[Met_help] [rt.rap.ucar.edu #88022] History for Question about MTD convolution

Tue Jul 9 12:07:13 MDT 2019

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

Hello,

I am aware that Mode Time-Domain (MTD) automatically applies both 
spatial and temporal convolution. Users are able to control the amount 
of spatial convolution through the conv_radius parameter in the MTD 
namelist. However, temporal convolution appears to automatically be 
applied over +/- 1 time-step according to the documentation here: 
https://dtcenter.org/met/users/support/OnLinePractical/OnLinePractical_218/mtd/run.php. 

For my work, I am interested in very small objects such that I am not 
applying any spatial convolution (conv_radius = 0). I would also like to 
see how the results differed if I did not apply any spatial convolution 
as well. Is there a way to configure the temporal convolution through 
the MTD namelist? If not, is there a way to disable that function 
altogether?

Thanks,
Samuel Degelia

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

Subject: Question about MTD convolution
From: John Halley Gotway
Time: Wed Dec 05 10:02:10 2018

Samuel,

Unfortunately no, the amount of time smoothing is currently had-coded
as
+/- 1 timestep.  There is currently no way in the configuration file
to
change this setting.  However, I used your request to create a
development
ticket for a future version of MET.  The MET software is open source.
For
now, you'd need to modify the code and recompile to change the
behavior.

Sorry I don't have a better answer for you.

Thanks,
John Halley Gotway

On Tue, Dec 4, 2018 at 12:09 PM Samuel Degelia via RT
<met_help at ucar.edu>
wrote:

>
> Tue Dec 04 12:09:02 2018: Request 88022 was acted upon.
> Transaction: Ticket created by sdegelia at gmail.com
>        Queue: met_help
>      Subject: Question about MTD convolution
>        Owner: Nobody
>   Requestors: sdegelia at gmail.com
>       Status: new
>  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=88022 >
>
>
> Hello,
>
> I am aware that Mode Time-Domain (MTD) automatically applies both
> spatial and temporal convolution. Users are able to control the
amount
> of spatial convolution through the conv_radius parameter in the MTD
> namelist. However, temporal convolution appears to automatically be
> applied over +/- 1 time-step according to the documentation here:
>
>
https://dtcenter.org/met/users/support/OnLinePractical/OnLinePractical_218/mtd/run.php.
>
>
>
> For my work, I am interested in very small objects such that I am
not
> applying any spatial convolution (conv_radius = 0). I would also
like to
> see how the results differed if I did not apply any spatial
convolution
> as well. Is there a way to configure the temporal convolution
through
> the MTD namelist? If not, is there a way to disable that function
> altogether?
>
> Thanks,
> Samuel Degelia
>
>
>
>

------------------------------------------------
Subject: Re: [rt.rap.ucar.edu #88022] Question about MTD convolution
From: Samuel Degelia
Time: Wed Dec 05 10:27:21 2018

John,

Thanks for the quick update. It is unfortunate that the time smoothing
is hard-coded in, but I don't imagine that this change would be too
hard
to make on my own. I looked through
src/tools/other/mode_time_domain/3d_conv.cc to determine an easy way
of
disabling the temporal smoothing. One way to do it might be to set
"scale" to 1.0 and then to set s_below and s_above to 0. I think those
latter two variables contain the variable data at +/- one timestep,
but
I am not 100% sure on that. I will explore this change to see if it
works for my purposes.

Thanks again!
Sam Degelia

On 12/5/2018 12:02 PM, John Halley Gotway via RT wrote:
> Samuel,
>
> Unfortunately no, the amount of time smoothing is currently had-
coded as
> +/- 1 timestep.  There is currently no way in the configuration file
to
> change this setting.  However, I used your request to create a
development
> ticket for a future version of MET.  The MET software is open
source.  For
> now, you'd need to modify the code and recompile to change the
behavior.
>
> Sorry I don't have a better answer for you.
>
> Thanks,
> John Halley Gotway
>
> On Tue, Dec 4, 2018 at 12:09 PM Samuel Degelia via RT
<met_help at ucar.edu>
> wrote:
>
>> Tue Dec 04 12:09:02 2018: Request 88022 was acted upon.
>> Transaction: Ticket created by sdegelia at gmail.com
>>         Queue: met_help
>>       Subject: Question about MTD convolution
>>         Owner: Nobody
>>    Requestors: sdegelia at gmail.com
>>        Status: new
>>   Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=88022 >
>>
>>
>> Hello,
>>
>> I am aware that Mode Time-Domain (MTD) automatically applies both
>> spatial and temporal convolution. Users are able to control the
amount
>> of spatial convolution through the conv_radius parameter in the MTD
>> namelist. However, temporal convolution appears to automatically be
>> applied over +/- 1 time-step according to the documentation here:
>>
>>
https://dtcenter.org/met/users/support/OnLinePractical/OnLinePractical_218/mtd/run.php.
>>
>>
>>
>> For my work, I am interested in very small objects such that I am
not
>> applying any spatial convolution (conv_radius = 0). I would also
like to
>> see how the results differed if I did not apply any spatial
convolution
>> as well. Is there a way to configure the temporal convolution
through
>> the MTD namelist? If not, is there a way to disable that function
>> altogether?
>>
>> Thanks,
>> Samuel Degelia
>>
>>
>>
>>

------------------------------------------------
Subject: Question about MTD convolution
From: John Halley Gotway
Time: Wed Dec 05 11:25:01 2018

Randy Bullock is the developer of the MODE Time-Domain tool and is
copied
on this message.  Randy can you please take a look at this email chain
and
offer any suggestions you might have about how the MTD source code
could be
modified to disable the time smoothing?

Thanks,
John

On Wed, Dec 5, 2018 at 10:27 AM Samuel Degelia via RT
<met_help at ucar.edu>
wrote:

>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=88022 >
>
> John,
>
> Thanks for the quick update. It is unfortunate that the time
smoothing
> is hard-coded in, but I don't imagine that this change would be too
hard
> to make on my own. I looked through
> src/tools/other/mode_time_domain/3d_conv.cc to determine an easy way
of
> disabling the temporal smoothing. One way to do it might be to set
> "scale" to 1.0 and then to set s_below and s_above to 0. I think
those
> latter two variables contain the variable data at +/- one timestep,
but
> I am not 100% sure on that. I will explore this change to see if it
> works for my purposes.
>
> Thanks again!
> Sam Degelia
>
> On 12/5/2018 12:02 PM, John Halley Gotway via RT wrote:
> > Samuel,
> >
> > Unfortunately no, the amount of time smoothing is currently had-
coded as
> > +/- 1 timestep.  There is currently no way in the configuration
file to
> > change this setting.  However, I used your request to create a
> development
> > ticket for a future version of MET.  The MET software is open
source.
> For
> > now, you'd need to modify the code and recompile to change the
behavior.
> >
> > Sorry I don't have a better answer for you.
> >
> > Thanks,
> > John Halley Gotway
> >
> > On Tue, Dec 4, 2018 at 12:09 PM Samuel Degelia via RT
<met_help at ucar.edu
> >
> > wrote:
> >
> >> Tue Dec 04 12:09:02 2018: Request 88022 was acted upon.
> >> Transaction: Ticket created by sdegelia at gmail.com
> >>         Queue: met_help
> >>       Subject: Question about MTD convolution
> >>         Owner: Nobody
> >>    Requestors: sdegelia at gmail.com
> >>        Status: new
> >>   Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=88022
> >
> >>
> >>
> >> Hello,
> >>
> >> I am aware that Mode Time-Domain (MTD) automatically applies both
> >> spatial and temporal convolution. Users are able to control the
amount
> >> of spatial convolution through the conv_radius parameter in the
MTD
> >> namelist. However, temporal convolution appears to automatically
be
> >> applied over +/- 1 time-step according to the documentation here:
> >>
> >>
>
https://dtcenter.org/met/users/support/OnLinePractical/OnLinePractical_218/mtd/run.php
> .
> >>
> >>
> >>
> >> For my work, I am interested in very small objects such that I am
not
> >> applying any spatial convolution (conv_radius = 0). I would also
like to
> >> see how the results differed if I did not apply any spatial
convolution
> >> as well. Is there a way to configure the temporal convolution
through
> >> the MTD namelist? If not, is there a way to disable that function
> >> altogether?
> >>
> >> Thanks,
> >> Samuel Degelia
> >>
> >>
> >>
> >>
>
>
>

------------------------------------------------
Subject: Question about MTD convolution
From: Randy Bullock
Time: Fri Dec 07 09:01:48 2018

Hi -

I've had a look at the code and I don't think that disabling the time
smoothing would be too difficult.  I must say that I'm not at all sure
it's
a good idea, however.

Randy

On Wed, Dec 5, 2018 at 11:24 AM John Halley Gotway <johnhg at ucar.edu>
wrote:

> Randy Bullock is the developer of the MODE Time-Domain tool and is
copied
> on this message.  Randy can you please take a look at this email
chain and
> offer any suggestions you might have about how the MTD source code
could be
> modified to disable the time smoothing?
>
> Thanks,
> John
>
> On Wed, Dec 5, 2018 at 10:27 AM Samuel Degelia via RT
<met_help at ucar.edu>
> wrote:
>
>>
>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=88022 >
>>
>> John,
>>
>> Thanks for the quick update. It is unfortunate that the time
smoothing
>> is hard-coded in, but I don't imagine that this change would be too
hard
>> to make on my own. I looked through
>> src/tools/other/mode_time_domain/3d_conv.cc to determine an easy
way of
>> disabling the temporal smoothing. One way to do it might be to set
>> "scale" to 1.0 and then to set s_below and s_above to 0. I think
those
>> latter two variables contain the variable data at +/- one timestep,
but
>> I am not 100% sure on that. I will explore this change to see if it
>> works for my purposes.
>>
>> Thanks again!
>> Sam Degelia
>>
>> On 12/5/2018 12:02 PM, John Halley Gotway via RT wrote:
>> > Samuel,
>> >
>> > Unfortunately no, the amount of time smoothing is currently had-
coded as
>> > +/- 1 timestep.  There is currently no way in the configuration
file to
>> > change this setting.  However, I used your request to create a
>> development
>> > ticket for a future version of MET.  The MET software is open
source.
>> For
>> > now, you'd need to modify the code and recompile to change the
behavior.
>> >
>> > Sorry I don't have a better answer for you.
>> >
>> > Thanks,
>> > John Halley Gotway
>> >
>> > On Tue, Dec 4, 2018 at 12:09 PM Samuel Degelia via RT <
>> met_help at ucar.edu>
>> > wrote:
>> >
>> >> Tue Dec 04 12:09:02 2018: Request 88022 was acted upon.
>> >> Transaction: Ticket created by sdegelia at gmail.com
>> >>         Queue: met_help
>> >>       Subject: Question about MTD convolution
>> >>         Owner: Nobody
>> >>    Requestors: sdegelia at gmail.com
>> >>        Status: new
>> >>   Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=88022
>> >
>> >>
>> >>
>> >> Hello,
>> >>
>> >> I am aware that Mode Time-Domain (MTD) automatically applies
both
>> >> spatial and temporal convolution. Users are able to control the
amount
>> >> of spatial convolution through the conv_radius parameter in the
MTD
>> >> namelist. However, temporal convolution appears to automatically
be
>> >> applied over +/- 1 time-step according to the documentation
here:
>> >>
>> >>
>>
https://dtcenter.org/met/users/support/OnLinePractical/OnLinePractical_218/mtd/run.php
>> .
>> >>
>> >>
>> >>
>> >> For my work, I am interested in very small objects such that I
am not
>> >> applying any spatial convolution (conv_radius = 0). I would also
like
>> to
>> >> see how the results differed if I did not apply any spatial
convolution
>> >> as well. Is there a way to configure the temporal convolution
through
>> >> the MTD namelist? If not, is there a way to disable that
function
>> >> altogether?
>> >>
>> >> Thanks,
>> >> Samuel Degelia
>> >>
>> >>
>> >>
>> >>
>>
>>
>>

------------------------------------------------
Subject: Re: [rt.rap.ucar.edu #88022] Question about MTD convolution
From: Samuel Degelia
Time: Fri Dec 07 09:11:11 2018

Randy and John,

I appreciate the feedback from you both! My only hesitation in using
the
time smoothing is that my results could be sensitive to the output
frequency of my forecasts/observations, as well as the speed of the
objects I am trying  to track. For my current case, I see a lot of
small
and fast objects that appear to be smoothed out even with the spatial
convolution set to zero. If you have the time, would either of you be
able to quickly summarize the goals/motivation for the temporal
smoothing? It is not covered in the user's guide online, as far as I
can
tell.

Thanks!
Samuel Degelia

On 12/7/2018 11:01 AM, Randy Bullock via RT wrote:
> Hi -
>
> I've had a look at the code and I don't think that disabling the
time
> smoothing would be too difficult.  I must say that I'm not at all
sure it's
> a good idea, however.
>
> Randy
>
> On Wed, Dec 5, 2018 at 11:24 AM John Halley Gotway <johnhg at ucar.edu>
wrote:
>
>> Randy Bullock is the developer of the MODE Time-Domain tool and is
copied
>> on this message.  Randy can you please take a look at this email
chain and
>> offer any suggestions you might have about how the MTD source code
could be
>> modified to disable the time smoothing?
>>
>> Thanks,
>> John
>>
>> On Wed, Dec 5, 2018 at 10:27 AM Samuel Degelia via RT
<met_help at ucar.edu>
>> wrote:
>>
>>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=88022 >
>>>
>>> John,
>>>
>>> Thanks for the quick update. It is unfortunate that the time
smoothing
>>> is hard-coded in, but I don't imagine that this change would be
too hard
>>> to make on my own. I looked through
>>> src/tools/other/mode_time_domain/3d_conv.cc to determine an easy
way of
>>> disabling the temporal smoothing. One way to do it might be to set
>>> "scale" to 1.0 and then to set s_below and s_above to 0. I think
those
>>> latter two variables contain the variable data at +/- one
timestep, but
>>> I am not 100% sure on that. I will explore this change to see if
it
>>> works for my purposes.
>>>
>>> Thanks again!
>>> Sam Degelia
>>>
>>> On 12/5/2018 12:02 PM, John Halley Gotway via RT wrote:
>>>> Samuel,
>>>>
>>>> Unfortunately no, the amount of time smoothing is currently had-
coded as
>>>> +/- 1 timestep.  There is currently no way in the configuration
file to
>>>> change this setting.  However, I used your request to create a
>>> development
>>>> ticket for a future version of MET.  The MET software is open
source.
>>> For
>>>> now, you'd need to modify the code and recompile to change the
behavior.
>>>>
>>>> Sorry I don't have a better answer for you.
>>>>
>>>> Thanks,
>>>> John Halley Gotway
>>>>
>>>> On Tue, Dec 4, 2018 at 12:09 PM Samuel Degelia via RT <
>>> met_help at ucar.edu>
>>>> wrote:
>>>>
>>>>> Tue Dec 04 12:09:02 2018: Request 88022 was acted upon.
>>>>> Transaction: Ticket created by sdegelia at gmail.com
>>>>>          Queue: met_help
>>>>>        Subject: Question about MTD convolution
>>>>>          Owner: Nobody
>>>>>     Requestors: sdegelia at gmail.com
>>>>>         Status: new
>>>>>    Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=88022
>>>>>
>>>>> Hello,
>>>>>
>>>>> I am aware that Mode Time-Domain (MTD) automatically applies
both
>>>>> spatial and temporal convolution. Users are able to control the
amount
>>>>> of spatial convolution through the conv_radius parameter in the
MTD
>>>>> namelist. However, temporal convolution appears to automatically
be
>>>>> applied over +/- 1 time-step according to the documentation
here:
>>>>>
>>>>>
>>>
https://dtcenter.org/met/users/support/OnLinePractical/OnLinePractical_218/mtd/run.php
>>> .
>>>>>
>>>>>
>>>>> For my work, I am interested in very small objects such that I
am not
>>>>> applying any spatial convolution (conv_radius = 0). I would also
like
>>> to
>>>>> see how the results differed if I did not apply any spatial
convolution
>>>>> as well. Is there a way to configure the temporal convolution
through
>>>>> the MTD namelist? If not, is there a way to disable that
function
>>>>> altogether?
>>>>>
>>>>> Thanks,
>>>>> Samuel Degelia
>>>>>
>>>>>
>>>>>
>>>>>
>>>
>>>

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