We introduce a non-linear Bayesian inversion approach (e.g., Tarantola, 2005; Rodgers, 2000) to estimate the basal properties, that is bedrock topography and basal slipperiness, along a flow line of Rutford ice stream from observations of surface topography and surface velocities. The forward function describing the relationship between the basal properties and the observations is solved numerically with a plane-strain finite-element model. As there exist no explicit solution for the inversion of this non-linear forward function, the solution of the inverse problem is sought numerically and iteratively by using a nonlinear Gauss-Newton procedure. The first order forward model derivatives needed for inversion are approximated by analytical linear transfer functions (Gudmundsson, 2003). This approximation is attractive as it greatly enhances the numerical efficiency of the method by sparing the time-consuming evaluation of the numerical derivatives (Chapter 5 in this thesis; Raymond and Gudmundsson (2007)). Inversions performed with synthetic data showed that the inversion procedure behaves correctly and converges to the correct solution. The basal properties we compute for Rutford ice stream are consistent both with the surface observations and the radar measurements of bedrock topography.
|Number of pages||33|
|Journal||Mitteilungen der Versuchsanstalt fur Wasserbau, Hydrologie und Glaziologie an der Eidgenossischen Technischen Hochschule Zurich|
|Publication status||Published - 1 Dec 2007|