[Met_help] [rt.rap.ucar.edu #78565] History for about the choice of the to_grid

[John Halley Gotway via RT]

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

Dear Mr. or Mrs.,

Thanks a lot in advance for your help on my problem.
In MET grid verification,  in my understanding,
both the fcst and obs would be regrided to a common grid , by to_grid
option,
for the verification.
What is the theoretically "best" choice for the common grid.
Should the original obs grids be used as the "common" grid and
the fcst should be regridded on it?
Your clarifications are appreciated.
Best Regards,
Ting


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

Subject: about the choice of the to_grid
From: John Halley Gotway
Time: Wed Nov 02 09:46:40 2016

Ting,

I see that you're wondering which grid should be used for
verification, the
forecast or the observation grid.  Unfortunately, I do not have a
simple
answer for you.  There are several factors to consider.

Look carefully at the resolution of the grids in question.  If one is
low
resolution, perhaps 40km, and another is high resolution, perhaps 4km,
I
would suggest using the lower resolution grid.  Otherwise, the
interpolation step would be "creating" most of the grid points that
you're
evaluating.

When the grid spacing is similar, some argue that you should use the
observation grid so as not to change your truth field with
interpolation.
However, others prefer using the forecast grid because they worry that
interpolation will degrade their forecast.  And others prefer using a
3rd
grid for verification to which the forecast and observation are both
interpolated.

The good news is that it's very easy to specify the regridding logic
in the
configuration file.  You could run Grid-Stat two or three times,
evaluating
on the forecast grid, observation grid, and a 3rd verification grid,
and
check to see if/how the scores change.

Sorry I don't have a more clear answer for you but there are arguments
for
doing this a few different ways.

Hope that helps.

Thanks,
John



On Wed, Nov 2, 2016 at 8:24 AM, Ting Lei - NOAA Affiliate via RT <
met_help at ucar.edu> wrote:

>
> Wed Nov 02 08:24:28 2016: Request 78565 was acted upon.
> Transaction: Ticket created by ting.lei at noaa.gov
>        Queue: met_help
>      Subject: about the choice of the to_grid
>        Owner: Nobody
>   Requestors: ting.lei at noaa.gov
>       Status: new
>  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=78565 >
>
>
> Dear Mr. or Mrs.,
>
> Thanks a lot in advance for your help on my problem.
> In MET grid verification,  in my understanding,
> both the fcst and obs would be regrided to a common grid , by
to_grid
> option,
> for the verification.
> What is the theoretically "best" choice for the common grid.
> Should the original obs grids be used as the "common" grid and
> the fcst should be regridded on it?
> Your clarifications are appreciated.
> Best Regards,
> Ting
>
>

------------------------------------------------
Subject: about the choice of the to_grid
From: Ting Lei - NOAA Affiliate
Time: Wed Nov 02 10:15:38 2016

Hi, John,
You has given very clear/complete description for the logic choosing
the
common grids.
It is great to know how you (MET developer) is thinking on it.
I will keep your points on mind when I do MET verification and in
other
similar situations, especially, when the forecasts to be compared are
on
different grids.
Your help is appreciated.
Best Wishes,
TIng

On Wed, Nov 2, 2016 at 11:46 AM, John Halley Gotway via RT <
met_help at ucar.edu> wrote:

> Ting,
>
> I see that you're wondering which grid should be used for
verification, the
> forecast or the observation grid.  Unfortunately, I do not have a
simple
> answer for you.  There are several factors to consider.
>
> Look carefully at the resolution of the grids in question.  If one
is low
> resolution, perhaps 40km, and another is high resolution, perhaps
4km, I
> would suggest using the lower resolution grid.  Otherwise, the
> interpolation step would be "creating" most of the grid points that
you're
> evaluating.
>
> When the grid spacing is similar, some argue that you should use the
> observation grid so as not to change your truth field with
interpolation.
> However, others prefer using the forecast grid because they worry
that
> interpolation will degrade their forecast.  And others prefer using
a 3rd
> grid for verification to which the forecast and observation are both
> interpolated.
>
> The good news is that it's very easy to specify the regridding logic
in the
> configuration file.  You could run Grid-Stat two or three times,
evaluating
> on the forecast grid, observation grid, and a 3rd verification grid,
and
> check to see if/how the scores change.
>
> Sorry I don't have a more clear answer for you but there are
arguments for
> doing this a few different ways.
>
> Hope that helps.
>
> Thanks,
> John
>
>
>
> On Wed, Nov 2, 2016 at 8:24 AM, Ting Lei - NOAA Affiliate via RT <
> met_help at ucar.edu> wrote:
>
> >
> > Wed Nov 02 08:24:28 2016: Request 78565 was acted upon.
> > Transaction: Ticket created by ting.lei at noaa.gov
> >        Queue: met_help
> >      Subject: about the choice of the to_grid
> >        Owner: Nobody
> >   Requestors: ting.lei at noaa.gov
> >       Status: new
> >  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=78565 >
> >
> >
> > Dear Mr. or Mrs.,
> >
> > Thanks a lot in advance for your help on my problem.
> > In MET grid verification,  in my understanding,
> > both the fcst and obs would be regrided to a common grid , by
to_grid
> > option,
> > for the verification.
> > What is the theoretically "best" choice for the common grid.
> > Should the original obs grids be used as the "common" grid and
> > the fcst should be regridded on it?
> > Your clarifications are appreciated.
> > Best Regards,
> > Ting
> >
> >
>
>

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