# [Met_help] [rt.rap.ucar.edu #64753] History for Convex hull

John Halley Gotway via RT met_help at ucar.edu
Fri Apr 11 12:25:40 MDT 2014

```----------------------------------------------------------------
Initial Request
----------------------------------------------------------------

Hi,

I have been using MODE and MODE-Analysis for my forecast verification.

At this point, we are investigation a way to display the forecast. We
want to display the raw gridded forecast, but also display some object
information computed by MODE.

So, we have a gridded forecast on which we use the MODE tool to merge
forecast objects that are close enough. What we would like to display is
the raw forecast with the overlying convex hulls computed by MODE. In
the netCDF file I can find the points for each object boundary, but
can't find the points of the convex hull. Where can I find the points of
each convex hull?

Thanks

Anna-Belle Filion

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

Subject: Re: [rt.rap.ucar.edu #64753] Convex hull
From: John Halley Gotway
Time: Mon Jan 06 11:53:53 2014

Anna-Belle,

The convex hull points are not currently included in the output of
MODE.  However, as you mention, the boundary points are in the NetCDF
output file.

As for the convex hulls, really there are two options.  Convex hulls
are computed for the individual (or what we call "simple") objects in
both fields.  Those convex hulls are used in the computation
of fuzzy logic interest values.  Then matching/merging is performed,
which results in the groups of objects (or what we call "cluster"
objects) being defined.  The convex hulls shown in the postscript
output file are drawn around the matched clusters of objects.  If you
look closely, you'll see that there are no convex hulls drawn around
unmatched objects since they are not part of any matched
cluster object.

Are you interested in seeing the convex hulls which are drawn around
the individual "simple" objects or the ones drawn around the matched
"cluster" objects?   Or both?

Once we define this, I can create a development task to add this
functionality to the next version of MET.

Thanks,
John Halley Gotway
met_help at ucar.edu

On 12/30/2013 03:37 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>
> Mon Dec 30 15:37:29 2013: Request 64753 was acted upon.
> Transaction: Ticket created by anna-belle.filion at ec.gc.ca
>         Queue: met_help
>       Subject: Convex hull
>         Owner: Nobody
>    Requestors: anna-belle.filion at ec.gc.ca
>        Status: new
>   Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>
>
> Hi,
>
>
>
> I have been using MODE and MODE-Analysis for my forecast
verification.
>
>
>
> At this point, we are investigation a way to display the forecast.
We
> want to display the raw gridded forecast, but also display some
object
> information computed by MODE.
>
>
>
> So, we have a gridded forecast on which we use the MODE tool to
merge
> forecast objects that are close enough. What we would like to
display is
> the raw forecast with the overlying convex hulls computed by MODE.
In
> the netCDF file I can find the points for each object boundary, but
> can't find the points of the convex hull. Where can I find the
points of
> each convex hull?
>
>
>
> Thanks
>
> Anna-Belle Filion
>
>
>

------------------------------------------------
Subject: RE: [rt.rap.ucar.edu #64753] Convex hull
From: Filion,Anna-Belle [Montreal]
Time: Mon Jan 06 12:46:02 2014

Hi John,

I am interested un both. I would like to have access to the convex
hull points of cluster objects and individual simple objects.

Do you knew how long it will take to implement this functionality in
the code?

Thanks
Anna-Belle

-----Message d'origine-----
De : John Halley Gotway via RT [mailto:met_help at ucar.edu]
Envoyé : 6 janvier 2014 13:54
À : Filion,Anna-Belle [Montreal]
Objet : Re: [rt.rap.ucar.edu #64753] Convex hull

Anna-Belle,

The convex hull points are not currently included in the output of
MODE.  However, as you mention, the boundary points are in the NetCDF
output file.

As for the convex hulls, really there are two options.  Convex hulls
are computed for the individual (or what we call "simple") objects in
both fields.  Those convex hulls are used in the computation
of fuzzy logic interest values.  Then matching/merging is performed,
which results in the groups of objects (or what we call "cluster"
objects) being defined.  The convex hulls shown in the postscript
output file are drawn around the matched clusters of objects.  If you
look closely, you'll see that there are no convex hulls drawn around
unmatched objects since they are not part of any matched
cluster object.

Are you interested in seeing the convex hulls which are drawn around
the individual "simple" objects or the ones drawn around the matched
"cluster" objects?   Or both?

Once we define this, I can create a development task to add this
functionality to the next version of MET.

Thanks,
John Halley Gotway
met_help at ucar.edu

On 12/30/2013 03:37 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>
> Mon Dec 30 15:37:29 2013: Request 64753 was acted upon.
> Transaction: Ticket created by anna-belle.filion at ec.gc.ca
>         Queue: met_help
>       Subject: Convex hull
>         Owner: Nobody
>    Requestors: anna-belle.filion at ec.gc.ca
>        Status: new
>   Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>
>
> Hi,
>
>
>
> I have been using MODE and MODE-Analysis for my forecast
verification.
>
>
>
> At this point, we are investigation a way to display the forecast.
We
> want to display the raw gridded forecast, but also display some
object
> information computed by MODE.
>
>
>
> So, we have a gridded forecast on which we use the MODE tool to
merge
> forecast objects that are close enough. What we would like to
display is
> the raw forecast with the overlying convex hulls computed by MODE.
In
> the netCDF file I can find the points for each object boundary, but
> can't find the points of the convex hull. Where can I find the
points of
> each convex hull?
>
>
>
> Thanks
>
> Anna-Belle Filion
>
>
>

------------------------------------------------
Subject: Re: [rt.rap.ucar.edu #64753] Convex hull
From: John Halley Gotway
Time: Mon Jan 06 14:13:25 2014

Anna-Belle,

Can you tell me what version of MET you're using?  Is it METv4.1, the
most recent one?

Since this is an enhancement and not a bugfix, we wouldn't release it
as a patch to METv4.1.  It would likely be distributed in the next
release, which has not yet been scheduled.

That being said, I could take a look at the code to see how difficult
it would be to dump out those points.  If it's easy, I could probably
add it and send you a patched file.  If it's more involved,
it'd have to wait until the next release.

Thanks,
John

On 01/06/2014 12:46 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>
> Hi John,
>
> I am interested un both. I would like to have access to the convex
hull points of cluster objects and individual simple objects.
>
> Do you knew how long it will take to implement this functionality in
the code?
>
> Thanks
> Anna-Belle
>
> -----Message d'origine-----
> De : John Halley Gotway via RT [mailto:met_help at ucar.edu]
> Envoyé : 6 janvier 2014 13:54
> À : Filion,Anna-Belle [Montreal]
> Objet : Re: [rt.rap.ucar.edu #64753] Convex hull
>
> Anna-Belle,
>
> The convex hull points are not currently included in the output of
MODE.  However, as you mention, the boundary points are in the NetCDF
output file.
>
> As for the convex hulls, really there are two options.  Convex hulls
are computed for the individual (or what we call "simple") objects in
both fields.  Those convex hulls are used in the computation
> of fuzzy logic interest values.  Then matching/merging is performed,
which results in the groups of objects (or what we call "cluster"
objects) being defined.  The convex hulls shown in the postscript
> output file are drawn around the matched clusters of objects.  If
you look closely, you'll see that there are no convex hulls drawn
around unmatched objects since they are not part of any matched
> cluster object.
>
> Are you interested in seeing the convex hulls which are drawn around
the individual "simple" objects or the ones drawn around the matched
"cluster" objects?   Or both?
>
> Once we define this, I can create a development task to add this
functionality to the next version of MET.
>
> Thanks,
> John Halley Gotway
> met_help at ucar.edu
>
> On 12/30/2013 03:37 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>>
>> Mon Dec 30 15:37:29 2013: Request 64753 was acted upon.
>> Transaction: Ticket created by anna-belle.filion at ec.gc.ca
>>          Queue: met_help
>>        Subject: Convex hull
>>          Owner: Nobody
>>     Requestors: anna-belle.filion at ec.gc.ca
>>         Status: new
>>    Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>
>>
>> Hi,
>>
>>
>>
>> I have been using MODE and MODE-Analysis for my forecast
verification.
>>
>>
>>
>> At this point, we are investigation a way to display the forecast.
We
>> want to display the raw gridded forecast, but also display some
object
>> information computed by MODE.
>>
>>
>>
>> So, we have a gridded forecast on which we use the MODE tool to
merge
>> forecast objects that are close enough. What we would like to
display is
>> the raw forecast with the overlying convex hulls computed by MODE.
In
>> the netCDF file I can find the points for each object boundary, but
>> can't find the points of the convex hull. Where can I find the
points of
>> each convex hull?
>>
>>
>>
>> Thanks
>>
>> Anna-Belle Filion
>>
>>
>>
>
>

------------------------------------------------
Subject: RE: [rt.rap.ucar.edu #64753] Convex hull
From: Filion,Anna-Belle [Montreal]
Time: Mon Jan 06 14:29:14 2014

Hi John,

Ok. Just let me know the solution you choose after analysing the code
and the approximate deadline of a patch if this is a possible
solution.

Thanks
Anna-Belle

-----Message d'origine-----
De : John Halley Gotway via RT [mailto:met_help at ucar.edu]
Envoyé : 6 janvier 2014 16:13
À : Filion,Anna-Belle [Montreal]
Objet : Re: [rt.rap.ucar.edu #64753] Convex hull

Anna-Belle,

Can you tell me what version of MET you're using?  Is it METv4.1, the
most recent one?

Since this is an enhancement and not a bugfix, we wouldn't release it
as a patch to METv4.1.  It would likely be distributed in the next
release, which has not yet been scheduled.

That being said, I could take a look at the code to see how difficult
it would be to dump out those points.  If it's easy, I could probably
add it and send you a patched file.  If it's more involved,
it'd have to wait until the next release.

Thanks,
John

On 01/06/2014 12:46 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>
> Hi John,
>
> I am interested un both. I would like to have access to the convex
hull points of cluster objects and individual simple objects.
>
> Do you knew how long it will take to implement this functionality in
the code?
>
> Thanks
> Anna-Belle
>
> -----Message d'origine-----
> De : John Halley Gotway via RT [mailto:met_help at ucar.edu]
> Envoyé : 6 janvier 2014 13:54
> À : Filion,Anna-Belle [Montreal]
> Objet : Re: [rt.rap.ucar.edu #64753] Convex hull
>
> Anna-Belle,
>
> The convex hull points are not currently included in the output of
MODE.  However, as you mention, the boundary points are in the NetCDF
output file.
>
> As for the convex hulls, really there are two options.  Convex hulls
are computed for the individual (or what we call "simple") objects in
both fields.  Those convex hulls are used in the computation
> of fuzzy logic interest values.  Then matching/merging is performed,
which results in the groups of objects (or what we call "cluster"
objects) being defined.  The convex hulls shown in the postscript
> output file are drawn around the matched clusters of objects.  If
you look closely, you'll see that there are no convex hulls drawn
around unmatched objects since they are not part of any matched
> cluster object.
>
> Are you interested in seeing the convex hulls which are drawn around
the individual "simple" objects or the ones drawn around the matched
"cluster" objects?   Or both?
>
> Once we define this, I can create a development task to add this
functionality to the next version of MET.
>
> Thanks,
> John Halley Gotway
> met_help at ucar.edu
>
> On 12/30/2013 03:37 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>>
>> Mon Dec 30 15:37:29 2013: Request 64753 was acted upon.
>> Transaction: Ticket created by anna-belle.filion at ec.gc.ca
>>          Queue: met_help
>>        Subject: Convex hull
>>          Owner: Nobody
>>     Requestors: anna-belle.filion at ec.gc.ca
>>         Status: new
>>    Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>
>>
>> Hi,
>>
>>
>>
>> I have been using MODE and MODE-Analysis for my forecast
verification.
>>
>>
>>
>> At this point, we are investigation a way to display the forecast.
We
>> want to display the raw gridded forecast, but also display some
object
>> information computed by MODE.
>>
>>
>>
>> So, we have a gridded forecast on which we use the MODE tool to
merge
>> forecast objects that are close enough. What we would like to
display is
>> the raw forecast with the overlying convex hulls computed by MODE.
In
>> the netCDF file I can find the points for each object boundary, but
>> can't find the points of the convex hull. Where can I find the
points of
>> each convex hull?
>>
>>
>>
>> Thanks
>>
>> Anna-Belle Filion
>>
>>
>>
>
>

------------------------------------------------
Subject: Convex hull
From: John Halley Gotway
Time: Thu Jan 09 14:39:26 2014

Anne-Belle,

I've attached a tar file that contains 4 updated files (met_file.h/.cc
and nc_utils.h/.cc).  You can place this in the top-level METv4.1
directory and extract it:
tar -xvzf METv4.1_patch_for_anna_belle.tar.gz

And then recompile METv4.1.

When you run MODE, you should notice many more variables in the NetCDF
object files.  I tweaked the naming convention of the existing
variables to be consistent with the new ones.  You'll now see data
for 6 different polyline types:
- Forecast and Observation Simple Boundary objects (fcst_simp_bdy
and obs_simp_bdy)
- Forecast and Observation Simple Convex Hull objects
(fcst_simp_hull and obs_simp_hull)
- Forecast and Observation Cluster Convex Hull objects
(fcst_clus_hull and obs_clus_hull)

Please take a look and let me know if this is what you were looking
for.

We'll plan to add these changes in to the next release of MET.
However, we want to make their output optional via the MODE
configuration file.

Thanks,
John Halley Gotway

On 01/06/2014 02:29 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>
> Hi John,
>
>
> Ok. Just let me know the solution you choose after analysing the
code and the approximate deadline of a patch if this is a possible
solution.
>
> Thanks
> Anna-Belle
>
>
>
> -----Message d'origine-----
> De : John Halley Gotway via RT [mailto:met_help at ucar.edu]
> Envoyé : 6 janvier 2014 16:13
> À : Filion,Anna-Belle [Montreal]
> Objet : Re: [rt.rap.ucar.edu #64753] Convex hull
>
> Anna-Belle,
>
> Can you tell me what version of MET you're using?  Is it METv4.1,
the most recent one?
>
> Since this is an enhancement and not a bugfix, we wouldn't release
it as a patch to METv4.1.  It would likely be distributed in the next
release, which has not yet been scheduled.
>
> That being said, I could take a look at the code to see how
difficult it would be to dump out those points.  If it's easy, I could
probably add it and send you a patched file.  If it's more involved,
> it'd have to wait until the next release.
>
> Thanks,
> John
>
> On 01/06/2014 12:46 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>>
>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>
>> Hi John,
>>
>> I am interested un both. I would like to have access to the convex
hull points of cluster objects and individual simple objects.
>>
>> Do you knew how long it will take to implement this functionality
in the code?
>>
>> Thanks
>> Anna-Belle
>>
>> -----Message d'origine-----
>> De : John Halley Gotway via RT [mailto:met_help at ucar.edu]
>> Envoyé : 6 janvier 2014 13:54
>> À : Filion,Anna-Belle [Montreal]
>> Objet : Re: [rt.rap.ucar.edu #64753] Convex hull
>>
>> Anna-Belle,
>>
>> The convex hull points are not currently included in the output of
MODE.  However, as you mention, the boundary points are in the NetCDF
output file.
>>
>> As for the convex hulls, really there are two options.  Convex
hulls are computed for the individual (or what we call "simple")
objects in both fields.  Those convex hulls are used in the
computation
>> of fuzzy logic interest values.  Then matching/merging is
performed, which results in the groups of objects (or what we call
"cluster" objects) being defined.  The convex hulls shown in the
postscript
>> output file are drawn around the matched clusters of objects.  If
you look closely, you'll see that there are no convex hulls drawn
around unmatched objects since they are not part of any matched
>> cluster object.
>>
>> Are you interested in seeing the convex hulls which are drawn
around the individual "simple" objects or the ones drawn around the
matched "cluster" objects?   Or both?
>>
>> Once we define this, I can create a development task to add this
functionality to the next version of MET.
>>
>> Thanks,
>> John Halley Gotway
>> met_help at ucar.edu
>>
>> On 12/30/2013 03:37 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>>>
>>> Mon Dec 30 15:37:29 2013: Request 64753 was acted upon.
>>> Transaction: Ticket created by anna-belle.filion at ec.gc.ca
>>>           Queue: met_help
>>>         Subject: Convex hull
>>>           Owner: Nobody
>>>      Requestors: anna-belle.filion at ec.gc.ca
>>>          Status: new
>>>     Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>>
>>>
>>> Hi,
>>>
>>>
>>>
>>> I have been using MODE and MODE-Analysis for my forecast
verification.
>>>
>>>
>>>
>>> At this point, we are investigation a way to display the forecast.
We
>>> want to display the raw gridded forecast, but also display some
object
>>> information computed by MODE.
>>>
>>>
>>>
>>> So, we have a gridded forecast on which we use the MODE tool to
merge
>>> forecast objects that are close enough. What we would like to
display is
>>> the raw forecast with the overlying convex hulls computed by MODE.
In
>>> the netCDF file I can find the points for each object boundary,
but
>>> can't find the points of the convex hull. Where can I find the
points of
>>> each convex hull?
>>>
>>>
>>>
>>> Thanks
>>>
>>> Anna-Belle Filion
>>>
>>>
>>>
>>
>>
>
>

------------------------------------------------
Subject: RE: [rt.rap.ucar.edu #64753] Convex hull
From: Filion,Anna-Belle [Montreal]
Time: Mon Jan 13 11:05:07 2014

Hi John,

Yes, this is perfect!

Thanks a lot!
Anna-Belle

-----Message d'origine-----
De : John Halley Gotway via RT [mailto:met_help at ucar.edu]
Envoyé : 9 janvier 2014 16:39
À : Filion,Anna-Belle [Montreal]
Objet : Re: [rt.rap.ucar.edu #64753] Convex hull

Anne-Belle,

I've attached a tar file that contains 4 updated files (met_file.h/.cc
and nc_utils.h/.cc).  You can place this in the top-level METv4.1
directory and extract it:
tar -xvzf METv4.1_patch_for_anna_belle.tar.gz

And then recompile METv4.1.

When you run MODE, you should notice many more variables in the NetCDF
object files.  I tweaked the naming convention of the existing
variables to be consistent with the new ones.  You'll now see data for
6 different polyline types:
- Forecast and Observation Simple Boundary objects (fcst_simp_bdy
and obs_simp_bdy)
- Forecast and Observation Simple Convex Hull objects
(fcst_simp_hull and obs_simp_hull)
- Forecast and Observation Cluster Convex Hull objects
(fcst_clus_hull and obs_clus_hull)

Please take a look and let me know if this is what you were looking
for.

We'll plan to add these changes in to the next release of MET.
However, we want to make their output optional via the MODE
configuration file.

Thanks,
John Halley Gotway

On 01/06/2014 02:29 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>
> Hi John,
>
>
> Ok. Just let me know the solution you choose after analysing the
code and the approximate deadline of a patch if this is a possible
solution.
>
> Thanks
> Anna-Belle
>
>
>
> -----Message d'origine-----
> De : John Halley Gotway via RT [mailto:met_help at ucar.edu] Envoyé : 6
> janvier 2014 16:13 À : Filion,Anna-Belle [Montreal] Objet : Re:
> [rt.rap.ucar.edu #64753] Convex hull
>
> Anna-Belle,
>
> Can you tell me what version of MET you're using?  Is it METv4.1,
the most recent one?
>
> Since this is an enhancement and not a bugfix, we wouldn't release
it as a patch to METv4.1.  It would likely be distributed in the next
release, which has not yet been scheduled.
>
> That being said, I could take a look at the code to see how
difficult
> it would be to dump out those points.  If it's easy, I could
probably add it and send you a patched file.  If it's more involved,
it'd have to wait until the next release.
>
> Thanks,
> John
>
> On 01/06/2014 12:46 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>>
>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>
>> Hi John,
>>
>> I am interested un both. I would like to have access to the convex
hull points of cluster objects and individual simple objects.
>>
>> Do you knew how long it will take to implement this functionality
in the code?
>>
>> Thanks
>> Anna-Belle
>>
>> -----Message d'origine-----
>> De : John Halley Gotway via RT [mailto:met_help at ucar.edu] Envoyé :
6
>> janvier 2014 13:54 À : Filion,Anna-Belle [Montreal] Objet : Re:
>> [rt.rap.ucar.edu #64753] Convex hull
>>
>> Anna-Belle,
>>
>> The convex hull points are not currently included in the output of
MODE.  However, as you mention, the boundary points are in the NetCDF
output file.
>>
>> As for the convex hulls, really there are two options.  Convex
hulls
>> are computed for the individual (or what we call "simple") objects
in
>> both fields.  Those convex hulls are used in the computation of
fuzzy
>> logic interest values.  Then matching/merging is performed, which
results in the groups of objects (or what we call "cluster" objects)
being defined.  The convex hulls shown in the postscript output file
are drawn around the matched clusters of objects.  If you look
closely, you'll see that there are no convex hulls drawn around
unmatched objects since they are not part of any matched cluster
object.
>>
>> Are you interested in seeing the convex hulls which are drawn
around the individual "simple" objects or the ones drawn around the
matched "cluster" objects?   Or both?
>>
>> Once we define this, I can create a development task to add this
functionality to the next version of MET.
>>
>> Thanks,
>> John Halley Gotway
>> met_help at ucar.edu
>>
>> On 12/30/2013 03:37 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>>>
>>> Mon Dec 30 15:37:29 2013: Request 64753 was acted upon.
>>> Transaction: Ticket created by anna-belle.filion at ec.gc.ca
>>>           Queue: met_help
>>>         Subject: Convex hull
>>>           Owner: Nobody
>>>      Requestors: anna-belle.filion at ec.gc.ca
>>>          Status: new
>>>     Ticket <URL:
>>> https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>>
>>>
>>> Hi,
>>>
>>>
>>>
>>> I have been using MODE and MODE-Analysis for my forecast
verification.
>>>
>>>
>>>
>>> At this point, we are investigation a way to display the forecast.
>>> We want to display the raw gridded forecast, but also display some
>>> object information computed by MODE.
>>>
>>>
>>>
>>> So, we have a gridded forecast on which we use the MODE tool to
>>> merge forecast objects that are close enough. What we would like
to
>>> display is the raw forecast with the overlying convex hulls
computed
>>> by MODE. In the netCDF file I can find the points for each object
>>> boundary, but can't find the points of the convex hull. Where can
I
>>> find the points of each convex hull?
>>>
>>>
>>>
>>> Thanks
>>>
>>> Anna-Belle Filion
>>>
>>>
>>>
>>
>>
>
>

------------------------------------------------
Subject: Re: [rt.rap.ucar.edu #64753] Convex hull
From: John Halley Gotway
Time: Tue Jan 14 16:11:42 2014

Anna-Belle,

Great.  Glad to hear it.  I'll include these changes in the trunk so
that they show up in the next release.

Thanks,
John

On 01/13/2014 11:05 AM, Filion,Anna-Belle [Montreal] via RT wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>
> Hi John,
>
> Yes, this is perfect!
>
> Thanks a lot!
> Anna-Belle
>
>
> -----Message d'origine-----
> De : John Halley Gotway via RT [mailto:met_help at ucar.edu]
> Envoyé : 9 janvier 2014 16:39
> À : Filion,Anna-Belle [Montreal]
> Objet : Re: [rt.rap.ucar.edu #64753] Convex hull
>
> Anne-Belle,
>
> I've attached a tar file that contains 4 updated files
(met_file.h/.cc and nc_utils.h/.cc).  You can place this in the top-
level METv4.1 directory and extract it:
>     tar -xvzf METv4.1_patch_for_anna_belle.tar.gz
>
> And then recompile METv4.1.
>
> When you run MODE, you should notice many more variables in the
NetCDF object files.  I tweaked the naming convention of the existing
variables to be consistent with the new ones.  You'll now see data for
6 different polyline types:
>    - Forecast and Observation Simple Boundary objects (fcst_simp_bdy
and obs_simp_bdy)
>    - Forecast and Observation Simple Convex Hull objects
(fcst_simp_hull and obs_simp_hull)
>    - Forecast and Observation Cluster Convex Hull objects
(fcst_clus_hull and obs_clus_hull)
>
> Please take a look and let me know if this is what you were looking
for.
>
> We'll plan to add these changes in to the next release of MET.
However, we want to make their output optional via the MODE
configuration file.
>
> Thanks,
> John Halley Gotway
>
> On 01/06/2014 02:29 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>>
>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>
>> Hi John,
>>
>>
>> Ok. Just let me know the solution you choose after analysing the
code and the approximate deadline of a patch if this is a possible
solution.
>>
>> Thanks
>> Anna-Belle
>>
>>
>>
>> -----Message d'origine-----
>> De : John Halley Gotway via RT [mailto:met_help at ucar.edu] Envoyé :
6
>> janvier 2014 16:13 À : Filion,Anna-Belle [Montreal] Objet : Re:
>> [rt.rap.ucar.edu #64753] Convex hull
>>
>> Anna-Belle,
>>
>> Can you tell me what version of MET you're using?  Is it METv4.1,
the most recent one?
>>
>> Since this is an enhancement and not a bugfix, we wouldn't release
it as a patch to METv4.1.  It would likely be distributed in the next
release, which has not yet been scheduled.
>>
>> That being said, I could take a look at the code to see how
difficult
>> it would be to dump out those points.  If it's easy, I could
probably add it and send you a patched file.  If it's more involved,
it'd have to wait until the next release.
>>
>> Thanks,
>> John
>>
>> On 01/06/2014 12:46 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>>>
>>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>>
>>> Hi John,
>>>
>>> I am interested un both. I would like to have access to the convex
hull points of cluster objects and individual simple objects.
>>>
>>> Do you knew how long it will take to implement this functionality
in the code?
>>>
>>> Thanks
>>> Anna-Belle
>>>
>>> -----Message d'origine-----
>>> De : John Halley Gotway via RT [mailto:met_help at ucar.edu] Envoyé :
6
>>> janvier 2014 13:54 À : Filion,Anna-Belle [Montreal] Objet : Re:
>>> [rt.rap.ucar.edu #64753] Convex hull
>>>
>>> Anna-Belle,
>>>
>>> The convex hull points are not currently included in the output of
MODE.  However, as you mention, the boundary points are in the NetCDF
output file.
>>>
>>> As for the convex hulls, really there are two options.  Convex
hulls
>>> are computed for the individual (or what we call "simple") objects
in
>>> both fields.  Those convex hulls are used in the computation of
fuzzy
>>> logic interest values.  Then matching/merging is performed, which
results in the groups of objects (or what we call "cluster" objects)
being defined.  The convex hulls shown in the postscript output file
are drawn around the matched clusters of objects.  If you look
closely, you'll see that there are no convex hulls drawn around
unmatched objects since they are not part of any matched cluster
object.
>>>
>>> Are you interested in seeing the convex hulls which are drawn
around the individual "simple" objects or the ones drawn around the
matched "cluster" objects?   Or both?
>>>
>>> Once we define this, I can create a development task to add this
functionality to the next version of MET.
>>>
>>> Thanks,
>>> John Halley Gotway
>>> met_help at ucar.edu
>>>
>>> On 12/30/2013 03:37 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>>>>
>>>> Mon Dec 30 15:37:29 2013: Request 64753 was acted upon.
>>>> Transaction: Ticket created by anna-belle.filion at ec.gc.ca
>>>>            Queue: met_help
>>>>          Subject: Convex hull
>>>>            Owner: Nobody
>>>>       Requestors: anna-belle.filion at ec.gc.ca
>>>>           Status: new
>>>>      Ticket <URL:
>>>> https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>>>
>>>>
>>>> Hi,
>>>>
>>>>
>>>>
>>>> I have been using MODE and MODE-Analysis for my forecast
verification.
>>>>
>>>>
>>>>
>>>> At this point, we are investigation a way to display the
forecast.
>>>> We want to display the raw gridded forecast, but also display
some
>>>> object information computed by MODE.
>>>>
>>>>
>>>>
>>>> So, we have a gridded forecast on which we use the MODE tool to
>>>> merge forecast objects that are close enough. What we would like
to
>>>> display is the raw forecast with the overlying convex hulls
computed
>>>> by MODE. In the netCDF file I can find the points for each object
>>>> boundary, but can't find the points of the convex hull. Where can
I
>>>> find the points of each convex hull?
>>>>
>>>>
>>>>
>>>> Thanks
>>>>
>>>> Anna-Belle Filion
>>>>
>>>>
>>>>
>>>
>>>
>>
>>
>
>

------------------------------------------------
Subject: RE: [rt.rap.ucar.edu #64753] Convex hull
From: Filion,Anna-Belle [Montreal]
Time: Thu Apr 10 11:30:32 2014

Hi john,

I tested this with a case where the forecast was matched with the
observation, so it was fine.

However, if the forecast is not matched to any observation, the
forecast cluster convex hull is not available only the forecast simple
convex hull is.

However, MODE still goes in the process of merging objects of the same
field to form bigger cluster object because we can see the cluster
convex hull in the .ps file.

Is there a way that the cluster convex hull could be available even
though the forecast was not matched to the observation?

Thanks
Anna-Belle Filion

-----Message d'origine-----
De : John Halley Gotway via RT [mailto:met_help at ucar.edu]
Envoyé : 9 janvier 2014 16:39
À : Filion,Anna-Belle [Montreal]
Objet : Re: [rt.rap.ucar.edu #64753] Convex hull

Anne-Belle,

I've attached a tar file that contains 4 updated files (met_file.h/.cc
and nc_utils.h/.cc).  You can place this in the top-level METv4.1
directory and extract it:
tar -xvzf METv4.1_patch_for_anna_belle.tar.gz

And then recompile METv4.1.

When you run MODE, you should notice many more variables in the NetCDF
object files.  I tweaked the naming convention of the existing
variables to be consistent with the new ones.  You'll now see data for
6 different polyline types:
- Forecast and Observation Simple Boundary objects (fcst_simp_bdy
and obs_simp_bdy)
- Forecast and Observation Simple Convex Hull objects
(fcst_simp_hull and obs_simp_hull)
- Forecast and Observation Cluster Convex Hull objects
(fcst_clus_hull and obs_clus_hull)

Please take a look and let me know if this is what you were looking
for.

We'll plan to add these changes in to the next release of MET.
However, we want to make their output optional via the MODE
configuration file.

Thanks,
John Halley Gotway

On 01/06/2014 02:29 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>
> Hi John,
>
>
> Ok. Just let me know the solution you choose after analysing the
code and the approximate deadline of a patch if this is a possible
solution.
>
> Thanks
> Anna-Belle
>
>
>
> -----Message d'origine-----
> De : John Halley Gotway via RT [mailto:met_help at ucar.edu] Envoyé : 6
> janvier 2014 16:13 À : Filion,Anna-Belle [Montreal] Objet : Re:
> [rt.rap.ucar.edu #64753] Convex hull
>
> Anna-Belle,
>
> Can you tell me what version of MET you're using?  Is it METv4.1,
the most recent one?
>
> Since this is an enhancement and not a bugfix, we wouldn't release
it as a patch to METv4.1.  It would likely be distributed in the next
release, which has not yet been scheduled.
>
> That being said, I could take a look at the code to see how
difficult
> it would be to dump out those points.  If it's easy, I could
probably add it and send you a patched file.  If it's more involved,
it'd have to wait until the next release.
>
> Thanks,
> John
>
> On 01/06/2014 12:46 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>>
>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>
>> Hi John,
>>
>> I am interested un both. I would like to have access to the convex
hull points of cluster objects and individual simple objects.
>>
>> Do you knew how long it will take to implement this functionality
in the code?
>>
>> Thanks
>> Anna-Belle
>>
>> -----Message d'origine-----
>> De : John Halley Gotway via RT [mailto:met_help at ucar.edu] Envoyé :
6
>> janvier 2014 13:54 À : Filion,Anna-Belle [Montreal] Objet : Re:
>> [rt.rap.ucar.edu #64753] Convex hull
>>
>> Anna-Belle,
>>
>> The convex hull points are not currently included in the output of
MODE.  However, as you mention, the boundary points are in the NetCDF
output file.
>>
>> As for the convex hulls, really there are two options.  Convex
hulls
>> are computed for the individual (or what we call "simple") objects
in
>> both fields.  Those convex hulls are used in the computation of
fuzzy
>> logic interest values.  Then matching/merging is performed, which
results in the groups of objects (or what we call "cluster" objects)
being defined.  The convex hulls shown in the postscript output file
are drawn around the matched clusters of objects.  If you look
closely, you'll see that there are no convex hulls drawn around
unmatched objects since they are not part of any matched cluster
object.
>>
>> Are you interested in seeing the convex hulls which are drawn
around the individual "simple" objects or the ones drawn around the
matched "cluster" objects?   Or both?
>>
>> Once we define this, I can create a development task to add this
functionality to the next version of MET.
>>
>> Thanks,
>> John Halley Gotway
>> met_help at ucar.edu
>>
>> On 12/30/2013 03:37 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>>>
>>> Mon Dec 30 15:37:29 2013: Request 64753 was acted upon.
>>> Transaction: Ticket created by anna-belle.filion at ec.gc.ca
>>>           Queue: met_help
>>>         Subject: Convex hull
>>>           Owner: Nobody
>>>      Requestors: anna-belle.filion at ec.gc.ca
>>>          Status: new
>>>     Ticket <URL:
>>> https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>>
>>>
>>> Hi,
>>>
>>>
>>>
>>> I have been using MODE and MODE-Analysis for my forecast
verification.
>>>
>>>
>>>
>>> At this point, we are investigation a way to display the forecast.
>>> We want to display the raw gridded forecast, but also display some
>>> object information computed by MODE.
>>>
>>>
>>>
>>> So, we have a gridded forecast on which we use the MODE tool to
>>> merge forecast objects that are close enough. What we would like
to
>>> display is the raw forecast with the overlying convex hulls
computed
>>> by MODE. In the netCDF file I can find the points for each object
>>> boundary, but can't find the points of the convex hull. Where can
I
>>> find the points of each convex hull?
>>>
>>>
>>>
>>> Thanks
>>>
>>> Anna-Belle Filion
>>>
>>>
>>>
>>
>>
>
>

------------------------------------------------
Subject: Re: [rt.rap.ucar.edu #64753] Convex hull
From: John Halley Gotway
Time: Thu Apr 10 12:53:27 2014

Anna,

By definition, a cluster object only exists when it is matched to
another cluster in the other field.  There is no such thing as an
unmatched cluster object.

Suppose you've chosen the double-threshold or fuzzy-engine method of
merging objects in each field.  After that merging step, MODE has
merged groups of objects in each field.  During the matching
step, MODE determines which objects from the forecast field match
objects in the observation field.  After the matching step, if there
are any groups of objects in the forecast or observation fields
whose members are all unmatched, that group is thrown away.  That's
why there are no unmatched clusters.

Does that make sense?

Thanks,
John

On 04/10/2014 11:30 AM, Filion,Anna-Belle [Montreal] via RT wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>
> Hi john,
>
> I tested this with a case where the forecast was matched with the
observation, so it was fine.
>
> However, if the forecast is not matched to any observation, the
forecast cluster convex hull is not available only the forecast simple
convex hull is.
>
> However, MODE still goes in the process of merging objects of the
same field to form bigger cluster object because we can see the
cluster convex hull in the .ps file.
>
> Is there a way that the cluster convex hull could be available even
though the forecast was not matched to the observation?
>
>
> Thanks
> Anna-Belle Filion
>
>
>
>
> -----Message d'origine-----
> De : John Halley Gotway via RT [mailto:met_help at ucar.edu]
> Envoyé : 9 janvier 2014 16:39
> À : Filion,Anna-Belle [Montreal]
> Objet : Re: [rt.rap.ucar.edu #64753] Convex hull
>
> Anne-Belle,
>
> I've attached a tar file that contains 4 updated files
(met_file.h/.cc and nc_utils.h/.cc).  You can place this in the top-
level METv4.1 directory and extract it:
>     tar -xvzf METv4.1_patch_for_anna_belle.tar.gz
>
> And then recompile METv4.1.
>
> When you run MODE, you should notice many more variables in the
NetCDF object files.  I tweaked the naming convention of the existing
variables to be consistent with the new ones.  You'll now see data for
6 different polyline types:
>    - Forecast and Observation Simple Boundary objects (fcst_simp_bdy
and obs_simp_bdy)
>    - Forecast and Observation Simple Convex Hull objects
(fcst_simp_hull and obs_simp_hull)
>    - Forecast and Observation Cluster Convex Hull objects
(fcst_clus_hull and obs_clus_hull)
>
> Please take a look and let me know if this is what you were looking
for.
>
> We'll plan to add these changes in to the next release of MET.
However, we want to make their output optional via the MODE
configuration file.
>
> Thanks,
> John Halley Gotway
>
> On 01/06/2014 02:29 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>>
>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>
>> Hi John,
>>
>>
>> Ok. Just let me know the solution you choose after analysing the
code and the approximate deadline of a patch if this is a possible
solution.
>>
>> Thanks
>> Anna-Belle
>>
>>
>>
>> -----Message d'origine-----
>> De : John Halley Gotway via RT [mailto:met_help at ucar.edu] Envoyé :
6
>> janvier 2014 16:13 À : Filion,Anna-Belle [Montreal] Objet : Re:
>> [rt.rap.ucar.edu #64753] Convex hull
>>
>> Anna-Belle,
>>
>> Can you tell me what version of MET you're using?  Is it METv4.1,
the most recent one?
>>
>> Since this is an enhancement and not a bugfix, we wouldn't release
it as a patch to METv4.1.  It would likely be distributed in the next
release, which has not yet been scheduled.
>>
>> That being said, I could take a look at the code to see how
difficult
>> it would be to dump out those points.  If it's easy, I could
probably add it and send you a patched file.  If it's more involved,
it'd have to wait until the next release.
>>
>> Thanks,
>> John
>>
>> On 01/06/2014 12:46 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>>>
>>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>>
>>> Hi John,
>>>
>>> I am interested un both. I would like to have access to the convex
hull points of cluster objects and individual simple objects.
>>>
>>> Do you knew how long it will take to implement this functionality
in the code?
>>>
>>> Thanks
>>> Anna-Belle
>>>
>>> -----Message d'origine-----
>>> De : John Halley Gotway via RT [mailto:met_help at ucar.edu] Envoyé :
6
>>> janvier 2014 13:54 À : Filion,Anna-Belle [Montreal] Objet : Re:
>>> [rt.rap.ucar.edu #64753] Convex hull
>>>
>>> Anna-Belle,
>>>
>>> The convex hull points are not currently included in the output of
MODE.  However, as you mention, the boundary points are in the NetCDF
output file.
>>>
>>> As for the convex hulls, really there are two options.  Convex
hulls
>>> are computed for the individual (or what we call "simple") objects
in
>>> both fields.  Those convex hulls are used in the computation of
fuzzy
>>> logic interest values.  Then matching/merging is performed, which
results in the groups of objects (or what we call "cluster" objects)
being defined.  The convex hulls shown in the postscript output file
are drawn around the matched clusters of objects.  If you look
closely, you'll see that there are no convex hulls drawn around
unmatched objects since they are not part of any matched cluster
object.
>>>
>>> Are you interested in seeing the convex hulls which are drawn
around the individual "simple" objects or the ones drawn around the
matched "cluster" objects?   Or both?
>>>
>>> Once we define this, I can create a development task to add this
functionality to the next version of MET.
>>>
>>> Thanks,
>>> John Halley Gotway
>>> met_help at ucar.edu
>>>
>>> On 12/30/2013 03:37 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>>>>
>>>> Mon Dec 30 15:37:29 2013: Request 64753 was acted upon.
>>>> Transaction: Ticket created by anna-belle.filion at ec.gc.ca
>>>>            Queue: met_help
>>>>          Subject: Convex hull
>>>>            Owner: Nobody
>>>>       Requestors: anna-belle.filion at ec.gc.ca
>>>>           Status: new
>>>>      Ticket <URL:
>>>> https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>>>
>>>>
>>>> Hi,
>>>>
>>>>
>>>>
>>>> I have been using MODE and MODE-Analysis for my forecast
verification.
>>>>
>>>>
>>>>
>>>> At this point, we are investigation a way to display the
forecast.
>>>> We want to display the raw gridded forecast, but also display
some
>>>> object information computed by MODE.
>>>>
>>>>
>>>>
>>>> So, we have a gridded forecast on which we use the MODE tool to
>>>> merge forecast objects that are close enough. What we would like
to
>>>> display is the raw forecast with the overlying convex hulls
computed
>>>> by MODE. In the netCDF file I can find the points for each object
>>>> boundary, but can't find the points of the convex hull. Where can
I
>>>> find the points of each convex hull?
>>>>
>>>>
>>>>
>>>> Thanks
>>>>
>>>> Anna-Belle Filion
>>>>
>>>>
>>>>
>>>
>>>
>>
>>
>
>

------------------------------------------------
Subject: Convex hull
From: Filion,Anna-Belle [Montreal]
Time: Thu Apr 10 14:01:59 2014

Hi John,

Yes, I understand, but when you look at the .ps file there is a convex
hull for merged object even if they were not matched

Is it possible to get the information on that convex hull?

If you look at the file .ps that is included in this email you might
understand better what I am looking for. In this case, there is no
forecasted object that is matched with the observed object. But on the
page 7,  there is a convex hull contour for the merged forecast
object.

Is there a way to easily modify the code to get this information?

Thanks
Anna-Belle

-----Message d'origine-----
De : John Halley Gotway via RT [mailto:met_help at ucar.edu]
Envoyé : 10 April 2014 2:53
À : Filion,Anna-Belle [Montreal]
Objet : Re: [rt.rap.ucar.edu #64753] Convex hull

Anna,

By definition, a cluster object only exists when it is matched to
another cluster in the other field.  There is no such thing as an
unmatched cluster object.

Suppose you've chosen the double-threshold or fuzzy-engine method of
merging objects in each field.  After that merging step, MODE has
merged groups of objects in each field.  During the matching step,
MODE determines which objects from the forecast field match objects in
the observation field.  After the matching step, if there are any
groups of objects in the forecast or observation fields whose members
are all unmatched, that group is thrown away.  That's why there are no
unmatched clusters.

Does that make sense?

Thanks,
John

On 04/10/2014 11:30 AM, Filion,Anna-Belle [Montreal] via RT wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>
> Hi john,
>
> I tested this with a case where the forecast was matched with the
observation, so it was fine.
>
> However, if the forecast is not matched to any observation, the
forecast cluster convex hull is not available only the forecast simple
convex hull is.
>
> However, MODE still goes in the process of merging objects of the
same field to form bigger cluster object because we can see the
cluster convex hull in the .ps file.
>
> Is there a way that the cluster convex hull could be available even
though the forecast was not matched to the observation?
>
>
> Thanks
> Anna-Belle Filion
>
>
>
>
> -----Message d'origine-----
> De : John Halley Gotway via RT [mailto:met_help at ucar.edu] Envoyé : 9
> janvier 2014 16:39 À : Filion,Anna-Belle [Montreal] Objet : Re:
> [rt.rap.ucar.edu #64753] Convex hull
>
> Anne-Belle,
>
> I've attached a tar file that contains 4 updated files
(met_file.h/.cc and nc_utils.h/.cc).  You can place this in the top-
level METv4.1 directory and extract it:
>     tar -xvzf METv4.1_patch_for_anna_belle.tar.gz
>
> And then recompile METv4.1.
>
> When you run MODE, you should notice many more variables in the
NetCDF object files.  I tweaked the naming convention of the existing
variables to be consistent with the new ones.  You'll now see data for
6 different polyline types:
>    - Forecast and Observation Simple Boundary objects (fcst_simp_bdy
and obs_simp_bdy)
>    - Forecast and Observation Simple Convex Hull objects
(fcst_simp_hull and obs_simp_hull)
>    - Forecast and Observation Cluster Convex Hull objects
> (fcst_clus_hull and obs_clus_hull)
>
> Please take a look and let me know if this is what you were looking
for.
>
> We'll plan to add these changes in to the next release of MET.
However, we want to make their output optional via the MODE
configuration file.
>
> Thanks,
> John Halley Gotway
>
> On 01/06/2014 02:29 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>>
>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>
>> Hi John,
>>
>>
>> Ok. Just let me know the solution you choose after analysing the
code and the approximate deadline of a patch if this is a possible
solution.
>>
>> Thanks
>> Anna-Belle
>>
>>
>>
>> -----Message d'origine-----
>> De : John Halley Gotway via RT [mailto:met_help at ucar.edu] Envoyé :
6
>> janvier 2014 16:13 À : Filion,Anna-Belle [Montreal] Objet : Re:
>> [rt.rap.ucar.edu #64753] Convex hull
>>
>> Anna-Belle,
>>
>> Can you tell me what version of MET you're using?  Is it METv4.1,
the most recent one?
>>
>> Since this is an enhancement and not a bugfix, we wouldn't release
it as a patch to METv4.1.  It would likely be distributed in the next
release, which has not yet been scheduled.
>>
>> That being said, I could take a look at the code to see how
difficult
>> it would be to dump out those points.  If it's easy, I could
probably add it and send you a patched file.  If it's more involved,
it'd have to wait until the next release.
>>
>> Thanks,
>> John
>>
>> On 01/06/2014 12:46 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>>>
>>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>>
>>> Hi John,
>>>
>>> I am interested un both. I would like to have access to the convex
hull points of cluster objects and individual simple objects.
>>>
>>> Do you knew how long it will take to implement this functionality
in the code?
>>>
>>> Thanks
>>> Anna-Belle
>>>
>>> -----Message d'origine-----
>>> De : John Halley Gotway via RT [mailto:met_help at ucar.edu] Envoyé :
6
>>> janvier 2014 13:54 À : Filion,Anna-Belle [Montreal] Objet : Re:
>>> [rt.rap.ucar.edu #64753] Convex hull
>>>
>>> Anna-Belle,
>>>
>>> The convex hull points are not currently included in the output of
MODE.  However, as you mention, the boundary points are in the NetCDF
output file.
>>>
>>> As for the convex hulls, really there are two options.  Convex
hulls
>>> are computed for the individual (or what we call "simple") objects
>>> in both fields.  Those convex hulls are used in the computation of
>>> fuzzy logic interest values.  Then matching/merging is performed,
which results in the groups of objects (or what we call "cluster"
objects) being defined.  The convex hulls shown in the postscript
output file are drawn around the matched clusters of objects.  If you
look closely, you'll see that there are no convex hulls drawn around
unmatched objects since they are not part of any matched cluster
object.
>>>
>>> Are you interested in seeing the convex hulls which are drawn
around the individual "simple" objects or the ones drawn around the
matched "cluster" objects?   Or both?
>>>
>>> Once we define this, I can create a development task to add this
functionality to the next version of MET.
>>>
>>> Thanks,
>>> John Halley Gotway
>>> met_help at ucar.edu
>>>
>>> On 12/30/2013 03:37 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>>>>
>>>> Mon Dec 30 15:37:29 2013: Request 64753 was acted upon.
>>>> Transaction: Ticket created by anna-belle.filion at ec.gc.ca
>>>>            Queue: met_help
>>>>          Subject: Convex hull
>>>>            Owner: Nobody
>>>>       Requestors: anna-belle.filion at ec.gc.ca
>>>>           Status: new
>>>>      Ticket <URL:
>>>> https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>>>
>>>>
>>>> Hi,
>>>>
>>>>
>>>>
>>>> I have been using MODE and MODE-Analysis for my forecast
verification.
>>>>
>>>>
>>>>
>>>> At this point, we are investigation a way to display the
forecast.
>>>> We want to display the raw gridded forecast, but also display
some
>>>> object information computed by MODE.
>>>>
>>>>
>>>>
>>>> So, we have a gridded forecast on which we use the MODE tool to
>>>> merge forecast objects that are close enough. What we would like
to
>>>> display is the raw forecast with the overlying convex hulls
>>>> computed by MODE. In the netCDF file I can find the points for
each
>>>> object boundary, but can't find the points of the convex hull.
>>>> Where can I find the points of each convex hull?
>>>>
>>>>
>>>>
>>>> Thanks
>>>>
>>>> Anna-Belle Filion
>>>>
>>>>
>>>>
>>>
>>>
>>
>>
>
>

------------------------------------------------
Subject: Re: [rt.rap.ucar.edu #64753] Convex hull
From: John Halley Gotway
Time: Fri Apr 11 10:10:15 2014

Anna-Belle,

OK, I understand your question better now.  You're referring to the
convex hulls that are shown on pages 6 and 8 of the PostScript file
you sent.

There are three methods available in MODE for merging objects in each
field: double-thresholding, fuzzy-engine, and allowing additional
merging during the matching process.
Pages 6 and 8 of the PostScript file you sent contain summary
information for the fuzzy-engine merging steps in the forecast and
observation fields, respectively.

In this step the objects within each field are compared to
*themselves*.  Obviously, each individual object should match itself
perfectly (with a total interest value of 1).  But objects that are
close to each other, may also match.  When multiple objects in one
field match the same object in the other field, they are merged
together.  So these cluster objects you see are created by comparing
the forecast to itself and the observation to itself.  The NetCDF and
ASCII output from MET only contain information about comparing the
forecast to the observation.  The output of the fuzzy engine
merging steps inform the merging that's done, but is otherwise not
written out.

But the solution here is simple.  If you'd like that convex hull
information from pages 6 and 8, just run MODE two more times, once
comparing the forecast to itself and once comparing the observation
to itself.  But I'm not sure how/why that information would be useful.

Hope that helps clarify.

Thanks,
John

On 04/10/2014 02:02 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>
> Hi John,
>
> Yes, I understand, but when you look at the .ps file there is a
convex hull for merged object even if they were not matched
>
> Is it possible to get the information on that convex hull?
>
> If you look at the file .ps that is included in this email you might
understand better what I am looking for. In this case, there is no
forecasted object that is matched with the observed object. But on the
page 7,  there is a convex hull contour for the merged forecast
object.
>
> Is there a way to easily modify the code to get this information?
>
> Thanks
> Anna-Belle
>
>
> -----Message d'origine-----
> De : John Halley Gotway via RT [mailto:met_help at ucar.edu]
> Envoyé : 10 April 2014 2:53
> À : Filion,Anna-Belle [Montreal]
> Objet : Re: [rt.rap.ucar.edu #64753] Convex hull
>
> Anna,
>
> By definition, a cluster object only exists when it is matched to
another cluster in the other field.  There is no such thing as an
unmatched cluster object.
>
> Suppose you've chosen the double-threshold or fuzzy-engine method of
merging objects in each field.  After that merging step, MODE has
merged groups of objects in each field.  During the matching step,
MODE determines which objects from the forecast field match objects in
the observation field.  After the matching step, if there are any
groups of objects in the forecast or observation fields whose members
are all unmatched, that group is thrown away.  That's why there are no
unmatched clusters.
>
> Does that make sense?
>
> Thanks,
> John
>
> On 04/10/2014 11:30 AM, Filion,Anna-Belle [Montreal] via RT wrote:
>>
>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>
>> Hi john,
>>
>> I tested this with a case where the forecast was matched with the
observation, so it was fine.
>>
>> However, if the forecast is not matched to any observation, the
forecast cluster convex hull is not available only the forecast simple
convex hull is.
>>
>> However, MODE still goes in the process of merging objects of the
same field to form bigger cluster object because we can see the
cluster convex hull in the .ps file.
>>
>> Is there a way that the cluster convex hull could be available even
though the forecast was not matched to the observation?
>>
>>
>> Thanks
>> Anna-Belle Filion
>>
>>
>>
>>
>> -----Message d'origine-----
>> De : John Halley Gotway via RT [mailto:met_help at ucar.edu] Envoyé :
9
>> janvier 2014 16:39 À : Filion,Anna-Belle [Montreal] Objet : Re:
>> [rt.rap.ucar.edu #64753] Convex hull
>>
>> Anne-Belle,
>>
>> I've attached a tar file that contains 4 updated files
(met_file.h/.cc and nc_utils.h/.cc).  You can place this in the top-
level METv4.1 directory and extract it:
>>      tar -xvzf METv4.1_patch_for_anna_belle.tar.gz
>>
>> And then recompile METv4.1.
>>
>> When you run MODE, you should notice many more variables in the
NetCDF object files.  I tweaked the naming convention of the existing
variables to be consistent with the new ones.  You'll now see data for
6 different polyline types:
>>     - Forecast and Observation Simple Boundary objects
(fcst_simp_bdy and obs_simp_bdy)
>>     - Forecast and Observation Simple Convex Hull objects
(fcst_simp_hull and obs_simp_hull)
>>     - Forecast and Observation Cluster Convex Hull objects
>> (fcst_clus_hull and obs_clus_hull)
>>
>> Please take a look and let me know if this is what you were looking
for.
>>
>> We'll plan to add these changes in to the next release of MET.
However, we want to make their output optional via the MODE
configuration file.
>>
>> Thanks,
>> John Halley Gotway
>>
>> On 01/06/2014 02:29 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>>>
>>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>>
>>> Hi John,
>>>
>>>
>>> Ok. Just let me know the solution you choose after analysing the
code and the approximate deadline of a patch if this is a possible
solution.
>>>
>>> Thanks
>>> Anna-Belle
>>>
>>>
>>>
>>> -----Message d'origine-----
>>> De : John Halley Gotway via RT [mailto:met_help at ucar.edu] Envoyé :
6
>>> janvier 2014 16:13 À : Filion,Anna-Belle [Montreal] Objet : Re:
>>> [rt.rap.ucar.edu #64753] Convex hull
>>>
>>> Anna-Belle,
>>>
>>> Can you tell me what version of MET you're using?  Is it METv4.1,
the most recent one?
>>>
>>> Since this is an enhancement and not a bugfix, we wouldn't release
it as a patch to METv4.1.  It would likely be distributed in the next
release, which has not yet been scheduled.
>>>
>>> That being said, I could take a look at the code to see how
difficult
>>> it would be to dump out those points.  If it's easy, I could
probably add it and send you a patched file.  If it's more involved,
it'd have to wait until the next release.
>>>
>>> Thanks,
>>> John
>>>
>>> On 01/06/2014 12:46 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>>>>
>>>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>>>
>>>> Hi John,
>>>>
>>>> I am interested un both. I would like to have access to the
convex hull points of cluster objects and individual simple objects.
>>>>
>>>> Do you knew how long it will take to implement this functionality
in the code?
>>>>
>>>> Thanks
>>>> Anna-Belle
>>>>
>>>> -----Message d'origine-----
>>>> De : John Halley Gotway via RT [mailto:met_help at ucar.edu] Envoyé
: 6
>>>> janvier 2014 13:54 À : Filion,Anna-Belle [Montreal] Objet : Re:
>>>> [rt.rap.ucar.edu #64753] Convex hull
>>>>
>>>> Anna-Belle,
>>>>
>>>> The convex hull points are not currently included in the output
of MODE.  However, as you mention, the boundary points are in the
NetCDF output file.
>>>>
>>>> As for the convex hulls, really there are two options.  Convex
hulls
>>>> are computed for the individual (or what we call "simple")
objects
>>>> in both fields.  Those convex hulls are used in the computation
of
>>>> fuzzy logic interest values.  Then matching/merging is performed,
which results in the groups of objects (or what we call "cluster"
objects) being defined.  The convex hulls shown in the postscript
output file are drawn around the matched clusters of objects.  If you
look closely, you'll see that there are no convex hulls drawn around
unmatched objects since they are not part of any matched cluster
object.
>>>>
>>>> Are you interested in seeing the convex hulls which are drawn
around the individual "simple" objects or the ones drawn around the
matched "cluster" objects?   Or both?
>>>>
>>>> Once we define this, I can create a development task to add this
functionality to the next version of MET.
>>>>
>>>> Thanks,
>>>> John Halley Gotway
>>>> met_help at ucar.edu
>>>>
>>>> On 12/30/2013 03:37 PM, Filion,Anna-Belle [Montreal] via RT
wrote:
>>>>>
>>>>> Mon Dec 30 15:37:29 2013: Request 64753 was acted upon.
>>>>> Transaction: Ticket created by anna-belle.filion at ec.gc.ca
>>>>>             Queue: met_help
>>>>>           Subject: Convex hull
>>>>>             Owner: Nobody
>>>>>        Requestors: anna-belle.filion at ec.gc.ca
>>>>>            Status: new
>>>>>       Ticket <URL:
>>>>> https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>>>>
>>>>>
>>>>> Hi,
>>>>>
>>>>>
>>>>>
>>>>> I have been using MODE and MODE-Analysis for my forecast
verification.
>>>>>
>>>>>
>>>>>
>>>>> At this point, we are investigation a way to display the
forecast.
>>>>> We want to display the raw gridded forecast, but also display
some
>>>>> object information computed by MODE.
>>>>>
>>>>>
>>>>>
>>>>> So, we have a gridded forecast on which we use the MODE tool to
>>>>> merge forecast objects that are close enough. What we would like
to
>>>>> display is the raw forecast with the overlying convex hulls
>>>>> computed by MODE. In the netCDF file I can find the points for
each
>>>>> object boundary, but can't find the points of the convex hull.
>>>>> Where can I find the points of each convex hull?
>>>>>
>>>>>
>>>>>
>>>>> Thanks
>>>>>
>>>>> Anna-Belle Filion
>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>
>>>
>>
>>
>
>

------------------------------------------------
Subject: Convex hull
From: Filion,Anna-Belle [Montreal]
Time: Fri Apr 11 11:11:02 2014

Hi John,

The reason that we need that information is that we will produce a
severe thunderstorm intensity index forecast this summer, but we want
to use the object approach in our product.

Have a look to the file produit.gif. So, we post-processed the fields
of a high-resolution model to produce a forecast and then we use mode
to merge convective cells that are close to each other and should
belong to the same systems/phenomena and produce the convex hull. So,
the forecast product is in a way pre-analyze and the convex hull gives
usually a good idea of the region that is expected to be affected by
severe convection.

Thanks
Anna-Belle

-----Message d'origine-----
De : John Halley Gotway via RT [mailto:met_help at ucar.edu]
Envoyé : 11 avril 2014 12:10
À : Filion,Anna-Belle [Montreal]
Objet : Re: [rt.rap.ucar.edu #64753] Convex hull

Anna-Belle,

OK, I understand your question better now.  You're referring to the
convex hulls that are shown on pages 6 and 8 of the PostScript file
you sent.

There are three methods available in MODE for merging objects in each
field: double-thresholding, fuzzy-engine, and allowing additional
merging during the matching process.
Pages 6 and 8 of the PostScript file you sent contain summary
information for the fuzzy-engine merging steps in the forecast and
observation fields, respectively.

In this step the objects within each field are compared to
*themselves*.  Obviously, each individual object should match itself
perfectly (with a total interest value of 1).  But objects that are
close to each other, may also match.  When multiple objects in one
field match the same object in the other field, they are merged
together.  So these cluster objects you see are created by comparing
the forecast to itself and the observation to itself.  The NetCDF and
ASCII output from MET only contain information about comparing the
forecast to the observation.  The output of the fuzzy engine
merging steps inform the merging that's done, but is otherwise not
written out.

But the solution here is simple.  If you'd like that convex hull
information from pages 6 and 8, just run MODE two more times, once
comparing the forecast to itself and once comparing the observation
to itself.  But I'm not sure how/why that information would be useful.

Hope that helps clarify.

Thanks,
John

On 04/10/2014 02:02 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>
> Hi John,
>
> Yes, I understand, but when you look at the .ps file there is a
convex hull for merged object even if they were not matched
>
> Is it possible to get the information on that convex hull?
>
> If you look at the file .ps that is included in this email you might
understand better what I am looking for. In this case, there is no
forecasted object that is matched with the observed object. But on the
page 7,  there is a convex hull contour for the merged forecast
object.
>
> Is there a way to easily modify the code to get this information?
>
> Thanks
> Anna-Belle
>
>
> -----Message d'origine-----
> De : John Halley Gotway via RT [mailto:met_help at ucar.edu]
> Envoyé : 10 April 2014 2:53
> À : Filion,Anna-Belle [Montreal]
> Objet : Re: [rt.rap.ucar.edu #64753] Convex hull
>
> Anna,
>
> By definition, a cluster object only exists when it is matched to
another cluster in the other field.  There is no such thing as an
unmatched cluster object.
>
> Suppose you've chosen the double-threshold or fuzzy-engine method of
merging objects in each field.  After that merging step, MODE has
merged groups of objects in each field.  During the matching step,
MODE determines which objects from the forecast field match objects in
the observation field.  After the matching step, if there are any
groups of objects in the forecast or observation fields whose members
are all unmatched, that group is thrown away.  That's why there are no
unmatched clusters.
>
> Does that make sense?
>
> Thanks,
> John
>
> On 04/10/2014 11:30 AM, Filion,Anna-Belle [Montreal] via RT wrote:
>>
>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>
>> Hi john,
>>
>> I tested this with a case where the forecast was matched with the
observation, so it was fine.
>>
>> However, if the forecast is not matched to any observation, the
forecast cluster convex hull is not available only the forecast simple
convex hull is.
>>
>> However, MODE still goes in the process of merging objects of the
same field to form bigger cluster object because we can see the
cluster convex hull in the .ps file.
>>
>> Is there a way that the cluster convex hull could be available even
though the forecast was not matched to the observation?
>>
>>
>> Thanks
>> Anna-Belle Filion
>>
>>
>>
>>
>> -----Message d'origine-----
>> De : John Halley Gotway via RT [mailto:met_help at ucar.edu] Envoyé :
9
>> janvier 2014 16:39 À : Filion,Anna-Belle [Montreal] Objet : Re:
>> [rt.rap.ucar.edu #64753] Convex hull
>>
>> Anne-Belle,
>>
>> I've attached a tar file that contains 4 updated files
(met_file.h/.cc and nc_utils.h/.cc).  You can place this in the top-
level METv4.1 directory and extract it:
>>      tar -xvzf METv4.1_patch_for_anna_belle.tar.gz
>>
>> And then recompile METv4.1.
>>
>> When you run MODE, you should notice many more variables in the
NetCDF object files.  I tweaked the naming convention of the existing
variables to be consistent with the new ones.  You'll now see data for
6 different polyline types:
>>     - Forecast and Observation Simple Boundary objects
(fcst_simp_bdy and obs_simp_bdy)
>>     - Forecast and Observation Simple Convex Hull objects
(fcst_simp_hull and obs_simp_hull)
>>     - Forecast and Observation Cluster Convex Hull objects
>> (fcst_clus_hull and obs_clus_hull)
>>
>> Please take a look and let me know if this is what you were looking
for.
>>
>> We'll plan to add these changes in to the next release of MET.
However, we want to make their output optional via the MODE
configuration file.
>>
>> Thanks,
>> John Halley Gotway
>>
>> On 01/06/2014 02:29 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>>>
>>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>>
>>> Hi John,
>>>
>>>
>>> Ok. Just let me know the solution you choose after analysing the
code and the approximate deadline of a patch if this is a possible
solution.
>>>
>>> Thanks
>>> Anna-Belle
>>>
>>>
>>>
>>> -----Message d'origine-----
>>> De : John Halley Gotway via RT [mailto:met_help at ucar.edu] Envoyé :
6
>>> janvier 2014 16:13 À : Filion,Anna-Belle [Montreal] Objet : Re:
>>> [rt.rap.ucar.edu #64753] Convex hull
>>>
>>> Anna-Belle,
>>>
>>> Can you tell me what version of MET you're using?  Is it METv4.1,
the most recent one?
>>>
>>> Since this is an enhancement and not a bugfix, we wouldn't release
it as a patch to METv4.1.  It would likely be distributed in the next
release, which has not yet been scheduled.
>>>
>>> That being said, I could take a look at the code to see how
difficult
>>> it would be to dump out those points.  If it's easy, I could
probably add it and send you a patched file.  If it's more involved,
it'd have to wait until the next release.
>>>
>>> Thanks,
>>> John
>>>
>>> On 01/06/2014 12:46 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>>>>
>>>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>>>
>>>> Hi John,
>>>>
>>>> I am interested un both. I would like to have access to the
convex hull points of cluster objects and individual simple objects.
>>>>
>>>> Do you knew how long it will take to implement this functionality
in the code?
>>>>
>>>> Thanks
>>>> Anna-Belle
>>>>
>>>> -----Message d'origine-----
>>>> De : John Halley Gotway via RT [mailto:met_help at ucar.edu] Envoyé
: 6
>>>> janvier 2014 13:54 À : Filion,Anna-Belle [Montreal] Objet : Re:
>>>> [rt.rap.ucar.edu #64753] Convex hull
>>>>
>>>> Anna-Belle,
>>>>
>>>> The convex hull points are not currently included in the output
of MODE.  However, as you mention, the boundary points are in the
NetCDF output file.
>>>>
>>>> As for the convex hulls, really there are two options.  Convex
hulls
>>>> are computed for the individual (or what we call "simple")
objects
>>>> in both fields.  Those convex hulls are used in the computation
of
>>>> fuzzy logic interest values.  Then matching/merging is performed,
which results in the groups of objects (or what we call "cluster"
objects) being defined.  The convex hulls shown in the postscript
output file are drawn around the matched clusters of objects.  If you
look closely, you'll see that there are no convex hulls drawn around
unmatched objects since they are not part of any matched cluster
object.
>>>>
>>>> Are you interested in seeing the convex hulls which are drawn
around the individual "simple" objects or the ones drawn around the
matched "cluster" objects?   Or both?
>>>>
>>>> Once we define this, I can create a development task to add this
functionality to the next version of MET.
>>>>
>>>> Thanks,
>>>> John Halley Gotway
>>>> met_help at ucar.edu
>>>>
>>>> On 12/30/2013 03:37 PM, Filion,Anna-Belle [Montreal] via RT
wrote:
>>>>>
>>>>> Mon Dec 30 15:37:29 2013: Request 64753 was acted upon.
>>>>> Transaction: Ticket created by anna-belle.filion at ec.gc.ca
>>>>>             Queue: met_help
>>>>>           Subject: Convex hull
>>>>>             Owner: Nobody
>>>>>        Requestors: anna-belle.filion at ec.gc.ca
>>>>>            Status: new
>>>>>       Ticket <URL:
>>>>> https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>>>>
>>>>>
>>>>> Hi,
>>>>>
>>>>>
>>>>>
>>>>> I have been using MODE and MODE-Analysis for my forecast
verification.
>>>>>
>>>>>
>>>>>
>>>>> At this point, we are investigation a way to display the
forecast.
>>>>> We want to display the raw gridded forecast, but also display
some
>>>>> object information computed by MODE.
>>>>>
>>>>>
>>>>>
>>>>> So, we have a gridded forecast on which we use the MODE tool to
>>>>> merge forecast objects that are close enough. What we would like
to
>>>>> display is the raw forecast with the overlying convex hulls
>>>>> computed by MODE. In the netCDF file I can find the points for
each
>>>>> object boundary, but can't find the points of the convex hull.
>>>>> Where can I find the points of each convex hull?
>>>>>
>>>>>
>>>>>
>>>>> Thanks
>>>>>
>>>>> Anna-Belle Filion
>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>
>>>
>>
>>
>
>

------------------------------------------------
Subject: Re: [rt.rap.ucar.edu #64753] Convex hull
From: John Halley Gotway
Time: Fri Apr 11 11:35:01 2014

Anna-Belle,

That sounds interesting - that's a novel way of using MODE.

Since you're running MODE on a forecast product, presumably you're
running it before observations are even available.  Wouldn't it work
to just run MODE to compare the forecast to itself?  Or is that
a problem because you'd end with too many convex hulls - since all the
objects match themselves?

I suspect that adding convex hull information for the groups of
objects prior to the matching step would just confuse most users.
Since this is a pretty specialized application of MODE, I don't think
that information would be generally useful.  That being said, the MET
software package is open source, and you're free to modify it in any
way you'd like.

Unless I'm missing something, I think running MODE to compare the
forecast to itself would be a good way to go.  If you want to skip
plotting convex hulls for cases where the object matches only
itself, you could either modify the MODE source code to do that, or

Thanks,
John

On 04/11/2014 11:11 AM, Filion,Anna-Belle [Montreal] via RT wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>
> Hi John,
>
> The reason that we need that information is that we will produce a
severe thunderstorm intensity index forecast this summer, but we want
to use the object approach in our product.
>
> Have a look to the file produit.gif. So, we post-processed the
fields of a high-resolution model to produce a forecast and then we
use mode to merge convective cells that are close to each other and
should belong to the same systems/phenomena and produce the convex
hull. So, the forecast product is in a way pre-analyze and the convex
hull gives usually a good idea of the region that is expected to be
affected by severe convection.
>
> Thanks
> Anna-Belle
>
>
>
> -----Message d'origine-----
> De : John Halley Gotway via RT [mailto:met_help at ucar.edu]
> Envoyé : 11 avril 2014 12:10
> À : Filion,Anna-Belle [Montreal]
> Objet : Re: [rt.rap.ucar.edu #64753] Convex hull
>
> Anna-Belle,
>
> OK, I understand your question better now.  You're referring to the
convex hulls that are shown on pages 6 and 8 of the PostScript file
you sent.
>
> There are three methods available in MODE for merging objects in
each field: double-thresholding, fuzzy-engine, and allowing additional
merging during the matching process.
> Pages 6 and 8 of the PostScript file you sent contain summary
information for the fuzzy-engine merging steps in the forecast and
observation fields, respectively.
>
> In this step the objects within each field are compared to
*themselves*.  Obviously, each individual object should match itself
perfectly (with a total interest value of 1).  But objects that are
> close to each other, may also match.  When multiple objects in one
field match the same object in the other field, they are merged
together.  So these cluster objects you see are created by comparing
> the forecast to itself and the observation to itself.  The NetCDF
and ASCII output from MET only contain information about comparing the
forecast to the observation.  The output of the fuzzy engine
> merging steps inform the merging that's done, but is otherwise not
written out.
>
> But the solution here is simple.  If you'd like that convex hull
information from pages 6 and 8, just run MODE two more times, once
comparing the forecast to itself and once comparing the observation
> to itself.  But I'm not sure how/why that information would be
useful.
>
> Hope that helps clarify.
>
> Thanks,
> John
>
>
> On 04/10/2014 02:02 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>>
>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>
>> Hi John,
>>
>> Yes, I understand, but when you look at the .ps file there is a
convex hull for merged object even if they were not matched
>>
>> Is it possible to get the information on that convex hull?
>>
>> If you look at the file .ps that is included in this email you
might understand better what I am looking for. In this case, there is
no forecasted object that is matched with the observed object. But on
the page 7,  there is a convex hull contour for the merged forecast
object.
>>
>> Is there a way to easily modify the code to get this information?
>>
>> Thanks
>> Anna-Belle
>>
>>
>> -----Message d'origine-----
>> De : John Halley Gotway via RT [mailto:met_help at ucar.edu]
>> Envoyé : 10 April 2014 2:53
>> À : Filion,Anna-Belle [Montreal]
>> Objet : Re: [rt.rap.ucar.edu #64753] Convex hull
>>
>> Anna,
>>
>> By definition, a cluster object only exists when it is matched to
another cluster in the other field.  There is no such thing as an
unmatched cluster object.
>>
>> Suppose you've chosen the double-threshold or fuzzy-engine method
of merging objects in each field.  After that merging step, MODE has
merged groups of objects in each field.  During the matching step,
MODE determines which objects from the forecast field match objects in
the observation field.  After the matching step, if there are any
groups of objects in the forecast or observation fields whose members
are all unmatched, that group is thrown away.  That's why there are no
unmatched clusters.
>>
>> Does that make sense?
>>
>> Thanks,
>> John
>>
>> On 04/10/2014 11:30 AM, Filion,Anna-Belle [Montreal] via RT wrote:
>>>
>>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>>
>>> Hi john,
>>>
>>> I tested this with a case where the forecast was matched with the
observation, so it was fine.
>>>
>>> However, if the forecast is not matched to any observation, the
forecast cluster convex hull is not available only the forecast simple
convex hull is.
>>>
>>> However, MODE still goes in the process of merging objects of the
same field to form bigger cluster object because we can see the
cluster convex hull in the .ps file.
>>>
>>> Is there a way that the cluster convex hull could be available
even though the forecast was not matched to the observation?
>>>
>>>
>>> Thanks
>>> Anna-Belle Filion
>>>
>>>
>>>
>>>
>>> -----Message d'origine-----
>>> De : John Halley Gotway via RT [mailto:met_help at ucar.edu] Envoyé :
9
>>> janvier 2014 16:39 À : Filion,Anna-Belle [Montreal] Objet : Re:
>>> [rt.rap.ucar.edu #64753] Convex hull
>>>
>>> Anne-Belle,
>>>
>>> I've attached a tar file that contains 4 updated files
(met_file.h/.cc and nc_utils.h/.cc).  You can place this in the top-
level METv4.1 directory and extract it:
>>>       tar -xvzf METv4.1_patch_for_anna_belle.tar.gz
>>>
>>> And then recompile METv4.1.
>>>
>>> When you run MODE, you should notice many more variables in the
NetCDF object files.  I tweaked the naming convention of the existing
variables to be consistent with the new ones.  You'll now see data for
6 different polyline types:
>>>      - Forecast and Observation Simple Boundary objects
(fcst_simp_bdy and obs_simp_bdy)
>>>      - Forecast and Observation Simple Convex Hull objects
(fcst_simp_hull and obs_simp_hull)
>>>      - Forecast and Observation Cluster Convex Hull objects
>>> (fcst_clus_hull and obs_clus_hull)
>>>
>>> Please take a look and let me know if this is what you were
looking for.
>>>
>>> We'll plan to add these changes in to the next release of MET.
However, we want to make their output optional via the MODE
configuration file.
>>>
>>> Thanks,
>>> John Halley Gotway
>>>
>>> On 01/06/2014 02:29 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>>>>
>>>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>>>
>>>> Hi John,
>>>>
>>>>
>>>> Ok. Just let me know the solution you choose after analysing the
code and the approximate deadline of a patch if this is a possible
solution.
>>>>
>>>> Thanks
>>>> Anna-Belle
>>>>
>>>>
>>>>
>>>> -----Message d'origine-----
>>>> De : John Halley Gotway via RT [mailto:met_help at ucar.edu] Envoyé
: 6
>>>> janvier 2014 16:13 À : Filion,Anna-Belle [Montreal] Objet : Re:
>>>> [rt.rap.ucar.edu #64753] Convex hull
>>>>
>>>> Anna-Belle,
>>>>
>>>> Can you tell me what version of MET you're using?  Is it METv4.1,
the most recent one?
>>>>
>>>> Since this is an enhancement and not a bugfix, we wouldn't
release it as a patch to METv4.1.  It would likely be distributed in
the next release, which has not yet been scheduled.
>>>>
>>>> That being said, I could take a look at the code to see how
difficult
>>>> it would be to dump out those points.  If it's easy, I could
probably add it and send you a patched file.  If it's more involved,
it'd have to wait until the next release.
>>>>
>>>> Thanks,
>>>> John
>>>>
>>>> On 01/06/2014 12:46 PM, Filion,Anna-Belle [Montreal] via RT
wrote:
>>>>>
>>>>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>>>>
>>>>> Hi John,
>>>>>
>>>>> I am interested un both. I would like to have access to the
convex hull points of cluster objects and individual simple objects.
>>>>>
>>>>> Do you knew how long it will take to implement this
functionality in the code?
>>>>>
>>>>> Thanks
>>>>> Anna-Belle
>>>>>
>>>>> -----Message d'origine-----
>>>>> De : John Halley Gotway via RT [mailto:met_help at ucar.edu] Envoyé
: 6
>>>>> janvier 2014 13:54 À : Filion,Anna-Belle [Montreal] Objet : Re:
>>>>> [rt.rap.ucar.edu #64753] Convex hull
>>>>>
>>>>> Anna-Belle,
>>>>>
>>>>> The convex hull points are not currently included in the output
of MODE.  However, as you mention, the boundary points are in the
NetCDF output file.
>>>>>
>>>>> As for the convex hulls, really there are two options.  Convex
hulls
>>>>> are computed for the individual (or what we call "simple")
objects
>>>>> in both fields.  Those convex hulls are used in the computation
of
>>>>> fuzzy logic interest values.  Then matching/merging is
performed, which results in the groups of objects (or what we call
"cluster" objects) being defined.  The convex hulls shown in the
postscript output file are drawn around the matched clusters of
objects.  If you look closely, you'll see that there are no convex
hulls drawn around unmatched objects since they are not part of any
matched cluster object.
>>>>>
>>>>> Are you interested in seeing the convex hulls which are drawn
around the individual "simple" objects or the ones drawn around the
matched "cluster" objects?   Or both?
>>>>>
>>>>> Once we define this, I can create a development task to add this
functionality to the next version of MET.
>>>>>
>>>>> Thanks,
>>>>> John Halley Gotway
>>>>> met_help at ucar.edu
>>>>>
>>>>> On 12/30/2013 03:37 PM, Filion,Anna-Belle [Montreal] via RT
wrote:
>>>>>>
>>>>>> Mon Dec 30 15:37:29 2013: Request 64753 was acted upon.
>>>>>> Transaction: Ticket created by anna-belle.filion at ec.gc.ca
>>>>>>              Queue: met_help
>>>>>>            Subject: Convex hull
>>>>>>              Owner: Nobody
>>>>>>         Requestors: anna-belle.filion at ec.gc.ca
>>>>>>             Status: new
>>>>>>        Ticket <URL:
>>>>>> https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>>>>>
>>>>>>
>>>>>> Hi,
>>>>>>
>>>>>>
>>>>>>
>>>>>> I have been using MODE and MODE-Analysis for my forecast
verification.
>>>>>>
>>>>>>
>>>>>>
>>>>>> At this point, we are investigation a way to display the
forecast.
>>>>>> We want to display the raw gridded forecast, but also display
some
>>>>>> object information computed by MODE.
>>>>>>
>>>>>>
>>>>>>
>>>>>> So, we have a gridded forecast on which we use the MODE tool to
>>>>>> merge forecast objects that are close enough. What we would
like to
>>>>>> display is the raw forecast with the overlying convex hulls
>>>>>> computed by MODE. In the netCDF file I can find the points for
each
>>>>>> object boundary, but can't find the points of the convex hull.
>>>>>> Where can I find the points of each convex hull?
>>>>>>
>>>>>>
>>>>>>
>>>>>> Thanks
>>>>>>
>>>>>> Anna-Belle Filion
>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>
>>>
>>
>>
>
>

------------------------------------------------
Subject: RE: [rt.rap.ucar.edu #64753] Convex hull
From: Filion,Anna-Belle [Montreal]
Time: Fri Apr 11 12:18:44 2014

Hi John,

Thanks a lot. This works. I sorry to have not think about that myself
and wasted some of your time!

Anna-Belle

-----Message d'origine-----
De : John Halley Gotway via RT [mailto:met_help at ucar.edu]
Envoyé : 11 avril 2014 13:35
À : Filion,Anna-Belle [Montreal]
Objet : Re: [rt.rap.ucar.edu #64753] Convex hull

Anna-Belle,

That sounds interesting - that's a novel way of using MODE.

Since you're running MODE on a forecast product, presumably you're
running it before observations are even available.  Wouldn't it work
to just run MODE to compare the forecast to itself?  Or is that
a problem because you'd end with too many convex hulls - since all the
objects match themselves?

I suspect that adding convex hull information for the groups of
objects prior to the matching step would just confuse most users.
Since this is a pretty specialized application of MODE, I don't think
that information would be generally useful.  That being said, the MET
software package is open source, and you're free to modify it in any
way you'd like.

Unless I'm missing something, I think running MODE to compare the
forecast to itself would be a good way to go.  If you want to skip
plotting convex hulls for cases where the object matches only
itself, you could either modify the MODE source code to do that, or

Thanks,
John

On 04/11/2014 11:11 AM, Filion,Anna-Belle [Montreal] via RT wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>
> Hi John,
>
> The reason that we need that information is that we will produce a
severe thunderstorm intensity index forecast this summer, but we want
to use the object approach in our product.
>
> Have a look to the file produit.gif. So, we post-processed the
fields of a high-resolution model to produce a forecast and then we
use mode to merge convective cells that are close to each other and
should belong to the same systems/phenomena and produce the convex
hull. So, the forecast product is in a way pre-analyze and the convex
hull gives usually a good idea of the region that is expected to be
affected by severe convection.
>
> Thanks
> Anna-Belle
>
>
>
> -----Message d'origine-----
> De : John Halley Gotway via RT [mailto:met_help at ucar.edu]
> Envoyé : 11 avril 2014 12:10
> À : Filion,Anna-Belle [Montreal]
> Objet : Re: [rt.rap.ucar.edu #64753] Convex hull
>
> Anna-Belle,
>
> OK, I understand your question better now.  You're referring to the
convex hulls that are shown on pages 6 and 8 of the PostScript file
you sent.
>
> There are three methods available in MODE for merging objects in
each field: double-thresholding, fuzzy-engine, and allowing additional
merging during the matching process.
> Pages 6 and 8 of the PostScript file you sent contain summary
information for the fuzzy-engine merging steps in the forecast and
observation fields, respectively.
>
> In this step the objects within each field are compared to
*themselves*.  Obviously, each individual object should match itself
perfectly (with a total interest value of 1).  But objects that are
> close to each other, may also match.  When multiple objects in one
field match the same object in the other field, they are merged
together.  So these cluster objects you see are created by comparing
> the forecast to itself and the observation to itself.  The NetCDF
and ASCII output from MET only contain information about comparing the
forecast to the observation.  The output of the fuzzy engine
> merging steps inform the merging that's done, but is otherwise not
written out.
>
> But the solution here is simple.  If you'd like that convex hull
information from pages 6 and 8, just run MODE two more times, once
comparing the forecast to itself and once comparing the observation
> to itself.  But I'm not sure how/why that information would be
useful.
>
> Hope that helps clarify.
>
> Thanks,
> John
>
>
> On 04/10/2014 02:02 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>>
>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>
>> Hi John,
>>
>> Yes, I understand, but when you look at the .ps file there is a
convex hull for merged object even if they were not matched
>>
>> Is it possible to get the information on that convex hull?
>>
>> If you look at the file .ps that is included in this email you
might understand better what I am looking for. In this case, there is
no forecasted object that is matched with the observed object. But on
the page 7,  there is a convex hull contour for the merged forecast
object.
>>
>> Is there a way to easily modify the code to get this information?
>>
>> Thanks
>> Anna-Belle
>>
>>
>> -----Message d'origine-----
>> De : John Halley Gotway via RT [mailto:met_help at ucar.edu]
>> Envoyé : 10 April 2014 2:53
>> À : Filion,Anna-Belle [Montreal]
>> Objet : Re: [rt.rap.ucar.edu #64753] Convex hull
>>
>> Anna,
>>
>> By definition, a cluster object only exists when it is matched to
another cluster in the other field.  There is no such thing as an
unmatched cluster object.
>>
>> Suppose you've chosen the double-threshold or fuzzy-engine method
of merging objects in each field.  After that merging step, MODE has
merged groups of objects in each field.  During the matching step,
MODE determines which objects from the forecast field match objects in
the observation field.  After the matching step, if there are any
groups of objects in the forecast or observation fields whose members
are all unmatched, that group is thrown away.  That's why there are no
unmatched clusters.
>>
>> Does that make sense?
>>
>> Thanks,
>> John
>>
>> On 04/10/2014 11:30 AM, Filion,Anna-Belle [Montreal] via RT wrote:
>>>
>>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>>
>>> Hi john,
>>>
>>> I tested this with a case where the forecast was matched with the
observation, so it was fine.
>>>
>>> However, if the forecast is not matched to any observation, the
forecast cluster convex hull is not available only the forecast simple
convex hull is.
>>>
>>> However, MODE still goes in the process of merging objects of the
same field to form bigger cluster object because we can see the
cluster convex hull in the .ps file.
>>>
>>> Is there a way that the cluster convex hull could be available
even though the forecast was not matched to the observation?
>>>
>>>
>>> Thanks
>>> Anna-Belle Filion
>>>
>>>
>>>
>>>
>>> -----Message d'origine-----
>>> De : John Halley Gotway via RT [mailto:met_help at ucar.edu] Envoyé :
9
>>> janvier 2014 16:39 À : Filion,Anna-Belle [Montreal] Objet : Re:
>>> [rt.rap.ucar.edu #64753] Convex hull
>>>
>>> Anne-Belle,
>>>
>>> I've attached a tar file that contains 4 updated files
(met_file.h/.cc and nc_utils.h/.cc).  You can place this in the top-
level METv4.1 directory and extract it:
>>>       tar -xvzf METv4.1_patch_for_anna_belle.tar.gz
>>>
>>> And then recompile METv4.1.
>>>
>>> When you run MODE, you should notice many more variables in the
NetCDF object files.  I tweaked the naming convention of the existing
variables to be consistent with the new ones.  You'll now see data for
6 different polyline types:
>>>      - Forecast and Observation Simple Boundary objects
(fcst_simp_bdy and obs_simp_bdy)
>>>      - Forecast and Observation Simple Convex Hull objects
(fcst_simp_hull and obs_simp_hull)
>>>      - Forecast and Observation Cluster Convex Hull objects
>>> (fcst_clus_hull and obs_clus_hull)
>>>
>>> Please take a look and let me know if this is what you were
looking for.
>>>
>>> We'll plan to add these changes in to the next release of MET.
However, we want to make their output optional via the MODE
configuration file.
>>>
>>> Thanks,
>>> John Halley Gotway
>>>
>>> On 01/06/2014 02:29 PM, Filion,Anna-Belle [Montreal] via RT wrote:
>>>>
>>>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>>>
>>>> Hi John,
>>>>
>>>>
>>>> Ok. Just let me know the solution you choose after analysing the
code and the approximate deadline of a patch if this is a possible
solution.
>>>>
>>>> Thanks
>>>> Anna-Belle
>>>>
>>>>
>>>>
>>>> -----Message d'origine-----
>>>> De : John Halley Gotway via RT [mailto:met_help at ucar.edu] Envoyé
: 6
>>>> janvier 2014 16:13 À : Filion,Anna-Belle [Montreal] Objet : Re:
>>>> [rt.rap.ucar.edu #64753] Convex hull
>>>>
>>>> Anna-Belle,
>>>>
>>>> Can you tell me what version of MET you're using?  Is it METv4.1,
the most recent one?
>>>>
>>>> Since this is an enhancement and not a bugfix, we wouldn't
release it as a patch to METv4.1.  It would likely be distributed in
the next release, which has not yet been scheduled.
>>>>
>>>> That being said, I could take a look at the code to see how
difficult
>>>> it would be to dump out those points.  If it's easy, I could
probably add it and send you a patched file.  If it's more involved,
it'd have to wait until the next release.
>>>>
>>>> Thanks,
>>>> John
>>>>
>>>> On 01/06/2014 12:46 PM, Filion,Anna-Belle [Montreal] via RT
wrote:
>>>>>
>>>>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>>>>
>>>>> Hi John,
>>>>>
>>>>> I am interested un both. I would like to have access to the
convex hull points of cluster objects and individual simple objects.
>>>>>
>>>>> Do you knew how long it will take to implement this
functionality in the code?
>>>>>
>>>>> Thanks
>>>>> Anna-Belle
>>>>>
>>>>> -----Message d'origine-----
>>>>> De : John Halley Gotway via RT [mailto:met_help at ucar.edu] Envoyé
: 6
>>>>> janvier 2014 13:54 À : Filion,Anna-Belle [Montreal] Objet : Re:
>>>>> [rt.rap.ucar.edu #64753] Convex hull
>>>>>
>>>>> Anna-Belle,
>>>>>
>>>>> The convex hull points are not currently included in the output
of MODE.  However, as you mention, the boundary points are in the
NetCDF output file.
>>>>>
>>>>> As for the convex hulls, really there are two options.  Convex
hulls
>>>>> are computed for the individual (or what we call "simple")
objects
>>>>> in both fields.  Those convex hulls are used in the computation
of
>>>>> fuzzy logic interest values.  Then matching/merging is
performed, which results in the groups of objects (or what we call
"cluster" objects) being defined.  The convex hulls shown in the
postscript output file are drawn around the matched clusters of
objects.  If you look closely, you'll see that there are no convex
hulls drawn around unmatched objects since they are not part of any
matched cluster object.
>>>>>
>>>>> Are you interested in seeing the convex hulls which are drawn
around the individual "simple" objects or the ones drawn around the
matched "cluster" objects?   Or both?
>>>>>
>>>>> Once we define this, I can create a development task to add this
functionality to the next version of MET.
>>>>>
>>>>> Thanks,
>>>>> John Halley Gotway
>>>>> met_help at ucar.edu
>>>>>
>>>>> On 12/30/2013 03:37 PM, Filion,Anna-Belle [Montreal] via RT
wrote:
>>>>>>
>>>>>> Mon Dec 30 15:37:29 2013: Request 64753 was acted upon.
>>>>>> Transaction: Ticket created by anna-belle.filion at ec.gc.ca
>>>>>>              Queue: met_help
>>>>>>            Subject: Convex hull
>>>>>>              Owner: Nobody
>>>>>>         Requestors: anna-belle.filion at ec.gc.ca
>>>>>>             Status: new
>>>>>>        Ticket <URL:
>>>>>> https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=64753 >
>>>>>>
>>>>>>
>>>>>> Hi,
>>>>>>
>>>>>>
>>>>>>
>>>>>> I have been using MODE and MODE-Analysis for my forecast
verification.
>>>>>>
>>>>>>
>>>>>>
>>>>>> At this point, we are investigation a way to display the
forecast.
>>>>>> We want to display the raw gridded forecast, but also display
some
>>>>>> object information computed by MODE.
>>>>>>
>>>>>>
>>>>>>
>>>>>> So, we have a gridded forecast on which we use the MODE tool to
>>>>>> merge forecast objects that are close enough. What we would
like to
>>>>>> display is the raw forecast with the overlying convex hulls
>>>>>> computed by MODE. In the netCDF file I can find the points for
each
>>>>>> object boundary, but can't find the points of the convex hull.
>>>>>> Where can I find the points of each convex hull?
>>>>>>
>>>>>>
>>>>>>
>>>>>> Thanks
>>>>>>
>>>>>> Anna-Belle Filion
>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>
>>>
>>
>>
>
>

------------------------------------------------
```