[Met_help] [rt.rap.ucar.edu #78073] History for Verification of Probabilistic Forecasts Using grid_stat

John Halley Gotway via RT met_help at ucar.edu
Fri Sep 23 11:12:11 MDT 2016


----------------------------------------------------------------
  Initial Request
----------------------------------------------------------------

Hello,

I am having trouble properly configuring the grid-stat configuration file
to verify a probabilistic forecast.  I am attempting to follow the
instructions in the online tutorial.  My inputs are two GRIB2 files, one
containing the forecast, the other contain the analysis.  I have set the
prob=TRUE flag in the forecast field.

I receive the error:
GridStatConfInfo::process_config() -> The number of thresholds for each
field in "fcst.cat_thresh" must match the number of thresholds for each
field in "obs.cat_thresh"

However, I believe I have configured my config file correctly based on
reading the documentation.

The relevant section of the config file is:
//
// Forecast and observation fields to be verified
//
fcst = {

   field = [
      {
        name       = "APCP";
        level      = [ "A24" ];
        prob       = TRUE;
        cat_thresh = [ >=0.0, >=0.1, >=0.2, >=0.3, >=0.4, >=0.5, >=0.6,
>=0.7, >=0.8, >=0.9, >=1.0 ];
      }
   ];

};
obs = {
   field = [
     {
       name       = "APCP";
       level      = [ "A24" ];
       cat_thresh = [ >=6.35 ];
     }
   ];
};


Thank you for any direction you can provide.

Bruce

-- 
Bruce Veenhuis
Meteorologist
NWS/NCEP/Weather Prediction Center


----------------------------------------------------------------
  Complete Ticket History
----------------------------------------------------------------

Subject: Verification of Probabilistic Forecasts Using grid_stat
From: John Halley Gotway
Time: Fri Sep 23 10:09:13 2016

Hello Bruce,

I see that you're having trouble configuring MET for probabilistic
verification.

Here's the short answer... setup the "fcst" field like this:

field = [ {
               name = "PROB"; level = "A24";
               prob = { name = "APCP"; thresh_lo = 0; };
            }
          ];


And here's the long answer...

The "prob" config file entry can either be set to a boolean, like
"prob =
TRUE;" or to a string which describes information about the
probabilistic
field.  Page 48 of the MET Users's Guide includes examples of both:

http://www.dtcenter.org/met/users/docs/users_guide/MET_Users_Guide_v5.2.pdf

In the first example, we're reading a field from a NetCDF file that
contains values between 0 and 1.  By setting "prob = TRUE;" we tell
MET to
interpret and verify that data as probabilities.

In the second example, we're reading data from a GRIB file.  In NCEP's
usage of GRIB, they encode all probability fields using GRIB code 191
(
http://www.nco.ncep.noaa.gov/pmb/docs/on388/table2.html).  So a single
GRIB
file often has many, many records with code = 191.  MET needs to be
told
additional information about which one you want.  NCEP uses the
"extended
product description section" to encode this extra information (
http://www.nco.ncep.noaa.gov/pmb/docs/on388/appendixc.html).  In the
"prob"
section, the "name" setting specifies the GRIB code defined in Octet
46.
The "thresh_lo" and "thresh_hi" are used to specify the values in
Octets 48
- 55.  For example, the following would be used for probability of 2m
temperature being between 273 and 283:
   { name = "PROB"; level = "Z2"; prob = { name = "TMP"; thresh_lo =
273;
thresh_hi = 283; } }

I realize that's confusing but hopefully the example helps clarify.

Thanks,
John Halley Gotway


On Fri, Sep 23, 2016 at 9:12 AM, Bruce Veenhuis - NOAA Federal via RT
<
met_help at ucar.edu> wrote:

>
> Fri Sep 23 09:12:48 2016: Request 78073 was acted upon.
> Transaction: Ticket created by bruce.veenhuis at noaa.gov
>        Queue: met_help
>      Subject: Verification of Probabilistic Forecasts Using
grid_stat
>        Owner: Nobody
>   Requestors: bruce.veenhuis at noaa.gov
>       Status: new
>  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=78073 >
>
>
> Hello,
>
> I am having trouble properly configuring the grid-stat configuration
file
> to verify a probabilistic forecast.  I am attempting to follow the
> instructions in the online tutorial.  My inputs are two GRIB2 files,
one
> containing the forecast, the other contain the analysis.  I have set
the
> prob=TRUE flag in the forecast field.
>
> I receive the error:
> GridStatConfInfo::process_config() -> The number of thresholds for
each
> field in "fcst.cat_thresh" must match the number of thresholds for
each
> field in "obs.cat_thresh"
>
> However, I believe I have configured my config file correctly based
on
> reading the documentation.
>
> The relevant section of the config file is:
> //
> // Forecast and observation fields to be verified
> //
> fcst = {
>
>    field = [
>       {
>         name       = "APCP";
>         level      = [ "A24" ];
>         prob       = TRUE;
>         cat_thresh = [ >=0.0, >=0.1, >=0.2, >=0.3, >=0.4, >=0.5,
>=0.6,
> >=0.7, >=0.8, >=0.9, >=1.0 ];
>       }
>    ];
>
> };
> obs = {
>    field = [
>      {
>        name       = "APCP";
>        level      = [ "A24" ];
>        cat_thresh = [ >=6.35 ];
>      }
>    ];
> };
>
>
> Thank you for any direction you can provide.
>
> Bruce
>
> --
> Bruce Veenhuis
> Meteorologist
> NWS/NCEP/Weather Prediction Center
>
>

------------------------------------------------
Subject: Verification of Probabilistic Forecasts Using grid_stat
From: Bruce Veenhuis - NOAA Federal
Time: Fri Sep 23 10:43:03 2016

Hi John,

That worked correctly and I followed your explanation.  Thank you very
much!

Best,
Bruce

On Fri, Sep 23, 2016 at 12:09 PM, John Halley Gotway via RT <
met_help at ucar.edu> wrote:

> Hello Bruce,
>
> I see that you're having trouble configuring MET for probabilistic
> verification.
>
> Here's the short answer... setup the "fcst" field like this:
>
> field = [ {
>                name = "PROB"; level = "A24";
>                prob = { name = "APCP"; thresh_lo = 0; };
>             }
>           ];
>
>
> And here's the long answer...
>
> The "prob" config file entry can either be set to a boolean, like
"prob =
> TRUE;" or to a string which describes information about the
probabilistic
> field.  Page 48 of the MET Users's Guide includes examples of both:
>
> http://www.dtcenter.org/met/users/docs/users_guide/MET_
> Users_Guide_v5.2.pdf
>
> In the first example, we're reading a field from a NetCDF file that
> contains values between 0 and 1.  By setting "prob = TRUE;" we tell
MET to
> interpret and verify that data as probabilities.
>
> In the second example, we're reading data from a GRIB file.  In
NCEP's
> usage of GRIB, they encode all probability fields using GRIB code
191 (
> http://www.nco.ncep.noaa.gov/pmb/docs/on388/table2.html).  So a
single
> GRIB
> file often has many, many records with code = 191.  MET needs to be
told
> additional information about which one you want.  NCEP uses the
"extended
> product description section" to encode this extra information (
> http://www.nco.ncep.noaa.gov/pmb/docs/on388/appendixc.html).  In the
> "prob"
> section, the "name" setting specifies the GRIB code defined in Octet
46.
> The "thresh_lo" and "thresh_hi" are used to specify the values in
Octets 48
> - 55.  For example, the following would be used for probability of
2m
> temperature being between 273 and 283:
>    { name = "PROB"; level = "Z2"; prob = { name = "TMP"; thresh_lo =
273;
> thresh_hi = 283; } }
>
> I realize that's confusing but hopefully the example helps clarify.
>
> Thanks,
> John Halley Gotway
>
>
> On Fri, Sep 23, 2016 at 9:12 AM, Bruce Veenhuis - NOAA Federal via
RT <
> met_help at ucar.edu> wrote:
>
> >
> > Fri Sep 23 09:12:48 2016: Request 78073 was acted upon.
> > Transaction: Ticket created by bruce.veenhuis at noaa.gov
> >        Queue: met_help
> >      Subject: Verification of Probabilistic Forecasts Using
grid_stat
> >        Owner: Nobody
> >   Requestors: bruce.veenhuis at noaa.gov
> >       Status: new
> >  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=78073 >
> >
> >
> > Hello,
> >
> > I am having trouble properly configuring the grid-stat
configuration file
> > to verify a probabilistic forecast.  I am attempting to follow the
> > instructions in the online tutorial.  My inputs are two GRIB2
files, one
> > containing the forecast, the other contain the analysis.  I have
set the
> > prob=TRUE flag in the forecast field.
> >
> > I receive the error:
> > GridStatConfInfo::process_config() -> The number of thresholds for
each
> > field in "fcst.cat_thresh" must match the number of thresholds for
each
> > field in "obs.cat_thresh"
> >
> > However, I believe I have configured my config file correctly
based on
> > reading the documentation.
> >
> > The relevant section of the config file is:
> > //
> > // Forecast and observation fields to be verified
> > //
> > fcst = {
> >
> >    field = [
> >       {
> >         name       = "APCP";
> >         level      = [ "A24" ];
> >         prob       = TRUE;
> >         cat_thresh = [ >=0.0, >=0.1, >=0.2, >=0.3, >=0.4, >=0.5,
>=0.6,
> > >=0.7, >=0.8, >=0.9, >=1.0 ];
> >       }
> >    ];
> >
> > };
> > obs = {
> >    field = [
> >      {
> >        name       = "APCP";
> >        level      = [ "A24" ];
> >        cat_thresh = [ >=6.35 ];
> >      }
> >    ];
> > };
> >
> >
> > Thank you for any direction you can provide.
> >
> > Bruce
> >
> > --
> > Bruce Veenhuis
> > Meteorologist
> > NWS/NCEP/Weather Prediction Center
> >
> >
>
>


--
Bruce Veenhuis
Meteorologist
NWS/NCEP/Weather Prediction Center

------------------------------------------------
Subject: Verification of Probabilistic Forecasts Using grid_stat
From: John Halley Gotway
Time: Fri Sep 23 11:11:37 2016

Great!  Thanks for letting me know that worked.

John

On Fri, Sep 23, 2016 at 10:43 AM, Bruce Veenhuis - NOAA Federal via RT
<
met_help at ucar.edu> wrote:

>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=78073 >
>
> Hi John,
>
> That worked correctly and I followed your explanation.  Thank you
very
> much!
>
> Best,
> Bruce
>
> On Fri, Sep 23, 2016 at 12:09 PM, John Halley Gotway via RT <
> met_help at ucar.edu> wrote:
>
> > Hello Bruce,
> >
> > I see that you're having trouble configuring MET for probabilistic
> > verification.
> >
> > Here's the short answer... setup the "fcst" field like this:
> >
> > field = [ {
> >                name = "PROB"; level = "A24";
> >                prob = { name = "APCP"; thresh_lo = 0; };
> >             }
> >           ];
> >
> >
> > And here's the long answer...
> >
> > The "prob" config file entry can either be set to a boolean, like
"prob =
> > TRUE;" or to a string which describes information about the
probabilistic
> > field.  Page 48 of the MET Users's Guide includes examples of
both:
> >
> > http://www.dtcenter.org/met/users/docs/users_guide/MET_
> > Users_Guide_v5.2.pdf
> >
> > In the first example, we're reading a field from a NetCDF file
that
> > contains values between 0 and 1.  By setting "prob = TRUE;" we
tell MET
> to
> > interpret and verify that data as probabilities.
> >
> > In the second example, we're reading data from a GRIB file.  In
NCEP's
> > usage of GRIB, they encode all probability fields using GRIB code
191 (
> > http://www.nco.ncep.noaa.gov/pmb/docs/on388/table2.html).  So a
single
> > GRIB
> > file often has many, many records with code = 191.  MET needs to
be told
> > additional information about which one you want.  NCEP uses the
"extended
> > product description section" to encode this extra information (
> > http://www.nco.ncep.noaa.gov/pmb/docs/on388/appendixc.html).  In
the
> > "prob"
> > section, the "name" setting specifies the GRIB code defined in
Octet 46.
> > The "thresh_lo" and "thresh_hi" are used to specify the values in
Octets
> 48
> > - 55.  For example, the following would be used for probability of
2m
> > temperature being between 273 and 283:
> >    { name = "PROB"; level = "Z2"; prob = { name = "TMP"; thresh_lo
= 273;
> > thresh_hi = 283; } }
> >
> > I realize that's confusing but hopefully the example helps
clarify.
> >
> > Thanks,
> > John Halley Gotway
> >
> >
> > On Fri, Sep 23, 2016 at 9:12 AM, Bruce Veenhuis - NOAA Federal via
RT <
> > met_help at ucar.edu> wrote:
> >
> > >
> > > Fri Sep 23 09:12:48 2016: Request 78073 was acted upon.
> > > Transaction: Ticket created by bruce.veenhuis at noaa.gov
> > >        Queue: met_help
> > >      Subject: Verification of Probabilistic Forecasts Using
grid_stat
> > >        Owner: Nobody
> > >   Requestors: bruce.veenhuis at noaa.gov
> > >       Status: new
> > >  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=78073
> >
> > >
> > >
> > > Hello,
> > >
> > > I am having trouble properly configuring the grid-stat
configuration
> file
> > > to verify a probabilistic forecast.  I am attempting to follow
the
> > > instructions in the online tutorial.  My inputs are two GRIB2
files,
> one
> > > containing the forecast, the other contain the analysis.  I have
set
> the
> > > prob=TRUE flag in the forecast field.
> > >
> > > I receive the error:
> > > GridStatConfInfo::process_config() -> The number of thresholds
for
> each
> > > field in "fcst.cat_thresh" must match the number of thresholds
for each
> > > field in "obs.cat_thresh"
> > >
> > > However, I believe I have configured my config file correctly
based on
> > > reading the documentation.
> > >
> > > The relevant section of the config file is:
> > > //
> > > // Forecast and observation fields to be verified
> > > //
> > > fcst = {
> > >
> > >    field = [
> > >       {
> > >         name       = "APCP";
> > >         level      = [ "A24" ];
> > >         prob       = TRUE;
> > >         cat_thresh = [ >=0.0, >=0.1, >=0.2, >=0.3, >=0.4, >=0.5,
>=0.6,
> > > >=0.7, >=0.8, >=0.9, >=1.0 ];
> > >       }
> > >    ];
> > >
> > > };
> > > obs = {
> > >    field = [
> > >      {
> > >        name       = "APCP";
> > >        level      = [ "A24" ];
> > >        cat_thresh = [ >=6.35 ];
> > >      }
> > >    ];
> > > };
> > >
> > >
> > > Thank you for any direction you can provide.
> > >
> > > Bruce
> > >
> > > --
> > > Bruce Veenhuis
> > > Meteorologist
> > > NWS/NCEP/Weather Prediction Center
> > >
> > >
> >
> >
>
>
> --
> Bruce Veenhuis
> Meteorologist
> NWS/NCEP/Weather Prediction Center
>
>

------------------------------------------------


More information about the Met_help mailing list