[Met_help] [rt.rap.ucar.edu #82793] History for Storm-relative verification with point observations in Hurricane Maria using shift_data_plane
John Halley Gotway via RT
met_help at ucar.edu
Tue Jul 9 12:04:11 MDT 2019
----------------------------------------------------------------
Initial Request
----------------------------------------------------------------
Dear MET-help,
First, many thanks for supporting this software to the community and in
advance for any advice that you can give me.
I am using MET v6.0 + patches in an attempt to compare HWRF forecasts of
temperature, moisture, and winds to the same data collected by an Uncrewed
Aircraft System (UAS) that flew through Hurricane Maria. The UAS made a
slow descent from ~1 km to ~100 m ASL over the course of an hour as it
sampled the storm.
So far, I've approached this task by doing the following:
1) Converted UAS data to the ascii MET point format
2) Used ASCII2NC to convert the data to netcdf format
3) Ran PointStat to get the statistics
This worked very well. But now I would like to remove the effect of the
model track error before I run PointStat, and this is where I'm a bit fuzzy
on how to proceed.
I read about the shift_data_plane tool, and thought that I could use it to
shift the HWRF forecast grid, which is in grib2 format, by the location
offset between the modeled and observed storm center. Let's say that the
model storm center is at 26.0 N, -72.0 W and the real storm in the
observations is at 26.0N, -73.0 W. I did this:
shift_data_plane maria15l.2017092318.hwrfprs.core.0p02.f000.grb2
> shift_temp.nc 'name="TMP";level="P900";' -from 26.0 -72.0 -to 26.0 -73.0
> -method BILIN
Works great. However, I cannot seem to translate the entire 3d-temperature
grid at once; the tool will only write one pressure level per file (I
think), even though I tried specifying a range (not shown here). If that is
the case, how do I create a new model grid that is georeferenced to the
observed storm location so that I can feed that grid to PointStat? Am I
using the correct tools to accomplish this?
Thanks in advance again for any advice that you can provide.
Cheers,
Evan Kalina
--
Evan Kalina
Research Scientist
Developmental Testbed Center (DTC)
Cooperative Institute for Research in Environmental Sciences (CIRES) at the
NOAA/Earth System Research Lab/Global Systems Division
M-W, F: NOAA DSRC 2B603, 303-497-5399 (t)
Thursdays: NCAR FL-3 1040, 303-497-8945 (t)
----------------------------------------------------------------
Complete Ticket History
----------------------------------------------------------------
Subject: Storm-relative verification with point observations in Hurricane Maria using shift_data_plane
From: John Halley Gotway
Time: Wed Nov 01 16:16:06 2017
Evan,
Good question. You're correct... shift_data_plane is currently set up
to
only process 1 field at a time. Using met-6.0, there is no way to
tell
shift_data_plane to process multiple input fields.
I have several thoughts about this... and can think of 2 potential
solutions.
(1) We could enhance shift_data_plane to process multiple fields. The
shift_data_plane tool is pretty analogous to the regrid_data_plane
tool...
which does support the use of multiple "-field" settings to specify
multiple fields.
Going down this path, we'd likely want 2 code changes....
(A) Modify shift_data_plane to use a "-field" command line option to
process multiple fields.
(B) Update logic of both regrid_data_plane and shift_data_plane to
support
the specification of a *range* of levels. That would enable "TMP"
from
"P500-1000" to work.
The complicating factor is how we'd format the NetCDF output file.
It'd be
easiest to write 1 output variable per level using 2 dimensions. For
example, TMP_P750(lat,lon) and TMP_P850(lat,lon). But that would
prevent
you from passing a range of levels to Point-Stat and doing vertical
interpolation.
(2) The second main option would be enhancing the MET statistics tools
(Point-Stat, Grid-Stat, and so on) to define the shift directly in
their
configuration files. That would prevent you from having to run the
separate shift_data_plane tool entirely. Instead, we'd do the
shifting in
memory after reading the data.
What are your thoughts on all of this? I can see all 3 of the code
changes
described above being potentially useful.
If you'd like to talk more about this, feel free to drop by my office.
Thanks,
John
On Wed, Nov 1, 2017 at 1:07 PM, Evan Kalina - NOAA Affiliate via RT <
met_help at ucar.edu> wrote:
>
> Wed Nov 01 13:07:35 2017: Request 82793 was acted upon.
> Transaction: Ticket created by evan.kalina at noaa.gov
> Queue: met_help
> Subject: Storm-relative verification with point observations in
> Hurricane Maria using shift_data_plane
> Owner: Nobody
> Requestors: evan.kalina at noaa.gov
> Status: new
> Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=82793 >
>
>
> Dear MET-help,
>
> First, many thanks for supporting this software to the community and
in
> advance for any advice that you can give me.
>
> I am using MET v6.0 + patches in an attempt to compare HWRF
forecasts of
> temperature, moisture, and winds to the same data collected by an
Uncrewed
> Aircraft System (UAS) that flew through Hurricane Maria. The UAS
made a
> slow descent from ~1 km to ~100 m ASL over the course of an hour as
it
> sampled the storm.
>
> So far, I've approached this task by doing the following:
>
> 1) Converted UAS data to the ascii MET point format
> 2) Used ASCII2NC to convert the data to netcdf format
> 3) Ran PointStat to get the statistics
>
> This worked very well. But now I would like to remove the effect of
the
> model track error before I run PointStat, and this is where I'm a
bit fuzzy
> on how to proceed.
>
> I read about the shift_data_plane tool, and thought that I could use
it to
> shift the HWRF forecast grid, which is in grib2 format, by the
location
> offset between the modeled and observed storm center. Let's say that
the
> model storm center is at 26.0 N, -72.0 W and the real storm in the
> observations is at 26.0N, -73.0 W. I did this:
>
> shift_data_plane maria15l.2017092318 <(201)%20709-2318>.hwrfprs.co
> re.0p02.f000.grb2
> > shift_temp.nc 'name="TMP";level="P900";' -from 26.0 -72.0 -to 26.0
-73.0
> > -method BILIN
>
>
> Works great. However, I cannot seem to translate the entire 3d-
temperature
> grid at once; the tool will only write one pressure level per file
(I
> think), even though I tried specifying a range (not shown here). If
that is
> the case, how do I create a new model grid that is georeferenced to
the
> observed storm location so that I can feed that grid to PointStat?
Am I
> using the correct tools to accomplish this?
>
> Thanks in advance again for any advice that you can provide.
>
> Cheers,
> Evan Kalina
>
> --
> Evan Kalina
> Research Scientist
> Developmental Testbed Center (DTC)
> Cooperative Institute for Research in Environmental Sciences (CIRES)
at the
> NOAA/Earth System Research Lab/Global Systems Division
>
> M-W, F: NOAA DSRC 2B603, 303-497-5399 (t)
> Thursdays: NCAR FL-3 1040, 303-497-8945 (t)
>
>
------------------------------------------------
Subject: Storm-relative verification with point observations in Hurricane Maria using shift_data_plane
From: Evan Kalina - NOAA Affiliate
Time: Wed Nov 01 16:52:54 2017
Hey John,
Thanks for the quick response. I see the power in doing this sort of
verification task in MET to be the vertical interpolation of the model
data
to the observation location, so I would definitely favor an approach
that
preserves that capability. In terms of option 1 versus 2, it seems
like
option 2 would be easiest/most intuitive for the user. In fact, after
I
tried the shift_data_plane tool, I went back to the Users Guide and
read
the section on Point-Stat carefully, thinking that there could already
be
an option in that tool to shift the model grid in memory.
As a side note, I am wondering if Point-Stat might be the only tool
that
truly requires this upgrade. It seems like if my observations were a
gridded analysis, I could use the approach described in this slide
deck (
https://www.weather.gov/media/sti/nggps/06%20Colle_PI_Meeting_2016.pdf,
slides 12–15) and use the gen_vx_mask tool and the -force option in
the
series_analysis tool to do the storm-relative verification. More
involved
than passing a couple of flags to Grid-Stat, but possible (I didn't
try it
though, so my interpretation of the slides could be off).
I'll stop by your office tomorrow to chat more. As an interim solution
for
today, I simply shifted the observation locations to match the storm
location in the model run that I'm looking at and wrote out some new
MET
point data.
Cheers,
evan
On Wed, Nov 1, 2017 at 4:16 PM, John Halley Gotway via RT
<met_help at ucar.edu
> wrote:
> Evan,
>
> Good question. You're correct... shift_data_plane is currently set
up to
> only process 1 field at a time. Using met-6.0, there is no way to
tell
> shift_data_plane to process multiple input fields.
>
> I have several thoughts about this... and can think of 2 potential
> solutions.
>
> (1) We could enhance shift_data_plane to process multiple fields.
The
> shift_data_plane tool is pretty analogous to the regrid_data_plane
tool...
> which does support the use of multiple "-field" settings to specify
> multiple fields.
>
> Going down this path, we'd likely want 2 code changes....
>
> (A) Modify shift_data_plane to use a "-field" command line option to
> process multiple fields.
> (B) Update logic of both regrid_data_plane and shift_data_plane to
support
> the specification of a *range* of levels. That would enable "TMP"
from
> "P500-1000" to work.
>
> The complicating factor is how we'd format the NetCDF output file.
It'd be
> easiest to write 1 output variable per level using 2 dimensions.
For
> example, TMP_P750(lat,lon) and TMP_P850(lat,lon). But that would
prevent
> you from passing a range of levels to Point-Stat and doing vertical
> interpolation.
>
> (2) The second main option would be enhancing the MET statistics
tools
> (Point-Stat, Grid-Stat, and so on) to define the shift directly in
their
> configuration files. That would prevent you from having to run the
> separate shift_data_plane tool entirely. Instead, we'd do the
shifting in
> memory after reading the data.
>
> What are your thoughts on all of this? I can see all 3 of the code
changes
> described above being potentially useful.
>
> If you'd like to talk more about this, feel free to drop by my
office.
>
> Thanks,
> John
>
> On Wed, Nov 1, 2017 at 1:07 PM, Evan Kalina - NOAA Affiliate via RT
<
> met_help at ucar.edu> wrote:
>
> >
> > Wed Nov 01 13:07:35 2017: Request 82793 was acted upon.
> > Transaction: Ticket created by evan.kalina at noaa.gov
> > Queue: met_help
> > Subject: Storm-relative verification with point observations
in
> > Hurricane Maria using shift_data_plane
> > Owner: Nobody
> > Requestors: evan.kalina at noaa.gov
> > Status: new
> > Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=82793 >
> >
> >
> > Dear MET-help,
> >
> > First, many thanks for supporting this software to the community
and in
> > advance for any advice that you can give me.
> >
> > I am using MET v6.0 + patches in an attempt to compare HWRF
forecasts of
> > temperature, moisture, and winds to the same data collected by an
> Uncrewed
> > Aircraft System (UAS) that flew through Hurricane Maria. The UAS
made a
> > slow descent from ~1 km to ~100 m ASL over the course of an hour
as it
> > sampled the storm.
> >
> > So far, I've approached this task by doing the following:
> >
> > 1) Converted UAS data to the ascii MET point format
> > 2) Used ASCII2NC to convert the data to netcdf format
> > 3) Ran PointStat to get the statistics
> >
> > This worked very well. But now I would like to remove the effect
of the
> > model track error before I run PointStat, and this is where I'm a
bit
> fuzzy
> > on how to proceed.
> >
> > I read about the shift_data_plane tool, and thought that I could
use it
> to
> > shift the HWRF forecast grid, which is in grib2 format, by the
location
> > offset between the modeled and observed storm center. Let's say
that the
> > model storm center is at 26.0 N, -72.0 W and the real storm in the
> > observations is at 26.0N, -73.0 W. I did this:
> >
> > shift_data_plane maria15l.2017092318 <(201)%20709-2318>.hwrfprs.co
> > re.0p02.f000.grb2
> > > shift_temp.nc 'name="TMP";level="P900";' -from 26.0 -72.0 -to
26.0
> -73.0
> > > -method BILIN
> >
> >
> > Works great. However, I cannot seem to translate the entire
> 3d-temperature
> > grid at once; the tool will only write one pressure level per file
(I
> > think), even though I tried specifying a range (not shown here).
If that
> is
> > the case, how do I create a new model grid that is georeferenced
to the
> > observed storm location so that I can feed that grid to PointStat?
Am I
> > using the correct tools to accomplish this?
> >
> > Thanks in advance again for any advice that you can provide.
> >
> > Cheers,
> > Evan Kalina
> >
> > --
> > Evan Kalina
> > Research Scientist
> > Developmental Testbed Center (DTC)
> > Cooperative Institute for Research in Environmental Sciences
(CIRES) at
> the
> > NOAA/Earth System Research Lab/Global Systems Division
> >
> > M-W, F: NOAA DSRC 2B603, 303-497-5399 (t)
> > Thursdays: NCAR FL-3 1040, 303-497-8945 (t)
> >
> >
>
>
--
Evan Kalina
Research Scientist
Developmental Testbed Center (DTC)
Cooperative Institute for Research in Environmental Sciences (CIRES)
at the
NOAA/Earth System Research Lab/Global Systems Division
M-W, F: NOAA DSRC 2B603, 303-497-5399 (t)
Thursdays: NCAR FL-3 1040, 303-497-8945 (t)
------------------------------------------------
More information about the Met_help
mailing list