[Met_help] [rt.rap.ucar.edu #91721] History for question about mask and point_stat output
John Halley Gotway via RT
met_help at ucar.edu
Thu Aug 29 13:09:25 MDT 2019
----------------------------------------------------------------
Initial Request
----------------------------------------------------------------
Hi John,
I am looking at the MET output for WRF and MPAS verification. I run
point_stat over three regions, i.e., FULL, CONUS, EAST, and then look at
T2 bias.
When I look at the EAST region, it seems that some stations outside the
EAST area still show in the output. For example, in the file
"point_stat_240000L_20170202_000000V_mpr.txt", if you search "71050",
which is located at (52.11N, -124.13W), it appears in all the three
regions, --- I expect it to be in CONUS and FULL but not in EAST.
I put one test case in /glade/scratch/chenming/MET, where the
"run_command.txt" and "PointStatConfig.TT" are what I used.
I wonder whether I specify something wrong in the config or the mask
files ( CONUS.nc, EAST.nc) are not correct?
Thank you !
Ming
(P.S.: I run METv7.0)
----------------------------------------------------------------
Complete Ticket History
----------------------------------------------------------------
Subject: question about mask and point_stat output
From: John Halley Gotway
Time: Wed Aug 28 16:06:00 2019
Ming,
I took a look at your data on cheyenne, and I finally understand
what's
happening and why.
Here's the punchline... a similar issue was found last month and is
fixed
in the upcoming met-8.1.2 bugfix release:
* https://github.com/NCAR/MET/issues/1163
<https://github.com/NCAR/MET/issues/1163>*
The fix for the issue listed above also fixes the issue you've found.
So what's going on? You are running Point-Stat using 3 masking
regions:
the FULL model domain and CONUS and EAST masking regions generated by
running gen_vx_mask. But those masks were NOT generated for the same
domain that you're currently evaluating. Since you pointed out
station
"71050" with (lat, lon) = (52.11, -124.13), I used that for testing.
I've attached two plots. One is the output of plot_point_obs showing
where
station 71050 lives on your model domain (obs_sid_71050.png). The
second
is the output of plot_data_plane which shows the contents of EAST.nc
(EAST.png).
Look at the extent of these domains and it's clear that they are
different. Next, look closely at the log messages from Point-Stat,
and
you'll see something like:
*DEBUG 2: Regridding mask grid to the verification grid using nearest
neighbor interpolation:DEBUG 2: Projection: Lambert Conformal Nx: 1799
Ny:
1059 Lat_LL: 21.138 Lon_LL: 122.720 Lon_orient: 97.500 Alpha: 4203.494
Cone: 0.623 Bx: 899.1693 By: 3198.9984!=DEBUG 2: Projection: Lambert
Conformal Nx: 414 Ny: 324 Lat_LL: 13.695 Lon_LL: 127.745 Lon_orient:
101.000 Alpha: 828.433 Cone: 0.631 Bx: 206.5000 By: 680.8028*
Since the EAST mask is on a different domain than the model, Point-
Stat
automatically regrids it for you to the model domain using nearest
neighbor
interpolation. BUT that's where the problem is that's described in
the
GitHub issue. All the points outside the domain are assigned the bad
data
value of -9999. Since that's not "0", those points are included in
the
masking region!
This will be fixed in met-8.1.2, but the fix for you is very easy.
Just
rerun gen_vx_mask for your actual model domain:
*module use /glade/p/ral/jntp/MET/MET_releases/modulefiles*
*module load met/7.0*
*gen_vx_mask WRFOUT_PRES_2017-02-01_12:00:00.nc <http://00.nc>
/glade/p/ral/jntp/MET/MET_releases/7.0/share/met/poly/EAST.poly
EAST.nc*
*gen_vx_mask WRFOUT_PRES_2017-02-01_12:00:00.nc <http://00.nc>
/glade/p/ral/jntp/MET/MET_releases/7.0/share/met/poly/CONUS.poly
CONUS.nc*
And then update your Point-Stat config file to point to those new
files:
*mask = { grid = [ "FULL" ]; poly =
["/glade/scratch/chenming/MET/CONUS.nc",
"/glade/u/home/chenming/met-tutorial/mask/EAST.nc"]; sid = [];};*
Sorry for this bug, but I hope this helps clarify the situation.
Thanks,
John
On Tue, Aug 27, 2019 at 3:42 PM Randy Bullock via RT
<met_help at ucar.edu>
wrote:
>
> Tue Aug 27 15:42:01 2019: Request 91721 was acted upon.
> Transaction: Given to johnhg (John Halley Gotway) by bullock
> Queue: met_help
> Subject: question about mask and point_stat output
> Owner: johnhg
> Requestors: chenming at ucar.edu
> Status: new
> Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=91721 >
>
>
> This transaction appears to have no content
>
------------------------------------------------
Subject: Re: [rt.rap.ucar.edu #91721] question about mask and point_stat output
From: chenming at ucar.edu
Time: Thu Aug 29 11:28:15 2019
Hi John,
I just re-produce the mask file and all works fine now. Thank you so
much for the help!
Best regards,
Ming
On 8/28/19 4:06 PM, John Halley Gotway via RT wrote:
> Ming,
>
> I took a look at your data on cheyenne, and I finally understand
what's
> happening and why.
>
> Here's the punchline... a similar issue was found last month and is
fixed
> in the upcoming met-8.1.2 bugfix release:
> * https://github.com/NCAR/MET/issues/1163
> <https://github.com/NCAR/MET/issues/1163>*
>
> The fix for the issue listed above also fixes the issue you've
found.
>
> So what's going on? You are running Point-Stat using 3 masking
regions:
> the FULL model domain and CONUS and EAST masking regions generated
by
> running gen_vx_mask. But those masks were NOT generated for the
same
> domain that you're currently evaluating. Since you pointed out
station
> "71050" with (lat, lon) = (52.11, -124.13), I used that for testing.
>
> I've attached two plots. One is the output of plot_point_obs
showing where
> station 71050 lives on your model domain (obs_sid_71050.png). The
second
> is the output of plot_data_plane which shows the contents of EAST.nc
> (EAST.png).
>
> Look at the extent of these domains and it's clear that they are
> different. Next, look closely at the log messages from Point-Stat,
and
> you'll see something like:
>
>
>
>
> *DEBUG 2: Regridding mask grid to the verification grid using
nearest
> neighbor interpolation:DEBUG 2: Projection: Lambert Conformal Nx:
1799 Ny:
> 1059 Lat_LL: 21.138 Lon_LL: 122.720 Lon_orient: 97.500 Alpha:
4203.494
> Cone: 0.623 Bx: 899.1693 By: 3198.9984!=DEBUG 2: Projection: Lambert
> Conformal Nx: 414 Ny: 324 Lat_LL: 13.695 Lon_LL: 127.745 Lon_orient:
> 101.000 Alpha: 828.433 Cone: 0.631 Bx: 206.5000 By: 680.8028*
>
> Since the EAST mask is on a different domain than the model, Point-
Stat
> automatically regrids it for you to the model domain using nearest
neighbor
> interpolation. BUT that's where the problem is that's described in
the
> GitHub issue. All the points outside the domain are assigned the
bad data
> value of -9999. Since that's not "0", those points are included in
the
> masking region!
>
> This will be fixed in met-8.1.2, but the fix for you is very easy.
Just
> rerun gen_vx_mask for your actual model domain:
>
> *module use /glade/p/ral/jntp/MET/MET_releases/modulefiles*
> *module load met/7.0*
> *gen_vx_mask WRFOUT_PRES_2017-02-01_12:00:00.nc <http://00.nc>
> /glade/p/ral/jntp/MET/MET_releases/7.0/share/met/poly/EAST.poly
EAST.nc*
> *gen_vx_mask WRFOUT_PRES_2017-02-01_12:00:00.nc <http://00.nc>
> /glade/p/ral/jntp/MET/MET_releases/7.0/share/met/poly/CONUS.poly
CONUS.nc*
>
> And then update your Point-Stat config file to point to those new
files:
>
>
>
>
>
> *mask = { grid = [ "FULL" ]; poly =
> ["/glade/scratch/chenming/MET/CONUS.nc",
> "/glade/u/home/chenming/met-tutorial/mask/EAST.nc"]; sid = [];};*
>
> Sorry for this bug, but I hope this helps clarify the situation.
>
> Thanks,
> John
>
>
> On Tue, Aug 27, 2019 at 3:42 PM Randy Bullock via RT
<met_help at ucar.edu>
> wrote:
>
>> Tue Aug 27 15:42:01 2019: Request 91721 was acted upon.
>> Transaction: Given to johnhg (John Halley Gotway) by bullock
>> Queue: met_help
>> Subject: question about mask and point_stat output
>> Owner: johnhg
>> Requestors: chenming at ucar.edu
>> Status: new
>> Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=91721 >
>>
>>
>> This transaction appears to have no content
>>
------------------------------------------------
More information about the Met_help
mailing list