From mn/geo/geoit
Jump to: navigation, search


This package is not yet available, contact Rona Thompson if you are interested in the development: rona.thompson@nilu.no

A Bayesian Inversion Framework

FLEXINVERT is a Bayesian inversion framework designed for use with the FLEXPART Lagrangian transport model. FLEXINVERT optimizes the fluxes of a given atmospheric species based on observations e.g. atmospheric mole fractions from flasks and/or in-situ measurements. The framework was developed by Rona Thompson at NILU and is under ongoing development.
FLEXINVERT is free software released under the GNU General Public License.


The transport model (FLEXPART) is used to relate the observations to changes in surface fluxes and defines the so-called source-receptor functions. Since problems of these kind are ill-conditioned, a prior estimate of the fluxes is required and this is only updated based on the information provided by the observations. The optimal solution is the one that minimizes the error between the modelled and observed mole fractions while remaining within the uncertainty limits set around the prior flux estimate. The fluxes are resolved spatially, on a grid of variable resolution, and at a predefined temporal resolution. Thus each flux variable represents the mean over each grid-cell and time interval. Currently, FLEXINVERT only handles species that may be considered inert within the period of the FLEXPART trajectory run, i.e. species that have no atmospheric loss or production during the particle’s lifetime.
Full documentation is available here PDF and a presentation is available on request.


FLEXINVERT is available from the Rona Thompson directly.

FLEXINVERT is writting in Fortran90 and requires the NetCDF and LAPACK libraries to be installed on your operating system. There is a makefile for use with the gfortran compiler included in the package.


Before running FLEXINVERT, you will need to have created all your FLEXPART grid_time files (and if you use the option for background calculation by coupling to a global model, the grid_initial files as well). Edit the file files.def for your own directory and files and control.def for your own inversion settings.