Get rid of DRIPCG#850
Conversation
RussTreadon-NOAA
left a comment
RussTreadon-NOAA
left a comment
There was a problem hiding this comment.
Changes are straight forward. Yes, g-w parm/config.gfs/config.atmanl will need to replace
export ATMVARYAML=${HOMEgfs}/sorc/gdas.cd/parm/atm/variational/3dvar_dripcg.yaml
with
export ATMVARYAML=${HOMEgfs}/sorc/gdas.cd/parm/atm/variational/3dvar_pcg.yaml
One question. Why do we switch to PCG instead of DRPCG. I thought the problem was with the I (inexact) part of the PCG. What do JCSDA and GMAO use for the variational DA runs?
|
Good question @RussTreadon-NOAA , and I have no good answer! I don't even know why we started with DRIPCG besides that is what a lot of the JCSDA test YAMLs use. |
|
@shoyokota may have some insights regarding which minimizer, among various options, we should consider using. Of course, we should test various minimizers in single obs, stand-alone, and cycled modes. Doing so is not a pre-requisite for this PR. |
|
@DavidNew-NOAA is going to test a few in the next week or so for his AMS poster. But that will be single analysis testing and not cycling. I think GMAO is using DRPCG, or at least Ricardo has a PR to OOPS to fix DRPCG and DRIPCG. |
andytangborn
left a comment
andytangborn
left a comment
There was a problem hiding this comment.
Seems simple enough. I can run a single analysis time experiment to see what the impact is going to be for aerosol DA.
DavidNew-NOAA
left a comment
DavidNew-NOAA
left a comment
There was a problem hiding this comment.
Seems pretty straightforward. I'll get some data on algorithm performance next week with the experiments for my AMS poster.
CoryMartin-NOAA
dismissed stale reviews from DavidNew-NOAA and andytangborn
via
41387bd
|
@DavidNew-NOAA found that PCG is slow, and NRL is apparently using DRPCG, so I've updated to use DRPCG. |
RussTreadon-NOAA
left a comment
RussTreadon-NOAA
left a comment
There was a problem hiding this comment.
Looks good to me. Approve.
|
Thanks @RussTreadon-NOAA you beat me to it! I think there may be some minor changes that will need to be made to g-w config files when this PR is merged. Like the 3dvar YAML template for example. I'll merge this now. |
A little (tall?) birdie has informed us that we apparently should not be using DRIPCG, and seems we are using DRIPCG in our tests. This PR removes that. Note that the global-workflow will likely need modified in `config.atmanl` to use the different YAML name.
4 participants
A little (tall?) birdie has informed us that we apparently should not be using DRIPCG, and seems we are using DRIPCG in our tests. This PR removes that. Note that the global-workflow will likely need modified in
config.atmanlto use the different YAML name.