Filling in DESC column

[CGenie]

Hello,

I'm trying to make a pointstat analysis using metplus. I have ascii files with station names, I use ascii2nc to convert them to .nc files. Then I run point stat and get _sl1l2.txt files. Currently, the DESC column has NA in them. Is there a way to pass the station description into the DESC column? I see the POINT_STAT_DESC variable in config files but I'm not sure how to fill that to use the observation description. If I specify that as POINT_STAT_DESC = {obs_sid} I get literal {obs_sid}.

[JohnHalleyGotway]

Hello @CGenie, I see that you have a question about setting the DESC column in the SL1L2 lines generated by the Point-Stat tool. In general, Point-Stat aggregates together multiple matched pairs when computing partial sums (e.g. SL1L2 line type) or statistics (e.g. CNT line type). That is to say that usually the TOTAL column in those lines has a value > 1.

When aggregating multiple lines together, the logic for picking a single station name to populate the DESC column doesn't make sense. Generally, you'd have many station names, one for each station being aggregated. However, based on your question, I'm guessing that you've configured Point-Stat in such a way that you're writing SL1L2 lines consisting of a single matched pair.

The short answer to your question is no, there's no mechanism directly within Point-Stat to detect the case when only a single matched pair is used to create an SL1L2 line and then use it's station name to populate the DESC column. However, there's a few other details/options that may be useful to you.

1. The most straight-forward info about individual matched pairs is contained in the MPR line type. In there, the OBS_SID column already contains the station name. However, I presume there's some good reason that you want SL1L2 lines instead of MPR lines.

2. You could run the Stat-Analysis tool on the MPR output from Point-Stat, converting it to SL1L2 and populating the DESC in exactly the way you want. The -set_hdr job command option can be used to override output columns either with a constant string or referencing one of the values from the input columns.

Here's an example:

stat_analysis -lookin mpr.txt \
-job aggregate_stat -line_type MPR -out_line_type SL1L2 \
-by OBS_SID -set_hdr DESC OBS_SID -out_stat sl1l2.txt

And here's the input mpr.txt and output
sl1l2.txt files.

So this does achieve the desired result but requires an extra post-processing step through Stat-Analysis. In this command, the -set_hdr DESC OBS_SID option is doing the work... reading the value from the input OBS_SID column of the MPR line type and writing it the output DESC column of the SL1L2 line type.

3. The Point-Stat mask configuration option defines how masking regions are defined. The sid (or station id) option enables you to list individual station names to be used. If you set sid to a list station names, then each will be processed as its own masking region. For example,

mask = {
...
   sid = [ "71069", "71108", "71114", "71118", "71123", "71207", "71450", "71455", "71457", "71473" ];
...
}

Then you'd get SL1L2 output separately for each one of these masks, where each mask consists of a single location. And this station id name will appear in the VX_MASK output column. If you have a relatively small set of stations and already know what they are by name, this could be a useful option. If the number of stations is large, this would be less useful and tedious to setup in a configuration file.

I can imagine how Point-Stat could be enhanced to do exactly what you're asking. However, I would still describe this as being an edge case. The intention of Point-Stat is to aggregate data across multiple points to compute partial sums and/or statistics. Information about individual matched pairs is available in the MPR line type. Perhaps we should step back and see why you're using SL1L2 lines rather than MPR lines and see if we can make it possible or easier to use the MPR lines instead?

[CGenie]

Hello,

Thank you for such a detailed answer!

Actually, I'm a beginner with metpluls so I'm not sure I know what I'm doing exactly :)

Anyways, about MPR, for some reason I get NA values for all the CLIMO_ columns. Can this be debugged in some way? It's strange because my sl1l2 values are OK.

[JohnHalleyGotway]

@CGenie, there are many choices of climatology that folks might use and, by default, none are used. Just like forecast and observation data, users can configure what climatological data is used. Generally speaking, climo data is specified in METplus using climatological mean fields and, optionally, climatological standard deviation fields.

Here's a METplus Use Case that demonstrates specifying the climo files:

GRID_STAT_CLIMO_MEAN_INPUT_DIR = {INPUT_BASE}/model_applications/medium_range/grid_to_grid/nwprod/fix
GRID_STAT_CLIMO_MEAN_INPUT_TEMPLATE = cmean_1d.1959{valid?fmt=%m%d}

Looks like we're actually pretty limited on METplus examples of specifying both climo mean and standard deviation. FYI, the climo files we use for testing (and that are also used by NOAA/EMC for verifying their model output) can be found here:
https://dtcenter.ucar.edu/dfiles/code/METplus/MET/MET_unit_test/unit_test/climatology_data/

In particular, NCEP is mostly using this Daily 1.0 degree global data. The cmean files contain climatological means and the cstdv files contain the standard deviations. The files are in GRIB1 format and contain data for several variables (HGT, PRMSL, PWAT, TMAX, TMIN, TMP, UGRD, VGRD) and vertical levels (10 m above gnd, 1000 mb, 2 m above gnd, 250 mb, 500 mb, 700 mb, 850 mb, MSL, atmos col).

[JohnHalleyGotway]

@CGenie wanted to followup on this discussion from a few months ago. Can we mark this as being answered or is more discussion needed?

[CGenie]

Yes, you can close this. Thank you.