Run JEDI-MPAS on regional mesh


I am trying to run JEDI-MPAS on a regional mesh, but I encountered the following error.

==> destruct MPAS corelist and domain:  0
BadValue: Invalid latitude 116.86172266098012 caught in 
Exception: saber::EstimateParams<MPAS> terminating...
Exception stack:
BadValue: Invalid latitude 116.86172266098012
backtrace [1] stack has 20 addresses

I suspect that when it is reading latCell during sampling, it samples a cell that is out of the boundary, so it loads a value in lonCell. Is there any setting that I need to be aware of if I want to use regional mesh? Thanks.

Since we are interested in high-resolution DA on regional mesh in MPAS,

There are doubts we encountered:

  1. Does JEDI-MPAS on regional mesh follow the same procedure as global mesh?
  2. Are there any specific settings of which I should be aware?

Thanks a lot for the help!

First, please note that, while it exists in the develop branch, the regional capability is neither released nor generally supported yet. You may well be exploring configurations that we have not tried.

Concerning the questions you pose:

  1. Yes, with the exception given next.
  2. You’ll need to ensure that observations outside the domain are rejected by the QC or remove all observations outside the domain from your files before ingest by IODA. We have used the first approach, via observation filters. (This may well no longer be necessary, as recent updates of the generic interpolation code should flag obs for which the interpolation fails. We haven’t checked this with a regional domain yet.)

Here’s an example yaml for the obs filter:

obs filters:
    - filter: Bounds Check
      filter variables:
      - name: eastward_wind
      - name: northward_wind
      test variables:
      - name:  LAMDomainCheck@ObsFunction
          map_projection: circle # an option currently supported
          save: true # will save the DerivedValue to IODA output file
          cenlat: 40.0    # central lat
          cenlon: 260.0  # central lon
          radius: 2750.0  # km
      minvalue: 1.0 # will filter all obs outside of the specified limited area domain

Our experiments to date have used pure EnVar and a regional mesh that is a “circular” subset of the x5.6488066 global mesh. We can’t comment yet on using BUMP to set up a static background covariance on a regional mesh or on using regional meshes with more complicated boundaries.

Finally, if you are attempting a configuration we’ve not tried, please let us know what seems to work and what seems to fail.