[Met_help] fm theo carter dubai
John Halley Gotway
johnhg at rap.ucar.edu
Thu Jun 18 08:38:26 MDT 2009
Theo,
Based on your message, it sounds like you really only have 6 matched pairs of data: 6 forecast values compared to 6 observation values. Statistically speaking, there isn't very much you can say about
such a small sample. I'd suggest performing the verification for many points in time. Suppose for example, that you've run the Point-Stat verification with those 6 stations once per day for 30 days.
For each one of those Point-Stat runs, you'd have a STAT output file. Here are some things you might do with that STAT output:
(1) You may want to simply look at a time series of RMSE values for mean sea level pressure over those 30 days. You wouldn't need to run STAT-Analysis for this, you'd just extract the RMSE column of
data from the CNT line type, and use whatever plotting tools you like to plot a time series of those values.
(2) You may want to look at a summary of those RMSE values for mean sea level pressure. Here's where you'd run the STAT-Analysis summary job. Something like this should do the trick:
"-job summary -line_type CNT -column RMSE -fcst_var PRMSL -dump_row summary_job.stat"
This is assuming that the sea-level pressure variable in your STAT file is named "PRMSL". All this job will do is look at the RMSE column of data from the CNT lines - in this case, it should
consist of 30 values - and it will compute summary info for that column of data. The summary info is described in the MET User's Guide and in the comments of the STAT-Analysis config file. When
running a job like this for the first time, you should use the "-dump_row" option which dumps the STAT lines that went into the computation to an output STAT file. Then you should look at those line
to verify that the job was run over the lines you intended. Suppose for example, your STAT output file contains statistics for PRMSL and Temperature... if you forget to specify "-fcst_var PRMSL",
then the summary job would be performed over the PRMSL AND the Temperature data. By looking in the dump_row file, you can make sure that you're filtering the STAT data the way you intended.
(3) Since you have such a small set of data, for each of the Point-Stat runs, you could turn ON the matched pair line (MPR) output. This will dump the raw fcst/obs matched pair values. And then in
STAT-Analysis, you could run many jobs on that matched pair data. In (2) above, we looked at a summary of the RMSE values for 30 days. However, you could also just compute an RMSE value directly
using the MPR data for those 30 days. In this example, we should have 30 days * 6 stations = 180 matched fcst/obs pairs of data. Here's a job you might run on that:
"-job aggregate_stat -fcst_var PRMSL -line_type MPR -out_line_type CNT -dump_row aggregate_stat_job.stat"
This job will look at the 180 matched pair lines in your Point-Stat output and compute a continuous statistics line type from those 180 matched pairs.
Hopefully that helps. I'd encourage you to do 4 things:
- make sure you read the MET User's Guide about STAT-Analysis
- take a look at the STAT-Analysis section of the MET online tutorial: http://www.dtcenter.org/met/users/support/online_tutorial/METv2.0/stat_analysis/index.php
- read the comments in the STAT-Analysis config file in METv2.0/scripts/config/STATAnalysisConfig
- and lastly, just start playing around with running STAT-Analysis jobs to see what it does
If you have questions about interpreting the statistical output of MET, I'll refer you to the statistician on our team.
Thanks,
John
Theo Carter wrote:
> Hi John,
>
>
>
> Thanks for the help so far! It is very appreciated. I have managed to
> find my way through pb2nc and pointstat, and am now looking at STAT to
> summarize (make a little more visible) the output from the pointstat
> tool for me. I have compared mean sea level pressure at 6 stations in
> the UAE between wrf output and prepbufr obs. Now I would like to
> explore how to summarize how the actuals compare with the forecast.
>
>
>
> So am trying to use stat analysis. Doing a filter job seemed fairly
> straight forward, and it wrote out a file which looked almost exactly
> the same as the one .stat which came from pointstat. But how does one
> call 'summary'? The documentation talks about using two options,
> -line_type and -column. Can you give me an example of a practical line
> for the
>
>
>
> jobs[] = [
>
> "-job summary ..............."
>
> ];
>
>
>
>
>
> line in the config file? I cannot figure it out,
>
> Thanks,
>
> Theo
>
>> Date: Tue, 16 Jun 2009 06:38:49 -0600
>> From: johnhg at rap.ucar.edu
>> To: thibidottwo at hotmail.com
>> CC: met_help at ucar.edu
>> Subject: Re: [Met_help] fm theo carter dubai
>>
>> Theo,
>>
>> I'm guessing that you have a trailing space in the first line of the polyline file. I tried adding a trailing space to the first line of a polyline file and running it through the Gen-Poly-Mask tool,
>> and I saw the exact same error message -> NetCDF: Name contains illegal characters
>>
>> So edit the file UAE.poly by removing any trailing spaces from the lines, and try rerunning it through Gen-Poly-Mask.
>>
>> Thanks,
>> John
>>
>> Theo Carter wrote:
>>> Hi John
>>>
>>>
>>>
>>> Thank you for the help so far. I have hit a snag trying to create the
>>> bitmap mask using Gen-Poly-Mask and cannot figure out what is going
>>> wrong.
>>>
>>>
>>>
>>> I use the .poly polyline file you sent through to create the mask with:
>>> It has been renamed UAE.poly and in it is the following (which came
>>> from you)
>>>
>>> UAE
>>>
>>>
>>> 26.242 56.074
>>>
>>>
>>> 25.669 56.503
>>>
>>>
>>> 25.013 56.492
>>>
>>>
>>> 24.054 56.107
>>>
>>>
>>> 22.418 55.283
>>>
>>>
>>> 22.885 52.394
>>>
>>>
>>> 24.134 51.526
>>>
>>>
>>> 24.534 51.537
>>>
>>>
>>> 24.714 54.053
>>>
>>>
>>>
>>>
>>>
>>> I then tried this on two different .grb files - firstly the wrf output
>>> from my wrf arw here and then on a gfs .grb file and got the same
>>> result using:
>>>
>>>
>>>
>>> bin/gen_poly_mask 20090614/wrfarwoutput/wrfpst_d01_20090613_1200_f02400.grb data/poly/UAE.poly 20090614/polybitmap.nc
>>>
>>> Input Data File: 20090614/wrfarwoutput/wrfpst_d01_20090613_1200_f02400.grb
>>>
>>> Input Poly File: data/poly/UAE.poly
>>>
>>> Parsed Grid: Lambert Conformal (256 x 192)
>>>
>>> containing 8 points 26.242 56.074
>>>
>>> NetCDF: Name contains illegal characters
>>>
>>> [wrf at localhost METv2.0]$ bin/gen_poly_mask
>>> 20090614/wrfarwoutput/wrfpst_d01_20090613_1200_f02400.grb
>>> data/poly/UAE.poly 20090614/polybitmap.nc
>>>
>>> Input Data File: 20090614/wrfarwoutput/wrfpst_d01_20090613_1200_f02400.grb
>>>
>>> Input Poly File: data/poly/UAE.poly
>>>
>>> Parsed Grid: Lambert Conformal (256 x 192)
>>>
>>> containing 8 points 26.242 56.074
>>>
>>> NetCDF: Name contains illegal characters
>>>
>>> [wrf at localhost METv2.0]$ bin/gen_poly_mask
>>> 20090614/wrfarwoutput/wrfpst_d01_20090613_1200_f02400.grb
>>> data/poly/UAE.poly 20090614/polybitmap.nc
>>>
>>> Input Data File: 20090614/wrfarwoutput/wrfpst_d01_20090613_1200_f02400.grb
>>>
>>> Input Poly File: data/poly/UAE.poly
>>>
>>> Parsed Grid: Lambert Conformal (256 x 192)
>>>
>>> containing 9 points UAE
>>>
>>> NetCDF: Name contains illegal characters
>>>
>>> [wrf at localhost METv2.0]$ bin/gen_poly_mask theotest2/gfs_3_20090608_0000_000.grb data/poly/UAE.poly 20090614/polybitmap.nc
>>>
>>> Input Data File: theotest2/gfs_3_20090608_0000_000.grb
>>>
>>> Input Poly File: data/poly/UAE.poly
>>>
>>> Parsed Grid: LatLon (360 x 181)
>>>
>>> containing 9 points UAE
>>>
>>> NetCDF: Name contains illegal characters
>>>
>>> [wrf at localhost METv2.0]$
>>>
>>>
>>>
>>> I tried renaming all the files and whatever I could think of, can you
>>> spot anything stupid or something I am doing wrong in my method here?
>>>
>>> Cheers,
>>>
>>> Theo
>>>
>>>
>
More information about the Met_help
mailing list