[Met_help] [rt.rap.ucar.edu #76465] History for Using MET MODE to match ensemble data objects

Tue May 24 09:31:58 MDT 2016

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

Greetings,

I am exploring MET MODE for displaying QPF objects of ensemble model data
operationally. I know that the common format of MET MODE is to match
forecast objects to observations, but I was wondering if it could also
match an ensemble of the most prominent forecast objects simultaneously
(i.e. no observation, but with multiple forecast inputs)?

This can likely be achieved by iteratively matching each unique forecast
couplet in the ensemble and then grouping the statistics, but I was
wondering if there was an easier way at the command line.

Thanks for your help!

Mike

-- 

Michael J. Erickson, Ph.D
Cooperative Institute for Research in Environmental Sciences (CIRES)
CIRES Contractor for the Weather Prediction Center (WPC)

Phone:  301-683-1546

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

Subject: Using MET MODE to match ensemble data objects
From: John Halley Gotway
Time: Mon May 23 15:38:23 2016

Mike,

Great question!  You're right that when you run MODE, it expects
exactly 2
input files: one forecast file and one observation file.  As you
mention,
you can easily run MODE to compare two different forecast fields.  The
best
way of running MODE on an ensemble of forecasts and making sense of
the
results is still an open question, but here are a couple of
thoughts...

(1) If you do have an observation field to which you'd like to
compare, you
could run MODE to compare each of the individual members to that
observation.  Then group the resulting objects based on the
observation
objects... meaning if objects from multiple ensemble members match the
*same* observation object, then they go together.

(2) If you do *not* have an observation, you could apply similar logic
but
compare each ensemble member to the ensemble mean field.  Ensemble
member
objects which match the same ensemble mean object "go together" as
well.
So you could first run the ensemble_stat tool to create an ensemble
mean
field and pass that to your MODE runs.

Ultimately, your decisions will likely be based on what data is
available
and what specific questions you're trying to answer.  If you're
ultimately
looking for a nice graphical summary of the MODE runs, unfortunately,
the
PostScript output MODE generates won't nicely summarize all of the
members.  However, I've attached two somewhat similar examples of how
you
might plot ensemble MODE data.  These are both animations, each frame
of
which is generated using NCL...

stamp_20160125* shows a plot of contoured raw probabilistic data (in
colors) overlaid with black, lightly gray shaded MODE objects.  The
MODE
objects were generated for a time lagged ensemble and the solid colors
are
the probabilities derived (by some other algorithm) for this case.

stamp_16km* shows solid-filled observation objects (red means it
matched an
ensemble mean object, blue means unmatched).  The ensemble mean
objects
have thick black lines and the slightly darker green fill.  The
individual
ensemble member objects have very thin gray lines and very light green
fill.

So these are two ways we have plotted ensembles of MODE objects in the
past, and we're happy to share our plotting scripts if you'd find that
useful.

Thanks,
John Halley Gotway

On Mon, May 23, 2016 at 2:51 PM, Michael Erickson - NOAA Affiliate via
RT <
met_help at ucar.edu> wrote:

>
> Mon May 23 14:51:24 2016: Request 76465 was acted upon.
> Transaction: Ticket created by michael.j.erickson at noaa.gov
>        Queue: met_help
>      Subject: Using MET MODE to match ensemble data objects
>        Owner: Nobody
>   Requestors: michael.j.erickson at noaa.gov
>       Status: new
>  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=76465 >
>
>
> Greetings,
>
> I am exploring MET MODE for displaying QPF objects of ensemble model
data
> operationally. I know that the common format of MET MODE is to match
> forecast objects to observations, but I was wondering if it could
also
> match an ensemble of the most prominent forecast objects
simultaneously
> (i.e. no observation, but with multiple forecast inputs)?
>
> This can likely be achieved by iteratively matching each unique
forecast
> couplet in the ensemble and then grouping the statistics, but I was
> wondering if there was an easier way at the command line.
>
> Thanks for your help!
>
> Mike
>
> --
>
> Michael J. Erickson, Ph.D
> Cooperative Institute for Research in Environmental Sciences (CIRES)
> CIRES Contractor for the Weather Prediction Center (WPC)
>
> Phone:  301-683-1546
>
>

------------------------------------------------
Subject: Using MET MODE to match ensemble data objects
From: Michael Erickson - NOAA Affiliate
Time: Tue May 24 06:32:17 2016

Hi John,

Thank you for your quick and very helpful response. We are hoping to
do
something similar to your option 2 by comparing several high
resolution
models' QPF objects in real-time. Comparing each member to the
ensemble
mean sounds like a good idea.

I think that's all the questions I have, but I'll let you know if I
can
think of anything else. Thanks again!

Mike

On Mon, May 23, 2016 at 9:38 PM, John Halley Gotway via RT <
met_help at ucar.edu> wrote:

> Mike,
>
> Great question!  You're right that when you run MODE, it expects
exactly 2
> input files: one forecast file and one observation file.  As you
mention,
> you can easily run MODE to compare two different forecast fields.
The best
> way of running MODE on an ensemble of forecasts and making sense of
the
> results is still an open question, but here are a couple of
thoughts...
>
> (1) If you do have an observation field to which you'd like to
compare, you
> could run MODE to compare each of the individual members to that
> observation.  Then group the resulting objects based on the
observation
> objects... meaning if objects from multiple ensemble members match
the
> *same* observation object, then they go together.
>
> (2) If you do *not* have an observation, you could apply similar
logic but
> compare each ensemble member to the ensemble mean field.  Ensemble
member
> objects which match the same ensemble mean object "go together" as
well.
> So you could first run the ensemble_stat tool to create an ensemble
mean
> field and pass that to your MODE runs.
>
> Ultimately, your decisions will likely be based on what data is
available
> and what specific questions you're trying to answer.  If you're
ultimately
> looking for a nice graphical summary of the MODE runs,
unfortunately, the
> PostScript output MODE generates won't nicely summarize all of the
> members.  However, I've attached two somewhat similar examples of
how you
> might plot ensemble MODE data.  These are both animations, each
frame of
> which is generated using NCL...
>
> stamp_20160125* shows a plot of contoured raw probabilistic data (in
> colors) overlaid with black, lightly gray shaded MODE objects.  The
MODE
> objects were generated for a time lagged ensemble and the solid
colors are
> the probabilities derived (by some other algorithm) for this case.
>
> stamp_16km* shows solid-filled observation objects (red means it
matched an
> ensemble mean object, blue means unmatched).  The ensemble mean
objects
> have thick black lines and the slightly darker green fill.  The
individual
> ensemble member objects have very thin gray lines and very light
green
> fill.
>
> So these are two ways we have plotted ensembles of MODE objects in
the
> past, and we're happy to share our plotting scripts if you'd find
that
> useful.
>
> Thanks,
> John Halley Gotway
>
>
> On Mon, May 23, 2016 at 2:51 PM, Michael Erickson - NOAA Affiliate
via RT <
> met_help at ucar.edu> wrote:
>
> >
> > Mon May 23 14:51:24 2016: Request 76465 was acted upon.
> > Transaction: Ticket created by michael.j.erickson at noaa.gov
> >        Queue: met_help
> >      Subject: Using MET MODE to match ensemble data objects
> >        Owner: Nobody
> >   Requestors: michael.j.erickson at noaa.gov
> >       Status: new
> >  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=76465 >
> >
> >
> > Greetings,
> >
> > I am exploring MET MODE for displaying QPF objects of ensemble
model data
> > operationally. I know that the common format of MET MODE is to
match
> > forecast objects to observations, but I was wondering if it could
also
> > match an ensemble of the most prominent forecast objects
simultaneously
> > (i.e. no observation, but with multiple forecast inputs)?
> >
> > This can likely be achieved by iteratively matching each unique
forecast
> > couplet in the ensemble and then grouping the statistics, but I
was
> > wondering if there was an easier way at the command line.
> >
> > Thanks for your help!
> >
> > Mike
> >
> > --
> >
> > Michael J. Erickson, Ph.D
> > Cooperative Institute for Research in Environmental Sciences
(CIRES)
> > CIRES Contractor for the Weather Prediction Center (WPC)
> >
> > Phone:  301-683-1546
> >
> >
>
>

--

Michael J. Erickson, Ph.D
Cooperative Institute for Research in Environmental Sciences (CIRES)
CIRES Contractor for the Weather Prediction Center (WPC)

Phone:  301-683-1546

------------------------------------------------
Subject: Using MET MODE to match ensemble data objects
From: John Halley Gotway
Time: Tue May 24 09:25:24 2016

Mike,

OK, sounds good.  Just let us know if any more issues or questions
arise.

I'll go ahead and resolve this ticket.

Thanks,
John

On Tue, May 24, 2016 at 6:32 AM, Michael Erickson - NOAA Affiliate via
RT <
met_help at ucar.edu> wrote:

>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=76465 >
>
> Hi John,
>
> Thank you for your quick and very helpful response. We are hoping to
do
> something similar to your option 2 by comparing several high
resolution
> models' QPF objects in real-time. Comparing each member to the
ensemble
> mean sounds like a good idea.
>
> I think that's all the questions I have, but I'll let you know if I
can
> think of anything else. Thanks again!
>
> Mike
>
>
>
>
>
> On Mon, May 23, 2016 at 9:38 PM, John Halley Gotway via RT <
> met_help at ucar.edu> wrote:
>
> > Mike,
> >
> > Great question!  You're right that when you run MODE, it expects
exactly
> 2
> > input files: one forecast file and one observation file.  As you
mention,
> > you can easily run MODE to compare two different forecast fields.
The
> best
> > way of running MODE on an ensemble of forecasts and making sense
of the
> > results is still an open question, but here are a couple of
thoughts...
> >
> > (1) If you do have an observation field to which you'd like to
compare,
> you
> > could run MODE to compare each of the individual members to that
> > observation.  Then group the resulting objects based on the
observation
> > objects... meaning if objects from multiple ensemble members match
the
> > *same* observation object, then they go together.
> >
> > (2) If you do *not* have an observation, you could apply similar
logic
> but
> > compare each ensemble member to the ensemble mean field.  Ensemble
member
> > objects which match the same ensemble mean object "go together" as
well.
> > So you could first run the ensemble_stat tool to create an
ensemble mean
> > field and pass that to your MODE runs.
> >
> > Ultimately, your decisions will likely be based on what data is
available
> > and what specific questions you're trying to answer.  If you're
> ultimately
> > looking for a nice graphical summary of the MODE runs,
unfortunately, the
> > PostScript output MODE generates won't nicely summarize all of the
> > members.  However, I've attached two somewhat similar examples of
how you
> > might plot ensemble MODE data.  These are both animations, each
frame of
> > which is generated using NCL...
> >
> > stamp_20160125* shows a plot of contoured raw probabilistic data
(in
> > colors) overlaid with black, lightly gray shaded MODE objects.
The MODE
> > objects were generated for a time lagged ensemble and the solid
colors
> are
> > the probabilities derived (by some other algorithm) for this case.
> >
> > stamp_16km* shows solid-filled observation objects (red means it
matched
> an
> > ensemble mean object, blue means unmatched).  The ensemble mean
objects
> > have thick black lines and the slightly darker green fill.  The
> individual
> > ensemble member objects have very thin gray lines and very light
green
> > fill.
> >
> > So these are two ways we have plotted ensembles of MODE objects in
the
> > past, and we're happy to share our plotting scripts if you'd find
that
> > useful.
> >
> > Thanks,
> > John Halley Gotway
> >
> >
> > On Mon, May 23, 2016 at 2:51 PM, Michael Erickson - NOAA Affiliate
via
> RT <
> > met_help at ucar.edu> wrote:
> >
> > >
> > > Mon May 23 14:51:24 2016: Request 76465 was acted upon.
> > > Transaction: Ticket created by michael.j.erickson at noaa.gov
> > >        Queue: met_help
> > >      Subject: Using MET MODE to match ensemble data objects
> > >        Owner: Nobody
> > >   Requestors: michael.j.erickson at noaa.gov
> > >       Status: new
> > >  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=76465
> >
> > >
> > >
> > > Greetings,
> > >
> > > I am exploring MET MODE for displaying QPF objects of ensemble
model
> data
> > > operationally. I know that the common format of MET MODE is to
match
> > > forecast objects to observations, but I was wondering if it
could also
> > > match an ensemble of the most prominent forecast objects
simultaneously
> > > (i.e. no observation, but with multiple forecast inputs)?
> > >
> > > This can likely be achieved by iteratively matching each unique
> forecast
> > > couplet in the ensemble and then grouping the statistics, but I
was
> > > wondering if there was an easier way at the command line.
> > >
> > > Thanks for your help!
> > >
> > > Mike
> > >
> > > --
> > >
> > > Michael J. Erickson, Ph.D
> > > Cooperative Institute for Research in Environmental Sciences
(CIRES)
> > > CIRES Contractor for the Weather Prediction Center (WPC)
> > >
> > > Phone:  301-683-1546
> > >
> > >
> >
> >
>
>
> --
>
> Michael J. Erickson, Ph.D
> Cooperative Institute for Research in Environmental Sciences (CIRES)
> CIRES Contractor for the Weather Prediction Center (WPC)
>
> Phone:  301-683-1546
>
>

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