 'eigenevctor analysis' may
have been the problem.

To mt;
    
you responded with'

"While I'm not one to claim that no drivel gets past peer review, the
statement that modellers don't "want" their output compared to
observational
data misrepresents the situation badly."

A peer review from the journal Science recently responded, 
"These simulations [transient GCM 2xCO2 runs] are not intended to be
'transient forecasts'".  .... Natural variability is certainly present,
and the model could not replicate this because of the way it has been
initialized.  ...  Those who work with GCMs know the problems
associated with their models.  This paper is not an objective critique
of those problems."

Peer reviews from the journal Nature were not dissimilar.  One author
tells us we have confused GCM output, designed as a sensitivity
analysis, with climate forecasts.  If we have confused them then so
have the myriad of published papers which use transient 2xCO2 GCM
output to predict the future responce of a number of physical and
ecological systems.

It is argued that GCMs cannot be compared to observations for their
output runs were not meant as forecasts.  Ecologists, economists, and
others are readily supplied with GCM output to be used in this way-and
they get published!!!  By not allowing statistical analyses of GCM
accuracy to demonstrate that GCM output should not be used as a
forecast, there is obvious bias in the climate change literature. 
Since Science and Nature both publish such work and GCMers distribute
their output freely to these authors,  both parties (or all three if
you count both journals) are supporting this bias.

And i dont understand how you have come to believe, as you wrote, " The
largest observed anomalies do in fact have a spatial pattern
consistent with that of the regions where most severe warming is
expected."
I would like to know what measurements or papers you have read which
have lead you to this conclusion.  The paper submitted to Science and
Nature about which i speak clealy demonstrates that the GFDL transient
(1%/yr) GCM run introduces error to the tune of 240% greater than the
climate change it correctly captures when compared to 20th century
ground-based temperatures. 

I agree with the modellers in that GCMs are great tools which can be
used to study the atmosphere and evaluate our ability to simulate it. 
However these same GCMs have no buisness being used to predict climate
change in any sort of quantitative manner.  When you use GCM output as
data you are assuming that they correctly simulate future climate in a
quantitative manner.  A number of studies have shown that GCMs do a
poor job of simulating today's climate especially on scales smaller
than regional.  The misuse of GCM output is the prime reason they have
gotten such a bad rap.

"People and critters" as you put it live in a world with pronounced
natural cliamte variability exhibited in a myriad of proxy climate
records (tree rings, Devils Hole calcite, ice cores, etc).  Since we
dont even understand yet how clouds, the ocean, or terrestrial
vegetation will respond to CO2 emissions how can we expect the models
we build to accurately forcast climates of the future????

I welcome your e-mail responce or postings.  I really think this is a
problem which is hidden in the closet.  On the otherhand please dont
take my writing as antagonlistic i enjoy this subject and would like to
encourage these discussions.

bob w
e-mail=     rjw9y at virginia.edu         



More information about the Ag-forst mailing list