[Met_help] [rt.rap.ucar.edu #99788] History for PBL height verification

John Halley Gotway via RT met_help at ucar.edu
Mon Jul 12 11:35:21 MDT 2021


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

Hello,

The PBL height verification is reported every 12 hours.  I'm looking to
plot the obs retrieved PBL height every 12 hours overlaid with hourly
output of PBL height from the model.  This would require passing in the
stats of PBL height for sites despite there being no obs reported outside
the 00z/12z time period for most instances.  Is there an option in
statanalysis and pointstat that allows me to produce PBL stats for model
output every hour so that I can overlay hour by hour stats of PBL height
overlaid with obs every 12 hours?  Here's an example of what it currently
looks like

Edward Strobach
EMC/NCEP/NWS/ IMSG
Coding Logic (if/then/else)


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

Subject: PBL height verification
From: John Halley Gotway
Time: Tue May 04 09:40:45 2021

Hi Ed,

I see that you have hourly forecast data but observations only every
12
hours. And you'd like to use METviewer to create a plot of hourly
forecasts
and 12-hourly observations.

Let's work backwards...
In order to plot the hourly forecast values with METviewer, we'll need
them
to live in the database. And presumably, we're talking about MPR
output
lines. So that means we'll need to run Point-Stat hourly to generate
that
hourly MPR output. But we need to pair the hourly forecast values to
*some*
actual PBL observations.

There are two ways that matching observation time window can be
defined in
Point-Stat:
- Either in the config file using: obs_window = { beg = -5400; end =
5400; }
  That defines a time window in second relative to the forecast valid
time
for which point obs should be used.
- Or on the command line using -obs_valid_beg and -obs_valid_end
command
line options. Those override the config file time window.

Honestly I think the second option is probably easier. Run Point-Stat
using
-obs_valid_beg and -obs_valid_end to define either the prior 12-hour
observation time, or perhaps the nearest 12-hour observation time.
Whichever it is, make sure you pass that data in the "obs_file".

I'm a little concerned here though. You're really only generating
these
pairs for plotting purposes. Any derived stats won't mean very much
because
of the large time offset. You might consider doing this run as a
second
call to Point-Stat, rather than mixing it with your statistics
generation
runs. And you could use the "desc" config option to indicate that the
resulting data is only for plotting... not for statistics generation.

Once you have the hourly MPR lines in METviewer, you should be able to
plot
that data hourly. But limiting the obs to only plot every 12 hours
might be
tricky. Perhaps you could plot the hourly forecast on the Y1 axis and
the
12-hourly obs on the Y2 axis? I'm not sure if that'd work. I've never
actually done that.

Hope that helps.

Thanks,
John

On Tue, May 4, 2021 at 8:16 AM George McCabe via RT
<met_help at ucar.edu>
wrote:

>
> Tue May 04 08:15:44 2021: Request 99788 was acted upon.
> Transaction: Given to johnhg (John Halley Gotway) by mccabe
>        Queue: met_help
>      Subject: PBL height verification
>        Owner: johnhg
>   Requestors: edward.strobach at noaa.gov
>       Status: new
>  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=99788 >
>
>
> This transaction appears to have no content
>

------------------------------------------------
Subject: PBL height verification
From: Edward Strobach - NOAA Affiliate
Time: Thu May 27 15:22:57 2021

Hello John,

I wanted to finish some things first before getting back to you on
this.
As you know I use metplus to generate stats. Lately I've been trying
to
reproduce the plotting verification as done using metviewer.  I
developed a
text file scanner that groups and orders the data in the same way that
is
done in metviewer which I have since confirmed.  I'm attaching an
example
of PBL verification using scripts that I developed to join model
output
with the verification stats.  Here's what I was trying to do which I
couldn't do in Metviewer.  I'm also attaching a second plot that shows
two
models - one going out to 48 hours and another that goes out to 72
hours.
Event equalization does not allow me to plot the forecast out to 72
hours
if the other model is only 48 hours.  My version of event equalization
allows me to do just that.  In addition, relying on Metviewer can take
7
hours of processing time.  I've cut that down to 20 to 30 minutes
since I
can now multi-process.



Edward Strobach
EMC/NCEP/NWS/ IMSG
Coding Logic (if/then/else)


On Tue, May 4, 2021 at 11:40 AM John Halley Gotway via RT
<met_help at ucar.edu>
wrote:

> Hi Ed,
>
> I see that you have hourly forecast data but observations only every
12
> hours. And you'd like to use METviewer to create a plot of hourly
forecasts
> and 12-hourly observations.
>
> Let's work backwards...
> In order to plot the hourly forecast values with METviewer, we'll
need them
> to live in the database. And presumably, we're talking about MPR
output
> lines. So that means we'll need to run Point-Stat hourly to generate
that
> hourly MPR output. But we need to pair the hourly forecast values to
*some*
> actual PBL observations.
>
> There are two ways that matching observation time window can be
defined in
> Point-Stat:
> - Either in the config file using: obs_window = { beg = -5400; end =
5400;
> }
>   That defines a time window in second relative to the forecast
valid time
> for which point obs should be used.
> - Or on the command line using -obs_valid_beg and -obs_valid_end
command
> line options. Those override the config file time window.
>
> Honestly I think the second option is probably easier. Run Point-
Stat using
> -obs_valid_beg and -obs_valid_end to define either the prior 12-hour
> observation time, or perhaps the nearest 12-hour observation time.
> Whichever it is, make sure you pass that data in the "obs_file".
>
> I'm a little concerned here though. You're really only generating
these
> pairs for plotting purposes. Any derived stats won't mean very much
because
> of the large time offset. You might consider doing this run as a
second
> call to Point-Stat, rather than mixing it with your statistics
generation
> runs. And you could use the "desc" config option to indicate that
the
> resulting data is only for plotting... not for statistics
generation.
>
> Once you have the hourly MPR lines in METviewer, you should be able
to plot
> that data hourly. But limiting the obs to only plot every 12 hours
might be
> tricky. Perhaps you could plot the hourly forecast on the Y1 axis
and the
> 12-hourly obs on the Y2 axis? I'm not sure if that'd work. I've
never
> actually done that.
>
> Hope that helps.
>
> Thanks,
> John
>
> On Tue, May 4, 2021 at 8:16 AM George McCabe via RT
<met_help at ucar.edu>
> wrote:
>
> >
> > Tue May 04 08:15:44 2021: Request 99788 was acted upon.
> > Transaction: Given to johnhg (John Halley Gotway) by mccabe
> >        Queue: met_help
> >      Subject: PBL height verification
> >        Owner: johnhg
> >   Requestors: edward.strobach at noaa.gov
> >       Status: new
> >  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=99788 >
> >
> >
> > This transaction appears to have no content
> >
>
>

------------------------------------------------
Subject: PBL height verification
From: John Halley Gotway
Time: Thu May 27 16:21:17 2021

Ed,

Well 7 hours sounds like way too long to wait! I'm copying Tatiana on
this
email to pass along this info to her.

So you're right that METviewer's event equalization is either on or
off.
And you technically can control which items are included in the event
equalization logic. But if the series are defined by model name, then
that's always included in the EE logic.

If I'm interpreting this correctly, I think you're saying the
following...
(1) Event equalization is applied separately for each independent
variable
value. In your example, that's forecast hour.
(2) For each value, only use cases that are present for each model
being
plotted.

BUT you'd like the ability to change the logic for (2) to be.
(2) For each value, only use cases that are present for each model
that has *at
least one value* present for that time.

So, since PROD and PROD_BC don't have values past 48 hours, they would
NOT
be used in the event equalization logic past that time.

Does that describe the option that you're looking for?

Thanks,
John


On Thu, May 27, 2021 at 3:23 PM Edward Strobach - NOAA Affiliate via
RT <
met_help at ucar.edu> wrote:

>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=99788 >
>
> Hello John,
>
> I wanted to finish some things first before getting back to you on
this.
> As you know I use metplus to generate stats. Lately I've been trying
to
> reproduce the plotting verification as done using metviewer.  I
developed a
> text file scanner that groups and orders the data in the same way
that is
> done in metviewer which I have since confirmed.  I'm attaching an
example
> of PBL verification using scripts that I developed to join model
output
> with the verification stats.  Here's what I was trying to do which I
> couldn't do in Metviewer.  I'm also attaching a second plot that
shows two
> models - one going out to 48 hours and another that goes out to 72
hours.
> Event equalization does not allow me to plot the forecast out to 72
hours
> if the other model is only 48 hours.  My version of event
equalization
> allows me to do just that.  In addition, relying on Metviewer can
take 7
> hours of processing time.  I've cut that down to 20 to 30 minutes
since I
> can now multi-process.
>
>
>
> Edward Strobach
> EMC/NCEP/NWS/ IMSG
> Coding Logic (if/then/else)
>
>
> On Tue, May 4, 2021 at 11:40 AM John Halley Gotway via RT <
> met_help at ucar.edu>
> wrote:
>
> > Hi Ed,
> >
> > I see that you have hourly forecast data but observations only
every 12
> > hours. And you'd like to use METviewer to create a plot of hourly
> forecasts
> > and 12-hourly observations.
> >
> > Let's work backwards...
> > In order to plot the hourly forecast values with METviewer, we'll
need
> them
> > to live in the database. And presumably, we're talking about MPR
output
> > lines. So that means we'll need to run Point-Stat hourly to
generate that
> > hourly MPR output. But we need to pair the hourly forecast values
to
> *some*
> > actual PBL observations.
> >
> > There are two ways that matching observation time window can be
defined
> in
> > Point-Stat:
> > - Either in the config file using: obs_window = { beg = -5400; end
=
> 5400;
> > }
> >   That defines a time window in second relative to the forecast
valid
> time
> > for which point obs should be used.
> > - Or on the command line using -obs_valid_beg and -obs_valid_end
command
> > line options. Those override the config file time window.
> >
> > Honestly I think the second option is probably easier. Run Point-
Stat
> using
> > -obs_valid_beg and -obs_valid_end to define either the prior 12-
hour
> > observation time, or perhaps the nearest 12-hour observation time.
> > Whichever it is, make sure you pass that data in the "obs_file".
> >
> > I'm a little concerned here though. You're really only generating
these
> > pairs for plotting purposes. Any derived stats won't mean very
much
> because
> > of the large time offset. You might consider doing this run as a
second
> > call to Point-Stat, rather than mixing it with your statistics
generation
> > runs. And you could use the "desc" config option to indicate that
the
> > resulting data is only for plotting... not for statistics
generation.
> >
> > Once you have the hourly MPR lines in METviewer, you should be
able to
> plot
> > that data hourly. But limiting the obs to only plot every 12 hours
might
> be
> > tricky. Perhaps you could plot the hourly forecast on the Y1 axis
and the
> > 12-hourly obs on the Y2 axis? I'm not sure if that'd work. I've
never
> > actually done that.
> >
> > Hope that helps.
> >
> > Thanks,
> > John
> >
> > On Tue, May 4, 2021 at 8:16 AM George McCabe via RT
<met_help at ucar.edu>
> > wrote:
> >
> > >
> > > Tue May 04 08:15:44 2021: Request 99788 was acted upon.
> > > Transaction: Given to johnhg (John Halley Gotway) by mccabe
> > >        Queue: met_help
> > >      Subject: PBL height verification
> > >        Owner: johnhg
> > >   Requestors: edward.strobach at noaa.gov
> > >       Status: new
> > >  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=99788
> >
> > >
> > >
> > > This transaction appears to have no content
> > >
> >
> >
>
>

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


More information about the Met_help mailing list