MET Online Tutorial for METv8.0 | GSI Tools > Output

Open up and examine the following output files we just generated:

• Conventional MPR data:

• Radiance MPR data:

• Conventional ORANK data:

• Radiance ORANK data:

Look at the header lines for the GSI2MPR output files and notice that the conventional and radiance headers are the same through the OBS_QC column. However, the remaining columns differ and are specific to the data type. The GSIDENS2ORANK output files include the standard ORANK header columns followed by extra columns specific to the data type. Next, we'll run STAT-Analysis aggregate_stat jobs to derive statistics from this data.

GSID2MPR Statistics

Run the following STAT-Analysis jobs to compute statistics using the GSID2MPR output:

• Read MPR lines and compute continuous statistics for each variable present. Write the output to the specified .stat file:

• Open up $MET_TUTORIAL_DATA/output/gsid2mpr/conv.mem001_CNT.stat and notice that multiple observation type values are written as a comma-separated list in the OBTYPE. Rerun this command by use the -set_hdr option to define the output for that column:

• The OBTYPE column in the output should now be set to ALL. With the right set of options, STAT-Analysis may be used to filter this data in any way you would like and derive many different types of statistics. For example, process only temperature data (-fcst_var t), using only the pairs that were actually assimilated (-column_thresh ANLY_USE eq1), and threshold them to define a contingency table (-out_line_type CTC -out_thresh ge273):

GSIDENS2ORANK Statistics

As we saw above, STAT-Analysis can read MPR lines and derive a variety of output line types, such as SL1L2, CNT, CTC, CTS, MCTC, MCTS, and so on). Similarly, STAT-Analysis can read the ORANK lines generated by GSIDENS2ORANK and derive ensemble output line types, such as RHIST, PHIST, and SSVAR. Run the following STAT-Analysis jobs to compute statistics using the GSIDENS2ORNK output:

• Read radiance ORANK lines and compute ranked histograms for each variable present and write the output to the screen:

• Now process conventional ORANK lines and derive PHIST output with a 0.10 bin size for each unique combination of FCST_VAR and N_USE column: