# [Met_help] [rt.rap.ucar.edu #94778] History for GridStatConfig

John Halley Gotway via RT met_help at ucar.edu
Wed Apr 1 11:08:27 MDT 2020

```----------------------------------------------------------------
Initial Request
----------------------------------------------------------------

Folks - I have a question regarding the regrid = {} object in the GridStatConfig file; I'm currently setting it to:

regrid = {
to_grid    = OBS;
method     = DW_MEAN;
width      = 4;
vld_thresh = 0.5;
shape      = SQUARE;
}

The obs field is 1.0 deg and the forecast probability field is 0.25 deg; are these values the best ones? I'm also creating my own box mask of 20x20 degrees positioned at the tropical cyclone center (lon,lat) point. Thank you.

Efren A. Serra (Contractor)
Physicist

DeVine Consulting, Inc.
Naval Research Laboratory
Marine Meteorology Division
7 Grace Hopper Ave., STOP 2
Monterey, CA 93943
Code 7542
Mobile: 408-425-5027

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

Subject: GridStatConfig
From: John Halley Gotway
Time: Mon Mar 30 17:27:03 2020

Hello Efren,

I see you are looking for advice when regridding your 0.25 degree
model
data to the 1.0 degree analysis grid.

I think I understand your logic.  Each 1.0 degree grid box consists of
4
0.25 degree points in both directions... so 16 points.  So you're
using a
4x4 with the distance-weighted mean method.  However, I suspect that
at the
1.0 degree grid points are really just a subset of the 0.25 degree
grid.
So for each 1.0 degree grid point, there is a 0.25 degree grid point
with
the exact same lat/lon.  The "distance" between the lat/lon's of the
source
and destination grids would be 0.  In the distance-weighted mean
logic, a
distance of 0 essentially means an infinite weight.  So I suspect
using
DW_MEAN in this way would yield the same result as nearest neighbor
(NEAREST).

To test this, I started with a 0.25 degree GFS field of 500mb
temperature.
I regridded it to NCEP grid 3, a 1.0 degree global grid.  But I did it
twice, once using 4x4 DW_MEAN and once using the NEAREST:

*regrid_data_plane gfs.pgrb20p25 G003 gfs.1p0_DW_MEAN.nc
<http://gfs.1p0_DW_MEAN.nc> -field 'name="TMP"; level="P500";' -method
DW_MEAN -width 4 -vld_thresh 0.5*
*regrid_data_plane gfs.pgrb20p25 G003 gfs.1p0_NEAREST.nc
<http://gfs.1p0_NEAREST.nc> -field 'name="TMP"; level="P500";' -method
NEAREST -width 1 -vld_thresh 0.5*

Next, I ran pcp_combine to compute their difference:

*pcp_combine -subtract gfs.1p0_DW_MEAN.nc <http://gfs.1p0_DW_MEAN.nc>
gfs.1p0_NEAREST.nc <http://gfs.1p0_NEAREST.nc> diff.nc
<http://diff.nc>
-field 'name="TMP_P500"; level="(*,*)";'*

The resulting difference field is all 0's.  If you really want to have
all
of the 16 nearest points in the interpolation, I'd recommend using the

*regrid_data_plane gfs.pgrb20p25 G003 gfs.1p0_UW_MEAN.nc
<http://gfs.1p0_UW_MEAN.nc> -field 'name="TMP"; level="P500";' -method
UW_MEAN -width 4 -vld_thresh 0.5*

I reran pcp_combine to diff this output with the nearest neighbor
output
and the result is slightly different.  Ultimately, it's up to you.
There's
no right or wrong here.  Hope this helps clarify what's going on.

Thanks,
John Halley Gotway

On Mon, Mar 30, 2020 at 3:08 PM efren.serra.ctr at nrlmry.navy.mil via RT
<
met_help at ucar.edu> wrote:

>
> Mon Mar 30 15:08:24 2020: Request 94778 was acted upon.
> Transaction: Ticket created by efren.serra.ctr at nrlmry.navy.mil
>        Queue: met_help
>      Subject: GridStatConfig
>        Owner: Nobody
>   Requestors: efren.serra.ctr at nrlmry.navy.mil
>       Status: new
>  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=94778 >
>
>
> Folks - I have a question regarding the regrid = {} object in the
> GridStatConfig file; I'm currently setting it to:
>
> regrid = {
>    to_grid    = OBS;
>    method     = DW_MEAN;
>    width      = 4;
>    vld_thresh = 0.5;
>    shape      = SQUARE;
> }
>
> The obs field is 1.0 deg and the forecast probability field is 0.25
deg;
> are these values the best ones? I'm also creating my own box mask of
20x20
> degrees positioned at the tropical cyclone center (lon,lat) point.
Thank
> you.
>
> Efren A. Serra (Contractor)
> Physicist
>
> DeVine Consulting, Inc.
> Naval Research Laboratory
> Marine Meteorology Division
> 7 Grace Hopper Ave., STOP 2
> Monterey, CA 93943
> Code 7542
> Mobile: 408-425-5027
>
>
>

------------------------------------------------
Subject: RE: [rt.rap.ucar.edu #94778] GridStatConfig
From: efren.serra.ctr at nrlmry.navy.mil
Time: Mon Mar 30 17:29:38 2020

John - How about if I just regrid to the mask file using the MAX
regridding method? Thanks for the excellent analysis below.

-----Original Message-----
From: John Halley Gotway via RT <met_help at ucar.edu>
Sent: Monday, March 30, 2020 4:27 PM
To: Serra, Mr. Efren, Contractor, Code 7531
<efren.serra.ctr at nrlmry.navy.mil>
Subject: Re: [rt.rap.ucar.edu #94778] GridStatConfig

Hello Efren,

I see you are looking for advice when regridding your 0.25 degree
model data to the 1.0 degree analysis grid.

I think I understand your logic.  Each 1.0 degree grid box consists of
4
0.25 degree points in both directions... so 16 points.  So you're
using a
4x4 with the distance-weighted mean method.  However, I suspect that
at the
1.0 degree grid points are really just a subset of the 0.25 degree
grid.
So for each 1.0 degree grid point, there is a 0.25 degree grid point
with the exact same lat/lon.  The "distance" between the lat/lon's of
the source and destination grids would be 0.  In the distance-weighted
mean logic, a distance of 0 essentially means an infinite weight.  So
I suspect using DW_MEAN in this way would yield the same result as
nearest neighbor (NEAREST).

To test this, I started with a 0.25 degree GFS field of 500mb
temperature.
I regridded it to NCEP grid 3, a 1.0 degree global grid.  But I did it
twice, once using 4x4 DW_MEAN and once using the NEAREST:

*regrid_data_plane gfs.pgrb20p25 G003 gfs.1p0_DW_MEAN.nc
<http://gfs.1p0_DW_MEAN.nc> -field 'name="TMP"; level="P500";' -method
DW_MEAN -width 4 -vld_thresh 0.5* *regrid_data_plane gfs.pgrb20p25
G003 gfs.1p0_NEAREST.nc <http://gfs.1p0_NEAREST.nc> -field
'name="TMP"; level="P500";' -method NEAREST -width 1 -vld_thresh 0.5*

Next, I ran pcp_combine to compute their difference:

*pcp_combine -subtract gfs.1p0_DW_MEAN.nc <http://gfs.1p0_DW_MEAN.nc>
gfs.1p0_NEAREST.nc <http://gfs.1p0_NEAREST.nc> diff.nc
<http://diff.nc> -field 'name="TMP_P500"; level="(*,*)";'*

The resulting difference field is all 0's.  If you really want to have
all of the 16 nearest points in the interpolation, I'd recommend using

*regrid_data_plane gfs.pgrb20p25 G003 gfs.1p0_UW_MEAN.nc
<http://gfs.1p0_UW_MEAN.nc> -field 'name="TMP"; level="P500";' -method
UW_MEAN -width 4 -vld_thresh 0.5*

I reran pcp_combine to diff this output with the nearest neighbor
output and the result is slightly different.  Ultimately, it's up to
you.  There's no right or wrong here.  Hope this helps clarify what's
going on.

Thanks,
John Halley Gotway

On Mon, Mar 30, 2020 at 3:08 PM efren.serra.ctr at nrlmry.navy.mil via RT
< met_help at ucar.edu> wrote:

>
> Mon Mar 30 15:08:24 2020: Request 94778 was acted upon.
> Transaction: Ticket created by efren.serra.ctr at nrlmry.navy.mil
>        Queue: met_help
>      Subject: GridStatConfig
>        Owner: Nobody
>   Requestors: efren.serra.ctr at nrlmry.navy.mil
>       Status: new
>  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=94778
> >
>
>
> Folks - I have a question regarding the regrid = {} object in the
> GridStatConfig file; I'm currently setting it to:
>
> regrid = {
>    to_grid    = OBS;
>    method     = DW_MEAN;
>    width      = 4;
>    vld_thresh = 0.5;
>    shape      = SQUARE;
> }
>
> The obs field is 1.0 deg and the forecast probability field is 0.25
> deg; are these values the best ones? I'm also creating my own box
> of 20x20 degrees positioned at the tropical cyclone center (lon,lat)
> point. Thank you.
>
> Efren A. Serra (Contractor)
> Physicist
>
> DeVine Consulting, Inc.
> Naval Research Laboratory
> Marine Meteorology Division
> 7 Grace Hopper Ave., STOP 2
> Monterey, CA 93943
> Code 7542
> Mobile: 408-425-5027
>
>
>

------------------------------------------------
Subject: GridStatConfig
From: John Halley Gotway
Time: Mon Mar 30 17:44:53 2020

Efren,

I just want to make sure you're clear on the difference between
regridding
and masking.  Grid-Stat computes its statistics on some grid, and
that's
controlled by the "regrid" dictionary.  You've chosen to use a 1.0
degree
analysis grid.  You can define as many subdomains (i.e. masks) of that
verification grid as you'd like.  And that's controlled by the "mask"
dictionary.

It sounds like you're interested in a 20x20 degree box centered on a
tropical cyclone location.  What isn't clear to me is whether that
region is defined on your 1.0 degree global grid... or defined as a
grid
itself.  You could choose to do either.

The only tricky part for the latter would be coming up with the grid
definition and specifying it in the Grid-Stat config file.  But you
certainly could.

Just search in this README file for the "to_grid" options:

John

On Mon, Mar 30, 2020 at 5:29 PM efren.serra.ctr at nrlmry.navy.mil via RT
<
met_help at ucar.edu> wrote:

>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=94778 >
>
> John - How about if I just regrid to the mask file using the MAX
> regridding method? Thanks for the excellent analysis below.
>
> -----Original Message-----
> From: John Halley Gotway via RT <met_help at ucar.edu>
> Sent: Monday, March 30, 2020 4:27 PM
> To: Serra, Mr. Efren, Contractor, Code 7531 <
> efren.serra.ctr at nrlmry.navy.mil>
> Subject: Re: [rt.rap.ucar.edu #94778] GridStatConfig
>
> Hello Efren,
>
> I see you are looking for advice when regridding your 0.25 degree
model
> data to the 1.0 degree analysis grid.
>
> I think I understand your logic.  Each 1.0 degree grid box consists
of 4
> 0.25 degree points in both directions... so 16 points.  So you're
using a
> 4x4 with the distance-weighted mean method.  However, I suspect that
at the
> 1.0 degree grid points are really just a subset of the 0.25 degree
grid.
> So for each 1.0 degree grid point, there is a 0.25 degree grid point
with
> the exact same lat/lon.  The "distance" between the lat/lon's of the
source
> and destination grids would be 0.  In the distance-weighted mean
logic, a
> distance of 0 essentially means an infinite weight.  So I suspect
using
> DW_MEAN in this way would yield the same result as nearest neighbor
> (NEAREST).
>
> To test this, I started with a 0.25 degree GFS field of 500mb
temperature.
> I regridded it to NCEP grid 3, a 1.0 degree global grid.  But I did
it
> twice, once using 4x4 DW_MEAN and once using the NEAREST:
>
> *regrid_data_plane gfs.pgrb20p25 G003 gfs.1p0_DW_MEAN.nc <
> http://gfs.1p0_DW_MEAN.nc> -field 'name="TMP"; level="P500";'
-method
> DW_MEAN -width 4 -vld_thresh 0.5* *regrid_data_plane gfs.pgrb20p25
G003
> gfs.1p0_NEAREST.nc <http://gfs.1p0_NEAREST.nc> -field 'name="TMP";
> level="P500";' -method NEAREST -width 1 -vld_thresh 0.5*
>
> Next, I ran pcp_combine to compute their difference:
>
> *pcp_combine -subtract gfs.1p0_DW_MEAN.nc
<http://gfs.1p0_DW_MEAN.nc>
> gfs.1p0_NEAREST.nc <http://gfs.1p0_NEAREST.nc> diff.nc
<http://diff.nc>
> -field 'name="TMP_P500"; level="(*,*)";'*
>
> The resulting difference field is all 0's.  If you really want to
have all
> of the 16 nearest points in the interpolation, I'd recommend using
the
>
> *regrid_data_plane gfs.pgrb20p25 G003 gfs.1p0_UW_MEAN.nc <
> http://gfs.1p0_UW_MEAN.nc> -field 'name="TMP"; level="P500";'
-method
> UW_MEAN -width 4 -vld_thresh 0.5*
>
> I reran pcp_combine to diff this output with the nearest neighbor
output
> and the result is slightly different.  Ultimately, it's up to you.
There's
> no right or wrong here.  Hope this helps clarify what's going on.
>
> Thanks,
> John Halley Gotway
>
> On Mon, Mar 30, 2020 at 3:08 PM efren.serra.ctr at nrlmry.navy.mil via
RT <
> met_help at ucar.edu> wrote:
>
> >
> > Mon Mar 30 15:08:24 2020: Request 94778 was acted upon.
> > Transaction: Ticket created by efren.serra.ctr at nrlmry.navy.mil
> >        Queue: met_help
> >      Subject: GridStatConfig
> >        Owner: Nobody
> >   Requestors: efren.serra.ctr at nrlmry.navy.mil
> >       Status: new
> >  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=94778
> > >
> >
> >
> > Folks - I have a question regarding the regrid = {} object in the
> > GridStatConfig file; I'm currently setting it to:
> >
> > regrid = {
> >    to_grid    = OBS;
> >    method     = DW_MEAN;
> >    width      = 4;
> >    vld_thresh = 0.5;
> >    shape      = SQUARE;
> > }
> >
> > The obs field is 1.0 deg and the forecast probability field is
0.25
> > deg; are these values the best ones? I'm also creating my own box
> > of 20x20 degrees positioned at the tropical cyclone center
(lon,lat)
> > point. Thank you.
> >
> > Efren A. Serra (Contractor)
> > Physicist
> >
> > DeVine Consulting, Inc.
> > Naval Research Laboratory
> > Marine Meteorology Division
> > 7 Grace Hopper Ave., STOP 2
> > Monterey, CA 93943
> > Code 7542
> > Mobile: 408-425-5027
> >
> >
> >
>
>
>
>

------------------------------------------------
Subject: RE: [rt.rap.ucar.edu #94778] GridStatConfig
From: efren.serra.ctr at nrlmry.navy.mil
Time: Mon Mar 30 18:06:38 2020

John

This is the context:
OBS ] (360 x 181) and (1440 x 721)
FCST] (360 x 181), (720 x 181) and (1440 x 721)

Hence, from your first email, I should go from higher resolution to
lower resolution and use the UW_MEAN "method" with the remaining
values as you indicated therein. Thanks.

-----Original Message-----
From: John Halley Gotway via RT <met_help at ucar.edu>
Sent: Monday, March 30, 2020 4:45 PM
To: Serra, Mr. Efren, Contractor, Code 7531
<efren.serra.ctr at nrlmry.navy.mil>
Subject: Re: [rt.rap.ucar.edu #94778] GridStatConfig

Efren,

I just want to make sure you're clear on the difference between
regridding and masking.  Grid-Stat computes its statistics on some
grid, and that's controlled by the "regrid" dictionary.  You've chosen
to use a 1.0 degree analysis grid.  You can define as many subdomains
(i.e. masks) of that verification grid as you'd like.  And that's
dictionary.

It sounds like you're interested in a 20x20 degree box centered on a
tropical cyclone location.  What isn't clear to me is whether that
masking region is defined on your 1.0 degree global grid... or defined
as a grid itself.  You could choose to do either.

The only tricky part for the latter would be coming up with the grid
definition and specifying it in the Grid-Stat config file.  But you
certainly could.

Just search in this README file for the "to_grid" options:

John

On Mon, Mar 30, 2020 at 5:29 PM efren.serra.ctr at nrlmry.navy.mil via RT
< met_help at ucar.edu> wrote:

>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=94778 >
>
> John - How about if I just regrid to the mask file using the MAX
> regridding method? Thanks for the excellent analysis below.
>
> -----Original Message-----
> From: John Halley Gotway via RT <met_help at ucar.edu>
> Sent: Monday, March 30, 2020 4:27 PM
> To: Serra, Mr. Efren, Contractor, Code 7531 <
> efren.serra.ctr at nrlmry.navy.mil>
> Subject: Re: [rt.rap.ucar.edu #94778] GridStatConfig
>
> Hello Efren,
>
> I see you are looking for advice when regridding your 0.25 degree
> model data to the 1.0 degree analysis grid.
>
> I think I understand your logic.  Each 1.0 degree grid box consists
of
> 4
> 0.25 degree points in both directions... so 16 points.  So you're
> using a
> 4x4 with the distance-weighted mean method.  However, I suspect that
> at the
> 1.0 degree grid points are really just a subset of the 0.25 degree
grid.
> So for each 1.0 degree grid point, there is a 0.25 degree grid point
> with the exact same lat/lon.  The "distance" between the lat/lon's
of
> the source and destination grids would be 0.  In the distance-
weighted
> mean logic, a distance of 0 essentially means an infinite weight.
So
> I suspect using DW_MEAN in this way would yield the same result as
> nearest neighbor (NEAREST).
>
> To test this, I started with a 0.25 degree GFS field of 500mb
temperature.
> I regridded it to NCEP grid 3, a 1.0 degree global grid.  But I did
it
> twice, once using 4x4 DW_MEAN and once using the NEAREST:
>
> *regrid_data_plane gfs.pgrb20p25 G003 gfs.1p0_DW_MEAN.nc <
> http://gfs.1p0_DW_MEAN.nc> -field 'name="TMP"; level="P500";'
-method
> DW_MEAN -width 4 -vld_thresh 0.5* *regrid_data_plane gfs.pgrb20p25
> G003 gfs.1p0_NEAREST.nc <http://gfs.1p0_NEAREST.nc> -field
> 'name="TMP"; level="P500";' -method NEAREST -width 1 -vld_thresh
0.5*
>
> Next, I ran pcp_combine to compute their difference:
>
> *pcp_combine -subtract gfs.1p0_DW_MEAN.nc
<http://gfs.1p0_DW_MEAN.nc>
> gfs.1p0_NEAREST.nc <http://gfs.1p0_NEAREST.nc> diff.nc
> <http://diff.nc> -field 'name="TMP_P500"; level="(*,*)";'*
>
> The resulting difference field is all 0's.  If you really want to
have
> all of the 16 nearest points in the interpolation, I'd recommend
using
> the unweighted-mean method (UW_MEAN) instead.
>
> *regrid_data_plane gfs.pgrb20p25 G003 gfs.1p0_UW_MEAN.nc <
> http://gfs.1p0_UW_MEAN.nc> -field 'name="TMP"; level="P500";'
-method
> UW_MEAN -width 4 -vld_thresh 0.5*
>
> I reran pcp_combine to diff this output with the nearest neighbor
> output and the result is slightly different.  Ultimately, it's up to
> you.  There's no right or wrong here.  Hope this helps clarify
what's going on.
>
> Thanks,
> John Halley Gotway
>
> On Mon, Mar 30, 2020 at 3:08 PM efren.serra.ctr at nrlmry.navy.mil via
RT
> < met_help at ucar.edu> wrote:
>
> >
> > Mon Mar 30 15:08:24 2020: Request 94778 was acted upon.
> > Transaction: Ticket created by efren.serra.ctr at nrlmry.navy.mil
> >        Queue: met_help
> >      Subject: GridStatConfig
> >        Owner: Nobody
> >   Requestors: efren.serra.ctr at nrlmry.navy.mil
> >       Status: new
> >  Ticket <URL:
> > https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=94778
> > >
> >
> >
> > Folks - I have a question regarding the regrid = {} object in the
> > GridStatConfig file; I'm currently setting it to:
> >
> > regrid = {
> >    to_grid    = OBS;
> >    method     = DW_MEAN;
> >    width      = 4;
> >    vld_thresh = 0.5;
> >    shape      = SQUARE;
> > }
> >
> > The obs field is 1.0 deg and the forecast probability field is
0.25
> > deg; are these values the best ones? I'm also creating my own box
> > mask of 20x20 degrees positioned at the tropical cyclone center
> > (lon,lat) point. Thank you.
> >
> > Efren A. Serra (Contractor)
> > Physicist
> >
> > DeVine Consulting, Inc.
> > Naval Research Laboratory
> > Marine Meteorology Division
> > 7 Grace Hopper Ave., STOP 2
> > Monterey, CA 93943
> > Code 7542
> > Mobile: 408-425-5027
> >
> >
> >
>
>
>
>

------------------------------------------------
```