Since PCP-Combine performs a simple operation and reformatting step, no configuration file is needed.

1. Start by making an output directory for PCP-Combine and changing directories:

2. Now let's run PCP-Combine twice using some sample data that's included with the MET tarball:

The "\" symbols in the commands above are used for ease of reading. They are line continuation markers enabling us to spread a long command line across multiple lines. They should be followed immediately by "Enter". You may copy and paste the command line OR type in the entire line with or without the "\".

Both commands run the sum command which searches the contents of the -pcpdir directory for the data required to create the requested accmululation interval.

In the first command, PCP-Combine summed up 4 3-hourly accumulation forecast files into a single 12-hour accumulation forecast. In the second command, PCP-Combine summed up 12 1-hourly accumulation observation files into a single 12-hour accumulation observation. PCP-Combine performs these tasks very quickly.

We'll use these PCP-Combine output files as input for Grid-Stat. So make sure that these commands have run successfully!

We have run examples of the PCP-Combine -sum command, but the tool also supports the -add, -subtract, and -derive commands. While the -sum command defines a directory to be searched, for -add, -subtract, and -derive we tell PCP-Combine exactly which files to read and what data to process. The following command adds together 3-hourly precipitation from 4 forecast files, just like we did in the previous step with the -sum command:

By default, PCP-Combine looks for accumulated precipitation, and the 03 tells it to look for 3-hourly accumulations. However, that 03 string can be replaced with a configuration string describing the data to be processed, which doesn't have to be accumulated precipation. The configuration string should be enclosed in single quotes. Below, we add together the U and V components of 10-meter wind from the same input file. You would not typically want to do this, but this demonstrates the functionality. We also use the -name command line option to define a descriptive output NetCDF variable name:

While the -add command can be run on one or more input files, the -subtract command requires exactly two. Let's rerun the wind example from above but do a subtraction instead:

Now run Plot-Data-Plane to visualize this output. Use the -plot_range option to specify a the desired plotting range, the -title option to add a title, and the -color_table option to switch from the default color table to one that's good for positive and negative values:

Now view the results:

While the PCP-Combine -add and -subtract commands compute exactly one output field of data, the -derive command can compute multiple output fields in a single run. This command reads data from one or more input files and derives the output fields requested on the command line (sum, min, max, range, mean, stdev, vld_count).

Run the following command to derive several summary metrics for both the 10-meter U and V wind components:

In the above example, we used a wildcard to list multiple input file names.  And we used the -field command line option twice to specify two input fields.  For each input field, PCP-Combine loops over the input files, derives the requested metrics, and writes them to the output NetCDF file.  Run ncview to visualize this output:

This output file contains 8 variables: 2 input fields * 4 metrics. Note the output variable names the tool chose.  You can still override those names using the -name command line argument, but you would have to specify a comma-separated list of 8 names, one for each output variable.

Start by making an output directory for Gen-Vx-Mask and changing directories:

Since Gen-Vx-Mask performs a simple masking step, no configuration file is needed.

We'll run the Gen-Vx-Mask tool to apply a polyline for the CONUS (Contiguous United States) to our model domain. Run Gen-Vx-Mask on the command line using the following command:

Re-run using verbosity level 3 and look closely at the log messages. How many grid points were included in this mask?

Gen-Vx-Mask should run very quickly since the grid is coarse (185x129 points) and there are 244 lat/lon points in the CONUS polyline. The more you increase the grid resolution and number of polyline points, the longer it will take to run. View the NetCDF bitmap file generated by executing the following command:

Notice that the bitmap has a value of 1 inside the CONUS polyline and 0 everywhere else. We'll use the CONUS mask we just defined in the next step.

You could try running plot_data_plane to create a PostScript image of this masking region. Can you remember how?

Notice that there are several ways that gen_vx_mask can be run to define regions of interest, some of which will be demonstrated over the next few pages.

Start by making an output directory for Grid-Stat and changing directories:

The behavior of Grid-Stat is controlled by the contents of the configuration file passed to it on the command line. The default Grid-Stat configuration file may be found in the data/config/GridStatConfig_default file. Prior to modifying the configuration file, users are advised to make a copy of the default:

Open up the GridStatConfig_tutorial file for editing with your preferred text editor.

The configurable items for Grid-Stat are used to specify how the verification is to be performed. The configurable items include specifications for the following:

• The forecast fields to be verified at the specified vertical level or accumulation interval
• The threshold values to be applied
• The areas over which to aggregate statistics - as predefined grids, configurable lat/lon polylines, or gridded data fields
• The confidence interval methods to be used
• The smoothing methods to be applied (as opposed to interpolation methods)
• The types of verification methods to be used

You may find a complete description of the configurable items in the grid_stat configuration file section of the MET User's Guide. Please take some time to review them.

For this tutorial, we'll configure Grid-Stat to verify the 12-hour accumulated precipitation output of PCP-Combine. We'll be using Grid-Stat to verify a single field using NetCDF input for both the forecast and observation files. However, Grid-Stat may in general be used to verify an arbitrary number of fields. Edit the GridStatConfig_tutorial file as follows:

• Set:
  
  fcst = {
     field = [
        {
          name       = "APCP_12";
          level      = [ "(*,*)" ];
          cat_thresh = [ >0.0, >=5.0, >=10.0 ];
        }
     ];
  }
  obs = fcst;

  To verify the field of precipitation accumulated over 12 hours using the 3 thresholds specified.

• Set:
  
  mask = {
     grid = [];
     poly = [ "../gen_vx_mask/CONUS_mask.nc",
              "MET_BASE/poly/NWC.poly",
              "MET_BASE/poly/SWC.poly",
              "MET_BASE/poly/GRB.poly",
              "MET_BASE/poly/SWD.poly",
              "MET_BASE/poly/NMT.poly",
              "MET_BASE/poly/SMT.poly",
              "MET_BASE/poly/NPL.poly",
              "MET_BASE/poly/SPL.poly",
              "MET_BASE/poly/MDW.poly",
              "MET_BASE/poly/LMV.poly",
              "MET_BASE/poly/GMC.poly",
              "MET_BASE/poly/APL.poly",
              "MET_BASE/poly/NEC.poly",
              "MET_BASE/poly/SEC.poly" ];
  }

  To accumulate statistics over the Continental United States (CONUS) and the 14 NCEP verification regions in the United States defined by the polylines specified. To see a plot of these regions, execute the following command:

  gv ${MET_BUILD_BASE}/share/met/poly/ncep_vx_regions.pdf &

• In the boot dictionary, set:
  
  n_rep = 500;

  To turn on the computation of bootstrap confidence intervals using 500 replicates.

• In the nbrhd dictionary, set:
  
  width          = [ 3, 5 ];
  cov_thresh = [ >=0.5, >=0.75 ];

  To define two neighborhood sizes and two fractional coverage field thresholds.

• Set:
  
  output_flag = {
     fho     = NONE;
     ctc      = BOTH;
     cts      = BOTH;
     mctc   = NONE;
     mcts   = NONE;
     cnt     = BOTH;
     sl1l2   = BOTH;
     sal1l2 = NONE;
     vl1l2   = NONE;
     val1l2 = NONE;
     pct     = NONE;
     pstd   = NONE;
     pjc     = NONE;
     prc     = NONE;
     eclv    = NONE;
     nbrctc = BOTH;
     nbrcts = BOTH;
     nbrcnt = BOTH;
     grad    = BOTH;
     dmap  = NONE;
  }

  To compute contingency table counts (CTC), contingency table statistics (CTS), continuous statistics (CNT), scalar partial sums (SL1L2), neighborhood contingency table counts (NBRCTC), neighborhood contingency table statistics (NBRCTS), and neighborhood continuous statistics (NBRCNT).

Next, run Grid-Stat on the command line using the following command:

Grid-Stat is now performing the verification tasks we requested in the configuration file. It should take a minute or two to run. The status messages written to the screen indicate progress.

In this example, Grid-Stat performs several verification tasks in evaluating the 12-hour accumulated precipiation field:

• For continuous statistics and partial sums (CNT and SL1L2), 15 output lines each:
  (1 field * 15 masking regions)
• For contingency table counts and statistics (CTC and CTS), 45 output lines each:
  (1 field * 3 raw thresholds * 15 masking regions)
• For neighborhood methods (NBRCNT, NBRCTC, and NBRCTS), 90 output lines each:
  (1 field * 3 raw thresholds * 2 neighborhood sizes * 15 masking regions)

To greatly increase the runtime performance of Grid-Stat, you could disable the computation of bootstrap confidence intervals in the configuration file. Edit the GridStatConfig_tutorial file as follows:

• In the boot dictionary, set:
  
  n_rep = 0;

  To disable the computation of bootstrap confidence intervals.

Now, try rerunning the Grid-Stat command listed above and notice how much faster it runs. While bootstrap confidence intervals are nice to have, they take a long time to compute, especially for gridded data.

The output of Grid-Stat is one or more ASCII files containing statistics summarizing the verification performed and a NetCDF file containing difference fields. In this example, the output is written to the current directory, as we requested on the command line. It should now contain 10 Grid-Stat output files beginning with the grid_stat_ prefix, one each for the CTC, CTS, CNT, SL1L2, GRAD, NBRCTC, NBRCTS, and NBRCNT ASCII files, a STAT file, and a NetCDF matched pairs file.

The format of the CTC, CTS, CNT, and SL1L2 ASCII files will be covered for the Point-Stat tool. The neighborhood method and gradient output are unique to the Grid-Stat tool.

• Rather than comparing forecast/observation values at individual grid points, the neighborhood method compares areas of forecast values to areas of observation values. At each grid box, a fractional coverage value is computed for each field as the number of grid points within the neighborhood (centered on the current grid point) that exceed the specified raw threshold value. The forecast/observation fractional coverage values are then compared rather than the raw values themselves.
• Gradient statistics are computed on the forecast and observation gradients in the X and Y directions.

Since the lines of data in these ASCII files are so long, we strongly recommend configuring your text editor to NOT use dynamic word wrapping. The files will be much easier to read that way.

Execute the following command to view the NetCDF output of Grid-Stat:

Click through the 2d vars variable names in the ncview window to see plots of the forecast, observation, and difference fields for each masking region. If you see a warning message about the min/max values being zero, just click OK.

Now dump the NetCDF header:

View the NetCDF header to see how the variable names are defined.

Notice how *MANY* variables there are, separate output for each of the masking regions defined. Try editing the config file again by setting apply_mask = FALSE; and gradient = TRUE; in the nc_pairs_flag dictionary. Re-run Grid-Stat and inspect the output NetCDF file. What affect did these changes have?

We have now successfully run the PCP-Combine and Grid-Stat tools to verify 12-hourly accumulated preciptation for a single output time. We did the following steps:

• Identified our forecast and observation datasets.
• Constructed PCP-Combine commands to put them into a common accumulation interval.
• Configured and ran Grid-Stat to compute our desired verification statistics.

Now that we've defined the logic for a single run, the next step would be writing a script to automate these steps for many model initializations and forecast lead times. Rather than every MET user rewriting the same type of scripts, use METplus to automate these steps in a use case!

PB2NC Tool: Configure

The behavior of PB2NC is controlled by the contents of the configuration file passed to it on the command line. The default PB2NC configuration may be found in the data/config/PB2NCConfig_default file.

The configurable items for PB2NC are used to filter out the PrepBufr observations that should be retained or derived. You may find a complete description of the configurable items in the pb2nc configuration file section of the MET User's Guide or in the Configuration File Overview.

For this tutorial, edit the PB2NCConfig_tutorial_run1 file as follows:

• Set:
  
  message_type = [ "ADPUPA", "ADPSFC" ];

  to retain only those 2 message types. Message types are described in:
  http://www.emc.ncep.noaa.gov/mmb/data_processing/prepbufr.doc/table_1.htm

• Set:
  
  obs_window = {
     beg = -1800;
     end =  1800;
  }

  so that only observations within 1800 second (30 minutes) of the file time will be retained.

• Set:
  
  mask = {
     grid = "G212";
     poly = "";
  }

  to retain only those observations residing within NCEP Grid 212, on which the forecast data resides.

• Set:
  
  obs_bufr_var = [ "QOB", "TOB", "UOB", "VOB", "D_WIND", "D_RH" ];

  to retain observations for specific humidity, temperature, the u-component of wind, and the v-component of wind and to derive observation values for wind speed and relative humidity.

  While we are request these observation variable names from the input file, the following corresponding strings will be written to the output file: SPFH, TMP, UGRD, VGRD, WIND, RH. This mapping of input PrepBufr variable names to output variable names is specified by the obs_prepbufr_map config file entry. This enables the new features in the current version of MET to be backward compatible with earlier versions.

PB2NC Tool: Run

PB2NC is now filtering the observations from the PrepBufr file using the configuration settings we specified and writing the output to the NetCDF file name we chose. This should take a few minutes to run. As it runs, you should see several status messages printed to the screen to indicate progress. You may use the -v command line option to turn off (-v 0) or change the amount of log information printed to the screen.

If you'd like to filter down the observations further, you may want to narrow the time window or modify other filtering criteria. We will do that after inspecting the resultant NetCDF file.

PB2NC Tool: Output

When PB2NC is finished, you may view the output NetCDF file it wrote using the ncdump utility.

In the NetCDF header, you'll see that the file contains nine dimensions and nine variables. The obs_arr variable contains the actual observation values. The obs_qty variable contains the corresponding quality flags. The four header variables (hdr_typ, hdr_sid, hdr_vld, hdr_arr) contain information about the observing locations.

The obs_var, obs_unit, and obs_desc variables describe the observation variables contained in the output. The second entry of the obs_arr variable (i.e. var_id) lists the index into these array for each observation. For example, for observations of temperature, you'd see TMP in obs_var, KELVIN in obs_unit, and TEMPERATURE OBSERVATION in obs_desc. For observations of temperature in obs_arr, the second entry (var_id) would list the index of that temperature information.

Plot-Point-Obs

The plot_point_obs tool plots the location of these NetCDF point observations. Just like plot_data_plane is useful to visualize gridded data, run plot_point_obs to make sure you have point observations where you expect.

Each red dot in the plot represents the location of at least one observation value. The plot_point_obs tool has additional command line options for filtering which observations get plotted and the area to be plotted.

By default, the points are plotted on the full globe.

MET extracts the grid information from the first record of that GRIB file and plots the points on that domain.

The plot_data_plane tool can be run on the NetCDF output of any of the MET point observation pre-processing tools (pb2nc, ascii2nc, madis2nc, and lidar2nc).

PB2NC Tool: Reconfigure and Rerun

Now we'll rerun PB2NC, but this time we'll tighten the observation acceptance criteria.

• Set:
  
  message_type = [];

  to retain all message types.

• Set:
  
  obs_window = {
     beg = -25*30;
     end =  25*30;
  }

  so that only observations 25 minutes before and 25 minutes after the top of the hour are retained.

• Set:
  
  quality_mark_thresh = 1;

  to retain only the observations marked "Good" by the NCEP quality control system.

The majority of the observations were rejected because their valid time no longer fell inside the tighter obs_window setting.

When configuring PB2NC for your own projects, you should err on the side of keeping more data rather than less. As you'll see, the grid-to-point verification tools (Point-Stat and Ensemble-Stat) allow you to further refine which point observations are actually used in the verification. However, keeping a lot of point observations that you'll never actually use will make the data files larger and slightly slow down the verification. For example, if you're using a Global Data Assimilation (GDAS) PREPBUFR file to verify a model over Europe, it would make sense to only keep those point observations that fall within your model domain.

ASCII2NC Tool: Run

Since ASCII2NC performs a simple reformatting step, typically no configuration file is needed. However, when processing high-frequency (1 or 3-minute) SURFRAD data, a configuration file may be used to define a time window and summary metric for each station. For example, you might compute the average observation value +/- 15 minutes at the top of each hour for each station. In this example, we will not use a configuration file.

The sample ASCII observations in the MET tarball are still identified by GRIB code rather than the newer variable name option.

For example, 52 corresponds to RH:

ASCII2NC should perform this reformatting step very quickly since the sample file only contains data for 5 stations.

ASCII2NC Tool: Output

When ASCII2NC is finished, you may view the output NetCDF file it wrote using the ncdump utility.

The NetCDF header should look nearly identical to the output of the NetCDF output of PB2NC. You can see the list of stations for which we have data by inspecting the hdr_sid_table variable:

This ASCII data only contains observations at a few locations.

Next, we'll use the NetCDF output of PB2NC and ASCII2NC to perform Grid-to-Point verification using the Point-Stat tool.

Point-Stat Tool: Configure

The behavior of Point-Stat is controlled by the contents of the configuration file passed to it on the command line. The default Point-Stat configuration file may be found in the data/config/PointStatConfig_default file.

The configurable items for Point-Stat are used to specify how the verification is to be performed. The configurable items include specifications for the following:

• The forecast fields to be verified at the specified vertical levels.
• The type of point observations to be matched to the forecasts.
• The threshold values to be applied.
• The areas over which to aggregate statistics - as predefined grids, lat/lon polylines, or individual stations.
• The confidence interval methods to be used.
• The interpolation methods to be used.
• The types of verification methods to be used.

• Set:
  
  fcst = {
    message_type = [ "ADPUPA" ];
    field = [
      {
        name     = "TMP";
        level      = [ "P850-1050", "P500-850" ];
        cat_thresh = [ <=273, >273 ];
      }
    ];
  }
  obs = fcst;

  to verify temperature over two different pressure ranges against ADPUPA observations using the thresholds specified.

• Set:
  
  ci_alpha = [ 0.05, 0.10 ];

  to compute confidence intervals using both a 5% and a 10% level of certainty.

• Set:
  
  output_flag = {
     fho    = BOTH;
     ctc    = BOTH;
     cts    = STAT;
     mctc   = NONE;
     mcts   = NONE;
     cnt    = BOTH;
     sl1l2  = STAT;
     sal1l2 = NONE;
     vl1l2  = NONE;
     val1l2 = NONE;
     pct    = NONE;
     pstd   = NONE;
     pjc    = NONE;
     prc    = NONE;
     ecnt   = NONE;
     eclv   = BOTH;
     mpr    = BOTH;
  }

  to indicate that the forecast-hit-observation (FHO) counts, contingency table counts (CTC), contingency table statistics (CTS), continuous statistics (CNT), partial sums (SL1L2), economic cost/loss value (ECLV), and the matched pair data (MPR) line types should be output. Setting SL1L2 and CTS to STAT causes those lines to only be written to the output .stat file, while setting others to BOTH causes them to be written to both the .stat file and the optional LINE_TYPE.txt file.

• Set:
  
  output_prefix = "run1";

  to customize the output file names for this run.

Note that in the mask dictionary, the grid entry is set to FULL. This instructs Point-Stat to compute statistics over the entire input model domain. Setting grid to FULL has this special meaning.

Point-Stat Tool: Run

Next, run Point-Stat to compare a GRIB forecast to the NetCDF point observation output of the ASCII2NC tool.

Point-Stat is now performing the verification tasks we requested in the configuration file. It should take less than a minute to run. You should see several status messages printed to the screen to indicate progress.

If you receive a syntax error such as the one listed below, review PointStatConfig_tutorial_run1 for an extra comma after the "}" on line number 59

Notice the more detailed information about which observations were used for each verification task. If you run Point-Stat and get fewer matched pairs than you expected, try using the -v 3 option to see why the observations were rejected.

Point-Stat Tool: Output

The output of Point-Stat is one or more ASCII files containing statistics summarizing the verification performed. Since we wrote output to the current directory, it should now contain 6 ASCII files that begin with the point_stat_ prefix, one each for the FHO, CTC, CNT, ECLV, and MPR types, and a sixth for the STAT file. The STAT file contains all of the output statistics while the other ASCII files contain the exact same data organized by line type.

• This is a simple ASCII file consisting of several rows of data.
• Each row contains data for a single verification task.
• The FCST_LEAD, FCST_VALID_BEG, and FCST_VALID_END columns indicate the timing information of the forecast field.
• The OBS_LEAD, OBS_VALID_BEG, and OBS_VALID_END columns indicate the timing information of the observation field.
• The FCST_VAR, FCST_UNITS, FCST_LEV, OBS_VAR, OBS_UNITS, and OBS_LEV columns indicate the two parts of the forecast and observation fields set in the configure file.
• The OBTYPE column indicates the PrepBufr message type used for this verification task.
• The VX_MASK column indicates the masking region over which the statistics were accumulated.
• The INTERP_MTHD and INTERP_PNTS columns indicate the method used to interpolate the forecast data to the observation location.
• The FCST_THRESH and OBS_THRESH columns indicate the thresholds applied to FCST_VAR and OBS_VAR.
• The COV_THRESH column is not applicable here and will always have NA when using point_stat.
• The ALPHA column indicates the alpha used for confidence intervals.
• The LINE_TYPE column indicates that these are CTC contingency table count lines.
• The TOTAL column indicates the total number of matched pairs.
• The remaining columns contain the counts for the contingency table computed by applying the threshold to the forecast/observation matched pairs. The FY_OY (forecast: yes, observation: yes), FY_ON (forecast: yes, observation: no), FN_OY (forecast: no, observation: yes), and FN_ON (forecast: no, observation: no) columns indicate those counts.

1. What do you notice about the structure of the contingency table counts with respect to the two thresholds used? Does this make sense?
2. Does the model appear to resolve relatively cold surface temperatures?
3. Based on these observations, are temperatures >273 relatively rare or common in the P850-500 range? How can this affect the ability to get a good score using contingency table statistics? What about temperatures <=273 at the surface?

• The columns prior to LINE_TYPE contain the same data as the previous file we viewed.
• The LINE_TYPE column indicates that these are CNT continuous lines.
• The remaining columns contain continuous statistics derived from the raw forecast/observation pairs. See the CNT OUTPUT FORMAT section in the Point-Stat section of the MET User's Guide for a thorough description of the output.
• Again, confidence intervals are given for each of these statistics as described above.

1. What conclusions can you draw about the model's performance at each level using continuous statistics? Justify your answer. Did you use a single metric in your evaluation? Why or why not?
2. Comparing the first line with an alpha value of 0.05 to the second line with an alpha value of 0.10, how does the level of confidence change the upper and lower bounds of the confidence intervals (CIs)?
3. Similarly, comparing the first line with few numbers of matched pairs in the TOTAL column to the third line with more, how does the sample size affect how you interpret your results?

• The columns prior to LINE_TYPE contain the same data as the previous file we viewed.
• The LINE_TYPE column indicates that these are FHO forecast-hit-observation rate lines.
• The remaining columns are similar to the contingency table output and contain the total number of matched pairs, the forecast rate, the hit rate, and observation rate.
• The forecast, hit, and observation rates should back up your answer to the third question about the contingency table output.

• The columns prior to LINE_TYPE contain the same data as the previous file we viewed.
• The LINE_TYPE column indicates that these are MPR matched pair lines.
• The remaining columns are similar to the contingency table output and contain the total number of matched pairs, the matched pair index, the latitude, longitude, and elevation of the observation, the forecasted value, the observed value, and the climatological value (if applicable).
• There is a lot of data here and it is recommended that the MPR line_type is used only to verify the tool is working properly.

Point-Stat Tool: Reconfigure

Now we'll reconfigure and rerun Point-Stat.

This time, we'll use two dictionary entries to specify the forecast field in order to set different thresholds for each vertical level. Point-Stat may be configured to verify as many or as few model variables and vertical levels as you desire.

• Set:
  
  fcst = {
     field = [
      {
        name       = "TMP";
        level      = [ "Z2" ];
        cat_thresh = [ >273, >278, >283, >288 ];
      },
      {
        name       = "TMP";
        level      = [ "P750-850" ];
        cat_thresh = [ >278 ];
      }
    ];
  }
  obs = fcst;

  to verify 2-meter temperature and temperature fields between 750hPa and 850hPa, using the thresholds specified.

• Set:
  
   message_type = ["ADPUPA","ADPSFC"]
   sid_inc = [];
   sid_exc = [];
   obs_quality = [];
   duplicate_flag = NONE;
   obs_summary = NONE;
   obs_perc_value = 50;

  to include the Upper Air (UPA) and Surface (SFC) observations in the evaluation

• Set:
  
  mask = {
     grid  = [ "G212" ];
     poly  = [ "MET_BASE/poly/EAST.poly",
               "MET_BASE/poly/WEST.poly" ];
     sid   = [];
     llpnt = [];
  }

  to compute statistics over the NCEP Grid 212 region and over the Eastern and Western United States, as defined by the polylines specified.

• Set:
  
  interp = {
     vld_thresh = 1.0;
     shape       = SQUARE;
     type = [
        {
           method = NEAREST;
           width  = 1;
        },
        {
           method = DW_MEAN;
           width  = 5;
        }
     ];
  }

  to indicate that the forecast values should be interpolated to the observation locations using the nearest neighbor method and by computing a distance-weighted average of the forecast values over the 5 by 5 box surrounding the observation location.

• Set:
  
  output_flag = {
     fho    = BOTH;
     ctc    = BOTH;
     cts    = BOTH;
     mctc   = NONE;
     mcts   = NONE;
     cnt    = BOTH;
     sl1l2  = BOTH;
     sal1l2 = NONE;
     vl1l2  = NONE;
     val1l2 = NONE;
     pct    = NONE;
     pstd   = NONE;
     pjc    = NONE;
     prc    = NONE;
     ecnt   = NONE;
     eclv   = BOTH;
     mpr    = BOTH;
  }

  to switch the SL1L2 and CTS output to BOTH and generate the optional ASCII output files for them.

• Set:
  
  output_prefix = "run2";

  to customize the output file names for this run.

• 2 fields: TMP/Z2 and TMP/P750-850
• 2 observing message types: ADPUPA and ADPSFC
• 3 masking regions: G212, EAST.poly, and WEST.poly
• 2 interpolations: UW_MEAN width 1 (nearest-neighbor) and DW_MEAN width 5

Multiplying 2 * 2 * 3 * 2 = 24. So in this example, Point-Stat will accumulate matched forecast/observation pairs into 24 groups. However, some of these groups will result in 0 matched pairs being found. To each non-zero group, the specified threshold(s) will be applied to compute contingency tables.

Point-Stat Tool: Rerun

Next, run Point-Stat to compare a GRIB forecast to the NetCDF point observation output of the PB2NC tool, as opposed to the much smaller ASCII2NC output we used in the first run.
 

Point-Stat is now performing the verification tasks we requested in the configuration file. It should take a minute or two to run. You should see several status messages printed to the screen to indicate progress. Note the number of matched pairs found for each verification task, some of which are 0.

Plot-Data-Plane Tool

In this step, we have verified 2-meter temperature. The Plot-Data-Plane tool within MET provides a way to visualize the gridded data fields that MET can read.

Plot-Data-Plane requires an input gridded data file, an output postscript image file name, and a configuration string defining which 2-D field is to be plotted.

1. Set the title to 2-m Temperature.
2. Set the plotting range as 250 to 305.
3. Use the color table named ${MET_BUILD_BASE}/share/met/colortables/NCL_colortables/wgne15.ctable

Next, we'll take a look at the Point-Stat output we just generated.

See the usage statement for all MET tools using the --help command line option or with no options at all.

Point-Stat Tool: Output

The format for the CTC, CTS, and CNT line types are the same. However, the numbers will be different as we used a different set of observations for the verification.

• The columns prior to LINE_TYPE contain header information.
• The LINE_TYPE column indicates that these are CTS lines.
• The remaining columns contain statistics derived from the threshold contingency table counts. See the point_stat output section of the MET User's Guide for a thorough description of the output.
• Confidence intervals are given for each of these statistics, computed using either one or two methods. The columns ending in _NCL(normal confidence lower) and _NCU (normal confidence upper) give lower and upper confidence limits computed using assumptions of normality. The columns ending in _BCL (bootstrap confidence lower) and _BCU (bootstrap confidence upper) give lower and upper confidence limits computed using bootstrapping.

• The columns prior to LINE_TYPE contain header information.
• The LINE_TYPE column indicates these are SL1L2 partial sums lines.

Lastly, the point_stat_run2_360000L_20070331_120000V.stat file contains all of the same data we just viewed but in a single file. The Stat-Analysis tool, which we'll use later in this tutorial, searches for the .stat output files by default but can also read the .txt output files.

Stat-Analysis Tool: Configure

The behavior of Stat-Analysis is controlled by the contents of the configuration file or the job command passed to it on the command line. The default Stat-Analysis configuration may be found in the data/config/StatAnalysisConfig_default file.

You will see that most options are left blank, so the tool will use whatever it finds or whatever is specified in the command or job line.  If you go down to the jobs[] section you will see a list of the jobs run for the test scripts.

The first job listed above will select out only the contingency table count lines (CTC) where the threshold applied is >273.0 over the FULL masking region. This should result in 2 lines, one for pressure levels P850-500 and one for pressure P1050-850. So this job will be aggregating contingency table counts across vertical levels.

The second job listed above will perform the same aggregation as the first. However, it'll dump out the corresponding contingency table statistics derived from the aggregated counts.

Stat-Analysis Tool: Run on Point-Stat output

The output for these two jobs are printed to the screen.

The output was written to aggr_ctc_lines.out. We'll look at this file in the next section.

Note that we had to put double quotes (") around the forecast theshold string for this to work.

Stat-Analysis Tool: Output

On the previous page, we generated the output file aggr_ctc_lines.out by using the -out command line argument.

This file contains the output for the two jobs we ran through the configuration file. The output for each job consists of 3 lines as follows:

1. The JOB_LIST line contains the job filtering parameters applied for this job.
2. The COL_NAME line contains the column names for the data to follow in the next line.
3. The third line consists of the line type generated (CTC and CTS in this case) followed by the values computed for that line type.

This job should aggregate 2 CTC lines for 2-meter temperature across the EAST and WEST regions.

1. Do the same aggregation as above but for the 5x5 interpolation output (i.e. 25 points instead of 1 point).
2. Do the aggregation listed in (1) but compute the corresponding contingency table statistics (CTS) line. Hint: you will need to change the job type to aggregate_stat and specify the desired -out_line_type.
   How do the scores change when you increase the number of interpolation points? Did you expect this?
3. Aggregate the scalar partial sums lines (SL1L2) for 2-meter temperature across the EAST and WEST masking regions.
   How does aggregating the East and West domains affect the output?
4. Do the aggregation listed in (3) but compute the corresponding continuous statistics (CNT) line. Hint: use the aggregate_stat job type.
5. Run an aggregate_stat job directly on the matched pair data (MPR lines), and use the -out_line_type command line argument to select the type of output to be generated. You'll likely have to supply additional command line arguments depending on what computation you request.

    

1. How do the scores compare to the original (separated by level) scores? What information is gained by aggregating the statistics?

1. Job Number 1:
    
   stat_analysis \
   -lookin ../point_stat/point_stat_run2_360000L_20070331_120000V.stat -v 2 \
   -job aggregate -fcst_var TMP -fcst_lev Z2 -vx_mask EAST -vx_mask WEST -interp_pnts 25 -fcst_thresh ">278.0" \
   -line_type CTC \
   -dump_row job1_ps.stat
2. Job Number 2:
    
   stat_analysis \
   -lookin ../point_stat/point_stat_run2_360000L_20070331_120000V.stat -v 2 \
   -job aggregate_stat -fcst_var TMP -fcst_lev Z2 -vx_mask EAST -vx_mask WEST -interp_pnts 25 -fcst_thresh ">278.0" \
   -line_type CTC -out_line_type CTS \
   -dump_row job2_ps.stat
3. Job Number 3:
    
   stat_analysis \
   -lookin ../point_stat/point_stat_run2_360000L_20070331_120000V.stat -v 2 \
   -job aggregate -fcst_var TMP -fcst_lev Z2 -vx_mask EAST -vx_mask WEST -interp_pnts 25 \
   -line_type SL1L2 \
   -dump_row job3_ps.stat
4. Job Number 4:
    
   stat_analysis \
   -lookin ../point_stat/point_stat_run2_360000L_20070331_120000V.stat -v 2 \
   -job aggregate_stat -fcst_var TMP -fcst_lev Z2 -vx_mask EAST -vx_mask WEST -interp_pnts 25 \
   -line_type SL1L2 -out_line_type CNT \
   -dump_row job4_ps.stat
5. This MPR job recomputes contingency table statistics for 2-meter temperature over G212 using a new threshold of ">=285":
    
   stat_analysis \
   -lookin ../point_stat/point_stat_run2_360000L_20070331_120000V.stat -v 2 \
   -job aggregate_stat -fcst_var TMP -fcst_lev Z2 -vx_mask G212 -interp_pnts 25 \
   -line_type MPR -out_line_type CTS \
   -out_fcst_thresh ge285 -out_obs_thresh ge285 \
   -dump_row job5_ps.stat

Series-Analysis Tool: Configure

Start by making an output directory for Series-Analysis and changing directories:

The behavior of Series-Analysis is controlled by the contents of the configuration file passed to it on the command line. The default Series-Analysis configuration file may be found in the data/config/SeriesAnalysisConfig_default file. Prior to modifying the configuration file, users are advised to make a copy of the default:

The configurable items for Series-Analysis are used to specify how the verification is to be performed. The configurable items include specifications for the following:

• The forecast fields to be verified at the specified vertical level or accumulation interval.
• The threshold values to be applied.
• The area over which to limit the computation of statistics - as predefined grids or configurable lat/lon polylines.
• The confidence interval methods to be used.
• The smoothing methods to be applied.
• The types of statistics to be computed.

You may find a complete description of the configurable items in the series_analysis configuration file section of the MET User's Guide. Please take some time to review them.

For this tutorial, we'll run Series-Analysis to verify a time series of 3-hour accumulated precipitation. We'll use GRIB1 for the forecast files and NetCDF for the observation files. Since the forecast and observations are different file formats, we'll specify the name and level information for them slightly differently.

Open up the SeriesAnalysisConfig_tutorial file for editing with your preferred text editor and edit it as follows:

• Set the fcst dictionary to
  
  fcst = {
     field = [
        {
          name  = "APCP";
          level = [ "A3" ];
        }
     ];
  }

  To request the GRIB abbreviation for precipitation (APCP) accumulated over 3 hours (A3).

• Delete obs = fcst; and insert
  
  obs = {
     field = [
        {
          name  = "APCP_03";
          level = [ "(*,*)" ];
        }
     ];
  }

  To request the NetCDF variable named APCP_03 where its two dimensions are the gridded dimensions (*,*).

• Look up a few lines above the fcst dictionary and set
  
  cat_thresh = [ >0.0, >=5.0 ];

  To define the categorical thresholds of interest. By defining this at the top level of config file context, these thresholds will be applied to both the fcst and obs settings.

• In the mask dictionary, set
  
  grid = "G212";

  To limit the computation of statistics to only those grid points falling inside the NCEP Grid 212 domain.

• Set
  
  block_size = 10000;

  To process 10,000 grid points in each pass through the data. Setting block_size larger should make the tool run faster but use more memory.

• In the output_stats dictionary, set
  
     fho    = [ "F_RATE", "O_RATE" ];
     ctc    = [ "FY_OY", "FN_ON" ];
     cts    = [ "CSI", "GSS" ];
     mctc   = [];
     mcts   = [];
     cnt    = [ "TOTAL", "RMSE" ];
     sl1l2  = [];
     pct    = [];
     pstd   = [];
     pjc    = [];
     prc    = [];

  For each line type, you can select statistics to be computed at each grid point over the series. These are the column names from those line types. Here, we select the forecast rate (FHO: F_RATE), observation rate (FHO: O_RATE), number of forecast yes and observation yes (CTC: FY_OY), number of forecast no and observation no (CTC: FN_ON), critical success index (CTS: CSI), and the Gilbert Skill Score (CTS: GSS) for each threshold, along with the root mean squared error (CNT: RMSE).

Save and close this file.

Series-Analysis Tool: Run

First, we need to prepare our observations by putting 1-hourly StageII precipitation forecasts into 3-hourly buckets. We already ran PCP-Combine prior to running MTD to prepare this data.  Check that the 8 expected NetCDF files exist:

If they do, copy that directory over to the current working directory:

If they do not, go back and run the PCP-Combine commands listed here.

Note that the previous set of PCP-Combine commands could easily be run by looping through times in a shell script or through a METplus use case! The MET tools are often run using a scripting language rather than typing individual commands by hand. You'll learn more about automation using the METplus python scripts.

Next, we'll run Series-Analysis using the following command:

The statistics we requested in the configuration file will be computed separately for each grid location and accumulated over a time series of eight three-hour accumulations over a 24-hour period. Each grid point will have up to 8 matched pair values.

Note how long this command line is. Imagine how long it would be for a series of 100 files! Instead of listing all of the input files on the command line, you can list them in an ASCII file and pass that to Series-Analysis using the -fcst and -obs options.

Series-Analysis Tool: Output

The output of Series-Analysis is one NetCDF file containing the requested output statistics for each grid location on the same grid as the input files.

You may view the output NetCDF file that Series-Analysis wrote using the ncdump utility. Run the following command to view the header of the NetCDF output file:

In the NetCDF header, we see that the file contains many arrays of data. For each threshold (>0.0 and >=5.0), there are values for the requested statistics: F_RATE, O_RATE, FY_OY, FN_ON, CSI, and GSS. The file also contains the requested RMSE and TOTAL number of matched pairs for each grid location over the 24-hour period.

Next, run the ncview utility to display the contents of the NetCDF output file:

Click through the different variables to see how the performance varies over the domain. Looking at the series_cnt_RMSEvariable, are the errors larger in the south eastern or north western regions of the United States?

Why does the extent of missing data increase for CSI for the higher threshold? Compare series_cts_CSI_gt0.0 to series_cts_CSI_ge5.0. (Hint: Find the definition of Critical Success index (CSI) in the MET User's Guide and look closely at the denominator.)

Try running Plot-Data-Plane to visualize the observation rate variable for non-zero precipitation (i.e. series_fho_O_RATE_gt0.0). Since the valid range of values for this data is 0 to 1, use that to set the -plot_range option.

Setting block_size to 10000 still required 3 passes through our 185x129 grid (= 23865 grid points). What happens when you increase block_size to 24000 and re-run? Does it run slower or faster?

The behavior of Ensemble-Stat is controlled by the contents of the configuration file passed to it on the command line. The default Ensemble-Stat configuration file may be found in the data/config/EnsembleStatConfig_default file. The configurations used by the test script may be found in the scripts/config/EnsembleStatConfig* files.

The configurable items for Ensemble-Stat are broken out into two sections. The first section specifies how the ensemble should be processed to derive summary fields, such as the ensemble mean and spread. The second section specifies how the ensemble should be verified directly, such as the computation of rank histograms and spread/skill. The configurable items include specifications for the following:

• Section 1: Ensemble Processing (ens dictionary)
  • The ensemble fields to be summarized at the specified vertical level or accumulation interval.
  • The threshold values to be applied in computing ensemble relative frequencies (e.g. the percent of ensemble members exceeding some threshold at each point).
  • Thresholds to specify how many of the ensemble members must actually be present with valid data.
• Section 2: Verification (fcst and obs dictionaries)
  • The forecast and observation fields to be verified at the specified vertical level or accumulation interval.
  • The matching time window for point observations.
  • The type of point observations to be matched to the forecasts.
  • The areas over which to aggregate statistics - as predefined grids or configurable lat/lon polylines.
  • The interpolation or smoothing methods to be used.

For this tutorial, we'll configure Ensemble-Stat to summarize and verify 24-hour accumulated precipitation. While we'll run Ensemble-Stat on a single field, please note that it may be configured to operate on multiple fields. The ensemble we're verifying consists of 6 members defined over the west coast of the United States.

• In the ens dictionary, set
     field = [
       {
         name       = "APCP";
         level      = [ "A24" ];
         cat_thresh = [ >0, >=5.0, >=10.0 ];
       }
     ];

  To read 24-hour accumulated precipitation from the input GRIB files and compute ensemble relative frequencies for the thresolds listed.

• In the fcst dictionary, set
     field = [
       {
         name     = "APCP";
         level      = [ "A24" ];
       }
     ];

  To also verify the 24-hour accumulated precipitation fields.

• In the fcst dictionary, set:
     message_type = [ "ADPSFC" ];

  To verify against surface observations.

• In the point observation filtering section, set:
     prob_cat_thresh= [ >=0, >=5.0, >=10.0 ];

  To specify thresholds to use for computation of the Ranked Probability Score (RPS).

• In the mask dictionary, set
     poly = [ "MET_BASE/poly/NWC.poly",
              "MET_BASE/poly/SWC.poly" ];

  To also verify over the northwest coast (NWC) and southwest coast (SWC) subregions.

• Set:
  output_flag = {
     ecnt  = BOTH;
     rhist = BOTH;
     phist = BOTH;
     orank = BOTH;
     ssvar = BOTH;
     relp  = BOTH;
  }

  To compute continuous ensemble statistics (ECNT), ranked histogram (RHIST), probability integral transform histogram (PHIST), observation ranks (ORANK), spread-skill variance (SSVAR), and relative position (RELP).

Ensemble-Stat is now performing the tasks we requested in the configuration file. Note that we've passed the input ensemble data directly on the command line by specifying the number of ensemble members (6) followed by their names using wildcards. We've also specified one gridded StageIV analysis field (-grid_obs) and one file containing point rain gauge observations (-point_obs) to be used in computing rank histograms. This tool should run pretty quickly.

When Ensemble-Stat is finished, it will have created 9 output files in the current directory: 7 ASCII statistics files (.stat, _ecnt.txt, _rhist.txt, _phist.txt, _orank.txt, _ssvar.txt , and _relp.txt ) , a NetCDF ensemble file (_ens.nc), and a NetCDF matched pairs file (_orank.nc).

The _ens.nc output from Ensemble-Stat is a NetCDF file containing the derived ensemble fields, one or more ASCII files containing statistics summarizing the verification performed, and a NetCDF file containing the gridded matched pairs.

All of the line types are written to the file ending in .stat. The Ensemble-Stat tool currently writes six output line types, ECNT, RHIST, PHIST, RELP, SSVAR, and ORANK.

1. The ECNT line type contains contains continuous ensemble statistics such as spread and skill. Ensemble-Stat uses assumed observation errors to compute both perturbed and unperturbed versions of these statistics. Statistics to which observation error have been applied can be found in columns which include the _OERR (for observation error) suffix.
2. The RHIST line type contains counts for a ranked histogram. This ranks each observation value relative to ensemble member values. Ideally, observation values would fall equally across all available ranks, yielding a flat rank histogram. In practice, ensembles are often under-(U shape) or over-(inverted U shape) dispersive. In the event of ties, ranks are randomly assigned.
3. The PHIST line type contains counts for a probability integral transform histogram. This scales the observation ranks to a range of values between 0 and 1 and allows ensembles of different size to be compared. Similarly, when ensemble members drop out, RHIST lines cannot be aggregated together but PHIST lines can.
4. The RELP line is the relative position, which indicates how often each ensemble member's value was closest to the observation's value. In the event of ties, credit is divided equally among the tied members.
5. The ORANK line type is similar to the matched pair (MPR) output of Point-Stat. For each point observation value, one ORANK line is written out containing the observation value, its rank, and the corresponding ensemble values for that point. When verifying against a griddedanalysis, the ranks can be written to the NetCDF output file.
6. The SSVAR line contains binned spread/skill information. For each observation location, the ensemble variance is computed at that point. Those variance values are binned based on the ens_ssvar_bin_size configuration setting. The skill is determined by comparing the ensemble mean value to the observation value. One SSVAR line is written for each bin summarizing the all the observation/ensemble mean pairs that it contains.

The STAT file contains all the ASCII output while the _ecnt.txt, _rhist.txt, _phist.txt, _orank.txt, _ssvar.txt, and _relp.txt files contain the same data but sorted by line type. Since so much data can be written for the ORANK line type, we recommend disabling the output of the optional text file using the output_flag parameter in the configuration file.

• There are 6 lines in this output file resulting from using 3 verification regions in the VX_MASK column (FULL, NWC, and SWC) and two observations datasets in the OBTYPE column (ADPSFC point observations and gridded observations).
• Each line contains columns for the observation ranks (RANK_#) and a handful of ensemble statistics (CRPS, CRPSS, IGN, and SPREAD).
• There is output for 7 ranks - since we verified a 6-member ensemble, there are 7 possible ranks the observation values could attain.

• There are 5 lines in this output file resulting from using 3 verification regions (FULL, NWC, and SWC) and two observations datasets (ADPSFC point observations and gridded observations), where the ADPSFC point observations for the SWC region were all zeros for which the probability integral transform is not defined.
• Each line contains columns for the BIN_SIZE and counts for each bin. The bin size is set in the configuration file using the ens_phist_bin_size field. In this case, it was set to .05, therefore creating 20 bins (1/ens_phist_bin_size).

• This file contains 1866 lines, 1 line for each observation value falling inside each verification region (VX_MASK).
• Each line contains 44 columns, including header information, the observation location and value, its rank, and the 6 values for the ensemble members at that point.
• When there are ties, Ensemble-Stat randomly assigns a rank from all the possible choices. This can be seen in the SWC masking region where all of the observed values are 0 and the ensemble forecasts are 0 as well. Ensemble-Stat randomly assigns a rank between 1 and 7.

This file contains variables for the following:

1. Ensemble Mean
2. Ensemble Standard Deviation
3. Ensemble Mean minus 1 Standard Deviation
4. Ensemble Mean plus 1 Standard Deviation
5. Ensemble Minimum
6. Ensemble Maximum
7. Ensemble Range
8. Ensemble Valid Data Count
9. Ensemble Relative Frequency (for 3 thresholds)

The output of any of these summary fields may be disabled using the output_flag parameter in the configuration file.

This file is only created when you've verified using gridded observations and have requested its output using the output_flag parameter in the configuration file. Click through the variables in this file. Note that for each of the three verification areas (FULL, NWC, and SWC) this file contains 4 variables:

1. The gridded observation value
2. The observation rank
3. The probability integral transform
4. The ensemble valid data count

In ncview, the random assignment of tied ranks is evident in areas of zero precipitation.

Feel free to explore using this dataset. Some options to try are:

• Try setting skip_const = TRUE; in the config file to discard points where all ensemble members and the observation are tied (i.e. zero precip).  If you want to save it to a different file, make sure you set output_prefix to something meaningful, such as "run2", or "skip-constant".
• Try setting obs_thresh = [ >0.01 ]; in the config file to only consider points where the observation meets this threshold. How does this differ from the using skip_const?
• Use wgrib to inventory the input files and add additional entries to the ens.field list. Can you process 10-meter U and V wind?

MODE Tool: Configure

The behavior of MODE is controlled by the contents of the configuration file passed to it on the command line. The default MODE configuration file may be found in the data/config/MODEConfig_default file.

We'll be using these three configuration files during this session.

The configuration items for MODE are used to specify how the object-based verification approach is to be performed. In MODE, as in the other MET statistics tools, you can compare any two fields. When necessary, the items in the configuration file are specified separately for the forecast and observation fields. In most cases though, users will be comparing the same forecast and observation fields. The configurable items include parameters for the following:

• The forecast and observation fields and vertical levels or accumulation intervals to be compared
• Options to mask out a portion of or threshold the raw fields
• The forecast and observation object definition parameters
• Options to filter out objects that don't meet a size or intensity criteria
• Flags to control the logic for matching/merging
• Weights to be applied for the fuzzy engine matching/merging algorithm
• Interest functions to be used for the fuzzy engine matching/merging algorithm
• Total interest threshold for matching/merging
• Various plotting options

We'll start here using by running the configuration files we copied over, as-is.

MODE Tool: Run

It is very possible you will receive and error message after running the APCP_24 case that includes this:

This is not an HDF5 error.

These commands make use of sample data that's distributed with the MET tarball. They run MODE on 12-hour accumulated precipitation, 24-hour accumulated precipitation, and on a field of relative humidity.

MODE Tool: Output

The output of MODE typically consists of 4 files: 2 ASCII statistics files, 1 NetCDF object file, and 1 PostScript summary plot. The output of any of these files may be disabled using the appropriate MODE command line argument. In this example, the output is written to the current mode directory, as we requested on the command line.

The MODE output file naming convention is similar to that of the other MET tools. It contains timing information about the forecast being evaluated (forecast valid, lead, and accumulation times).

The 4 MODE output files are described briefly below:

• The PostScript file ends in .ps and is described below.
• The NetCDF object file ends in _obj.nc and contains the object indices.
• The ASCII contingency table statistics file and ends in _cts.txt.
• The ASCII object statistics file ends in _obj.txt and contains all of the object and object comparison data.

1. Page 1 summarizes the entire MODE run. Thumbnail images show the input data, resolved objects, and numbers identifying each object for both the forecast and observation fields. The color indicates object matching between the forecast and observation fields. Royal blue indicates an unmatched object. The object definition information is listed at the bottom of the page, and a sorted list of total object interest is listed on the right side.

2. Page 2 is an expanded view of the forecast thumbnail images.
3. Page 3 is an expanded view of the observation thumbnail images.
4. Page 4 has images showing the forecast objects with observation object outlines overlaid, and vice-versa.
5. Page 5 shows images and statistics for matching object clusters (i.e. one or more forecast objects matching one or more observation objects). These statistics also appear in the ASCII output from MODE.
6. When double-thresholding or fuzzy engine merging is enabled, additional PostScript pages are added to illustrate those methods.

What are the benefits of spatial methods over traditional statistics? The weaknesses? What are some examples where an object-based verification would be inappropriate?

To accumulate the output of the object based verification, you use the MODE-Analysis Tool. We will use this next.

MTD: Configure

The behavior of MTD is controlled by the contents of the configuration file passed to it on the command line. The default MTD configuration file may be found in the data/config/MTDConfig_default file.

The configuration items for MTD are used to specify how the space-time-object-based verification approach is to be performed. Just as MODE may be used to compare any two fields, the same is true of MTD. When necessary, the items in the configuration file are specified separately for the forecast and observation fields. In most cases though, users will be comparing the same forecast and observation fields. The configurable items include specifications for the following:

• The verification domain.
• The forecast and observation fields and vertical levels or accumulation intervals to be compared.
• The forecast and observation object definition parameters.
• Options to filter out objects that don't meet a minimum volume.
• Matching/merging weights and interest functions.
• Total interest threshold for matching/merging.
• Flags to control output files.

For this tutorial, we'll configure MTD to process the same series of data we ran through the Series-Analysis tool. Just like MODE, MTD compares a single forecast field to a single observation field in each run.

Set:

To retain all objects regardless of their volume.

MTD: Run

Rather than listing 8 input forecast and observation files on the command line, we will write them to a file list first. Since the 0-hour forecast does not contain 3-hourly accumulated precip, we will exclude that from the list. We will use the 3-hourly APCP output from PCP-Combine that we prepared above:

Just as with MODE, MTD applies a convolution operation to smooth the data. However, there are two important differences. In MODE, the convolution shape is a circle (radius = conv_radius). In MTD, the convolution shape is a square (width = 2*conv_radius+1) and for time t, the values in that square are averaged for times t-1, t, and t+1. Convolving in space plus time enables MTD to identify more continuous space-time objects.

MTD: Output

The MTD output typically consists of 6 files: 5 ASCII statistics files and 1 NetCDF object file. MTD does not create any graphical output. In this example, the output is written to the current mtd directory as we requested on the command line.

The MTD output file naming convention begins with mtd_ followed by the last valid time it encountered. The output file names may be modified using the output_prefix option in the configuration file, which should be used to prevent the output of one run from over-writing the output of a previous run. The 6 MTD output files are described briefly below:

• The NetCDF object file ends in .nc and contains gridded fields of the raw data, simple object indices, and cluster object indices for each forecast and observed timestep.
• The ASCII file ending with _2D.txt contains many columns similar to the output of MODE. This data summarizes the 2-dimensional object attributes for each individual time slice of the 3D forecast and observation objects.
• The ASCII files ending with _single_simple.txt and _single_cluster.txt contain 3D space-time attributes for simple and cluster objects, respectively.
• The ASCII files ending with _pair_simple.txt and _pair_cluster.txt contain 3D space-time attributes for pairs of simple and cluster objects, respectively.

Notice that the objects are defined in the active areas in the raw fields. Also notice some features merging (i.e. combining) as time passes while other features split (i.e. break apart). While they may be disconnected at a particular timestep, they remain part of the same space-time object.

Notice the generalization of the 2D MODE object attributes to 3 dimensions. Area measure becomes volume. MTD measures the object speed. Each object has a beginning and ending time.

TC-Pairs Tool: General

TC-Pairs Usage

View the usage statement for TC-Pairs by simply typing the following:

At a minimum, the input ATCF format -adeck (or -edeck) source, the input ATCF format -bdeck source, and the configuration -config file must be passed in on the command line.

The -adeck, -edeck, and -edeck options can be set to either a specific file name or a top-level directory which will be searched for files ending in ".dat".

TC-Pairs Tool: Configure

Start by making an output directory for TC-Pairs and changing directories:

Like the Point-Stat tool, the TC-Pairs tool is controlled by the contents of the configuration file passed to it on the command line. The default TC-Pairs configuration file may be found in the data/config/TCPairsConfig_default file.

The configurable items for TC-Pairs are used to specify which data are ingested and how the input data are filtered. The configurable items include specifications for the following:

• All model names to include in matched pairs
• Storms to include in verification: parsed by storm_id, basin, storm_name, cyclone
• Date/Time/Space restrictions: initialization time, valid time, lead time, masking
• User defined consensus forecasts
• Thresholds over watch/warning status of storm

The first step is to set-up the configuration file. Keep in mind, TC-Pairs should be run with the largest desired sample in mind to minimize re-running this tool! First, make a copy of the default:

Next, open up the TCPairsConfig_tutorial_run file for editing and modify it as follows:

Set:

To identify the models that we want to verify, as well as the basin. The three models identified are all U.S. operational models: HWRF, GFS (AVNO), and GFDL. Any field left blank will assume no restictions.

Set:

To compute a user defined consensus named "CONS". The membership of the consensus are the three listed models, none are required, but 2 must be present to calculate the consensus.

Set:

To only keep pairs that are the intersection of the model forecast and observation.

Next, save the TCPairsConfig_tutorial_run file and exit the text editor.

This run uses the default distance to land output file.  Run ncview to see it:

Water points have distance to land values greater than 0 while land points have distances <= 0.

TC-Pairs Tool: Run

Next, run TC-Pairs to compare all three ATCF forecast models specified in configuration file to the ATCF format best track analysis. Run the following command line:

TC-Pairs is now grabbing the model data we requested in the configuration file, generating the consensus, and performing the additional filtering criteria. It should take approximate 20 seconds to run. You should see several status messages printed to the screen to indicate progress.

There should be an output file "tc_pairs.tcst" in the directory, where .tcst stands for TC statistics. The next page will describe what is in the file.

If you are running many models over many storms/seasons, it is best to run TC-Pairs using a script to call TC-Pairs for each storm. This avoids potential memory issues when parsing very large datasets.

TC-Pairs Tool: Output

The output of TC-Pairs is an ASCII file containing matched pairs for each of the models requested. In this example, the output is written to the tc_pairs.tcst file as we requested on the command line. This output file is in TCST format, which is similar to the STAT output from the Point-Stat and Grid-Stat tools. For more header information on the TCST format, see the tc_pairs output section of the MET User's Guide.

Remember to configure your text editor to NOT use dynamic word wrapping. The files will be much easier to read that way:

• In the kwrite editor, select Settings->Configure Editor, de-select Dynamic Word Wrap and click OK.
• In the vi editor, type the command :set nowrap. To set this as the default behavior, run the following command:

Open tc_pairs.tcst

• Notice this is a simple ASCII file with rows for each matched pair for a single valid time and similar to the MPR line type written by other tools.
• Any field that has A in the column name (such as AMAX_WIND, ALAT) indicate the model forecast.
• Any field that has B in the column name (such as BMAX_WIND, BLAT) indicate the observation field (Best Track).
• The TK_ERR, X_ERR, Y_ERR, ALTK_ERR, CRTK_ERR columns are calculated track error, X component position error, Y component postion error, along track error, and cross track error, respectively.
• Columns 34-63 are the 34-, 50-, and 64-kt wind radii for each quadrant.

TC-Stat Tool: Configure

Start by making an output directory for TC-Stat and changing directories:

The behavior of TC-Stat is controlled by the contents of the configuration file or the job command passed to it on the command line. The default TC-Stat configuration may be found in the data/config/TCStatConfig_default file. Make a copy of the default configuration file and make following modifications:

Open up the TCStatConfig_tutorial file for editing with your preferred text editor.

Set:

To only parse over two of the three model names in the tc_pairs.tcst file. The event_equal=TRUE flag will keep pairs for specified forecast valid and lead times that are only in both HWRF and GFDL pairs.

Many of the filter options are left blank, indicating TC-Stat should parse over all available fields. You will notice many more available filter options beyond what was available with the TCPairsConfig.

Now, scroll all the way to the bottom of the TCStatConfig, and you will find the jobs[] section. Edit this section as follows, then save and close the editor:

TC-Stat: Run on TC-Pairs output

Run the TC-Stat using the following command:

Open the output file tc_stat.tcst. We can see that this filter job simply event equalized the two models specified in amodel.

Let's try to filter further, this time using the command line rather than the configuration file:

Here, we ran a filter job at the command line: only keeping tracks over water (not encountering land) and with categories Hurricane (HU), Tropical Storm (TS), Tropical Depression (TD), Subtropical Storm (SS), and Subtropical Depression (SD).

Open the output file tc_stat2.tcst: notice fewer lines have been kept. Look at the "LEVEL" column ... note all the non-tropical and subtropical level classifications have been filtered out of the sample.

Also, find the columns ADLAND and BDLAND. All these values are now positive, meaning the tracks over land (negative values) have been filtered.

With the filtering jobs mastered, lets give the second type of job - summary jobs - a try!

TC-Stat: Run on TC-Pairs output

Now, we will run a summary job using TC-Stat on the command line using the following command:

Open up the file tc_stat_summary.tcst. Notice this output is much different from the filter jobs.

The track data is event equalized for the HWRF and GFDL models, and summary statistics are produced for the TK_ERR column for each model by lead time.

TC-Stat: Plotting with R

In this section, you will use the R statistics software package to produce a plot of a few results. R was introduced in practical session 1.

The MET release includes a number of plotting tools for TC graphics. All of the Rscripts are included with the MET distribution in the Rscripts directory. The script for TC graphics is plot_tcmpr.R, which uses the TCST output files from TC-Pairs as input. At this time, there are two additional environment variables that need to be set to make this work.  They are MET_INSTALL_DIR and MET_BASE. To get the usage statement, type:

For Bash:

For C-shell:

The TC-Stat tool can be called from the Rscript to do additional filter jobs on the TCST output from TC-Pairs. This can be done on the command line by calling a filter job (following tc-stat), or a configuraton file can be used to select filtering criteria. A default configuration file can be found in Rscripts/include/plot_tcmpr_config_default.R.

The configuration file also includes various plot types to generate: MEAN, MEDIAN, SCATTER, REFPERF, BOXPLOT, and RANK. All of the plot commands can be called on the command line as well as with the configuration file. Plots are configurable (title, colors, axis labels, etc) either by modifying the configuration file or setting options on the command line. To run from the command line:

This plots the track error for two models: HWRF and CONS. CONS is the user defined consensus you generated in TC-Pairs.

Next, open up the output *png files:

The script produces the three plot types called in the configuration file:

1. Boxplot showing distribution of errors for a homogeneous sample of the two models (TK_ERR_boxplot.png).
2. Mean Errors with 95% CI for the same sample (TK_ERR_mean.png).
3. Rank plot indicating performance of HWRF model relative to CONS (TK_ERR_rank.png).

METviewer: Time Series of Categorical Statistics Plot

The following list shows the setting name and the setting value for each of the controls on the web application. Please go through them in order, starting at the top, setting and checking the controls as suggested. The settings that you need to change are marked in blue. The settings are grouped according to the areas on the web app. If you need more information about the controls, please click the little i in a circle to the right of METviewer X.Y at the very top of the page. This will take you to the METviewer documentation.

• Database: Click on the box named Select databases and from the Verification group, select mv_met_tutorial to make a plot with sample tutorial data.
• Tab: Series - Create a time-series (as we will do here) or threshold-series line plot. The different tabs correspond to the types of plots that METviewer can generate, such as box plots, bar plots, histograms, and ensemble plots.
• Plot Data: Stat - Plot traditional statistics as opposed to MODE object attributes.
• Y1 Axis variables - These controls specify the statistics and curves for the Y1 axis of the plot.
  • Y1 Dependent (Forecast) Variables - pairs of forecast variables and statistics to be plotted
    • Click the dropdown list and select: APCP_03 to select 3-hourly Accumulated Precipitation
    • Click Select attribute stat dropdown list and select: ACC and CSI to plot Accuracy and Critical Success Index
  • Y1 Series Variables - specifies the lines on the plot for each dependent variable
    • Click the dropdown list and select: MODEL
    • Click the Select value dropdown list and select: QPF to output the one model present
    • Next, click the + Series Variable button, click the dropdown list, and select: FCST_THRESH
    • Click the Select value dropdown list and select: >0.0, >2.54, and >6.35 to plot statistics for 3 categorical thresholds
• Y2 Axis Tab - these controls specify the statistics and curves for the Y2 axis of the plot; it is not used in this demonstration
• Fixed Values - field/value pairs that are held constant for all data on the plot
  • Click the + Fixed Value button - to add a fixed value selection option
  • Click the dropdown list and select VX_MASK
  • Click the Select value dropdown list and select: FULL to plot data for FULL model domain (the only one present in this dataset)
• Independent Variable - the x-axis field and values over which the data are displayed on the plot
  • Click the dropdown list and select FCST_LEAD
  • Click the Select value dropdown list and click the Check all option to select all items
• Statistics
  • The Summary radio button tells MET to plot the mean or median of the statistics pulled directly from the database.
  • The Aggregation statistics radio button tells METviewer to aggregate contingency table counts or partial sums and compute aggregate summary statistics and confidence intervals.
  • Use the default Summary radio button selection
• Series Formatting - settings to control colors, line types, plotting symbols, and confidence intervals
  • There should be settings for 6 lines: 3 lines for ACC followed by 3 lines for CSI
  • In the Line Color column:
    • Set the 1st and 4th lines to the same RED color
    • Set the 2nd and 5th lines to the same GREEN color
    • Set the 3rd and 6th lines to the same BLUE color
  • In the Line Type column:
    • Set the 4th, 5th, and 6th lines to dashed
• Plot Formatting - settings that control the appearance of the plot, along with several miscellaneous features
  • On the Titles & Labels tab
    • set Plot Title to ACC / CSI vs. Lead Time
    • set X label to Forecast Lead
    • set Y1 label to ACC / CSI
  • On the Common tab, select Display Number of Stats
• Click the Generate Plot button at the top of the page to submit your plot request to the METviewer plotting engine.