WRF 3D-Var (wrfvar)-based verification package

Yong-Run Guo

30 December 2005

 

1, What means the WRF 3D-Var (wrfvar)-based verification?

á      The observation operators: the spatial interpolation and the physical transformation for each type of observations, are already developed within the wrfvar code;

á      The statistics of the innovations (O-B) for each type of observations have been computed and outputted as a diagnostic file from wrfvar code;

á      There is quality control procedure for each type of the observations implemented in wrfvar code, which includes the Background check and the 3-dimenssional model domain check, so a set of the quality-controlled or filtered observations can be easily obtained by using a standard first guess fields, such as NCEP GFS.   

Given a set of the quality-controlled observations, it is easy to obtain the RMS and Bias errors by using the first guess fields from the forecasts at the different time, 6-h, 12-h,É, 72-h from the different experiments.

2, What are the more works for verifications against the observations?

á      Modify the wrfvar code for the special option: ANALYSIS_TYPE = QC-OBS, to write out the filtered observation files. With this quality-controlled observation file, the WRF 3D-Var (wrfvar) gave the identical results by setting check_Mac_iv = .TRUE. and .FALSE.

á      Modify the wrfvar code for the special option: ANALYSIS_TYPE = VERIFY, a) no further minimization need to be completed; b) allow the extrapolation to be done when the observations located below the lowest model level and above the highest model level so that the exactly same observation data will be verified against for all the first guess fields from the different experiments.

á      Develop the utility program to split the WRF model outputted historic files, such as wrfout_d01, etc., to be the Single-time forecast files because now wrfvar code can ONLY read one first guess file once (no search reading function);

á      Develop an utility program to process the wrfvar output statistics files, a) read the statistics files for the first guess fields from the different forecasts times and different initial times; b) plot the RMS and Bias errors as a function of forecast length, and as a function of the initial times for the different type of the observations, such as SOUND, and SYNOP, METAR, SHIPS, and BUOY, etc.

á      Develop a flexible shell script being able to complete all the steps for verifications sequentially or just to do the selected steps. 

3, What are the advantages and disadvantages of this VERIFY package?

The advantages:

á      A variety of the observations consistently quality-controlled (filtered) by WRF-Var (wrfvar) code are used in this package;

á      The verifications can be done not only against the conventional data but also the non-conventional data;

á      The exactly same observations are verified against for all the experiments;

á      It is easy to be used for the comparison of the results from the different experiments. The errors can

depend on the forecast lengths for all initial times, or

depend on the initial times at the different forecast times

with a flexible interval.

Disadvantages:

á      Because the different SOUND data may have the different vertical distribution of the data points, the statistics were only calculated over the entire 3-dimenssional space, not over the 2-dimenssional space on the specific levels. No vertical profiles of the error statistics could be provided and plotted;

á      Because the quality control procedure and verifications are depended on the observation operators implemented in wrfvar code, the different observation operators used, such as SFC_ASSI_OPTIONS =1 or 2, may give the different results of the verifications.

4, The procedure of VERIFY

This package completes the verifications against the quality-controlled observations used in WRF 3D-Var (wrfvar). There are several steps needed to be completed.

1) To obtained the quality-controlled observations

Preparation:

á      Original observation files generated by 3DVAR_OBSPROC;

á      Standard first guess fields generated by WRFSI, or other analysis;

Run var_qc_obs.csh

2) To split the WRF model outputted historic files: wrfout_d01, etc.

Preparation:

á      Acquire the WRF model forecast files to the local disk

Run the utility program: time_split_nc.exe

3) To generate the innovation (O-B) statistics files:

 DAProg_WRF-Var.statistics_inittime_fcsttime

Run var_verify.csh

4) To plot the figures of the verifications

Preparation:

á      Edit the namelist: namelist.var_stats

Run the utility program: plot_var_stats.exe

 

All above four steps can be completed by shell script WRFVAR_verify.csh. Users only need to edit this shell script to provide the proper directories. Usually Step 1 and 2 just need to do once and Step 3 and 4 can be repeatedly executed with the different settings of the options.