Ensemble Kalman Inversion
Abstract
We study the use of ensemble Kalman methods for inversion in the context of geophysical applications. The formulation of the algorithms as a time-continuous derivative-free optimization method is emphasized. We also demonstrate a methodology for imposing constraints on the output of the algorithm, and a methodology for employing time averages of data to improve the inversion for parameters in chaotic dynamical systems. Illustrations are given for the estimation of parameters in a sub-grid scale cloud model, within a GCM, and for the estimation of soil properties from surface response to earthquakes.
Contact information
Prof. Andrew Mark Stuart California Institute of Technology