Preparation

Atmospheric data

The atmospheric data as provided by the measuring institutions need to be converted into the internal format used in the inversion system.

• For CO₂, available data records are collated in internal format in "~m212046/INVERSION/DATA/TS.co2/VERSION/". A new version of the CO₂ data collection will normally be available in July each year (containing data until the previous year, plus some preliminary data until spring of the current year).
• For some other species (CH₄), data collections are maintained by various colleages.
• If you consider a species for which data in internal format do not yet exist, data collection/conversion needs to be set up.

Inversion set-up

The set-up comprises various parameters and input fields defining the state space (including its a-priori probability distribution) of the inversion, as well as further parameters concerning data treatment etc. All the parameters of a given set-up are set in user-specific file "parameters.SPECIES.SETUP.F" where SETUP is the name of the set-up chosen by the user (as for all user-defined names in the inversion system, choose short and simple set-up names without special characters, because the set-up name will become part of the directory name). The set-up parameter files reside in the user's "~/INVERSION/DATA/RUN/" directory. They contain several FORTRAN subroutines that will be called by the inversion code at pertinent stages during the initialization and the postprocessing:

• Subroutine parameters:
  Setting the set-up parameters. The names, default values, and meanings (short explanations) of the set-up parameters can be found in the beginning of the code file "~/INVERSION/DATA/RUN/src/include.f90".
• Subroutine setpriors_special (optional):
  User-defined code to calculate the prior and shape (and possibly truth) fields during the construction of the flux model at the beginning of an inversion run. In particular, this is place to do grid conversion (see below).
• Subroutine modpriors_special (optional):
  Possibly modify shape after its normalization.
• Subroutines integrate_special, integrate_special2, integrate_special3 (optional):
  User-defined code for postprocessing (for calculations not part of the standard postprocessing output).

Tip: The parameter file "parameters.SPECIES.SETUP.F" for a new set-up is best developed starting from a pre-existing parameter file as a template. For example, the standard CO₂ inversion set-up "~m212046/INVERSION/RUN/parameters.co2.ATMv2021.F" can be used as a template. Note that the Jena inversion system is constantly evolving, introducing new set-up parameters for further options or new core subroutines to make the definition of set-ups easier.

Input fields

In most cases, inversion set-ups involve external spatio-temporal data sets for prior fluxes, shapes, extra constraints, or other fields. In general, the external data sets need to be converted to the grid and time step of the inversion. The inversion core code contains subroutines for grid conversion to be called "on the fly" during the initialization stage of the inversion run in the user-defined subroutine setpriors_special. (By that, all the processing steps are contained in the single parameter file in a transparent and easily reproducable way. In particular, no additional steps are needed if the inversion is run on several grids, e.g., both for global and regional domains.)

The usage of the grid conversion subroutines is best taken from a suitable template set-up (see above). Moreover, the core code file "~/INVERSION/RUN/src/ingest.f90" contains subroutines to grid-convert various important fields (such as fossil fuel CO₂ emissions from GCP [Jones et al.]) which can be used as such or as templates as well.

For many input data sets, grid conversion is fast enough to be done "on the fly" in every single run. If it takes inconveniently long (which may happen if the original data set has very high resolution), it is recommended to do a separate "ingestion" run (see the INGEST option in the template set-up). Feel free to contact Christian Rödenbeck for assistence.

Atmospheric transport model

The default global atmospheric transport model of the Jena inversion system is TM3. The meteorological driver fields needed for TM3 are maintained as part of the system (yearly updates, normally done in June or July). Currently, meteorological driver fields based on NCEP are available from 1950. Meteorological driver fields based on ERA-Interim are available for a limited period; it is planned to add meteorological driver fields based on ERA5 as soon as possible.

For regional inversions, pre-calculated transport footprints from the STILT model are used. The transport footprints are specific for the chosen regional domain as well as for the atmospheric data points to be used. For various regional domains footprints are already available from current projects. For new regional domains or years without footprints, it is necessary to calculate the required footprints before a regional inversion can be done.

The Jena inversion system could also be coupled with other transport models, provided these have an adjoint. For Eulerian models, some new coding would be needed to interface between the inversion and transport model codes. Alternatively, the new transport model could be used to calculate footprints, avoiding the need of a direct code coupling.