[Met_help] [rt.rap.ucar.edu #40156] History for 2 questions regarding precipitation verification

Wed Aug 18 14:39:07 MDT 2010

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

Good morning,

I have 2 questions regarding verification of precipitation observations.

*[Question 1]*

I am trying to validate 1-hour accumulated precipitation from ADPSFC,
MESONET, and SFCSHP observations to compare against WRF. However, point_stat
does not seem to be calculating any comparisons. I verified both observation
and model files contain the APCP field. The observation file contains 122
entries for this field inside my domain. I tried setting the following
parameters

fcst_field[]  = [ "APCP/A1" ];
fcst_field[]  = [ ];
fcst_thresh[] = [ "ge0" ];
fcst_thresh[] = [ ];

but all the output *txt files contain only headers...no statistics info. I
tried changing the variable to "APCP/L0", but that caused the program to
stop with an error. Could you please take a look at my input files (see
tarball attached) and see if there is anything obviously wrong? I am stumped
as to why no matches are occurring. The output file "pointStat.log" was
created with the following command

point_stat NR05_pcp_001.nc madisObs_20060802_010000.nc PointStatConfig >&
pointStat.log

*[Question 2]*

I am also trying to validate WRF against NCEP Stage_IV 1-hr accumulated
precipitation. First, I converted my WRF domain (~18km grid spacing) to
Stage_IV (~4km grid spacing) to preserve the observation accuracy.
Unfortunately, the interpolated grid is spatially larger than my original
domain containing ~32x more data points. This results in long computation
times! I tried applying a polygon mask in the shape of my original domain
hoping to filter the extraneous grid points, but it still takes just as
long. Is there any advice you could give to slash computation times?

Thanks,
Scott

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

Subject: Re: [rt.rap.ucar.edu #40156] 2 questions regarding precipitation verification
From: Paul Oldenburg
Time: Wed Aug 18 11:30:44 2010

Scott,

Question #1:  The reason that you are getting no matched pairs for the
observation type ADPSFC is that the accumulation
interval in your madis observation text file is not set correctly.
Please refer to the MET User's Guide pg. 3-12.  In
the table that describes the input file format, it specifies that
field 8 should contain "Pressure level in hPa or
accumulation interval in hour for the observation value".  Since you
have 1 hour accumulations in the forecast file, you
should specify 1 in that field instead of the value of -999999.00 that
is currently set.  Also, please note that
forecast surface accumulations are only matched with observation types
of ADPSFC and SFCSHP.  The observations you have
with type ADPUPA and AIRCAR will not be matched for this reason.

Question #2:  Two ways to speed up grid_stat are to turn off
bootstrapping and computation of rank correlation
coefficients.  In the grid_stat configuration file, the following
settings are used to do this.

n_boot_rep = 0;

rank_corr_flag = 0;

Please let us know if you have any other questions.

Thanks,

Paul

RAL HelpDesk {for Scott Rabenhorst} wrote:
> Wed Aug 18 09:35:35 2010: Request 40156 was acted upon.
> Transaction: Ticket created by scott.rabenhorst at gmail.com
>        Queue: met_help
>      Subject: 2 questions regarding precipitation verification
>        Owner: Nobody
>   Requestors: scott.rabenhorst at gmail.com
>       Status: new
>  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=40156 >
>
>
> Good morning,
>
> I have 2 questions regarding verification of precipitation
observations.
>
> *[Question 1]*
>
> I am trying to validate 1-hour accumulated precipitation from
ADPSFC,
> MESONET, and SFCSHP observations to compare against WRF. However,
point_stat
> does not seem to be calculating any comparisons. I verified both
observation
> and model files contain the APCP field. The observation file
contains 122
> entries for this field inside my domain. I tried setting the
following
> parameters
>
> fcst_field[]  = [ "APCP/A1" ];
> fcst_field[]  = [ ];
> fcst_thresh[] = [ "ge0" ];
> fcst_thresh[] = [ ];
>
> but all the output *txt files contain only headers...no statistics
info. I
> tried changing the variable to "APCP/L0", but that caused the
program to
> stop with an error. Could you please take a look at my input files
(see
> tarball attached) and see if there is anything obviously wrong? I am
stumped
> as to why no matches are occurring. The output file "pointStat.log"
was
> created with the following command
>
> point_stat NR05_pcp_001.nc madisObs_20060802_010000.nc
PointStatConfig >&
> pointStat.log
>
>
> *[Question 2]*
>
> I am also trying to validate WRF against NCEP Stage_IV 1-hr
accumulated
> precipitation. First, I converted my WRF domain (~18km grid spacing)
to
> Stage_IV (~4km grid spacing) to preserve the observation accuracy.
> Unfortunately, the interpolated grid is spatially larger than my
original
> domain containing ~32x more data points. This results in long
computation
> times! I tried applying a polygon mask in the shape of my original
domain
> hoping to filter the extraneous grid points, but it still takes just
as
> long. Is there any advice you could give to slash computation times?
>
> Thanks,
> Scott
>

------------------------------------------------
Subject: 2 questions regarding precipitation verification
From: Scott Rabenhorst
Time: Wed Aug 18 11:39:47 2010

Hi Paul,

Thanks for this clarification on accumulation vs pressure in the
observation
file. That makes sense. As for the efficiency problem, I will try
turning
off the bootstrapping as you suggest and see how that performs.

Thanks!
Scott

On Wed, Aug 18, 2010 at 1:30 PM, RAL HelpDesk {for Paul Oldenburg} <
met_help at ucar.edu> wrote:

> Scott,
>
> Question #1:  The reason that you are getting no matched pairs for
the
> observation type ADPSFC is that the accumulation
> interval in your madis observation text file is not set correctly.
Please
> refer to the MET User's Guide pg. 3-12.  In
> the table that describes the input file format, it specifies that
field 8
> should contain "Pressure level in hPa or
> accumulation interval in hour for the observation value".  Since you
have 1
> hour accumulations in the forecast file, you
> should specify 1 in that field instead of the value of -999999.00
that is
> currently set.  Also, please note that
> forecast surface accumulations are only matched with observation
types of
> ADPSFC and SFCSHP.  The observations you have
> with type ADPUPA and AIRCAR will not be matched for this reason.
>
> Question #2:  Two ways to speed up grid_stat are to turn off
bootstrapping
> and computation of rank correlation
> coefficients.  In the grid_stat configuration file, the following
settings
> are used to do this.
>
> n_boot_rep = 0;
>
> rank_corr_flag = 0;
>
> Please let us know if you have any other questions.
>
> Thanks,
>
> Paul
>
> RAL HelpDesk {for Scott Rabenhorst} wrote:
> > Wed Aug 18 09:35:35 2010: Request 40156 was acted upon.
> > Transaction: Ticket created by scott.rabenhorst at gmail.com
> >        Queue: met_help
> >      Subject: 2 questions regarding precipitation verification
> >        Owner: Nobody
> >   Requestors: scott.rabenhorst at gmail.com
> >       Status: new
> >  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=40156 >
> >
> >
> > Good morning,
> >
> > I have 2 questions regarding verification of precipitation
observations.
> >
> > *[Question 1]*
> >
> > I am trying to validate 1-hour accumulated precipitation from
ADPSFC,
> > MESONET, and SFCSHP observations to compare against WRF. However,
> point_stat
> > does not seem to be calculating any comparisons. I verified both
> observation
> > and model files contain the APCP field. The observation file
contains 122
> > entries for this field inside my domain. I tried setting the
following
> > parameters
> >
> > fcst_field[]  = [ "APCP/A1" ];
> > fcst_field[]  = [ ];
> > fcst_thresh[] = [ "ge0" ];
> > fcst_thresh[] = [ ];
> >
> > but all the output *txt files contain only headers...no statistics
info.
> I
> > tried changing the variable to "APCP/L0", but that caused the
program to
> > stop with an error. Could you please take a look at my input files
(see
> > tarball attached) and see if there is anything obviously wrong? I
am
> stumped
> > as to why no matches are occurring. The output file
"pointStat.log" was
> > created with the following command
> >
> > point_stat NR05_pcp_001.nc madisObs_20060802_010000.nc
PointStatConfig >&
> > pointStat.log
> >
> >
> > *[Question 2]*
> >
> > I am also trying to validate WRF against NCEP Stage_IV 1-hr
accumulated
> > precipitation. First, I converted my WRF domain (~18km grid
spacing) to
> > Stage_IV (~4km grid spacing) to preserve the observation accuracy.
> > Unfortunately, the interpolated grid is spatially larger than my
original
> > domain containing ~32x more data points. This results in long
computation
> > times! I tried applying a polygon mask in the shape of my original
domain
> > hoping to filter the extraneous grid points, but it still takes
just as
> > long. Is there any advice you could give to slash computation
times?
> >
> > Thanks,
> > Scott
> >
>
>
>

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