Thursday, November 1, 2007
Dynamical climate reconstruction
Historical properties of Earth's climate are obtained from proxy
measurements at select locations. Time series analysis of these
measurements is often used to infer changes about climate properties,
such as temperature, on larger scales, such as the global-mean. Further
understanding is sometimes derived by using information from the observations as a basis for numerical simulation.
An alternative approach to studying historical climate variability is
proposed here by fusing models and observations using state estimation.
Unlike methods using observations or models alone, this method has the
attractive attribute of providing the dynamically consistent
three-dimensional structure that is consistent with available observations. Moreover, Gaussian error statistics for the recovered states are included. Since paleoclimate measurements usually represent averages, or integrals, in time, standard methods require generalization. Here we present a method that accounts for these averages and for the possibility that climate statistics are not stationary in time. Results from experiments using this method on idealized models of Earth's climate will be discussed, including tests of a theory to optimally determine locations for high-impact observations. Plans to implement the technique with real observations will also be discussed.
Greg Hakim is an Associate Professor at the University of Washington's Department of Atmospheric Sciences.