Inverse Methods

Two variational inverse methods have been implemented within Elmer/Ice to inverse either the basal friction coefficient field beta(x,y) or the fluidity parameter A(x,y,z). Both methods are based on the minimisation of a cost function that measures the mismatch between modelled and observed velocities. These two methods are:

  • the linear adjoint method, following Morlighem et al. (2010)
  • a Robin type method, from Arthen and Gudmundsson (2010)

some info about the Robin type method, in the outputfile related to the costfunction the columns follow this convention:
1st column: number of iteration
2nd column: total cost function
3rd column: surface term of the cost function measuring the mismatch between Neumann and Dirichlet problem
4th column: integral on the beta of the derivative of beta (penalisation term)
For the Lambda-curve plot column 4 has to be plotted against column 3 for different values of lambda. Note, that it is $2=$3+0.5*lambda$4“.

Adjoint inverse methods have been re-organised in Spring 2020; documentation is available with the elmerice sources here.

problems/inverse.txt · Last modified: 2020/05/12 12:55 by fgillet
CC Attribution-Share Alike 4.0 International Valid CSS Driven by DokuWiki do yourself a favour and use a real browser - get firefox!! Recent changes RSS feed Valid XHTML 1.0