[extropy-chat] A Bayesian Looks at Climate Change

[Hal Finney]

As Eliezer points out, it is only correct to combine estimates as
independent if they really are independent.  The blog entry
http://julesandjames.blogspot.com/2006/03/climate-sensitivity-is-3c.html
and paper
http://www.jamstec.go.jp/frcgc/research/d5/jdannan/GRL_sensitivity.pdf
describe the following methods for estimating climate sensitivity:

1. Attempting to model and reproduce 20th century temperature records
using climate models.  This requires modelers to fit numerous parameters
including solar and environmental effects, but sensitivity to changes
in energy inputs is one of the most important.  This produces a certain
value and range for climate sensitivity.

2. Looking at actual cooling from volcanic eruptions.  If sensitivity is
high we would see more temperature changes than if sensitivity is low.
The authors note, "Although it might appear that this information
is already implicit in the 20th century reconstructions, those papers
generally did not consider the short-term temperature changes in detail,
instead relying largely on a long-term energy balance. Therefore we
consider it reasonable to treat this constraint as a physically and
observationally independent one."

3. Looking at temperatures during the last ice age.  The temperature
changes were in the opposite direction but the concept of climate
sensitivity still applies, and this leads to an independent estimate.

They also considered some other methods, such as looking at temperature
changes during the "Maunder Minimum" in the late 17th century, when
solar output was thought to be decreased, but they decided that these
were not completely independent of the ones above.

Their main result came from combining the three methods listed above.
In addition, they did some robustness tests which reinforce the result.
The bottom line is that the conventionally quoted uncertainty range can
only be seen as an exaggeration.  The fact that multiple independent
lines of analysis produce similar uncertainties should not be taken to
reinforce the validity of that uncertainty range!  But that seems to be
what has generally been the view in climatology.

Hopefully this new (for this field anyway) way of looking at things
will have a larger influence.  Scientists may be uncomfortable with a
narrow range because it makes their predictions much more falsifiable,
but it is what logically follows from this work.

BTW James Annan has a blog entry out today discussing a report in Nature
that produces a new and somewhat narrower estimate of sensitivity:
http://julesandjames.blogspot.com/2006/04/hegerl-et-al-on-climate-sensitivity.html
However the new range is not nearly as narrow as what James came up with
so he is a little miffed that Nature rejected his paper.  He suggests
that Nature has been excessively alarmist on some of its global warming
positions and was perhaps not too happy with his result throwing cold
water on the catastrophist position.  The paper they published instead
lets them begin to back off more gracefully from the extremes.

Hal