Title: 'Humpty Dumpty and regional climate change'

Abstract: A climate model is a dynamical system of interacting processes. Climate
modelers disassemble the system into it's individual pieces and try to
improve each one in the hope that the result will aggregate into an
improved representation of nature. Confidence in the predictions of such
models depend on answering such questions as: how does uncertainty in one
particular process contribute to uncertainty in the climate response? How
well must the connections between processes be known to constrain the
response within desired bounds? These questions are naturally addressed
within a feedback analysis, which is well-developed for the global-average
climate response. However impacts from global-mean changes are limited and
the real goal is regional climate prediction. The same questions arise, and
another can be added: how does uncertainty in the processes in one region
contribute to the uncertainty in the response of another region? I'll
present a careful analysis of the spatial pattern of local climate
feedbacks, and address the basic challenge: can we put the pieces back
together again?

Speaker's website: http://earthweb.ess.washington.edu/roe/GerardWeb/Home.html