[Met_help] [rt.rap.ucar.edu #74813] History for Regrid Options
John Halley Gotway via RT
met_help at ucar.edu
Wed Jan 27 12:02:23 MST 2016
----------------------------------------------------------------
Initial Request
----------------------------------------------------------------
John, I am using the regrid option in grid_stat to remap my observation grid (Stage IV) to a specific FCST grid. In the documentation it says you can specify the grid definition in the to_grid parameter but it doesn't say the format to use. I assume it is something like this 255 3 342 210 20586 -122243 8 -96000 15000 15000 0 64 60000 30000 but I don't know the specific syntax to use?
Also, science question. I want to generate FSS stats for the output of our 4km ensemble compared to our 20km ensemble. The obs data is Stage IV. What is the smartest way to do this - should I remap the 20km FCST to the 4km FCST grid along with the OBS or should I remap the 20km and 4km to the OBS (Stage IV grid)?
Thanks
----------------------------------------------------------------
Complete Ticket History
----------------------------------------------------------------
Subject: Regrid Options
From: John Halley Gotway
Time: Thu Jan 21 10:23:25 2016
Bob,
I see that you're running Grid-Stat and are trying to use the
regridding
option included in the met-5.1 release. I think this is simpler than
you
expect. You don't need to explicitly specify the grid definition
yourself. Instead, all you need to put in the config file is
something
like this:
//
// Verification grid
//
regrid = {
to_grid = OBS; // Or use "FCST" to verify on the forecast grid.
vld_thresh = 0.5;
method = NEAREST;
width = 1;
}
Grid-Stat reads the forecast and observation files, extracts their
grid
information, and checks to see if they're the same. If not, it looks
in
the "to_grid" setting in the "regrid" section of the config file. If
it
says "FCST", it regrids the observation data onto the forecast grid.
If is
says "OBS", it puts the forecast on the observation grid. And there
are
some more options listed on page 46 of the User's Guide:
http://www.dtcenter.org/met/users/docs/users_guide/MET_Users_Guide_v5.1.pdf
FYI, the default regridding method is just nearest neighbor, as shown
above. You might want to use "BUDGET" for budget interpolation when
using
precip. Budget interpolation is designed to preserve mass better than
other methods, which is generally recommended for precip.
Lastly, I see you've asked the age-old question... regrid to the
higher-res
or lower-res domain? I don't have a good answer for you. People do
it
both ways for different reasons. Some folks prefer to evaluate on the
model domain, others prefer to use the observation domain so as not
to corrupt the obs with interpolation error. My guess is that more
people
would say evaluate on the 20km rather than the 4km because it's better
to
throw data points away than to create new ones through interpolation.
But
people do it both ways.
Hope that helps.
Thanks
John
On Thu, Jan 21, 2016 at 9:26 AM, robert.craig.2 at us.af.mil
<javascript:_e(%7B%7D,'cvml','robert.craig.2 at us.af.mil');> via RT <
met_help at ucar.edu <javascript:_e(%7B%7D,'cvml','met_help at ucar.edu');>>
wrote:
>
> Thu Jan 21 09:26:14 2016: Request 74813 was acted upon.
> Transaction: Ticket created by robert.craig.2 at us.af.mil
> <javascript:_e(%7B%7D,'cvml','robert.craig.2 at us.af.mil');>
> Queue: met_help
> Subject: Regrid Options
> Owner: Nobody
> Requestors: robert.craig.2 at us.af.mil
> <javascript:_e(%7B%7D,'cvml','robert.craig.2 at us.af.mil');>
> Status: new
> Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=74813 >
>
>
> John, I am using the regrid option in grid_stat to remap my
observation
> grid (Stage IV) to a specific FCST grid. In the documentation it
says you
> can specify the grid definition in the to_grid parameter but it
doesn't say
> the format to use. I assume it is something like this 255 3 342
210 20586
> -122243 8 -96000 15000 15000 0 64 60000 30000 but I don't know the
specific
> syntax to use?
>
> Also, science question. I want to generate FSS stats for the output
of
> our 4km ensemble compared to our 20km ensemble. The obs data is
Stage
> IV. What is the smartest way to do this - should I remap the 20km
FCST to
> the 4km FCST grid along with the OBS or should I remap the 20km and
4km to
> the OBS (Stage IV grid)?
>
> Thanks
>
>
>
------------------------------------------------
Subject: RE: [rt.rap.ucar.edu #74813] Regrid Options
From: robert.craig.2 at us.af.mil
Time: Thu Jan 21 10:57:12 2016
Thanks John, for some reason I missed page 46. The regrid option is
nice - makes it much easier.
Bob
-----Original Message-----
From: John Halley Gotway via RT [mailto:met_help at ucar.edu]
Sent: Thursday, January 21, 2016 11:23 AM
To: CRAIG, ROBERT J GS-12 USAF ACC 16 WS/WXN
Subject: [rt.rap.ucar.edu #74813] Regrid Options
Bob,
I see that you're running Grid-Stat and are trying to use the
regridding option included in the met-5.1 release. I think this is
simpler than you expect. You don't need to explicitly specify the
grid definition yourself. Instead, all you need to put in the config
file is something like this:
//
// Verification grid
//
regrid = {
to_grid = OBS; // Or use "FCST" to verify on the forecast grid.
vld_thresh = 0.5;
method = NEAREST;
width = 1;
}
Grid-Stat reads the forecast and observation files, extracts their
grid information, and checks to see if they're the same. If not, it
looks in the "to_grid" setting in the "regrid" section of the config
file. If it says "FCST", it regrids the observation data onto the
forecast grid. If is says "OBS", it puts the forecast on the
observation grid. And there are some more options listed on page 46
of the User's Guide:
http://www.dtcenter.org/met/users/docs/users_guide/MET_Users_Guide_v5.1.pdf
FYI, the default regridding method is just nearest neighbor, as shown
above. You might want to use "BUDGET" for budget interpolation when
using precip. Budget interpolation is designed to preserve mass
better than other methods, which is generally recommended for precip.
Lastly, I see you've asked the age-old question... regrid to the
higher-res or lower-res domain? I don't have a good answer for you.
People do it both ways for different reasons. Some folks prefer to
evaluate on the model domain, others prefer to use the observation
domain so as not to corrupt the obs with interpolation error. My
guess is that more people would say evaluate on the 20km rather than
the 4km because it's better to throw data points away than to create
new ones through interpolation. But people do it both ways.
Hope that helps.
Thanks
John
On Thu, Jan 21, 2016 at 9:26 AM, robert.craig.2 at us.af.mil
<javascript:_e(%7B%7D,'cvml','robert.craig.2 at us.af.mil');> via RT <
met_help at ucar.edu <javascript:_e(%7B%7D,'cvml','met_help at ucar.edu');>>
wrote:
>
> Thu Jan 21 09:26:14 2016: Request 74813 was acted upon.
> Transaction: Ticket created by robert.craig.2 at us.af.mil
> <javascript:_e(%7B%7D,'cvml','robert.craig.2 at us.af.mil');>
> Queue: met_help
> Subject: Regrid Options
> Owner: Nobody
> Requestors: robert.craig.2 at us.af.mil
> <javascript:_e(%7B%7D,'cvml','robert.craig.2 at us.af.mil');>
> Status: new
> Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=74813
> >
>
>
> John, I am using the regrid option in grid_stat to remap my
> observation grid (Stage IV) to a specific FCST grid. In the
> documentation it says you can specify the grid definition in the
to_grid parameter but it doesn't say
> the format to use. I assume it is something like this 255 3 342
210 20586
> -122243 8 -96000 15000 15000 0 64 60000 30000 but I don't know the
> specific syntax to use?
>
> Also, science question. I want to generate FSS stats for the output
> of our 4km ensemble compared to our 20km ensemble. The obs data
is
> Stage IV. What is the smartest way to do this - should I remap the
> 20km FCST to the 4km FCST grid along with the OBS or should I remap
> the 20km and 4km to the OBS (Stage IV grid)?
>
> Thanks
>
>
>
------------------------------------------------
Subject: RE: [rt.rap.ucar.edu #74813] Regrid Options
From: robert.craig.2 at us.af.mil
Time: Wed Jan 27 07:59:51 2016
John, With the regridding budget option, what is typically used for
the width?
Thanks
-----Original Message-----
From: John Halley Gotway via RT [mailto:met_help at ucar.edu]
Sent: Thursday, January 21, 2016 11:23 AM
To: CRAIG, ROBERT J GS-12 USAF ACC 16 WS/WXN
Subject: [rt.rap.ucar.edu #74813] Regrid Options
Bob,
I see that you're running Grid-Stat and are trying to use the
regridding option included in the met-5.1 release. I think this is
simpler than you expect. You don't need to explicitly specify the
grid definition yourself. Instead, all you need to put in the config
file is something like this:
//
// Verification grid
//
regrid = {
to_grid = OBS; // Or use "FCST" to verify on the forecast grid.
vld_thresh = 0.5;
method = NEAREST;
width = 1;
}
Grid-Stat reads the forecast and observation files, extracts their
grid information, and checks to see if they're the same. If not, it
looks in the "to_grid" setting in the "regrid" section of the config
file. If it says "FCST", it regrids the observation data onto the
forecast grid. If is says "OBS", it puts the forecast on the
observation grid. And there are some more options listed on page 46
of the User's Guide:
http://www.dtcenter.org/met/users/docs/users_guide/MET_Users_Guide_v5.1.pdf
FYI, the default regridding method is just nearest neighbor, as shown
above. You might want to use "BUDGET" for budget interpolation when
using precip. Budget interpolation is designed to preserve mass
better than other methods, which is generally recommended for precip.
Lastly, I see you've asked the age-old question... regrid to the
higher-res or lower-res domain? I don't have a good answer for you.
People do it both ways for different reasons. Some folks prefer to
evaluate on the model domain, others prefer to use the observation
domain so as not to corrupt the obs with interpolation error. My
guess is that more people would say evaluate on the 20km rather than
the 4km because it's better to throw data points away than to create
new ones through interpolation. But people do it both ways.
Hope that helps.
Thanks
John
On Thu, Jan 21, 2016 at 9:26 AM, robert.craig.2 at us.af.mil
<javascript:_e(%7B%7D,'cvml','robert.craig.2 at us.af.mil');> via RT <
met_help at ucar.edu <javascript:_e(%7B%7D,'cvml','met_help at ucar.edu');>>
wrote:
>
> Thu Jan 21 09:26:14 2016: Request 74813 was acted upon.
> Transaction: Ticket created by robert.craig.2 at us.af.mil
> <javascript:_e(%7B%7D,'cvml','robert.craig.2 at us.af.mil');>
> Queue: met_help
> Subject: Regrid Options
> Owner: Nobody
> Requestors: robert.craig.2 at us.af.mil
> <javascript:_e(%7B%7D,'cvml','robert.craig.2 at us.af.mil');>
> Status: new
> Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=74813
> >
>
>
> John, I am using the regrid option in grid_stat to remap my
> observation grid (Stage IV) to a specific FCST grid. In the
> documentation it says you can specify the grid definition in the
to_grid parameter but it doesn't say
> the format to use. I assume it is something like this 255 3 342
210 20586
> -122243 8 -96000 15000 15000 0 64 60000 30000 but I don't know the
> specific syntax to use?
>
> Also, science question. I want to generate FSS stats for the output
> of our 4km ensemble compared to our 20km ensemble. The obs data
is
> Stage IV. What is the smartest way to do this - should I remap the
> 20km FCST to the 4km FCST grid along with the OBS or should I remap
> the 20km and 4km to the OBS (Stage IV grid)?
>
> Thanks
>
>
>
------------------------------------------------
Subject: Regrid Options
From: John Halley Gotway
Time: Wed Jan 27 09:37:26 2016
Bob,
You should use 2 as the width for the budget interpolation method. In
fact, if you try to use anything other than 2, you'll get an error
message
similar to this:
WARNING: process_command_line() -> Resetting the regridding width
from 6
to 2 for regridding method "BUDGET".
Budget interpolation is very similar to bilinear interpolation, which
also
can only be used with a width of 2. Bilinear interpolation performs
the
interpolation using the closest 4 points. In budget interpolation,
each
grid box is decomposed into a 5x5 mesh and bilinear interpolation is
performed for each of those 25 sub-grid points. The final
interpolated
value for that grid box is just average of those 25 sub-grid points.
That
makes budget interpolation much slower, but (presumably) better able
to
conserve accumulated variables, like precip.
Hope that helps.
Thanks,
John
On Wed, Jan 27, 2016 at 7:59 AM, robert.craig.2 at us.af.mil via RT <
met_help at ucar.edu> wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=74813 >
>
> John, With the regridding budget option, what is typically used for
the
> width?
>
> Thanks
>
> -----Original Message-----
> From: John Halley Gotway via RT [mailto:met_help at ucar.edu]
> Sent: Thursday, January 21, 2016 11:23 AM
> To: CRAIG, ROBERT J GS-12 USAF ACC 16 WS/WXN
> Subject: [rt.rap.ucar.edu #74813] Regrid Options
>
> Bob,
>
> I see that you're running Grid-Stat and are trying to use the
regridding
> option included in the met-5.1 release. I think this is simpler
than you
> expect. You don't need to explicitly specify the grid definition
> yourself. Instead, all you need to put in the config file is
something
> like this:
>
> //
> // Verification grid
> //
> regrid = {
> to_grid = OBS; // Or use "FCST" to verify on the forecast
grid.
> vld_thresh = 0.5;
> method = NEAREST;
> width = 1;
> }
>
> Grid-Stat reads the forecast and observation files, extracts their
grid
> information, and checks to see if they're the same. If not, it
looks in
> the "to_grid" setting in the "regrid" section of the config file.
If it
> says "FCST", it regrids the observation data onto the forecast grid.
If is
> says "OBS", it puts the forecast on the observation grid. And there
are
> some more options listed on page 46 of the User's Guide:
>
>
http://www.dtcenter.org/met/users/docs/users_guide/MET_Users_Guide_v5.1.pdf
>
> FYI, the default regridding method is just nearest neighbor, as
shown
> above. You might want to use "BUDGET" for budget interpolation when
using
> precip. Budget interpolation is designed to preserve mass better
than
> other methods, which is generally recommended for precip.
>
> Lastly, I see you've asked the age-old question... regrid to the
> higher-res or lower-res domain? I don't have a good answer for you.
> People do it both ways for different reasons. Some folks prefer to
> evaluate on the model domain, others prefer to use the observation
domain
> so as not to corrupt the obs with interpolation error. My guess is
that
> more people would say evaluate on the 20km rather than the 4km
because it's
> better to throw data points away than to create new ones through
> interpolation. But people do it both ways.
>
> Hope that helps.
>
> Thanks
> John
>
> On Thu, Jan 21, 2016 at 9:26 AM, robert.craig.2 at us.af.mil
> <javascript:_e(%7B%7D,'cvml','robert.craig.2 at us.af.mil');> via RT <
> met_help at ucar.edu
<javascript:_e(%7B%7D,'cvml','met_help at ucar.edu');>>
> wrote:
>
> >
> > Thu Jan 21 09:26:14 2016: Request 74813 was acted upon.
> > Transaction: Ticket created by robert.craig.2 at us.af.mil
> > <javascript:_e(%7B%7D,'cvml','robert.craig.2 at us.af.mil');>
> > Queue: met_help
> > Subject: Regrid Options
> > Owner: Nobody
> > Requestors: robert.craig.2 at us.af.mil
> > <javascript:_e(%7B%7D,'cvml','robert.craig.2 at us.af.mil');>
> > Status: new
> > Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=74813
> > >
> >
> >
> > John, I am using the regrid option in grid_stat to remap my
> > observation grid (Stage IV) to a specific FCST grid. In the
> > documentation it says you can specify the grid definition in the
to_grid
> parameter but it doesn't say
> > the format to use. I assume it is something like this 255 3 342
210
> 20586
> > -122243 8 -96000 15000 15000 0 64 60000 30000 but I don't know the
> > specific syntax to use?
> >
> > Also, science question. I want to generate FSS stats for the
output
> > of our 4km ensemble compared to our 20km ensemble. The obs data
is
> > Stage IV. What is the smartest way to do this - should I remap
the
> > 20km FCST to the 4km FCST grid along with the OBS or should I
remap
> > the 20km and 4km to the OBS (Stage IV grid)?
> >
> > Thanks
> >
> >
> >
>
>
>
>
------------------------------------------------
Subject: RE: [rt.rap.ucar.edu #74813] Regrid Options
From: robert.craig.2 at us.af.mil
Time: Wed Jan 27 10:30:59 2016
That helps. Below is a listing of my run. Note the warning which was
explained by you. Down several lines is says interpolation method
NEAREST(1). Should this also say BUDGET?
Thanks
DEBUG 1: Default Config File: /home/qcteam/met-
5.1/share/met/config/GridStatConfig_default
DEBUG 1: User Config File:
/h/data/global/WXQC/data/fss/cfg/GridStatConfig_pop_updated
WARNING:
WARNING: parse_conf_regrid() -> Resetting the regridding width from 4
to 2 for regridding method "BUDGET".
WARNING:
GSL_RNG_TYPE=mt19937
GSL_RNG_SEED=18446744072165042262
DEBUG 1: Forecast File:
/h/data/global/WXQC/data/temp/grib.2016011500.0072
DEBUG 1: Observation File:
/h/data/global/WXQC/data/fss/pcpcombout/2016011800/rgST4.2016011800.06h.nc
DEBUG 2:
DEBUG 2:
--------------------------------------------------------------------------------
DEBUG 2:
DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
method NEAREST(1), over region FULL, using 402704 pairs.
DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
method NBRHD(1), raw thresholds of >0 and >0, over region FULL, using
402704 pairs.
DEBUG 2: Computing Neighborhood Continuous Statistics.
DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
method NBRHD(9), raw thresholds of >0 and >0, over region FULL, using
393481 pairs.
DEBUG 2: Computing Neighborhood Continuous Statistics.
DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
method NBRHD(25), raw thresholds of >0 and >0, over region FULL, using
384982 pairs.
DEBUG 2: Computing Neighborhood Continuous Statistics.
DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
method NBRHD(81), raw thresholds of >0 and >0, over region FULL, using
368893 pairs.
DEBUG 2: Computing Neighborhood Continuous Statistics.
DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
method NBRHD(121), raw thresholds of >0 and >0, over region FULL,
using 361134 pairs.
DEBUG 2: Computing Neighborhood Continuous Statistics.
DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
method NBRHD(169), raw thresholds of >0 and >0, over region FULL,
using 353530 pairs.
DEBUG 2: Computing Neighborhood Continuous Statistics.
DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
method NBRHD(225), raw thresholds of >0 and >0, over region FULL,
using 346075 pairs.
DEBUG 2: Computing Neighborhood Continuous Statistics.
DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
method NBRHD(289), raw thresholds of >0 and >0, over region FULL,
using 338782 pairs.
DEBUG 2: Computing Neighborhood Continuous Statistics.
DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
method NBRHD(361), raw thresholds of >0 and >0, over region FULL,
using 331650 pairs.
DEBUG 2: Computing Neighborhood Continuous Statistics.
DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
method NBRHD(441), raw thresholds of >0 and >0, over region FULL,
using 324655 pairs.
DEBUG 2: Computing Neighborhood Continuous Statistics.
DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
method NBRHD(529), raw thresholds of >0 and >0, over region FULL,
using 317763 pairs.
DEBUG 2: Computing Neighborhood Continuous Statistics.
DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
method NBRHD(625), raw thresholds of >0 and >0, over region FULL,
using 311011 pairs.
DEBUG 2: Computing Neighborhood Continuous Statistics.
DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
method NBRHD(729), raw thresholds of >0 and >0, over region FULL,
using 304407 pairs.
DEBUG 2: Computing Neighborhood Continuous Statistics.
DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
method NBRHD(841), raw thresholds of >0 and >0, over region FULL,
using 297911 pairs.
DEBUG 2: Computing Neighborhood Continuous Statistics.
DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
method NBRHD(961), raw thresholds of >0 and >0, over region FULL,
using 291521 pairs.
DEBUG 2: Computing Neighborhood Continuous Statistics.
DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
method NBRHD(1089), raw thresholds of >0 and >0, over region FULL,
using 285215 pairs.
DEBUG 2: Computing Neighborhood Continuous Statistics.
DEBUG 2:
DEBUG 2:
--------------------------------------------------------------------------------
DEBUG 2:
DEBUG 1: Output file:
/h/data/global/WXQC/data/fss/metoutput_fss/2016011500/2016011500.0072/grid_stat_50_APCP_6_720000L_20160118_000000V.stat
DEBUG 1: Output file:
/h/data/global/WXQC/data/fss/metoutput_fss/2016011500/2016011500.0072/grid_stat_50_APCP_6_720000L_20160118_000000V_pairs.nc
1.0 QP100
['/h/WXQC/met-5.1/bin/grid_stat',
'/h/data/global/WXQC/data/temp/grib.2016011500.0072',
'/h/data/global/WXQC/data/fss/pcpcombout/2016011800/rgST4.2016011800.06h.nc',
'/h/data/global/WXQC/data/fss/cfg/GridStatConfig_pop_updated', '-
outdir',
'/h/data/global/WXQC/data/fss/metoutput_fss/2016011500/2016011500.0072']
DEBUG 1: Default Config File: /home/qcteam/met-
5.1/share/met/config/GridStatConfig_default
DEBUG 1: User Config File:
/h/data/global/WXQC/data/fss/cfg/GridStatConfig_pop_updated
WARNING:
-----Original Message-----
From: John Halley Gotway via RT [mailto:met_help at ucar.edu]
Sent: Wednesday, January 27, 2016 10:37 AM
To: CRAIG, ROBERT J GS-12 USAF ACC 16 WS/WXN
Subject: Re: [rt.rap.ucar.edu #74813] Regrid Options
Bob,
You should use 2 as the width for the budget interpolation method. In
fact, if you try to use anything other than 2, you'll get an error
message similar to this:
WARNING: process_command_line() -> Resetting the regridding width
from 6 to 2 for regridding method "BUDGET".
Budget interpolation is very similar to bilinear interpolation, which
also can only be used with a width of 2. Bilinear interpolation
performs the interpolation using the closest 4 points. In budget
interpolation, each grid box is decomposed into a 5x5 mesh and
bilinear interpolation is performed for each of those 25 sub-grid
points. The final interpolated value for that grid box is just
average of those 25 sub-grid points. That makes budget interpolation
much slower, but (presumably) better able to conserve accumulated
variables, like precip.
Hope that helps.
Thanks,
John
On Wed, Jan 27, 2016 at 7:59 AM, robert.craig.2 at us.af.mil via RT <
met_help at ucar.edu> wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=74813 >
>
> John, With the regridding budget option, what is typically used for
> the width?
>
> Thanks
>
> -----Original Message-----
> From: John Halley Gotway via RT [mailto:met_help at ucar.edu]
> Sent: Thursday, January 21, 2016 11:23 AM
> To: CRAIG, ROBERT J GS-12 USAF ACC 16 WS/WXN
> Subject: [rt.rap.ucar.edu #74813] Regrid Options
>
> Bob,
>
> I see that you're running Grid-Stat and are trying to use the
> regridding option included in the met-5.1 release. I think this is
> simpler than you expect. You don't need to explicitly specify the
> grid definition yourself. Instead, all you need to put in the
config
> file is something like this:
>
> //
> // Verification grid
> //
> regrid = {
> to_grid = OBS; // Or use "FCST" to verify on the forecast
grid.
> vld_thresh = 0.5;
> method = NEAREST;
> width = 1;
> }
>
> Grid-Stat reads the forecast and observation files, extracts their
> grid information, and checks to see if they're the same. If not, it
> looks in the "to_grid" setting in the "regrid" section of the config
> file. If it says "FCST", it regrids the observation data onto the
> forecast grid. If is says "OBS", it puts the forecast on the
> observation grid. And there are some more options listed on page 46
of the User's Guide:
>
>
http://www.dtcenter.org/met/users/docs/users_guide/MET_Users_Guide_v5.
> 1.pdf
>
> FYI, the default regridding method is just nearest neighbor, as
shown
> above. You might want to use "BUDGET" for budget interpolation when
> using precip. Budget interpolation is designed to preserve mass
> better than other methods, which is generally recommended for
precip.
>
> Lastly, I see you've asked the age-old question... regrid to the
> higher-res or lower-res domain? I don't have a good answer for you.
> People do it both ways for different reasons. Some folks prefer to
> evaluate on the model domain, others prefer to use the observation
> domain so as not to corrupt the obs with interpolation error. My
> guess is that more people would say evaluate on the 20km rather than
> the 4km because it's better to throw data points away than to create
> new ones through interpolation. But people do it both ways.
>
> Hope that helps.
>
> Thanks
> John
>
> On Thu, Jan 21, 2016 at 9:26 AM, robert.craig.2 at us.af.mil
> <javascript:_e(%7B%7D,'cvml','robert.craig.2 at us.af.mil');> via RT <
> met_help at ucar.edu
<javascript:_e(%7B%7D,'cvml','met_help at ucar.edu');>>
> wrote:
>
> >
> > Thu Jan 21 09:26:14 2016: Request 74813 was acted upon.
> > Transaction: Ticket created by robert.craig.2 at us.af.mil
> > <javascript:_e(%7B%7D,'cvml','robert.craig.2 at us.af.mil');>
> > Queue: met_help
> > Subject: Regrid Options
> > Owner: Nobody
> > Requestors: robert.craig.2 at us.af.mil
> > <javascript:_e(%7B%7D,'cvml','robert.craig.2 at us.af.mil');>
> > Status: new
> > Ticket <URL:
> > https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=74813
> > >
> >
> >
> > John, I am using the regrid option in grid_stat to remap my
> > observation grid (Stage IV) to a specific FCST grid. In the
> > documentation it says you can specify the grid definition in the
> > to_grid
> parameter but it doesn't say
> > the format to use. I assume it is something like this 255 3 342
210
> 20586
> > -122243 8 -96000 15000 15000 0 64 60000 30000 but I don't know the
> > specific syntax to use?
> >
> > Also, science question. I want to generate FSS stats for the
output
> > of our 4km ensemble compared to our 20km ensemble. The obs data
> > is Stage IV. What is the smartest way to do this - should I remap
> > the 20km FCST to the 4km FCST grid along with the OBS or should I
> > remap the 20km and 4km to the OBS (Stage IV grid)?
> >
> > Thanks
> >
> >
> >
>
>
>
>
------------------------------------------------
Subject: Regrid Options
From: John Halley Gotway
Time: Wed Jan 27 11:10:15 2016
Bob,
Thanks for send me that output from Grid-Stat. This help me
understand
what you're seeing.
Let me start by saying this all looks fine to me. Here's why...
You're passing Grid-Stat two input files that are on different
domains.
Grid-Stat recognizes that the grids differ and uses the logic defined
in
the "regrid" section of the configuration file to put them onto a
common
grid. In there, you've chosen the "BUDGET" interpolation method with
a
width of 4. The warning message explains that for BUDGET
interpolation the
width must be set to 2, and it resets that for you. Increasing the
verbosity level (-v 3 or -v 4) may print out a message about the
automated
regridding that was done.
The forecast and observation data are now on a common domain. Looking
in
the Grid-Stat configuration file, you'll also see a section named
"interp"
which includes settings for interpolation methods and widths. The
name
"interp" is a little confusing... perhaps we should have named it
"smooth"
instead because really it controls if/how the forecast and/or
observation
data is smoothed prior to computing statistics. We kept the name
"interp"
to be consistent with similar options in the Point-Stat tool.
When doing verification, some folks apply upscaling (or smoothing) to
check
how smoothing their data effects the verification scores. And that
can be
done using the "interp" section of the config file. However, users
typically do not want to smooth their data prior to computing stats,
so the
default is to perform no smoothing. Nearest neighbor with a width of
1
means "no smoothing".
And that's the explanation for the string "for interpolation method
NEAREST(1)". All it's basically saying is that all grid points are
being
compared with no smoothing applied.
Hopefully that helps.
Thanks,
John
On Wed, Jan 27, 2016 at 10:30 AM, robert.craig.2 at us.af.mil via RT <
met_help at ucar.edu> wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=74813 >
>
> That helps. Below is a listing of my run. Note the warning which
was
> explained by you. Down several lines is says interpolation method
> NEAREST(1). Should this also say BUDGET?
>
> Thanks
>
> DEBUG 1: Default Config File:
> /home/qcteam/met-5.1/share/met/config/GridStatConfig_default
> DEBUG 1: User Config File:
> /h/data/global/WXQC/data/fss/cfg/GridStatConfig_pop_updated
> WARNING:
> WARNING: parse_conf_regrid() -> Resetting the regridding width from
4 to 2
> for regridding method "BUDGET".
> WARNING:
> GSL_RNG_TYPE=mt19937
> GSL_RNG_SEED=18446744072165042262
> DEBUG 1: Forecast File:
/h/data/global/WXQC/data/temp/grib.2016011500.0072
> DEBUG 1: Observation File:
>
/h/data/global/WXQC/data/fss/pcpcombout/2016011800/rgST4.2016011800.06h.nc
> DEBUG 2:
> DEBUG 2:
>
--------------------------------------------------------------------------------
> DEBUG 2:
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NEAREST(1), over region FULL, using 402704 pairs.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(1), raw thresholds of >0 and >0, over region FULL,
using
> 402704 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(9), raw thresholds of >0 and >0, over region FULL,
using
> 393481 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(25), raw thresholds of >0 and >0, over region FULL,
using
> 384982 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(81), raw thresholds of >0 and >0, over region FULL,
using
> 368893 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(121), raw thresholds of >0 and >0, over region FULL,
using
> 361134 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(169), raw thresholds of >0 and >0, over region FULL,
using
> 353530 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(225), raw thresholds of >0 and >0, over region FULL,
using
> 346075 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(289), raw thresholds of >0 and >0, over region FULL,
using
> 338782 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(361), raw thresholds of >0 and >0, over region FULL,
using
> 331650 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(441), raw thresholds of >0 and >0, over region FULL,
using
> 324655 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(529), raw thresholds of >0 and >0, over region FULL,
using
> 317763 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(625), raw thresholds of >0 and >0, over region FULL,
using
> 311011 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(729), raw thresholds of >0 and >0, over region FULL,
using
> 304407 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(841), raw thresholds of >0 and >0, over region FULL,
using
> 297911 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(961), raw thresholds of >0 and >0, over region FULL,
using
> 291521 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(1089), raw thresholds of >0 and >0, over region FULL,
using
> 285215 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2:
> DEBUG 2:
>
--------------------------------------------------------------------------------
> DEBUG 2:
> DEBUG 1: Output file:
>
/h/data/global/WXQC/data/fss/metoutput_fss/2016011500/2016011500.0072/grid_stat_50_APCP_6_720000L_20160118_000000V.stat
> DEBUG 1: Output file:
>
/h/data/global/WXQC/data/fss/metoutput_fss/2016011500/2016011500.0072/grid_stat_50_APCP_6_720000L_20160118_000000V_pairs.nc
> 1.0 QP100
> ['/h/WXQC/met-5.1/bin/grid_stat',
> '/h/data/global/WXQC/data/temp/grib.2016011500.0072',
> '/h/data/global/WXQC/data/fss/pcpcombout/2016011800/
> rgST4.2016011800.06h.nc',
> '/h/data/global/WXQC/data/fss/cfg/GridStatConfig_pop_updated', '-
outdir',
>
'/h/data/global/WXQC/data/fss/metoutput_fss/2016011500/2016011500.0072']
> DEBUG 1: Default Config File:
> /home/qcteam/met-5.1/share/met/config/GridStatConfig_default
> DEBUG 1: User Config File:
> /h/data/global/WXQC/data/fss/cfg/GridStatConfig_pop_updated
> WARNING:
>
> -----Original Message-----
> From: John Halley Gotway via RT [mailto:met_help at ucar.edu]
> Sent: Wednesday, January 27, 2016 10:37 AM
> To: CRAIG, ROBERT J GS-12 USAF ACC 16 WS/WXN
> Subject: Re: [rt.rap.ucar.edu #74813] Regrid Options
>
> Bob,
>
> You should use 2 as the width for the budget interpolation method.
In
> fact, if you try to use anything other than 2, you'll get an error
message
> similar to this:
>
> WARNING: process_command_line() -> Resetting the regridding width
from
> 6 to 2 for regridding method "BUDGET".
>
> Budget interpolation is very similar to bilinear interpolation,
which also
> can only be used with a width of 2. Bilinear interpolation performs
the
> interpolation using the closest 4 points. In budget interpolation,
each
> grid box is decomposed into a 5x5 mesh and bilinear interpolation is
> performed for each of those 25 sub-grid points. The final
interpolated
> value for that grid box is just average of those 25 sub-grid points.
That
> makes budget interpolation much slower, but (presumably) better able
to
> conserve accumulated variables, like precip.
>
> Hope that helps.
>
> Thanks,
> John
>
> On Wed, Jan 27, 2016 at 7:59 AM, robert.craig.2 at us.af.mil via RT <
> met_help at ucar.edu> wrote:
>
> >
> > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=74813 >
> >
> > John, With the regridding budget option, what is typically used
for
> > the width?
> >
> > Thanks
> >
> > -----Original Message-----
> > From: John Halley Gotway via RT [mailto:met_help at ucar.edu]
> > Sent: Thursday, January 21, 2016 11:23 AM
> > To: CRAIG, ROBERT J GS-12 USAF ACC 16 WS/WXN
> > Subject: [rt.rap.ucar.edu #74813] Regrid Options
> >
> > Bob,
> >
> > I see that you're running Grid-Stat and are trying to use the
> > regridding option included in the met-5.1 release. I think this
is
> > simpler than you expect. You don't need to explicitly specify the
> > grid definition yourself. Instead, all you need to put in the
config
> > file is something like this:
> >
> > //
> > // Verification grid
> > //
> > regrid = {
> > to_grid = OBS; // Or use "FCST" to verify on the forecast
grid.
> > vld_thresh = 0.5;
> > method = NEAREST;
> > width = 1;
> > }
> >
> > Grid-Stat reads the forecast and observation files, extracts their
> > grid information, and checks to see if they're the same. If not,
it
> > looks in the "to_grid" setting in the "regrid" section of the
config
> > file. If it says "FCST", it regrids the observation data onto the
> > forecast grid. If is says "OBS", it puts the forecast on the
> > observation grid. And there are some more options listed on page
46 of
> the User's Guide:
> >
> >
http://www.dtcenter.org/met/users/docs/users_guide/MET_Users_Guide_v5.
> > 1.pdf
> >
> > FYI, the default regridding method is just nearest neighbor, as
shown
> > above. You might want to use "BUDGET" for budget interpolation
when
> > using precip. Budget interpolation is designed to preserve mass
> > better than other methods, which is generally recommended for
precip.
> >
> > Lastly, I see you've asked the age-old question... regrid to the
> > higher-res or lower-res domain? I don't have a good answer for
you.
> > People do it both ways for different reasons. Some folks prefer
to
> > evaluate on the model domain, others prefer to use the observation
> > domain so as not to corrupt the obs with interpolation error. My
> > guess is that more people would say evaluate on the 20km rather
than
> > the 4km because it's better to throw data points away than to
create
> > new ones through interpolation. But people do it both ways.
> >
> > Hope that helps.
> >
> > Thanks
> > John
> >
> > On Thu, Jan 21, 2016 at 9:26 AM, robert.craig.2 at us.af.mil
> > <javascript:_e(%7B%7D,'cvml','robert.craig.2 at us.af.mil');> via RT
<
> > met_help at ucar.edu
<javascript:_e(%7B%7D,'cvml','met_help at ucar.edu');>>
> > wrote:
> >
> > >
> > > Thu Jan 21 09:26:14 2016: Request 74813 was acted upon.
> > > Transaction: Ticket created by robert.craig.2 at us.af.mil
> > > <javascript:_e(%7B%7D,'cvml','robert.craig.2 at us.af.mil');>
> > > Queue: met_help
> > > Subject: Regrid Options
> > > Owner: Nobody
> > > Requestors: robert.craig.2 at us.af.mil
> > > <javascript:_e(%7B%7D,'cvml','robert.craig.2 at us.af.mil');>
> > > Status: new
> > > Ticket <URL:
> > > https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=74813
> > > >
> > >
> > >
> > > John, I am using the regrid option in grid_stat to remap my
> > > observation grid (Stage IV) to a specific FCST grid. In the
> > > documentation it says you can specify the grid definition in the
> > > to_grid
> > parameter but it doesn't say
> > > the format to use. I assume it is something like this 255 3
342 210
> > 20586
> > > -122243 8 -96000 15000 15000 0 64 60000 30000 but I don't know
the
> > > specific syntax to use?
> > >
> > > Also, science question. I want to generate FSS stats for the
output
> > > of our 4km ensemble compared to our 20km ensemble. The obs
data
> > > is Stage IV. What is the smartest way to do this - should I
remap
> > > the 20km FCST to the 4km FCST grid along with the OBS or should
I
> > > remap the 20km and 4km to the OBS (Stage IV grid)?
> > >
> > > Thanks
> > >
> > >
> > >
> >
> >
> >
> >
>
>
>
>
------------------------------------------------
Subject: RE: [rt.rap.ucar.edu #74813] Regrid Options
From: robert.craig.2 at us.af.mil
Time: Wed Jan 27 11:44:52 2016
Okay, makes sense.
Thanks
-----Original Message-----
From: John Halley Gotway via RT [mailto:met_help at ucar.edu]
Sent: Wednesday, January 27, 2016 12:10 PM
To: CRAIG, ROBERT J GS-12 USAF ACC 16 WS/WXN
Subject: Re: [rt.rap.ucar.edu #74813] Regrid Options
Bob,
Thanks for send me that output from Grid-Stat. This help me
understand what you're seeing.
Let me start by saying this all looks fine to me. Here's why...
You're passing Grid-Stat two input files that are on different
domains.
Grid-Stat recognizes that the grids differ and uses the logic defined
in the "regrid" section of the configuration file to put them onto a
common grid. In there, you've chosen the "BUDGET" interpolation
method with a width of 4. The warning message explains that for
BUDGET interpolation the width must be set to 2, and it resets that
for you. Increasing the verbosity level (-v 3 or -v 4) may print out
a message about the automated regridding that was done.
The forecast and observation data are now on a common domain. Looking
in the Grid-Stat configuration file, you'll also see a section named
"interp"
which includes settings for interpolation methods and widths. The
name "interp" is a little confusing... perhaps we should have named it
"smooth"
instead because really it controls if/how the forecast and/or
observation data is smoothed prior to computing statistics. We kept
the name "interp"
to be consistent with similar options in the Point-Stat tool.
When doing verification, some folks apply upscaling (or smoothing) to
check how smoothing their data effects the verification scores. And
that can be done using the "interp" section of the config file.
However, users typically do not want to smooth their data prior to
computing stats, so the default is to perform no smoothing. Nearest
neighbor with a width of 1 means "no smoothing".
And that's the explanation for the string "for interpolation method
NEAREST(1)". All it's basically saying is that all grid points are
being compared with no smoothing applied.
Hopefully that helps.
Thanks,
John
On Wed, Jan 27, 2016 at 10:30 AM, robert.craig.2 at us.af.mil via RT <
met_help at ucar.edu> wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=74813 >
>
> That helps. Below is a listing of my run. Note the warning which
was
> explained by you. Down several lines is says interpolation method
> NEAREST(1). Should this also say BUDGET?
>
> Thanks
>
> DEBUG 1: Default Config File:
> /home/qcteam/met-5.1/share/met/config/GridStatConfig_default
> DEBUG 1: User Config File:
> /h/data/global/WXQC/data/fss/cfg/GridStatConfig_pop_updated
> WARNING:
> WARNING: parse_conf_regrid() -> Resetting the regridding width from
4
> to 2 for regridding method "BUDGET".
> WARNING:
> GSL_RNG_TYPE=mt19937
> GSL_RNG_SEED=18446744072165042262
> DEBUG 1: Forecast File:
> /h/data/global/WXQC/data/temp/grib.2016011500.0072
> DEBUG 1: Observation File:
>
/h/data/global/WXQC/data/fss/pcpcombout/2016011800/rgST4.2016011800.06
> h.nc
> DEBUG 2:
> DEBUG 2:
>
----------------------------------------------------------------------
> ----------
> DEBUG 2:
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NEAREST(1), over region FULL, using 402704 pairs.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(1), raw thresholds of >0 and >0, over region FULL,
using
> 402704 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(9), raw thresholds of >0 and >0, over region FULL,
using
> 393481 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(25), raw thresholds of >0 and >0, over region FULL,
using
> 384982 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(81), raw thresholds of >0 and >0, over region FULL,
using
> 368893 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(121), raw thresholds of >0 and >0, over region FULL,
> using
> 361134 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(169), raw thresholds of >0 and >0, over region FULL,
> using
> 353530 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(225), raw thresholds of >0 and >0, over region FULL,
> using
> 346075 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(289), raw thresholds of >0 and >0, over region FULL,
> using
> 338782 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(361), raw thresholds of >0 and >0, over region FULL,
> using
> 331650 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(441), raw thresholds of >0 and >0, over region FULL,
> using
> 324655 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(529), raw thresholds of >0 and >0, over region FULL,
> using
> 317763 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(625), raw thresholds of >0 and >0, over region FULL,
> using
> 311011 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(729), raw thresholds of >0 and >0, over region FULL,
> using
> 304407 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(841), raw thresholds of >0 and >0, over region FULL,
> using
> 297911 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(961), raw thresholds of >0 and >0, over region FULL,
> using
> 291521 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2: Processing QP050/A06 versus APCP_06(*,*), for interpolation
> method NBRHD(1089), raw thresholds of >0 and >0, over region FULL,
> using
> 285215 pairs.
> DEBUG 2: Computing Neighborhood Continuous Statistics.
> DEBUG 2:
> DEBUG 2:
>
----------------------------------------------------------------------
> ----------
> DEBUG 2:
> DEBUG 1: Output file:
>
/h/data/global/WXQC/data/fss/metoutput_fss/2016011500/2016011500.0072/
> grid_stat_50_APCP_6_720000L_20160118_000000V.stat
> DEBUG 1: Output file:
>
/h/data/global/WXQC/data/fss/metoutput_fss/2016011500/2016011500.0072/
> grid_stat_50_APCP_6_720000L_20160118_000000V_pairs.nc
> 1.0 QP100
> ['/h/WXQC/met-5.1/bin/grid_stat',
> '/h/data/global/WXQC/data/temp/grib.2016011500.0072',
> '/h/data/global/WXQC/data/fss/pcpcombout/2016011800/
> rgST4.2016011800.06h.nc',
> '/h/data/global/WXQC/data/fss/cfg/GridStatConfig_pop_updated',
> '-outdir',
>
'/h/data/global/WXQC/data/fss/metoutput_fss/2016011500/2016011500.0072
> ']
> DEBUG 1: Default Config File:
> /home/qcteam/met-5.1/share/met/config/GridStatConfig_default
> DEBUG 1: User Config File:
> /h/data/global/WXQC/data/fss/cfg/GridStatConfig_pop_updated
> WARNING:
>
> -----Original Message-----
> From: John Halley Gotway via RT [mailto:met_help at ucar.edu]
> Sent: Wednesday, January 27, 2016 10:37 AM
> To: CRAIG, ROBERT J GS-12 USAF ACC 16 WS/WXN
> Subject: Re: [rt.rap.ucar.edu #74813] Regrid Options
>
> Bob,
>
> You should use 2 as the width for the budget interpolation method.
In
> fact, if you try to use anything other than 2, you'll get an error
> message similar to this:
>
> WARNING: process_command_line() -> Resetting the regridding width
> from
> 6 to 2 for regridding method "BUDGET".
>
> Budget interpolation is very similar to bilinear interpolation,
which
> also can only be used with a width of 2. Bilinear interpolation
> performs the interpolation using the closest 4 points. In budget
> interpolation, each grid box is decomposed into a 5x5 mesh and
> bilinear interpolation is performed for each of those 25 sub-grid
> points. The final interpolated value for that grid box is just
> average of those 25 sub-grid points. That makes budget
interpolation
> much slower, but (presumably) better able to conserve accumulated
variables, like precip.
>
> Hope that helps.
>
> Thanks,
> John
>
> On Wed, Jan 27, 2016 at 7:59 AM, robert.craig.2 at us.af.mil via RT <
> met_help at ucar.edu> wrote:
>
> >
> > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=74813 >
> >
> > John, With the regridding budget option, what is typically used
for
> > the width?
> >
> > Thanks
> >
> > -----Original Message-----
> > From: John Halley Gotway via RT [mailto:met_help at ucar.edu]
> > Sent: Thursday, January 21, 2016 11:23 AM
> > To: CRAIG, ROBERT J GS-12 USAF ACC 16 WS/WXN
> > Subject: [rt.rap.ucar.edu #74813] Regrid Options
> >
> > Bob,
> >
> > I see that you're running Grid-Stat and are trying to use the
> > regridding option included in the met-5.1 release. I think this
is
> > simpler than you expect. You don't need to explicitly specify the
> > grid definition yourself. Instead, all you need to put in the
> > config file is something like this:
> >
> > //
> > // Verification grid
> > //
> > regrid = {
> > to_grid = OBS; // Or use "FCST" to verify on the forecast
grid.
> > vld_thresh = 0.5;
> > method = NEAREST;
> > width = 1;
> > }
> >
> > Grid-Stat reads the forecast and observation files, extracts their
> > grid information, and checks to see if they're the same. If not,
it
> > looks in the "to_grid" setting in the "regrid" section of the
config
> > file. If it says "FCST", it regrids the observation data onto the
> > forecast grid. If is says "OBS", it puts the forecast on the
> > observation grid. And there are some more options listed on page
46
> > of
> the User's Guide:
> >
> >
http://www.dtcenter.org/met/users/docs/users_guide/MET_Users_Guide_v5.
> > 1.pdf
> >
> > FYI, the default regridding method is just nearest neighbor, as
> > shown above. You might want to use "BUDGET" for budget
> > interpolation when using precip. Budget interpolation is designed
> > to preserve mass better than other methods, which is generally
recommended for precip.
> >
> > Lastly, I see you've asked the age-old question... regrid to the
> > higher-res or lower-res domain? I don't have a good answer for
you.
> > People do it both ways for different reasons. Some folks prefer
to
> > evaluate on the model domain, others prefer to use the observation
> > domain so as not to corrupt the obs with interpolation error. My
> > guess is that more people would say evaluate on the 20km rather
than
> > the 4km because it's better to throw data points away than to
create
> > new ones through interpolation. But people do it both ways.
> >
> > Hope that helps.
> >
> > Thanks
> > John
> >
> > On Thu, Jan 21, 2016 at 9:26 AM, robert.craig.2 at us.af.mil
> > <javascript:_e(%7B%7D,'cvml','robert.craig.2 at us.af.mil');> via RT
<
> > met_help at ucar.edu
> > <javascript:_e(%7B%7D,'cvml','met_help at ucar.edu');>>
> > wrote:
> >
> > >
> > > Thu Jan 21 09:26:14 2016: Request 74813 was acted upon.
> > > Transaction: Ticket created by robert.craig.2 at us.af.mil
> > > <javascript:_e(%7B%7D,'cvml','robert.craig.2 at us.af.mil');>
> > > Queue: met_help
> > > Subject: Regrid Options
> > > Owner: Nobody
> > > Requestors: robert.craig.2 at us.af.mil
> > > <javascript:_e(%7B%7D,'cvml','robert.craig.2 at us.af.mil');>
> > > Status: new
> > > Ticket <URL:
> > > https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=74813
> > > >
> > >
> > >
> > > John, I am using the regrid option in grid_stat to remap my
> > > observation grid (Stage IV) to a specific FCST grid. In the
> > > documentation it says you can specify the grid definition in the
> > > to_grid
> > parameter but it doesn't say
> > > the format to use. I assume it is something like this 255 3
342 210
> > 20586
> > > -122243 8 -96000 15000 15000 0 64 60000 30000 but I don't know
the
> > > specific syntax to use?
> > >
> > > Also, science question. I want to generate FSS stats for the
> > > output of our 4km ensemble compared to our 20km ensemble. The
> > > obs data is Stage IV. What is the smartest way to do this -
> > > should I remap the 20km FCST to the 4km FCST grid along with the
> > > OBS or should I remap the 20km and 4km to the OBS (Stage IV
grid)?
> > >
> > > Thanks
> > >
> > >
> > >
> >
> >
> >
> >
>
>
>
>
------------------------------------------------
More information about the Met_help
mailing list