6. Introduction to more GSI Applications¶
Introduction to Global GSI analysis¶
The Global Forecast System (GFS) is a global numerical weather prediction system containing a global computer model and variational analysis run by the U.S. National Weather Service (NWS). As of February 2015, the numerical model is run four times a day, and produces forecasts for up to 16 days in advance, with decreased spatial resolution after 10 days. It is a spectral model with a resolution of T1534 from 0 to 240 hours (0-10 days) and T574 from 240 to 384 hours (10-16 days). In the vertical, the model is divided into 64 layers and temporally, it produces forecast output every hour for the first 12 hours, every 3 hours out to 10 days, and every 12 hours after that. Its data assimilation system runs 6-hourly continuous cycles using the GSI-hybrid.
GSI has many functions specifically designed and tuned for GFS. Although the release version of the community GSI includes all the functions used by the operational systems, the DTC can only support the GSI regional applications because the DTC is not able to run GFS on community computers. Beginning with release version 3.2, the DTC began to introduce the use of GSI for global applications, assuming users can obtain the GFS background through the NCEP data hub or by running GFS themselves.
The Difference between Global and Regional GSI¶
As mentioned above, all of the NCEP operational systems use GSI as their analysis system. The majority of the GSI code is common to these operational systems. Very little source code is specific to a particular operational system. The main differences in the GSI operational application come from the configuration the run scripts and namelist parameters.
The different GSI applications need different backgrounds, observations, and fixed files. For the GFS system, GSI needs:
GFS Backgrounds: Typically, GSI uses 6-h GFS forecasts as the background. GFS 3-h and 9-h forecasts are also needed for the FGAT function in the GSI analysis. Both surface and atmosphere forecasts are needed.
Observations: NCEP has several sets of BUFR/prepBUFR observation files with global coverage for global systems. The files that start with the prefix GDAS are for the 6-hourly global data assimilation system. These files have more data available for the analysis, but have a longer delay for use in real-time. The files that start with gfs are for the GFS forecasts. Different operational systems need different observation data files because they require different kinds of observations with different coverage, cut-off times, and quality control processes. All these observation files are read in and processed in GSI by the same section of code. Therefore, there is no problem using GFS observation data files for regional GSI applications, as is described in the practice cases and the GSI Users Guide. Using regional BUFR files for global applications will cause problems since the data only cover part of the analysis domain, but GSI can still read in the observations and perform the analysis.
- Fixed files: Section 3.1 of the GSI Users Guide introduced the notion that different operational systems have their own fixed files. These global fixed files can be downloaded as a separate tar ball from the GSI users website (http://www.dtcenter.org/com-GSI/users/downloads/index.php). For the GFS GSI application, the big difference is the background error covariance (BE). Different resolutions of the GFS backgrounds use their matched BE files, which are different from the BE files used by the regional GSI applications. In release version 3.4, ten new BE files were provided for users (in addition to the two for release version 3.3):
global_berror.l64y386.f77global_berror.l64y96.f77global_berror.l64y1154.f77global_berror.l64y290.f77global_berror.l64y882.f77global_berror.l64y130.f77global_berror.l64y192.f77global_berror.l64y578.f77global_berror.l64y258.f77global_berror.l64y674.f77
Global GFS Scripts¶
Starting with release version 3.3, support for running GSI global
applications was added to the community release. Specifically a sample
global analysis run script was added to the ./run directory named
run_gsi_global.ksh of the community release. This script
run_gsi_global.ksh is based on the GSI GFS regression tests. This
script retains the same basic structure as the regional run script
run_gsi_regional.ksh, but includes additional details needed for the
global analysis.
| EULERIAN | SEMI-LAGRANGIAN |
| SPECTRAL RESOLUTION | LONB | LATB | LONB | LATB |
|---|---|---|---|---|
| T62 | 192 | 94 | 128 | 64 |
| T126 | 384 | 190 | 256 | 128 |
| T170 | 512 | 256 | 352 | 176 |
| T190 | 576 | 288 | 384 | 192 |
| T254 | 768 | 384 | 512 | 256 |
| T382 | 1152 | 576 | 768 | 384 |
| T574 | 1760 | 880 | 1152 | 576 |
| T878 | 2304 | 1152 | 1760 | 880 |
| T1148 | 2304 | 1152 | ||
| T1534 | 3072 | 1536 | ||
| T2014 | 4032 | 2016 | ||
| T2046 | 4096 | 2048 | ||
| T3070 | 6144 | 3072 |
[ch6_table1]
The first part of the global analysis run script, just as in the regional script, sets up the computer environment and case configuration. The primary differences between the global and regional scripts are the specification of the GFS case and the global application namelist.
GFSCASE=T126
GSI_NAMELIST=${GSI_ROOT}/run/comgsi_namelist_gfs.sh
While the regional script simply specifies the background and BE files, the global script needs to know the background resolution by defining the following parameters:
# Set the JCAP resolution which you want.
# All resolutions use LEVS=64
if [[ "$GFSCASE" = "T62" ]]; then
JCAP=62
JCAP_B=62
elif [[ "$GFSCASE" = "T126" ]]; then
JCAP=126
JCAP_B=126
elif [[ "$GFSCASE" = "enkf_glb_t254" ]]; then
JCAP=254
JCAP_B=254
elif [[ "$GFSCASE" = "T254" ]]; then
JCAP=254
JCAP_B=574
elif [[ "$GFSCASE" = "T574" ]]; then
JCAP=574
JCAP_B=1534
else
echo "INVALID case = $GFSCASE"
exit
fi
LEVS=64
#
Just as with the regional analysis run script, the global script double checks the needed parameters, creates a run directory, and copies the background, observations, and fixed files into the run directory. It generates the namelist, and places it in the run directory as well.
Specify the values of
LATA,LONA,DELTIME,resolbased on the choice ofJCAP:# Given the requested resolution, set dependent resolution parameters if [[ "$JCAP" = "382" ]]; then LONA=768 LATA=384 DELTIM=180 resol=1 elif [[ "$JCAP" = "574" ]]; then LONA=1152 LATA=576 DELTIM=1200 resol=2 elif [[ "$JCAP" = "254" ]]; then LONA=512 LATA=256 DELTIM=1200 resol=2 elif [[ "$JCAP" = "126" ]]; then LONA=256 LATA=128 DELTIM=1200 resol=2 elif [[ "$JCAP" = "62" ]]; then LONA=192 LATA=94 DELTIM=1200 resol=2 else echo "INVALID JCAP = $JCAP" exit fi NLAT=` expr $LATA + 2 `
Set up CO2, CH4, N2O, CO file decisions:
ICO2=${ICO2:-0} if [ $ICO2 -gt 0 ] ; then # Copy co2 files to $workdir co2dir=${FIX_ROOT} yyyy=`echo $ANAL_TIME | cut -c1-4` rm ./global_co2_data.txt co2=$co2dir/global_co2.gcmscl_$yyyy.txt while [ ! -s $co2 ] ; do ((yyyy-=1)) co2=$co2dir/global_co2.gcmscl_$yyyy.txt done if [ -s $co2 ] ; then cp $co2 ./global_co2_data.txt fi if [ ! -s ./global_co2_data.txt ] ; then echo "\./global_co2_data.txt" not created exit 1 fi fi #CH4 file decision ...Set up the namelist parameters and generate the namelist:
# Set some parameters for use by the GSI executable and to build the namelist echo " Build the namelist " vs_op='0.7,' hzscl_op='1.7,0.8,0.5,' ... ... # Build the GSI namelist on-the-fly . $GSI_NAMELIST cat << EOF > gsiparm.anl $comgsi_namelist EOF
Multiple time level backgrounds are needed:
if [[ "$GFSCASE" = "enkf_glb_t254" ]]; then cp $OBS_ROOT/gdas1.t12z.abias ./satbias_in cp $OBS_ROOT/gdas1.t12z.satang ./satbias_angle cp $BK_ROOT/sfcanl_2014040506_fhr03_ensmean ./sfcf03 cp $BK_ROOT/sfcanl_2014040506_fhr06_ensmean ./sfcf06 cp $BK_ROOT/sfcanl_2014040506_fhr06_ensmean ./sfcf09 cp $BK_ROOT/sfg_2014040506_fhr03_mem001 ./sigf03 cp $BK_ROOT/sfg_2014040506_fhr06_mem001 ./sigf06 cp $BK_ROOT/sfg_2014040506_fhr09_mem001 ./sigf09 else cp $OBS_ROOT/satbias_in ./satbias_in cp $OBS_ROOT/satbias_angle ./satbias_angle cp $BK_ROOT/sfcf03 ./sfcf03 cp $BK_ROOT/sfcf06 ./sfcf06 cp $BK_ROOT/sfcf09 ./sfcf09 cp $BK_ROOT/sigf03 ./sigf03 cp $BK_ROOT/sigf06 ./sigf06 cp $BK_ROOT/sigf09 ./sigf09 fi
Both surface and atmosphere files at 03, 06, and 09 hour forecasts are needed.- More observations files are availableIn the sample run script, many more observations are listed for use:
# Link to the other observation data if [[ "$GFSCASE" = "enkf_glb_t254" ]]; then obsfile_amua=gdas1.t12z.1bamua.tm00.bufr_d obsfile_amub=gdas1.t12z.1bamub.tm00.bufr_d else obsfile_amua=amsuabufr obsfile_amub=amsubbufr fi if [ -r "${OBS_ROOT}/satwnd" ]; then ln -s ${OBS_ROOT}/satwnd . fi if [ -r "${OBS_ROOT}/gpsrobufr" ]; then ln -s ${OBS_ROOT}/gpsrobufr . fi ...... if [ -r "${OBS_ROOT}/amsubbufrears" ]; then ln -s ${OBS_ROOT}/amsubbufrears . fi if [ -r "${OBS_ROOT}/tcvitl" ]; then ln -s ${OBS_ROOT}/tcvitl . fi
Table [ch6_table1] lists the grid dimensions for GFS of different resolutions from Running Global Model Parallel Experiments Version 6.0 from NCEP/EMC, which may provide users more information on the above mentioned resolution parameters:
Sample Results¶
After a successful run of the GSI GFS analysis, the contents of the run directory, with the clean option turned on, will look something like this:
amsrebufr diag_mhs_n18_anl.2011080100.gz fort.213
amsuabufr diag_mhs_n18_ges.2011080100.gz fort.214
amsubbufr diag_mhs_n19_anl.2011080100.gz fort.215
anavinfo diag_mhs_n19_ges.2011080100.gz fort.217
atms_beamwidth.txt diag_omi_aura_anl.2011080100.gz fort.218
berror_stats diag_omi_aura_ges.2011080100.gz fort.219
bftab_sstphr diag_pcp_tmi_trmm_anl.2011080100.gz fort.220
convinfo diag_pcp_tmi_trmm_ges.2011080100.gz fort.221
diag_amsre_hig_aqua_anl.2011080100.gz diag_sbuv2_n16_anl.2011080100.gz gomebufr
diag_amsre_hig_aqua_ges.2011080100.gz diag_sbuv2_n16_ges.2011080100.gz gpsrobufr
diag_amsre_low_aqua_anl.2011080100.gz diag_sbuv2_n17_anl.2011080100.gz gsi.exe
diag_amsre_low_aqua_ges.2011080100.gz diag_sbuv2_n17_ges.2011080100.gz gsiparm.anl
diag_amsre_mid_aqua_anl.2011080100.gz diag_sbuv2_n18_anl.2011080100.gz hirs3bufr
diag_amsre_mid_aqua_ges.2011080100.gz diag_sbuv2_n18_ges.2011080100.gz hirs4bufr
diag_amsua_metop-a_anl.2011080100.gz diag_sbuv2_n19_anl.2011080100.gz mhsbufr
diag_amsua_metop-a_ges.2011080100.gz diag_sbuv2_n19_ges.2011080100.gz omibufr
diag_amsua_n15_anl.2011080100.gz diag_seviri_m09_anl.2011080100.gz ozinfo
diag_amsua_n15_ges.2011080100.gz diag_seviri_m09_ges.2011080100.gz pcpbias_out
diag_amsua_n18_anl.2011080100.gz errtable pcpinfo
diag_amsua_n18_ges.2011080100.gz fit_p1.2011080100 prepbufr
diag_amsua_n19_anl.2011080100.gz fit_q1.2011080100 prepobs_prep.bufrtable
diag_amsua_n19_ges.2011080100.gz fit_rad1.2011080100 satbias_angle
diag_amsub_n17_anl.2011080100.gz fit_t1.2011080100 satbias_in
diag_amsub_n17_ges.2011080100.gz fit_w1.2011080100 satbias_out
diag_conv_anl.2011080100.gz fort.201 satinfo
diag_conv_ges.2011080100.gz fort.202 satwnd
diag_gome_metop-a_anl.2011080100.gz fort.203 sbuvbufr
diag_gome_metop-a_ges.2011080100.gz fort.204 scaninfo
diag_hirs3_n17_anl.2011080100.gz fort.205 seviribufr
diag_hirs3_n17_ges.2011080100.gz fort.206 sfcanl.gsi
diag_hirs4_metop-a_anl.2011080100.gz fort.207 siganl
diag_hirs4_metop-a_ges.2011080100.gz fort.208 stdout
diag_hirs4_n19_anl.2011080100.gz fort.209 stdout.anl.2011080100
diag_hirs4_n19_ges.2011080100.gz fort.210 tcvitl
diag_mhs_metop-a_anl.2011080100.gz fort.211 tmirrbufr
diag_mhs_metop-a_ges.2011080100.gz fort.212
The majority of these files existed after running the GSI regional
analysis examples in section [sec3.2.3] of the Basic
Users Guide, and they provide the same information about the GSI run. Of
note, the GSI global analysis run includes more radiance observations,
resulting in more radiance diag files in this list. Instead of the
single background file wrf_inout as seen with the regional analysis,
the global analysis background is split between the two files
siganl, for the atmosphere, and sfcanl.gsi for the surface. A
quick check of the standard output file stdout shows information
similar to that generated by the regional runs for the namelist, data
ingest, and minimization, but is quite different with respect to
information on the background IO.
Please visit our online tutorial for more details regarding how to conduct a global GSI run.
Introduction to Chemical Analysis¶
The GSI has also been developed to analyze chemical observations, such as MODIS AOD or PM2.5, to improve the pollution forecasts with chemical models. In this release, GSI can do the following chemical analyses:
| case | Chemical Model | background species | Observations |
|---|---|---|---|
| 1 | WRF-Chem | GOCART | MODIS AOD |
| 2 | WRF-Chem | GOCART | PM2.5 |
| 3 | WRF-Chem | PM2.5 | PM2.5 |
| 4 | CMAQ | CMAQ | PM2.5 |
[tab62]
The GSI run script for a chemical analysis (./run/run_gsi_chem.ksh )
and associated namelist (./run/comgsi_namelist_chem.sh) are provided
with this release. Sample background and observation files are provided
through the on-line tutorial.
Setup GSI Run Scripts for Chemical Analysis¶
The script run_gsi_chem.ksh was built based on regional GSI run
scripts and has a similar structure to the regional run script
run_gsi_regional.ksh, but include a couple of differences.
The first part of the run script sets up the computer environment and
case configuration. This is similar to the regional analysis run
scripts, except for the specification of (bk_core and obs_type)
for a given chemical case, and the namelist for the chemical
application:
GSI_NAMELIST=${GSI_ROOT}/run/comgsi_namelist_chem.sh
#------------------------------------------------
# bk_core= set background (WRFCHEM_GOCART WRFCHEM_PM25 or CMAQ)
# obs_type= set observation type (MODISAOD or PM25)
bk_core=CMAQ
obs_type=PM25
The choices of (bk_core and obs_type) for a chemical case need
to match with the options PREPBUFR and BK_FILE, which set
background and observation files. Table [tab63] shows how
to set up these two options for each case:
| case | background (BK_FILE ; bk_core ) | Observation (PREPBUFR; obs_type) |
|---|---|---|
| 1 | wrfinput_enkf_d01_2012-06-03_18:00:00; | Aqua_Terra_AOD_BUFR:2012-06-03_00:00:00; |
| WRFCHEM_GOCART | MODISAOD | |
| 2 | wrfinput_enkf_d01_2012-06-03_18:00:00; | anow.2012060318.bufr; |
| WRFCHEM_GOCART | PM25 | |
| 3 | wrfinput_enkf_d01_2012-06-03_18:00:00; | anow.2012060318.bufr; |
| WRFCHEM_PM25 | PM25 | |
| 4 | cmaq2gsi_4.7_20130621_120000.bin; | anow.2013062112.bufr; |
| CMAQ | PM25 |
[tab63]
Similar to the regional run script, this chemical run script will also double check the needed parameters. Then it creates a run directory, generates the namelist, and copies the background, observations, and fixed files into the run directory. Users who run the cases listed in table [tab62] do not need to change the rest of the run script. But users who need to build new cases may need to know the differences between chemical and regional applications, which is shown below.
Specify the name of the background and observations:
# Bring over background field (it's modified by GSI so we can't link to it) if [ ${bk_core} = WRFCHEM_GOCART ] ; then cp ${BK_FILE} ./wrf_inout fi if [ ${bk_core} = WRFCHEM_PM25 ] ; then cp ${BK_FILE} ./wrf_inout fi if [ ${bk_core} = CMAQ ] ; then cp ${BK_FILE} ./cmaq_in.bin fi # Link to the observation data if [ ${obs_type} = MODISAOD ] ; then ln -s ${PREPBUFR} ./modisbufr fi if [ ${obs_type} = PM25 ] ; then ln -s ${PREPBUFR} ./pm25bufr fiSpecify the background error file and anavinfo:
if [ ${bk_core} = WRFCHEM_GOCART ] ; then BERROR=${FIX_ROOT}/wrf_chem_berror_little_endian BERROR_CHEM=${FIX_ROOT}/wrf_chem_berror_little_endian ANAVINFO=${FIX_ROOT}/anavinfo_wrfchem_gocart fi if [ ${bk_core} = WRFCHEM_PM25 ] ; then BERROR=${FIX_ROOT}/wrf_chem_berror_little_endian BERROR_CHEM=${FIX_ROOT}/wrf_chem_berror_little_endian ANAVINFO=${FIX_ROOT}/anavinfo_wrfchem_pm25 fi if [ ${bk_core} = CMAQ ] ; then BERROR=${FIX_ROOT}/cmaq_berror_little_endian BERROR_CHEM=${FIX_ROOT}/cmaq_berror_little_endian ANAVINFO=${FIX_ROOT}/anavinfo_cmaq_pm25 fiSpecify the options for building the namelist:
if [ ${bk_core} = WRFCHEM_GOCART ] ; then bk_core_arw='.true.' bk_if_netcdf='.true.' bk_core_cmaq='.false.' bk_wrf_pm2_5='.false.' bk_laeroana_gocart='.true.' fi if [ ${bk_core} = WRFCHEM_PM25 ] ; then bk_core_arw='.true.' bk_if_netcdf='.true.' bk_core_cmaq='.false.' bk_wrf_pm2_5='.true.' bk_laeroana_gocart='.false.' fi if [ ${bk_core} = CMAQ ] ; then bk_core_arw='.false.' bk_if_netcdf='.false.' bk_core_cmaq='.true.' bk_wrf_pm2_5='.false.' bk_laeroana_gocart='.false.' fi
Sample Results¶
In this section, case one in Table [tab62] will be used as an example. After a successful run of the GSI Chem analysis, the contents of the run directory, with the clean option turned on, will look something like this:
aeroinfo fit_w1.2012060318 fort.212 fort.226 pcpinfo
anavinfo fort.201 fort.213 fort.227 prepobs_prep.bufrtable
berror_stats fort.202 fort.214 fort.228 satbias_angle
berror_stats_chem fort.203 fort.215 fort.229 satbias_in
convinfo fort.204 fort.217 fort.230 satbias_out
diag_conv_anl.2012060318 fort.205 fort.218 gsi.exe satinfo
diag_conv_ges.2012060318 fort.206 fort.219 gsiparm.anl stdout
errtable fort.207 fort.220 l2rwbufr stdout.anl.2012060318
fit_p1.2012060318 fort.208 fort.221 list_run_directory wrfanl.2012060318
fit_q1.2012060318 fort.209 fort.223 modisbufr wrf_inout
fit_rad1.2012060318 fort.210 fort.224 ozinfo
fit_t1.2012060318 fort.211 fort.225 pcpbias_out
Following instructions from Chapter 5, the following steps are conducted to check the results of this GSI chemical analysis:
- Check the standard output file:
Read in chemical background fields:
rmse_var=sulf ordering=XYZ WrfType,WRF_REAL= 104 104 ndim1= 3 staggering= N/A start_index= 1 1 1 0 end_index= 112 122 40 0 k,max,min,mid var=sulf 1 9.622933 3.3167184E-13 0.7885093 k,max,min,mid var=sulf 2 9.687045 3.3167214E-13 0.7910572 ... ... rmse_var=BC1 ... rmse_var=BC2 ... rmse_var=OC1 ... rmse_var=OC2 ...
Read in modis AOD observations:
OBS_PARA: modis_aod terra 7 64 50 11 34 25 7 81 40 31 50 27 38 46 56 25 7 0 26 53 62 60 19 0 OBS_PARA: modis_aod aqua 29 34 44 23 50 18 55 76 33 30 49 22 20 67 76 30 2 0 36 76 38 50 14 0
Minimizations:
Begin J table inner/outer loop 0 1 J term J aerosol aod 7.8897778549100276E+03 ----------------------------------------------------- J Global 7.8897778549100276E+03 End Jo table inner/outer loop 0 1 Initial cost function = 7.88977E+03 Initial gradient norm = 1.531557E+02 cost,grad,step,b,step? = 1 0 7.8897778E+03 1.5315573E+02 1.3701618E-01 0.0000000E+00 good pcgsoi: gnorm(1:2),b= 5.2207374E+03 5.220737604157E+03 2.225693512490E-01 cost,grad,step,b,step? = 1 1 4.67583332E+03 7.2254671E+01 1.4308398E-01 2.2256935E-01 good ... pcgsoi: gnorm(1:2),b= 1.254432996913E-06 1.254434051E-06 2.97203315E-01 cost,grad,step,b,step? = 2 16 3.15885215E+03 1.12001473E-03 2.3966297E-01 2.9720331E-01 good PCGSOI: WARNING **** Stopping inner iteration *** gnorm 0.534787149781733860E-10 less than 0.100000004E-09 update_guess: successfully complete \end{scriptsize} \end{tiny} \item Update chemical background fields: \begin{scriptsize} \begin{verbatim} ... k,max,min,mid var=sulf 38 9.1036431E-02 4.5203370E-07 1.8112745E-02 k,max,min,mid var=sulf 39 2.5694052E-02 6.2565640E-08 5.7557132E-03 k,max,min,mid var=sulf 40 1.1622031E-02 1.4633585E-09 8.4061045E-03 rmse_var=sulf ...
Analysis increments: After successfully running GSI, the analysis increments should be checked to see if data impacts are reasonable.
Analysis increments in the lowest level for SEAS_1 (left) and BC1 (right).