WRF 4DVAR cost function finding maximum

Issues with running 3DVAR.

WRF 4DVAR cost function finding maximum

Postby Louise » Mon Dec 09, 2013 5:13 pm

Hi,

I'm running WRF 4DVAR for a domain with dx=6km. I'm trying to assimilate standard obs from ds461.0 and ds351.0 as well as ASCAT ocean winds, all of which are in little_r format. For the purpose of the assimilation, I'm calling the ASCAT winds QSCAT.

If I just use surface obs from ds461.0, then the assimilation converges properly, but if I run the assimilation including the upper obs from ds351.0 and the surface obs from ds461.0, or alternately using ONLY the ASCAT winds, then my cost function seems to converge to a maximum instead of a minimum. The data assimilation completes successfully, but I the solution is no good, since it is a maximum of the cost function and I have U increments of up to 30m/s at the surface.

The output in the cost_fn file is shown below, and my namelist is pasted after that. Does anyone have any advice on how to start debugging the problem, or experience with how to solve it?

Thanks in advance for the help,
Louise

Cost function:
Outer EPS Inner J Jb Jo Jc Je Jp Js jl
Iter Iter
1 0.500E-01 0 2160.562 0.000 2160.562 0.000 0.000 0.000 0.000 0.000
1 0.500E-01 1 2130.272 0.019 2130.253 0.031 0.000 0.000 0.000 0.000
1 0.500E-01 2 2060.118 0.331 2059.787 0.519 0.000 0.000 0.000 0.000
1 0.500E-01 3 1995.857 0.843 1995.015 1.371 0.000 0.000 0.000 0.000
1 0.500E-01 4 1912.742 2.023 1910.719 2.963 0.000 0.000 0.000 0.000
1 0.500E-01 5 1833.195 4.765 1828.430 6.246 0.000 0.000 0.000 0.000
1 0.500E-01 6 1933.488 10.606 1922.882 12.733 0.000 0.000 0.000 0.000
1 0.500E-01 7 2294.504 22.980 2271.524 25.232 0.000 0.000 0.000 0.000
1 0.500E-01 8 3427.213 40.933 3386.279 41.311 0.000 0.000 0.000 0.000
1 0.500E-01 9 8426.193 101.362 8324.831 96.975 0.000 0.000 0.000 0.000
1 0.500E-01 10 26174.750 330.688 25844.062 292.479 0.000 0.000 0.000 0.000
1 0.500E-01 11 51159.875 657.920 50501.955 535.991 0.000 0.000 0.000 0.000
1 0.500E-01 12 95966.373 1148.703 94817.670 880.820 0.000 0.000 0.000 0.000
1 0.500E-01 13 244268.931 2999.588 241269.343 1988.155 0.000 0.000 0.000 0.000
1 0.500E-01 14 407833.889 5027.121 402806.768 3130.884 0.000 0.000 0.000 0.000
1 0.500E-01 15 ********** 17989.414 ********** 9603.862 0.000 0.000 0.000 0.000
1 0.500E-01 16 ********** 55138.531 ********** 27617.976 0.000 0.000 0.000 0.000
1 0.500E-01 17 ********** 59512.056 ********** 29758.189 0.000 0.000 0.000 0.000
1 0.500E-01 18 ********** 64141.156 ********** 31891.487 0.000 0.000 0.000 0.000
1 0.500E-01 19 ********** 70298.618 ********** 34901.158 0.000 0.000 0.000 0.000
1 0.500E-01 20 ********** 84600.708 ********** 41492.124 0.000 0.000 0.000 0.000
1 0.500E-01 21 ********** 113419.260 ********** 54479.609 0.000 0.000 0.000 0.000
1 0.500E-01 22 ********** 123291.864 ********** 58345.373 0.000 0.000 0.000 0.000
1 0.500E-01 23 ********** 134659.746 ********** 63055.051 0.000 0.000 0.000 0.000
1 0.500E-01 24 ********** 154474.221 ********** 71340.625 0.000 0.000 0.000 0.000
1 0.500E-01 25 ********** 154474.221 ********** 71340.625 0.000 0.000 0.000 0.000
1 0.500E-01 25 ********** 154474.221 ********** 71340.625 0.000 0.000 0.000 0.000


Namelist:
&wrfvar1
var4d=true,
var4d_lbc=false,
var4d_bin=3600,
var4d_bin_rain=21600,
print_detail_outerloop=false,
print_detail_grad=false,
print_detail_obs=true
/
&wrfvar2
/
&wrfvar3
ob_format=2,
/
&wrfvar4
use_synopobs=true,
use_shipsobs=true,
use_metarobs=true,
use_soundobs=true,
use_pilotobs=true,
use_airepobs=true,
use_geoamvobs=true,
use_polaramvobs=true,
use_bogusobs=true,
use_buoyobs=true,
use_profilerobs=true,
use_satemobs=true,
use_gpspwobs=true,
use_gpsrefobs=true,
use_qscatobs=true,
use_rainobs=false,
/
&wrfvar5
check_max_iv=true,
/
&wrfvar6
max_ext_its=1,
eps = 0.05,
ntmax=50,
orthonorm_gradient=true,
/
&wrfvar7
cv_options=5,
/
&wrfvar8
/
&wrfvar9
/
&wrfvar10
test_transforms=false,
test_gradient=false,
/
&wrfvar11
cv_options_hum=1,
check_rh=0,
sfc_assi_options=1,
calculate_cg_cost_fn=true,
/
&wrfvar12
/
&wrfvar13
/
&wrfvar14
/
&wrfvar15
/
&wrfvar16
/
&wrfvar17
/
&wrfvar18
analysis_date="2010-10-20_06:00:00.0000",
/
&wrfvar19
/
&wrfvar20
/
&wrfvar21
time_window_min="2010-10-20_06:00:00.0000",
/
&wrfvar22
time_window_max="2010-10-20_12:00:00.0000",
/
&wrfvar23
/
&time_control
run_hours=6,
start_year=2010,
start_month=10,
start_day=20,
start_hour=06,
end_year=2010,
end_month=10,
end_day=20,
end_hour=12,
interval_seconds=21600,
debug_level=1000,
/
&domains
time_step=36,
e_we=164,
e_sn=166,
e_vert=36,
p_top_requested=5000,
interp_type=1,
t_extrap_type=1,
dx=6000,
dy=6000,
smooth_option=0,
nproc_x=0,
hypsometric_opt=2,
/
&fdda
/
&dfi_control
/
&tc
/
&physics
mp_physics=8,
ra_lw_physics=1,
ra_sw_physics=1,
radt=30,
sf_sfclay_physics=1,
sf_surface_physics=2,
bl_pbl_physics=1,
cu_physics=1,
cudt=5,
num_soil_layers=4,
num_land_cat = 20,
mp_zero_out=2,
co2tf=0,
/
&scm
/
&dynamics
w_damping=0,
diff_opt=1,
km_opt=4,
dampcoef=0.15,
time_step_sound=4,
base_temp=290.0,
use_baseparam_fr_nml=true,
/
&bdy_control
specified=true,
real_data_init_type=3,
/
&grib2
/
&namelist_quilt
/
&perturbation
trajectory_io=true,
enable_identity=false,
jcdfi_use=false,
jcdfi_diag=1,
jcdfi_penalty=1000.0,
/
Louise
 
Posts: 4
Joined: Wed Dec 05, 2012 3:51 pm

Re: WRF 4DVAR cost function finding maximum

Postby kam3545 » Sun May 10, 2015 9:31 pm

Dear All
I tested WRF version 3.4 to 3.7, but unfortunately in all these version WRF 4DVAR is not able to assimilate radiance (HIRS, now in 3.7 SEVIRI) correctly.
After one iteration gradient value is unrealistic (zero, NaN).
I used intel compiler for WRF installation
Prashant Kumar
Scientist
Atmospheric Sciences Division
ISRO
kam3545
 
Posts: 7
Joined: Sun Apr 12, 2009 2:16 am


Return to Runtime Problems

Who is online

Users browsing this forum: No registered users and 2 guests