I tested a 3DVAR DA with surface in-situ measurement over C401 regional domain. It returned status of zero, but all its output are nan value. The yaml input looks like
cost function:
cost type: 3D-Var
window begin: ‘2020-08-13T03:00:00Z’
window length: PT6H
analysis variables: [pm25_tot]
geometry:
nml_file: Data/fv3files/input_rrfs_cmaq.nml
nml_file_mpp: Data/fv3files/fmsmpp.nml
fv3_input_dir: Data/inputs/rrfs_cmaq/INPUT
trc_file: Data/fv3files/field_table
akbk: Data/fv3files/akbk64.nc4
npz: 64
fieldsets:
- fieldset: Data/fieldsets/dynamics_lam_cmaq.yaml
- fieldset: Data/fieldsets/rrfs_cmaq.yaml
- fieldset: Data/fieldsets/ufo.yaml
background:
filetype: gfs
datapath: Data/inputs/rrfs_cmaq/bkg/
filename_core: 20200813.060000.fv_core.res.tile1.nc
filename_trcr: 20200813.060000.fv_tracer.res.tile1.nc
filename_sfcd: 20200813.060000.sfc_data.nc
filename_sfcw: 20200813.060000.fv_srf_wnd.res.tile1.nc
filename_cplr: 20200813.060000.coupler.res
state variables: [u,v,T,DELP,ps,sphum,phis,pm25_tot]
background error:
covariance model: BUMP
date: ‘2020-08-13T06:00:00Z’
bump:
prefix: Data/bump/fv3jedi_bumpparameters_nicas_rrfs_cmaq
method: cor
strategy: specific_univariate
load_nicas: 1
mpicom: 2
verbosity: main
io_keys:
- “pm25_tot-pm25_tot”
io_values:
- “fixed_300km_0.1”
observations:
-
obs space:
name: PM25_TOT
obsdatain:
obsfile: Data/obs/testinput_tier_1/pm25_dummy_obs_2020081306.nc
obsdataout:
obsfile: Data/hofx/pm25_hofx_dummy_rrfs_cmaq_2020081306_m.nc
simulated variables: [pm25_tot]
obs filters:- filter: Bounds Check
filter variables:- name: [pm25_tot]
test variables: - name: LAMDomainCheck@ObsFunction
options:
map_projection: gnomonic_ed # only option currently supported
save: true # will save the DerivedValue to IODA output filethe size of the FV3-LAM grid (mass, not staggered)
npx: 200
npy: 110below comes from the global attributes in the FV3-LAM grid netCDF file
plat: 38.5
plon: -97.5
a: 0.21423
k: -0.23209
dx: 0.1124152007
dy: 0.1124152007
minvalue: 1.0 # will filter all obs outside of the specified limited area domain
- name: [pm25_tot]
obs operator:
name: VertInterp
obs error:
covariance model: diagonal
variational:
minimizer:
algorithm: PCG
iterations: - filter: Bounds Check
-
ninner: 10
gradient norm reduction: 1e-10
test: on
geometry:
nml_file: Data/fv3files/input_rrfs_cmaq.nml
nml_file_mpp: Data/fv3files/fmsmpp.nml
fv3_input_dir: Data/inputs/rrfs_cmaq/INPUT
trc_file: Data/fv3files/field_table
akbk: Data/fv3files/akbk64.nc4
npz: 64
fieldsets:- fieldset: Data/fieldsets/dynamics_lam_cmaq.yaml
- fieldset: Data/fieldsets/rrfs_cmaq.yaml
- fieldset: Data/fieldsets/ufo.yaml
diagnostics:
departures: ombg
final:
diagnostics:
departures: oman
output:
filetype: gfs
datapath: Data/analysis/
filename_core: 3dvar_rrfs_cmaq_pm25.fv_core.res.nc
filename_trcr: 3dvar_rrfs_cmaq_pm25.fv_tracer.res.nc
filename_sfcd: 3dvar_rrfs_cmaq_pm25.sfc_data.nc
filename_sfcw: 3dvar_rrfs_cmaq_pm25.fv_srf_wnd.res.nc
filename_cplr: 3dvar_rrfs_cmaq_pm25.coupler.res
first: PT0H
frequency: PT1H
Here are some print output
…
Increment print | number of fields = 1 | cube sphere face size: C200
pm25_tot : Min = +0.000000e+00, Max = +0.000000e+00, RMS = +0.000000e+00
CostJb : Nonlinear Jb State = 0
CostJb : Nonlinear Jb Model Aux = 0
CostJb : Nonlinear Jb Obs Aux = 0
CostJb : Nonlinear Jb = 0
Test : CostJb : Nonlinear Jb = 0
Jo Observations:
PM25_TOT nobs= 108 Min=0.5, Max=0.5, RMS=0.5
End Jo Observations
Jo Observations Equivalent:
PM25_TOT nobs= 108 Min=0.00210068, Max=1.20638, RMS=0.207984
End Jo Observations Equivalent
Jo Bias Corrected Departures:
PM25_TOT nobs= 108 Min=-0.497899, Max=0.706379, RMS=0.462786
End Jo Bias Corrected Departures
Test : CostJo : Nonlinear Jo(PM25_TOT) = inf, nobs = 108, Jo/n = inf, err = 0
Test : CostFunction: Nonlinear J = inf
CostJb: FG-BG
The costJo went to “inf” value. What could cause this issue? I ran the prior unit tests: geometry, state, hofx, bumpparamater etc, and each of them looked normal. However, the 3dVar gave weird result.
Thanks