[Met_help] [rt.rap.ucar.edu #79224] History for calculating ensemble spread

John Halley Gotway via RT met_help at ucar.edu
Tue Feb 7 17:17:29 MST 2017


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

Hi, for my project I need to calculate some measure of the spread of two
different ensembles (one "large" and one "small" ensemble). I have model
output data (the large ensemble is in grib2 format, the small ensemble is
WRF NetCDF), and I have MADIS point observations that have already been
processed for use in MET.

I previously used the ensemble-stat tool on the large ensemble to dump out
matched pair info for all ensemble member forecasts at each observation
location by printing out the ORANK line type. Given the inputs that I
described above, will MET calculate the ensemble spread automatically if I
turn on the SSVAR line type? In the user guide (v5.2) it mentions that an
"observation field" is needed to calculate things like spread/skill
variance. Do point obs from MADIS count as an "observation field"?

Guidance would be much appreciated!

Jared

-- 
===============================
Jared A. Lee, Ph.D.
Project Scientist I
Research Applications Laboratory
National Center for Atmospheric Research
Boulder, Colorado, USA

Email: jaredlee at ucar.edu (w)
Phone: 303.497.8485 (w)
Web: https://staff.ucar.edu/users/jaredlee
===============================


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

Subject: calculating ensemble spread
From: John Halley Gotway
Time: Fri Jan 20 17:07:14 2017

Hi Jared,

I see a couple of questions in your email.

First, yes, Ensemble-Stat is set up to process both point and gridded
observations.  And the SSVAR (spread/skill variance) line type can be
produced using either of these.

The SSVAR line type stores information about the relationship between
ensemble spread and skill.  Here's how it works.  Ensemble-Stat
processes
each observation value... either from a point location (i.e. for point
obs)
or from a gridded location (i.e. gridded obs).  For each observation,
it
gathers the N ensemble forecast values which correspond to it.  It
computes
the variance of those ensemble values, and it also computes the error
by
subtracting the mean ensemble value minus the observation value.
(Note,
you can pass in an ensemble mean field instead of computing it on the
fly
like this).  After processing N observations, we have N pairs of
variance
values and error values.  It groups those N pairs into buckets by
binning
the variance using the bucket size defined in the config file.  You'll
see
in the output one SSVAR line for each of those variance buckets.

The METViewer database and display system (or Stat-Analysis) can
aggregate
those SSVAR lines intelligently across multiple cases and create plots
which show the relationship between spread and skill.

If that's what you need, great.

However, if your looking for a much simpler computation of mean spread
over
some spatial region, unfortunately, the answer is no.  For example,
Ensemble-Stat doesn't write out the mean spread value over the entire
domain.

Just let me know what questions you still have.

Thanks,
John Halley Gotway


On Fri, Jan 20, 2017 at 11:41 AM, Jared Lee via RT <met_help at ucar.edu>
wrote:

>
> Fri Jan 20 11:41:03 2017: Request 79224 was acted upon.
> Transaction: Ticket created by jaredlee at ucar.edu
>        Queue: met_help
>      Subject: calculating ensemble spread
>        Owner: Nobody
>   Requestors: jaredlee at ucar.edu
>       Status: new
>  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
>
>
> Hi, for my project I need to calculate some measure of the spread of
two
> different ensembles (one "large" and one "small" ensemble). I have
model
> output data (the large ensemble is in grib2 format, the small
ensemble is
> WRF NetCDF), and I have MADIS point observations that have already
been
> processed for use in MET.
>
> I previously used the ensemble-stat tool on the large ensemble to
dump out
> matched pair info for all ensemble member forecasts at each
observation
> location by printing out the ORANK line type. Given the inputs that
I
> described above, will MET calculate the ensemble spread
automatically if I
> turn on the SSVAR line type? In the user guide (v5.2) it mentions
that an
> "observation field" is needed to calculate things like spread/skill
> variance. Do point obs from MADIS count as an "observation field"?
>
> Guidance would be much appreciated!
>
> Jared
>
> --
> ===============================
> Jared A. Lee, Ph.D.
> Project Scientist I
> Research Applications Laboratory
> National Center for Atmospheric Research
> Boulder, Colorado, USA
>
> Email: jaredlee at ucar.edu (w)
> Phone: 303.497.8485 (w)
> Web: https://staff.ucar.edu/users/jaredlee
> ===============================
>
>

------------------------------------------------
Subject: calculating ensemble spread
From: Jared Lee
Time: Fri Jan 20 17:31:24 2017

Hi John,

Thanks for clarifying what ensemble-stat does to produce the SSVAR
line
type. That helps.

As for what I ultimately want, I guess that's still somewhat vaguely
defined. I think making plots that show the spread/skill relationship
across the date ranges, lead times, and variables of interest for both
ensembles would be a useful way to illustrate and compare the spread
of the
two ensembles. That seems like it would be more useful than simply
calculating the mean spread of the ensemble (which I could do by
writing a
new NCL script to read in all the ensemble fields and not even
bothering
with MET...).

Another side question... Can MET read in native-grid NetCDF WRF output
files? Or do I need to run them through some pre-processor (like UPP)
first? After going through the user guide I'm not clear whether MET
accepts
NetCDF wrfout files or not without pre-processing. Also, if all the
output
times are in a single wrfout file, can MET open the one file and find
the
correct time it's supposed to read, or do all the wrfout files need to
be
for a single time only? The NetCDF wrfout files I've been given for
the
small ensemble have all the output times in a single file, but I can
write
a script to parse them out to single-time files with ncks. I'd just
rather
not go through that hassle if I don't have to!

Jared

On Fri, Jan 20, 2017 at 5:07 PM, John Halley Gotway via RT <
met_help at ucar.edu> wrote:

> Hi Jared,
>
> I see a couple of questions in your email.
>
> First, yes, Ensemble-Stat is set up to process both point and
gridded
> observations.  And the SSVAR (spread/skill variance) line type can
be
> produced using either of these.
>
> The SSVAR line type stores information about the relationship
between
> ensemble spread and skill.  Here's how it works.  Ensemble-Stat
processes
> each observation value... either from a point location (i.e. for
point obs)
> or from a gridded location (i.e. gridded obs).  For each
observation, it
> gathers the N ensemble forecast values which correspond to it.  It
computes
> the variance of those ensemble values, and it also computes the
error by
> subtracting the mean ensemble value minus the observation value.
(Note,
> you can pass in an ensemble mean field instead of computing it on
the fly
> like this).  After processing N observations, we have N pairs of
variance
> values and error values.  It groups those N pairs into buckets by
binning
> the variance using the bucket size defined in the config file.
You'll see
> in the output one SSVAR line for each of those variance buckets.
>
> The METViewer database and display system (or Stat-Analysis) can
aggregate
> those SSVAR lines intelligently across multiple cases and create
plots
> which show the relationship between spread and skill.
>
> If that's what you need, great.
>
> However, if your looking for a much simpler computation of mean
spread over
> some spatial region, unfortunately, the answer is no.  For example,
> Ensemble-Stat doesn't write out the mean spread value over the
entire
> domain.
>
> Just let me know what questions you still have.
>
> Thanks,
> John Halley Gotway
>
>
> On Fri, Jan 20, 2017 at 11:41 AM, Jared Lee via RT
<met_help at ucar.edu>
> wrote:
>
> >
> > Fri Jan 20 11:41:03 2017: Request 79224 was acted upon.
> > Transaction: Ticket created by jaredlee at ucar.edu
> >        Queue: met_help
> >      Subject: calculating ensemble spread
> >        Owner: Nobody
> >   Requestors: jaredlee at ucar.edu
> >       Status: new
> >  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
> >
> >
> > Hi, for my project I need to calculate some measure of the spread
of two
> > different ensembles (one "large" and one "small" ensemble). I have
model
> > output data (the large ensemble is in grib2 format, the small
ensemble is
> > WRF NetCDF), and I have MADIS point observations that have already
been
> > processed for use in MET.
> >
> > I previously used the ensemble-stat tool on the large ensemble to
dump
> out
> > matched pair info for all ensemble member forecasts at each
observation
> > location by printing out the ORANK line type. Given the inputs
that I
> > described above, will MET calculate the ensemble spread
automatically if
> I
> > turn on the SSVAR line type? In the user guide (v5.2) it mentions
that an
> > "observation field" is needed to calculate things like
spread/skill
> > variance. Do point obs from MADIS count as an "observation field"?
> >
> > Guidance would be much appreciated!
> >
> > Jared
> >
> > --
> > ===============================
> > Jared A. Lee, Ph.D.
> > Project Scientist I
> > Research Applications Laboratory
> > National Center for Atmospheric Research
> > Boulder, Colorado, USA
> >
> > Email: jaredlee at ucar.edu (w)
> > Phone: 303.497.8485 (w)
> > Web: https://staff.ucar.edu/users/jaredlee
> > ===============================
> >
> >
>
>


--
===============================
Jared A. Lee, Ph.D.
Project Scientist I
Research Applications Laboratory
National Center for Atmospheric Research
Boulder, Colorado, USA

Email: jaredlee at ucar.edu (w)
Phone: 303.497.8485 (w)
Web: https://staff.ucar.edu/users/jaredlee
===============================

------------------------------------------------
Subject: calculating ensemble spread
From: John Halley Gotway
Time: Mon Jan 23 17:01:59 2017

Jared,

Unfortunately, no, MET is not set up to read wrfout files directly.
As for
post-processing, you have 2 options... either UPP or wrf_interp (only
for
WRF-ARW core).

The former writes GRIB output which MET handles well. The latter
writes a
flavor of NetCDF that MET is able to read... with the caveat that it
can
only read fields that are defined on non-staggered dimensions.
Basically,
it can't read the wind fields.

I always encourage people to use UPP since MET supports GRIB well.

Being able to read wrfout directly has often been requested.  The two
issues are the staggered dimensions and the hybrid vertical
coordinate.
Generally, matching to point observations is done in terms of pressure
or
height.  So some interpolation to pressure/height levels in inevitable
along the way.  We've resisted re-implementing that post-processing
logic
directly in MET to avoid duplication of effort with UPP.  And UPP
supports
the derivation of many other variable types that are pretty useful.

Perhaps there is some middle ground where we could enable MET to call
UPP
behind the scenes?  But we'd need to work out many details.

Thanks,
John

On Fri, Jan 20, 2017 at 5:31 PM, Jared Lee via RT <met_help at ucar.edu>
wrote:

>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
>
> Hi John,
>
> Thanks for clarifying what ensemble-stat does to produce the SSVAR
line
> type. That helps.
>
> As for what I ultimately want, I guess that's still somewhat vaguely
> defined. I think making plots that show the spread/skill
relationship
> across the date ranges, lead times, and variables of interest for
both
> ensembles would be a useful way to illustrate and compare the spread
of the
> two ensembles. That seems like it would be more useful than simply
> calculating the mean spread of the ensemble (which I could do by
writing a
> new NCL script to read in all the ensemble fields and not even
bothering
> with MET...).
>
> Another side question... Can MET read in native-grid NetCDF WRF
output
> files? Or do I need to run them through some pre-processor (like
UPP)
> first? After going through the user guide I'm not clear whether MET
accepts
> NetCDF wrfout files or not without pre-processing. Also, if all the
output
> times are in a single wrfout file, can MET open the one file and
find the
> correct time it's supposed to read, or do all the wrfout files need
to be
> for a single time only? The NetCDF wrfout files I've been given for
the
> small ensemble have all the output times in a single file, but I can
write
> a script to parse them out to single-time files with ncks. I'd just
rather
> not go through that hassle if I don't have to!
>
> Jared
>
> On Fri, Jan 20, 2017 at 5:07 PM, John Halley Gotway via RT <
> met_help at ucar.edu> wrote:
>
> > Hi Jared,
> >
> > I see a couple of questions in your email.
> >
> > First, yes, Ensemble-Stat is set up to process both point and
gridded
> > observations.  And the SSVAR (spread/skill variance) line type can
be
> > produced using either of these.
> >
> > The SSVAR line type stores information about the relationship
between
> > ensemble spread and skill.  Here's how it works.  Ensemble-Stat
processes
> > each observation value... either from a point location (i.e. for
point
> obs)
> > or from a gridded location (i.e. gridded obs).  For each
observation, it
> > gathers the N ensemble forecast values which correspond to it.  It
> computes
> > the variance of those ensemble values, and it also computes the
error by
> > subtracting the mean ensemble value minus the observation value.
(Note,
> > you can pass in an ensemble mean field instead of computing it on
the fly
> > like this).  After processing N observations, we have N pairs of
variance
> > values and error values.  It groups those N pairs into buckets by
binning
> > the variance using the bucket size defined in the config file.
You'll
> see
> > in the output one SSVAR line for each of those variance buckets.
> >
> > The METViewer database and display system (or Stat-Analysis) can
> aggregate
> > those SSVAR lines intelligently across multiple cases and create
plots
> > which show the relationship between spread and skill.
> >
> > If that's what you need, great.
> >
> > However, if your looking for a much simpler computation of mean
spread
> over
> > some spatial region, unfortunately, the answer is no.  For
example,
> > Ensemble-Stat doesn't write out the mean spread value over the
entire
> > domain.
> >
> > Just let me know what questions you still have.
> >
> > Thanks,
> > John Halley Gotway
> >
> >
> > On Fri, Jan 20, 2017 at 11:41 AM, Jared Lee via RT
<met_help at ucar.edu>
> > wrote:
> >
> > >
> > > Fri Jan 20 11:41:03 2017: Request 79224 was acted upon.
> > > Transaction: Ticket created by jaredlee at ucar.edu
> > >        Queue: met_help
> > >      Subject: calculating ensemble spread
> > >        Owner: Nobody
> > >   Requestors: jaredlee at ucar.edu
> > >       Status: new
> > >  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224
> >
> > >
> > >
> > > Hi, for my project I need to calculate some measure of the
spread of
> two
> > > different ensembles (one "large" and one "small" ensemble). I
have
> model
> > > output data (the large ensemble is in grib2 format, the small
ensemble
> is
> > > WRF NetCDF), and I have MADIS point observations that have
already been
> > > processed for use in MET.
> > >
> > > I previously used the ensemble-stat tool on the large ensemble
to dump
> > out
> > > matched pair info for all ensemble member forecasts at each
observation
> > > location by printing out the ORANK line type. Given the inputs
that I
> > > described above, will MET calculate the ensemble spread
automatically
> if
> > I
> > > turn on the SSVAR line type? In the user guide (v5.2) it
mentions that
> an
> > > "observation field" is needed to calculate things like
spread/skill
> > > variance. Do point obs from MADIS count as an "observation
field"?
> > >
> > > Guidance would be much appreciated!
> > >
> > > Jared
> > >
> > > --
> > > ===============================
> > > Jared A. Lee, Ph.D.
> > > Project Scientist I
> > > Research Applications Laboratory
> > > National Center for Atmospheric Research
> > > Boulder, Colorado, USA
> > >
> > > Email: jaredlee at ucar.edu (w)
> > > Phone: 303.497.8485 (w)
> > > Web: https://staff.ucar.edu/users/jaredlee
> > > ===============================
> > >
> > >
> >
> >
>
>
> --
> ===============================
> Jared A. Lee, Ph.D.
> Project Scientist I
> Research Applications Laboratory
> National Center for Atmospheric Research
> Boulder, Colorado, USA
>
> Email: jaredlee at ucar.edu (w)
> Phone: 303.497.8485 (w)
> Web: https://staff.ucar.edu/users/jaredlee
> ===============================
>
>

------------------------------------------------
Subject: calculating ensemble spread
From: Jared Lee
Time: Fri Feb 03 15:09:30 2017

Hi John,

Now I finally have finished running UPP (after having to split apart
the
wrfout files into single-time files with ncks) and ensemble-stat on
all the
model data for my three ensembles. How do I go about getting that into
METViewer so that I can plot spread-skill information? My data is on
yellowstone. I also looked through the stat-analysis section of the
MET
Users Guide, but I'm not clear on what exactly I'd need to do to get
what I
need with that...

Jared

On Mon, Jan 23, 2017 at 5:01 PM, John Halley Gotway via RT <
met_help at ucar.edu> wrote:

> Jared,
>
> Unfortunately, no, MET is not set up to read wrfout files directly.
As for
> post-processing, you have 2 options... either UPP or wrf_interp
(only for
> WRF-ARW core).
>
> The former writes GRIB output which MET handles well. The latter
writes a
> flavor of NetCDF that MET is able to read... with the caveat that it
can
> only read fields that are defined on non-staggered dimensions.
Basically,
> it can't read the wind fields.
>
> I always encourage people to use UPP since MET supports GRIB well.
>
> Being able to read wrfout directly has often been requested.  The
two
> issues are the staggered dimensions and the hybrid vertical
coordinate.
> Generally, matching to point observations is done in terms of
pressure or
> height.  So some interpolation to pressure/height levels in
inevitable
> along the way.  We've resisted re-implementing that post-processing
logic
> directly in MET to avoid duplication of effort with UPP.  And UPP
supports
> the derivation of many other variable types that are pretty useful.
>
> Perhaps there is some middle ground where we could enable MET to
call UPP
> behind the scenes?  But we'd need to work out many details.
>
> Thanks,
> John
>
> On Fri, Jan 20, 2017 at 5:31 PM, Jared Lee via RT
<met_help at ucar.edu>
> wrote:
>
> >
> > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
> >
> > Hi John,
> >
> > Thanks for clarifying what ensemble-stat does to produce the SSVAR
line
> > type. That helps.
> >
> > As for what I ultimately want, I guess that's still somewhat
vaguely
> > defined. I think making plots that show the spread/skill
relationship
> > across the date ranges, lead times, and variables of interest for
both
> > ensembles would be a useful way to illustrate and compare the
spread of
> the
> > two ensembles. That seems like it would be more useful than simply
> > calculating the mean spread of the ensemble (which I could do by
writing
> a
> > new NCL script to read in all the ensemble fields and not even
bothering
> > with MET...).
> >
> > Another side question... Can MET read in native-grid NetCDF WRF
output
> > files? Or do I need to run them through some pre-processor (like
UPP)
> > first? After going through the user guide I'm not clear whether
MET
> accepts
> > NetCDF wrfout files or not without pre-processing. Also, if all
the
> output
> > times are in a single wrfout file, can MET open the one file and
find the
> > correct time it's supposed to read, or do all the wrfout files
need to be
> > for a single time only? The NetCDF wrfout files I've been given
for the
> > small ensemble have all the output times in a single file, but I
can
> write
> > a script to parse them out to single-time files with ncks. I'd
just
> rather
> > not go through that hassle if I don't have to!
> >
> > Jared
> >
> > On Fri, Jan 20, 2017 at 5:07 PM, John Halley Gotway via RT <
> > met_help at ucar.edu> wrote:
> >
> > > Hi Jared,
> > >
> > > I see a couple of questions in your email.
> > >
> > > First, yes, Ensemble-Stat is set up to process both point and
gridded
> > > observations.  And the SSVAR (spread/skill variance) line type
can be
> > > produced using either of these.
> > >
> > > The SSVAR line type stores information about the relationship
between
> > > ensemble spread and skill.  Here's how it works.  Ensemble-Stat
> processes
> > > each observation value... either from a point location (i.e. for
point
> > obs)
> > > or from a gridded location (i.e. gridded obs).  For each
observation,
> it
> > > gathers the N ensemble forecast values which correspond to it.
It
> > computes
> > > the variance of those ensemble values, and it also computes the
error
> by
> > > subtracting the mean ensemble value minus the observation value.
> (Note,
> > > you can pass in an ensemble mean field instead of computing it
on the
> fly
> > > like this).  After processing N observations, we have N pairs of
> variance
> > > values and error values.  It groups those N pairs into buckets
by
> binning
> > > the variance using the bucket size defined in the config file.
You'll
> > see
> > > in the output one SSVAR line for each of those variance buckets.
> > >
> > > The METViewer database and display system (or Stat-Analysis) can
> > aggregate
> > > those SSVAR lines intelligently across multiple cases and create
plots
> > > which show the relationship between spread and skill.
> > >
> > > If that's what you need, great.
> > >
> > > However, if your looking for a much simpler computation of mean
spread
> > over
> > > some spatial region, unfortunately, the answer is no.  For
example,
> > > Ensemble-Stat doesn't write out the mean spread value over the
entire
> > > domain.
> > >
> > > Just let me know what questions you still have.
> > >
> > > Thanks,
> > > John Halley Gotway
> > >
> > >
> > > On Fri, Jan 20, 2017 at 11:41 AM, Jared Lee via RT
<met_help at ucar.edu>
> > > wrote:
> > >
> > > >
> > > > Fri Jan 20 11:41:03 2017: Request 79224 was acted upon.
> > > > Transaction: Ticket created by jaredlee at ucar.edu
> > > >        Queue: met_help
> > > >      Subject: calculating ensemble spread
> > > >        Owner: Nobody
> > > >   Requestors: jaredlee at ucar.edu
> > > >       Status: new
> > > >  Ticket <URL: https://rt.rap.ucar.edu/rt/
> Ticket/Display.html?id=79224
> > >
> > > >
> > > >
> > > > Hi, for my project I need to calculate some measure of the
spread of
> > two
> > > > different ensembles (one "large" and one "small" ensemble). I
have
> > model
> > > > output data (the large ensemble is in grib2 format, the small
> ensemble
> > is
> > > > WRF NetCDF), and I have MADIS point observations that have
already
> been
> > > > processed for use in MET.
> > > >
> > > > I previously used the ensemble-stat tool on the large ensemble
to
> dump
> > > out
> > > > matched pair info for all ensemble member forecasts at each
> observation
> > > > location by printing out the ORANK line type. Given the inputs
that I
> > > > described above, will MET calculate the ensemble spread
automatically
> > if
> > > I
> > > > turn on the SSVAR line type? In the user guide (v5.2) it
mentions
> that
> > an
> > > > "observation field" is needed to calculate things like
spread/skill
> > > > variance. Do point obs from MADIS count as an "observation
field"?
> > > >
> > > > Guidance would be much appreciated!
> > > >
> > > > Jared
> > > >
> > > > --
> > > > ===============================
> > > > Jared A. Lee, Ph.D.
> > > > Project Scientist I
> > > > Research Applications Laboratory
> > > > National Center for Atmospheric Research
> > > > Boulder, Colorado, USA
> > > >
> > > > Email: jaredlee at ucar.edu (w)
> > > > Phone: 303.497.8485 (w)
> > > > Web: https://staff.ucar.edu/users/jaredlee
> > > > ===============================
> > > >
> > > >
> > >
> > >
> >
> >
> > --
> > ===============================
> > Jared A. Lee, Ph.D.
> > Project Scientist I
> > Research Applications Laboratory
> > National Center for Atmospheric Research
> > Boulder, Colorado, USA
> >
> > Email: jaredlee at ucar.edu (w)
> > Phone: 303.497.8485 (w)
> > Web: https://staff.ucar.edu/users/jaredlee
> > ===============================
> >
> >
>
>


--
===============================
Jared A. Lee, Ph.D.
Project Scientist I
Research Applications Laboratory
National Center for Atmospheric Research
Boulder, Colorado, USA

Email: jaredlee at ucar.edu (w)
Phone: 303.497.8485 (w)
Web: https://staff.ucar.edu/users/jaredlee
===============================

------------------------------------------------
Subject: calculating ensemble spread
From: John Halley Gotway
Time: Fri Feb 03 16:36:06 2017

Hi Jared,

I'm surprised you need to use ncks to split out multiple times.  Isn't
there a script for running UPP on a wrfout file with multiple times?
I
believe it's called run_unipost_frames.

As for your question about METViewer... you'd need to pull the
ensemble-stat data down to a machine inside the RAL firewall.  And
then I
could demonstrate to you how to load it into METViewer and start
creating
plots.

Unfortunately, you can't load it directly from yellowstone into
METViewer,
which lives on a machine named "mandan".

John

On Fri, Feb 3, 2017 at 3:09 PM, Jared Lee via RT <met_help at ucar.edu>
wrote:

>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
>
> Hi John,
>
> Now I finally have finished running UPP (after having to split apart
the
> wrfout files into single-time files with ncks) and ensemble-stat on
all the
> model data for my three ensembles. How do I go about getting that
into
> METViewer so that I can plot spread-skill information? My data is on
> yellowstone. I also looked through the stat-analysis section of the
MET
> Users Guide, but I'm not clear on what exactly I'd need to do to get
what I
> need with that...
>
> Jared
>
> On Mon, Jan 23, 2017 at 5:01 PM, John Halley Gotway via RT <
> met_help at ucar.edu> wrote:
>
> > Jared,
> >
> > Unfortunately, no, MET is not set up to read wrfout files
directly.  As
> for
> > post-processing, you have 2 options... either UPP or wrf_interp
(only for
> > WRF-ARW core).
> >
> > The former writes GRIB output which MET handles well. The latter
writes a
> > flavor of NetCDF that MET is able to read... with the caveat that
it can
> > only read fields that are defined on non-staggered dimensions.
> Basically,
> > it can't read the wind fields.
> >
> > I always encourage people to use UPP since MET supports GRIB well.
> >
> > Being able to read wrfout directly has often been requested.  The
two
> > issues are the staggered dimensions and the hybrid vertical
coordinate.
> > Generally, matching to point observations is done in terms of
pressure or
> > height.  So some interpolation to pressure/height levels in
inevitable
> > along the way.  We've resisted re-implementing that post-
processing logic
> > directly in MET to avoid duplication of effort with UPP.  And UPP
> supports
> > the derivation of many other variable types that are pretty
useful.
> >
> > Perhaps there is some middle ground where we could enable MET to
call UPP
> > behind the scenes?  But we'd need to work out many details.
> >
> > Thanks,
> > John
> >
> > On Fri, Jan 20, 2017 at 5:31 PM, Jared Lee via RT
<met_help at ucar.edu>
> > wrote:
> >
> > >
> > > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
> > >
> > > Hi John,
> > >
> > > Thanks for clarifying what ensemble-stat does to produce the
SSVAR line
> > > type. That helps.
> > >
> > > As for what I ultimately want, I guess that's still somewhat
vaguely
> > > defined. I think making plots that show the spread/skill
relationship
> > > across the date ranges, lead times, and variables of interest
for both
> > > ensembles would be a useful way to illustrate and compare the
spread of
> > the
> > > two ensembles. That seems like it would be more useful than
simply
> > > calculating the mean spread of the ensemble (which I could do by
> writing
> > a
> > > new NCL script to read in all the ensemble fields and not even
> bothering
> > > with MET...).
> > >
> > > Another side question... Can MET read in native-grid NetCDF WRF
output
> > > files? Or do I need to run them through some pre-processor (like
UPP)
> > > first? After going through the user guide I'm not clear whether
MET
> > accepts
> > > NetCDF wrfout files or not without pre-processing. Also, if all
the
> > output
> > > times are in a single wrfout file, can MET open the one file and
find
> the
> > > correct time it's supposed to read, or do all the wrfout files
need to
> be
> > > for a single time only? The NetCDF wrfout files I've been given
for the
> > > small ensemble have all the output times in a single file, but I
can
> > write
> > > a script to parse them out to single-time files with ncks. I'd
just
> > rather
> > > not go through that hassle if I don't have to!
> > >
> > > Jared
> > >
> > > On Fri, Jan 20, 2017 at 5:07 PM, John Halley Gotway via RT <
> > > met_help at ucar.edu> wrote:
> > >
> > > > Hi Jared,
> > > >
> > > > I see a couple of questions in your email.
> > > >
> > > > First, yes, Ensemble-Stat is set up to process both point and
gridded
> > > > observations.  And the SSVAR (spread/skill variance) line type
can be
> > > > produced using either of these.
> > > >
> > > > The SSVAR line type stores information about the relationship
between
> > > > ensemble spread and skill.  Here's how it works.  Ensemble-
Stat
> > processes
> > > > each observation value... either from a point location (i.e.
for
> point
> > > obs)
> > > > or from a gridded location (i.e. gridded obs).  For each
observation,
> > it
> > > > gathers the N ensemble forecast values which correspond to it.
It
> > > computes
> > > > the variance of those ensemble values, and it also computes
the error
> > by
> > > > subtracting the mean ensemble value minus the observation
value.
> > (Note,
> > > > you can pass in an ensemble mean field instead of computing it
on the
> > fly
> > > > like this).  After processing N observations, we have N pairs
of
> > variance
> > > > values and error values.  It groups those N pairs into buckets
by
> > binning
> > > > the variance using the bucket size defined in the config file.
> You'll
> > > see
> > > > in the output one SSVAR line for each of those variance
buckets.
> > > >
> > > > The METViewer database and display system (or Stat-Analysis)
can
> > > aggregate
> > > > those SSVAR lines intelligently across multiple cases and
create
> plots
> > > > which show the relationship between spread and skill.
> > > >
> > > > If that's what you need, great.
> > > >
> > > > However, if your looking for a much simpler computation of
mean
> spread
> > > over
> > > > some spatial region, unfortunately, the answer is no.  For
example,
> > > > Ensemble-Stat doesn't write out the mean spread value over the
entire
> > > > domain.
> > > >
> > > > Just let me know what questions you still have.
> > > >
> > > > Thanks,
> > > > John Halley Gotway
> > > >
> > > >
> > > > On Fri, Jan 20, 2017 at 11:41 AM, Jared Lee via RT <
> met_help at ucar.edu>
> > > > wrote:
> > > >
> > > > >
> > > > > Fri Jan 20 11:41:03 2017: Request 79224 was acted upon.
> > > > > Transaction: Ticket created by jaredlee at ucar.edu
> > > > >        Queue: met_help
> > > > >      Subject: calculating ensemble spread
> > > > >        Owner: Nobody
> > > > >   Requestors: jaredlee at ucar.edu
> > > > >       Status: new
> > > > >  Ticket <URL: https://rt.rap.ucar.edu/rt/
> > Ticket/Display.html?id=79224
> > > >
> > > > >
> > > > >
> > > > > Hi, for my project I need to calculate some measure of the
spread
> of
> > > two
> > > > > different ensembles (one "large" and one "small" ensemble).
I have
> > > model
> > > > > output data (the large ensemble is in grib2 format, the
small
> > ensemble
> > > is
> > > > > WRF NetCDF), and I have MADIS point observations that have
already
> > been
> > > > > processed for use in MET.
> > > > >
> > > > > I previously used the ensemble-stat tool on the large
ensemble to
> > dump
> > > > out
> > > > > matched pair info for all ensemble member forecasts at each
> > observation
> > > > > location by printing out the ORANK line type. Given the
inputs
> that I
> > > > > described above, will MET calculate the ensemble spread
> automatically
> > > if
> > > > I
> > > > > turn on the SSVAR line type? In the user guide (v5.2) it
mentions
> > that
> > > an
> > > > > "observation field" is needed to calculate things like
spread/skill
> > > > > variance. Do point obs from MADIS count as an "observation
field"?
> > > > >
> > > > > Guidance would be much appreciated!
> > > > >
> > > > > Jared
> > > > >
> > > > > --
> > > > > ===============================
> > > > > Jared A. Lee, Ph.D.
> > > > > Project Scientist I
> > > > > Research Applications Laboratory
> > > > > National Center for Atmospheric Research
> > > > > Boulder, Colorado, USA
> > > > >
> > > > > Email: jaredlee at ucar.edu (w)
> > > > > Phone: 303.497.8485 (w)
> > > > > Web: https://staff.ucar.edu/users/jaredlee
> > > > > ===============================
> > > > >
> > > > >
> > > >
> > > >
> > >
> > >
> > > --
> > > ===============================
> > > Jared A. Lee, Ph.D.
> > > Project Scientist I
> > > Research Applications Laboratory
> > > National Center for Atmospheric Research
> > > Boulder, Colorado, USA
> > >
> > > Email: jaredlee at ucar.edu (w)
> > > Phone: 303.497.8485 (w)
> > > Web: https://staff.ucar.edu/users/jaredlee
> > > ===============================
> > >
> > >
> >
> >
>
>
> --
> ===============================
> Jared A. Lee, Ph.D.
> Project Scientist I
> Research Applications Laboratory
> National Center for Atmospheric Research
> Boulder, Colorado, USA
>
> Email: jaredlee at ucar.edu (w)
> Phone: 303.497.8485 (w)
> Web: https://staff.ucar.edu/users/jaredlee
> ===============================
>
>

------------------------------------------------
Subject: calculating ensemble spread
From: Jared Lee
Time: Fri Feb 03 21:22:25 2017

Hi John,

I was unaware that UPP had a script to do that. As it turns out, some
members had all 48 hours in one file, and some had them split between
two
different files, so as it turns out it probably was simpler to break
them
up first with ncks anyway. In the future I'll have the person running
WRF
to output single-time files from the get-go, so now my scripts will
work
for that.

Does my laptop 'shadowfax' count as a machine within the RAL firewall
if
I'm physically at NCAR? What if I use VPN when I'm offsite? Otherwise
I
could try to find some other machine to temporarily house my data for
METViewer to access.

Jared

On Fri, Feb 3, 2017 at 4:36 PM, John Halley Gotway via RT
<met_help at ucar.edu
> wrote:

> Hi Jared,
>
> I'm surprised you need to use ncks to split out multiple times.
Isn't
> there a script for running UPP on a wrfout file with multiple times?
I
> believe it's called run_unipost_frames.
>
> As for your question about METViewer... you'd need to pull the
> ensemble-stat data down to a machine inside the RAL firewall.  And
then I
> could demonstrate to you how to load it into METViewer and start
creating
> plots.
>
> Unfortunately, you can't load it directly from yellowstone into
METViewer,
> which lives on a machine named "mandan".
>
> John
>
> On Fri, Feb 3, 2017 at 3:09 PM, Jared Lee via RT <met_help at ucar.edu>
> wrote:
>
> >
> > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
> >
> > Hi John,
> >
> > Now I finally have finished running UPP (after having to split
apart the
> > wrfout files into single-time files with ncks) and ensemble-stat
on all
> the
> > model data for my three ensembles. How do I go about getting that
into
> > METViewer so that I can plot spread-skill information? My data is
on
> > yellowstone. I also looked through the stat-analysis section of
the MET
> > Users Guide, but I'm not clear on what exactly I'd need to do to
get
> what I
> > need with that...
> >
> > Jared
> >
> > On Mon, Jan 23, 2017 at 5:01 PM, John Halley Gotway via RT <
> > met_help at ucar.edu> wrote:
> >
> > > Jared,
> > >
> > > Unfortunately, no, MET is not set up to read wrfout files
directly.  As
> > for
> > > post-processing, you have 2 options... either UPP or wrf_interp
(only
> for
> > > WRF-ARW core).
> > >
> > > The former writes GRIB output which MET handles well. The latter
> writes a
> > > flavor of NetCDF that MET is able to read... with the caveat
that it
> can
> > > only read fields that are defined on non-staggered dimensions.
> > Basically,
> > > it can't read the wind fields.
> > >
> > > I always encourage people to use UPP since MET supports GRIB
well.
> > >
> > > Being able to read wrfout directly has often been requested.
The two
> > > issues are the staggered dimensions and the hybrid vertical
coordinate.
> > > Generally, matching to point observations is done in terms of
pressure
> or
> > > height.  So some interpolation to pressure/height levels in
inevitable
> > > along the way.  We've resisted re-implementing that post-
processing
> logic
> > > directly in MET to avoid duplication of effort with UPP.  And
UPP
> > supports
> > > the derivation of many other variable types that are pretty
useful.
> > >
> > > Perhaps there is some middle ground where we could enable MET to
call
> UPP
> > > behind the scenes?  But we'd need to work out many details.
> > >
> > > Thanks,
> > > John
> > >
> > > On Fri, Jan 20, 2017 at 5:31 PM, Jared Lee via RT
<met_help at ucar.edu>
> > > wrote:
> > >
> > > >
> > > > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224
>
> > > >
> > > > Hi John,
> > > >
> > > > Thanks for clarifying what ensemble-stat does to produce the
SSVAR
> line
> > > > type. That helps.
> > > >
> > > > As for what I ultimately want, I guess that's still somewhat
vaguely
> > > > defined. I think making plots that show the spread/skill
relationship
> > > > across the date ranges, lead times, and variables of interest
for
> both
> > > > ensembles would be a useful way to illustrate and compare the
spread
> of
> > > the
> > > > two ensembles. That seems like it would be more useful than
simply
> > > > calculating the mean spread of the ensemble (which I could do
by
> > writing
> > > a
> > > > new NCL script to read in all the ensemble fields and not even
> > bothering
> > > > with MET...).
> > > >
> > > > Another side question... Can MET read in native-grid NetCDF
WRF
> output
> > > > files? Or do I need to run them through some pre-processor
(like UPP)
> > > > first? After going through the user guide I'm not clear
whether MET
> > > accepts
> > > > NetCDF wrfout files or not without pre-processing. Also, if
all the
> > > output
> > > > times are in a single wrfout file, can MET open the one file
and find
> > the
> > > > correct time it's supposed to read, or do all the wrfout files
need
> to
> > be
> > > > for a single time only? The NetCDF wrfout files I've been
given for
> the
> > > > small ensemble have all the output times in a single file, but
I can
> > > write
> > > > a script to parse them out to single-time files with ncks. I'd
just
> > > rather
> > > > not go through that hassle if I don't have to!
> > > >
> > > > Jared
> > > >
> > > > On Fri, Jan 20, 2017 at 5:07 PM, John Halley Gotway via RT <
> > > > met_help at ucar.edu> wrote:
> > > >
> > > > > Hi Jared,
> > > > >
> > > > > I see a couple of questions in your email.
> > > > >
> > > > > First, yes, Ensemble-Stat is set up to process both point
and
> gridded
> > > > > observations.  And the SSVAR (spread/skill variance) line
type can
> be
> > > > > produced using either of these.
> > > > >
> > > > > The SSVAR line type stores information about the
relationship
> between
> > > > > ensemble spread and skill.  Here's how it works.  Ensemble-
Stat
> > > processes
> > > > > each observation value... either from a point location (i.e.
for
> > point
> > > > obs)
> > > > > or from a gridded location (i.e. gridded obs).  For each
> observation,
> > > it
> > > > > gathers the N ensemble forecast values which correspond to
it.  It
> > > > computes
> > > > > the variance of those ensemble values, and it also computes
the
> error
> > > by
> > > > > subtracting the mean ensemble value minus the observation
value.
> > > (Note,
> > > > > you can pass in an ensemble mean field instead of computing
it on
> the
> > > fly
> > > > > like this).  After processing N observations, we have N
pairs of
> > > variance
> > > > > values and error values.  It groups those N pairs into
buckets by
> > > binning
> > > > > the variance using the bucket size defined in the config
file.
> > You'll
> > > > see
> > > > > in the output one SSVAR line for each of those variance
buckets.
> > > > >
> > > > > The METViewer database and display system (or Stat-Analysis)
can
> > > > aggregate
> > > > > those SSVAR lines intelligently across multiple cases and
create
> > plots
> > > > > which show the relationship between spread and skill.
> > > > >
> > > > > If that's what you need, great.
> > > > >
> > > > > However, if your looking for a much simpler computation of
mean
> > spread
> > > > over
> > > > > some spatial region, unfortunately, the answer is no.  For
example,
> > > > > Ensemble-Stat doesn't write out the mean spread value over
the
> entire
> > > > > domain.
> > > > >
> > > > > Just let me know what questions you still have.
> > > > >
> > > > > Thanks,
> > > > > John Halley Gotway
> > > > >
> > > > >
> > > > > On Fri, Jan 20, 2017 at 11:41 AM, Jared Lee via RT <
> > met_help at ucar.edu>
> > > > > wrote:
> > > > >
> > > > > >
> > > > > > Fri Jan 20 11:41:03 2017: Request 79224 was acted upon.
> > > > > > Transaction: Ticket created by jaredlee at ucar.edu
> > > > > >        Queue: met_help
> > > > > >      Subject: calculating ensemble spread
> > > > > >        Owner: Nobody
> > > > > >   Requestors: jaredlee at ucar.edu
> > > > > >       Status: new
> > > > > >  Ticket <URL: https://rt.rap.ucar.edu/rt/
> > > Ticket/Display.html?id=79224
> > > > >
> > > > > >
> > > > > >
> > > > > > Hi, for my project I need to calculate some measure of the
spread
> > of
> > > > two
> > > > > > different ensembles (one "large" and one "small"
ensemble). I
> have
> > > > model
> > > > > > output data (the large ensemble is in grib2 format, the
small
> > > ensemble
> > > > is
> > > > > > WRF NetCDF), and I have MADIS point observations that have
> already
> > > been
> > > > > > processed for use in MET.
> > > > > >
> > > > > > I previously used the ensemble-stat tool on the large
ensemble to
> > > dump
> > > > > out
> > > > > > matched pair info for all ensemble member forecasts at
each
> > > observation
> > > > > > location by printing out the ORANK line type. Given the
inputs
> > that I
> > > > > > described above, will MET calculate the ensemble spread
> > automatically
> > > > if
> > > > > I
> > > > > > turn on the SSVAR line type? In the user guide (v5.2) it
mentions
> > > that
> > > > an
> > > > > > "observation field" is needed to calculate things like
> spread/skill
> > > > > > variance. Do point obs from MADIS count as an "observation
> field"?
> > > > > >
> > > > > > Guidance would be much appreciated!
> > > > > >
> > > > > > Jared
> > > > > >
> > > > > > --
> > > > > > ===============================
> > > > > > Jared A. Lee, Ph.D.
> > > > > > Project Scientist I
> > > > > > Research Applications Laboratory
> > > > > > National Center for Atmospheric Research
> > > > > > Boulder, Colorado, USA
> > > > > >
> > > > > > Email: jaredlee at ucar.edu (w)
> > > > > > Phone: 303.497.8485 (w)
> > > > > > Web: https://staff.ucar.edu/users/jaredlee
> > > > > > ===============================
> > > > > >
> > > > > >
> > > > >
> > > > >
> > > >
> > > >
> > > > --
> > > > ===============================
> > > > Jared A. Lee, Ph.D.
> > > > Project Scientist I
> > > > Research Applications Laboratory
> > > > National Center for Atmospheric Research
> > > > Boulder, Colorado, USA
> > > >
> > > > Email: jaredlee at ucar.edu (w)
> > > > Phone: 303.497.8485 (w)
> > > > Web: https://staff.ucar.edu/users/jaredlee
> > > > ===============================
> > > >
> > > >
> > >
> > >
> >
> >
> > --
> > ===============================
> > Jared A. Lee, Ph.D.
> > Project Scientist I
> > Research Applications Laboratory
> > National Center for Atmospheric Research
> > Boulder, Colorado, USA
> >
> > Email: jaredlee at ucar.edu (w)
> > Phone: 303.497.8485 (w)
> > Web: https://staff.ucar.edu/users/jaredlee
> > ===============================
> >
> >
>
>


--
===============================
Jared A. Lee, Ph.D.
Project Scientist I
Research Applications Laboratory
National Center for Atmospheric Research
Boulder, Colorado, USA

Email: jaredlee at ucar.edu (w)
Phone: 303.497.8485 (w)
Web: https://staff.ucar.edu/users/jaredlee
===============================

------------------------------------------------
Subject: calculating ensemble spread
From: John Halley Gotway
Time: Sat Feb 04 18:47:47 2017

Jared,

Just come by my office sometime next week and we'll copy your data
down to
a project machine and load it into METViewer.

My calendar should be up to date if you want to schedule something
through
google calendar.

John

On Fri, Feb 3, 2017 at 9:22 PM Jared Lee via RT <met_help at ucar.edu>
wrote:

>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
>
> Hi John,
>
> I was unaware that UPP had a script to do that. As it turns out,
some
> members had all 48 hours in one file, and some had them split
between two
> different files, so as it turns out it probably was simpler to break
them
> up first with ncks anyway. In the future I'll have the person
running WRF
> to output single-time files from the get-go, so now my scripts will
work
> for that.
>
> Does my laptop 'shadowfax' count as a machine within the RAL
firewall if
> I'm physically at NCAR? What if I use VPN when I'm offsite?
Otherwise I
> could try to find some other machine to temporarily house my data
for
> METViewer to access.
>
> Jared
>
> On Fri, Feb 3, 2017 at 4:36 PM, John Halley Gotway via RT <
> met_help at ucar.edu
> > wrote:
>
> > Hi Jared,
> >
> > I'm surprised you need to use ncks to split out multiple times.
Isn't
> > there a script for running UPP on a wrfout file with multiple
times?  I
> > believe it's called run_unipost_frames.
> >
> > As for your question about METViewer... you'd need to pull the
> > ensemble-stat data down to a machine inside the RAL firewall.  And
then I
> > could demonstrate to you how to load it into METViewer and start
creating
> > plots.
> >
> > Unfortunately, you can't load it directly from yellowstone into
> METViewer,
> > which lives on a machine named "mandan".
> >
> > John
> >
> > On Fri, Feb 3, 2017 at 3:09 PM, Jared Lee via RT
<met_help at ucar.edu>
> > wrote:
> >
> > >
> > > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
> > >
> > > Hi John,
> > >
> > > Now I finally have finished running UPP (after having to split
apart
> the
> > > wrfout files into single-time files with ncks) and ensemble-stat
on all
> > the
> > > model data for my three ensembles. How do I go about getting
that into
> > > METViewer so that I can plot spread-skill information? My data
is on
> > > yellowstone. I also looked through the stat-analysis section of
the MET
> > > Users Guide, but I'm not clear on what exactly I'd need to do to
get
> > what I
> > > need with that...
> > >
> > > Jared
> > >
> > > On Mon, Jan 23, 2017 at 5:01 PM, John Halley Gotway via RT <
> > > met_help at ucar.edu> wrote:
> > >
> > > > Jared,
> > > >
> > > > Unfortunately, no, MET is not set up to read wrfout files
directly.
> As
> > > for
> > > > post-processing, you have 2 options... either UPP or
wrf_interp (only
> > for
> > > > WRF-ARW core).
> > > >
> > > > The former writes GRIB output which MET handles well. The
latter
> > writes a
> > > > flavor of NetCDF that MET is able to read... with the caveat
that it
> > can
> > > > only read fields that are defined on non-staggered dimensions.
> > > Basically,
> > > > it can't read the wind fields.
> > > >
> > > > I always encourage people to use UPP since MET supports GRIB
well.
> > > >
> > > > Being able to read wrfout directly has often been requested.
The two
> > > > issues are the staggered dimensions and the hybrid vertical
> coordinate.
> > > > Generally, matching to point observations is done in terms of
> pressure
> > or
> > > > height.  So some interpolation to pressure/height levels in
> inevitable
> > > > along the way.  We've resisted re-implementing that post-
processing
> > logic
> > > > directly in MET to avoid duplication of effort with UPP.  And
UPP
> > > supports
> > > > the derivation of many other variable types that are pretty
useful.
> > > >
> > > > Perhaps there is some middle ground where we could enable MET
to call
> > UPP
> > > > behind the scenes?  But we'd need to work out many details.
> > > >
> > > > Thanks,
> > > > John
> > > >
> > > > On Fri, Jan 20, 2017 at 5:31 PM, Jared Lee via RT
<met_help at ucar.edu
> >
> > > > wrote:
> > > >
> > > > >
> > > > > <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
> > > > >
> > > > > Hi John,
> > > > >
> > > > > Thanks for clarifying what ensemble-stat does to produce the
SSVAR
> > line
> > > > > type. That helps.
> > > > >
> > > > > As for what I ultimately want, I guess that's still somewhat
> vaguely
> > > > > defined. I think making plots that show the spread/skill
> relationship
> > > > > across the date ranges, lead times, and variables of
interest for
> > both
> > > > > ensembles would be a useful way to illustrate and compare
the
> spread
> > of
> > > > the
> > > > > two ensembles. That seems like it would be more useful than
simply
> > > > > calculating the mean spread of the ensemble (which I could
do by
> > > writing
> > > > a
> > > > > new NCL script to read in all the ensemble fields and not
even
> > > bothering
> > > > > with MET...).
> > > > >
> > > > > Another side question... Can MET read in native-grid NetCDF
WRF
> > output
> > > > > files? Or do I need to run them through some pre-processor
(like
> UPP)
> > > > > first? After going through the user guide I'm not clear
whether MET
> > > > accepts
> > > > > NetCDF wrfout files or not without pre-processing. Also, if
all the
> > > > output
> > > > > times are in a single wrfout file, can MET open the one file
and
> find
> > > the
> > > > > correct time it's supposed to read, or do all the wrfout
files need
> > to
> > > be
> > > > > for a single time only? The NetCDF wrfout files I've been
given for
> > the
> > > > > small ensemble have all the output times in a single file,
but I
> can
> > > > write
> > > > > a script to parse them out to single-time files with ncks.
I'd just
> > > > rather
> > > > > not go through that hassle if I don't have to!
> > > > >
> > > > > Jared
> > > > >
> > > > > On Fri, Jan 20, 2017 at 5:07 PM, John Halley Gotway via RT <
> > > > > met_help at ucar.edu> wrote:
> > > > >
> > > > > > Hi Jared,
> > > > > >
> > > > > > I see a couple of questions in your email.
> > > > > >
> > > > > > First, yes, Ensemble-Stat is set up to process both point
and
> > gridded
> > > > > > observations.  And the SSVAR (spread/skill variance) line
type
> can
> > be
> > > > > > produced using either of these.
> > > > > >
> > > > > > The SSVAR line type stores information about the
relationship
> > between
> > > > > > ensemble spread and skill.  Here's how it works.
Ensemble-Stat
> > > > processes
> > > > > > each observation value... either from a point location
(i.e. for
> > > point
> > > > > obs)
> > > > > > or from a gridded location (i.e. gridded obs).  For each
> > observation,
> > > > it
> > > > > > gathers the N ensemble forecast values which correspond to
it.
> It
> > > > > computes
> > > > > > the variance of those ensemble values, and it also
computes the
> > error
> > > > by
> > > > > > subtracting the mean ensemble value minus the observation
value.
> > > > (Note,
> > > > > > you can pass in an ensemble mean field instead of
computing it on
> > the
> > > > fly
> > > > > > like this).  After processing N observations, we have N
pairs of
> > > > variance
> > > > > > values and error values.  It groups those N pairs into
buckets by
> > > > binning
> > > > > > the variance using the bucket size defined in the config
file.
> > > You'll
> > > > > see
> > > > > > in the output one SSVAR line for each of those variance
buckets.
> > > > > >
> > > > > > The METViewer database and display system (or Stat-
Analysis) can
> > > > > aggregate
> > > > > > those SSVAR lines intelligently across multiple cases and
create
> > > plots
> > > > > > which show the relationship between spread and skill.
> > > > > >
> > > > > > If that's what you need, great.
> > > > > >
> > > > > > However, if your looking for a much simpler computation of
mean
> > > spread
> > > > > over
> > > > > > some spatial region, unfortunately, the answer is no.  For
> example,
> > > > > > Ensemble-Stat doesn't write out the mean spread value over
the
> > entire
> > > > > > domain.
> > > > > >
> > > > > > Just let me know what questions you still have.
> > > > > >
> > > > > > Thanks,
> > > > > > John Halley Gotway
> > > > > >
> > > > > >
> > > > > > On Fri, Jan 20, 2017 at 11:41 AM, Jared Lee via RT <
> > > met_help at ucar.edu>
> > > > > > wrote:
> > > > > >
> > > > > > >
> > > > > > > Fri Jan 20 11:41:03 2017: Request 79224 was acted upon.
> > > > > > > Transaction: Ticket created by jaredlee at ucar.edu
> > > > > > >        Queue: met_help
> > > > > > >      Subject: calculating ensemble spread
> > > > > > >        Owner: Nobody
> > > > > > >   Requestors: jaredlee at ucar.edu
> > > > > > >       Status: new
> > > > > > >  Ticket <URL: https://rt.rap.ucar.edu/rt/
> > > > Ticket/Display.html?id=79224
> > > > > >
> > > > > > >
> > > > > > >
> > > > > > > Hi, for my project I need to calculate some measure of
the
> spread
> > > of
> > > > > two
> > > > > > > different ensembles (one "large" and one "small"
ensemble). I
> > have
> > > > > model
> > > > > > > output data (the large ensemble is in grib2 format, the
small
> > > > ensemble
> > > > > is
> > > > > > > WRF NetCDF), and I have MADIS point observations that
have
> > already
> > > > been
> > > > > > > processed for use in MET.
> > > > > > >
> > > > > > > I previously used the ensemble-stat tool on the large
ensemble
> to
> > > > dump
> > > > > > out
> > > > > > > matched pair info for all ensemble member forecasts at
each
> > > > observation
> > > > > > > location by printing out the ORANK line type. Given the
inputs
> > > that I
> > > > > > > described above, will MET calculate the ensemble spread
> > > automatically
> > > > > if
> > > > > > I
> > > > > > > turn on the SSVAR line type? In the user guide (v5.2) it
> mentions
> > > > that
> > > > > an
> > > > > > > "observation field" is needed to calculate things like
> > spread/skill
> > > > > > > variance. Do point obs from MADIS count as an
"observation
> > field"?
> > > > > > >
> > > > > > > Guidance would be much appreciated!
> > > > > > >
> > > > > > > Jared
> > > > > > >
> > > > > > > --
> > > > > > > ===============================
> > > > > > > Jared A. Lee, Ph.D.
> > > > > > > Project Scientist I
> > > > > > > Research Applications Laboratory
> > > > > > > National Center for Atmospheric Research
> > > > > > > Boulder, Colorado, USA
> > > > > > >
> > > > > > > Email: jaredlee at ucar.edu (w)
> > > > > > > Phone: 303.497.8485 (w)
> > > > > > > Web: https://staff.ucar.edu/users/jaredlee
> > > > > > > ===============================
> > > > > > >
> > > > > > >
> > > > > >
> > > > > >
> > > > >
> > > > >
> > > > > --
> > > > > ===============================
> > > > > Jared A. Lee, Ph.D.
> > > > > Project Scientist I
> > > > > Research Applications Laboratory
> > > > > National Center for Atmospheric Research
> > > > > Boulder, Colorado, USA
> > > > >
> > > > > Email: jaredlee at ucar.edu (w)
> > > > > Phone: 303.497.8485 (w)
> > > > > Web: https://staff.ucar.edu/users/jaredlee
> > > > > ===============================
> > > > >
> > > > >
> > > >
> > > >
> > >
> > >
> > > --
> > > ===============================
> > > Jared A. Lee, Ph.D.
> > > Project Scientist I
> > > Research Applications Laboratory
> > > National Center for Atmospheric Research
> > > Boulder, Colorado, USA
> > >
> > > Email: jaredlee at ucar.edu (w)
> > > Phone: 303.497.8485 (w)
> > > Web: https://staff.ucar.edu/users/jaredlee
> > > ===============================
> > >
> > >
> >
> >
>
>
> --
> ===============================
> Jared A. Lee, Ph.D.
> Project Scientist I
> Research Applications Laboratory
> National Center for Atmospheric Research
> Boulder, Colorado, USA
>
> Email: jaredlee at ucar.edu (w)
> Phone: 303.497.8485 (w)
> Web: https://staff.ucar.edu/users/jaredlee
> ===============================
>
>

------------------------------------------------
Subject: calculating ensemble spread
From: Jared Lee
Time: Tue Feb 07 12:04:18 2017

Hi John,

I've finished re-generating all the .stat files (including with the
"model"
labels I want). The zipped file is 405 MB and is located here on
yellowstone:

/glade/p/ral/wsap/jaredlee/NOAA_AQ_Ensemble/MET/15day_data.tar.gz

Could you please replace all the data in the "noaa_aq_15day" METViewer
database with that? Thanks!

Jared

On Sat, Feb 4, 2017 at 6:47 PM, John Halley Gotway via RT
<met_help at ucar.edu
> wrote:

> Jared,
>
> Just come by my office sometime next week and we'll copy your data
down to
> a project machine and load it into METViewer.
>
> My calendar should be up to date if you want to schedule something
through
> google calendar.
>
> John
>
> On Fri, Feb 3, 2017 at 9:22 PM Jared Lee via RT <met_help at ucar.edu>
wrote:
>
> >
> > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
> >
> > Hi John,
> >
> > I was unaware that UPP had a script to do that. As it turns out,
some
> > members had all 48 hours in one file, and some had them split
between two
> > different files, so as it turns out it probably was simpler to
break them
> > up first with ncks anyway. In the future I'll have the person
running WRF
> > to output single-time files from the get-go, so now my scripts
will work
> > for that.
> >
> > Does my laptop 'shadowfax' count as a machine within the RAL
firewall if
> > I'm physically at NCAR? What if I use VPN when I'm offsite?
Otherwise I
> > could try to find some other machine to temporarily house my data
for
> > METViewer to access.
> >
> > Jared
> >
> > On Fri, Feb 3, 2017 at 4:36 PM, John Halley Gotway via RT <
> > met_help at ucar.edu
> > > wrote:
> >
> > > Hi Jared,
> > >
> > > I'm surprised you need to use ncks to split out multiple times.
Isn't
> > > there a script for running UPP on a wrfout file with multiple
times?  I
> > > believe it's called run_unipost_frames.
> > >
> > > As for your question about METViewer... you'd need to pull the
> > > ensemble-stat data down to a machine inside the RAL firewall.
And
> then I
> > > could demonstrate to you how to load it into METViewer and start
> creating
> > > plots.
> > >
> > > Unfortunately, you can't load it directly from yellowstone into
> > METViewer,
> > > which lives on a machine named "mandan".
> > >
> > > John
> > >
> > > On Fri, Feb 3, 2017 at 3:09 PM, Jared Lee via RT
<met_help at ucar.edu>
> > > wrote:
> > >
> > > >
> > > > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224
>
> > > >
> > > > Hi John,
> > > >
> > > > Now I finally have finished running UPP (after having to split
apart
> > the
> > > > wrfout files into single-time files with ncks) and ensemble-
stat on
> all
> > > the
> > > > model data for my three ensembles. How do I go about getting
that
> into
> > > > METViewer so that I can plot spread-skill information? My data
is on
> > > > yellowstone. I also looked through the stat-analysis section
of the
> MET
> > > > Users Guide, but I'm not clear on what exactly I'd need to do
to get
> > > what I
> > > > need with that...
> > > >
> > > > Jared
> > > >
> > > > On Mon, Jan 23, 2017 at 5:01 PM, John Halley Gotway via RT <
> > > > met_help at ucar.edu> wrote:
> > > >
> > > > > Jared,
> > > > >
> > > > > Unfortunately, no, MET is not set up to read wrfout files
directly.
> > As
> > > > for
> > > > > post-processing, you have 2 options... either UPP or
wrf_interp
> (only
> > > for
> > > > > WRF-ARW core).
> > > > >
> > > > > The former writes GRIB output which MET handles well. The
latter
> > > writes a
> > > > > flavor of NetCDF that MET is able to read... with the caveat
that
> it
> > > can
> > > > > only read fields that are defined on non-staggered
dimensions.
> > > > Basically,
> > > > > it can't read the wind fields.
> > > > >
> > > > > I always encourage people to use UPP since MET supports GRIB
well.
> > > > >
> > > > > Being able to read wrfout directly has often been requested.
The
> two
> > > > > issues are the staggered dimensions and the hybrid vertical
> > coordinate.
> > > > > Generally, matching to point observations is done in terms
of
> > pressure
> > > or
> > > > > height.  So some interpolation to pressure/height levels in
> > inevitable
> > > > > along the way.  We've resisted re-implementing that post-
processing
> > > logic
> > > > > directly in MET to avoid duplication of effort with UPP.
And UPP
> > > > supports
> > > > > the derivation of many other variable types that are pretty
useful.
> > > > >
> > > > > Perhaps there is some middle ground where we could enable
MET to
> call
> > > UPP
> > > > > behind the scenes?  But we'd need to work out many details.
> > > > >
> > > > > Thanks,
> > > > > John
> > > > >
> > > > > On Fri, Jan 20, 2017 at 5:31 PM, Jared Lee via RT <
> met_help at ucar.edu
> > >
> > > > > wrote:
> > > > >
> > > > > >
> > > > > > <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
> > > > > >
> > > > > > Hi John,
> > > > > >
> > > > > > Thanks for clarifying what ensemble-stat does to produce
the
> SSVAR
> > > line
> > > > > > type. That helps.
> > > > > >
> > > > > > As for what I ultimately want, I guess that's still
somewhat
> > vaguely
> > > > > > defined. I think making plots that show the spread/skill
> > relationship
> > > > > > across the date ranges, lead times, and variables of
interest for
> > > both
> > > > > > ensembles would be a useful way to illustrate and compare
the
> > spread
> > > of
> > > > > the
> > > > > > two ensembles. That seems like it would be more useful
than
> simply
> > > > > > calculating the mean spread of the ensemble (which I could
do by
> > > > writing
> > > > > a
> > > > > > new NCL script to read in all the ensemble fields and not
even
> > > > bothering
> > > > > > with MET...).
> > > > > >
> > > > > > Another side question... Can MET read in native-grid
NetCDF WRF
> > > output
> > > > > > files? Or do I need to run them through some pre-processor
(like
> > UPP)
> > > > > > first? After going through the user guide I'm not clear
whether
> MET
> > > > > accepts
> > > > > > NetCDF wrfout files or not without pre-processing. Also,
if all
> the
> > > > > output
> > > > > > times are in a single wrfout file, can MET open the one
file and
> > find
> > > > the
> > > > > > correct time it's supposed to read, or do all the wrfout
files
> need
> > > to
> > > > be
> > > > > > for a single time only? The NetCDF wrfout files I've been
given
> for
> > > the
> > > > > > small ensemble have all the output times in a single file,
but I
> > can
> > > > > write
> > > > > > a script to parse them out to single-time files with ncks.
I'd
> just
> > > > > rather
> > > > > > not go through that hassle if I don't have to!
> > > > > >
> > > > > > Jared
> > > > > >
> > > > > > On Fri, Jan 20, 2017 at 5:07 PM, John Halley Gotway via RT
<
> > > > > > met_help at ucar.edu> wrote:
> > > > > >
> > > > > > > Hi Jared,
> > > > > > >
> > > > > > > I see a couple of questions in your email.
> > > > > > >
> > > > > > > First, yes, Ensemble-Stat is set up to process both
point and
> > > gridded
> > > > > > > observations.  And the SSVAR (spread/skill variance)
line type
> > can
> > > be
> > > > > > > produced using either of these.
> > > > > > >
> > > > > > > The SSVAR line type stores information about the
relationship
> > > between
> > > > > > > ensemble spread and skill.  Here's how it works.
Ensemble-Stat
> > > > > processes
> > > > > > > each observation value... either from a point location
(i.e.
> for
> > > > point
> > > > > > obs)
> > > > > > > or from a gridded location (i.e. gridded obs).  For each
> > > observation,
> > > > > it
> > > > > > > gathers the N ensemble forecast values which correspond
to it.
> > It
> > > > > > computes
> > > > > > > the variance of those ensemble values, and it also
computes the
> > > error
> > > > > by
> > > > > > > subtracting the mean ensemble value minus the
observation
> value.
> > > > > (Note,
> > > > > > > you can pass in an ensemble mean field instead of
computing it
> on
> > > the
> > > > > fly
> > > > > > > like this).  After processing N observations, we have N
pairs
> of
> > > > > variance
> > > > > > > values and error values.  It groups those N pairs into
buckets
> by
> > > > > binning
> > > > > > > the variance using the bucket size defined in the config
file.
> > > > You'll
> > > > > > see
> > > > > > > in the output one SSVAR line for each of those variance
> buckets.
> > > > > > >
> > > > > > > The METViewer database and display system (or Stat-
Analysis)
> can
> > > > > > aggregate
> > > > > > > those SSVAR lines intelligently across multiple cases
and
> create
> > > > plots
> > > > > > > which show the relationship between spread and skill.
> > > > > > >
> > > > > > > If that's what you need, great.
> > > > > > >
> > > > > > > However, if your looking for a much simpler computation
of mean
> > > > spread
> > > > > > over
> > > > > > > some spatial region, unfortunately, the answer is no.
For
> > example,
> > > > > > > Ensemble-Stat doesn't write out the mean spread value
over the
> > > entire
> > > > > > > domain.
> > > > > > >
> > > > > > > Just let me know what questions you still have.
> > > > > > >
> > > > > > > Thanks,
> > > > > > > John Halley Gotway
> > > > > > >
> > > > > > >
> > > > > > > On Fri, Jan 20, 2017 at 11:41 AM, Jared Lee via RT <
> > > > met_help at ucar.edu>
> > > > > > > wrote:
> > > > > > >
> > > > > > > >
> > > > > > > > Fri Jan 20 11:41:03 2017: Request 79224 was acted
upon.
> > > > > > > > Transaction: Ticket created by jaredlee at ucar.edu
> > > > > > > >        Queue: met_help
> > > > > > > >      Subject: calculating ensemble spread
> > > > > > > >        Owner: Nobody
> > > > > > > >   Requestors: jaredlee at ucar.edu
> > > > > > > >       Status: new
> > > > > > > >  Ticket <URL: https://rt.rap.ucar.edu/rt/
> > > > > Ticket/Display.html?id=79224
> > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > > > Hi, for my project I need to calculate some measure of
the
> > spread
> > > > of
> > > > > > two
> > > > > > > > different ensembles (one "large" and one "small"
ensemble). I
> > > have
> > > > > > model
> > > > > > > > output data (the large ensemble is in grib2 format,
the small
> > > > > ensemble
> > > > > > is
> > > > > > > > WRF NetCDF), and I have MADIS point observations that
have
> > > already
> > > > > been
> > > > > > > > processed for use in MET.
> > > > > > > >
> > > > > > > > I previously used the ensemble-stat tool on the large
> ensemble
> > to
> > > > > dump
> > > > > > > out
> > > > > > > > matched pair info for all ensemble member forecasts at
each
> > > > > observation
> > > > > > > > location by printing out the ORANK line type. Given
the
> inputs
> > > > that I
> > > > > > > > described above, will MET calculate the ensemble
spread
> > > > automatically
> > > > > > if
> > > > > > > I
> > > > > > > > turn on the SSVAR line type? In the user guide (v5.2)
it
> > mentions
> > > > > that
> > > > > > an
> > > > > > > > "observation field" is needed to calculate things like
> > > spread/skill
> > > > > > > > variance. Do point obs from MADIS count as an
"observation
> > > field"?
> > > > > > > >
> > > > > > > > Guidance would be much appreciated!
> > > > > > > >
> > > > > > > > Jared
> > > > > > > >
> > > > > > > > --
> > > > > > > > ===============================
> > > > > > > > Jared A. Lee, Ph.D.
> > > > > > > > Project Scientist I
> > > > > > > > Research Applications Laboratory
> > > > > > > > National Center for Atmospheric Research
> > > > > > > > Boulder, Colorado, USA
> > > > > > > >
> > > > > > > > Email: jaredlee at ucar.edu (w)
> > > > > > > > Phone: 303.497.8485 (w)
> > > > > > > > Web: https://staff.ucar.edu/users/jaredlee
> > > > > > > > ===============================
> > > > > > > >
> > > > > > > >
> > > > > > >
> > > > > > >
> > > > > >
> > > > > >
> > > > > > --
> > > > > > ===============================
> > > > > > Jared A. Lee, Ph.D.
> > > > > > Project Scientist I
> > > > > > Research Applications Laboratory
> > > > > > National Center for Atmospheric Research
> > > > > > Boulder, Colorado, USA
> > > > > >
> > > > > > Email: jaredlee at ucar.edu (w)
> > > > > > Phone: 303.497.8485 (w)
> > > > > > Web: https://staff.ucar.edu/users/jaredlee
> > > > > > ===============================
> > > > > >
> > > > > >
> > > > >
> > > > >
> > > >
> > > >
> > > > --
> > > > ===============================
> > > > Jared A. Lee, Ph.D.
> > > > Project Scientist I
> > > > Research Applications Laboratory
> > > > National Center for Atmospheric Research
> > > > Boulder, Colorado, USA
> > > >
> > > > Email: jaredlee at ucar.edu (w)
> > > > Phone: 303.497.8485 (w)
> > > > Web: https://staff.ucar.edu/users/jaredlee
> > > > ===============================
> > > >
> > > >
> > >
> > >
> >
> >
> > --
> > ===============================
> > Jared A. Lee, Ph.D.
> > Project Scientist I
> > Research Applications Laboratory
> > National Center for Atmospheric Research
> > Boulder, Colorado, USA
> >
> > Email: jaredlee at ucar.edu (w)
> > Phone: 303.497.8485 (w)
> > Web: https://staff.ucar.edu/users/jaredlee
> > ===============================
> >
> >
>
>


--
===============================
Jared A. Lee, Ph.D.
Project Scientist I
Research Applications Laboratory
National Center for Atmospheric Research
Boulder, Colorado, USA

Email: jaredlee at ucar.edu (w)
Phone: 303.497.8485 (w)
Web: https://staff.ucar.edu/users/jaredlee
===============================

------------------------------------------------
Subject: calculating ensemble spread
From: John Halley Gotway
Time: Tue Feb 07 12:10:51 2017

Jared,

Will do.  Do I need to update the model names again?

SREF -> 3MEM
SREF -> ARW_ONLY

John

On Tue, Feb 7, 2017 at 12:04 PM, Jared Lee via RT <met_help at ucar.edu>
wrote:

>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
>
> Hi John,
>
> I've finished re-generating all the .stat files (including with the
"model"
> labels I want). The zipped file is 405 MB and is located here on
> yellowstone:
>
> /glade/p/ral/wsap/jaredlee/NOAA_AQ_Ensemble/MET/15day_data.tar.gz
>
> Could you please replace all the data in the "noaa_aq_15day"
METViewer
> database with that? Thanks!
>
> Jared
>
> On Sat, Feb 4, 2017 at 6:47 PM, John Halley Gotway via RT <
> met_help at ucar.edu
> > wrote:
>
> > Jared,
> >
> > Just come by my office sometime next week and we'll copy your data
down
> to
> > a project machine and load it into METViewer.
> >
> > My calendar should be up to date if you want to schedule something
> through
> > google calendar.
> >
> > John
> >
> > On Fri, Feb 3, 2017 at 9:22 PM Jared Lee via RT
<met_help at ucar.edu>
> wrote:
> >
> > >
> > > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
> > >
> > > Hi John,
> > >
> > > I was unaware that UPP had a script to do that. As it turns out,
some
> > > members had all 48 hours in one file, and some had them split
between
> two
> > > different files, so as it turns out it probably was simpler to
break
> them
> > > up first with ncks anyway. In the future I'll have the person
running
> WRF
> > > to output single-time files from the get-go, so now my scripts
will
> work
> > > for that.
> > >
> > > Does my laptop 'shadowfax' count as a machine within the RAL
firewall
> if
> > > I'm physically at NCAR? What if I use VPN when I'm offsite?
Otherwise I
> > > could try to find some other machine to temporarily house my
data for
> > > METViewer to access.
> > >
> > > Jared
> > >
> > > On Fri, Feb 3, 2017 at 4:36 PM, John Halley Gotway via RT <
> > > met_help at ucar.edu
> > > > wrote:
> > >
> > > > Hi Jared,
> > > >
> > > > I'm surprised you need to use ncks to split out multiple
times.
> Isn't
> > > > there a script for running UPP on a wrfout file with multiple
> times?  I
> > > > believe it's called run_unipost_frames.
> > > >
> > > > As for your question about METViewer... you'd need to pull the
> > > > ensemble-stat data down to a machine inside the RAL firewall.
And
> > then I
> > > > could demonstrate to you how to load it into METViewer and
start
> > creating
> > > > plots.
> > > >
> > > > Unfortunately, you can't load it directly from yellowstone
into
> > > METViewer,
> > > > which lives on a machine named "mandan".
> > > >
> > > > John
> > > >
> > > > On Fri, Feb 3, 2017 at 3:09 PM, Jared Lee via RT
<met_help at ucar.edu>
> > > > wrote:
> > > >
> > > > >
> > > > > <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
> > > > >
> > > > > Hi John,
> > > > >
> > > > > Now I finally have finished running UPP (after having to
split
> apart
> > > the
> > > > > wrfout files into single-time files with ncks) and ensemble-
stat on
> > all
> > > > the
> > > > > model data for my three ensembles. How do I go about getting
that
> > into
> > > > > METViewer so that I can plot spread-skill information? My
data is
> on
> > > > > yellowstone. I also looked through the stat-analysis section
of the
> > MET
> > > > > Users Guide, but I'm not clear on what exactly I'd need to
do to
> get
> > > > what I
> > > > > need with that...
> > > > >
> > > > > Jared
> > > > >
> > > > > On Mon, Jan 23, 2017 at 5:01 PM, John Halley Gotway via RT <
> > > > > met_help at ucar.edu> wrote:
> > > > >
> > > > > > Jared,
> > > > > >
> > > > > > Unfortunately, no, MET is not set up to read wrfout files
> directly.
> > > As
> > > > > for
> > > > > > post-processing, you have 2 options... either UPP or
wrf_interp
> > (only
> > > > for
> > > > > > WRF-ARW core).
> > > > > >
> > > > > > The former writes GRIB output which MET handles well. The
latter
> > > > writes a
> > > > > > flavor of NetCDF that MET is able to read... with the
caveat that
> > it
> > > > can
> > > > > > only read fields that are defined on non-staggered
dimensions.
> > > > > Basically,
> > > > > > it can't read the wind fields.
> > > > > >
> > > > > > I always encourage people to use UPP since MET supports
GRIB
> well.
> > > > > >
> > > > > > Being able to read wrfout directly has often been
requested.  The
> > two
> > > > > > issues are the staggered dimensions and the hybrid
vertical
> > > coordinate.
> > > > > > Generally, matching to point observations is done in terms
of
> > > pressure
> > > > or
> > > > > > height.  So some interpolation to pressure/height levels
in
> > > inevitable
> > > > > > along the way.  We've resisted re-implementing that
> post-processing
> > > > logic
> > > > > > directly in MET to avoid duplication of effort with UPP.
And UPP
> > > > > supports
> > > > > > the derivation of many other variable types that are
pretty
> useful.
> > > > > >
> > > > > > Perhaps there is some middle ground where we could enable
MET to
> > call
> > > > UPP
> > > > > > behind the scenes?  But we'd need to work out many
details.
> > > > > >
> > > > > > Thanks,
> > > > > > John
> > > > > >
> > > > > > On Fri, Jan 20, 2017 at 5:31 PM, Jared Lee via RT <
> > met_help at ucar.edu
> > > >
> > > > > > wrote:
> > > > > >
> > > > > > >
> > > > > > > <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224
> >
> > > > > > >
> > > > > > > Hi John,
> > > > > > >
> > > > > > > Thanks for clarifying what ensemble-stat does to produce
the
> > SSVAR
> > > > line
> > > > > > > type. That helps.
> > > > > > >
> > > > > > > As for what I ultimately want, I guess that's still
somewhat
> > > vaguely
> > > > > > > defined. I think making plots that show the spread/skill
> > > relationship
> > > > > > > across the date ranges, lead times, and variables of
interest
> for
> > > > both
> > > > > > > ensembles would be a useful way to illustrate and
compare the
> > > spread
> > > > of
> > > > > > the
> > > > > > > two ensembles. That seems like it would be more useful
than
> > simply
> > > > > > > calculating the mean spread of the ensemble (which I
could do
> by
> > > > > writing
> > > > > > a
> > > > > > > new NCL script to read in all the ensemble fields and
not even
> > > > > bothering
> > > > > > > with MET...).
> > > > > > >
> > > > > > > Another side question... Can MET read in native-grid
NetCDF WRF
> > > > output
> > > > > > > files? Or do I need to run them through some pre-
processor
> (like
> > > UPP)
> > > > > > > first? After going through the user guide I'm not clear
whether
> > MET
> > > > > > accepts
> > > > > > > NetCDF wrfout files or not without pre-processing. Also,
if all
> > the
> > > > > > output
> > > > > > > times are in a single wrfout file, can MET open the one
file
> and
> > > find
> > > > > the
> > > > > > > correct time it's supposed to read, or do all the wrfout
files
> > need
> > > > to
> > > > > be
> > > > > > > for a single time only? The NetCDF wrfout files I've
been given
> > for
> > > > the
> > > > > > > small ensemble have all the output times in a single
file, but
> I
> > > can
> > > > > > write
> > > > > > > a script to parse them out to single-time files with
ncks. I'd
> > just
> > > > > > rather
> > > > > > > not go through that hassle if I don't have to!
> > > > > > >
> > > > > > > Jared
> > > > > > >
> > > > > > > On Fri, Jan 20, 2017 at 5:07 PM, John Halley Gotway via
RT <
> > > > > > > met_help at ucar.edu> wrote:
> > > > > > >
> > > > > > > > Hi Jared,
> > > > > > > >
> > > > > > > > I see a couple of questions in your email.
> > > > > > > >
> > > > > > > > First, yes, Ensemble-Stat is set up to process both
point and
> > > > gridded
> > > > > > > > observations.  And the SSVAR (spread/skill variance)
line
> type
> > > can
> > > > be
> > > > > > > > produced using either of these.
> > > > > > > >
> > > > > > > > The SSVAR line type stores information about the
relationship
> > > > between
> > > > > > > > ensemble spread and skill.  Here's how it works.
> Ensemble-Stat
> > > > > > processes
> > > > > > > > each observation value... either from a point location
(i.e.
> > for
> > > > > point
> > > > > > > obs)
> > > > > > > > or from a gridded location (i.e. gridded obs).  For
each
> > > > observation,
> > > > > > it
> > > > > > > > gathers the N ensemble forecast values which
correspond to
> it.
> > > It
> > > > > > > computes
> > > > > > > > the variance of those ensemble values, and it also
computes
> the
> > > > error
> > > > > > by
> > > > > > > > subtracting the mean ensemble value minus the
observation
> > value.
> > > > > > (Note,
> > > > > > > > you can pass in an ensemble mean field instead of
computing
> it
> > on
> > > > the
> > > > > > fly
> > > > > > > > like this).  After processing N observations, we have
N pairs
> > of
> > > > > > variance
> > > > > > > > values and error values.  It groups those N pairs into
> buckets
> > by
> > > > > > binning
> > > > > > > > the variance using the bucket size defined in the
config
> file.
> > > > > You'll
> > > > > > > see
> > > > > > > > in the output one SSVAR line for each of those
variance
> > buckets.
> > > > > > > >
> > > > > > > > The METViewer database and display system (or Stat-
Analysis)
> > can
> > > > > > > aggregate
> > > > > > > > those SSVAR lines intelligently across multiple cases
and
> > create
> > > > > plots
> > > > > > > > which show the relationship between spread and skill.
> > > > > > > >
> > > > > > > > If that's what you need, great.
> > > > > > > >
> > > > > > > > However, if your looking for a much simpler
computation of
> mean
> > > > > spread
> > > > > > > over
> > > > > > > > some spatial region, unfortunately, the answer is no.
For
> > > example,
> > > > > > > > Ensemble-Stat doesn't write out the mean spread value
over
> the
> > > > entire
> > > > > > > > domain.
> > > > > > > >
> > > > > > > > Just let me know what questions you still have.
> > > > > > > >
> > > > > > > > Thanks,
> > > > > > > > John Halley Gotway
> > > > > > > >
> > > > > > > >
> > > > > > > > On Fri, Jan 20, 2017 at 11:41 AM, Jared Lee via RT <
> > > > > met_help at ucar.edu>
> > > > > > > > wrote:
> > > > > > > >
> > > > > > > > >
> > > > > > > > > Fri Jan 20 11:41:03 2017: Request 79224 was acted
upon.
> > > > > > > > > Transaction: Ticket created by jaredlee at ucar.edu
> > > > > > > > >        Queue: met_help
> > > > > > > > >      Subject: calculating ensemble spread
> > > > > > > > >        Owner: Nobody
> > > > > > > > >   Requestors: jaredlee at ucar.edu
> > > > > > > > >       Status: new
> > > > > > > > >  Ticket <URL: https://rt.rap.ucar.edu/rt/
> > > > > > Ticket/Display.html?id=79224
> > > > > > > >
> > > > > > > > >
> > > > > > > > >
> > > > > > > > > Hi, for my project I need to calculate some measure
of the
> > > spread
> > > > > of
> > > > > > > two
> > > > > > > > > different ensembles (one "large" and one "small"
> ensemble). I
> > > > have
> > > > > > > model
> > > > > > > > > output data (the large ensemble is in grib2 format,
the
> small
> > > > > > ensemble
> > > > > > > is
> > > > > > > > > WRF NetCDF), and I have MADIS point observations
that have
> > > > already
> > > > > > been
> > > > > > > > > processed for use in MET.
> > > > > > > > >
> > > > > > > > > I previously used the ensemble-stat tool on the
large
> > ensemble
> > > to
> > > > > > dump
> > > > > > > > out
> > > > > > > > > matched pair info for all ensemble member forecasts
at each
> > > > > > observation
> > > > > > > > > location by printing out the ORANK line type. Given
the
> > inputs
> > > > > that I
> > > > > > > > > described above, will MET calculate the ensemble
spread
> > > > > automatically
> > > > > > > if
> > > > > > > > I
> > > > > > > > > turn on the SSVAR line type? In the user guide
(v5.2) it
> > > mentions
> > > > > > that
> > > > > > > an
> > > > > > > > > "observation field" is needed to calculate things
like
> > > > spread/skill
> > > > > > > > > variance. Do point obs from MADIS count as an
"observation
> > > > field"?
> > > > > > > > >
> > > > > > > > > Guidance would be much appreciated!
> > > > > > > > >
> > > > > > > > > Jared
> > > > > > > > >
> > > > > > > > > --
> > > > > > > > > ===============================
> > > > > > > > > Jared A. Lee, Ph.D.
> > > > > > > > > Project Scientist I
> > > > > > > > > Research Applications Laboratory
> > > > > > > > > National Center for Atmospheric Research
> > > > > > > > > Boulder, Colorado, USA
> > > > > > > > >
> > > > > > > > > Email: jaredlee at ucar.edu (w)
> > > > > > > > > Phone: 303.497.8485 (w)
> > > > > > > > > Web: https://staff.ucar.edu/users/jaredlee
> > > > > > > > > ===============================
> > > > > > > > >
> > > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > --
> > > > > > > ===============================
> > > > > > > Jared A. Lee, Ph.D.
> > > > > > > Project Scientist I
> > > > > > > Research Applications Laboratory
> > > > > > > National Center for Atmospheric Research
> > > > > > > Boulder, Colorado, USA
> > > > > > >
> > > > > > > Email: jaredlee at ucar.edu (w)
> > > > > > > Phone: 303.497.8485 (w)
> > > > > > > Web: https://staff.ucar.edu/users/jaredlee
> > > > > > > ===============================
> > > > > > >
> > > > > > >
> > > > > >
> > > > > >
> > > > >
> > > > >
> > > > > --
> > > > > ===============================
> > > > > Jared A. Lee, Ph.D.
> > > > > Project Scientist I
> > > > > Research Applications Laboratory
> > > > > National Center for Atmospheric Research
> > > > > Boulder, Colorado, USA
> > > > >
> > > > > Email: jaredlee at ucar.edu (w)
> > > > > Phone: 303.497.8485 (w)
> > > > > Web: https://staff.ucar.edu/users/jaredlee
> > > > > ===============================
> > > > >
> > > > >
> > > >
> > > >
> > >
> > >
> > > --
> > > ===============================
> > > Jared A. Lee, Ph.D.
> > > Project Scientist I
> > > Research Applications Laboratory
> > > National Center for Atmospheric Research
> > > Boulder, Colorado, USA
> > >
> > > Email: jaredlee at ucar.edu (w)
> > > Phone: 303.497.8485 (w)
> > > Web: https://staff.ucar.edu/users/jaredlee
> > > ===============================
> > >
> > >
> >
> >
>
>
> --
> ===============================
> Jared A. Lee, Ph.D.
> Project Scientist I
> Research Applications Laboratory
> National Center for Atmospheric Research
> Boulder, Colorado, USA
>
> Email: jaredlee at ucar.edu (w)
> Phone: 303.497.8485 (w)
> Web: https://staff.ucar.edu/users/jaredlee
> ===============================
>
>

------------------------------------------------
Subject: calculating ensemble spread
From: John Halley Gotway
Time: Tue Feb 07 12:12:05 2017

Sorry... should have checked first.  Looks like you already updated
them.

I'm reloading now.

On Tue, Feb 7, 2017 at 12:10 PM, John Halley Gotway <johnhg at ucar.edu>
wrote:

> Jared,
>
> Will do.  Do I need to update the model names again?
>
> SREF -> 3MEM
> SREF -> ARW_ONLY
>
> John
>
> On Tue, Feb 7, 2017 at 12:04 PM, Jared Lee via RT
<met_help at ucar.edu>
> wrote:
>
>>
>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
>>
>> Hi John,
>>
>> I've finished re-generating all the .stat files (including with the
>> "model"
>> labels I want). The zipped file is 405 MB and is located here on
>> yellowstone:
>>
>> /glade/p/ral/wsap/jaredlee/NOAA_AQ_Ensemble/MET/15day_data.tar.gz
>>
>> Could you please replace all the data in the "noaa_aq_15day"
METViewer
>> database with that? Thanks!
>>
>> Jared
>>
>> On Sat, Feb 4, 2017 at 6:47 PM, John Halley Gotway via RT <
>> met_help at ucar.edu
>> > wrote:
>>
>> > Jared,
>> >
>> > Just come by my office sometime next week and we'll copy your
data down
>> to
>> > a project machine and load it into METViewer.
>> >
>> > My calendar should be up to date if you want to schedule
something
>> through
>> > google calendar.
>> >
>> > John
>> >
>> > On Fri, Feb 3, 2017 at 9:22 PM Jared Lee via RT
<met_help at ucar.edu>
>> wrote:
>> >
>> > >
>> > > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
>> > >
>> > > Hi John,
>> > >
>> > > I was unaware that UPP had a script to do that. As it turns
out, some
>> > > members had all 48 hours in one file, and some had them split
between
>> two
>> > > different files, so as it turns out it probably was simpler to
break
>> them
>> > > up first with ncks anyway. In the future I'll have the person
running
>> WRF
>> > > to output single-time files from the get-go, so now my scripts
will
>> work
>> > > for that.
>> > >
>> > > Does my laptop 'shadowfax' count as a machine within the RAL
firewall
>> if
>> > > I'm physically at NCAR? What if I use VPN when I'm offsite?
Otherwise
>> I
>> > > could try to find some other machine to temporarily house my
data for
>> > > METViewer to access.
>> > >
>> > > Jared
>> > >
>> > > On Fri, Feb 3, 2017 at 4:36 PM, John Halley Gotway via RT <
>> > > met_help at ucar.edu
>> > > > wrote:
>> > >
>> > > > Hi Jared,
>> > > >
>> > > > I'm surprised you need to use ncks to split out multiple
times.
>> Isn't
>> > > > there a script for running UPP on a wrfout file with multiple
>> times?  I
>> > > > believe it's called run_unipost_frames.
>> > > >
>> > > > As for your question about METViewer... you'd need to pull
the
>> > > > ensemble-stat data down to a machine inside the RAL firewall.
And
>> > then I
>> > > > could demonstrate to you how to load it into METViewer and
start
>> > creating
>> > > > plots.
>> > > >
>> > > > Unfortunately, you can't load it directly from yellowstone
into
>> > > METViewer,
>> > > > which lives on a machine named "mandan".
>> > > >
>> > > > John
>> > > >
>> > > > On Fri, Feb 3, 2017 at 3:09 PM, Jared Lee via RT
<met_help at ucar.edu
>> >
>> > > > wrote:
>> > > >
>> > > > >
>> > > > > <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
>> > > > >
>> > > > > Hi John,
>> > > > >
>> > > > > Now I finally have finished running UPP (after having to
split
>> apart
>> > > the
>> > > > > wrfout files into single-time files with ncks) and
ensemble-stat
>> on
>> > all
>> > > > the
>> > > > > model data for my three ensembles. How do I go about
getting that
>> > into
>> > > > > METViewer so that I can plot spread-skill information? My
data is
>> on
>> > > > > yellowstone. I also looked through the stat-analysis
section of
>> the
>> > MET
>> > > > > Users Guide, but I'm not clear on what exactly I'd need to
do to
>> get
>> > > > what I
>> > > > > need with that...
>> > > > >
>> > > > > Jared
>> > > > >
>> > > > > On Mon, Jan 23, 2017 at 5:01 PM, John Halley Gotway via RT
<
>> > > > > met_help at ucar.edu> wrote:
>> > > > >
>> > > > > > Jared,
>> > > > > >
>> > > > > > Unfortunately, no, MET is not set up to read wrfout files
>> directly.
>> > > As
>> > > > > for
>> > > > > > post-processing, you have 2 options... either UPP or
wrf_interp
>> > (only
>> > > > for
>> > > > > > WRF-ARW core).
>> > > > > >
>> > > > > > The former writes GRIB output which MET handles well. The
latter
>> > > > writes a
>> > > > > > flavor of NetCDF that MET is able to read... with the
caveat
>> that
>> > it
>> > > > can
>> > > > > > only read fields that are defined on non-staggered
dimensions.
>> > > > > Basically,
>> > > > > > it can't read the wind fields.
>> > > > > >
>> > > > > > I always encourage people to use UPP since MET supports
GRIB
>> well.
>> > > > > >
>> > > > > > Being able to read wrfout directly has often been
requested.
>> The
>> > two
>> > > > > > issues are the staggered dimensions and the hybrid
vertical
>> > > coordinate.
>> > > > > > Generally, matching to point observations is done in
terms of
>> > > pressure
>> > > > or
>> > > > > > height.  So some interpolation to pressure/height levels
in
>> > > inevitable
>> > > > > > along the way.  We've resisted re-implementing that
>> post-processing
>> > > > logic
>> > > > > > directly in MET to avoid duplication of effort with UPP.
And
>> UPP
>> > > > > supports
>> > > > > > the derivation of many other variable types that are
pretty
>> useful.
>> > > > > >
>> > > > > > Perhaps there is some middle ground where we could enable
MET to
>> > call
>> > > > UPP
>> > > > > > behind the scenes?  But we'd need to work out many
details.
>> > > > > >
>> > > > > > Thanks,
>> > > > > > John
>> > > > > >
>> > > > > > On Fri, Jan 20, 2017 at 5:31 PM, Jared Lee via RT <
>> > met_help at ucar.edu
>> > > >
>> > > > > > wrote:
>> > > > > >
>> > > > > > >
>> > > > > > > <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224
>> >
>> > > > > > >
>> > > > > > > Hi John,
>> > > > > > >
>> > > > > > > Thanks for clarifying what ensemble-stat does to
produce the
>> > SSVAR
>> > > > line
>> > > > > > > type. That helps.
>> > > > > > >
>> > > > > > > As for what I ultimately want, I guess that's still
somewhat
>> > > vaguely
>> > > > > > > defined. I think making plots that show the
spread/skill
>> > > relationship
>> > > > > > > across the date ranges, lead times, and variables of
interest
>> for
>> > > > both
>> > > > > > > ensembles would be a useful way to illustrate and
compare the
>> > > spread
>> > > > of
>> > > > > > the
>> > > > > > > two ensembles. That seems like it would be more useful
than
>> > simply
>> > > > > > > calculating the mean spread of the ensemble (which I
could do
>> by
>> > > > > writing
>> > > > > > a
>> > > > > > > new NCL script to read in all the ensemble fields and
not even
>> > > > > bothering
>> > > > > > > with MET...).
>> > > > > > >
>> > > > > > > Another side question... Can MET read in native-grid
NetCDF
>> WRF
>> > > > output
>> > > > > > > files? Or do I need to run them through some pre-
processor
>> (like
>> > > UPP)
>> > > > > > > first? After going through the user guide I'm not clear
>> whether
>> > MET
>> > > > > > accepts
>> > > > > > > NetCDF wrfout files or not without pre-processing.
Also, if
>> all
>> > the
>> > > > > > output
>> > > > > > > times are in a single wrfout file, can MET open the one
file
>> and
>> > > find
>> > > > > the
>> > > > > > > correct time it's supposed to read, or do all the
wrfout files
>> > need
>> > > > to
>> > > > > be
>> > > > > > > for a single time only? The NetCDF wrfout files I've
been
>> given
>> > for
>> > > > the
>> > > > > > > small ensemble have all the output times in a single
file,
>> but I
>> > > can
>> > > > > > write
>> > > > > > > a script to parse them out to single-time files with
ncks. I'd
>> > just
>> > > > > > rather
>> > > > > > > not go through that hassle if I don't have to!
>> > > > > > >
>> > > > > > > Jared
>> > > > > > >
>> > > > > > > On Fri, Jan 20, 2017 at 5:07 PM, John Halley Gotway via
RT <
>> > > > > > > met_help at ucar.edu> wrote:
>> > > > > > >
>> > > > > > > > Hi Jared,
>> > > > > > > >
>> > > > > > > > I see a couple of questions in your email.
>> > > > > > > >
>> > > > > > > > First, yes, Ensemble-Stat is set up to process both
point
>> and
>> > > > gridded
>> > > > > > > > observations.  And the SSVAR (spread/skill variance)
line
>> type
>> > > can
>> > > > be
>> > > > > > > > produced using either of these.
>> > > > > > > >
>> > > > > > > > The SSVAR line type stores information about the
>> relationship
>> > > > between
>> > > > > > > > ensemble spread and skill.  Here's how it works.
>> Ensemble-Stat
>> > > > > > processes
>> > > > > > > > each observation value... either from a point
location (i.e.
>> > for
>> > > > > point
>> > > > > > > obs)
>> > > > > > > > or from a gridded location (i.e. gridded obs).  For
each
>> > > > observation,
>> > > > > > it
>> > > > > > > > gathers the N ensemble forecast values which
correspond to
>> it.
>> > > It
>> > > > > > > computes
>> > > > > > > > the variance of those ensemble values, and it also
computes
>> the
>> > > > error
>> > > > > > by
>> > > > > > > > subtracting the mean ensemble value minus the
observation
>> > value.
>> > > > > > (Note,
>> > > > > > > > you can pass in an ensemble mean field instead of
computing
>> it
>> > on
>> > > > the
>> > > > > > fly
>> > > > > > > > like this).  After processing N observations, we have
N
>> pairs
>> > of
>> > > > > > variance
>> > > > > > > > values and error values.  It groups those N pairs
into
>> buckets
>> > by
>> > > > > > binning
>> > > > > > > > the variance using the bucket size defined in the
config
>> file.
>> > > > > You'll
>> > > > > > > see
>> > > > > > > > in the output one SSVAR line for each of those
variance
>> > buckets.
>> > > > > > > >
>> > > > > > > > The METViewer database and display system (or Stat-
Analysis)
>> > can
>> > > > > > > aggregate
>> > > > > > > > those SSVAR lines intelligently across multiple cases
and
>> > create
>> > > > > plots
>> > > > > > > > which show the relationship between spread and skill.
>> > > > > > > >
>> > > > > > > > If that's what you need, great.
>> > > > > > > >
>> > > > > > > > However, if your looking for a much simpler
computation of
>> mean
>> > > > > spread
>> > > > > > > over
>> > > > > > > > some spatial region, unfortunately, the answer is no.
For
>> > > example,
>> > > > > > > > Ensemble-Stat doesn't write out the mean spread value
over
>> the
>> > > > entire
>> > > > > > > > domain.
>> > > > > > > >
>> > > > > > > > Just let me know what questions you still have.
>> > > > > > > >
>> > > > > > > > Thanks,
>> > > > > > > > John Halley Gotway
>> > > > > > > >
>> > > > > > > >
>> > > > > > > > On Fri, Jan 20, 2017 at 11:41 AM, Jared Lee via RT <
>> > > > > met_help at ucar.edu>
>> > > > > > > > wrote:
>> > > > > > > >
>> > > > > > > > >
>> > > > > > > > > Fri Jan 20 11:41:03 2017: Request 79224 was acted
upon.
>> > > > > > > > > Transaction: Ticket created by jaredlee at ucar.edu
>> > > > > > > > >        Queue: met_help
>> > > > > > > > >      Subject: calculating ensemble spread
>> > > > > > > > >        Owner: Nobody
>> > > > > > > > >   Requestors: jaredlee at ucar.edu
>> > > > > > > > >       Status: new
>> > > > > > > > >  Ticket <URL: https://rt.rap.ucar.edu/rt/
>> > > > > > Ticket/Display.html?id=79224
>> > > > > > > >
>> > > > > > > > >
>> > > > > > > > >
>> > > > > > > > > Hi, for my project I need to calculate some measure
of the
>> > > spread
>> > > > > of
>> > > > > > > two
>> > > > > > > > > different ensembles (one "large" and one "small"
>> ensemble). I
>> > > > have
>> > > > > > > model
>> > > > > > > > > output data (the large ensemble is in grib2 format,
the
>> small
>> > > > > > ensemble
>> > > > > > > is
>> > > > > > > > > WRF NetCDF), and I have MADIS point observations
that have
>> > > > already
>> > > > > > been
>> > > > > > > > > processed for use in MET.
>> > > > > > > > >
>> > > > > > > > > I previously used the ensemble-stat tool on the
large
>> > ensemble
>> > > to
>> > > > > > dump
>> > > > > > > > out
>> > > > > > > > > matched pair info for all ensemble member forecasts
at
>> each
>> > > > > > observation
>> > > > > > > > > location by printing out the ORANK line type. Given
the
>> > inputs
>> > > > > that I
>> > > > > > > > > described above, will MET calculate the ensemble
spread
>> > > > > automatically
>> > > > > > > if
>> > > > > > > > I
>> > > > > > > > > turn on the SSVAR line type? In the user guide
(v5.2) it
>> > > mentions
>> > > > > > that
>> > > > > > > an
>> > > > > > > > > "observation field" is needed to calculate things
like
>> > > > spread/skill
>> > > > > > > > > variance. Do point obs from MADIS count as an
"observation
>> > > > field"?
>> > > > > > > > >
>> > > > > > > > > Guidance would be much appreciated!
>> > > > > > > > >
>> > > > > > > > > Jared
>> > > > > > > > >
>> > > > > > > > > --
>> > > > > > > > > ===============================
>> > > > > > > > > Jared A. Lee, Ph.D.
>> > > > > > > > > Project Scientist I
>> > > > > > > > > Research Applications Laboratory
>> > > > > > > > > National Center for Atmospheric Research
>> > > > > > > > > Boulder, Colorado, USA
>> > > > > > > > >
>> > > > > > > > > Email: jaredlee at ucar.edu (w)
>> > > > > > > > > Phone: 303.497.8485 (w)
>> > > > > > > > > Web: https://staff.ucar.edu/users/jaredlee
>> > > > > > > > > ===============================
>> > > > > > > > >
>> > > > > > > > >
>> > > > > > > >
>> > > > > > > >
>> > > > > > >
>> > > > > > >
>> > > > > > > --
>> > > > > > > ===============================
>> > > > > > > Jared A. Lee, Ph.D.
>> > > > > > > Project Scientist I
>> > > > > > > Research Applications Laboratory
>> > > > > > > National Center for Atmospheric Research
>> > > > > > > Boulder, Colorado, USA
>> > > > > > >
>> > > > > > > Email: jaredlee at ucar.edu (w)
>> > > > > > > Phone: 303.497.8485 (w)
>> > > > > > > Web: https://staff.ucar.edu/users/jaredlee
>> > > > > > > ===============================
>> > > > > > >
>> > > > > > >
>> > > > > >
>> > > > > >
>> > > > >
>> > > > >
>> > > > > --
>> > > > > ===============================
>> > > > > Jared A. Lee, Ph.D.
>> > > > > Project Scientist I
>> > > > > Research Applications Laboratory
>> > > > > National Center for Atmospheric Research
>> > > > > Boulder, Colorado, USA
>> > > > >
>> > > > > Email: jaredlee at ucar.edu (w)
>> > > > > Phone: 303.497.8485 (w)
>> > > > > Web: https://staff.ucar.edu/users/jaredlee
>> > > > > ===============================
>> > > > >
>> > > > >
>> > > >
>> > > >
>> > >
>> > >
>> > > --
>> > > ===============================
>> > > Jared A. Lee, Ph.D.
>> > > Project Scientist I
>> > > Research Applications Laboratory
>> > > National Center for Atmospheric Research
>> > > Boulder, Colorado, USA
>> > >
>> > > Email: jaredlee at ucar.edu (w)
>> > > Phone: 303.497.8485 (w)
>> > > Web: https://staff.ucar.edu/users/jaredlee
>> > > ===============================
>> > >
>> > >
>> >
>> >
>>
>>
>> --
>> ===============================
>> Jared A. Lee, Ph.D.
>> Project Scientist I
>> Research Applications Laboratory
>> National Center for Atmospheric Research
>> Boulder, Colorado, USA
>>
>> Email: jaredlee at ucar.edu (w)
>> Phone: 303.497.8485 (w)
>> Web: https://staff.ucar.edu/users/jaredlee
>> ===============================
>>
>>
>

------------------------------------------------
Subject: calculating ensemble spread
From: Jared Lee
Time: Tue Feb 07 12:13:31 2017

No worries. I suppose you might need to update the xml again, unless
it
picks up the model names automatically from the stat files. Anyway,
the
model names I have in the stat files are now:

ALL_SREF
ARW_ONLY
3MEM

Jared

On Tue, Feb 7, 2017 at 12:12 PM, John Halley Gotway via RT <
met_help at ucar.edu> wrote:

> Sorry... should have checked first.  Looks like you already updated
them.
>
> I'm reloading now.
>
> On Tue, Feb 7, 2017 at 12:10 PM, John Halley Gotway
<johnhg at ucar.edu>
> wrote:
>
> > Jared,
> >
> > Will do.  Do I need to update the model names again?
> >
> > SREF -> 3MEM
> > SREF -> ARW_ONLY
> >
> > John
> >
> > On Tue, Feb 7, 2017 at 12:04 PM, Jared Lee via RT
<met_help at ucar.edu>
> > wrote:
> >
> >>
> >> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
> >>
> >> Hi John,
> >>
> >> I've finished re-generating all the .stat files (including with
the
> >> "model"
> >> labels I want). The zipped file is 405 MB and is located here on
> >> yellowstone:
> >>
> >> /glade/p/ral/wsap/jaredlee/NOAA_AQ_Ensemble/MET/15day_data.tar.gz
> >>
> >> Could you please replace all the data in the "noaa_aq_15day"
METViewer
> >> database with that? Thanks!
> >>
> >> Jared
> >>
> >> On Sat, Feb 4, 2017 at 6:47 PM, John Halley Gotway via RT <
> >> met_help at ucar.edu
> >> > wrote:
> >>
> >> > Jared,
> >> >
> >> > Just come by my office sometime next week and we'll copy your
data
> down
> >> to
> >> > a project machine and load it into METViewer.
> >> >
> >> > My calendar should be up to date if you want to schedule
something
> >> through
> >> > google calendar.
> >> >
> >> > John
> >> >
> >> > On Fri, Feb 3, 2017 at 9:22 PM Jared Lee via RT
<met_help at ucar.edu>
> >> wrote:
> >> >
> >> > >
> >> > > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224
>
> >> > >
> >> > > Hi John,
> >> > >
> >> > > I was unaware that UPP had a script to do that. As it turns
out,
> some
> >> > > members had all 48 hours in one file, and some had them split
> between
> >> two
> >> > > different files, so as it turns out it probably was simpler
to break
> >> them
> >> > > up first with ncks anyway. In the future I'll have the person
> running
> >> WRF
> >> > > to output single-time files from the get-go, so now my
scripts will
> >> work
> >> > > for that.
> >> > >
> >> > > Does my laptop 'shadowfax' count as a machine within the RAL
> firewall
> >> if
> >> > > I'm physically at NCAR? What if I use VPN when I'm offsite?
> Otherwise
> >> I
> >> > > could try to find some other machine to temporarily house my
data
> for
> >> > > METViewer to access.
> >> > >
> >> > > Jared
> >> > >
> >> > > On Fri, Feb 3, 2017 at 4:36 PM, John Halley Gotway via RT <
> >> > > met_help at ucar.edu
> >> > > > wrote:
> >> > >
> >> > > > Hi Jared,
> >> > > >
> >> > > > I'm surprised you need to use ncks to split out multiple
times.
> >> Isn't
> >> > > > there a script for running UPP on a wrfout file with
multiple
> >> times?  I
> >> > > > believe it's called run_unipost_frames.
> >> > > >
> >> > > > As for your question about METViewer... you'd need to pull
the
> >> > > > ensemble-stat data down to a machine inside the RAL
firewall.  And
> >> > then I
> >> > > > could demonstrate to you how to load it into METViewer and
start
> >> > creating
> >> > > > plots.
> >> > > >
> >> > > > Unfortunately, you can't load it directly from yellowstone
into
> >> > > METViewer,
> >> > > > which lives on a machine named "mandan".
> >> > > >
> >> > > > John
> >> > > >
> >> > > > On Fri, Feb 3, 2017 at 3:09 PM, Jared Lee via RT <
> met_help at ucar.edu
> >> >
> >> > > > wrote:
> >> > > >
> >> > > > >
> >> > > > > <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
> >> > > > >
> >> > > > > Hi John,
> >> > > > >
> >> > > > > Now I finally have finished running UPP (after having to
split
> >> apart
> >> > > the
> >> > > > > wrfout files into single-time files with ncks) and
ensemble-stat
> >> on
> >> > all
> >> > > > the
> >> > > > > model data for my three ensembles. How do I go about
getting
> that
> >> > into
> >> > > > > METViewer so that I can plot spread-skill information? My
data
> is
> >> on
> >> > > > > yellowstone. I also looked through the stat-analysis
section of
> >> the
> >> > MET
> >> > > > > Users Guide, but I'm not clear on what exactly I'd need
to do to
> >> get
> >> > > > what I
> >> > > > > need with that...
> >> > > > >
> >> > > > > Jared
> >> > > > >
> >> > > > > On Mon, Jan 23, 2017 at 5:01 PM, John Halley Gotway via
RT <
> >> > > > > met_help at ucar.edu> wrote:
> >> > > > >
> >> > > > > > Jared,
> >> > > > > >
> >> > > > > > Unfortunately, no, MET is not set up to read wrfout
files
> >> directly.
> >> > > As
> >> > > > > for
> >> > > > > > post-processing, you have 2 options... either UPP or
> wrf_interp
> >> > (only
> >> > > > for
> >> > > > > > WRF-ARW core).
> >> > > > > >
> >> > > > > > The former writes GRIB output which MET handles well.
The
> latter
> >> > > > writes a
> >> > > > > > flavor of NetCDF that MET is able to read... with the
caveat
> >> that
> >> > it
> >> > > > can
> >> > > > > > only read fields that are defined on non-staggered
dimensions.
> >> > > > > Basically,
> >> > > > > > it can't read the wind fields.
> >> > > > > >
> >> > > > > > I always encourage people to use UPP since MET supports
GRIB
> >> well.
> >> > > > > >
> >> > > > > > Being able to read wrfout directly has often been
requested.
> >> The
> >> > two
> >> > > > > > issues are the staggered dimensions and the hybrid
vertical
> >> > > coordinate.
> >> > > > > > Generally, matching to point observations is done in
terms of
> >> > > pressure
> >> > > > or
> >> > > > > > height.  So some interpolation to pressure/height
levels in
> >> > > inevitable
> >> > > > > > along the way.  We've resisted re-implementing that
> >> post-processing
> >> > > > logic
> >> > > > > > directly in MET to avoid duplication of effort with
UPP.  And
> >> UPP
> >> > > > > supports
> >> > > > > > the derivation of many other variable types that are
pretty
> >> useful.
> >> > > > > >
> >> > > > > > Perhaps there is some middle ground where we could
enable MET
> to
> >> > call
> >> > > > UPP
> >> > > > > > behind the scenes?  But we'd need to work out many
details.
> >> > > > > >
> >> > > > > > Thanks,
> >> > > > > > John
> >> > > > > >
> >> > > > > > On Fri, Jan 20, 2017 at 5:31 PM, Jared Lee via RT <
> >> > met_help at ucar.edu
> >> > > >
> >> > > > > > wrote:
> >> > > > > >
> >> > > > > > >
> >> > > > > > > <URL: https://rt.rap.ucar.edu/rt/
> Ticket/Display.html?id=79224
> >> >
> >> > > > > > >
> >> > > > > > > Hi John,
> >> > > > > > >
> >> > > > > > > Thanks for clarifying what ensemble-stat does to
produce the
> >> > SSVAR
> >> > > > line
> >> > > > > > > type. That helps.
> >> > > > > > >
> >> > > > > > > As for what I ultimately want, I guess that's still
somewhat
> >> > > vaguely
> >> > > > > > > defined. I think making plots that show the
spread/skill
> >> > > relationship
> >> > > > > > > across the date ranges, lead times, and variables of
> interest
> >> for
> >> > > > both
> >> > > > > > > ensembles would be a useful way to illustrate and
compare
> the
> >> > > spread
> >> > > > of
> >> > > > > > the
> >> > > > > > > two ensembles. That seems like it would be more
useful than
> >> > simply
> >> > > > > > > calculating the mean spread of the ensemble (which I
could
> do
> >> by
> >> > > > > writing
> >> > > > > > a
> >> > > > > > > new NCL script to read in all the ensemble fields and
not
> even
> >> > > > > bothering
> >> > > > > > > with MET...).
> >> > > > > > >
> >> > > > > > > Another side question... Can MET read in native-grid
NetCDF
> >> WRF
> >> > > > output
> >> > > > > > > files? Or do I need to run them through some pre-
processor
> >> (like
> >> > > UPP)
> >> > > > > > > first? After going through the user guide I'm not
clear
> >> whether
> >> > MET
> >> > > > > > accepts
> >> > > > > > > NetCDF wrfout files or not without pre-processing.
Also, if
> >> all
> >> > the
> >> > > > > > output
> >> > > > > > > times are in a single wrfout file, can MET open the
one file
> >> and
> >> > > find
> >> > > > > the
> >> > > > > > > correct time it's supposed to read, or do all the
wrfout
> files
> >> > need
> >> > > > to
> >> > > > > be
> >> > > > > > > for a single time only? The NetCDF wrfout files I've
been
> >> given
> >> > for
> >> > > > the
> >> > > > > > > small ensemble have all the output times in a single
file,
> >> but I
> >> > > can
> >> > > > > > write
> >> > > > > > > a script to parse them out to single-time files with
ncks.
> I'd
> >> > just
> >> > > > > > rather
> >> > > > > > > not go through that hassle if I don't have to!
> >> > > > > > >
> >> > > > > > > Jared
> >> > > > > > >
> >> > > > > > > On Fri, Jan 20, 2017 at 5:07 PM, John Halley Gotway
via RT <
> >> > > > > > > met_help at ucar.edu> wrote:
> >> > > > > > >
> >> > > > > > > > Hi Jared,
> >> > > > > > > >
> >> > > > > > > > I see a couple of questions in your email.
> >> > > > > > > >
> >> > > > > > > > First, yes, Ensemble-Stat is set up to process both
point
> >> and
> >> > > > gridded
> >> > > > > > > > observations.  And the SSVAR (spread/skill
variance) line
> >> type
> >> > > can
> >> > > > be
> >> > > > > > > > produced using either of these.
> >> > > > > > > >
> >> > > > > > > > The SSVAR line type stores information about the
> >> relationship
> >> > > > between
> >> > > > > > > > ensemble spread and skill.  Here's how it works.
> >> Ensemble-Stat
> >> > > > > > processes
> >> > > > > > > > each observation value... either from a point
location
> (i.e.
> >> > for
> >> > > > > point
> >> > > > > > > obs)
> >> > > > > > > > or from a gridded location (i.e. gridded obs).  For
each
> >> > > > observation,
> >> > > > > > it
> >> > > > > > > > gathers the N ensemble forecast values which
correspond to
> >> it.
> >> > > It
> >> > > > > > > computes
> >> > > > > > > > the variance of those ensemble values, and it also
> computes
> >> the
> >> > > > error
> >> > > > > > by
> >> > > > > > > > subtracting the mean ensemble value minus the
observation
> >> > value.
> >> > > > > > (Note,
> >> > > > > > > > you can pass in an ensemble mean field instead of
> computing
> >> it
> >> > on
> >> > > > the
> >> > > > > > fly
> >> > > > > > > > like this).  After processing N observations, we
have N
> >> pairs
> >> > of
> >> > > > > > variance
> >> > > > > > > > values and error values.  It groups those N pairs
into
> >> buckets
> >> > by
> >> > > > > > binning
> >> > > > > > > > the variance using the bucket size defined in the
config
> >> file.
> >> > > > > You'll
> >> > > > > > > see
> >> > > > > > > > in the output one SSVAR line for each of those
variance
> >> > buckets.
> >> > > > > > > >
> >> > > > > > > > The METViewer database and display system (or
> Stat-Analysis)
> >> > can
> >> > > > > > > aggregate
> >> > > > > > > > those SSVAR lines intelligently across multiple
cases and
> >> > create
> >> > > > > plots
> >> > > > > > > > which show the relationship between spread and
skill.
> >> > > > > > > >
> >> > > > > > > > If that's what you need, great.
> >> > > > > > > >
> >> > > > > > > > However, if your looking for a much simpler
computation of
> >> mean
> >> > > > > spread
> >> > > > > > > over
> >> > > > > > > > some spatial region, unfortunately, the answer is
no.  For
> >> > > example,
> >> > > > > > > > Ensemble-Stat doesn't write out the mean spread
value over
> >> the
> >> > > > entire
> >> > > > > > > > domain.
> >> > > > > > > >
> >> > > > > > > > Just let me know what questions you still have.
> >> > > > > > > >
> >> > > > > > > > Thanks,
> >> > > > > > > > John Halley Gotway
> >> > > > > > > >
> >> > > > > > > >
> >> > > > > > > > On Fri, Jan 20, 2017 at 11:41 AM, Jared Lee via RT
<
> >> > > > > met_help at ucar.edu>
> >> > > > > > > > wrote:
> >> > > > > > > >
> >> > > > > > > > >
> >> > > > > > > > > Fri Jan 20 11:41:03 2017: Request 79224 was acted
upon.
> >> > > > > > > > > Transaction: Ticket created by jaredlee at ucar.edu
> >> > > > > > > > >        Queue: met_help
> >> > > > > > > > >      Subject: calculating ensemble spread
> >> > > > > > > > >        Owner: Nobody
> >> > > > > > > > >   Requestors: jaredlee at ucar.edu
> >> > > > > > > > >       Status: new
> >> > > > > > > > >  Ticket <URL: https://rt.rap.ucar.edu/rt/
> >> > > > > > Ticket/Display.html?id=79224
> >> > > > > > > >
> >> > > > > > > > >
> >> > > > > > > > >
> >> > > > > > > > > Hi, for my project I need to calculate some
measure of
> the
> >> > > spread
> >> > > > > of
> >> > > > > > > two
> >> > > > > > > > > different ensembles (one "large" and one "small"
> >> ensemble). I
> >> > > > have
> >> > > > > > > model
> >> > > > > > > > > output data (the large ensemble is in grib2
format, the
> >> small
> >> > > > > > ensemble
> >> > > > > > > is
> >> > > > > > > > > WRF NetCDF), and I have MADIS point observations
that
> have
> >> > > > already
> >> > > > > > been
> >> > > > > > > > > processed for use in MET.
> >> > > > > > > > >
> >> > > > > > > > > I previously used the ensemble-stat tool on the
large
> >> > ensemble
> >> > > to
> >> > > > > > dump
> >> > > > > > > > out
> >> > > > > > > > > matched pair info for all ensemble member
forecasts at
> >> each
> >> > > > > > observation
> >> > > > > > > > > location by printing out the ORANK line type.
Given the
> >> > inputs
> >> > > > > that I
> >> > > > > > > > > described above, will MET calculate the ensemble
spread
> >> > > > > automatically
> >> > > > > > > if
> >> > > > > > > > I
> >> > > > > > > > > turn on the SSVAR line type? In the user guide
(v5.2) it
> >> > > mentions
> >> > > > > > that
> >> > > > > > > an
> >> > > > > > > > > "observation field" is needed to calculate things
like
> >> > > > spread/skill
> >> > > > > > > > > variance. Do point obs from MADIS count as an
> "observation
> >> > > > field"?
> >> > > > > > > > >
> >> > > > > > > > > Guidance would be much appreciated!
> >> > > > > > > > >
> >> > > > > > > > > Jared
> >> > > > > > > > >
> >> > > > > > > > > --
> >> > > > > > > > > ===============================
> >> > > > > > > > > Jared A. Lee, Ph.D.
> >> > > > > > > > > Project Scientist I
> >> > > > > > > > > Research Applications Laboratory
> >> > > > > > > > > National Center for Atmospheric Research
> >> > > > > > > > > Boulder, Colorado, USA
> >> > > > > > > > >
> >> > > > > > > > > Email: jaredlee at ucar.edu (w)
> >> > > > > > > > > Phone: 303.497.8485 (w)
> >> > > > > > > > > Web: https://staff.ucar.edu/users/jaredlee
> >> > > > > > > > > ===============================
> >> > > > > > > > >
> >> > > > > > > > >
> >> > > > > > > >
> >> > > > > > > >
> >> > > > > > >
> >> > > > > > >
> >> > > > > > > --
> >> > > > > > > ===============================
> >> > > > > > > Jared A. Lee, Ph.D.
> >> > > > > > > Project Scientist I
> >> > > > > > > Research Applications Laboratory
> >> > > > > > > National Center for Atmospheric Research
> >> > > > > > > Boulder, Colorado, USA
> >> > > > > > >
> >> > > > > > > Email: jaredlee at ucar.edu (w)
> >> > > > > > > Phone: 303.497.8485 (w)
> >> > > > > > > Web: https://staff.ucar.edu/users/jaredlee
> >> > > > > > > ===============================
> >> > > > > > >
> >> > > > > > >
> >> > > > > >
> >> > > > > >
> >> > > > >
> >> > > > >
> >> > > > > --
> >> > > > > ===============================
> >> > > > > Jared A. Lee, Ph.D.
> >> > > > > Project Scientist I
> >> > > > > Research Applications Laboratory
> >> > > > > National Center for Atmospheric Research
> >> > > > > Boulder, Colorado, USA
> >> > > > >
> >> > > > > Email: jaredlee at ucar.edu (w)
> >> > > > > Phone: 303.497.8485 (w)
> >> > > > > Web: https://staff.ucar.edu/users/jaredlee
> >> > > > > ===============================
> >> > > > >
> >> > > > >
> >> > > >
> >> > > >
> >> > >
> >> > >
> >> > > --
> >> > > ===============================
> >> > > Jared A. Lee, Ph.D.
> >> > > Project Scientist I
> >> > > Research Applications Laboratory
> >> > > National Center for Atmospheric Research
> >> > > Boulder, Colorado, USA
> >> > >
> >> > > Email: jaredlee at ucar.edu (w)
> >> > > Phone: 303.497.8485 (w)
> >> > > Web: https://staff.ucar.edu/users/jaredlee
> >> > > ===============================
> >> > >
> >> > >
> >> >
> >> >
> >>
> >>
> >> --
> >> ===============================
> >> Jared A. Lee, Ph.D.
> >> Project Scientist I
> >> Research Applications Laboratory
> >> National Center for Atmospheric Research
> >> Boulder, Colorado, USA
> >>
> >> Email: jaredlee at ucar.edu (w)
> >> Phone: 303.497.8485 (w)
> >> Web: https://staff.ucar.edu/users/jaredlee
> >> ===============================
> >>
> >>
> >
>
>


--
===============================
Jared A. Lee, Ph.D.
Project Scientist I
Research Applications Laboratory
National Center for Atmospheric Research
Boulder, Colorado, USA

Email: jaredlee at ucar.edu (w)
Phone: 303.497.8485 (w)
Web: https://staff.ucar.edu/users/jaredlee
===============================

------------------------------------------------
Subject: calculating ensemble spread
From: John Halley Gotway
Time: Tue Feb 07 15:43:35 2017

Jared,

Forgot to write back earlier... the reload finished at 12:16pm today.

Please take a look and let me know if you run into any issues.

I'll go ahead and resolve this ticket... and hopefully Tatiana will
fix the
issue with the spread-skill legend in the next METViewer release.

Thanks,
John

On Tue, Feb 7, 2017 at 12:13 PM, Jared Lee via RT <met_help at ucar.edu>
wrote:

>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
>
> No worries. I suppose you might need to update the xml again, unless
it
> picks up the model names automatically from the stat files. Anyway,
the
> model names I have in the stat files are now:
>
> ALL_SREF
> ARW_ONLY
> 3MEM
>
> Jared
>
> On Tue, Feb 7, 2017 at 12:12 PM, John Halley Gotway via RT <
> met_help at ucar.edu> wrote:
>
> > Sorry... should have checked first.  Looks like you already
updated them.
> >
> > I'm reloading now.
> >
> > On Tue, Feb 7, 2017 at 12:10 PM, John Halley Gotway
<johnhg at ucar.edu>
> > wrote:
> >
> > > Jared,
> > >
> > > Will do.  Do I need to update the model names again?
> > >
> > > SREF -> 3MEM
> > > SREF -> ARW_ONLY
> > >
> > > John
> > >
> > > On Tue, Feb 7, 2017 at 12:04 PM, Jared Lee via RT
<met_help at ucar.edu>
> > > wrote:
> > >
> > >>
> > >> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
> > >>
> > >> Hi John,
> > >>
> > >> I've finished re-generating all the .stat files (including with
the
> > >> "model"
> > >> labels I want). The zipped file is 405 MB and is located here
on
> > >> yellowstone:
> > >>
> > >>
/glade/p/ral/wsap/jaredlee/NOAA_AQ_Ensemble/MET/15day_data.tar.gz
> > >>
> > >> Could you please replace all the data in the "noaa_aq_15day"
METViewer
> > >> database with that? Thanks!
> > >>
> > >> Jared
> > >>
> > >> On Sat, Feb 4, 2017 at 6:47 PM, John Halley Gotway via RT <
> > >> met_help at ucar.edu
> > >> > wrote:
> > >>
> > >> > Jared,
> > >> >
> > >> > Just come by my office sometime next week and we'll copy your
data
> > down
> > >> to
> > >> > a project machine and load it into METViewer.
> > >> >
> > >> > My calendar should be up to date if you want to schedule
something
> > >> through
> > >> > google calendar.
> > >> >
> > >> > John
> > >> >
> > >> > On Fri, Feb 3, 2017 at 9:22 PM Jared Lee via RT
<met_help at ucar.edu>
> > >> wrote:
> > >> >
> > >> > >
> > >> > > <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
> > >> > >
> > >> > > Hi John,
> > >> > >
> > >> > > I was unaware that UPP had a script to do that. As it turns
out,
> > some
> > >> > > members had all 48 hours in one file, and some had them
split
> > between
> > >> two
> > >> > > different files, so as it turns out it probably was simpler
to
> break
> > >> them
> > >> > > up first with ncks anyway. In the future I'll have the
person
> > running
> > >> WRF
> > >> > > to output single-time files from the get-go, so now my
scripts
> will
> > >> work
> > >> > > for that.
> > >> > >
> > >> > > Does my laptop 'shadowfax' count as a machine within the
RAL
> > firewall
> > >> if
> > >> > > I'm physically at NCAR? What if I use VPN when I'm offsite?
> > Otherwise
> > >> I
> > >> > > could try to find some other machine to temporarily house
my data
> > for
> > >> > > METViewer to access.
> > >> > >
> > >> > > Jared
> > >> > >
> > >> > > On Fri, Feb 3, 2017 at 4:36 PM, John Halley Gotway via RT <
> > >> > > met_help at ucar.edu
> > >> > > > wrote:
> > >> > >
> > >> > > > Hi Jared,
> > >> > > >
> > >> > > > I'm surprised you need to use ncks to split out multiple
times.
> > >> Isn't
> > >> > > > there a script for running UPP on a wrfout file with
multiple
> > >> times?  I
> > >> > > > believe it's called run_unipost_frames.
> > >> > > >
> > >> > > > As for your question about METViewer... you'd need to
pull the
> > >> > > > ensemble-stat data down to a machine inside the RAL
firewall.
> And
> > >> > then I
> > >> > > > could demonstrate to you how to load it into METViewer
and start
> > >> > creating
> > >> > > > plots.
> > >> > > >
> > >> > > > Unfortunately, you can't load it directly from
yellowstone into
> > >> > > METViewer,
> > >> > > > which lives on a machine named "mandan".
> > >> > > >
> > >> > > > John
> > >> > > >
> > >> > > > On Fri, Feb 3, 2017 at 3:09 PM, Jared Lee via RT <
> > met_help at ucar.edu
> > >> >
> > >> > > > wrote:
> > >> > > >
> > >> > > > >
> > >> > > > > <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224
> >
> > >> > > > >
> > >> > > > > Hi John,
> > >> > > > >
> > >> > > > > Now I finally have finished running UPP (after having
to split
> > >> apart
> > >> > > the
> > >> > > > > wrfout files into single-time files with ncks) and
> ensemble-stat
> > >> on
> > >> > all
> > >> > > > the
> > >> > > > > model data for my three ensembles. How do I go about
getting
> > that
> > >> > into
> > >> > > > > METViewer so that I can plot spread-skill information?
My data
> > is
> > >> on
> > >> > > > > yellowstone. I also looked through the stat-analysis
section
> of
> > >> the
> > >> > MET
> > >> > > > > Users Guide, but I'm not clear on what exactly I'd need
to do
> to
> > >> get
> > >> > > > what I
> > >> > > > > need with that...
> > >> > > > >
> > >> > > > > Jared
> > >> > > > >
> > >> > > > > On Mon, Jan 23, 2017 at 5:01 PM, John Halley Gotway via
RT <
> > >> > > > > met_help at ucar.edu> wrote:
> > >> > > > >
> > >> > > > > > Jared,
> > >> > > > > >
> > >> > > > > > Unfortunately, no, MET is not set up to read wrfout
files
> > >> directly.
> > >> > > As
> > >> > > > > for
> > >> > > > > > post-processing, you have 2 options... either UPP or
> > wrf_interp
> > >> > (only
> > >> > > > for
> > >> > > > > > WRF-ARW core).
> > >> > > > > >
> > >> > > > > > The former writes GRIB output which MET handles well.
The
> > latter
> > >> > > > writes a
> > >> > > > > > flavor of NetCDF that MET is able to read... with the
caveat
> > >> that
> > >> > it
> > >> > > > can
> > >> > > > > > only read fields that are defined on non-staggered
> dimensions.
> > >> > > > > Basically,
> > >> > > > > > it can't read the wind fields.
> > >> > > > > >
> > >> > > > > > I always encourage people to use UPP since MET
supports GRIB
> > >> well.
> > >> > > > > >
> > >> > > > > > Being able to read wrfout directly has often been
requested.
> > >> The
> > >> > two
> > >> > > > > > issues are the staggered dimensions and the hybrid
vertical
> > >> > > coordinate.
> > >> > > > > > Generally, matching to point observations is done in
terms
> of
> > >> > > pressure
> > >> > > > or
> > >> > > > > > height.  So some interpolation to pressure/height
levels in
> > >> > > inevitable
> > >> > > > > > along the way.  We've resisted re-implementing that
> > >> post-processing
> > >> > > > logic
> > >> > > > > > directly in MET to avoid duplication of effort with
UPP.
> And
> > >> UPP
> > >> > > > > supports
> > >> > > > > > the derivation of many other variable types that are
pretty
> > >> useful.
> > >> > > > > >
> > >> > > > > > Perhaps there is some middle ground where we could
enable
> MET
> > to
> > >> > call
> > >> > > > UPP
> > >> > > > > > behind the scenes?  But we'd need to work out many
details.
> > >> > > > > >
> > >> > > > > > Thanks,
> > >> > > > > > John
> > >> > > > > >
> > >> > > > > > On Fri, Jan 20, 2017 at 5:31 PM, Jared Lee via RT <
> > >> > met_help at ucar.edu
> > >> > > >
> > >> > > > > > wrote:
> > >> > > > > >
> > >> > > > > > >
> > >> > > > > > > <URL: https://rt.rap.ucar.edu/rt/
> > Ticket/Display.html?id=79224
> > >> >
> > >> > > > > > >
> > >> > > > > > > Hi John,
> > >> > > > > > >
> > >> > > > > > > Thanks for clarifying what ensemble-stat does to
produce
> the
> > >> > SSVAR
> > >> > > > line
> > >> > > > > > > type. That helps.
> > >> > > > > > >
> > >> > > > > > > As for what I ultimately want, I guess that's still
> somewhat
> > >> > > vaguely
> > >> > > > > > > defined. I think making plots that show the
spread/skill
> > >> > > relationship
> > >> > > > > > > across the date ranges, lead times, and variables
of
> > interest
> > >> for
> > >> > > > both
> > >> > > > > > > ensembles would be a useful way to illustrate and
compare
> > the
> > >> > > spread
> > >> > > > of
> > >> > > > > > the
> > >> > > > > > > two ensembles. That seems like it would be more
useful
> than
> > >> > simply
> > >> > > > > > > calculating the mean spread of the ensemble (which
I could
> > do
> > >> by
> > >> > > > > writing
> > >> > > > > > a
> > >> > > > > > > new NCL script to read in all the ensemble fields
and not
> > even
> > >> > > > > bothering
> > >> > > > > > > with MET...).
> > >> > > > > > >
> > >> > > > > > > Another side question... Can MET read in native-
grid
> NetCDF
> > >> WRF
> > >> > > > output
> > >> > > > > > > files? Or do I need to run them through some pre-
processor
> > >> (like
> > >> > > UPP)
> > >> > > > > > > first? After going through the user guide I'm not
clear
> > >> whether
> > >> > MET
> > >> > > > > > accepts
> > >> > > > > > > NetCDF wrfout files or not without pre-processing.
Also,
> if
> > >> all
> > >> > the
> > >> > > > > > output
> > >> > > > > > > times are in a single wrfout file, can MET open the
one
> file
> > >> and
> > >> > > find
> > >> > > > > the
> > >> > > > > > > correct time it's supposed to read, or do all the
wrfout
> > files
> > >> > need
> > >> > > > to
> > >> > > > > be
> > >> > > > > > > for a single time only? The NetCDF wrfout files
I've been
> > >> given
> > >> > for
> > >> > > > the
> > >> > > > > > > small ensemble have all the output times in a
single file,
> > >> but I
> > >> > > can
> > >> > > > > > write
> > >> > > > > > > a script to parse them out to single-time files
with ncks.
> > I'd
> > >> > just
> > >> > > > > > rather
> > >> > > > > > > not go through that hassle if I don't have to!
> > >> > > > > > >
> > >> > > > > > > Jared
> > >> > > > > > >
> > >> > > > > > > On Fri, Jan 20, 2017 at 5:07 PM, John Halley Gotway
via
> RT <
> > >> > > > > > > met_help at ucar.edu> wrote:
> > >> > > > > > >
> > >> > > > > > > > Hi Jared,
> > >> > > > > > > >
> > >> > > > > > > > I see a couple of questions in your email.
> > >> > > > > > > >
> > >> > > > > > > > First, yes, Ensemble-Stat is set up to process
both
> point
> > >> and
> > >> > > > gridded
> > >> > > > > > > > observations.  And the SSVAR (spread/skill
variance)
> line
> > >> type
> > >> > > can
> > >> > > > be
> > >> > > > > > > > produced using either of these.
> > >> > > > > > > >
> > >> > > > > > > > The SSVAR line type stores information about the
> > >> relationship
> > >> > > > between
> > >> > > > > > > > ensemble spread and skill.  Here's how it works.
> > >> Ensemble-Stat
> > >> > > > > > processes
> > >> > > > > > > > each observation value... either from a point
location
> > (i.e.
> > >> > for
> > >> > > > > point
> > >> > > > > > > obs)
> > >> > > > > > > > or from a gridded location (i.e. gridded obs).
For each
> > >> > > > observation,
> > >> > > > > > it
> > >> > > > > > > > gathers the N ensemble forecast values which
correspond
> to
> > >> it.
> > >> > > It
> > >> > > > > > > computes
> > >> > > > > > > > the variance of those ensemble values, and it
also
> > computes
> > >> the
> > >> > > > error
> > >> > > > > > by
> > >> > > > > > > > subtracting the mean ensemble value minus the
> observation
> > >> > value.
> > >> > > > > > (Note,
> > >> > > > > > > > you can pass in an ensemble mean field instead of
> > computing
> > >> it
> > >> > on
> > >> > > > the
> > >> > > > > > fly
> > >> > > > > > > > like this).  After processing N observations, we
have N
> > >> pairs
> > >> > of
> > >> > > > > > variance
> > >> > > > > > > > values and error values.  It groups those N pairs
into
> > >> buckets
> > >> > by
> > >> > > > > > binning
> > >> > > > > > > > the variance using the bucket size defined in the
config
> > >> file.
> > >> > > > > You'll
> > >> > > > > > > see
> > >> > > > > > > > in the output one SSVAR line for each of those
variance
> > >> > buckets.
> > >> > > > > > > >
> > >> > > > > > > > The METViewer database and display system (or
> > Stat-Analysis)
> > >> > can
> > >> > > > > > > aggregate
> > >> > > > > > > > those SSVAR lines intelligently across multiple
cases
> and
> > >> > create
> > >> > > > > plots
> > >> > > > > > > > which show the relationship between spread and
skill.
> > >> > > > > > > >
> > >> > > > > > > > If that's what you need, great.
> > >> > > > > > > >
> > >> > > > > > > > However, if your looking for a much simpler
computation
> of
> > >> mean
> > >> > > > > spread
> > >> > > > > > > over
> > >> > > > > > > > some spatial region, unfortunately, the answer is
no.
> For
> > >> > > example,
> > >> > > > > > > > Ensemble-Stat doesn't write out the mean spread
value
> over
> > >> the
> > >> > > > entire
> > >> > > > > > > > domain.
> > >> > > > > > > >
> > >> > > > > > > > Just let me know what questions you still have.
> > >> > > > > > > >
> > >> > > > > > > > Thanks,
> > >> > > > > > > > John Halley Gotway
> > >> > > > > > > >
> > >> > > > > > > >
> > >> > > > > > > > On Fri, Jan 20, 2017 at 11:41 AM, Jared Lee via
RT <
> > >> > > > > met_help at ucar.edu>
> > >> > > > > > > > wrote:
> > >> > > > > > > >
> > >> > > > > > > > >
> > >> > > > > > > > > Fri Jan 20 11:41:03 2017: Request 79224 was
acted
> upon.
> > >> > > > > > > > > Transaction: Ticket created by
jaredlee at ucar.edu
> > >> > > > > > > > >        Queue: met_help
> > >> > > > > > > > >      Subject: calculating ensemble spread
> > >> > > > > > > > >        Owner: Nobody
> > >> > > > > > > > >   Requestors: jaredlee at ucar.edu
> > >> > > > > > > > >       Status: new
> > >> > > > > > > > >  Ticket <URL: https://rt.rap.ucar.edu/rt/
> > >> > > > > > Ticket/Display.html?id=79224
> > >> > > > > > > >
> > >> > > > > > > > >
> > >> > > > > > > > >
> > >> > > > > > > > > Hi, for my project I need to calculate some
measure of
> > the
> > >> > > spread
> > >> > > > > of
> > >> > > > > > > two
> > >> > > > > > > > > different ensembles (one "large" and one
"small"
> > >> ensemble). I
> > >> > > > have
> > >> > > > > > > model
> > >> > > > > > > > > output data (the large ensemble is in grib2
format,
> the
> > >> small
> > >> > > > > > ensemble
> > >> > > > > > > is
> > >> > > > > > > > > WRF NetCDF), and I have MADIS point
observations that
> > have
> > >> > > > already
> > >> > > > > > been
> > >> > > > > > > > > processed for use in MET.
> > >> > > > > > > > >
> > >> > > > > > > > > I previously used the ensemble-stat tool on the
large
> > >> > ensemble
> > >> > > to
> > >> > > > > > dump
> > >> > > > > > > > out
> > >> > > > > > > > > matched pair info for all ensemble member
forecasts at
> > >> each
> > >> > > > > > observation
> > >> > > > > > > > > location by printing out the ORANK line type.
Given
> the
> > >> > inputs
> > >> > > > > that I
> > >> > > > > > > > > described above, will MET calculate the
ensemble
> spread
> > >> > > > > automatically
> > >> > > > > > > if
> > >> > > > > > > > I
> > >> > > > > > > > > turn on the SSVAR line type? In the user guide
(v5.2)
> it
> > >> > > mentions
> > >> > > > > > that
> > >> > > > > > > an
> > >> > > > > > > > > "observation field" is needed to calculate
things like
> > >> > > > spread/skill
> > >> > > > > > > > > variance. Do point obs from MADIS count as an
> > "observation
> > >> > > > field"?
> > >> > > > > > > > >
> > >> > > > > > > > > Guidance would be much appreciated!
> > >> > > > > > > > >
> > >> > > > > > > > > Jared
> > >> > > > > > > > >
> > >> > > > > > > > > --
> > >> > > > > > > > > ===============================
> > >> > > > > > > > > Jared A. Lee, Ph.D.
> > >> > > > > > > > > Project Scientist I
> > >> > > > > > > > > Research Applications Laboratory
> > >> > > > > > > > > National Center for Atmospheric Research
> > >> > > > > > > > > Boulder, Colorado, USA
> > >> > > > > > > > >
> > >> > > > > > > > > Email: jaredlee at ucar.edu (w)
> > >> > > > > > > > > Phone: 303.497.8485 (w)
> > >> > > > > > > > > Web: https://staff.ucar.edu/users/jaredlee
> > >> > > > > > > > > ===============================
> > >> > > > > > > > >
> > >> > > > > > > > >
> > >> > > > > > > >
> > >> > > > > > > >
> > >> > > > > > >
> > >> > > > > > >
> > >> > > > > > > --
> > >> > > > > > > ===============================
> > >> > > > > > > Jared A. Lee, Ph.D.
> > >> > > > > > > Project Scientist I
> > >> > > > > > > Research Applications Laboratory
> > >> > > > > > > National Center for Atmospheric Research
> > >> > > > > > > Boulder, Colorado, USA
> > >> > > > > > >
> > >> > > > > > > Email: jaredlee at ucar.edu (w)
> > >> > > > > > > Phone: 303.497.8485 (w)
> > >> > > > > > > Web: https://staff.ucar.edu/users/jaredlee
> > >> > > > > > > ===============================
> > >> > > > > > >
> > >> > > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > >
> > >> > > > >
> > >> > > > > --
> > >> > > > > ===============================
> > >> > > > > Jared A. Lee, Ph.D.
> > >> > > > > Project Scientist I
> > >> > > > > Research Applications Laboratory
> > >> > > > > National Center for Atmospheric Research
> > >> > > > > Boulder, Colorado, USA
> > >> > > > >
> > >> > > > > Email: jaredlee at ucar.edu (w)
> > >> > > > > Phone: 303.497.8485 (w)
> > >> > > > > Web: https://staff.ucar.edu/users/jaredlee
> > >> > > > > ===============================
> > >> > > > >
> > >> > > > >
> > >> > > >
> > >> > > >
> > >> > >
> > >> > >
> > >> > > --
> > >> > > ===============================
> > >> > > Jared A. Lee, Ph.D.
> > >> > > Project Scientist I
> > >> > > Research Applications Laboratory
> > >> > > National Center for Atmospheric Research
> > >> > > Boulder, Colorado, USA
> > >> > >
> > >> > > Email: jaredlee at ucar.edu (w)
> > >> > > Phone: 303.497.8485 (w)
> > >> > > Web: https://staff.ucar.edu/users/jaredlee
> > >> > > ===============================
> > >> > >
> > >> > >
> > >> >
> > >> >
> > >>
> > >>
> > >> --
> > >> ===============================
> > >> Jared A. Lee, Ph.D.
> > >> Project Scientist I
> > >> Research Applications Laboratory
> > >> National Center for Atmospheric Research
> > >> Boulder, Colorado, USA
> > >>
> > >> Email: jaredlee at ucar.edu (w)
> > >> Phone: 303.497.8485 (w)
> > >> Web: https://staff.ucar.edu/users/jaredlee
> > >> ===============================
> > >>
> > >>
> > >
> >
> >
>
>
> --
> ===============================
> Jared A. Lee, Ph.D.
> Project Scientist I
> Research Applications Laboratory
> National Center for Atmospheric Research
> Boulder, Colorado, USA
>
> Email: jaredlee at ucar.edu (w)
> Phone: 303.497.8485 (w)
> Web: https://staff.ucar.edu/users/jaredlee
> ===============================
>
>

------------------------------------------------
Subject: calculating ensemble spread
From: Jared Lee
Time: Tue Feb 07 15:52:45 2017

Hi John,

Thanks! I just tried to make a plot, and I think something with the
underlying xml specifying the database needs to be updated. I updated
the
name of one of the ensembles from "SREF" to "ALL_SREF" to be
consistent
with my naming conventions elsewhere in my scripts and file structure.
Right now METViewer is expecting to see model="SREF" as a holdover
from
yesterday but is not finding anything in the database that matches.

Jared

On Tue, Feb 7, 2017 at 3:43 PM, John Halley Gotway via RT
<met_help at ucar.edu
> wrote:

> Jared,
>
> Forgot to write back earlier... the reload finished at 12:16pm
today.
>
> Please take a look and let me know if you run into any issues.
>
> I'll go ahead and resolve this ticket... and hopefully Tatiana will
fix the
> issue with the spread-skill legend in the next METViewer release.
>
> Thanks,
> John
>
> On Tue, Feb 7, 2017 at 12:13 PM, Jared Lee via RT
<met_help at ucar.edu>
> wrote:
>
> >
> > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
> >
> > No worries. I suppose you might need to update the xml again,
unless it
> > picks up the model names automatically from the stat files.
Anyway, the
> > model names I have in the stat files are now:
> >
> > ALL_SREF
> > ARW_ONLY
> > 3MEM
> >
> > Jared
> >
> > On Tue, Feb 7, 2017 at 12:12 PM, John Halley Gotway via RT <
> > met_help at ucar.edu> wrote:
> >
> > > Sorry... should have checked first.  Looks like you already
updated
> them.
> > >
> > > I'm reloading now.
> > >
> > > On Tue, Feb 7, 2017 at 12:10 PM, John Halley Gotway
<johnhg at ucar.edu>
> > > wrote:
> > >
> > > > Jared,
> > > >
> > > > Will do.  Do I need to update the model names again?
> > > >
> > > > SREF -> 3MEM
> > > > SREF -> ARW_ONLY
> > > >
> > > > John
> > > >
> > > > On Tue, Feb 7, 2017 at 12:04 PM, Jared Lee via RT
<met_help at ucar.edu
> >
> > > > wrote:
> > > >
> > > >>
> > > >> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224
>
> > > >>
> > > >> Hi John,
> > > >>
> > > >> I've finished re-generating all the .stat files (including
with the
> > > >> "model"
> > > >> labels I want). The zipped file is 405 MB and is located here
on
> > > >> yellowstone:
> > > >>
> > > >>
/glade/p/ral/wsap/jaredlee/NOAA_AQ_Ensemble/MET/15day_data.tar.gz
> > > >>
> > > >> Could you please replace all the data in the "noaa_aq_15day"
> METViewer
> > > >> database with that? Thanks!
> > > >>
> > > >> Jared
> > > >>
> > > >> On Sat, Feb 4, 2017 at 6:47 PM, John Halley Gotway via RT <
> > > >> met_help at ucar.edu
> > > >> > wrote:
> > > >>
> > > >> > Jared,
> > > >> >
> > > >> > Just come by my office sometime next week and we'll copy
your data
> > > down
> > > >> to
> > > >> > a project machine and load it into METViewer.
> > > >> >
> > > >> > My calendar should be up to date if you want to schedule
something
> > > >> through
> > > >> > google calendar.
> > > >> >
> > > >> > John
> > > >> >
> > > >> > On Fri, Feb 3, 2017 at 9:22 PM Jared Lee via RT <
> met_help at ucar.edu>
> > > >> wrote:
> > > >> >
> > > >> > >
> > > >> > > <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
> > > >> > >
> > > >> > > Hi John,
> > > >> > >
> > > >> > > I was unaware that UPP had a script to do that. As it
turns out,
> > > some
> > > >> > > members had all 48 hours in one file, and some had them
split
> > > between
> > > >> two
> > > >> > > different files, so as it turns out it probably was
simpler to
> > break
> > > >> them
> > > >> > > up first with ncks anyway. In the future I'll have the
person
> > > running
> > > >> WRF
> > > >> > > to output single-time files from the get-go, so now my
scripts
> > will
> > > >> work
> > > >> > > for that.
> > > >> > >
> > > >> > > Does my laptop 'shadowfax' count as a machine within the
RAL
> > > firewall
> > > >> if
> > > >> > > I'm physically at NCAR? What if I use VPN when I'm
offsite?
> > > Otherwise
> > > >> I
> > > >> > > could try to find some other machine to temporarily house
my
> data
> > > for
> > > >> > > METViewer to access.
> > > >> > >
> > > >> > > Jared
> > > >> > >
> > > >> > > On Fri, Feb 3, 2017 at 4:36 PM, John Halley Gotway via RT
<
> > > >> > > met_help at ucar.edu
> > > >> > > > wrote:
> > > >> > >
> > > >> > > > Hi Jared,
> > > >> > > >
> > > >> > > > I'm surprised you need to use ncks to split out
multiple
> times.
> > > >> Isn't
> > > >> > > > there a script for running UPP on a wrfout file with
multiple
> > > >> times?  I
> > > >> > > > believe it's called run_unipost_frames.
> > > >> > > >
> > > >> > > > As for your question about METViewer... you'd need to
pull the
> > > >> > > > ensemble-stat data down to a machine inside the RAL
firewall.
> > And
> > > >> > then I
> > > >> > > > could demonstrate to you how to load it into METViewer
and
> start
> > > >> > creating
> > > >> > > > plots.
> > > >> > > >
> > > >> > > > Unfortunately, you can't load it directly from
yellowstone
> into
> > > >> > > METViewer,
> > > >> > > > which lives on a machine named "mandan".
> > > >> > > >
> > > >> > > > John
> > > >> > > >
> > > >> > > > On Fri, Feb 3, 2017 at 3:09 PM, Jared Lee via RT <
> > > met_help at ucar.edu
> > > >> >
> > > >> > > > wrote:
> > > >> > > >
> > > >> > > > >
> > > >> > > > > <URL: https://rt.rap.ucar.edu/rt/
> Ticket/Display.html?id=79224
> > >
> > > >> > > > >
> > > >> > > > > Hi John,
> > > >> > > > >
> > > >> > > > > Now I finally have finished running UPP (after having
to
> split
> > > >> apart
> > > >> > > the
> > > >> > > > > wrfout files into single-time files with ncks) and
> > ensemble-stat
> > > >> on
> > > >> > all
> > > >> > > > the
> > > >> > > > > model data for my three ensembles. How do I go about
getting
> > > that
> > > >> > into
> > > >> > > > > METViewer so that I can plot spread-skill
information? My
> data
> > > is
> > > >> on
> > > >> > > > > yellowstone. I also looked through the stat-analysis
section
> > of
> > > >> the
> > > >> > MET
> > > >> > > > > Users Guide, but I'm not clear on what exactly I'd
need to
> do
> > to
> > > >> get
> > > >> > > > what I
> > > >> > > > > need with that...
> > > >> > > > >
> > > >> > > > > Jared
> > > >> > > > >
> > > >> > > > > On Mon, Jan 23, 2017 at 5:01 PM, John Halley Gotway
via RT <
> > > >> > > > > met_help at ucar.edu> wrote:
> > > >> > > > >
> > > >> > > > > > Jared,
> > > >> > > > > >
> > > >> > > > > > Unfortunately, no, MET is not set up to read wrfout
files
> > > >> directly.
> > > >> > > As
> > > >> > > > > for
> > > >> > > > > > post-processing, you have 2 options... either UPP
or
> > > wrf_interp
> > > >> > (only
> > > >> > > > for
> > > >> > > > > > WRF-ARW core).
> > > >> > > > > >
> > > >> > > > > > The former writes GRIB output which MET handles
well. The
> > > latter
> > > >> > > > writes a
> > > >> > > > > > flavor of NetCDF that MET is able to read... with
the
> caveat
> > > >> that
> > > >> > it
> > > >> > > > can
> > > >> > > > > > only read fields that are defined on non-staggered
> > dimensions.
> > > >> > > > > Basically,
> > > >> > > > > > it can't read the wind fields.
> > > >> > > > > >
> > > >> > > > > > I always encourage people to use UPP since MET
supports
> GRIB
> > > >> well.
> > > >> > > > > >
> > > >> > > > > > Being able to read wrfout directly has often been
> requested.
> > > >> The
> > > >> > two
> > > >> > > > > > issues are the staggered dimensions and the hybrid
> vertical
> > > >> > > coordinate.
> > > >> > > > > > Generally, matching to point observations is done
in terms
> > of
> > > >> > > pressure
> > > >> > > > or
> > > >> > > > > > height.  So some interpolation to pressure/height
levels
> in
> > > >> > > inevitable
> > > >> > > > > > along the way.  We've resisted re-implementing that
> > > >> post-processing
> > > >> > > > logic
> > > >> > > > > > directly in MET to avoid duplication of effort with
UPP.
> > And
> > > >> UPP
> > > >> > > > > supports
> > > >> > > > > > the derivation of many other variable types that
are
> pretty
> > > >> useful.
> > > >> > > > > >
> > > >> > > > > > Perhaps there is some middle ground where we could
enable
> > MET
> > > to
> > > >> > call
> > > >> > > > UPP
> > > >> > > > > > behind the scenes?  But we'd need to work out many
> details.
> > > >> > > > > >
> > > >> > > > > > Thanks,
> > > >> > > > > > John
> > > >> > > > > >
> > > >> > > > > > On Fri, Jan 20, 2017 at 5:31 PM, Jared Lee via RT <
> > > >> > met_help at ucar.edu
> > > >> > > >
> > > >> > > > > > wrote:
> > > >> > > > > >
> > > >> > > > > > >
> > > >> > > > > > > <URL: https://rt.rap.ucar.edu/rt/
> > > Ticket/Display.html?id=79224
> > > >> >
> > > >> > > > > > >
> > > >> > > > > > > Hi John,
> > > >> > > > > > >
> > > >> > > > > > > Thanks for clarifying what ensemble-stat does to
produce
> > the
> > > >> > SSVAR
> > > >> > > > line
> > > >> > > > > > > type. That helps.
> > > >> > > > > > >
> > > >> > > > > > > As for what I ultimately want, I guess that's
still
> > somewhat
> > > >> > > vaguely
> > > >> > > > > > > defined. I think making plots that show the
spread/skill
> > > >> > > relationship
> > > >> > > > > > > across the date ranges, lead times, and variables
of
> > > interest
> > > >> for
> > > >> > > > both
> > > >> > > > > > > ensembles would be a useful way to illustrate and
> compare
> > > the
> > > >> > > spread
> > > >> > > > of
> > > >> > > > > > the
> > > >> > > > > > > two ensembles. That seems like it would be more
useful
> > than
> > > >> > simply
> > > >> > > > > > > calculating the mean spread of the ensemble
(which I
> could
> > > do
> > > >> by
> > > >> > > > > writing
> > > >> > > > > > a
> > > >> > > > > > > new NCL script to read in all the ensemble fields
and
> not
> > > even
> > > >> > > > > bothering
> > > >> > > > > > > with MET...).
> > > >> > > > > > >
> > > >> > > > > > > Another side question... Can MET read in native-
grid
> > NetCDF
> > > >> WRF
> > > >> > > > output
> > > >> > > > > > > files? Or do I need to run them through some
> pre-processor
> > > >> (like
> > > >> > > UPP)
> > > >> > > > > > > first? After going through the user guide I'm not
clear
> > > >> whether
> > > >> > MET
> > > >> > > > > > accepts
> > > >> > > > > > > NetCDF wrfout files or not without pre-
processing. Also,
> > if
> > > >> all
> > > >> > the
> > > >> > > > > > output
> > > >> > > > > > > times are in a single wrfout file, can MET open
the one
> > file
> > > >> and
> > > >> > > find
> > > >> > > > > the
> > > >> > > > > > > correct time it's supposed to read, or do all the
wrfout
> > > files
> > > >> > need
> > > >> > > > to
> > > >> > > > > be
> > > >> > > > > > > for a single time only? The NetCDF wrfout files
I've
> been
> > > >> given
> > > >> > for
> > > >> > > > the
> > > >> > > > > > > small ensemble have all the output times in a
single
> file,
> > > >> but I
> > > >> > > can
> > > >> > > > > > write
> > > >> > > > > > > a script to parse them out to single-time files
with
> ncks.
> > > I'd
> > > >> > just
> > > >> > > > > > rather
> > > >> > > > > > > not go through that hassle if I don't have to!
> > > >> > > > > > >
> > > >> > > > > > > Jared
> > > >> > > > > > >
> > > >> > > > > > > On Fri, Jan 20, 2017 at 5:07 PM, John Halley
Gotway via
> > RT <
> > > >> > > > > > > met_help at ucar.edu> wrote:
> > > >> > > > > > >
> > > >> > > > > > > > Hi Jared,
> > > >> > > > > > > >
> > > >> > > > > > > > I see a couple of questions in your email.
> > > >> > > > > > > >
> > > >> > > > > > > > First, yes, Ensemble-Stat is set up to process
both
> > point
> > > >> and
> > > >> > > > gridded
> > > >> > > > > > > > observations.  And the SSVAR (spread/skill
variance)
> > line
> > > >> type
> > > >> > > can
> > > >> > > > be
> > > >> > > > > > > > produced using either of these.
> > > >> > > > > > > >
> > > >> > > > > > > > The SSVAR line type stores information about
the
> > > >> relationship
> > > >> > > > between
> > > >> > > > > > > > ensemble spread and skill.  Here's how it
works.
> > > >> Ensemble-Stat
> > > >> > > > > > processes
> > > >> > > > > > > > each observation value... either from a point
location
> > > (i.e.
> > > >> > for
> > > >> > > > > point
> > > >> > > > > > > obs)
> > > >> > > > > > > > or from a gridded location (i.e. gridded obs).
For
> each
> > > >> > > > observation,
> > > >> > > > > > it
> > > >> > > > > > > > gathers the N ensemble forecast values which
> correspond
> > to
> > > >> it.
> > > >> > > It
> > > >> > > > > > > computes
> > > >> > > > > > > > the variance of those ensemble values, and it
also
> > > computes
> > > >> the
> > > >> > > > error
> > > >> > > > > > by
> > > >> > > > > > > > subtracting the mean ensemble value minus the
> > observation
> > > >> > value.
> > > >> > > > > > (Note,
> > > >> > > > > > > > you can pass in an ensemble mean field instead
of
> > > computing
> > > >> it
> > > >> > on
> > > >> > > > the
> > > >> > > > > > fly
> > > >> > > > > > > > like this).  After processing N observations,
we have
> N
> > > >> pairs
> > > >> > of
> > > >> > > > > > variance
> > > >> > > > > > > > values and error values.  It groups those N
pairs into
> > > >> buckets
> > > >> > by
> > > >> > > > > > binning
> > > >> > > > > > > > the variance using the bucket size defined in
the
> config
> > > >> file.
> > > >> > > > > You'll
> > > >> > > > > > > see
> > > >> > > > > > > > in the output one SSVAR line for each of those
> variance
> > > >> > buckets.
> > > >> > > > > > > >
> > > >> > > > > > > > The METViewer database and display system (or
> > > Stat-Analysis)
> > > >> > can
> > > >> > > > > > > aggregate
> > > >> > > > > > > > those SSVAR lines intelligently across multiple
cases
> > and
> > > >> > create
> > > >> > > > > plots
> > > >> > > > > > > > which show the relationship between spread and
skill.
> > > >> > > > > > > >
> > > >> > > > > > > > If that's what you need, great.
> > > >> > > > > > > >
> > > >> > > > > > > > However, if your looking for a much simpler
> computation
> > of
> > > >> mean
> > > >> > > > > spread
> > > >> > > > > > > over
> > > >> > > > > > > > some spatial region, unfortunately, the answer
is no.
> > For
> > > >> > > example,
> > > >> > > > > > > > Ensemble-Stat doesn't write out the mean spread
value
> > over
> > > >> the
> > > >> > > > entire
> > > >> > > > > > > > domain.
> > > >> > > > > > > >
> > > >> > > > > > > > Just let me know what questions you still have.
> > > >> > > > > > > >
> > > >> > > > > > > > Thanks,
> > > >> > > > > > > > John Halley Gotway
> > > >> > > > > > > >
> > > >> > > > > > > >
> > > >> > > > > > > > On Fri, Jan 20, 2017 at 11:41 AM, Jared Lee via
RT <
> > > >> > > > > met_help at ucar.edu>
> > > >> > > > > > > > wrote:
> > > >> > > > > > > >
> > > >> > > > > > > > >
> > > >> > > > > > > > > Fri Jan 20 11:41:03 2017: Request 79224 was
acted
> > upon.
> > > >> > > > > > > > > Transaction: Ticket created by
jaredlee at ucar.edu
> > > >> > > > > > > > >        Queue: met_help
> > > >> > > > > > > > >      Subject: calculating ensemble spread
> > > >> > > > > > > > >        Owner: Nobody
> > > >> > > > > > > > >   Requestors: jaredlee at ucar.edu
> > > >> > > > > > > > >       Status: new
> > > >> > > > > > > > >  Ticket <URL: https://rt.rap.ucar.edu/rt/
> > > >> > > > > > Ticket/Display.html?id=79224
> > > >> > > > > > > >
> > > >> > > > > > > > >
> > > >> > > > > > > > >
> > > >> > > > > > > > > Hi, for my project I need to calculate some
measure
> of
> > > the
> > > >> > > spread
> > > >> > > > > of
> > > >> > > > > > > two
> > > >> > > > > > > > > different ensembles (one "large" and one
"small"
> > > >> ensemble). I
> > > >> > > > have
> > > >> > > > > > > model
> > > >> > > > > > > > > output data (the large ensemble is in grib2
format,
> > the
> > > >> small
> > > >> > > > > > ensemble
> > > >> > > > > > > is
> > > >> > > > > > > > > WRF NetCDF), and I have MADIS point
observations
> that
> > > have
> > > >> > > > already
> > > >> > > > > > been
> > > >> > > > > > > > > processed for use in MET.
> > > >> > > > > > > > >
> > > >> > > > > > > > > I previously used the ensemble-stat tool on
the
> large
> > > >> > ensemble
> > > >> > > to
> > > >> > > > > > dump
> > > >> > > > > > > > out
> > > >> > > > > > > > > matched pair info for all ensemble member
forecasts
> at
> > > >> each
> > > >> > > > > > observation
> > > >> > > > > > > > > location by printing out the ORANK line type.
Given
> > the
> > > >> > inputs
> > > >> > > > > that I
> > > >> > > > > > > > > described above, will MET calculate the
ensemble
> > spread
> > > >> > > > > automatically
> > > >> > > > > > > if
> > > >> > > > > > > > I
> > > >> > > > > > > > > turn on the SSVAR line type? In the user
guide
> (v5.2)
> > it
> > > >> > > mentions
> > > >> > > > > > that
> > > >> > > > > > > an
> > > >> > > > > > > > > "observation field" is needed to calculate
things
> like
> > > >> > > > spread/skill
> > > >> > > > > > > > > variance. Do point obs from MADIS count as an
> > > "observation
> > > >> > > > field"?
> > > >> > > > > > > > >
> > > >> > > > > > > > > Guidance would be much appreciated!
> > > >> > > > > > > > >
> > > >> > > > > > > > > Jared
> > > >> > > > > > > > >
> > > >> > > > > > > > > --
> > > >> > > > > > > > > ===============================
> > > >> > > > > > > > > Jared A. Lee, Ph.D.
> > > >> > > > > > > > > Project Scientist I
> > > >> > > > > > > > > Research Applications Laboratory
> > > >> > > > > > > > > National Center for Atmospheric Research
> > > >> > > > > > > > > Boulder, Colorado, USA
> > > >> > > > > > > > >
> > > >> > > > > > > > > Email: jaredlee at ucar.edu (w)
> > > >> > > > > > > > > Phone: 303.497.8485 (w)
> > > >> > > > > > > > > Web: https://staff.ucar.edu/users/jaredlee
> > > >> > > > > > > > > ===============================
> > > >> > > > > > > > >
> > > >> > > > > > > > >
> > > >> > > > > > > >
> > > >> > > > > > > >
> > > >> > > > > > >
> > > >> > > > > > >
> > > >> > > > > > > --
> > > >> > > > > > > ===============================
> > > >> > > > > > > Jared A. Lee, Ph.D.
> > > >> > > > > > > Project Scientist I
> > > >> > > > > > > Research Applications Laboratory
> > > >> > > > > > > National Center for Atmospheric Research
> > > >> > > > > > > Boulder, Colorado, USA
> > > >> > > > > > >
> > > >> > > > > > > Email: jaredlee at ucar.edu (w)
> > > >> > > > > > > Phone: 303.497.8485 (w)
> > > >> > > > > > > Web: https://staff.ucar.edu/users/jaredlee
> > > >> > > > > > > ===============================
> > > >> > > > > > >
> > > >> > > > > > >
> > > >> > > > > >
> > > >> > > > > >
> > > >> > > > >
> > > >> > > > >
> > > >> > > > > --
> > > >> > > > > ===============================
> > > >> > > > > Jared A. Lee, Ph.D.
> > > >> > > > > Project Scientist I
> > > >> > > > > Research Applications Laboratory
> > > >> > > > > National Center for Atmospheric Research
> > > >> > > > > Boulder, Colorado, USA
> > > >> > > > >
> > > >> > > > > Email: jaredlee at ucar.edu (w)
> > > >> > > > > Phone: 303.497.8485 (w)
> > > >> > > > > Web: https://staff.ucar.edu/users/jaredlee
> > > >> > > > > ===============================
> > > >> > > > >
> > > >> > > > >
> > > >> > > >
> > > >> > > >
> > > >> > >
> > > >> > >
> > > >> > > --
> > > >> > > ===============================
> > > >> > > Jared A. Lee, Ph.D.
> > > >> > > Project Scientist I
> > > >> > > Research Applications Laboratory
> > > >> > > National Center for Atmospheric Research
> > > >> > > Boulder, Colorado, USA
> > > >> > >
> > > >> > > Email: jaredlee at ucar.edu (w)
> > > >> > > Phone: 303.497.8485 (w)
> > > >> > > Web: https://staff.ucar.edu/users/jaredlee
> > > >> > > ===============================
> > > >> > >
> > > >> > >
> > > >> >
> > > >> >
> > > >>
> > > >>
> > > >> --
> > > >> ===============================
> > > >> Jared A. Lee, Ph.D.
> > > >> Project Scientist I
> > > >> Research Applications Laboratory
> > > >> National Center for Atmospheric Research
> > > >> Boulder, Colorado, USA
> > > >>
> > > >> Email: jaredlee at ucar.edu (w)
> > > >> Phone: 303.497.8485 (w)
> > > >> Web: https://staff.ucar.edu/users/jaredlee
> > > >> ===============================
> > > >>
> > > >>
> > > >
> > >
> > >
> >
> >
> > --
> > ===============================
> > Jared A. Lee, Ph.D.
> > Project Scientist I
> > Research Applications Laboratory
> > National Center for Atmospheric Research
> > Boulder, Colorado, USA
> >
> > Email: jaredlee at ucar.edu (w)
> > Phone: 303.497.8485 (w)
> > Web: https://staff.ucar.edu/users/jaredlee
> > ===============================
> >
> >
>
>


--
===============================
Jared A. Lee, Ph.D.
Project Scientist I
Research Applications Laboratory
National Center for Atmospheric Research
Boulder, Colorado, USA

Email: jaredlee at ucar.edu (w)
Phone: 303.497.8485 (w)
Web: https://staff.ucar.edu/users/jaredlee
===============================

------------------------------------------------
Subject: calculating ensemble spread
From: John Halley Gotway
Time: Tue Feb 07 16:33:20 2017

After clicking the reload databases button and doing a cntrl-refresh
on my
browser (about 5 times), I was able to get SREF to switch to ALL_SREF
in
the names of the models listed.

If you're loading an XML that uses "SREF", you can just edit it to say
"ALL_SREF" instead prior to loading.

Hope that fixes it for you.

John

On Tue, Feb 7, 2017 at 3:52 PM, Jared Lee via RT <met_help at ucar.edu>
wrote:

>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
>
> Hi John,
>
> Thanks! I just tried to make a plot, and I think something with the
> underlying xml specifying the database needs to be updated. I
updated the
> name of one of the ensembles from "SREF" to "ALL_SREF" to be
consistent
> with my naming conventions elsewhere in my scripts and file
structure.
> Right now METViewer is expecting to see model="SREF" as a holdover
from
> yesterday but is not finding anything in the database that matches.
>
> Jared
>
> On Tue, Feb 7, 2017 at 3:43 PM, John Halley Gotway via RT <
> met_help at ucar.edu
> > wrote:
>
> > Jared,
> >
> > Forgot to write back earlier... the reload finished at 12:16pm
today.
> >
> > Please take a look and let me know if you run into any issues.
> >
> > I'll go ahead and resolve this ticket... and hopefully Tatiana
will fix
> the
> > issue with the spread-skill legend in the next METViewer release.
> >
> > Thanks,
> > John
> >
> > On Tue, Feb 7, 2017 at 12:13 PM, Jared Lee via RT
<met_help at ucar.edu>
> > wrote:
> >
> > >
> > > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
> > >
> > > No worries. I suppose you might need to update the xml again,
unless it
> > > picks up the model names automatically from the stat files.
Anyway, the
> > > model names I have in the stat files are now:
> > >
> > > ALL_SREF
> > > ARW_ONLY
> > > 3MEM
> > >
> > > Jared
> > >
> > > On Tue, Feb 7, 2017 at 12:12 PM, John Halley Gotway via RT <
> > > met_help at ucar.edu> wrote:
> > >
> > > > Sorry... should have checked first.  Looks like you already
updated
> > them.
> > > >
> > > > I'm reloading now.
> > > >
> > > > On Tue, Feb 7, 2017 at 12:10 PM, John Halley Gotway
<johnhg at ucar.edu
> >
> > > > wrote:
> > > >
> > > > > Jared,
> > > > >
> > > > > Will do.  Do I need to update the model names again?
> > > > >
> > > > > SREF -> 3MEM
> > > > > SREF -> ARW_ONLY
> > > > >
> > > > > John
> > > > >
> > > > > On Tue, Feb 7, 2017 at 12:04 PM, Jared Lee via RT <
> met_help at ucar.edu
> > >
> > > > > wrote:
> > > > >
> > > > >>
> > > > >> <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
> > > > >>
> > > > >> Hi John,
> > > > >>
> > > > >> I've finished re-generating all the .stat files (including
with
> the
> > > > >> "model"
> > > > >> labels I want). The zipped file is 405 MB and is located
here on
> > > > >> yellowstone:
> > > > >>
> > > > >>
/glade/p/ral/wsap/jaredlee/NOAA_AQ_Ensemble/MET/15day_data.tar.gz
> > > > >>
> > > > >> Could you please replace all the data in the
"noaa_aq_15day"
> > METViewer
> > > > >> database with that? Thanks!
> > > > >>
> > > > >> Jared
> > > > >>
> > > > >> On Sat, Feb 4, 2017 at 6:47 PM, John Halley Gotway via RT <
> > > > >> met_help at ucar.edu
> > > > >> > wrote:
> > > > >>
> > > > >> > Jared,
> > > > >> >
> > > > >> > Just come by my office sometime next week and we'll copy
your
> data
> > > > down
> > > > >> to
> > > > >> > a project machine and load it into METViewer.
> > > > >> >
> > > > >> > My calendar should be up to date if you want to schedule
> something
> > > > >> through
> > > > >> > google calendar.
> > > > >> >
> > > > >> > John
> > > > >> >
> > > > >> > On Fri, Feb 3, 2017 at 9:22 PM Jared Lee via RT <
> > met_help at ucar.edu>
> > > > >> wrote:
> > > > >> >
> > > > >> > >
> > > > >> > > <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224
> >
> > > > >> > >
> > > > >> > > Hi John,
> > > > >> > >
> > > > >> > > I was unaware that UPP had a script to do that. As it
turns
> out,
> > > > some
> > > > >> > > members had all 48 hours in one file, and some had them
split
> > > > between
> > > > >> two
> > > > >> > > different files, so as it turns out it probably was
simpler to
> > > break
> > > > >> them
> > > > >> > > up first with ncks anyway. In the future I'll have the
person
> > > > running
> > > > >> WRF
> > > > >> > > to output single-time files from the get-go, so now my
scripts
> > > will
> > > > >> work
> > > > >> > > for that.
> > > > >> > >
> > > > >> > > Does my laptop 'shadowfax' count as a machine within
the RAL
> > > > firewall
> > > > >> if
> > > > >> > > I'm physically at NCAR? What if I use VPN when I'm
offsite?
> > > > Otherwise
> > > > >> I
> > > > >> > > could try to find some other machine to temporarily
house my
> > data
> > > > for
> > > > >> > > METViewer to access.
> > > > >> > >
> > > > >> > > Jared
> > > > >> > >
> > > > >> > > On Fri, Feb 3, 2017 at 4:36 PM, John Halley Gotway via
RT <
> > > > >> > > met_help at ucar.edu
> > > > >> > > > wrote:
> > > > >> > >
> > > > >> > > > Hi Jared,
> > > > >> > > >
> > > > >> > > > I'm surprised you need to use ncks to split out
multiple
> > times.
> > > > >> Isn't
> > > > >> > > > there a script for running UPP on a wrfout file with
> multiple
> > > > >> times?  I
> > > > >> > > > believe it's called run_unipost_frames.
> > > > >> > > >
> > > > >> > > > As for your question about METViewer... you'd need to
pull
> the
> > > > >> > > > ensemble-stat data down to a machine inside the RAL
> firewall.
> > > And
> > > > >> > then I
> > > > >> > > > could demonstrate to you how to load it into
METViewer and
> > start
> > > > >> > creating
> > > > >> > > > plots.
> > > > >> > > >
> > > > >> > > > Unfortunately, you can't load it directly from
yellowstone
> > into
> > > > >> > > METViewer,
> > > > >> > > > which lives on a machine named "mandan".
> > > > >> > > >
> > > > >> > > > John
> > > > >> > > >
> > > > >> > > > On Fri, Feb 3, 2017 at 3:09 PM, Jared Lee via RT <
> > > > met_help at ucar.edu
> > > > >> >
> > > > >> > > > wrote:
> > > > >> > > >
> > > > >> > > > >
> > > > >> > > > > <URL: https://rt.rap.ucar.edu/rt/
> > Ticket/Display.html?id=79224
> > > >
> > > > >> > > > >
> > > > >> > > > > Hi John,
> > > > >> > > > >
> > > > >> > > > > Now I finally have finished running UPP (after
having to
> > split
> > > > >> apart
> > > > >> > > the
> > > > >> > > > > wrfout files into single-time files with ncks) and
> > > ensemble-stat
> > > > >> on
> > > > >> > all
> > > > >> > > > the
> > > > >> > > > > model data for my three ensembles. How do I go
about
> getting
> > > > that
> > > > >> > into
> > > > >> > > > > METViewer so that I can plot spread-skill
information? My
> > data
> > > > is
> > > > >> on
> > > > >> > > > > yellowstone. I also looked through the stat-
analysis
> section
> > > of
> > > > >> the
> > > > >> > MET
> > > > >> > > > > Users Guide, but I'm not clear on what exactly I'd
need to
> > do
> > > to
> > > > >> get
> > > > >> > > > what I
> > > > >> > > > > need with that...
> > > > >> > > > >
> > > > >> > > > > Jared
> > > > >> > > > >
> > > > >> > > > > On Mon, Jan 23, 2017 at 5:01 PM, John Halley Gotway
via
> RT <
> > > > >> > > > > met_help at ucar.edu> wrote:
> > > > >> > > > >
> > > > >> > > > > > Jared,
> > > > >> > > > > >
> > > > >> > > > > > Unfortunately, no, MET is not set up to read
wrfout
> files
> > > > >> directly.
> > > > >> > > As
> > > > >> > > > > for
> > > > >> > > > > > post-processing, you have 2 options... either UPP
or
> > > > wrf_interp
> > > > >> > (only
> > > > >> > > > for
> > > > >> > > > > > WRF-ARW core).
> > > > >> > > > > >
> > > > >> > > > > > The former writes GRIB output which MET handles
well.
> The
> > > > latter
> > > > >> > > > writes a
> > > > >> > > > > > flavor of NetCDF that MET is able to read... with
the
> > caveat
> > > > >> that
> > > > >> > it
> > > > >> > > > can
> > > > >> > > > > > only read fields that are defined on non-
staggered
> > > dimensions.
> > > > >> > > > > Basically,
> > > > >> > > > > > it can't read the wind fields.
> > > > >> > > > > >
> > > > >> > > > > > I always encourage people to use UPP since MET
supports
> > GRIB
> > > > >> well.
> > > > >> > > > > >
> > > > >> > > > > > Being able to read wrfout directly has often been
> > requested.
> > > > >> The
> > > > >> > two
> > > > >> > > > > > issues are the staggered dimensions and the
hybrid
> > vertical
> > > > >> > > coordinate.
> > > > >> > > > > > Generally, matching to point observations is done
in
> terms
> > > of
> > > > >> > > pressure
> > > > >> > > > or
> > > > >> > > > > > height.  So some interpolation to pressure/height
levels
> > in
> > > > >> > > inevitable
> > > > >> > > > > > along the way.  We've resisted re-implementing
that
> > > > >> post-processing
> > > > >> > > > logic
> > > > >> > > > > > directly in MET to avoid duplication of effort
with UPP.
> > > And
> > > > >> UPP
> > > > >> > > > > supports
> > > > >> > > > > > the derivation of many other variable types that
are
> > pretty
> > > > >> useful.
> > > > >> > > > > >
> > > > >> > > > > > Perhaps there is some middle ground where we
could
> enable
> > > MET
> > > > to
> > > > >> > call
> > > > >> > > > UPP
> > > > >> > > > > > behind the scenes?  But we'd need to work out
many
> > details.
> > > > >> > > > > >
> > > > >> > > > > > Thanks,
> > > > >> > > > > > John
> > > > >> > > > > >
> > > > >> > > > > > On Fri, Jan 20, 2017 at 5:31 PM, Jared Lee via RT
<
> > > > >> > met_help at ucar.edu
> > > > >> > > >
> > > > >> > > > > > wrote:
> > > > >> > > > > >
> > > > >> > > > > > >
> > > > >> > > > > > > <URL: https://rt.rap.ucar.edu/rt/
> > > > Ticket/Display.html?id=79224
> > > > >> >
> > > > >> > > > > > >
> > > > >> > > > > > > Hi John,
> > > > >> > > > > > >
> > > > >> > > > > > > Thanks for clarifying what ensemble-stat does
to
> produce
> > > the
> > > > >> > SSVAR
> > > > >> > > > line
> > > > >> > > > > > > type. That helps.
> > > > >> > > > > > >
> > > > >> > > > > > > As for what I ultimately want, I guess that's
still
> > > somewhat
> > > > >> > > vaguely
> > > > >> > > > > > > defined. I think making plots that show the
> spread/skill
> > > > >> > > relationship
> > > > >> > > > > > > across the date ranges, lead times, and
variables of
> > > > interest
> > > > >> for
> > > > >> > > > both
> > > > >> > > > > > > ensembles would be a useful way to illustrate
and
> > compare
> > > > the
> > > > >> > > spread
> > > > >> > > > of
> > > > >> > > > > > the
> > > > >> > > > > > > two ensembles. That seems like it would be more
useful
> > > than
> > > > >> > simply
> > > > >> > > > > > > calculating the mean spread of the ensemble
(which I
> > could
> > > > do
> > > > >> by
> > > > >> > > > > writing
> > > > >> > > > > > a
> > > > >> > > > > > > new NCL script to read in all the ensemble
fields and
> > not
> > > > even
> > > > >> > > > > bothering
> > > > >> > > > > > > with MET...).
> > > > >> > > > > > >
> > > > >> > > > > > > Another side question... Can MET read in
native-grid
> > > NetCDF
> > > > >> WRF
> > > > >> > > > output
> > > > >> > > > > > > files? Or do I need to run them through some
> > pre-processor
> > > > >> (like
> > > > >> > > UPP)
> > > > >> > > > > > > first? After going through the user guide I'm
not
> clear
> > > > >> whether
> > > > >> > MET
> > > > >> > > > > > accepts
> > > > >> > > > > > > NetCDF wrfout files or not without pre-
processing.
> Also,
> > > if
> > > > >> all
> > > > >> > the
> > > > >> > > > > > output
> > > > >> > > > > > > times are in a single wrfout file, can MET open
the
> one
> > > file
> > > > >> and
> > > > >> > > find
> > > > >> > > > > the
> > > > >> > > > > > > correct time it's supposed to read, or do all
the
> wrfout
> > > > files
> > > > >> > need
> > > > >> > > > to
> > > > >> > > > > be
> > > > >> > > > > > > for a single time only? The NetCDF wrfout files
I've
> > been
> > > > >> given
> > > > >> > for
> > > > >> > > > the
> > > > >> > > > > > > small ensemble have all the output times in a
single
> > file,
> > > > >> but I
> > > > >> > > can
> > > > >> > > > > > write
> > > > >> > > > > > > a script to parse them out to single-time files
with
> > ncks.
> > > > I'd
> > > > >> > just
> > > > >> > > > > > rather
> > > > >> > > > > > > not go through that hassle if I don't have to!
> > > > >> > > > > > >
> > > > >> > > > > > > Jared
> > > > >> > > > > > >
> > > > >> > > > > > > On Fri, Jan 20, 2017 at 5:07 PM, John Halley
Gotway
> via
> > > RT <
> > > > >> > > > > > > met_help at ucar.edu> wrote:
> > > > >> > > > > > >
> > > > >> > > > > > > > Hi Jared,
> > > > >> > > > > > > >
> > > > >> > > > > > > > I see a couple of questions in your email.
> > > > >> > > > > > > >
> > > > >> > > > > > > > First, yes, Ensemble-Stat is set up to
process both
> > > point
> > > > >> and
> > > > >> > > > gridded
> > > > >> > > > > > > > observations.  And the SSVAR (spread/skill
variance)
> > > line
> > > > >> type
> > > > >> > > can
> > > > >> > > > be
> > > > >> > > > > > > > produced using either of these.
> > > > >> > > > > > > >
> > > > >> > > > > > > > The SSVAR line type stores information about
the
> > > > >> relationship
> > > > >> > > > between
> > > > >> > > > > > > > ensemble spread and skill.  Here's how it
works.
> > > > >> Ensemble-Stat
> > > > >> > > > > > processes
> > > > >> > > > > > > > each observation value... either from a point
> location
> > > > (i.e.
> > > > >> > for
> > > > >> > > > > point
> > > > >> > > > > > > obs)
> > > > >> > > > > > > > or from a gridded location (i.e. gridded
obs).  For
> > each
> > > > >> > > > observation,
> > > > >> > > > > > it
> > > > >> > > > > > > > gathers the N ensemble forecast values which
> > correspond
> > > to
> > > > >> it.
> > > > >> > > It
> > > > >> > > > > > > computes
> > > > >> > > > > > > > the variance of those ensemble values, and it
also
> > > > computes
> > > > >> the
> > > > >> > > > error
> > > > >> > > > > > by
> > > > >> > > > > > > > subtracting the mean ensemble value minus the
> > > observation
> > > > >> > value.
> > > > >> > > > > > (Note,
> > > > >> > > > > > > > you can pass in an ensemble mean field
instead of
> > > > computing
> > > > >> it
> > > > >> > on
> > > > >> > > > the
> > > > >> > > > > > fly
> > > > >> > > > > > > > like this).  After processing N observations,
we
> have
> > N
> > > > >> pairs
> > > > >> > of
> > > > >> > > > > > variance
> > > > >> > > > > > > > values and error values.  It groups those N
pairs
> into
> > > > >> buckets
> > > > >> > by
> > > > >> > > > > > binning
> > > > >> > > > > > > > the variance using the bucket size defined in
the
> > config
> > > > >> file.
> > > > >> > > > > You'll
> > > > >> > > > > > > see
> > > > >> > > > > > > > in the output one SSVAR line for each of
those
> > variance
> > > > >> > buckets.
> > > > >> > > > > > > >
> > > > >> > > > > > > > The METViewer database and display system (or
> > > > Stat-Analysis)
> > > > >> > can
> > > > >> > > > > > > aggregate
> > > > >> > > > > > > > those SSVAR lines intelligently across
multiple
> cases
> > > and
> > > > >> > create
> > > > >> > > > > plots
> > > > >> > > > > > > > which show the relationship between spread
and
> skill.
> > > > >> > > > > > > >
> > > > >> > > > > > > > If that's what you need, great.
> > > > >> > > > > > > >
> > > > >> > > > > > > > However, if your looking for a much simpler
> > computation
> > > of
> > > > >> mean
> > > > >> > > > > spread
> > > > >> > > > > > > over
> > > > >> > > > > > > > some spatial region, unfortunately, the
answer is
> no.
> > > For
> > > > >> > > example,
> > > > >> > > > > > > > Ensemble-Stat doesn't write out the mean
spread
> value
> > > over
> > > > >> the
> > > > >> > > > entire
> > > > >> > > > > > > > domain.
> > > > >> > > > > > > >
> > > > >> > > > > > > > Just let me know what questions you still
have.
> > > > >> > > > > > > >
> > > > >> > > > > > > > Thanks,
> > > > >> > > > > > > > John Halley Gotway
> > > > >> > > > > > > >
> > > > >> > > > > > > >
> > > > >> > > > > > > > On Fri, Jan 20, 2017 at 11:41 AM, Jared Lee
via RT <
> > > > >> > > > > met_help at ucar.edu>
> > > > >> > > > > > > > wrote:
> > > > >> > > > > > > >
> > > > >> > > > > > > > >
> > > > >> > > > > > > > > Fri Jan 20 11:41:03 2017: Request 79224 was
acted
> > > upon.
> > > > >> > > > > > > > > Transaction: Ticket created by
jaredlee at ucar.edu
> > > > >> > > > > > > > >        Queue: met_help
> > > > >> > > > > > > > >      Subject: calculating ensemble spread
> > > > >> > > > > > > > >        Owner: Nobody
> > > > >> > > > > > > > >   Requestors: jaredlee at ucar.edu
> > > > >> > > > > > > > >       Status: new
> > > > >> > > > > > > > >  Ticket <URL: https://rt.rap.ucar.edu/rt/
> > > > >> > > > > > Ticket/Display.html?id=79224
> > > > >> > > > > > > >
> > > > >> > > > > > > > >
> > > > >> > > > > > > > >
> > > > >> > > > > > > > > Hi, for my project I need to calculate some
> measure
> > of
> > > > the
> > > > >> > > spread
> > > > >> > > > > of
> > > > >> > > > > > > two
> > > > >> > > > > > > > > different ensembles (one "large" and one
"small"
> > > > >> ensemble). I
> > > > >> > > > have
> > > > >> > > > > > > model
> > > > >> > > > > > > > > output data (the large ensemble is in grib2
> format,
> > > the
> > > > >> small
> > > > >> > > > > > ensemble
> > > > >> > > > > > > is
> > > > >> > > > > > > > > WRF NetCDF), and I have MADIS point
observations
> > that
> > > > have
> > > > >> > > > already
> > > > >> > > > > > been
> > > > >> > > > > > > > > processed for use in MET.
> > > > >> > > > > > > > >
> > > > >> > > > > > > > > I previously used the ensemble-stat tool on
the
> > large
> > > > >> > ensemble
> > > > >> > > to
> > > > >> > > > > > dump
> > > > >> > > > > > > > out
> > > > >> > > > > > > > > matched pair info for all ensemble member
> forecasts
> > at
> > > > >> each
> > > > >> > > > > > observation
> > > > >> > > > > > > > > location by printing out the ORANK line
type.
> Given
> > > the
> > > > >> > inputs
> > > > >> > > > > that I
> > > > >> > > > > > > > > described above, will MET calculate the
ensemble
> > > spread
> > > > >> > > > > automatically
> > > > >> > > > > > > if
> > > > >> > > > > > > > I
> > > > >> > > > > > > > > turn on the SSVAR line type? In the user
guide
> > (v5.2)
> > > it
> > > > >> > > mentions
> > > > >> > > > > > that
> > > > >> > > > > > > an
> > > > >> > > > > > > > > "observation field" is needed to calculate
things
> > like
> > > > >> > > > spread/skill
> > > > >> > > > > > > > > variance. Do point obs from MADIS count as
an
> > > > "observation
> > > > >> > > > field"?
> > > > >> > > > > > > > >
> > > > >> > > > > > > > > Guidance would be much appreciated!
> > > > >> > > > > > > > >
> > > > >> > > > > > > > > Jared
> > > > >> > > > > > > > >
> > > > >> > > > > > > > > --
> > > > >> > > > > > > > > ===============================
> > > > >> > > > > > > > > Jared A. Lee, Ph.D.
> > > > >> > > > > > > > > Project Scientist I
> > > > >> > > > > > > > > Research Applications Laboratory
> > > > >> > > > > > > > > National Center for Atmospheric Research
> > > > >> > > > > > > > > Boulder, Colorado, USA
> > > > >> > > > > > > > >
> > > > >> > > > > > > > > Email: jaredlee at ucar.edu (w)
> > > > >> > > > > > > > > Phone: 303.497.8485 (w)
> > > > >> > > > > > > > > Web: https://staff.ucar.edu/users/jaredlee
> > > > >> > > > > > > > > ===============================
> > > > >> > > > > > > > >
> > > > >> > > > > > > > >
> > > > >> > > > > > > >
> > > > >> > > > > > > >
> > > > >> > > > > > >
> > > > >> > > > > > >
> > > > >> > > > > > > --
> > > > >> > > > > > > ===============================
> > > > >> > > > > > > Jared A. Lee, Ph.D.
> > > > >> > > > > > > Project Scientist I
> > > > >> > > > > > > Research Applications Laboratory
> > > > >> > > > > > > National Center for Atmospheric Research
> > > > >> > > > > > > Boulder, Colorado, USA
> > > > >> > > > > > >
> > > > >> > > > > > > Email: jaredlee at ucar.edu (w)
> > > > >> > > > > > > Phone: 303.497.8485 (w)
> > > > >> > > > > > > Web: https://staff.ucar.edu/users/jaredlee
> > > > >> > > > > > > ===============================
> > > > >> > > > > > >
> > > > >> > > > > > >
> > > > >> > > > > >
> > > > >> > > > > >
> > > > >> > > > >
> > > > >> > > > >
> > > > >> > > > > --
> > > > >> > > > > ===============================
> > > > >> > > > > Jared A. Lee, Ph.D.
> > > > >> > > > > Project Scientist I
> > > > >> > > > > Research Applications Laboratory
> > > > >> > > > > National Center for Atmospheric Research
> > > > >> > > > > Boulder, Colorado, USA
> > > > >> > > > >
> > > > >> > > > > Email: jaredlee at ucar.edu (w)
> > > > >> > > > > Phone: 303.497.8485 (w)
> > > > >> > > > > Web: https://staff.ucar.edu/users/jaredlee
> > > > >> > > > > ===============================
> > > > >> > > > >
> > > > >> > > > >
> > > > >> > > >
> > > > >> > > >
> > > > >> > >
> > > > >> > >
> > > > >> > > --
> > > > >> > > ===============================
> > > > >> > > Jared A. Lee, Ph.D.
> > > > >> > > Project Scientist I
> > > > >> > > Research Applications Laboratory
> > > > >> > > National Center for Atmospheric Research
> > > > >> > > Boulder, Colorado, USA
> > > > >> > >
> > > > >> > > Email: jaredlee at ucar.edu (w)
> > > > >> > > Phone: 303.497.8485 (w)
> > > > >> > > Web: https://staff.ucar.edu/users/jaredlee
> > > > >> > > ===============================
> > > > >> > >
> > > > >> > >
> > > > >> >
> > > > >> >
> > > > >>
> > > > >>
> > > > >> --
> > > > >> ===============================
> > > > >> Jared A. Lee, Ph.D.
> > > > >> Project Scientist I
> > > > >> Research Applications Laboratory
> > > > >> National Center for Atmospheric Research
> > > > >> Boulder, Colorado, USA
> > > > >>
> > > > >> Email: jaredlee at ucar.edu (w)
> > > > >> Phone: 303.497.8485 (w)
> > > > >> Web: https://staff.ucar.edu/users/jaredlee
> > > > >> ===============================
> > > > >>
> > > > >>
> > > > >
> > > >
> > > >
> > >
> > >
> > > --
> > > ===============================
> > > Jared A. Lee, Ph.D.
> > > Project Scientist I
> > > Research Applications Laboratory
> > > National Center for Atmospheric Research
> > > Boulder, Colorado, USA
> > >
> > > Email: jaredlee at ucar.edu (w)
> > > Phone: 303.497.8485 (w)
> > > Web: https://staff.ucar.edu/users/jaredlee
> > > ===============================
> > >
> > >
> >
> >
>
>
> --
> ===============================
> Jared A. Lee, Ph.D.
> Project Scientist I
> Research Applications Laboratory
> National Center for Atmospheric Research
> Boulder, Colorado, USA
>
> Email: jaredlee at ucar.edu (w)
> Phone: 303.497.8485 (w)
> Web: https://staff.ucar.edu/users/jaredlee
> ===============================
>
>

------------------------------------------------
Subject: calculating ensemble spread
From: Jared Lee
Time: Tue Feb 07 16:36:00 2017

Thanks, it works now!

Jared

On Tue, Feb 7, 2017 at 4:33 PM, John Halley Gotway via RT
<met_help at ucar.edu
> wrote:

> After clicking the reload databases button and doing a cntrl-refresh
on my
> browser (about 5 times), I was able to get SREF to switch to
ALL_SREF in
> the names of the models listed.
>
> If you're loading an XML that uses "SREF", you can just edit it to
say
> "ALL_SREF" instead prior to loading.
>
> Hope that fixes it for you.
>
> John
>
> On Tue, Feb 7, 2017 at 3:52 PM, Jared Lee via RT <met_help at ucar.edu>
> wrote:
>
> >
> > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
> >
> > Hi John,
> >
> > Thanks! I just tried to make a plot, and I think something with
the
> > underlying xml specifying the database needs to be updated. I
updated the
> > name of one of the ensembles from "SREF" to "ALL_SREF" to be
consistent
> > with my naming conventions elsewhere in my scripts and file
structure.
> > Right now METViewer is expecting to see model="SREF" as a holdover
from
> > yesterday but is not finding anything in the database that
matches.
> >
> > Jared
> >
> > On Tue, Feb 7, 2017 at 3:43 PM, John Halley Gotway via RT <
> > met_help at ucar.edu
> > > wrote:
> >
> > > Jared,
> > >
> > > Forgot to write back earlier... the reload finished at 12:16pm
today.
> > >
> > > Please take a look and let me know if you run into any issues.
> > >
> > > I'll go ahead and resolve this ticket... and hopefully Tatiana
will fix
> > the
> > > issue with the spread-skill legend in the next METViewer
release.
> > >
> > > Thanks,
> > > John
> > >
> > > On Tue, Feb 7, 2017 at 12:13 PM, Jared Lee via RT
<met_help at ucar.edu>
> > > wrote:
> > >
> > > >
> > > > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224
>
> > > >
> > > > No worries. I suppose you might need to update the xml again,
unless
> it
> > > > picks up the model names automatically from the stat files.
Anyway,
> the
> > > > model names I have in the stat files are now:
> > > >
> > > > ALL_SREF
> > > > ARW_ONLY
> > > > 3MEM
> > > >
> > > > Jared
> > > >
> > > > On Tue, Feb 7, 2017 at 12:12 PM, John Halley Gotway via RT <
> > > > met_help at ucar.edu> wrote:
> > > >
> > > > > Sorry... should have checked first.  Looks like you already
updated
> > > them.
> > > > >
> > > > > I'm reloading now.
> > > > >
> > > > > On Tue, Feb 7, 2017 at 12:10 PM, John Halley Gotway <
> johnhg at ucar.edu
> > >
> > > > > wrote:
> > > > >
> > > > > > Jared,
> > > > > >
> > > > > > Will do.  Do I need to update the model names again?
> > > > > >
> > > > > > SREF -> 3MEM
> > > > > > SREF -> ARW_ONLY
> > > > > >
> > > > > > John
> > > > > >
> > > > > > On Tue, Feb 7, 2017 at 12:04 PM, Jared Lee via RT <
> > met_help at ucar.edu
> > > >
> > > > > > wrote:
> > > > > >
> > > > > >>
> > > > > >> <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=79224 >
> > > > > >>
> > > > > >> Hi John,
> > > > > >>
> > > > > >> I've finished re-generating all the .stat files
(including with
> > the
> > > > > >> "model"
> > > > > >> labels I want). The zipped file is 405 MB and is located
here on
> > > > > >> yellowstone:
> > > > > >>
> > > > > >> /glade/p/ral/wsap/jaredlee/NOAA_AQ_Ensemble/MET/15day_
> data.tar.gz
> > > > > >>
> > > > > >> Could you please replace all the data in the
"noaa_aq_15day"
> > > METViewer
> > > > > >> database with that? Thanks!
> > > > > >>
> > > > > >> Jared
> > > > > >>
> > > > > >> On Sat, Feb 4, 2017 at 6:47 PM, John Halley Gotway via RT
<
> > > > > >> met_help at ucar.edu
> > > > > >> > wrote:
> > > > > >>
> > > > > >> > Jared,
> > > > > >> >
> > > > > >> > Just come by my office sometime next week and we'll
copy your
> > data
> > > > > down
> > > > > >> to
> > > > > >> > a project machine and load it into METViewer.
> > > > > >> >
> > > > > >> > My calendar should be up to date if you want to
schedule
> > something
> > > > > >> through
> > > > > >> > google calendar.
> > > > > >> >
> > > > > >> > John
> > > > > >> >
> > > > > >> > On Fri, Feb 3, 2017 at 9:22 PM Jared Lee via RT <
> > > met_help at ucar.edu>
> > > > > >> wrote:
> > > > > >> >
> > > > > >> > >
> > > > > >> > > <URL: https://rt.rap.ucar.edu/rt/
> Ticket/Display.html?id=79224
> > >
> > > > > >> > >
> > > > > >> > > Hi John,
> > > > > >> > >
> > > > > >> > > I was unaware that UPP had a script to do that. As it
turns
> > out,
> > > > > some
> > > > > >> > > members had all 48 hours in one file, and some had
them
> split
> > > > > between
> > > > > >> two
> > > > > >> > > different files, so as it turns out it probably was
simpler
> to
> > > > break
> > > > > >> them
> > > > > >> > > up first with ncks anyway. In the future I'll have
the
> person
> > > > > running
> > > > > >> WRF
> > > > > >> > > to output single-time files from the get-go, so now
my
> scripts
> > > > will
> > > > > >> work
> > > > > >> > > for that.
> > > > > >> > >
> > > > > >> > > Does my laptop 'shadowfax' count as a machine within
the RAL
> > > > > firewall
> > > > > >> if
> > > > > >> > > I'm physically at NCAR? What if I use VPN when I'm
offsite?
> > > > > Otherwise
> > > > > >> I
> > > > > >> > > could try to find some other machine to temporarily
house my
> > > data
> > > > > for
> > > > > >> > > METViewer to access.
> > > > > >> > >
> > > > > >> > > Jared
> > > > > >> > >
> > > > > >> > > On Fri, Feb 3, 2017 at 4:36 PM, John Halley Gotway
via RT <
> > > > > >> > > met_help at ucar.edu
> > > > > >> > > > wrote:
> > > > > >> > >
> > > > > >> > > > Hi Jared,
> > > > > >> > > >
> > > > > >> > > > I'm surprised you need to use ncks to split out
multiple
> > > times.
> > > > > >> Isn't
> > > > > >> > > > there a script for running UPP on a wrfout file
with
> > multiple
> > > > > >> times?  I
> > > > > >> > > > believe it's called run_unipost_frames.
> > > > > >> > > >
> > > > > >> > > > As for your question about METViewer... you'd need
to pull
> > the
> > > > > >> > > > ensemble-stat data down to a machine inside the RAL
> > firewall.
> > > > And
> > > > > >> > then I
> > > > > >> > > > could demonstrate to you how to load it into
METViewer and
> > > start
> > > > > >> > creating
> > > > > >> > > > plots.
> > > > > >> > > >
> > > > > >> > > > Unfortunately, you can't load it directly from
yellowstone
> > > into
> > > > > >> > > METViewer,
> > > > > >> > > > which lives on a machine named "mandan".
> > > > > >> > > >
> > > > > >> > > > John
> > > > > >> > > >
> > > > > >> > > > On Fri, Feb 3, 2017 at 3:09 PM, Jared Lee via RT <
> > > > > met_help at ucar.edu
> > > > > >> >
> > > > > >> > > > wrote:
> > > > > >> > > >
> > > > > >> > > > >
> > > > > >> > > > > <URL: https://rt.rap.ucar.edu/rt/
> > > Ticket/Display.html?id=79224
> > > > >
> > > > > >> > > > >
> > > > > >> > > > > Hi John,
> > > > > >> > > > >
> > > > > >> > > > > Now I finally have finished running UPP (after
having to
> > > split
> > > > > >> apart
> > > > > >> > > the
> > > > > >> > > > > wrfout files into single-time files with ncks)
and
> > > > ensemble-stat
> > > > > >> on
> > > > > >> > all
> > > > > >> > > > the
> > > > > >> > > > > model data for my three ensembles. How do I go
about
> > getting
> > > > > that
> > > > > >> > into
> > > > > >> > > > > METViewer so that I can plot spread-skill
information?
> My
> > > data
> > > > > is
> > > > > >> on
> > > > > >> > > > > yellowstone. I also looked through the stat-
analysis
> > section
> > > > of
> > > > > >> the
> > > > > >> > MET
> > > > > >> > > > > Users Guide, but I'm not clear on what exactly
I'd need
> to
> > > do
> > > > to
> > > > > >> get
> > > > > >> > > > what I
> > > > > >> > > > > need with that...
> > > > > >> > > > >
> > > > > >> > > > > Jared
> > > > > >> > > > >
> > > > > >> > > > > On Mon, Jan 23, 2017 at 5:01 PM, John Halley
Gotway via
> > RT <
> > > > > >> > > > > met_help at ucar.edu> wrote:
> > > > > >> > > > >
> > > > > >> > > > > > Jared,
> > > > > >> > > > > >
> > > > > >> > > > > > Unfortunately, no, MET is not set up to read
wrfout
> > files
> > > > > >> directly.
> > > > > >> > > As
> > > > > >> > > > > for
> > > > > >> > > > > > post-processing, you have 2 options... either
UPP or
> > > > > wrf_interp
> > > > > >> > (only
> > > > > >> > > > for
> > > > > >> > > > > > WRF-ARW core).
> > > > > >> > > > > >
> > > > > >> > > > > > The former writes GRIB output which MET handles
well.
> > The
> > > > > latter
> > > > > >> > > > writes a
> > > > > >> > > > > > flavor of NetCDF that MET is able to read...
with the
> > > caveat
> > > > > >> that
> > > > > >> > it
> > > > > >> > > > can
> > > > > >> > > > > > only read fields that are defined on non-
staggered
> > > > dimensions.
> > > > > >> > > > > Basically,
> > > > > >> > > > > > it can't read the wind fields.
> > > > > >> > > > > >
> > > > > >> > > > > > I always encourage people to use UPP since MET
> supports
> > > GRIB
> > > > > >> well.
> > > > > >> > > > > >
> > > > > >> > > > > > Being able to read wrfout directly has often
been
> > > requested.
> > > > > >> The
> > > > > >> > two
> > > > > >> > > > > > issues are the staggered dimensions and the
hybrid
> > > vertical
> > > > > >> > > coordinate.
> > > > > >> > > > > > Generally, matching to point observations is
done in
> > terms
> > > > of
> > > > > >> > > pressure
> > > > > >> > > > or
> > > > > >> > > > > > height.  So some interpolation to
pressure/height
> levels
> > > in
> > > > > >> > > inevitable
> > > > > >> > > > > > along the way.  We've resisted re-implementing
that
> > > > > >> post-processing
> > > > > >> > > > logic
> > > > > >> > > > > > directly in MET to avoid duplication of effort
with
> UPP.
> > > > And
> > > > > >> UPP
> > > > > >> > > > > supports
> > > > > >> > > > > > the derivation of many other variable types
that are
> > > pretty
> > > > > >> useful.
> > > > > >> > > > > >
> > > > > >> > > > > > Perhaps there is some middle ground where we
could
> > enable
> > > > MET
> > > > > to
> > > > > >> > call
> > > > > >> > > > UPP
> > > > > >> > > > > > behind the scenes?  But we'd need to work out
many
> > > details.
> > > > > >> > > > > >
> > > > > >> > > > > > Thanks,
> > > > > >> > > > > > John
> > > > > >> > > > > >
> > > > > >> > > > > > On Fri, Jan 20, 2017 at 5:31 PM, Jared Lee via
RT <
> > > > > >> > met_help at ucar.edu
> > > > > >> > > >
> > > > > >> > > > > > wrote:
> > > > > >> > > > > >
> > > > > >> > > > > > >
> > > > > >> > > > > > > <URL: https://rt.rap.ucar.edu/rt/
> > > > > Ticket/Display.html?id=79224
> > > > > >> >
> > > > > >> > > > > > >
> > > > > >> > > > > > > Hi John,
> > > > > >> > > > > > >
> > > > > >> > > > > > > Thanks for clarifying what ensemble-stat does
to
> > produce
> > > > the
> > > > > >> > SSVAR
> > > > > >> > > > line
> > > > > >> > > > > > > type. That helps.
> > > > > >> > > > > > >
> > > > > >> > > > > > > As for what I ultimately want, I guess that's
still
> > > > somewhat
> > > > > >> > > vaguely
> > > > > >> > > > > > > defined. I think making plots that show the
> > spread/skill
> > > > > >> > > relationship
> > > > > >> > > > > > > across the date ranges, lead times, and
variables of
> > > > > interest
> > > > > >> for
> > > > > >> > > > both
> > > > > >> > > > > > > ensembles would be a useful way to illustrate
and
> > > compare
> > > > > the
> > > > > >> > > spread
> > > > > >> > > > of
> > > > > >> > > > > > the
> > > > > >> > > > > > > two ensembles. That seems like it would be
more
> useful
> > > > than
> > > > > >> > simply
> > > > > >> > > > > > > calculating the mean spread of the ensemble
(which I
> > > could
> > > > > do
> > > > > >> by
> > > > > >> > > > > writing
> > > > > >> > > > > > a
> > > > > >> > > > > > > new NCL script to read in all the ensemble
fields
> and
> > > not
> > > > > even
> > > > > >> > > > > bothering
> > > > > >> > > > > > > with MET...).
> > > > > >> > > > > > >
> > > > > >> > > > > > > Another side question... Can MET read in
native-grid
> > > > NetCDF
> > > > > >> WRF
> > > > > >> > > > output
> > > > > >> > > > > > > files? Or do I need to run them through some
> > > pre-processor
> > > > > >> (like
> > > > > >> > > UPP)
> > > > > >> > > > > > > first? After going through the user guide I'm
not
> > clear
> > > > > >> whether
> > > > > >> > MET
> > > > > >> > > > > > accepts
> > > > > >> > > > > > > NetCDF wrfout files or not without pre-
processing.
> > Also,
> > > > if
> > > > > >> all
> > > > > >> > the
> > > > > >> > > > > > output
> > > > > >> > > > > > > times are in a single wrfout file, can MET
open the
> > one
> > > > file
> > > > > >> and
> > > > > >> > > find
> > > > > >> > > > > the
> > > > > >> > > > > > > correct time it's supposed to read, or do all
the
> > wrfout
> > > > > files
> > > > > >> > need
> > > > > >> > > > to
> > > > > >> > > > > be
> > > > > >> > > > > > > for a single time only? The NetCDF wrfout
files I've
> > > been
> > > > > >> given
> > > > > >> > for
> > > > > >> > > > the
> > > > > >> > > > > > > small ensemble have all the output times in a
single
> > > file,
> > > > > >> but I
> > > > > >> > > can
> > > > > >> > > > > > write
> > > > > >> > > > > > > a script to parse them out to single-time
files with
> > > ncks.
> > > > > I'd
> > > > > >> > just
> > > > > >> > > > > > rather
> > > > > >> > > > > > > not go through that hassle if I don't have
to!
> > > > > >> > > > > > >
> > > > > >> > > > > > > Jared
> > > > > >> > > > > > >
> > > > > >> > > > > > > On Fri, Jan 20, 2017 at 5:07 PM, John Halley
Gotway
> > via
> > > > RT <
> > > > > >> > > > > > > met_help at ucar.edu> wrote:
> > > > > >> > > > > > >
> > > > > >> > > > > > > > Hi Jared,
> > > > > >> > > > > > > >
> > > > > >> > > > > > > > I see a couple of questions in your email.
> > > > > >> > > > > > > >
> > > > > >> > > > > > > > First, yes, Ensemble-Stat is set up to
process
> both
> > > > point
> > > > > >> and
> > > > > >> > > > gridded
> > > > > >> > > > > > > > observations.  And the SSVAR (spread/skill
> variance)
> > > > line
> > > > > >> type
> > > > > >> > > can
> > > > > >> > > > be
> > > > > >> > > > > > > > produced using either of these.
> > > > > >> > > > > > > >
> > > > > >> > > > > > > > The SSVAR line type stores information
about the
> > > > > >> relationship
> > > > > >> > > > between
> > > > > >> > > > > > > > ensemble spread and skill.  Here's how it
works.
> > > > > >> Ensemble-Stat
> > > > > >> > > > > > processes
> > > > > >> > > > > > > > each observation value... either from a
point
> > location
> > > > > (i.e.
> > > > > >> > for
> > > > > >> > > > > point
> > > > > >> > > > > > > obs)
> > > > > >> > > > > > > > or from a gridded location (i.e. gridded
obs).
> For
> > > each
> > > > > >> > > > observation,
> > > > > >> > > > > > it
> > > > > >> > > > > > > > gathers the N ensemble forecast values
which
> > > correspond
> > > > to
> > > > > >> it.
> > > > > >> > > It
> > > > > >> > > > > > > computes
> > > > > >> > > > > > > > the variance of those ensemble values, and
it also
> > > > > computes
> > > > > >> the
> > > > > >> > > > error
> > > > > >> > > > > > by
> > > > > >> > > > > > > > subtracting the mean ensemble value minus
the
> > > > observation
> > > > > >> > value.
> > > > > >> > > > > > (Note,
> > > > > >> > > > > > > > you can pass in an ensemble mean field
instead of
> > > > > computing
> > > > > >> it
> > > > > >> > on
> > > > > >> > > > the
> > > > > >> > > > > > fly
> > > > > >> > > > > > > > like this).  After processing N
observations, we
> > have
> > > N
> > > > > >> pairs
> > > > > >> > of
> > > > > >> > > > > > variance
> > > > > >> > > > > > > > values and error values.  It groups those N
pairs
> > into
> > > > > >> buckets
> > > > > >> > by
> > > > > >> > > > > > binning
> > > > > >> > > > > > > > the variance using the bucket size defined
in the
> > > config
> > > > > >> file.
> > > > > >> > > > > You'll
> > > > > >> > > > > > > see
> > > > > >> > > > > > > > in the output one SSVAR line for each of
those
> > > variance
> > > > > >> > buckets.
> > > > > >> > > > > > > >
> > > > > >> > > > > > > > The METViewer database and display system
(or
> > > > > Stat-Analysis)
> > > > > >> > can
> > > > > >> > > > > > > aggregate
> > > > > >> > > > > > > > those SSVAR lines intelligently across
multiple
> > cases
> > > > and
> > > > > >> > create
> > > > > >> > > > > plots
> > > > > >> > > > > > > > which show the relationship between spread
and
> > skill.
> > > > > >> > > > > > > >
> > > > > >> > > > > > > > If that's what you need, great.
> > > > > >> > > > > > > >
> > > > > >> > > > > > > > However, if your looking for a much simpler
> > > computation
> > > > of
> > > > > >> mean
> > > > > >> > > > > spread
> > > > > >> > > > > > > over
> > > > > >> > > > > > > > some spatial region, unfortunately, the
answer is
> > no.
> > > > For
> > > > > >> > > example,
> > > > > >> > > > > > > > Ensemble-Stat doesn't write out the mean
spread
> > value
> > > > over
> > > > > >> the
> > > > > >> > > > entire
> > > > > >> > > > > > > > domain.
> > > > > >> > > > > > > >
> > > > > >> > > > > > > > Just let me know what questions you still
have.
> > > > > >> > > > > > > >
> > > > > >> > > > > > > > Thanks,
> > > > > >> > > > > > > > John Halley Gotway
> > > > > >> > > > > > > >
> > > > > >> > > > > > > >
> > > > > >> > > > > > > > On Fri, Jan 20, 2017 at 11:41 AM, Jared Lee
via
> RT <
> > > > > >> > > > > met_help at ucar.edu>
> > > > > >> > > > > > > > wrote:
> > > > > >> > > > > > > >
> > > > > >> > > > > > > > >
> > > > > >> > > > > > > > > Fri Jan 20 11:41:03 2017: Request 79224
was
> acted
> > > > upon.
> > > > > >> > > > > > > > > Transaction: Ticket created by
> jaredlee at ucar.edu
> > > > > >> > > > > > > > >        Queue: met_help
> > > > > >> > > > > > > > >      Subject: calculating ensemble spread
> > > > > >> > > > > > > > >        Owner: Nobody
> > > > > >> > > > > > > > >   Requestors: jaredlee at ucar.edu
> > > > > >> > > > > > > > >       Status: new
> > > > > >> > > > > > > > >  Ticket <URL: https://rt.rap.ucar.edu/rt/
> > > > > >> > > > > > Ticket/Display.html?id=79224
> > > > > >> > > > > > > >
> > > > > >> > > > > > > > >
> > > > > >> > > > > > > > >
> > > > > >> > > > > > > > > Hi, for my project I need to calculate
some
> > measure
> > > of
> > > > > the
> > > > > >> > > spread
> > > > > >> > > > > of
> > > > > >> > > > > > > two
> > > > > >> > > > > > > > > different ensembles (one "large" and one
"small"
> > > > > >> ensemble). I
> > > > > >> > > > have
> > > > > >> > > > > > > model
> > > > > >> > > > > > > > > output data (the large ensemble is in
grib2
> > format,
> > > > the
> > > > > >> small
> > > > > >> > > > > > ensemble
> > > > > >> > > > > > > is
> > > > > >> > > > > > > > > WRF NetCDF), and I have MADIS point
observations
> > > that
> > > > > have
> > > > > >> > > > already
> > > > > >> > > > > > been
> > > > > >> > > > > > > > > processed for use in MET.
> > > > > >> > > > > > > > >
> > > > > >> > > > > > > > > I previously used the ensemble-stat tool
on the
> > > large
> > > > > >> > ensemble
> > > > > >> > > to
> > > > > >> > > > > > dump
> > > > > >> > > > > > > > out
> > > > > >> > > > > > > > > matched pair info for all ensemble member
> > forecasts
> > > at
> > > > > >> each
> > > > > >> > > > > > observation
> > > > > >> > > > > > > > > location by printing out the ORANK line
type.
> > Given
> > > > the
> > > > > >> > inputs
> > > > > >> > > > > that I
> > > > > >> > > > > > > > > described above, will MET calculate the
ensemble
> > > > spread
> > > > > >> > > > > automatically
> > > > > >> > > > > > > if
> > > > > >> > > > > > > > I
> > > > > >> > > > > > > > > turn on the SSVAR line type? In the user
guide
> > > (v5.2)
> > > > it
> > > > > >> > > mentions
> > > > > >> > > > > > that
> > > > > >> > > > > > > an
> > > > > >> > > > > > > > > "observation field" is needed to
calculate
> things
> > > like
> > > > > >> > > > spread/skill
> > > > > >> > > > > > > > > variance. Do point obs from MADIS count
as an
> > > > > "observation
> > > > > >> > > > field"?
> > > > > >> > > > > > > > >
> > > > > >> > > > > > > > > Guidance would be much appreciated!
> > > > > >> > > > > > > > >
> > > > > >> > > > > > > > > Jared
> > > > > >> > > > > > > > >
> > > > > >> > > > > > > > > --
> > > > > >> > > > > > > > > ===============================
> > > > > >> > > > > > > > > Jared A. Lee, Ph.D.
> > > > > >> > > > > > > > > Project Scientist I
> > > > > >> > > > > > > > > Research Applications Laboratory
> > > > > >> > > > > > > > > National Center for Atmospheric Research
> > > > > >> > > > > > > > > Boulder, Colorado, USA
> > > > > >> > > > > > > > >
> > > > > >> > > > > > > > > Email: jaredlee at ucar.edu (w)
> > > > > >> > > > > > > > > Phone: 303.497.8485 (w)
> > > > > >> > > > > > > > > Web:
https://staff.ucar.edu/users/jaredlee
> > > > > >> > > > > > > > > ===============================
> > > > > >> > > > > > > > >
> > > > > >> > > > > > > > >
> > > > > >> > > > > > > >
> > > > > >> > > > > > > >
> > > > > >> > > > > > >
> > > > > >> > > > > > >
> > > > > >> > > > > > > --
> > > > > >> > > > > > > ===============================
> > > > > >> > > > > > > Jared A. Lee, Ph.D.
> > > > > >> > > > > > > Project Scientist I
> > > > > >> > > > > > > Research Applications Laboratory
> > > > > >> > > > > > > National Center for Atmospheric Research
> > > > > >> > > > > > > Boulder, Colorado, USA
> > > > > >> > > > > > >
> > > > > >> > > > > > > Email: jaredlee at ucar.edu (w)
> > > > > >> > > > > > > Phone: 303.497.8485 (w)
> > > > > >> > > > > > > Web: https://staff.ucar.edu/users/jaredlee
> > > > > >> > > > > > > ===============================
> > > > > >> > > > > > >
> > > > > >> > > > > > >
> > > > > >> > > > > >
> > > > > >> > > > > >
> > > > > >> > > > >
> > > > > >> > > > >
> > > > > >> > > > > --
> > > > > >> > > > > ===============================
> > > > > >> > > > > Jared A. Lee, Ph.D.
> > > > > >> > > > > Project Scientist I
> > > > > >> > > > > Research Applications Laboratory
> > > > > >> > > > > National Center for Atmospheric Research
> > > > > >> > > > > Boulder, Colorado, USA
> > > > > >> > > > >
> > > > > >> > > > > Email: jaredlee at ucar.edu (w)
> > > > > >> > > > > Phone: 303.497.8485 (w)
> > > > > >> > > > > Web: https://staff.ucar.edu/users/jaredlee
> > > > > >> > > > > ===============================
> > > > > >> > > > >
> > > > > >> > > > >
> > > > > >> > > >
> > > > > >> > > >
> > > > > >> > >
> > > > > >> > >
> > > > > >> > > --
> > > > > >> > > ===============================
> > > > > >> > > Jared A. Lee, Ph.D.
> > > > > >> > > Project Scientist I
> > > > > >> > > Research Applications Laboratory
> > > > > >> > > National Center for Atmospheric Research
> > > > > >> > > Boulder, Colorado, USA
> > > > > >> > >
> > > > > >> > > Email: jaredlee at ucar.edu (w)
> > > > > >> > > Phone: 303.497.8485 (w)
> > > > > >> > > Web: https://staff.ucar.edu/users/jaredlee
> > > > > >> > > ===============================
> > > > > >> > >
> > > > > >> > >
> > > > > >> >
> > > > > >> >
> > > > > >>
> > > > > >>
> > > > > >> --
> > > > > >> ===============================
> > > > > >> Jared A. Lee, Ph.D.
> > > > > >> Project Scientist I
> > > > > >> Research Applications Laboratory
> > > > > >> National Center for Atmospheric Research
> > > > > >> Boulder, Colorado, USA
> > > > > >>
> > > > > >> Email: jaredlee at ucar.edu (w)
> > > > > >> Phone: 303.497.8485 (w)
> > > > > >> Web: https://staff.ucar.edu/users/jaredlee
> > > > > >> ===============================
> > > > > >>
> > > > > >>
> > > > > >
> > > > >
> > > > >
> > > >
> > > >
> > > > --
> > > > ===============================
> > > > Jared A. Lee, Ph.D.
> > > > Project Scientist I
> > > > Research Applications Laboratory
> > > > National Center for Atmospheric Research
> > > > Boulder, Colorado, USA
> > > >
> > > > Email: jaredlee at ucar.edu (w)
> > > > Phone: 303.497.8485 (w)
> > > > Web: https://staff.ucar.edu/users/jaredlee
> > > > ===============================
> > > >
> > > >
> > >
> > >
> >
> >
> > --
> > ===============================
> > Jared A. Lee, Ph.D.
> > Project Scientist I
> > Research Applications Laboratory
> > National Center for Atmospheric Research
> > Boulder, Colorado, USA
> >
> > Email: jaredlee at ucar.edu (w)
> > Phone: 303.497.8485 (w)
> > Web: https://staff.ucar.edu/users/jaredlee
> > ===============================
> >
> >
>
>


--
===============================
Jared A. Lee, Ph.D.
Project Scientist I
Research Applications Laboratory
National Center for Atmospheric Research
Boulder, Colorado, USA

Email: jaredlee at ucar.edu (w)
Phone: 303.497.8485 (w)
Web: https://staff.ucar.edu/users/jaredlee
===============================

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


More information about the Met_help mailing list