[Met_help] [rt.rap.ucar.edu #80413] History for FCST_LEAD time in Grid-stat

[John Halley Gotway via RT]

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

Thanks for giving the tutorial at NCEP a few weeks ago.  I learned a lot of new ways I want to use the tools, especially stat-analysis.  However, to do this I really need my grid-stat output files to have the correct FCST_LEAD time.  Currently it is all zeros.  I see you used to have a "-fcst_lead" flag you could set while running grid-stat, but that is not supported in MET5.2.  Do I need to set the fcst_lead time manually in some other way?  Perhaps my forecast NetCDF file has some inconsistent formatting or the level is not specified correctly.
  
I have uploaded a forecast file (a 3 day forecast in 24 hour increments), the observation file for the 3rd day,and my config file on the ftp site.  The FCST_LEAD time should be 72 hours.

Best,
Shannon

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

Subject: FCST_LEAD time in Grid-stat
From: John Halley Gotway
Time: Wed May 10 11:35:41 2017

Shannon,

I see that you're having trouble populating the FCST_LEAD output
column when you run Grid-Stat.  Thanks for sending your sample data
files.  I was able to run the met-5.2 version of Grid-Stat and
replicate the behavior your describe.

You are passing to Grid-Stat a CF-compliant NetCDF file.  There is a
specific way in the CF-convention of specifying the initialization
time of the model.  You do so by defining a variable named
"forecast_reference_time".  When that variable is not present, MET
sets the forecast initialization time to the same value as the valid
time... which results in a lead time of 0.

To demonstrate adding "forecast_reference_time", I did the following:

# Run ncdump
ncdump PRATEsfc24all.nc > PRATEsfc24all.ncdump

# Edit the ncdump file by adding:
float forecast_reference_time ;
      forecast_reference_time:long_name = "reference time" ;
      forecast_reference_time:units = "hours since 2015-10-28
00:00:00" ;
 ...
 forecast_reference_time = 0 ;

# Run ncgen
ncgen -o PRATEsfc24all_NEW.nc PRATEsfc24all.ncdump

Rerun Grid-Stat, pointing to the NEW file, and check that the
FCST_LEAD column is now populated with non-zero values (720000).

I also posted that updated NetCDF file to the met_help ftp site:
ftp://ftp.rap.ucar.edu/incoming/irap/met_help/Rees_data/PRATEsfc24all_NEW.nc

Hopefully that'll solve the issue.

Thanks,
John

------------------------------------------------
Subject: FCST_LEAD time in Grid-stat
From: Shannon Rees - NOAA Affiliate
Time: Wed May 10 12:06:40 2017

John,

That worked perfectly.  You can close the ticket now.  Thanks!

Shannon

On Wed, May 10, 2017 at 1:35 PM, John Halley Gotway via RT <
met_help at ucar.edu> wrote:

> Shannon,
>
> I see that you're having trouble populating the FCST_LEAD output
column
> when you run Grid-Stat.  Thanks for sending your sample data files.
I was
> able to run the met-5.2 version of Grid-Stat and replicate the
behavior
> your describe.
>
> You are passing to Grid-Stat a CF-compliant NetCDF file.  There is a
> specific way in the CF-convention of specifying the initialization
time of
> the model.  You do so by defining a variable named
> "forecast_reference_time".  When that variable is not present, MET
sets the
> forecast initialization time to the same value as the valid time...
which
> results in a lead time of 0.
>
> To demonstrate adding "forecast_reference_time", I did the
following:
>
> # Run ncdump
> ncdump PRATEsfc24all.nc > PRATEsfc24all.ncdump
>
> # Edit the ncdump file by adding:
> float forecast_reference_time ;
>       forecast_reference_time:long_name = "reference time" ;
>       forecast_reference_time:units = "hours since 2015-10-28
00:00:00" ;
>  ...
>  forecast_reference_time = 0 ;
>
> # Run ncgen
> ncgen -o PRATEsfc24all_NEW.nc PRATEsfc24all.ncdump
>
> Rerun Grid-Stat, pointing to the NEW file, and check that the
FCST_LEAD
> column is now populated with non-zero values (720000).
>
> I also posted that updated NetCDF file to the met_help ftp site:
> ftp://ftp.rap.ucar.edu/incoming/irap/met_help/Rees_
> data/PRATEsfc24all_NEW.nc
>
> Hopefully that'll solve the issue.
>
> Thanks,
> John
>
>


--
Shannon Rees
Meteorologist
Engility
Geophysical Fluid Dynamics Lab
Princeton University Forrestal Campus
201 Forrestal Rd Princeton, NJ
609-452-5384

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