[Met_help] [rt.rap.ucar.edu #56648] History for Defining Interpolation Method in the GridStat's Neighbourhood Approach
Paul Oldenburg via RT
met_help at ucar.edu
Mon Jun 11 09:02:09 MDT 2012
----------------------------------------------------------------
Initial Request
----------------------------------------------------------------
Hi,
I came across this statement in MET_Users_Guide_v3.0.3 (page 5-3), "... The
user must specify several parameters in the grid_stat configuration file to
utilize the neighbourhood approach, such as the *interpolation method*,
size of smoothing window,..."
I would like to confirm that specifying the interpolation method is only
required if one wants to upscale the information and use traditional
verification metrics on the coarser grid, isn't it? So, if one were to use
the FSS and the BSS to investigate the results of the neighbourhood
approach, one will not be required to define the "interpolation method" and
other parameters related to the interpolation/smoothing approach. Am I
right to say this?
Thanks,
Raizan
----------------------------------------------------------------
Complete Ticket History
----------------------------------------------------------------
Subject: Re: [rt.rap.ucar.edu #56648] Defining Interpolation Method in the GridStat's Neighbourhood Approach
From: Paul Oldenburg
Time: Tue May 29 09:29:19 2012
Raizan,
You are basically correct, but you still do need to specify an
interpolation method in the grid_stat config file. The
following interpolation settings basically mean "no interpolation":
interp_method[] = [ "UW_MEAN" ];
interp_width[] = [ 1 ];
You should make sure that the nbr_width and nbr_thresh settings have
values that are appropriate to your verification
experiment. The the nbr_width setting specifies the size of the
neighborhood(s) over which FSS and other statistics are
calculated. Finally, you should set the output_flag switches at
indexes 13-15 to 1 or 2 to trigger the calculation of
neighborhood statistics.
I hope this answers your question.
Thanks,
Paul
On 05/29/2012 03:13 AM, Zan Rahmat via RT wrote:
>
> Tue May 29 03:13:33 2012: Request 56648 was acted upon.
> Transaction: Ticket created by raizan.rahmat at gmail.com
> Queue: met_help
> Subject: Defining Interpolation Method in the GridStat's
Neighbourhood Approach
> Owner: Nobody
> Requestors: raizan.rahmat at gmail.com
> Status: new
> Ticket<URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=56648>
>
>
> Hi,
>
> I came across this statement in MET_Users_Guide_v3.0.3 (page 5-3),
"... The
> user must specify several parameters in the grid_stat configuration
file to
> utilize the neighbourhood approach, such as the *interpolation
method*,
> size of smoothing window,..."
>
> I would like to confirm that specifying the interpolation method is
only
> required if one wants to upscale the information and use traditional
> verification metrics on the coarser grid, isn't it? So, if one were
to use
> the FSS and the BSS to investigate the results of the neighbourhood
> approach, one will not be required to define the "interpolation
method" and
> other parameters related to the interpolation/smoothing approach. Am
I
> right to say this?
>
> Thanks,
> Raizan
------------------------------------------------
Subject: Defining Interpolation Method in the GridStat's Neighbourhood Approach
From: Zan Rahmat
Time: Tue May 29 10:11:42 2012
Dear Paul,
Thanks for the clarifications.
If I may ask another question related to the neighbourhood method of
verification, the NBRCTC and NBRCTS (Neighbourhood Contigency Table
Counts
and Statistics) are calculated based only on the:
1. cov_thres[],
2. nbr_width[],
3. nbr_thresh and also,
4. *fcst_thres[]* isn't it?
I'm quite certain of the first 3, and would like to confirm the 4th
one is
used as well. Because there's a mention of "threshold of 2" in the
example
given on page 5-11 of MET_Users_Guide_v3.0.3, but I am not sure where
this
threshold is supposed to come from. I'm referring to this statement in
the
user guide:
".. For example, if 10 of the 25 forecast grid points contain values
larger
than a *threshold of 2*, then the coverage is 10/25 = 0.4. If the
coverage
threshold is set to 0.5, then this neighborhood is considered to be a
“No”
forecast...."
Thanks,
Raizan
On 29 May 2012 23:29, Paul Oldenburg via RT <met_help at ucar.edu> wrote:
> Raizan,
>
> You are basically correct, but you still do need to specify an
> interpolation method in the grid_stat config file. The
> following interpolation settings basically mean "no interpolation":
>
> interp_method[] = [ "UW_MEAN" ];
> interp_width[] = [ 1 ];
>
> You should make sure that the nbr_width and nbr_thresh settings have
> values that are appropriate to your verification
> experiment. The the nbr_width setting specifies the size of the
> neighborhood(s) over which FSS and other statistics are
> calculated. Finally, you should set the output_flag switches at
indexes
> 13-15 to 1 or 2 to trigger the calculation of
> neighborhood statistics.
>
> I hope this answers your question.
>
> Thanks,
>
> Paul
>
>
> On 05/29/2012 03:13 AM, Zan Rahmat via RT wrote:
> >
> > Tue May 29 03:13:33 2012: Request 56648 was acted upon.
> > Transaction: Ticket created by raizan.rahmat at gmail.com
> > Queue: met_help
> > Subject: Defining Interpolation Method in the GridStat's
> Neighbourhood Approach
> > Owner: Nobody
> > Requestors: raizan.rahmat at gmail.com
> > Status: new
> > Ticket<URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=56648>
> >
> >
> > Hi,
> >
> > I came across this statement in MET_Users_Guide_v3.0.3 (page 5-3),
"...
> The
> > user must specify several parameters in the grid_stat
configuration file
> to
> > utilize the neighbourhood approach, such as the *interpolation
method*,
> > size of smoothing window,..."
> >
> > I would like to confirm that specifying the interpolation method
is only
> > required if one wants to upscale the information and use
traditional
> > verification metrics on the coarser grid, isn't it? So, if one
were to
> use
> > the FSS and the BSS to investigate the results of the
neighbourhood
> > approach, one will not be required to define the "interpolation
method"
> and
> > other parameters related to the interpolation/smoothing approach.
Am I
> > right to say this?
> >
> > Thanks,
> > Raizan
>
>
>
------------------------------------------------
Subject: Re: [rt.rap.ucar.edu #56648] Defining Interpolation Method in the GridStat's Neighbourhood Approach
From: Paul Oldenburg
Time: Tue May 29 10:15:57 2012
Raizan,
You are correct that fcst_thresh must be specified. This threshold
defines an event at a particular point in a
neighborhood.
Paul
On 05/29/2012 10:11 AM, Zan Rahmat via RT wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=56648>
>
> Dear Paul,
>
> Thanks for the clarifications.
>
> If I may ask another question related to the neighbourhood method of
> verification, the NBRCTC and NBRCTS (Neighbourhood Contigency Table
Counts
> and Statistics) are calculated based only on the:
>
> 1. cov_thres[],
> 2. nbr_width[],
> 3. nbr_thresh and also,
> 4. *fcst_thres[]* isn't it?
>
> I'm quite certain of the first 3, and would like to confirm the 4th
one is
> used as well. Because there's a mention of "threshold of 2" in the
example
> given on page 5-11 of MET_Users_Guide_v3.0.3, but I am not sure
where this
> threshold is supposed to come from. I'm referring to this statement
in the
> user guide:
>
> ".. For example, if 10 of the 25 forecast grid points contain values
larger
> than a *threshold of 2*, then the coverage is 10/25 = 0.4. If the
coverage
> threshold is set to 0.5, then this neighborhood is considered to be
a “No”
> forecast...."
>
> Thanks,
> Raizan
>
> On 29 May 2012 23:29, Paul Oldenburg via RT<met_help at ucar.edu>
wrote:
>
>> Raizan,
>>
>> You are basically correct, but you still do need to specify an
>> interpolation method in the grid_stat config file. The
>> following interpolation settings basically mean "no interpolation":
>>
>> interp_method[] = [ "UW_MEAN" ];
>> interp_width[] = [ 1 ];
>>
>> You should make sure that the nbr_width and nbr_thresh settings
have
>> values that are appropriate to your verification
>> experiment. The the nbr_width setting specifies the size of the
>> neighborhood(s) over which FSS and other statistics are
>> calculated. Finally, you should set the output_flag switches at
indexes
>> 13-15 to 1 or 2 to trigger the calculation of
>> neighborhood statistics.
>>
>> I hope this answers your question.
>>
>> Thanks,
>>
>> Paul
>>
>>
>> On 05/29/2012 03:13 AM, Zan Rahmat via RT wrote:
>>>
>>> Tue May 29 03:13:33 2012: Request 56648 was acted upon.
>>> Transaction: Ticket created by raizan.rahmat at gmail.com
>>> Queue: met_help
>>> Subject: Defining Interpolation Method in the GridStat's
>> Neighbourhood Approach
>>> Owner: Nobody
>>> Requestors: raizan.rahmat at gmail.com
>>> Status: new
>>> Ticket<URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=56648>
>>>
>>>
>>> Hi,
>>>
>>> I came across this statement in MET_Users_Guide_v3.0.3 (page 5-3),
"...
>> The
>>> user must specify several parameters in the grid_stat
configuration file
>> to
>>> utilize the neighbourhood approach, such as the *interpolation
method*,
>>> size of smoothing window,..."
>>>
>>> I would like to confirm that specifying the interpolation method
is only
>>> required if one wants to upscale the information and use
traditional
>>> verification metrics on the coarser grid, isn't it? So, if one
were to
>> use
>>> the FSS and the BSS to investigate the results of the
neighbourhood
>>> approach, one will not be required to define the "interpolation
method"
>> and
>>> other parameters related to the interpolation/smoothing approach.
Am I
>>> right to say this?
>>>
>>> Thanks,
>>> Raizan
>>
>>
>>
------------------------------------------------
Subject: Defining Interpolation Method in the GridStat's Neighbourhood Approach
From: Zan Rahmat
Time: Tue May 29 10:20:39 2012
Thanks Paul,
It's is clear now.
Warm Regards,
Raizan
On 30 May 2012 00:15, Paul Oldenburg via RT <met_help at ucar.edu> wrote:
> Raizan,
>
> You are correct that fcst_thresh must be specified. This threshold
> defines an event at a particular point in a
> neighborhood.
>
> Paul
>
>
> On 05/29/2012 10:11 AM, Zan Rahmat via RT wrote:
> >
> > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=56648>
> >
> > Dear Paul,
> >
> > Thanks for the clarifications.
> >
> > If I may ask another question related to the neighbourhood method
of
> > verification, the NBRCTC and NBRCTS (Neighbourhood Contigency
Table
> Counts
> > and Statistics) are calculated based only on the:
> >
> > 1. cov_thres[],
> > 2. nbr_width[],
> > 3. nbr_thresh and also,
> > 4. *fcst_thres[]* isn't it?
> >
> > I'm quite certain of the first 3, and would like to confirm the
4th one
> is
> > used as well. Because there's a mention of "threshold of 2" in the
> example
> > given on page 5-11 of MET_Users_Guide_v3.0.3, but I am not sure
where
> this
> > threshold is supposed to come from. I'm referring to this
statement in
> the
> > user guide:
> >
> > ".. For example, if 10 of the 25 forecast grid points contain
values
> larger
> > than a *threshold of 2*, then the coverage is 10/25 = 0.4. If the
> coverage
> > threshold is set to 0.5, then this neighborhood is considered to
be a
> “No”
> > forecast...."
> >
> > Thanks,
> > Raizan
> >
> > On 29 May 2012 23:29, Paul Oldenburg via RT<met_help at ucar.edu>
wrote:
> >
> >> Raizan,
> >>
> >> You are basically correct, but you still do need to specify an
> >> interpolation method in the grid_stat config file. The
> >> following interpolation settings basically mean "no
interpolation":
> >>
> >> interp_method[] = [ "UW_MEAN" ];
> >> interp_width[] = [ 1 ];
> >>
> >> You should make sure that the nbr_width and nbr_thresh settings
have
> >> values that are appropriate to your verification
> >> experiment. The the nbr_width setting specifies the size of the
> >> neighborhood(s) over which FSS and other statistics are
> >> calculated. Finally, you should set the output_flag switches at
indexes
> >> 13-15 to 1 or 2 to trigger the calculation of
> >> neighborhood statistics.
> >>
> >> I hope this answers your question.
> >>
> >> Thanks,
> >>
> >> Paul
> >>
> >>
> >> On 05/29/2012 03:13 AM, Zan Rahmat via RT wrote:
> >>>
> >>> Tue May 29 03:13:33 2012: Request 56648 was acted upon.
> >>> Transaction: Ticket created by raizan.rahmat at gmail.com
> >>> Queue: met_help
> >>> Subject: Defining Interpolation Method in the GridStat's
> >> Neighbourhood Approach
> >>> Owner: Nobody
> >>> Requestors: raizan.rahmat at gmail.com
> >>> Status: new
> >>> Ticket<URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=56648
> >
> >>>
> >>>
> >>> Hi,
> >>>
> >>> I came across this statement in MET_Users_Guide_v3.0.3 (page 5-
3), "...
> >> The
> >>> user must specify several parameters in the grid_stat
configuration
> file
> >> to
> >>> utilize the neighbourhood approach, such as the *interpolation
method*,
> >>> size of smoothing window,..."
> >>>
> >>> I would like to confirm that specifying the interpolation method
is
> only
> >>> required if one wants to upscale the information and use
traditional
> >>> verification metrics on the coarser grid, isn't it? So, if one
were to
> >> use
> >>> the FSS and the BSS to investigate the results of the
neighbourhood
> >>> approach, one will not be required to define the "interpolation
method"
> >> and
> >>> other parameters related to the interpolation/smoothing
approach. Am I
> >>> right to say this?
> >>>
> >>> Thanks,
> >>> Raizan
> >>
> >>
> >>
>
>
>
------------------------------------------------
More information about the Met_help
mailing list