[Met_help] Running point stat
John Halley Gotway
johnhg at rap.ucar.edu
Wed May 6 16:03:24 MDT 2009
I took a look at the data you sent. I believe there are two likely
culprits that may explain why you're not getting any matched pairs.
(1) Timing problems.
It may be the case that none of your observations are falling within the
matching time window for this file - resulting in 0 matched pairs. We
actually recently posted a bug fix for MET about parsing initialization
times from GRIB files. The code was assuming that all times were after
the year 2000. When processing forecasts from before 2000, Point-Stat
computed the wrong initilization time resulting in 0 matched pairs. I
looked at the wgrib output you sent and it appears that the data is from
01/01/2000 00:00:00. So this bug shouldn't be the culprit.
But I would suggest retrieving all of the latest bug fix by downloading
the file "METv2.0_patches_20090505.tar.gz" from:
(2) Spatial problems.
It may be possible that none of your forecasts and/or observations are
falling within your spatial verification region.
I suggest setting the "mask_grid" in the configuration file as:
mask_grid = [ "FULL" ];
That means to verify over the full input forecast domain. I checked the
locations of your point observations and they all appear to be in
California. Are you sure that the domain of the forecast data includes
California, or is it possible that in writing the GRIB file, the domain
definition was messed up? You could try running your forecast file
through MODE, even just compare it to itself by passing the same file as
the forecast and observation. Then take a look at MODE's output
PostScript plot to see where the MET tools think the forecast grid resides
on the earth. If it doesn't include CA, then that's the problem. If it
does, it's not the problem.
If these two things don't help, I'd be happy to take a look at your GRIB
file to see if I can figure out what's going on. I'd suggest copying it
to our anonymous ftp site:
username = anonymous
password = your email address
put "file names"
John Halley Gotway
johnhg at ucar.edu
> Dear sirs,
> I am in the learning stage of MET, so please excuse my inexperience in
> using the software, and questions that I will have.
> I am trying to run point_stat by following the instructions within the
> user documentation; however, the output files are still empty although
> there is no error message. When I installed the MET, I ran the test
> script, which didn't indicate any error that MET wasn't working
> properly. I also updated the MET modules by installing the patch tar
> file, and compiling the code. I am comparing MM5 output with point
> observations for time period between January 1, 2000 and January 4,
> 2000, as an example. I first converted MM5 sigma levels to MM5 pressure
> levels, and created grib1 file from there as forecast file. Then, I
> converted ascii point observations to netcdf file. When I check the
> contents of each file, they all have reasonable values (to the best of
> my knowledge). But, obviously there is either some data that is not
> quite right, or I am doing something wrong, including the setup of
> configuration file, and as a result I don't get any error or any result.
> Could you please help me in identifying my error by reviewing the
> contents of the attached tar file, which has the input (obs netcdf ,
> content of gribfile input, output, and configuration files, and my
> command at the prompt, and its output?
> Thank you very much in advance. With regards,
> Met_help mailing list
> Met_help at mailman.ucar.edu
More information about the Met_help