Update prepobs_errtable.global for GSI based soil DA.#33
Conversation
CoryMartin-NOAA
left a comment
CoryMartin-NOAA
left a comment
There was a problem hiding this comment.
Changes look good from what I can tell but we should wait for @ClaraDraper-NOAA to confirm these errors are correct/what she is expecting them to be set at
|
I've had another look at this: with the new obs table we're not assimilating observation type 183 for T (nor, I assume q) From Table 2, Obs Type 183 are marine and land obs with missing pressure/altimeter, and for which the station height is > 7.5 m ASL. For these obs, a pressure is approximated from the station height (it's also approximated for obs < 7.5 m ASL, however these are then converted to obs type 181, presumably because we have more confidence in the approximated pressure). The 183 obs are being rejected with a prep QC mark of 9. From Table 7, QC Mark 9 means missing obs erorr, or missing pressure obs error (then all obs in the record get flagged). Since we've added the obs errors for 183 in the new obs table, I'm guessing that the issue is that there's no obs pressure error (since there's no obs pressure either, this would make sense). @jiaruidong2017 Can you please double check that the new obs table has the T and q errors for type 183? Obs type 183 are about 5% of our total screen-level obs. My initial thought was that we could over-ride the no-pressure-obs-error flag for type 183, since we don't really need the pressure for 2m obs - setup[tq] places these obs on the surface (dlnp = 1). However, somewhere in the setup[tq], it is using the pressure when inflating the obs error but I haven't yet found where this is happening. For now, I'm leaning towards just not using these obs. @ilianagenkova - do you have anything to add to this? |
|
Update: obs_type 183 are largely over Canada, The Great Lakes, and India. Over land, we will not lose any regional coverage by discarding type 183 (as in, there are obs type 187 and/or 181 already present at/near the same locations). @CoryMartin-NOAA Removing the 183 obs type will also remove the need to discard the obs over Great Lakes. Unless anyone has any objections, I suggest that @jiaruidong2017 removes the 183 obs errors from the new obs table, and we proceed with 181 and 187 only. Note: there are some differences in the final obs_errors that I'd like to track down before approving this PR - these differences are being introduced in setup[tq], so I'm not expecting them to be related to this PR. |
|
Sounds like this solves multiple problems by excluding 183 so that sounds good to me |
|
@ClaraDraper-NOAA I did an initial checks, and found the new error table does have the T and q errors for type 183. See the attached file for two reports with tq errors highlighted in red color from 183 obstype. I will do more investigations and let you know then. https://docs.google.com/document/d/1Uyo1HtvxBU-gKh8DNR4v61sY-zNUXktyzq52aXm7i10/edit?tab=t.0 |
Jiarui - I don't think there's any need to do any more investigation. Can you please just remove the obs errors for T and q for type 183 in the new obs table. |
|
Below is what I got for the obs type 183. There are about 3181 reports with error for t2m and 2922 reports for q2m.
Okay, I will remove the obs errors for T and q for type 183 in the new obs table. |
|
If we remove the obs error for obs type 183, do we still need to add the reject list, because there is only one station left in the list? |
|
What is the one station? I'm thinking not |
|
That single station is on the island of US/Canada boundary as below.
|
ClaraDraper-NOAA
left a comment
ClaraDraper-NOAA
left a comment
There was a problem hiding this comment.
@jiaruidong2017 The new obs table that is here:
/scratch1/NCEPDEV/global/Jiarui.Dong/JEDI/GlobalWorkflow/global-workflow.bufrtest4/fix/prepobs_errtable.global
Currently only has changed obs errors for 183. We want to not change 183 and change 181 and 187.
ClaraDraper-NOAA
left a comment
ClaraDraper-NOAA
left a comment
There was a problem hiding this comment.
The new obs table that is here:
/scratch1/NCEPDEV/global/Jiarui.Dong/JEDI/GlobalWorkflow/global-workflow.bufrtest4/fix/prepobs_errtable.global
Currently only has changed obs errors for 183. We want to not change 183 and change 181 and 187.
Likewise, the version in this PR still has the obs errors inserted for 183. can you please remove these, then make sure that the version on hera and the version here are the same?
|
@ClaraDraper-NOAA , your interpretation of the documented tables is as good as mine. The /fix directory error tables look good (i.e. each column is for a different atmo state variable - t,q,etc). |
|
Removed the changes in error table for obs type 183. |
|
Should the ObsProc team be aware of and/or review these changes? |
|
Sure. @ilianagenkova Please review this PR. |
ilianagenkova
left a comment
ilianagenkova
left a comment
There was a problem hiding this comment.
The changes look good to me.
RussTreadon-NOAA
left a comment
RussTreadon-NOAA
left a comment
There was a problem hiding this comment.
Changes tested in conjunction with GSI PR #866 with acceptable behavior.
Approve.
|
Thanks @RussTreadon-NOAA for your efforts. |
@RussTreadon-NOAA , obsproc's prepobs code reads the error tables from: $COMROOT/gfs/v16.*/fix/prepobs_errtable.global , thus when you guys update any of the tables we will pick it up by default. |
|
Thank you @ilianagenkova . Just want to ensure ObsProc and DA team are on the same page. Given other tasks I increasingly rely upon developers and peer reviewers to fully vet GSI related changes. |
5 participants
We propose to replace the current "prepobs_errtable.global" with a new error table for the GSI base soil DA contributing to GFSv17 implementation.
The main changes are made for surface layers including to add temperature observation error, moisture obs error for 181/183/187 surface types, and to add pressure obs error for 183/187 surface types.
Issue: #32
The linked issue is NOAA-EMC/global-workflow#3546