introducing perturbations to GOCART2G

[mpagowski]

At NOAA we are developing ensemble DA for UFS-Aerosols model that relies on GOCART2G. In the past we developed perturbations to emissions for GEFS-Aerosols model (an older NOAA model) using a scheme called SPPT that creates spatially and temporally correlated patterns. This scheme was developed at ECMWF to parameterize model errors and was shown to have benefits for ensemble data assimilation at this institution and also at NCEP. We demonstrated similar benefits for aerosol DA with GEFS-Aerosols. We are considering two, possibly three options to include perturbations in UFS-Aerosols: 1) passing SPPT perturbations to GOCART code (i.e. in-line); 2) perturbing emission input files, distinct for each ensemble member (i.e. off-line); 3) using *.rc files to provide 2D perturbations (i.e. off-line but not sure if at all possible with 2D files). We think that option 1) is the most robust for various reasons that I could elaborate on if there is interest. In short, we would like to pass SPPT perturbations from the physics part of the NOAA model to GOCART in the form of 2D multiplicative coefficients. The coefficients would be applied to sources of dust, sea-salt, carbon/dust from wildfires and anthropogenic aerosols. As a default there would be no perturbations and the results would be identical to the original. Please let me know what you think about these ideas and if NASA would support the concept and let us introduce perturbation codes to the repository.

[mathomp4]

@mpagowski A note: we do have a code in GMAO_Shared, GMAO_stoch, that has an sppt.rc file in it:

https://github.com/GEOS-ESM/GMAO_Shared/tree/main/GMAO_stoch

so perhaps GEOS as a version of SPPT, if not the same version you use.

NOTE: You might not see it in the GMAO_Shared that GOCART brings in because it is currently sparsed out, but it could be brought in if necessary.

[mpagowski]

@mathomp4: thanks for pointing me to the stochastic code. It looks very similar to what exists at NOAA. It seems that bringing perturbations to GOCART2G would be much simpler than I expected. Then the question is if NASA is interested in such modifications to GOCART2G and would allow us to do it.

[mathomp4]

@mathomp4: thanks for pointing me to the stochastic code. It looks very similar to what exists at NOAA. It seems that bringing perturbations to GOCART2G would be much simpler than I expected. Then the question is if NASA is interested in such modifications to GOCART2G and would allow us to do it.

I suppose this is a question for @amdasilva and @pcolarco at least. Maybe even @elakkraoui and @rtodling as it is DA-related (though maybe not in their bailiwick).

[pcolarco]

Add @vbuchard to this as she is the one hands on with the ensemble DA work. I'm not disinterested, but not knowledgable about the need.

[amdasilva]

Putting this on hold while we make a release for GEOS-FP.