[Met_help] Zero output

John Halley Gotway johnhg at ucar.edu
Fri Dec 4 11:45:56 MST 2009


Joe,

OK, I see what's going on, and it's kind of subtle.  In the Point-Stat configuration file, I changed the "fcst_field" line:
FROM THIS: fcst_field[] = [ "UGRD/Z10", "VGRD/Z10", "TMP/Z2" ];
  TO THIS: fcst_field[] = [ "UGRD/L10", "VGRD/L10", "TMP/L2" ];

Once I did that, it ran fine and did find matched pairs.  Here's why.

The Point-Stat tool does matching for surface variables in a very simplistic way.  It's all based on message type.  This is inherited from the way NCEP does this type of matching.  So by convention,
any observation with a message type of ADPSFC or SFCSHP is considered to be at the surface.  And by convention, 10-meter winds and 2-meter temperatures are considered to be surface variables.  So when
you say, "UGRD/Z10" that tells Point-Stat to get the U-component of wind at 10-meters and treat it as a surface variable... i.e. only match it to observations of type ADPSFC or SFCSHP.

By changing it to "UGRD/L10" that tells Point-Stat to get the U-component of wind with a level value of 10 and not to worry about only matching it to surface observations.  Instead, it matches it to
all observation types you request - MSONET in your case.  I assume all of your observations are actually at the surface which makes this work fine.

So going forward, you're welcome to use "L10" and "L2".  Or when building your observation sets with ASCII2NC, you could use the "ADPSFC" message type and keep using "Z10" and "Z2".

Based on your data, here's a couple more suggestions:

- I see that you have a whole month's worth of observations in your NetCDF file.  With data every 5 minutes, that's 4,773,360 observations for the month!  Point-Stat seems to handle it fine, but it
sure does make Point-Stat run slowly.  Instead of storing a whole month's worth of observations in each file, I'd suggest doing it daily or hourly instead.  That'd make each run of Point-Stat much
faster since it won't have to parse through millions of observations each time it runs.

- In the Point-Stat config file, you're using the "FULL" mask and the "CONUS.poly" mask.  I looked at your domain, and it's basically just all of Oklahoma.  Using "FULL" and "CONUS.poly" will have the
same effect, they both just give you the full domain.  So get rid of the "CONUS.poly" mask, and just use "FULL" by itself.

Hope that helps.

John

Joe Eastman wrote:
> Hi John,
> 
> The data should be up there shortly. I put a WRF grib file and the mesonet netCDF file in the specified directory.
> 
> Thanks,
> Joe
> 
> Dr. Joseph L. Eastman
> 
> Senior Atmospheric Scientist 
> WindLogics Inc.
> 
> 201 4th St NW
> 
> Grand Rapids, MN
> 
> 55744
> 
>  
> 
> c - 410.279.9702
> 
> w - 651.556.4297
> 
> jeastman at windlogics.com
> 
>  
> 
>  
> 
> 
> -----Original Message-----
> From: John Halley Gotway [mailto:johnhg at ucar.edu] 
> Sent: Friday, December 04, 2009 10:16 AM
> To: Joe Eastman
> Cc: met_help at ucar.edu
> Subject: Re: [Met_help] Zero output
> 
> Joe,
> 
> I see that you're running the Point-Stat tool but are getting zero matched pairs.  There are many reasons why this could be happening, such as...
> - Perhaps the observations don't fall into the region you're verifying - maybe you have a sign wrong in the lat/lon location.
> - Perhaps the forecast and observation times don't match up.
> - Perhaps the observation message type doesn't match what you're trying to verify.
> 
> It may be easiest for you to just send me a sample forecast and observation file, and I'll take a look and see what's going on.  Could you post a sample forecast and observation file to our anonymous
> ftp site?  Here's how you'd do that:
> 
> ftp ftp.rap.ucar.edu
> username = anonymous
> password = "your email address"
> cd incoming/irap/met_help/eastman_data
> put "your forecast file"
> put "your observation file"
> bye
> 
> To answer you're second question... the MET stats tools (Grid-Stat, Point-Stat, Wavelet-Stat, and MODE) were designed to run at a single point in time.  So you should run Point-Stat once for each
> output time... in your case, every 10 minutes.  To aggregate statistics through time, you'll need to run the STAT-Analysis tool.  Now I see that you're trying to verify over the whole continental US
> as well as several individual stations.  When verifying at stations, each individual station is treated as it's own "masking region".  For each run of Point-Stat, you should only be getting 1 matched
> pair for each station.  So you really will need to run your output through the STAT-Analysis tool to aggregate through time, which is what I assume you're really interested in.  I realize that this
> design is cumbersome for what you're trying to do - but that's how it currently works.
> 
> Based on what you appear to be doing, here are some suggestions:
> - If you haven't already done so, retrieve and install the latest set of patches for METv2.0 from:
>      http://www.dtcenter.org/met/users/support/known_issues/METv2.0/index.php
> - Do NOT combine the GRIB files for multiple times into a single file.  MET expects that GRIB output from separate times be in separate files.
> - I see that you've set "beg_ds" and "end_ds" to +/- 60 seconds in the config file which will match +/- 1 minute around the forecast valid time.  That looks like a good choice to me for your data.
> 
> Let's try to get a single case running for now, and once that's working well, I'll be happy to help you aggregate them through time using STAT-Analysis.
> 
> Please let me know if you're able to post those forecast/observation files.  And I'll take a look to see if I can provide better advice.
> 
> Thanks,
> John Halley Gotway
> johnhg at ucar.edu
> 
> Joe Eastman wrote:
>> Hello,
>>
>> I am new to using METv2.0, so this is probably something silly. I run with the command:
>>
>> ~/METv2.0-update_bin/point_stat myj_d03_2002-05-05_06:00:00 okmet.nc pointstat -outdir . -v 1000 >&m.out&
>>
>> What I end up with is zero output in the txt and stat files. It prints the headers and nothing else. I have converted some Oklahoma mesonet data successfully using the ascii2nc program. I included my station file too. The other question that is not obvious to is how does one do multiple time periods. I have data every 5 minutes. I write the grib WRF output every 10 minutes and would like evaluation over all forecast times in a single file. Is this possible? Do I need to write all WRF output to a single grib file?
>>
>> Thanks for your time.
>>
>> Take care,
>> Joe
>>
>>
>> Dr. Joseph L. Eastman
>> Senior Atmospheric Scientist
>> WindLogics Inc.
>> 201 4th St NW
>> Grand Rapids, MN
>> 55744
>>
>> c - 410.279.9702
>> w - 651.556.4297
>> jeastman at windlogics.com<mailto:jeastman at windlogics.com>
>>
>>
>>
>> ________________________________
>> CONFIDENTIALITY NOTICE: This e-mail message is for the sole use of the intended recipient(s) and may contain confidential and privileged information. Any unauthorized review, use, disclosure or distribution of any kind is strictly prohibited. If you are not the intended recipient, please contact the sender via reply e-mail and destroy all copies of the original message. Thank you.
>>
>>
>>
>> ------------------------------------------------------------------------
>>
>> _______________________________________________
>> Met_help mailing list
>> Met_help at mailman.ucar.edu
>> http://mailman.ucar.edu/mailman/listinfo/met_help


More information about the Met_help mailing list