Curtailing the da_state variables

[clark-evans]

The mpasout files produced by the da_state stream are rather large, an issue that is exacerbated when doing ensemble cycling. There is a desire from our AVID colleagues to reduce their sizes as much as possible.

The current stream_list.atmosphere.da_state contents are provided below. It can be found on ursa at /scratch4/BMC/wrfruc/chunhua/code/RRFSv2/dev3/rrfs-workflow.20260106/fix/stream_list/hrrrv5/. We need to review and suggest variables that can be removed. It may be helpful to review the "blame" mode of core_atmosphere/Registry.xml to see when variables entered the Registry and to what model component(s) they connect.

scalars
initial_time
xtime
Time
u
w
rho_zz
theta_m
rho
pressure_p
pressure_base
theta
relhum
rho_base
theta_base
uReconstructZonal
uReconstructMeridional
circulation
exner
exner_base
rtheta_base
ru
ru_p
rw
rw_p
rtheta_p
rho_p
surface_pressure
rainprod
evapprod
isltyp
soilf
ivgtyp
landusef
mminlu
landmask
shdmin
shdmax
snoalb
albedo12m
greenfrac
lai12m
cldfrac
re_cloud
re_ice
re_snow
nwfa2d
nifa2d
nt_c
mu_c
tend_sfc_pressure
rt_diabatic_tend
nsteps_accum
ndays_accum
tlag
tday_accum
tyear_mean
tyear_accum
refl10cm
cuprec
nca
cubot
cutop
w0avg
qcg
sh3d
sm3d
qke
qke_adv
cov
qsq
tsq
el_pbl
cldfrac_bl
qc_bl
qi_bl
ch
kpbl
hpbl
hfx
cldfrac_cu
mol
qfx
qsfc
zol
br
cd
cda
chs
chs2
cqs2
cqs
ck
cka
cpm
flhc
flqc
gz1oz0
lh
psim
psih
qgh
regime
rmol
wspd
fh
fm
th2m
dusfcg
dvsfcg
dtaux3d
dtauy3d
rubldiff
rvbldiff
gsw
glw
o3clim
o3vmr
acsnom
acsnow
canwat
chklowq
grdflx
noahres
potevp
sfcrunoff
smstav
smstot
snopcx
snotime
sstsk
sstsk_dtc
sstsk_dtw
udrunoff
rhosnf
precipfr
qsg
qvg
dew
soilt1
tsnav
acrunoff
snowfallac
smfr3d
k22_shallow
kbcon_shallow
ktop_shallow
ktop_deep
xmb_total
xmb_shallow
qc_cu
qi_cu
rthratensw
rthratenlw
vcpool
sub3d_rthcuten
sub3d_rqvcuten
sub3d_rucuten
sub3d_rvcuten
sigma_deep
smcrel
t_oml
t_oml_initial
t_oml_200m_initial
h_oml
hu_oml
hv_oml
lai
isice_lu
iswater_lu
sfc_albbck
sfc_albedo
sfc_emibck
mavail
sfc_emiss
thc
ust
xicem
z0
znt
skintemp
snow
snowc
snowh
sst
tmn
vegfra
seaice
xice
xland
u10
v10
q2
t2m
precipw
dzs
zs
ter
sh2o
smois
tslb
h_oml_initial

[clark-evans]

Some initial thoughts:

• Anything with _oml in the variable name could be removed, as it relates to the MPAS 1-D mixed-layer ocean model that we are not currently using. This would need to change should we begin doing so, however.
• noahres is the residual of the Noah LSM budget and can be removed. Many of the variables surrounding noahres entered the MPAS code with the Noah LSM, but we'd need subject-matter experts (for NoahMP and RUC LSM) to inform which need to stay and which can be removed.
• Apart from the shortwave (rthratensw) and longwave (rthratenlw) radiation tendencies, most of the variables between k22_shallow and sigma_deep can be removed for <= 3 km applications, as these variables all relate to the Grell-Freitas-Li deep cumulus parameterization.

[clark-evans]

Tagging @joeolson42 @AndersJensen-NOAA @barlage @dustinswales @mdtoyNOAA @XiaSun-Atmos @haiqinli @guoqing-noaa for awareness. I'd like to take some time at the 1/14 code discussion meeting to discuss and get back to Ming et al.

[guoqing-noaa]

@clark-evans Thanks a lot for creating this issue and helping with reducing mpasout file size.
FYI, here is Ruifang's google sheet about her latest progress:
https://docs.google.com/spreadsheets/d/1xqZPabbjebLZ8ghxgmlbCCIOgjKjEJCLEKmQ4F-av6g/edit?gid=0#gid=0

We plan to do 3~6 more retro experiments to remove variables incrementally and your second post greatly helps us to plan our experiments. Look forward to more discussions on this. Thanks!

[clark-evans]

@haiqinli believes that the variables between k22_shallow and sigma_deep, except for rthratensw, rthratenlw, and the sub3d_* variables, can be removed for storm-scale DA experiments.

[guoqing-noaa]

@clark-evans @haiqinli
We will also try some data assimilation on the global scale (global uniform or global variable resolution).
We may need to have two stream_list.atmosphere.da_state files for different applications, but we want to have them in the overall da_state stream.