MOM6 DA history collection

[guillaumevernieres]

Description

As an alternative to dumping intermittent MOM6 restarts, add a DA history collection to the diag_table that would include hourly snapshots of:

- Potential temp
- salt
- currents
- ave_ssh
- h
- mld
- ...

Deliverable

• A diag_table that will dump the above collection of variables in a file prefixed ocn_da
• multiple option for the vertical coordinate:
  • save on the evolving layers
  • define a vertical coordinate and use that to save the fields
• Check if the initial conditions are saved
• Check if the missing value can be set by the user. Currently an issue when trying to read with soca,

Definition of done

Not clear as of now. Options:

• [x] Generate low-res hourly ocn_da diag files for 2018-04-15-09 to 2018-04-15-18
• [x] Test reading diag files with soca
• [x] Add the diag_table to this repository under .../parm/mom6/diag_table

[guillaumevernieres]

@hyunchul386 , happy to discuss this issue when you have time.

[hyunchul386]

@guillaumevernieres Following is a collection in diag_table for the history of ODA. This diag_table gave six sets of one hourly snap shop in each files. for example, ocn_2021_03_22_13.nc have the one hourly snap shots from 2021-03-22 13Z to 2021-03-22 18Z.

###################### "ocn%4yr%2mo%2dy%2hr", 0, "hours", 1, "hours", "time", 6, "hours", "1901 1 1 0 0 0" ##############################################

static fields

"ocean_model", "geolon", "geolon", "ocn%4yr%2mo%2dy%2hr", "all", .false., "none", 2 "ocean_model", "geolat", "geolat", "ocn%4yr%2mo%2dy%2hr", "all", .false., "none", 2

ocean output TSUV and others

"ocean_model", "SSH", "SSH", "ocn%4yr%2mo%2dy%2hr","all",.false.,"none",2 "ocean_model", "SST", "SST", "ocn%4yr%2mo%2dy%2hr","all",.false.,"none",2 "ocean_model", "SSS", "SSS", "ocn%4yr%2mo%2dy%2hr","all",.false.,"none",2 "ocean_model", "MLD_0125", "MLD", "ocn%4yr%2mo%2dy%2hr","all",.false.,"none",2

Z-Space Fields Provided for CMIP6 (CMOR Names):

===============================================

"ocean_model_z","uo","u" ,"ocn%4yr%2mo%2dy%2hr","all",.false.,"none",2 "ocean_model_z","vo","v" ,"ocn%4yr%2mo%2dy%2hr","all",.false.,"none",2 "ocean_model_z","so","salt" ,"ocn%4yr%2mo%2dy%2hr","all",.false.,"none",2 "ocean_model_z","temp","temp" ,"ocn%4yr%2mo%2dy%2hr","all",.false.,"none",2 "ocean_model_z","h","h" ,"ocn%4yr%2mo%2dy%2hr","all",.false.,"none",2

=============================================================================================

The variables in each file are GEOLON Longitude of tracer (T) points 1:72 1:35 ... ... GEOLAT Latitude of tracer (T) points 1:72 1:35 ... ... SSH Sea Surface Height 1:72 1:35 ... 1:6 SST Sea Surface Temperature 1:72 1:35 ... 1:6 SSS Sea Surface Salinity 1:72 1:35 ... 1:6 MLD Mixed layer depth (delta rho = 1:72 1:35 ... 1:6 U Sea Water X Velocity 1:72 1:35 1:33 1:6 V Sea Water Y Velocity 1:72 1:35 1:33 1:6 SALT Sea Water Salinity 1:72 1:35 1:33 1:6 TEMP Potential Temperature 1:72 1:35 1:33 1:6 H Layer Thickness 1:72 1:35 1:33 1:6

The data from ncdump are netcdf ocn_2021_03_22_13 { dimensions: xh = 72 ; yh = 35 ; time = UNLIMITED ; // (6 currently) xq = 72 ; z_l = 33 ; z_i = 34 ; yq = 35 ; variables: double xh(xh) ; xh:units = "degrees_east" ; xh:long_name = "h point nominal longitude" ; xh:axis = "X" ; double yh(yh) ; yh:units = "degrees_north" ; yh:long_name = "h point nominal latitude" ; yh:axis = "Y" ; double time(time) ; time:units = "hours since 2021-03-22 06:00:00" ; time:long_name = "time" ; time:axis = "T" ; time:calendar_type = "JULIAN" ; time:calendar = "julian" ; double xq(xq) ; xq:units = "degrees_east" ; xq:long_name = "q point nominal longitude" ; xq:axis = "X" ; double z_l(z_l) ; z_l:units = "meters" ; z_l:long_name = "Depth at cell center" ; z_l:axis = "Z" ; z_l:positive = "down" ; z_l:edges = "z_i" ; double z_i(z_i) ; z_i:units = "meters" ; z_i:long_name = "Depth at interface" ; z_i:axis = "Z" ; z_i:positive = "down" ; double yq(yq) ; yq:units = "degrees_north" ; yq:long_name = "q point nominal latitude" ; yq:axis = "Y" ; float geolon(yh, xh) ; geolon:_FillValue = 1.e+20f ; geolon:missing_value = 1.e+20f ; geolon:units = "degrees_east" ; geolon:long_name = "Longitude of tracer (T) points" ; geolon:cell_methods = "time: point" ; float geolat(yh, xh) ; geolat:_FillValue = 1.e+20f ; geolat:missing_value = 1.e+20f ; geolat:units = "degrees_north" ; geolat:long_name = "Latitude of tracer (T) points" ; geolat:cell_methods = "time: point" ; float SSH(time, yh, xh) ; SSH:_FillValue = 1.e+20f ; SSH:missing_value = 1.e+20f ; SSH:units = "m" ; SSH:long_name = "Sea Surface Height" ; SSH:cell_methods = "area:mean yh:mean xh:mean time: point" ; float SST(time, yh, xh) ; SST:_FillValue = 1.e+20f ; SST:missing_value = 1.e+20f ; SST:units = "degC" ; SST:long_name = "Sea Surface Temperature" ; SST:cell_methods = "area:mean yh:mean xh:mean time: point" ; float SSS(time, yh, xh) ; SSS:_FillValue = 1.e+20f ; SSS:missing_value = 1.e+20f ; SSS:units = "psu" ; SSS:long_name = "Sea Surface Salinity" ; SSS:cell_methods = "area:mean yh:mean xh:mean time: point" ; float MLD(time, yh, xh) ; MLD:_FillValue = 1.e+20f ; MLD:missing_value = 1.e+20f ; MLD:units = "m" ; MLD:long_name = "Mixed layer depth (delta rho = 0.125)" ; MLD:cell_methods = "area:mean yh:mean xh:mean time: point" ; float u(time, z_l, yh, xq) ; u:_FillValue = 1.e+20f ; u:missing_value = 1.e+20f ; u:units = "m s-1" ; u:long_name = "Sea Water X Velocity" ; u:cell_methods = "z_l:mean yh:mean xq:point time: point" ; u:standard_name = "sea_water_x_velocity" ; u:interp_method = "none" ; float v(time, z_l, yq, xh) ; v:_FillValue = 1.e+20f ; v:missing_value = 1.e+20f ; v:units = "m s-1" ; v:long_name = "Sea Water Y Velocity" ; v:cell_methods = "z_l:mean yq:point xh:mean time: point" ; v:standard_name = "sea_water_y_velocity" ; v:interp_method = "none" ; float salt(time, z_l, yh, xh) ; salt:_FillValue = 1.e+20f ; salt:missing_value = 1.e+20f ; salt:units = "psu" ; salt:long_name = "Sea Water Salinity" ; salt:cell_methods = "area:mean z_l:mean yh:mean xh:mean time: point" ; salt:standard_name = "sea_water_salinity" ; float temp(time, z_l, yh, xh) ; temp:_FillValue = 1.e+20f ; temp:missing_value = 1.e+20f ; temp:units = "degC" ; temp:long_name = "Potential Temperature" ; temp:cell_methods = "area:mean z_l:mean yh:mean xh:mean time: point" ; float h(time, z_l, yh, xh) ; h:_FillValue = 1.e+20f ; h:missing_value = 1.e+20f ; h:units = "m" ; h:long_name = "Layer Thickness" ; h:cell_methods = "area:mean z_l:sum yh:mean xh:mean time: point" ;

// global attributes: :NumFilesInSet = 1 ; :title = "20210322.06Z.C48.64bit.non-mono" ; :grid_type = "regular" ; :grid_tile = "N/A" ;

The diag_table and outputs can be found at /work/noaa/stmp/hlee/stmp/hlee/FV3_RT/cpld_c48/cpld_control_c48/diag_table_ocn_history /work/noaa/stmp/hlee/stmp/hlee/FV3_RT/cpld_c48/cpld_control_c48/ocn_202103??_??.nc

[ShastriPaturi]

@hyunchul386 you could add: "ocean_model_z","umo","umo" (for transport) "ocean_model_z","vmo","vmo" "ocean_model","MLD_003" , "MLD_003"

and other variables.

[guillaumevernieres]

@hyunchul386 you could add: "ocean_model_z","umo","umo" (for transport) "ocean_model_z","vmo","vmo" "ocean_model","MLD_003" , "MLD_003"

and other variables.

We're building a collection for DA, not diagnostic @ShastriPaturi . But we do need an mld estimate.

[hyunchul386]

The diag_table includes MLD, based on density. Do we need both MLD's?

[guillaumevernieres]

Thanks @hyunchul386 . A few comments:

• We need to call the history collection ocn_da ... or something.
• Dumb question: is there an option to only save 1 snapshot per file and have multiple files? instead of having a Time dimension>1 in the file?
• We need to keep our options open for the vertical coordinate:
  • save on the evolving layers
  • define a vertical coordinate and use that to save the fields
• Can we save the initial conditions?

The diag_table includes MLD, based on density. Do we need both MLD's?

Nope, we only need one.

[ShastriPaturi]

For CICE6 variables:

"icemodel","siconc", "sic", "sicm%4yr%3dy_%2hr","all","mean","none",2

"icemodel","sithick", "sih", "sicm%4yr%3dy_%2hr","all","mean","none",2

"sicm%4yr%3dy%2hr", 1,"days", 1, "days","Time", 1,"days" for daily.

[hyunchul386]

@ShastriPaturi, I will take a look at "siconc" and "sithick", and what is the variable for "hsnon"?

[ShastriPaturi]

@ShastriPaturi, I will take a look at "siconc" and "sithick", and what is the variable for "hsnon"?

Good question. I will need to look.

[guillaumevernieres]

@hyunchul386 & @ShastriPaturi , I'm not too interested in cice for now, sorting out the issues in my comment above would be more useful.

[hyunchul386]

The ocean_da history files by /work/noaa/stmp/hlee/stmp/hlee/FV3_RT/cpld_c48/cpld_control_c48/diag_table_ocn_da :

• the history collections are ocn_da_yyy_mm_dd_hh.nc
• the history files are snap shots at each time
• the 3-D variables (T,S,U,V) are saving in the layers (h) which is evolving with time
• the evolving layers of "h" is saving in the file
• the initial conditioned are saved in 3-hourly
• cice history is excluded in the table, and would need to re-visit later.

[guillaumevernieres]

Thanks @hyunchul386 . A few questions/requests:

• couldn't we just append the new collection to the existing diag_table while still dumping the ocean da variables in a separate file? (thinking longer term when we will want to have that collection in the RT)
• what does mean means here: `temp:cell_methods = "area:mean zl:mean yh:mean xh:mean time: point" ;`
• it looks like we can name the variable whichever way we want. Can you use the naming convention used in the restarts? Or possibly what we use in soca, but I'll need to check this would be readable as a state ... so temp should be Temp, salt Salt, tec ...
• are the initial conditions saved in the history file as well?
• for ssh we've been using ave_ssh which is the average ssh over 1 baroclinic time step. What's the definition of SSH in your history collection?

[hyunchul386]

• the update table (/work/noaa/stmp/hlee/stmp/hlee/FV3_RT/cpld_c48/cpld_control_c48/diag_table) is appended the ocn_da to the existing one
• "mean" in cell methods means the method to get the variable at the cell, e.g. ``temp:cell_methods = "area:mean zl:mean yh:mean xh:mean time: point" ==> temp has been obtained by the average (mean) of xh,yh,zl (so volume averages for the cell) at he time point (snap shot). e.g. : v:cell_methods = "zl:mean yq:point xh:mean time: point" ; ==> v has been obtained by the average of xh, zl (area average) at yh and time.
• fixed : temp -> Temp, salt ->Salt
• SSH in this file is snap shot at the time point, so corresponding to sfc, and in my look, the ave_ssh ( actually ave_ssh_ibc ) is not included in the table in the current MOM6.
• for the initial conditions in the history files, we can save some variables but cannot some variables. So, we cannot save the full initial conditions in the history file.

[guillaumevernieres]

@hyunchul386 it all sounds good. Please double check, but it seems that (u,v) are saved on the native grid, but are marked as "mean" because they have been rotated (meridional/zonal as opposed to following the logical grid). Can you confirm?

[guillaumevernieres]

@hyunchul386 , I generated the diag files for the period specified in the description using your diag_table. I was also able to read them with soca after modifying the layer thickness field.

[hyunchul386]

@guillaumevernieres I am checking the native grid representation and will update it

[hyunchul386]

@guillaumevernieres, in my view,

• the (u,v) fields are only converted in the units from L_T-2 ~> m s-1 at c-grid of (U,V) points, which are the native grids.
• the "mean" is the finite-volume representation of (u,v) in MOM6, in which "zonal" and "meridional" is referring the native grid, not the real East and North.
• it seems that there is no the rotational process from the native grid to the Cartesian grid in the diag field output process.

[hyunchul386]

• create feature/ocn_da_diag_table to add the diag_table as ~/parm/soca/mom6/diag_table

[guillaumevernieres]

• create feature/ocn_da_diag_table to add the diag_table as ~/parm/soca/mom6/diag_table

Sounds good @hyunchul386 . Create a pull request when ready.

[hyunchul386]

@guillaumevernieres There was the confirmation from Jiande, modeling group that the diag_table provide fields at the MOM6 native grid without rotational process. A PR was created.