What obs operator should I use to assimilate surface observation?


I am trying to assimilate surface observation using JEDI-MPAS 1.0.0.

I tried a few operators listed in this page (ufo/src/ufo/operators at master · JCSDA/ufo · GitHub), such as VertInterp, Identity, GSISfcModel. However, I either got wrong results or encountered errors with these operators. I wonder what obs operator should I use to assimilate surface observation.


1 Like

All three of these operators have worked fine when using FV3-jedi. They should also work with MPAS, but you will have to describe the problem(s) in more detail. Simply provide the yaml file you are attempting and more complete diagnostic info and we can likely help you.

1 Like

Hi @gthompsnJCSDA ,

Thanks for your help. Actually I found out that I forgot to specify vertical coordinate: height for VertInterp so segmentation fault occurred. I will stick to VertInterp for now because it seems to produce the best result.

1 Like

One thing to note is that for surface pressure observations, JEDI-MPAS developers use SfcPCorrected as shown in this YAML stub:

It includes a station height correction to account for smoothing in the model height field.

Hi @jjguerrette ,

Thanks for your reminder. I wanted to use SfcPCorrected for surface pressure and VertInterp for other variables. However, SfcPCorrected requires linear obs operator, and I couldn’t find a way to apply linear obs operator to surface pressure only and not other variables. How can I write in the YAML to achieve that? Thanks.

This is the obs operator part of my YAML:

    obs operator:
      name: Composite
      - name: VertInterp
        vertical coordinate: height
        - name: air_temperature
        - name: eastward_wind
        - name: northward_wind
      - name: SfcPCorrected
        da_psfc_scheme: UKMO
        - name: surface_pressure
    linear obs operator:
      name: Identity

Unfortunately I do not have any experience with the composite operator. My usual practice is to use the SfcPCorrected operator with a separate obs space than what is used for other profile type variables. Then there will only be surface_pressure as a simulated variable. Note that when reading the same observation file twice for two different obs space sections, only the simulated variables associated with each obs space are stored in memory. There will be duplication of MetaData entries, however.

I see. I will definitely give it a try. Thanks!