DOWNLOAD AND BUILD THE COMMUNITY ENKF SYSTEM

Download GSI/EnKF system

This exercise is intended to give you practice building the GSI/EnKF system. This version of GSI/EnKF system includes a WRF I/O library. There are no need to build the WRF model for this version GSI building.

The community GSI resources, including source code, build system, utilities, and fix files, are available for download from the DTC community GSI users website: downloads

The code downloads will result in the following tar file:

comGSIv3.7_EnKFv1.3.tar.gz

GSI requires use of CRTM coefficients to analyze satellite radiance observations. Due to their large size, these are available as a separate tarfile available for download by selecting the link: CRTM 2.3.0 Big_Endian coefficients tarball

The tars file may be unpacked using the UNIX commands:

tar -zxvf comGSIv3.7_EnKFv1.3.tar.gz

 

This creates the top level GSI directory: comGSIv3.7_EnKFv1.3

Setting up the machine environment

Set the necessary paths for using your selected compiler, such as loading the appropriate modules or modifying your path variable.

Before configuring the GSI/EnKF code to be built, at least one, and no more than two environment variables must be set.

• NETCDF: the path to the NETCDF libraries
• LAPACK_PATH: the path to the LAPACK math libraries

For more detail, see EnKFv1.3 Users' Guide Chapter 2.

Building GSI v3.7 as observer and EnKF v1.3

This version GSI can ONLY be compiled with Cmake ( the same process as building GSI only). 

First create a building directory (any name works) outside the GSI directory:

mkdir build

Then, get into this building directory

cd build

Now run cmake to configure makefile by typing:

cmake ${Path To GSI Home Directory}/comGSIv3.7_EnKFv1.3

 

Cmake will check your computer environment and generate the makefile. A successful cmake run should end with:

-- Configuring done 
-- Generating done 
-- Build files have been written to: (Path_of_this_directory)/build

The building directory includes many new file and directories generated from Cmake:

bin cmake_install.cmake include Makefile Testing 
CMakeCache.txt CTestTestfile.cmake lib regression_var.out util 
CMakeFiles DartConfiguration.tcl libsrc src 

Now, we can simply run make to compile the GSI/EnKF and its untilities:

make

If the compilation is successful, the executable gsi.x, enkf.x, will be created in the ./bin directory. If the compilation is not successful, run make with make VERBOSE=1 to collect more information on errors.

After successfully compiling GSI and EnKF, the user can now run practice cases.

PREPARE ENKF RUN SCRIPT FOR TUTORIAL CASES

Setting up the Run Script

The EnKF run script comenkf_run_regional.ksh, which is provided with the source code, must be customized to the local environment. These changes include things such as the analysis time and date, the path to files, and MPI specific information, which are necessary for running all of the basic tutorial cases.

For this tutorial, start by copying the original run script to a working copy
cd ${PATH of working directory}/run/ 
cp ${PATH}/comGSIv3.7_EnKFv1.3/ush/comenkf_run_regional.ksh run_enkf_wrf.ksh_basic 
that will be used as the template for the scripts used in the tutorial cases.

Make the following modifications to the script run_enkf_wrf.ksh_basic

• set the job header if running in MPI mode (consult local IT support for more information about the local environment).
• set the value of variable GSIPROC
• set the variable ARCH to the value appropriate for your system.
• In the case set up section of the script, set the environment variables to values appropriate for the tutorial case. Below lists the lines that might need modifications.
  • ANAL_TIME=2014021300
  • JOB_DIR=$(PATH)/run
  • RUN_NAME=enkf_baisc
  • OBS_ROOT=$(PATH)/enkf_arw_2014021300/obs
  • BK_ROOT=$(PATH)/enkf_arw_2014021300/bkg
  • GSI_ROOT=$(where you put the)/comGSIv3.7_EnKFv1.3
  • CRTM_ROOT=${PATH}/CRTM_2.2.3
  • diag_ROOT=${PATH}/run/observer_basic
  • ENKF_EXE=${JOB_DIR}/enkf_wrf.x
  • WORK_ROOT=${JOB_DIR}/${RUN_NAME}
  • FIX_ROOT=${GSI_ROOT}/fix
  • EnKF_NAMELIST=${GSI_ROOT}/ush/comenkf_namelist.sh
• In the ensemble parameters setion, check for proper case/ensemble settings:
  • NMEM_ENKF=20
  • BK_FILE_mem=${BK_ROOT}/wrfarw
  • NLONS=129
  • NLATS=70
  • NLEVS=50
  • IF_ARW=.true.
  • IF_NMM=.false.
• Directly following, check the list to select observations (from the diags files):
• list="conv" 
# list="conv amsua_n18 hirs4_n19" 

 

•  
• On NCAR Cheyenne, the basic scripts are submitted to the queue by typing:
• qsub run_enkf_wrf.ksh_basic 

 

•  

An example of this basic run script is available from the link run_enkf_wrf.ksh_basic

ENKF ONLINE EXERCISE DATA

Introduction

The EnKF online exercise data sets are available for download from this website, though the data size has been reduced by using only 20 members for an ARW case and 10 members for the global case. There are three tarballs: one for the ARW case (427M) and two for the background ensemble and observations for the global case (3.0G, 453M)

Content of online exercise data

Inside the /data directory (Yellowstone) are separate directories for /arw_2014021300 and /enkf_glb_t254. Inside each of these directories are subdirectories for ensemble member and mean (/bk) and observations (/obs).

The tar files available for download are arw_2014021300.tar.gz, enkf_glb_bk.tar.gz and enkf_glb_obs.tar.gz. Untaring all four results in the following directories:

Inside the arw directory is subdirectories for ensemble members and mean (/bk) and observations (/obs).

In the ARW ensemble background directory are files follow the naming convention wrfarw.memXXX where in this instance XXX ranges from 001 to 020. In addition, the ensemble mean of the backgrounds is namedwrfarw.ensmean. The observation directory (/obs) contains a collection of gdas bufr files for various observations types.

In the global model background directory (./enkf_glb_bk) are files following the naming convention sfg_yyyymmddhh_fhrhh_memXXX where in this instance XXX ranges from 001 to 010. In addition, the ensemble mean of the backgrounds is named sfg_yyyymmddhh_fhrhh_ensmean. The observation directory (./enkf_glb_obs) contains a collection of gdas bufr files for various observations types.

After successfully locating or downloading the case data, users can now run the practice cases cases.