Tropical Mooring Data Appears Mostly Limited to Surface Observations

[guillaumevernieres]

This issue has existed since the beginning of the BUFR-to-IODA project. The figure below shows a comparison of the RTOFS and obsForge profiles.

The orange lines correspond to the tropical moorings.

One possible explanation is that resolving this issue may simply require looking farther back in time, outside the DA window, to access the deeper observations.

The information below was provided by @sanAkel. It lists the files used to prepare the in situ observations for RTOFS:

$ for file in /lfs/h1/ops/prod/com/rtofs/v2.5/rtofs.20260327/dump/*out.20260326; do name="${file##*/}"; clean_name="${name%%.out.*}"; grep 'dcom' "$file" | head -n 1 | sed "s|^|${clean_name}: |"; done
bathy:  -1- /lfs/h1/ops/prod/dcom/202603/b031/xx001
dbuoyb:  -1- /lfs/h1/ops/prod/dcom/20260320/b001/xx102
mbuoyb:  -1- /lfs/h1/ops/prod/dcom/20260320/b001/xx103
shipsb:  -1- /lfs/h1/ops/prod/dcom/20260324/b001/xx101
ships:  -1- /lfs/h1/ops/prod/dcom/20260324/b001/xx001
shipsu:  -1- /lfs/h1/ops/prod/dcom/20260324/b001/xx013
shipub:  -1- /lfs/h1/ops/prod/dcom/20260324/b001/xx113
subpfl:  -1- /lfs/h1/ops/prod/dcom/202603/b031/xx005
tesac:  -1- /lfs/h1/ops/prod/dcom/202603/b031/xx002
xbtctd:  -1- /lfs/h1/ops/prod/dcom/202603/b031/xx006

Location of the processing script:

/lfs/h1/ops/prod/packages/rtofs.v2.5.4/ush/rtofs_ncoda_prep_sfc_sfcr_prof.sh

Example log file:

/lfs/h1/ops/prod/com/rtofs/v2.5/rtofs.20260327/rtofs_glo.t00z.ncoda_qc.OUTPUT.972249

Example usage:

for dtyp in shipsu ships shipub shipsb; do
  /apps/ops/prod/nco/intel/19.1.3.304/bufr_dump.v1.2.1/ush/dumpjb 2026032412 11.999 $dtyp
done

A key difference is that RTOFS extracts observations from the previous 12 hours, whereas our workflow likely only considers observations within the 6-hour DA window. This may explain why deeper tropical mooring observations are missing.

[sanAkel]

@guillaumevernieres Some more info reg RTOFS processing ⬇️

1. The script makes the dumps for *all ob types.
2. For profiles.
3. For surface, e.g., drifters.

• Log from ⬆️ 1 can be found here.
• The lines/info that @guillaumevernieres copy/pasted can be found in the following files:

santha.akella@clogin07:~> ls /lfs/h1/ops/prod/com/rtofs/v2.5/rtofs.20260327/dump/
bathy.20260326      dbuoyb.out.20260326          mbuoyb.20260326      shipsb.20260326      shipsu.20260326      shipub.out.20260326  subpfl.20260326      tesac.out.20260326
bathy.out.20260326  l2out.f285.51.2026032500.nc  mbuoyb.out.20260326  shipsb.out.20260326  shipsu.out.20260326  ssmisu.20260325.out  subpfl.out.20260326  xbtctd.20260326
dbuoyb.20260326     l2out.f285.51.2026032600.nc  ships.20260326       ships.out.20260326   shipub.20260326      ssmisu.20260326.out  tesac.20260326       xbtctd.out.20260326

• The dir structure of what was finally used:

santha.akella@clogin07:~> ls /lfs/h1/ops/prod/com/rtofs/v2.5/rtofs.20260327/ncoda/
glbl_var  hycom_var  logs  nhem_var  ocnqc  shem_var
santha.akella@clogin07:~> ls /lfs/h1/ops/prod/com/rtofs/v2.5/rtofs.20260327/ncoda/ocnqc/
amsr  goes  himawari  ice  incoming  metop  msg  profile  sfc  ssh  sss  velocity  viirs
santha.akella@clogin07:~> ls /lfs/h1/ops/prod/com/rtofs/v2.5/rtofs.20260327/ncoda/ocnqc/profile/
2026031200.profile  2026031400.profile  2026031600.profile  2026031800.profile  2026032000.profile  2026032200.profile  2026032400.profile  2026032600.profile
2026031300.profile  2026031500.profile  2026031700.profile  2026031900.profile  2026032100.profile  2026032300.profile  2026032500.profile
santha.akella@clogin07:~> 
santha.akella@clogin07:~> ls /lfs/h1/ops/prod/com/rtofs/v2.5/rtofs.20260327/ncoda/ocnqc/sfc/
2026031200.sfc  2026031300.sfc  2026031400.sfc  2026031500.sfc  2026031600.sfc  2026031700.sfc  2026031800.sfc  2026031900.sfc  2026032000.sfc  2026032100.sfc  2026032200.sfc  

• Perhaps a self-descriptive format helps!

santha.akella@clogin07:~> ls large_fs/RTOFS_OM/v2.5/obs_stat/2026032*/*prof*
large_fs/RTOFS_OM/v2.5/obs_stat/20260321/20260320.profile.nc  large_fs/RTOFS_OM/v2.5/obs_stat/20260324/20260323.profile.nc  large_fs/RTOFS_OM/v2.5/obs_stat/20260327/20260326.profile.nc
large_fs/RTOFS_OM/v2.5/obs_stat/20260322/20260321.profile.nc  large_fs/RTOFS_OM/v2.5/obs_stat/20260325/20260324.profile.nc
large_fs/RTOFS_OM/v2.5/obs_stat/20260323/20260322.profile.nc  large_fs/RTOFS_OM/v2.5/obs_stat/20260326/20260325.profile.nc
santha.akella@clogin07:~> 
santha.akella@clogin07:~> ls large_fs/RTOFS_OM/v2.5/obs_stat/2026032*/*sfc*
large_fs/RTOFS_OM/v2.5/obs_stat/20260321/20260320.sfc.nc  large_fs/RTOFS_OM/v2.5/obs_stat/20260323/20260322.sfc.nc  large_fs/RTOFS_OM/v2.5/obs_stat/20260325/20260324.sfc.nc  large_fs/RTOFS_OM/v2.5/obs_stat/20260327/20260326.sfc.nc
large_fs/RTOFS_OM/v2.5/obs_stat/20260322/20260321.sfc.nc  large_fs/RTOFS_OM/v2.5/obs_stat/20260324/20260323.sfc.nc  large_fs/RTOFS_OM/v2.5/obs_stat/20260326/20260325.sfc.nc

[sanAkel]

@guillaumevernieres if the ⬆️ info suffices, please unassign me.

[ilianagenkova]

@sanAkel , could you pls clarify - are the /lfs/h1/ops/prod/com/rtofs/v2.5/rtofs.20260327/ncoda/ocnqc/profile/ binary non-BUFR files the QC-ed version of the /lfs/h1/ops/prod/com/rtofs/v2.5/rtofs.20260327/dump/ files?

[guillaumevernieres]

@sanAkel , could you pls clarify - are the /lfs/h1/ops/prod/com/rtofs/v2.5/rtofs.20260327/ncoda/ocnqc/profile/ binary non-BUFR files the QC-ed version of the /lfs/h1/ops/prod/com/rtofs/v2.5/rtofs.20260327/dump/ files?

@ilianagenkova : Yes.

[ilianagenkova]

@guillaumevernieres , I have been saying for a while now that for GFS dumps we use stricter cutoff time than RTOFS uses. If we need to expand the dump time windows to the left (look further in the past), we will have to update the legacy obsproc code.
How far back do you want to go?
At the moment most data sets look 3h back, except proflr (4h), axbt (6h), abtctd (6h), altkob (15h).

If I read the plots above correctly, you want:
TAO,TRITON,PIRATA,RAMA profiles to be deeper.
Glider reaches ~500m, is that the expected depth?
Ship/CTD (TESAC) , Drifter SST, NDBC buoy SST - do these need to go deeper too?

@givelberg @apchoiCMD, which dump files, or which tanks, are these data - I need your help matching tanks numbers or dump filenames to the data type descriptions Guillaume tends to use (which I still don't know by heart)

[apchoiCMD]

@ilianagenkova here it is : /lfs/h1/ops/prod/com/obsproc/v1.2

[guillaumevernieres]

@guillaumevernieres , I have been saying for a while now that for GFS dumps we use stricter cutoff time than RTOFS uses. If we need to expand the dump time windows to the left (look further in the past), we will have to update the legacy obsproc code. How far back do you want to go? At the moment most data sets look 3h back, except proflr (4h), axbt (6h), abtctd (6h), altkob (15h).

If I read the plots above correctly, you want: TAO,TRITON,PIRATA,RAMA profiles to be deeper. Glider reaches ~500m, is that the expected depth? Ship/CTD (TESAC) , Drifter SST, NDBC buoy SST - do these need to go deeper too?

@givelberg @apchoiCMD, which dump files, or which tanks, are these data - I need your help matching tanks numbers or dump filenames to the data type descriptions Guillaume tends to use (which I still don't know by heart)

@ilianagenkova Gilders ~600+m deep, Tropical moorings ~700+m deep.
12 hours seems like a reasonable number to start with for the moorings, gliders and xbt/ctd's. I'll let you translate to whatever that means in terms of tanks!

[sanAkel]

FWIW, RTOFS goes -3 days!

[sanAkel]

@ilianagenkova

Drifter SST, NDBC buoy SST - do these need to go deeper too?

No, these are drogue to a depth of 20 m, measure near surface, typically < 1 m (drifter), 1- 2 m (NDBC).

Ship can be deep, but as far as I seen < 50 m max- I personally would advise not to assimilate them!

CTD can go to near bottom.

[JohnSteffen-NOAA]

A useful webpage from PMEL describes the expected data latency: https://www.pmel.noaa.gov/gtmba/gts-data-distribution

"Over half of the TAO data placed on the GTS are available within 3 hours or less and over 90% are available within 7 hours."