#################################################################
How to run a forward simulation using the results of an inversion
#################################################################

After :doc:`running an inversion</usertutos/first-inversion/index>`, it can be useful — for example to evaluate the results — to run a forward simulation using the inversion output as input.

This is done by modifying the inversion YAML as follows:

  - change :bash:`WORKDIR`
  - change the :doc:`mode</documentation/configuration/modes/index>` to :bash:`forward` (and remove arguments that are only relevant for the inversion, such as :bash:`minimizer`). Add the optional argument :bash:`use_xb: False` so that output files are not generated from the prior state.
  - in :doc:`controlvect</documentation/plugins/controlvects/index>`, add :bash:`reload_xb: True` and :bash:`reload_file` pointing to the pickle file storing the control vectors, typically :bash:`$WORKDIR_inversion/controlvect_final.pickle`.
  - update :doc:`datavect</documentation/plugins/datavects/index>` components as needed, for example :doc:`concs</documentation/plugins/datavects/standard>` or :doc:`satellites</documentation/plugins/transforms/complex/satellites>` to compare against observations that were not assimilated.