[Met_help] [rt.rap.ucar.edu #87357] History for met-tc wind radius

John Halley Gotway via RT met_help at ucar.edu
Tue Jul 9 12:07:21 MDT 2019


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

This is the first time I've used MET to verify TC wind radius.  I'm using
met-7.0

AAL_WIND_34 is the maximum radius from all 4 quadrants, right?

It seems like the wind radius from all 4 quadrants (AAL_WIND_34) is always
"NA", even though the wind radii for individual quadrants (ANE_WIND_34,
ASE_WIND_34, ASW_WIND_34, ANW_WIND_34) are valid.

Even if one of the quadrants has a value of "NA" it should be treated as a
zero when calculating the maximum, AAL_WIND_34.

For example,
ANE_WIND_34 = NA
ASE_WIND_34 = NA
ASW_WIND_34 = NA
ANW_WIND_34 = 55
AAL_WIND_34 = 55


This is probably an issue with 50 and 64 knots too, but I haven't checked.
Also the bdeck (or best track) counterparts to the adeck.

I attached an example of this behavior from tc_pairs.

Dave


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

Subject: met-tc wind radius
From: John Halley Gotway
Time: Thu Oct 11 15:15:28 2018

Dave,

I see you have a question about wind quadrants in MET.  MET-TC is not
actually computing these values itself.  Instead, it's just passing
the
values through from the input ATCF files to the output .tcst output
file.

Take a look at the ATCF format:
https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt

While the "full circle" is listed as an option, it's typically not
actually
specified in the ATCF files we see.  Therefore, it shows up as NA in
the
.tcst output.

Presumably, tc-pairs could be enhanced to use the quadrants info and
compute something intelligible for the full circle and write that to
the
output.  But it doesn't currently do that because the NHC vx tools
didn't
do that.

Thanks,
John

On Thu, Oct 11, 2018 at 11:57 AM David Ahijevych via RT
<met_help at ucar.edu>
wrote:

>
> Thu Oct 11 11:57:31 2018: Request 87357 was acted upon.
> Transaction: Ticket created by ahijevyc
>        Queue: met_help
>      Subject: met-tc wind radius
>        Owner: Nobody
>   Requestors: ahijevyc at ucar.edu
>       Status: new
>  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357 >
>
>
> This is the first time I've used MET to verify TC wind radius.  I'm
using
> met-7.0
>
> AAL_WIND_34 is the maximum radius from all 4 quadrants, right?
>
> It seems like the wind radius from all 4 quadrants (AAL_WIND_34) is
always
> "NA", even though the wind radii for individual quadrants
(ANE_WIND_34,
> ASE_WIND_34, ASW_WIND_34, ANW_WIND_34) are valid.
>
> Even if one of the quadrants has a value of "NA" it should be
treated as a
> zero when calculating the maximum, AAL_WIND_34.
>
> For example,
> ANE_WIND_34 = NA
> ASE_WIND_34 = NA
> ASW_WIND_34 = NA
> ANW_WIND_34 = 55
> AAL_WIND_34 = 55
>
>
> This is probably an issue with 50 and 64 knots too, but I haven't
checked.
> Also the bdeck (or best track) counterparts to the adeck.
>
> I attached an example of this behavior from tc_pairs.
>
> Dave
>
>

------------------------------------------------
Subject: met-tc wind radius
From: David Ahijevych
Time: Thu Oct 11 15:45:02 2018

Hi John

I see.  Thanks for the quick response.   It makes total sense to
derive the
full circle radius from the quadrants, but I also understand your
mandate
to replicate NHC vx tools exactly.

Given your helpful explanation, I think I can work around this by
deriving
my own full circle ATCF lines.

Dave


On Thu, Oct 11, 2018 at 3:15 PM John Halley Gotway via RT
<met_help at ucar.edu>
wrote:

> Dave,
>
> I see you have a question about wind quadrants in MET.  MET-TC is
not
> actually computing these values itself.  Instead, it's just passing
the
> values through from the input ATCF files to the output .tcst output
file.
>
> Take a look at the ATCF format:
> https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt
>
> While the "full circle" is listed as an option, it's typically not
actually
> specified in the ATCF files we see.  Therefore, it shows up as NA in
the
> .tcst output.
>
> Presumably, tc-pairs could be enhanced to use the quadrants info and
> compute something intelligible for the full circle and write that to
the
> output.  But it doesn't currently do that because the NHC vx tools
didn't
> do that.
>
> Thanks,
> John
>
> On Thu, Oct 11, 2018 at 11:57 AM David Ahijevych via RT
<met_help at ucar.edu
> >
> wrote:
>
> >
> > Thu Oct 11 11:57:31 2018: Request 87357 was acted upon.
> > Transaction: Ticket created by ahijevyc
> >        Queue: met_help
> >      Subject: met-tc wind radius
> >        Owner: Nobody
> >   Requestors: ahijevyc at ucar.edu
> >       Status: new
> >  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357 >
> >
> >
> > This is the first time I've used MET to verify TC wind radius.
I'm using
> > met-7.0
> >
> > AAL_WIND_34 is the maximum radius from all 4 quadrants, right?
> >
> > It seems like the wind radius from all 4 quadrants (AAL_WIND_34)
is
> always
> > "NA", even though the wind radii for individual quadrants
(ANE_WIND_34,
> > ASE_WIND_34, ASW_WIND_34, ANW_WIND_34) are valid.
> >
> > Even if one of the quadrants has a value of "NA" it should be
treated as
> a
> > zero when calculating the maximum, AAL_WIND_34.
> >
> > For example,
> > ANE_WIND_34 = NA
> > ASE_WIND_34 = NA
> > ASW_WIND_34 = NA
> > ANW_WIND_34 = 55
> > AAL_WIND_34 = 55
> >
> >
> > This is probably an issue with 50 and 64 knots too, but I haven't
> checked.
> > Also the bdeck (or best track) counterparts to the adeck.
> >
> > I attached an example of this behavior from tc_pairs.
> >
> > Dave
> >
> >
>
>

------------------------------------------------
Subject: met-tc wind radius
From: David Ahijevych
Time: Mon Oct 15 10:08:33 2018

Hi John,

I thought the solution might be to derive AAA (full circle) wind radii
and
add those lines to the atcf file. But tc_pairs will ignore the AAA
full-circle wind radii lines when the NEQ lines are present. If I
manually
remove the NEQ lines, then tc_pairs appears to work--it doesn't throw
away
the full-circle wind radii.

tc_pairs should probably allow both types of lines, and process them
both,
because tc_pairs doesn't derive the full circle radii on its own.

The best solution would be to derive AAA full-circle from the NEQ
individual quadrants, as you mentioned earlier.

But barring that, processing both full-circle and individual quadrant
wind
radii makes sense.

Perhaps this was done to replicate the old NHC software exactly, but
it
makes tc_pairs less useful. It makes it hard to do verification on
hurricane size.

Dave





On Thu, Oct 11, 2018 at 3:44 PM David Ahijevych <ahijevyc at ucar.edu>
wrote:

> Hi John
>
> I see.  Thanks for the quick response.   It makes total sense to
derive
> the full circle radius from the quadrants, but I also understand
your
> mandate to replicate NHC vx tools exactly.
>
> Given your helpful explanation, I think I can work around this by
deriving
> my own full circle ATCF lines.
>
> Dave
>
>
> On Thu, Oct 11, 2018 at 3:15 PM John Halley Gotway via RT <
> met_help at ucar.edu> wrote:
>
>> Dave,
>>
>> I see you have a question about wind quadrants in MET.  MET-TC is
not
>> actually computing these values itself.  Instead, it's just passing
the
>> values through from the input ATCF files to the output .tcst output
file.
>>
>> Take a look at the ATCF format:
>> https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt
>>
>> While the "full circle" is listed as an option, it's typically not
>> actually
>> specified in the ATCF files we see.  Therefore, it shows up as NA
in the
>> .tcst output.
>>
>> Presumably, tc-pairs could be enhanced to use the quadrants info
and
>> compute something intelligible for the full circle and write that
to the
>> output.  But it doesn't currently do that because the NHC vx tools
didn't
>> do that.
>>
>> Thanks,
>> John
>>
>> On Thu, Oct 11, 2018 at 11:57 AM David Ahijevych via RT <
>> met_help at ucar.edu>
>> wrote:
>>
>> >
>> > Thu Oct 11 11:57:31 2018: Request 87357 was acted upon.
>> > Transaction: Ticket created by ahijevyc
>> >        Queue: met_help
>> >      Subject: met-tc wind radius
>> >        Owner: Nobody
>> >   Requestors: ahijevyc at ucar.edu
>> >       Status: new
>> >  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357 >
>> >
>> >
>> > This is the first time I've used MET to verify TC wind radius.
I'm
>> using
>> > met-7.0
>> >
>> > AAL_WIND_34 is the maximum radius from all 4 quadrants, right?
>> >
>> > It seems like the wind radius from all 4 quadrants (AAL_WIND_34)
is
>> always
>> > "NA", even though the wind radii for individual quadrants
(ANE_WIND_34,
>> > ASE_WIND_34, ASW_WIND_34, ANW_WIND_34) are valid.
>> >
>> > Even if one of the quadrants has a value of "NA" it should be
treated
>> as a
>> > zero when calculating the maximum, AAL_WIND_34.
>> >
>> > For example,
>> > ANE_WIND_34 = NA
>> > ASE_WIND_34 = NA
>> > ASW_WIND_34 = NA
>> > ANW_WIND_34 = 55
>> > AAL_WIND_34 = 55
>> >
>> >
>> > This is probably an issue with 50 and 64 knots too, but I haven't
>> checked.
>> > Also the bdeck (or best track) counterparts to the adeck.
>> >
>> > I attached an example of this behavior from tc_pairs.
>> >
>> > Dave
>> >
>> >
>>
>>

------------------------------------------------
Subject: met-tc wind radius
From: John Halley Gotway
Time: Tue Oct 16 13:19:27 2018

Dave,

I looked into this a bit more and did some testing similar to what
you've
done.

Looking at the ATCF documentation, the WINDCODE is either set to AAA
*or*
NEQ (for NE Quadrant).  I asked around the Jonathan Vigh (cc'ed here)
and
Kathryn Newman, and they agree that the ATCF data should contain
either the
full circle winds or the quadrant info, not both.  Technically, the
starting quadrant can be specified as any of the 4, but in practice,
I've
only ever seen NEQ.

So there's currently no way of running tc_pairs to populate both the
all-winds columns (AAL_WIND* and BAL_WIND*) and the wind quadrant
columns
for the same track.

If tc_pairs were enhanced to derive the AAL_WIND* and BAL_WIND*
columns
from the 4 quadrants, how would you do it?  Should it be the max of
the 4
quadrants... the mean of them... or something else?

And let me step back and ask what exactly are you trying to do?  We
did
make a minor tweak to met-8.0 based on a request from someone at NHC.
In
tc_stat (and stat_analysis), the new "-column_union" option tells
those
tools to run the summary job on the union of multiple input columns.
Specifically, he wanted to the see wind errors summarized across all 4
quadrants.

For example, running the following job results in 15 output lines...
ADECK
Wind - BDECK wind for full-circle and the 4 quadrants:
   met-8.0/bin/tc_stat -lookin tc_pairs -job summary -column WIND

But adding "-column_union true" results in 1 output line representing
the
full distribution of all the 15 error types from above:
      met-8.0/bin/tc_stat -lookin tc_pairs -job summary -column WIND
-column_union true

Does that help?

Thanks,
John


On Mon, Oct 15, 2018 at 10:09 AM David Ahijevych via RT
<met_help at ucar.edu>
wrote:

>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357 >
>
> Hi John,
>
> I thought the solution might be to derive AAA (full circle) wind
radii and
> add those lines to the atcf file. But tc_pairs will ignore the AAA
> full-circle wind radii lines when the NEQ lines are present. If I
manually
> remove the NEQ lines, then tc_pairs appears to work--it doesn't
throw away
> the full-circle wind radii.
>
> tc_pairs should probably allow both types of lines, and process them
both,
> because tc_pairs doesn't derive the full circle radii on its own.
>
> The best solution would be to derive AAA full-circle from the NEQ
> individual quadrants, as you mentioned earlier.
>
> But barring that, processing both full-circle and individual
quadrant wind
> radii makes sense.
>
> Perhaps this was done to replicate the old NHC software exactly, but
it
> makes tc_pairs less useful. It makes it hard to do verification on
> hurricane size.
>
> Dave
>
>
>
>
>
> On Thu, Oct 11, 2018 at 3:44 PM David Ahijevych <ahijevyc at ucar.edu>
wrote:
>
> > Hi John
> >
> > I see.  Thanks for the quick response.   It makes total sense to
derive
> > the full circle radius from the quadrants, but I also understand
your
> > mandate to replicate NHC vx tools exactly.
> >
> > Given your helpful explanation, I think I can work around this by
> deriving
> > my own full circle ATCF lines.
> >
> > Dave
> >
> >
> > On Thu, Oct 11, 2018 at 3:15 PM John Halley Gotway via RT <
> > met_help at ucar.edu> wrote:
> >
> >> Dave,
> >>
> >> I see you have a question about wind quadrants in MET.  MET-TC is
not
> >> actually computing these values itself.  Instead, it's just
passing the
> >> values through from the input ATCF files to the output .tcst
output
> file.
> >>
> >> Take a look at the ATCF format:
> >> https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt
> >>
> >> While the "full circle" is listed as an option, it's typically
not
> >> actually
> >> specified in the ATCF files we see.  Therefore, it shows up as NA
in the
> >> .tcst output.
> >>
> >> Presumably, tc-pairs could be enhanced to use the quadrants info
and
> >> compute something intelligible for the full circle and write that
to the
> >> output.  But it doesn't currently do that because the NHC vx
tools
> didn't
> >> do that.
> >>
> >> Thanks,
> >> John
> >>
> >> On Thu, Oct 11, 2018 at 11:57 AM David Ahijevych via RT <
> >> met_help at ucar.edu>
> >> wrote:
> >>
> >> >
> >> > Thu Oct 11 11:57:31 2018: Request 87357 was acted upon.
> >> > Transaction: Ticket created by ahijevyc
> >> >        Queue: met_help
> >> >      Subject: met-tc wind radius
> >> >        Owner: Nobody
> >> >   Requestors: ahijevyc at ucar.edu
> >> >       Status: new
> >> >  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357
> >
> >> >
> >> >
> >> > This is the first time I've used MET to verify TC wind radius.
I'm
> >> using
> >> > met-7.0
> >> >
> >> > AAL_WIND_34 is the maximum radius from all 4 quadrants, right?
> >> >
> >> > It seems like the wind radius from all 4 quadrants
(AAL_WIND_34) is
> >> always
> >> > "NA", even though the wind radii for individual quadrants
> (ANE_WIND_34,
> >> > ASE_WIND_34, ASW_WIND_34, ANW_WIND_34) are valid.
> >> >
> >> > Even if one of the quadrants has a value of "NA" it should be
treated
> >> as a
> >> > zero when calculating the maximum, AAL_WIND_34.
> >> >
> >> > For example,
> >> > ANE_WIND_34 = NA
> >> > ASE_WIND_34 = NA
> >> > ASW_WIND_34 = NA
> >> > ANW_WIND_34 = 55
> >> > AAL_WIND_34 = 55
> >> >
> >> >
> >> > This is probably an issue with 50 and 64 knots too, but I
haven't
> >> checked.
> >> > Also the bdeck (or best track) counterparts to the adeck.
> >> >
> >> > I attached an example of this behavior from tc_pairs.
> >> >
> >> > Dave
> >> >
> >> >
> >>
> >>
>
>

------------------------------------------------
Subject: met-tc wind radius
From: David Ahijevych
Time: Tue Oct 16 16:42:12 2018

Hi John,
I'd like to verify the max radii of 34/50/64kt winds in any quadrant.

The document
https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt uses
the
same wording to describe full circle radius as quadrant radius, so I
interpreted that as the maximum radius over the full circle.

I suppose it could be the mean, but I interpreted the full-circle wind
radii to the maximum of the 4 quadrants.

Unfortunately, the -column_union option doesn't really help.  I would
like
separate error statistics for each of the 3 different wind radii.

Dave


On Tue, Oct 16, 2018 at 1:19 PM John Halley Gotway via RT
<met_help at ucar.edu>
wrote:

> Dave,
>
> I looked into this a bit more and did some testing similar to what
you've
> done.
>
> Looking at the ATCF documentation, the WINDCODE is either set to AAA
*or*
> NEQ (for NE Quadrant).  I asked around the Jonathan Vigh (cc'ed
here) and
> Kathryn Newman, and they agree that the ATCF data should contain
either the
> full circle winds or the quadrant info, not both.  Technically, the
> starting quadrant can be specified as any of the 4, but in practice,
I've
> only ever seen NEQ.
>
> So there's currently no way of running tc_pairs to populate both the
> all-winds columns (AAL_WIND* and BAL_WIND*) and the wind quadrant
columns
> for the same track.
>
> If tc_pairs were enhanced to derive the AAL_WIND* and BAL_WIND*
columns
> from the 4 quadrants, how would you do it?  Should it be the max of
the 4
> quadrants... the mean of them... or something else?
>
> And let me step back and ask what exactly are you trying to do?  We
did
> make a minor tweak to met-8.0 based on a request from someone at
NHC.  In
> tc_stat (and stat_analysis), the new "-column_union" option tells
those
> tools to run the summary job on the union of multiple input columns.
> Specifically, he wanted to the see wind errors summarized across all
4
> quadrants.
>
> For example, running the following job results in 15 output lines...
ADECK
> Wind - BDECK wind for full-circle and the 4 quadrants:
>    met-8.0/bin/tc_stat -lookin tc_pairs -job summary -column WIND
>
> But adding "-column_union true" results in 1 output line
representing the
> full distribution of all the 15 error types from above:
>       met-8.0/bin/tc_stat -lookin tc_pairs -job summary -column WIND
> -column_union true
>
> Does that help?
>
> Thanks,
> John
>
>
> On Mon, Oct 15, 2018 at 10:09 AM David Ahijevych via RT
<met_help at ucar.edu
> >
> wrote:
>
> >
> > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357 >
> >
> > Hi John,
> >
> > I thought the solution might be to derive AAA (full circle) wind
radii
> and
> > add those lines to the atcf file. But tc_pairs will ignore the AAA
> > full-circle wind radii lines when the NEQ lines are present. If I
> manually
> > remove the NEQ lines, then tc_pairs appears to work--it doesn't
throw
> away
> > the full-circle wind radii.
> >
> > tc_pairs should probably allow both types of lines, and process
them
> both,
> > because tc_pairs doesn't derive the full circle radii on its own.
> >
> > The best solution would be to derive AAA full-circle from the NEQ
> > individual quadrants, as you mentioned earlier.
> >
> > But barring that, processing both full-circle and individual
quadrant
> wind
> > radii makes sense.
> >
> > Perhaps this was done to replicate the old NHC software exactly,
but it
> > makes tc_pairs less useful. It makes it hard to do verification on
> > hurricane size.
> >
> > Dave
> >
> >
> >
> >
> >
> > On Thu, Oct 11, 2018 at 3:44 PM David Ahijevych
<ahijevyc at ucar.edu>
> wrote:
> >
> > > Hi John
> > >
> > > I see.  Thanks for the quick response.   It makes total sense to
derive
> > > the full circle radius from the quadrants, but I also understand
your
> > > mandate to replicate NHC vx tools exactly.
> > >
> > > Given your helpful explanation, I think I can work around this
by
> > deriving
> > > my own full circle ATCF lines.
> > >
> > > Dave
> > >
> > >
> > > On Thu, Oct 11, 2018 at 3:15 PM John Halley Gotway via RT <
> > > met_help at ucar.edu> wrote:
> > >
> > >> Dave,
> > >>
> > >> I see you have a question about wind quadrants in MET.  MET-TC
is not
> > >> actually computing these values itself.  Instead, it's just
passing
> the
> > >> values through from the input ATCF files to the output .tcst
output
> > file.
> > >>
> > >> Take a look at the ATCF format:
> > >>
https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt
> > >>
> > >> While the "full circle" is listed as an option, it's typically
not
> > >> actually
> > >> specified in the ATCF files we see.  Therefore, it shows up as
NA in
> the
> > >> .tcst output.
> > >>
> > >> Presumably, tc-pairs could be enhanced to use the quadrants
info and
> > >> compute something intelligible for the full circle and write
that to
> the
> > >> output.  But it doesn't currently do that because the NHC vx
tools
> > didn't
> > >> do that.
> > >>
> > >> Thanks,
> > >> John
> > >>
> > >> On Thu, Oct 11, 2018 at 11:57 AM David Ahijevych via RT <
> > >> met_help at ucar.edu>
> > >> wrote:
> > >>
> > >> >
> > >> > Thu Oct 11 11:57:31 2018: Request 87357 was acted upon.
> > >> > Transaction: Ticket created by ahijevyc
> > >> >        Queue: met_help
> > >> >      Subject: met-tc wind radius
> > >> >        Owner: Nobody
> > >> >   Requestors: ahijevyc at ucar.edu
> > >> >       Status: new
> > >> >  Ticket <URL:
> https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357
> > >
> > >> >
> > >> >
> > >> > This is the first time I've used MET to verify TC wind
radius.  I'm
> > >> using
> > >> > met-7.0
> > >> >
> > >> > AAL_WIND_34 is the maximum radius from all 4 quadrants,
right?
> > >> >
> > >> > It seems like the wind radius from all 4 quadrants
(AAL_WIND_34) is
> > >> always
> > >> > "NA", even though the wind radii for individual quadrants
> > (ANE_WIND_34,
> > >> > ASE_WIND_34, ASW_WIND_34, ANW_WIND_34) are valid.
> > >> >
> > >> > Even if one of the quadrants has a value of "NA" it should be
> treated
> > >> as a
> > >> > zero when calculating the maximum, AAL_WIND_34.
> > >> >
> > >> > For example,
> > >> > ANE_WIND_34 = NA
> > >> > ASE_WIND_34 = NA
> > >> > ASW_WIND_34 = NA
> > >> > ANW_WIND_34 = 55
> > >> > AAL_WIND_34 = 55
> > >> >
> > >> >
> > >> > This is probably an issue with 50 and 64 knots too, but I
haven't
> > >> checked.
> > >> > Also the bdeck (or best track) counterparts to the adeck.
> > >> >
> > >> > I attached an example of this behavior from tc_pairs.
> > >> >
> > >> > Dave
> > >> >
> > >> >
> > >>
> > >>
> >
> >
>
>

------------------------------------------------
Subject: met-tc wind radius
From: John Halley Gotway
Time: Wed Oct 17 09:02:49 2018

Dave,

OK, so the behavior you'd like to see is this...

MET can only process the *ONE* wind code anyway.  AAA for full circle
or
NEQ (typically) for quadrants.  So when the quadrant wind code is
used,
populate the AAL_WIND_## and BAL_WIND_## columns (where ## is 34, 50,
and
64) with the maximum of the 4 quadrants from the ADeck and BDeck,
respectively.

And compute it as the maximum of the *non-missing* quadrant values.
So if
2 quadrants have missing data, report the max of the 2 with non-
missing
data.

Jonathan and Kathryn, do you agree that that's a reasonable way of
populating an output column that would otherwise be set to "NA" for
missing
data?

>From a coding perspective, this should be really easy... just a few
lines
of code in the right spot.

Thanks,
John

On Tue, Oct 16, 2018 at 4:42 PM David Ahijevych via RT
<met_help at ucar.edu>
wrote:

>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357 >
>
> Hi John,
> I'd like to verify the max radii of 34/50/64kt winds in any
quadrant.
>
> The document
> https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt
uses the
> same wording to describe full circle radius as quadrant radius, so I
> interpreted that as the maximum radius over the full circle.
>
> I suppose it could be the mean, but I interpreted the full-circle
wind
> radii to the maximum of the 4 quadrants.
>
> Unfortunately, the -column_union option doesn't really help.  I
would like
> separate error statistics for each of the 3 different wind radii.
>
> Dave
>
>
> On Tue, Oct 16, 2018 at 1:19 PM John Halley Gotway via RT <
> met_help at ucar.edu>
> wrote:
>
> > Dave,
> >
> > I looked into this a bit more and did some testing similar to what
you've
> > done.
> >
> > Looking at the ATCF documentation, the WINDCODE is either set to
AAA *or*
> > NEQ (for NE Quadrant).  I asked around the Jonathan Vigh (cc'ed
here) and
> > Kathryn Newman, and they agree that the ATCF data should contain
either
> the
> > full circle winds or the quadrant info, not both.  Technically,
the
> > starting quadrant can be specified as any of the 4, but in
practice, I've
> > only ever seen NEQ.
> >
> > So there's currently no way of running tc_pairs to populate both
the
> > all-winds columns (AAL_WIND* and BAL_WIND*) and the wind quadrant
columns
> > for the same track.
> >
> > If tc_pairs were enhanced to derive the AAL_WIND* and BAL_WIND*
columns
> > from the 4 quadrants, how would you do it?  Should it be the max
of the 4
> > quadrants... the mean of them... or something else?
> >
> > And let me step back and ask what exactly are you trying to do?
We did
> > make a minor tweak to met-8.0 based on a request from someone at
NHC.  In
> > tc_stat (and stat_analysis), the new "-column_union" option tells
those
> > tools to run the summary job on the union of multiple input
columns.
> > Specifically, he wanted to the see wind errors summarized across
all 4
> > quadrants.
> >
> > For example, running the following job results in 15 output
lines...
> ADECK
> > Wind - BDECK wind for full-circle and the 4 quadrants:
> >    met-8.0/bin/tc_stat -lookin tc_pairs -job summary -column WIND
> >
> > But adding "-column_union true" results in 1 output line
representing the
> > full distribution of all the 15 error types from above:
> >       met-8.0/bin/tc_stat -lookin tc_pairs -job summary -column
WIND
> > -column_union true
> >
> > Does that help?
> >
> > Thanks,
> > John
> >
> >
> > On Mon, Oct 15, 2018 at 10:09 AM David Ahijevych via RT <
> met_help at ucar.edu
> > >
> > wrote:
> >
> > >
> > > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357 >
> > >
> > > Hi John,
> > >
> > > I thought the solution might be to derive AAA (full circle) wind
radii
> > and
> > > add those lines to the atcf file. But tc_pairs will ignore the
AAA
> > > full-circle wind radii lines when the NEQ lines are present. If
I
> > manually
> > > remove the NEQ lines, then tc_pairs appears to work--it doesn't
throw
> > away
> > > the full-circle wind radii.
> > >
> > > tc_pairs should probably allow both types of lines, and process
them
> > both,
> > > because tc_pairs doesn't derive the full circle radii on its
own.
> > >
> > > The best solution would be to derive AAA full-circle from the
NEQ
> > > individual quadrants, as you mentioned earlier.
> > >
> > > But barring that, processing both full-circle and individual
quadrant
> > wind
> > > radii makes sense.
> > >
> > > Perhaps this was done to replicate the old NHC software exactly,
but it
> > > makes tc_pairs less useful. It makes it hard to do verification
on
> > > hurricane size.
> > >
> > > Dave
> > >
> > >
> > >
> > >
> > >
> > > On Thu, Oct 11, 2018 at 3:44 PM David Ahijevych
<ahijevyc at ucar.edu>
> > wrote:
> > >
> > > > Hi John
> > > >
> > > > I see.  Thanks for the quick response.   It makes total sense
to
> derive
> > > > the full circle radius from the quadrants, but I also
understand your
> > > > mandate to replicate NHC vx tools exactly.
> > > >
> > > > Given your helpful explanation, I think I can work around this
by
> > > deriving
> > > > my own full circle ATCF lines.
> > > >
> > > > Dave
> > > >
> > > >
> > > > On Thu, Oct 11, 2018 at 3:15 PM John Halley Gotway via RT <
> > > > met_help at ucar.edu> wrote:
> > > >
> > > >> Dave,
> > > >>
> > > >> I see you have a question about wind quadrants in MET.  MET-
TC is
> not
> > > >> actually computing these values itself.  Instead, it's just
passing
> > the
> > > >> values through from the input ATCF files to the output .tcst
output
> > > file.
> > > >>
> > > >> Take a look at the ATCF format:
> > > >>
https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt
> > > >>
> > > >> While the "full circle" is listed as an option, it's
typically not
> > > >> actually
> > > >> specified in the ATCF files we see.  Therefore, it shows up
as NA in
> > the
> > > >> .tcst output.
> > > >>
> > > >> Presumably, tc-pairs could be enhanced to use the quadrants
info and
> > > >> compute something intelligible for the full circle and write
that to
> > the
> > > >> output.  But it doesn't currently do that because the NHC vx
tools
> > > didn't
> > > >> do that.
> > > >>
> > > >> Thanks,
> > > >> John
> > > >>
> > > >> On Thu, Oct 11, 2018 at 11:57 AM David Ahijevych via RT <
> > > >> met_help at ucar.edu>
> > > >> wrote:
> > > >>
> > > >> >
> > > >> > Thu Oct 11 11:57:31 2018: Request 87357 was acted upon.
> > > >> > Transaction: Ticket created by ahijevyc
> > > >> >        Queue: met_help
> > > >> >      Subject: met-tc wind radius
> > > >> >        Owner: Nobody
> > > >> >   Requestors: ahijevyc at ucar.edu
> > > >> >       Status: new
> > > >> >  Ticket <URL:
> > https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357
> > > >
> > > >> >
> > > >> >
> > > >> > This is the first time I've used MET to verify TC wind
radius.
> I'm
> > > >> using
> > > >> > met-7.0
> > > >> >
> > > >> > AAL_WIND_34 is the maximum radius from all 4 quadrants,
right?
> > > >> >
> > > >> > It seems like the wind radius from all 4 quadrants
(AAL_WIND_34)
> is
> > > >> always
> > > >> > "NA", even though the wind radii for individual quadrants
> > > (ANE_WIND_34,
> > > >> > ASE_WIND_34, ASW_WIND_34, ANW_WIND_34) are valid.
> > > >> >
> > > >> > Even if one of the quadrants has a value of "NA" it should
be
> > treated
> > > >> as a
> > > >> > zero when calculating the maximum, AAL_WIND_34.
> > > >> >
> > > >> > For example,
> > > >> > ANE_WIND_34 = NA
> > > >> > ASE_WIND_34 = NA
> > > >> > ASW_WIND_34 = NA
> > > >> > ANW_WIND_34 = 55
> > > >> > AAL_WIND_34 = 55
> > > >> >
> > > >> >
> > > >> > This is probably an issue with 50 and 64 knots too, but I
haven't
> > > >> checked.
> > > >> > Also the bdeck (or best track) counterparts to the adeck.
> > > >> >
> > > >> > I attached an example of this behavior from tc_pairs.
> > > >> >
> > > >> > Dave
> > > >> >
> > > >> >
> > > >>
> > > >>
> > >
> > >
> >
> >
>
>

------------------------------------------------
Subject: met-tc wind radius
From: David Ahijevych
Time: Wed Oct 17 10:18:49 2018

Hi John,
Yes, that sounds right.

You had mentioned that Looking at the ATCF documentation, the WINDCODE
is
either set to AAA *or* NEQ (for NE Quadrant).

I'm probably missing something, but I didn't see anything in
https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt
precluding
'AAA' and 'NEQ' in the same file.

It describes the WINDCODE in the same way as it describes RAD, and we
know
RAD could be '34', '50', or '64' with all the different lines being in
the
same file.  I figured the same could be said for WINDCODE='NEQ' or
'AAA'
with the different lines being in the same file.

Dave






On Wed, Oct 17, 2018 at 9:03 AM John Halley Gotway via RT
<met_help at ucar.edu>
wrote:

> Dave,
>
> OK, so the behavior you'd like to see is this...
>
> MET can only process the *ONE* wind code anyway.  AAA for full
circle or
> NEQ (typically) for quadrants.  So when the quadrant wind code is
used,
> populate the AAL_WIND_## and BAL_WIND_## columns (where ## is 34,
50, and
> 64) with the maximum of the 4 quadrants from the ADeck and BDeck,
> respectively.
>
> And compute it as the maximum of the *non-missing* quadrant values.
So if
> 2 quadrants have missing data, report the max of the 2 with non-
missing
> data.
>
> Jonathan and Kathryn, do you agree that that's a reasonable way of
> populating an output column that would otherwise be set to "NA" for
missing
> data?
>
> From a coding perspective, this should be really easy... just a few
lines
> of code in the right spot.
>
> Thanks,
> John
>
> On Tue, Oct 16, 2018 at 4:42 PM David Ahijevych via RT
<met_help at ucar.edu>
> wrote:
>
> >
> > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357 >
> >
> > Hi John,
> > I'd like to verify the max radii of 34/50/64kt winds in any
quadrant.
> >
> > The document
> > https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt
uses
> the
> > same wording to describe full circle radius as quadrant radius, so
I
> > interpreted that as the maximum radius over the full circle.
> >
> > I suppose it could be the mean, but I interpreted the full-circle
wind
> > radii to the maximum of the 4 quadrants.
> >
> > Unfortunately, the -column_union option doesn't really help.  I
would
> like
> > separate error statistics for each of the 3 different wind radii.
> >
> > Dave
> >
> >
> > On Tue, Oct 16, 2018 at 1:19 PM John Halley Gotway via RT <
> > met_help at ucar.edu>
> > wrote:
> >
> > > Dave,
> > >
> > > I looked into this a bit more and did some testing similar to
what
> you've
> > > done.
> > >
> > > Looking at the ATCF documentation, the WINDCODE is either set to
AAA
> *or*
> > > NEQ (for NE Quadrant).  I asked around the Jonathan Vigh (cc'ed
here)
> and
> > > Kathryn Newman, and they agree that the ATCF data should contain
either
> > the
> > > full circle winds or the quadrant info, not both.  Technically,
the
> > > starting quadrant can be specified as any of the 4, but in
practice,
> I've
> > > only ever seen NEQ.
> > >
> > > So there's currently no way of running tc_pairs to populate both
the
> > > all-winds columns (AAL_WIND* and BAL_WIND*) and the wind
quadrant
> columns
> > > for the same track.
> > >
> > > If tc_pairs were enhanced to derive the AAL_WIND* and BAL_WIND*
columns
> > > from the 4 quadrants, how would you do it?  Should it be the max
of
> the 4
> > > quadrants... the mean of them... or something else?
> > >
> > > And let me step back and ask what exactly are you trying to do?
We did
> > > make a minor tweak to met-8.0 based on a request from someone at
NHC.
> In
> > > tc_stat (and stat_analysis), the new "-column_union" option
tells those
> > > tools to run the summary job on the union of multiple input
columns.
> > > Specifically, he wanted to the see wind errors summarized across
all 4
> > > quadrants.
> > >
> > > For example, running the following job results in 15 output
lines...
> > ADECK
> > > Wind - BDECK wind for full-circle and the 4 quadrants:
> > >    met-8.0/bin/tc_stat -lookin tc_pairs -job summary -column
WIND
> > >
> > > But adding "-column_union true" results in 1 output line
representing
> the
> > > full distribution of all the 15 error types from above:
> > >       met-8.0/bin/tc_stat -lookin tc_pairs -job summary -column
WIND
> > > -column_union true
> > >
> > > Does that help?
> > >
> > > Thanks,
> > > John
> > >
> > >
> > > On Mon, Oct 15, 2018 at 10:09 AM David Ahijevych via RT <
> > met_help at ucar.edu
> > > >
> > > wrote:
> > >
> > > >
> > > > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357
>
> > > >
> > > > Hi John,
> > > >
> > > > I thought the solution might be to derive AAA (full circle)
wind
> radii
> > > and
> > > > add those lines to the atcf file. But tc_pairs will ignore the
AAA
> > > > full-circle wind radii lines when the NEQ lines are present.
If I
> > > manually
> > > > remove the NEQ lines, then tc_pairs appears to work--it
doesn't throw
> > > away
> > > > the full-circle wind radii.
> > > >
> > > > tc_pairs should probably allow both types of lines, and
process them
> > > both,
> > > > because tc_pairs doesn't derive the full circle radii on its
own.
> > > >
> > > > The best solution would be to derive AAA full-circle from the
NEQ
> > > > individual quadrants, as you mentioned earlier.
> > > >
> > > > But barring that, processing both full-circle and individual
quadrant
> > > wind
> > > > radii makes sense.
> > > >
> > > > Perhaps this was done to replicate the old NHC software
exactly, but
> it
> > > > makes tc_pairs less useful. It makes it hard to do
verification on
> > > > hurricane size.
> > > >
> > > > Dave
> > > >
> > > >
> > > >
> > > >
> > > >
> > > > On Thu, Oct 11, 2018 at 3:44 PM David Ahijevych
<ahijevyc at ucar.edu>
> > > wrote:
> > > >
> > > > > Hi John
> > > > >
> > > > > I see.  Thanks for the quick response.   It makes total
sense to
> > derive
> > > > > the full circle radius from the quadrants, but I also
understand
> your
> > > > > mandate to replicate NHC vx tools exactly.
> > > > >
> > > > > Given your helpful explanation, I think I can work around
this by
> > > > deriving
> > > > > my own full circle ATCF lines.
> > > > >
> > > > > Dave
> > > > >
> > > > >
> > > > > On Thu, Oct 11, 2018 at 3:15 PM John Halley Gotway via RT <
> > > > > met_help at ucar.edu> wrote:
> > > > >
> > > > >> Dave,
> > > > >>
> > > > >> I see you have a question about wind quadrants in MET.
MET-TC is
> > not
> > > > >> actually computing these values itself.  Instead, it's just
> passing
> > > the
> > > > >> values through from the input ATCF files to the output
.tcst
> output
> > > > file.
> > > > >>
> > > > >> Take a look at the ATCF format:
> > > > >>
https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt
> > > > >>
> > > > >> While the "full circle" is listed as an option, it's
typically not
> > > > >> actually
> > > > >> specified in the ATCF files we see.  Therefore, it shows up
as NA
> in
> > > the
> > > > >> .tcst output.
> > > > >>
> > > > >> Presumably, tc-pairs could be enhanced to use the quadrants
info
> and
> > > > >> compute something intelligible for the full circle and
write that
> to
> > > the
> > > > >> output.  But it doesn't currently do that because the NHC
vx tools
> > > > didn't
> > > > >> do that.
> > > > >>
> > > > >> Thanks,
> > > > >> John
> > > > >>
> > > > >> On Thu, Oct 11, 2018 at 11:57 AM David Ahijevych via RT <
> > > > >> met_help at ucar.edu>
> > > > >> wrote:
> > > > >>
> > > > >> >
> > > > >> > Thu Oct 11 11:57:31 2018: Request 87357 was acted upon.
> > > > >> > Transaction: Ticket created by ahijevyc
> > > > >> >        Queue: met_help
> > > > >> >      Subject: met-tc wind radius
> > > > >> >        Owner: Nobody
> > > > >> >   Requestors: ahijevyc at ucar.edu
> > > > >> >       Status: new
> > > > >> >  Ticket <URL:
> > > https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357
> > > > >
> > > > >> >
> > > > >> >
> > > > >> > This is the first time I've used MET to verify TC wind
radius.
> > I'm
> > > > >> using
> > > > >> > met-7.0
> > > > >> >
> > > > >> > AAL_WIND_34 is the maximum radius from all 4 quadrants,
right?
> > > > >> >
> > > > >> > It seems like the wind radius from all 4 quadrants
(AAL_WIND_34)
> > is
> > > > >> always
> > > > >> > "NA", even though the wind radii for individual quadrants
> > > > (ANE_WIND_34,
> > > > >> > ASE_WIND_34, ASW_WIND_34, ANW_WIND_34) are valid.
> > > > >> >
> > > > >> > Even if one of the quadrants has a value of "NA" it
should be
> > > treated
> > > > >> as a
> > > > >> > zero when calculating the maximum, AAL_WIND_34.
> > > > >> >
> > > > >> > For example,
> > > > >> > ANE_WIND_34 = NA
> > > > >> > ASE_WIND_34 = NA
> > > > >> > ASW_WIND_34 = NA
> > > > >> > ANW_WIND_34 = 55
> > > > >> > AAL_WIND_34 = 55
> > > > >> >
> > > > >> >
> > > > >> > This is probably an issue with 50 and 64 knots too, but I
> haven't
> > > > >> checked.
> > > > >> > Also the bdeck (or best track) counterparts to the adeck.
> > > > >> >
> > > > >> > I attached an example of this behavior from tc_pairs.
> > > > >> >
> > > > >> > Dave
> > > > >> >
> > > > >> >
> > > > >>
> > > > >>
> > > >
> > > >
> > >
> > >
> >
> >
>
>

------------------------------------------------
Subject: met-tc wind radius
From: David Ahijevych
Time: Wed Oct 17 10:22:09 2018

But maybe that's not important.  If MET derives full-circle wind
radius
from quadrants, we won't have to fiddle with full-circle 'AAA' lines
in
ATCF files.




On Wed, Oct 17, 2018 at 10:18 AM David Ahijevych <ahijevyc at ucar.edu>
wrote:

> Hi John,
> Yes, that sounds right.
>
> You had mentioned that Looking at the ATCF documentation, the
WINDCODE is
> either set to AAA *or* NEQ (for NE Quadrant).
>
> I'm probably missing something, but I didn't see anything in
> https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt
precluding
> 'AAA' and 'NEQ' in the same file.
>
> It describes the WINDCODE in the same way as it describes RAD, and
we know
> RAD could be '34', '50', or '64' with all the different lines being
in the
> same file.  I figured the same could be said for WINDCODE='NEQ' or
'AAA'
> with the different lines being in the same file.
>
> Dave
>
>
>
>
>
>
> On Wed, Oct 17, 2018 at 9:03 AM John Halley Gotway via RT <
> met_help at ucar.edu> wrote:
>
>> Dave,
>>
>> OK, so the behavior you'd like to see is this...
>>
>> MET can only process the *ONE* wind code anyway.  AAA for full
circle or
>> NEQ (typically) for quadrants.  So when the quadrant wind code is
used,
>> populate the AAL_WIND_## and BAL_WIND_## columns (where ## is 34,
50, and
>> 64) with the maximum of the 4 quadrants from the ADeck and BDeck,
>> respectively.
>>
>> And compute it as the maximum of the *non-missing* quadrant values.
So if
>> 2 quadrants have missing data, report the max of the 2 with non-
missing
>> data.
>>
>> Jonathan and Kathryn, do you agree that that's a reasonable way of
>> populating an output column that would otherwise be set to "NA" for
>> missing
>> data?
>>
>> From a coding perspective, this should be really easy... just a few
lines
>> of code in the right spot.
>>
>> Thanks,
>> John
>>
>> On Tue, Oct 16, 2018 at 4:42 PM David Ahijevych via RT
<met_help at ucar.edu
>> >
>> wrote:
>>
>> >
>> > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357 >
>> >
>> > Hi John,
>> > I'd like to verify the max radii of 34/50/64kt winds in any
quadrant.
>> >
>> > The document
>> > https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt
uses
>> the
>> > same wording to describe full circle radius as quadrant radius,
so I
>> > interpreted that as the maximum radius over the full circle.
>> >
>> > I suppose it could be the mean, but I interpreted the full-circle
wind
>> > radii to the maximum of the 4 quadrants.
>> >
>> > Unfortunately, the -column_union option doesn't really help.  I
would
>> like
>> > separate error statistics for each of the 3 different wind radii.
>> >
>> > Dave
>> >
>> >
>> > On Tue, Oct 16, 2018 at 1:19 PM John Halley Gotway via RT <
>> > met_help at ucar.edu>
>> > wrote:
>> >
>> > > Dave,
>> > >
>> > > I looked into this a bit more and did some testing similar to
what
>> you've
>> > > done.
>> > >
>> > > Looking at the ATCF documentation, the WINDCODE is either set
to AAA
>> *or*
>> > > NEQ (for NE Quadrant).  I asked around the Jonathan Vigh (cc'ed
here)
>> and
>> > > Kathryn Newman, and they agree that the ATCF data should
contain
>> either
>> > the
>> > > full circle winds or the quadrant info, not both.  Technically,
the
>> > > starting quadrant can be specified as any of the 4, but in
practice,
>> I've
>> > > only ever seen NEQ.
>> > >
>> > > So there's currently no way of running tc_pairs to populate
both the
>> > > all-winds columns (AAL_WIND* and BAL_WIND*) and the wind
quadrant
>> columns
>> > > for the same track.
>> > >
>> > > If tc_pairs were enhanced to derive the AAL_WIND* and BAL_WIND*
>> columns
>> > > from the 4 quadrants, how would you do it?  Should it be the
max of
>> the 4
>> > > quadrants... the mean of them... or something else?
>> > >
>> > > And let me step back and ask what exactly are you trying to do?
We
>> did
>> > > make a minor tweak to met-8.0 based on a request from someone
at
>> NHC.  In
>> > > tc_stat (and stat_analysis), the new "-column_union" option
tells
>> those
>> > > tools to run the summary job on the union of multiple input
columns.
>> > > Specifically, he wanted to the see wind errors summarized
across all 4
>> > > quadrants.
>> > >
>> > > For example, running the following job results in 15 output
lines...
>> > ADECK
>> > > Wind - BDECK wind for full-circle and the 4 quadrants:
>> > >    met-8.0/bin/tc_stat -lookin tc_pairs -job summary -column
WIND
>> > >
>> > > But adding "-column_union true" results in 1 output line
representing
>> the
>> > > full distribution of all the 15 error types from above:
>> > >       met-8.0/bin/tc_stat -lookin tc_pairs -job summary -column
WIND
>> > > -column_union true
>> > >
>> > > Does that help?
>> > >
>> > > Thanks,
>> > > John
>> > >
>> > >
>> > > On Mon, Oct 15, 2018 at 10:09 AM David Ahijevych via RT <
>> > met_help at ucar.edu
>> > > >
>> > > wrote:
>> > >
>> > > >
>> > > > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357
>
>> > > >
>> > > > Hi John,
>> > > >
>> > > > I thought the solution might be to derive AAA (full circle)
wind
>> radii
>> > > and
>> > > > add those lines to the atcf file. But tc_pairs will ignore
the AAA
>> > > > full-circle wind radii lines when the NEQ lines are present.
If I
>> > > manually
>> > > > remove the NEQ lines, then tc_pairs appears to work--it
doesn't
>> throw
>> > > away
>> > > > the full-circle wind radii.
>> > > >
>> > > > tc_pairs should probably allow both types of lines, and
process them
>> > > both,
>> > > > because tc_pairs doesn't derive the full circle radii on its
own.
>> > > >
>> > > > The best solution would be to derive AAA full-circle from the
NEQ
>> > > > individual quadrants, as you mentioned earlier.
>> > > >
>> > > > But barring that, processing both full-circle and individual
>> quadrant
>> > > wind
>> > > > radii makes sense.
>> > > >
>> > > > Perhaps this was done to replicate the old NHC software
exactly,
>> but it
>> > > > makes tc_pairs less useful. It makes it hard to do
verification on
>> > > > hurricane size.
>> > > >
>> > > > Dave
>> > > >
>> > > >
>> > > >
>> > > >
>> > > >
>> > > > On Thu, Oct 11, 2018 at 3:44 PM David Ahijevych
<ahijevyc at ucar.edu>
>> > > wrote:
>> > > >
>> > > > > Hi John
>> > > > >
>> > > > > I see.  Thanks for the quick response.   It makes total
sense to
>> > derive
>> > > > > the full circle radius from the quadrants, but I also
understand
>> your
>> > > > > mandate to replicate NHC vx tools exactly.
>> > > > >
>> > > > > Given your helpful explanation, I think I can work around
this by
>> > > > deriving
>> > > > > my own full circle ATCF lines.
>> > > > >
>> > > > > Dave
>> > > > >
>> > > > >
>> > > > > On Thu, Oct 11, 2018 at 3:15 PM John Halley Gotway via RT <
>> > > > > met_help at ucar.edu> wrote:
>> > > > >
>> > > > >> Dave,
>> > > > >>
>> > > > >> I see you have a question about wind quadrants in MET.
MET-TC is
>> > not
>> > > > >> actually computing these values itself.  Instead, it's
just
>> passing
>> > > the
>> > > > >> values through from the input ATCF files to the output
.tcst
>> output
>> > > > file.
>> > > > >>
>> > > > >> Take a look at the ATCF format:
>> > > > >>
>> https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt
>> > > > >>
>> > > > >> While the "full circle" is listed as an option, it's
typically
>> not
>> > > > >> actually
>> > > > >> specified in the ATCF files we see.  Therefore, it shows
up as
>> NA in
>> > > the
>> > > > >> .tcst output.
>> > > > >>
>> > > > >> Presumably, tc-pairs could be enhanced to use the
quadrants info
>> and
>> > > > >> compute something intelligible for the full circle and
write
>> that to
>> > > the
>> > > > >> output.  But it doesn't currently do that because the NHC
vx
>> tools
>> > > > didn't
>> > > > >> do that.
>> > > > >>
>> > > > >> Thanks,
>> > > > >> John
>> > > > >>
>> > > > >> On Thu, Oct 11, 2018 at 11:57 AM David Ahijevych via RT <
>> > > > >> met_help at ucar.edu>
>> > > > >> wrote:
>> > > > >>
>> > > > >> >
>> > > > >> > Thu Oct 11 11:57:31 2018: Request 87357 was acted upon.
>> > > > >> > Transaction: Ticket created by ahijevyc
>> > > > >> >        Queue: met_help
>> > > > >> >      Subject: met-tc wind radius
>> > > > >> >        Owner: Nobody
>> > > > >> >   Requestors: ahijevyc at ucar.edu
>> > > > >> >       Status: new
>> > > > >> >  Ticket <URL:
>> > > https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357
>> > > > >
>> > > > >> >
>> > > > >> >
>> > > > >> > This is the first time I've used MET to verify TC wind
radius.
>> > I'm
>> > > > >> using
>> > > > >> > met-7.0
>> > > > >> >
>> > > > >> > AAL_WIND_34 is the maximum radius from all 4 quadrants,
right?
>> > > > >> >
>> > > > >> > It seems like the wind radius from all 4 quadrants
>> (AAL_WIND_34)
>> > is
>> > > > >> always
>> > > > >> > "NA", even though the wind radii for individual
quadrants
>> > > > (ANE_WIND_34,
>> > > > >> > ASE_WIND_34, ASW_WIND_34, ANW_WIND_34) are valid.
>> > > > >> >
>> > > > >> > Even if one of the quadrants has a value of "NA" it
should be
>> > > treated
>> > > > >> as a
>> > > > >> > zero when calculating the maximum, AAL_WIND_34.
>> > > > >> >
>> > > > >> > For example,
>> > > > >> > ANE_WIND_34 = NA
>> > > > >> > ASE_WIND_34 = NA
>> > > > >> > ASW_WIND_34 = NA
>> > > > >> > ANW_WIND_34 = 55
>> > > > >> > AAL_WIND_34 = 55
>> > > > >> >
>> > > > >> >
>> > > > >> > This is probably an issue with 50 and 64 knots too, but
I
>> haven't
>> > > > >> checked.
>> > > > >> > Also the bdeck (or best track) counterparts to the
adeck.
>> > > > >> >
>> > > > >> > I attached an example of this behavior from tc_pairs.
>> > > > >> >
>> > > > >> > Dave
>> > > > >> >
>> > > > >> >
>> > > > >>
>> > > > >>
>> > > >
>> > > >
>> > >
>> > >
>> >
>> >
>>
>>

------------------------------------------------
Subject: met-tc wind radius
From: John Halley Gotway
Time: Wed Oct 17 12:01:28 2018

Dave,

OK I did some more testing and understand what's going on.

The problem with ATCF data that we're really trying to avoid in MET is
handling duplicate track data.  In the DTC Hurricane testing we were
getting multiple sources of ATCF data and some of it was duplicates
that we
wanted to avoid.  For example, the OFCL forecast may have shown up in
the
Stream 1.0 data source AND the Stream 1.5 data source.  And we wanted
MET
to discard those duplicates.

So we added the following bit of logic.  When parsing a hurricane
track
from ATCF, the timestamps of the data should be monotonically
increasing.
If a track is going backwards in time, that's a good indication that
something is wrong... likely, you have duplicate data.

When I tested with wind quadrant data *AND* full circle winds, I just
replaced all the NEQ's with AAA's and passed two ADECK files into MET-
TC.
But that violates the monotonically increasing track times and
resulted in
many warning messages like this:

WARNING: TrackInfo::add(const ATCFTrackLine &) -> skipping
ATCFTrackLine
since the valid time is not increasing (20100629_060000 <
20100703_180000):

If you intersperse the NEQ lines followed by AAA lines for the same
timestamp, then it process them as you'd expect.

You are right that there's nothing in the ATCF spec indicating that
QQQ and
AAA windcodes couldn't both be specified.  But in practice, we've
never
seen that with real data.

So the question is where and when in the logic to add the computation
of
the maximum of the quadrants?  I currently have it implemented when
processing the winds... if QQQ, then set AAA = max(QQQ).   But I'm
thinking
we should change it to after parsing the whole track.  Only compute
the
maximums if AAA is NOT specified for any of the track points?

John

John



On Wed, Oct 17, 2018 at 10:22 AM David Ahijevych via RT
<met_help at ucar.edu>
wrote:

>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357 >
>
> But maybe that's not important.  If MET derives full-circle wind
radius
> from quadrants, we won't have to fiddle with full-circle 'AAA' lines
in
> ATCF files.
>
>
>
>
> On Wed, Oct 17, 2018 at 10:18 AM David Ahijevych <ahijevyc at ucar.edu>
> wrote:
>
> > Hi John,
> > Yes, that sounds right.
> >
> > You had mentioned that Looking at the ATCF documentation, the
WINDCODE is
> > either set to AAA *or* NEQ (for NE Quadrant).
> >
> > I'm probably missing something, but I didn't see anything in
> > https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt
> precluding
> > 'AAA' and 'NEQ' in the same file.
> >
> > It describes the WINDCODE in the same way as it describes RAD, and
we
> know
> > RAD could be '34', '50', or '64' with all the different lines
being in
> the
> > same file.  I figured the same could be said for WINDCODE='NEQ' or
'AAA'
> > with the different lines being in the same file.
> >
> > Dave
> >
> >
> >
> >
> >
> >
> > On Wed, Oct 17, 2018 at 9:03 AM John Halley Gotway via RT <
> > met_help at ucar.edu> wrote:
> >
> >> Dave,
> >>
> >> OK, so the behavior you'd like to see is this...
> >>
> >> MET can only process the *ONE* wind code anyway.  AAA for full
circle or
> >> NEQ (typically) for quadrants.  So when the quadrant wind code is
used,
> >> populate the AAL_WIND_## and BAL_WIND_## columns (where ## is 34,
50,
> and
> >> 64) with the maximum of the 4 quadrants from the ADeck and BDeck,
> >> respectively.
> >>
> >> And compute it as the maximum of the *non-missing* quadrant
values.  So
> if
> >> 2 quadrants have missing data, report the max of the 2 with non-
missing
> >> data.
> >>
> >> Jonathan and Kathryn, do you agree that that's a reasonable way
of
> >> populating an output column that would otherwise be set to "NA"
for
> >> missing
> >> data?
> >>
> >> From a coding perspective, this should be really easy... just a
few
> lines
> >> of code in the right spot.
> >>
> >> Thanks,
> >> John
> >>
> >> On Tue, Oct 16, 2018 at 4:42 PM David Ahijevych via RT <
> met_help at ucar.edu
> >> >
> >> wrote:
> >>
> >> >
> >> > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357 >
> >> >
> >> > Hi John,
> >> > I'd like to verify the max radii of 34/50/64kt winds in any
quadrant.
> >> >
> >> > The document
> >> >
https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt
> uses
> >> the
> >> > same wording to describe full circle radius as quadrant radius,
so I
> >> > interpreted that as the maximum radius over the full circle.
> >> >
> >> > I suppose it could be the mean, but I interpreted the full-
circle wind
> >> > radii to the maximum of the 4 quadrants.
> >> >
> >> > Unfortunately, the -column_union option doesn't really help.  I
would
> >> like
> >> > separate error statistics for each of the 3 different wind
radii.
> >> >
> >> > Dave
> >> >
> >> >
> >> > On Tue, Oct 16, 2018 at 1:19 PM John Halley Gotway via RT <
> >> > met_help at ucar.edu>
> >> > wrote:
> >> >
> >> > > Dave,
> >> > >
> >> > > I looked into this a bit more and did some testing similar to
what
> >> you've
> >> > > done.
> >> > >
> >> > > Looking at the ATCF documentation, the WINDCODE is either set
to AAA
> >> *or*
> >> > > NEQ (for NE Quadrant).  I asked around the Jonathan Vigh
(cc'ed
> here)
> >> and
> >> > > Kathryn Newman, and they agree that the ATCF data should
contain
> >> either
> >> > the
> >> > > full circle winds or the quadrant info, not both.
Technically, the
> >> > > starting quadrant can be specified as any of the 4, but in
practice,
> >> I've
> >> > > only ever seen NEQ.
> >> > >
> >> > > So there's currently no way of running tc_pairs to populate
both the
> >> > > all-winds columns (AAL_WIND* and BAL_WIND*) and the wind
quadrant
> >> columns
> >> > > for the same track.
> >> > >
> >> > > If tc_pairs were enhanced to derive the AAL_WIND* and
BAL_WIND*
> >> columns
> >> > > from the 4 quadrants, how would you do it?  Should it be the
max of
> >> the 4
> >> > > quadrants... the mean of them... or something else?
> >> > >
> >> > > And let me step back and ask what exactly are you trying to
do?  We
> >> did
> >> > > make a minor tweak to met-8.0 based on a request from someone
at
> >> NHC.  In
> >> > > tc_stat (and stat_analysis), the new "-column_union" option
tells
> >> those
> >> > > tools to run the summary job on the union of multiple input
columns.
> >> > > Specifically, he wanted to the see wind errors summarized
across
> all 4
> >> > > quadrants.
> >> > >
> >> > > For example, running the following job results in 15 output
lines...
> >> > ADECK
> >> > > Wind - BDECK wind for full-circle and the 4 quadrants:
> >> > >    met-8.0/bin/tc_stat -lookin tc_pairs -job summary -column
WIND
> >> > >
> >> > > But adding "-column_union true" results in 1 output line
> representing
> >> the
> >> > > full distribution of all the 15 error types from above:
> >> > >       met-8.0/bin/tc_stat -lookin tc_pairs -job summary
-column WIND
> >> > > -column_union true
> >> > >
> >> > > Does that help?
> >> > >
> >> > > Thanks,
> >> > > John
> >> > >
> >> > >
> >> > > On Mon, Oct 15, 2018 at 10:09 AM David Ahijevych via RT <
> >> > met_help at ucar.edu
> >> > > >
> >> > > wrote:
> >> > >
> >> > > >
> >> > > > <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357 >
> >> > > >
> >> > > > Hi John,
> >> > > >
> >> > > > I thought the solution might be to derive AAA (full circle)
wind
> >> radii
> >> > > and
> >> > > > add those lines to the atcf file. But tc_pairs will ignore
the AAA
> >> > > > full-circle wind radii lines when the NEQ lines are
present. If I
> >> > > manually
> >> > > > remove the NEQ lines, then tc_pairs appears to work--it
doesn't
> >> throw
> >> > > away
> >> > > > the full-circle wind radii.
> >> > > >
> >> > > > tc_pairs should probably allow both types of lines, and
process
> them
> >> > > both,
> >> > > > because tc_pairs doesn't derive the full circle radii on
its own.
> >> > > >
> >> > > > The best solution would be to derive AAA full-circle from
the NEQ
> >> > > > individual quadrants, as you mentioned earlier.
> >> > > >
> >> > > > But barring that, processing both full-circle and
individual
> >> quadrant
> >> > > wind
> >> > > > radii makes sense.
> >> > > >
> >> > > > Perhaps this was done to replicate the old NHC software
exactly,
> >> but it
> >> > > > makes tc_pairs less useful. It makes it hard to do
verification on
> >> > > > hurricane size.
> >> > > >
> >> > > > Dave
> >> > > >
> >> > > >
> >> > > >
> >> > > >
> >> > > >
> >> > > > On Thu, Oct 11, 2018 at 3:44 PM David Ahijevych <
> ahijevyc at ucar.edu>
> >> > > wrote:
> >> > > >
> >> > > > > Hi John
> >> > > > >
> >> > > > > I see.  Thanks for the quick response.   It makes total
sense to
> >> > derive
> >> > > > > the full circle radius from the quadrants, but I also
understand
> >> your
> >> > > > > mandate to replicate NHC vx tools exactly.
> >> > > > >
> >> > > > > Given your helpful explanation, I think I can work around
this
> by
> >> > > > deriving
> >> > > > > my own full circle ATCF lines.
> >> > > > >
> >> > > > > Dave
> >> > > > >
> >> > > > >
> >> > > > > On Thu, Oct 11, 2018 at 3:15 PM John Halley Gotway via RT
<
> >> > > > > met_help at ucar.edu> wrote:
> >> > > > >
> >> > > > >> Dave,
> >> > > > >>
> >> > > > >> I see you have a question about wind quadrants in MET.
MET-TC
> is
> >> > not
> >> > > > >> actually computing these values itself.  Instead, it's
just
> >> passing
> >> > > the
> >> > > > >> values through from the input ATCF files to the output
.tcst
> >> output
> >> > > > file.
> >> > > > >>
> >> > > > >> Take a look at the ATCF format:
> >> > > > >>
> >> https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt
> >> > > > >>
> >> > > > >> While the "full circle" is listed as an option, it's
typically
> >> not
> >> > > > >> actually
> >> > > > >> specified in the ATCF files we see.  Therefore, it shows
up as
> >> NA in
> >> > > the
> >> > > > >> .tcst output.
> >> > > > >>
> >> > > > >> Presumably, tc-pairs could be enhanced to use the
quadrants
> info
> >> and
> >> > > > >> compute something intelligible for the full circle and
write
> >> that to
> >> > > the
> >> > > > >> output.  But it doesn't currently do that because the
NHC vx
> >> tools
> >> > > > didn't
> >> > > > >> do that.
> >> > > > >>
> >> > > > >> Thanks,
> >> > > > >> John
> >> > > > >>
> >> > > > >> On Thu, Oct 11, 2018 at 11:57 AM David Ahijevych via RT
<
> >> > > > >> met_help at ucar.edu>
> >> > > > >> wrote:
> >> > > > >>
> >> > > > >> >
> >> > > > >> > Thu Oct 11 11:57:31 2018: Request 87357 was acted
upon.
> >> > > > >> > Transaction: Ticket created by ahijevyc
> >> > > > >> >        Queue: met_help
> >> > > > >> >      Subject: met-tc wind radius
> >> > > > >> >        Owner: Nobody
> >> > > > >> >   Requestors: ahijevyc at ucar.edu
> >> > > > >> >       Status: new
> >> > > > >> >  Ticket <URL:
> >> > > https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357
> >> > > > >
> >> > > > >> >
> >> > > > >> >
> >> > > > >> > This is the first time I've used MET to verify TC wind
> radius.
> >> > I'm
> >> > > > >> using
> >> > > > >> > met-7.0
> >> > > > >> >
> >> > > > >> > AAL_WIND_34 is the maximum radius from all 4
quadrants,
> right?
> >> > > > >> >
> >> > > > >> > It seems like the wind radius from all 4 quadrants
> >> (AAL_WIND_34)
> >> > is
> >> > > > >> always
> >> > > > >> > "NA", even though the wind radii for individual
quadrants
> >> > > > (ANE_WIND_34,
> >> > > > >> > ASE_WIND_34, ASW_WIND_34, ANW_WIND_34) are valid.
> >> > > > >> >
> >> > > > >> > Even if one of the quadrants has a value of "NA" it
should be
> >> > > treated
> >> > > > >> as a
> >> > > > >> > zero when calculating the maximum, AAL_WIND_34.
> >> > > > >> >
> >> > > > >> > For example,
> >> > > > >> > ANE_WIND_34 = NA
> >> > > > >> > ASE_WIND_34 = NA
> >> > > > >> > ASW_WIND_34 = NA
> >> > > > >> > ANW_WIND_34 = 55
> >> > > > >> > AAL_WIND_34 = 55
> >> > > > >> >
> >> > > > >> >
> >> > > > >> > This is probably an issue with 50 and 64 knots too,
but I
> >> haven't
> >> > > > >> checked.
> >> > > > >> > Also the bdeck (or best track) counterparts to the
adeck.
> >> > > > >> >
> >> > > > >> > I attached an example of this behavior from tc_pairs.
> >> > > > >> >
> >> > > > >> > Dave
> >> > > > >> >
> >> > > > >> >
> >> > > > >>
> >> > > > >>
> >> > > >
> >> > > >
> >> > >
> >> > >
> >> >
> >> >
> >>
> >>
>
>

------------------------------------------------
Subject: met-tc wind radius
From: David Ahijevych
Time: Wed Oct 17 15:24:41 2018

Good question. I'd say don't allow AAA and QQQ in the same track. If
AAA is
specified, then don't allow QQQ.  If QQQ is specified, then don't
allow AAA.

Otherwise you would need a sanity check to ensure  AAA= max(QQQ).

I can't wrap my head around duplicates and where the AAA derivation
logic
should be, so I'll have to defer to you.  I think you are right about
putting the logic after the whole track is parsed.

Dave









On Wed, Oct 17, 2018 at 12:01 PM John Halley Gotway via RT <
met_help at ucar.edu> wrote:

> Dave,
>
> OK I did some more testing and understand what's going on.
>
> The problem with ATCF data that we're really trying to avoid in MET
is
> handling duplicate track data.  In the DTC Hurricane testing we were
> getting multiple sources of ATCF data and some of it was duplicates
that we
> wanted to avoid.  For example, the OFCL forecast may have shown up
in the
> Stream 1.0 data source AND the Stream 1.5 data source.  And we
wanted MET
> to discard those duplicates.
>
> So we added the following bit of logic.  When parsing a hurricane
track
> from ATCF, the timestamps of the data should be monotonically
increasing.
> If a track is going backwards in time, that's a good indication that
> something is wrong... likely, you have duplicate data.
>
> When I tested with wind quadrant data *AND* full circle winds, I
just
> replaced all the NEQ's with AAA's and passed two ADECK files into
MET-TC.
> But that violates the monotonically increasing track times and
resulted in
> many warning messages like this:
>
> WARNING: TrackInfo::add(const ATCFTrackLine &) -> skipping
ATCFTrackLine
> since the valid time is not increasing (20100629_060000 <
20100703_180000):
>
> If you intersperse the NEQ lines followed by AAA lines for the same
> timestamp, then it process them as you'd expect.
>
> You are right that there's nothing in the ATCF spec indicating that
QQQ and
> AAA windcodes couldn't both be specified.  But in practice, we've
never
> seen that with real data.
>
> So the question is where and when in the logic to add the
computation of
> the maximum of the quadrants?  I currently have it implemented when
> processing the winds... if QQQ, then set AAA = max(QQQ).   But I'm
thinking
> we should change it to after parsing the whole track.  Only compute
the
> maximums if AAA is NOT specified for any of the track points?
>
> John
>
> John
>
>
>
> On Wed, Oct 17, 2018 at 10:22 AM David Ahijevych via RT
<met_help at ucar.edu
> >
> wrote:
>
> >
> > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357 >
> >
> > But maybe that's not important.  If MET derives full-circle wind
radius
> > from quadrants, we won't have to fiddle with full-circle 'AAA'
lines in
> > ATCF files.
> >
> >
> >
> >
> > On Wed, Oct 17, 2018 at 10:18 AM David Ahijevych
<ahijevyc at ucar.edu>
> > wrote:
> >
> > > Hi John,
> > > Yes, that sounds right.
> > >
> > > You had mentioned that Looking at the ATCF documentation, the
WINDCODE
> is
> > > either set to AAA *or* NEQ (for NE Quadrant).
> > >
> > > I'm probably missing something, but I didn't see anything in
> > >
https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt
> > precluding
> > > 'AAA' and 'NEQ' in the same file.
> > >
> > > It describes the WINDCODE in the same way as it describes RAD,
and we
> > know
> > > RAD could be '34', '50', or '64' with all the different lines
being in
> > the
> > > same file.  I figured the same could be said for WINDCODE='NEQ'
or
> 'AAA'
> > > with the different lines being in the same file.
> > >
> > > Dave
> > >
> > >
> > >
> > >
> > >
> > >
> > > On Wed, Oct 17, 2018 at 9:03 AM John Halley Gotway via RT <
> > > met_help at ucar.edu> wrote:
> > >
> > >> Dave,
> > >>
> > >> OK, so the behavior you'd like to see is this...
> > >>
> > >> MET can only process the *ONE* wind code anyway.  AAA for full
circle
> or
> > >> NEQ (typically) for quadrants.  So when the quadrant wind code
is
> used,
> > >> populate the AAL_WIND_## and BAL_WIND_## columns (where ## is
34, 50,
> > and
> > >> 64) with the maximum of the 4 quadrants from the ADeck and
BDeck,
> > >> respectively.
> > >>
> > >> And compute it as the maximum of the *non-missing* quadrant
values.
> So
> > if
> > >> 2 quadrants have missing data, report the max of the 2 with
> non-missing
> > >> data.
> > >>
> > >> Jonathan and Kathryn, do you agree that that's a reasonable way
of
> > >> populating an output column that would otherwise be set to "NA"
for
> > >> missing
> > >> data?
> > >>
> > >> From a coding perspective, this should be really easy... just a
few
> > lines
> > >> of code in the right spot.
> > >>
> > >> Thanks,
> > >> John
> > >>
> > >> On Tue, Oct 16, 2018 at 4:42 PM David Ahijevych via RT <
> > met_help at ucar.edu
> > >> >
> > >> wrote:
> > >>
> > >> >
> > >> > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357
>
> > >> >
> > >> > Hi John,
> > >> > I'd like to verify the max radii of 34/50/64kt winds in any
> quadrant.
> > >> >
> > >> > The document
> > >> >
https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt
> > uses
> > >> the
> > >> > same wording to describe full circle radius as quadrant
radius, so I
> > >> > interpreted that as the maximum radius over the full circle.
> > >> >
> > >> > I suppose it could be the mean, but I interpreted the full-
circle
> wind
> > >> > radii to the maximum of the 4 quadrants.
> > >> >
> > >> > Unfortunately, the -column_union option doesn't really help.
I
> would
> > >> like
> > >> > separate error statistics for each of the 3 different wind
radii.
> > >> >
> > >> > Dave
> > >> >
> > >> >
> > >> > On Tue, Oct 16, 2018 at 1:19 PM John Halley Gotway via RT <
> > >> > met_help at ucar.edu>
> > >> > wrote:
> > >> >
> > >> > > Dave,
> > >> > >
> > >> > > I looked into this a bit more and did some testing similar
to what
> > >> you've
> > >> > > done.
> > >> > >
> > >> > > Looking at the ATCF documentation, the WINDCODE is either
set to
> AAA
> > >> *or*
> > >> > > NEQ (for NE Quadrant).  I asked around the Jonathan Vigh
(cc'ed
> > here)
> > >> and
> > >> > > Kathryn Newman, and they agree that the ATCF data should
contain
> > >> either
> > >> > the
> > >> > > full circle winds or the quadrant info, not both.
Technically,
> the
> > >> > > starting quadrant can be specified as any of the 4, but in
> practice,
> > >> I've
> > >> > > only ever seen NEQ.
> > >> > >
> > >> > > So there's currently no way of running tc_pairs to populate
both
> the
> > >> > > all-winds columns (AAL_WIND* and BAL_WIND*) and the wind
quadrant
> > >> columns
> > >> > > for the same track.
> > >> > >
> > >> > > If tc_pairs were enhanced to derive the AAL_WIND* and
BAL_WIND*
> > >> columns
> > >> > > from the 4 quadrants, how would you do it?  Should it be
the max
> of
> > >> the 4
> > >> > > quadrants... the mean of them... or something else?
> > >> > >
> > >> > > And let me step back and ask what exactly are you trying to
do?
> We
> > >> did
> > >> > > make a minor tweak to met-8.0 based on a request from
someone at
> > >> NHC.  In
> > >> > > tc_stat (and stat_analysis), the new "-column_union" option
tells
> > >> those
> > >> > > tools to run the summary job on the union of multiple input
> columns.
> > >> > > Specifically, he wanted to the see wind errors summarized
across
> > all 4
> > >> > > quadrants.
> > >> > >
> > >> > > For example, running the following job results in 15 output
> lines...
> > >> > ADECK
> > >> > > Wind - BDECK wind for full-circle and the 4 quadrants:
> > >> > >    met-8.0/bin/tc_stat -lookin tc_pairs -job summary
-column WIND
> > >> > >
> > >> > > But adding "-column_union true" results in 1 output line
> > representing
> > >> the
> > >> > > full distribution of all the 15 error types from above:
> > >> > >       met-8.0/bin/tc_stat -lookin tc_pairs -job summary
-column
> WIND
> > >> > > -column_union true
> > >> > >
> > >> > > Does that help?
> > >> > >
> > >> > > Thanks,
> > >> > > John
> > >> > >
> > >> > >
> > >> > > On Mon, Oct 15, 2018 at 10:09 AM David Ahijevych via RT <
> > >> > met_help at ucar.edu
> > >> > > >
> > >> > > wrote:
> > >> > >
> > >> > > >
> > >> > > > <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357 >
> > >> > > >
> > >> > > > Hi John,
> > >> > > >
> > >> > > > I thought the solution might be to derive AAA (full
circle) wind
> > >> radii
> > >> > > and
> > >> > > > add those lines to the atcf file. But tc_pairs will
ignore the
> AAA
> > >> > > > full-circle wind radii lines when the NEQ lines are
present. If
> I
> > >> > > manually
> > >> > > > remove the NEQ lines, then tc_pairs appears to work--it
doesn't
> > >> throw
> > >> > > away
> > >> > > > the full-circle wind radii.
> > >> > > >
> > >> > > > tc_pairs should probably allow both types of lines, and
process
> > them
> > >> > > both,
> > >> > > > because tc_pairs doesn't derive the full circle radii on
its
> own.
> > >> > > >
> > >> > > > The best solution would be to derive AAA full-circle from
the
> NEQ
> > >> > > > individual quadrants, as you mentioned earlier.
> > >> > > >
> > >> > > > But barring that, processing both full-circle and
individual
> > >> quadrant
> > >> > > wind
> > >> > > > radii makes sense.
> > >> > > >
> > >> > > > Perhaps this was done to replicate the old NHC software
exactly,
> > >> but it
> > >> > > > makes tc_pairs less useful. It makes it hard to do
verification
> on
> > >> > > > hurricane size.
> > >> > > >
> > >> > > > Dave
> > >> > > >
> > >> > > >
> > >> > > >
> > >> > > >
> > >> > > >
> > >> > > > On Thu, Oct 11, 2018 at 3:44 PM David Ahijevych <
> > ahijevyc at ucar.edu>
> > >> > > wrote:
> > >> > > >
> > >> > > > > Hi John
> > >> > > > >
> > >> > > > > I see.  Thanks for the quick response.   It makes total
sense
> to
> > >> > derive
> > >> > > > > the full circle radius from the quadrants, but I also
> understand
> > >> your
> > >> > > > > mandate to replicate NHC vx tools exactly.
> > >> > > > >
> > >> > > > > Given your helpful explanation, I think I can work
around this
> > by
> > >> > > > deriving
> > >> > > > > my own full circle ATCF lines.
> > >> > > > >
> > >> > > > > Dave
> > >> > > > >
> > >> > > > >
> > >> > > > > On Thu, Oct 11, 2018 at 3:15 PM John Halley Gotway via
RT <
> > >> > > > > met_help at ucar.edu> wrote:
> > >> > > > >
> > >> > > > >> Dave,
> > >> > > > >>
> > >> > > > >> I see you have a question about wind quadrants in MET.
> MET-TC
> > is
> > >> > not
> > >> > > > >> actually computing these values itself.  Instead, it's
just
> > >> passing
> > >> > > the
> > >> > > > >> values through from the input ATCF files to the output
.tcst
> > >> output
> > >> > > > file.
> > >> > > > >>
> > >> > > > >> Take a look at the ATCF format:
> > >> > > > >>
> > >>
https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt
> > >> > > > >>
> > >> > > > >> While the "full circle" is listed as an option, it's
> typically
> > >> not
> > >> > > > >> actually
> > >> > > > >> specified in the ATCF files we see.  Therefore, it
shows up
> as
> > >> NA in
> > >> > > the
> > >> > > > >> .tcst output.
> > >> > > > >>
> > >> > > > >> Presumably, tc-pairs could be enhanced to use the
quadrants
> > info
> > >> and
> > >> > > > >> compute something intelligible for the full circle and
write
> > >> that to
> > >> > > the
> > >> > > > >> output.  But it doesn't currently do that because the
NHC vx
> > >> tools
> > >> > > > didn't
> > >> > > > >> do that.
> > >> > > > >>
> > >> > > > >> Thanks,
> > >> > > > >> John
> > >> > > > >>
> > >> > > > >> On Thu, Oct 11, 2018 at 11:57 AM David Ahijevych via
RT <
> > >> > > > >> met_help at ucar.edu>
> > >> > > > >> wrote:
> > >> > > > >>
> > >> > > > >> >
> > >> > > > >> > Thu Oct 11 11:57:31 2018: Request 87357 was acted
upon.
> > >> > > > >> > Transaction: Ticket created by ahijevyc
> > >> > > > >> >        Queue: met_help
> > >> > > > >> >      Subject: met-tc wind radius
> > >> > > > >> >        Owner: Nobody
> > >> > > > >> >   Requestors: ahijevyc at ucar.edu
> > >> > > > >> >       Status: new
> > >> > > > >> >  Ticket <URL:
> > >> > > https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357
> > >> > > > >
> > >> > > > >> >
> > >> > > > >> >
> > >> > > > >> > This is the first time I've used MET to verify TC
wind
> > radius.
> > >> > I'm
> > >> > > > >> using
> > >> > > > >> > met-7.0
> > >> > > > >> >
> > >> > > > >> > AAL_WIND_34 is the maximum radius from all 4
quadrants,
> > right?
> > >> > > > >> >
> > >> > > > >> > It seems like the wind radius from all 4 quadrants
> > >> (AAL_WIND_34)
> > >> > is
> > >> > > > >> always
> > >> > > > >> > "NA", even though the wind radii for individual
quadrants
> > >> > > > (ANE_WIND_34,
> > >> > > > >> > ASE_WIND_34, ASW_WIND_34, ANW_WIND_34) are valid.
> > >> > > > >> >
> > >> > > > >> > Even if one of the quadrants has a value of "NA" it
should
> be
> > >> > > treated
> > >> > > > >> as a
> > >> > > > >> > zero when calculating the maximum, AAL_WIND_34.
> > >> > > > >> >
> > >> > > > >> > For example,
> > >> > > > >> > ANE_WIND_34 = NA
> > >> > > > >> > ASE_WIND_34 = NA
> > >> > > > >> > ASW_WIND_34 = NA
> > >> > > > >> > ANW_WIND_34 = 55
> > >> > > > >> > AAL_WIND_34 = 55
> > >> > > > >> >
> > >> > > > >> >
> > >> > > > >> > This is probably an issue with 50 and 64 knots too,
but I
> > >> haven't
> > >> > > > >> checked.
> > >> > > > >> > Also the bdeck (or best track) counterparts to the
adeck.
> > >> > > > >> >
> > >> > > > >> > I attached an example of this behavior from
tc_pairs.
> > >> > > > >> >
> > >> > > > >> > Dave
> > >> > > > >> >
> > >> > > > >> >
> > >> > > > >>
> > >> > > > >>
> > >> > > >
> > >> > > >
> > >> > >
> > >> > >
> > >> >
> > >> >
> > >>
> > >>
> >
> >
>
>

------------------------------------------------
Subject: met-tc wind radius
From: John Halley Gotway
Time: Wed Oct 17 16:23:53 2018

Dave,

As long as tracks can only have AAA or QQQ, but not both, a simple
solution
is this.

When processing a single ATCF line, if the windcode is QQQ, derive and
store AAA = max(QQQ).  That adds only 5 lines of code to a single
function.  So if you give it QQQ, you'll get QQQ and AAA in the
output.  If
you give it AAA, you get only AAA in the output.

Waiting until the track is fully parsed and applying logic across all
the
track points is more complex.  After thinking through it, I'm inclined
to
go with the simple approach.

Last question, can this wait until the met-8.1 release in a few
months?  Or
should I get you a beta version of that release sooner?

We could call this a bugfix for met-8.0, but I'm pretty leery of that
since
it changes the output.  And that change is really a new feature, not a
bug.

Thanks,
John


On Wed, Oct 17, 2018 at 3:25 PM David Ahijevych via RT
<met_help at ucar.edu>
wrote:

>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357 >
>
> Good question. I'd say don't allow AAA and QQQ in the same track. If
AAA is
> specified, then don't allow QQQ.  If QQQ is specified, then don't
allow
> AAA.
>
> Otherwise you would need a sanity check to ensure  AAA= max(QQQ).
>
> I can't wrap my head around duplicates and where the AAA derivation
logic
> should be, so I'll have to defer to you.  I think you are right
about
> putting the logic after the whole track is parsed.
>
> Dave
>
>
>
>
>
>
>
>
>
> On Wed, Oct 17, 2018 at 12:01 PM John Halley Gotway via RT <
> met_help at ucar.edu> wrote:
>
> > Dave,
> >
> > OK I did some more testing and understand what's going on.
> >
> > The problem with ATCF data that we're really trying to avoid in
MET is
> > handling duplicate track data.  In the DTC Hurricane testing we
were
> > getting multiple sources of ATCF data and some of it was
duplicates that
> we
> > wanted to avoid.  For example, the OFCL forecast may have shown up
in the
> > Stream 1.0 data source AND the Stream 1.5 data source.  And we
wanted MET
> > to discard those duplicates.
> >
> > So we added the following bit of logic.  When parsing a hurricane
track
> > from ATCF, the timestamps of the data should be monotonically
increasing.
> > If a track is going backwards in time, that's a good indication
that
> > something is wrong... likely, you have duplicate data.
> >
> > When I tested with wind quadrant data *AND* full circle winds, I
just
> > replaced all the NEQ's with AAA's and passed two ADECK files into
MET-TC.
> > But that violates the monotonically increasing track times and
resulted
> in
> > many warning messages like this:
> >
> > WARNING: TrackInfo::add(const ATCFTrackLine &) -> skipping
ATCFTrackLine
> > since the valid time is not increasing (20100629_060000 <
> 20100703_180000):
> >
> > If you intersperse the NEQ lines followed by AAA lines for the
same
> > timestamp, then it process them as you'd expect.
> >
> > You are right that there's nothing in the ATCF spec indicating
that QQQ
> and
> > AAA windcodes couldn't both be specified.  But in practice, we've
never
> > seen that with real data.
> >
> > So the question is where and when in the logic to add the
computation of
> > the maximum of the quadrants?  I currently have it implemented
when
> > processing the winds... if QQQ, then set AAA = max(QQQ).   But I'm
> thinking
> > we should change it to after parsing the whole track.  Only
compute the
> > maximums if AAA is NOT specified for any of the track points?
> >
> > John
> >
> > John
> >
> >
> >
> > On Wed, Oct 17, 2018 at 10:22 AM David Ahijevych via RT <
> met_help at ucar.edu
> > >
> > wrote:
> >
> > >
> > > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357 >
> > >
> > > But maybe that's not important.  If MET derives full-circle wind
radius
> > > from quadrants, we won't have to fiddle with full-circle 'AAA'
lines in
> > > ATCF files.
> > >
> > >
> > >
> > >
> > > On Wed, Oct 17, 2018 at 10:18 AM David Ahijevych
<ahijevyc at ucar.edu>
> > > wrote:
> > >
> > > > Hi John,
> > > > Yes, that sounds right.
> > > >
> > > > You had mentioned that Looking at the ATCF documentation, the
> WINDCODE
> > is
> > > > either set to AAA *or* NEQ (for NE Quadrant).
> > > >
> > > > I'm probably missing something, but I didn't see anything in
> > > >
https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt
> > > precluding
> > > > 'AAA' and 'NEQ' in the same file.
> > > >
> > > > It describes the WINDCODE in the same way as it describes RAD,
and we
> > > know
> > > > RAD could be '34', '50', or '64' with all the different lines
being
> in
> > > the
> > > > same file.  I figured the same could be said for
WINDCODE='NEQ' or
> > 'AAA'
> > > > with the different lines being in the same file.
> > > >
> > > > Dave
> > > >
> > > >
> > > >
> > > >
> > > >
> > > >
> > > > On Wed, Oct 17, 2018 at 9:03 AM John Halley Gotway via RT <
> > > > met_help at ucar.edu> wrote:
> > > >
> > > >> Dave,
> > > >>
> > > >> OK, so the behavior you'd like to see is this...
> > > >>
> > > >> MET can only process the *ONE* wind code anyway.  AAA for
full
> circle
> > or
> > > >> NEQ (typically) for quadrants.  So when the quadrant wind
code is
> > used,
> > > >> populate the AAL_WIND_## and BAL_WIND_## columns (where ## is
34,
> 50,
> > > and
> > > >> 64) with the maximum of the 4 quadrants from the ADeck and
BDeck,
> > > >> respectively.
> > > >>
> > > >> And compute it as the maximum of the *non-missing* quadrant
values.
> > So
> > > if
> > > >> 2 quadrants have missing data, report the max of the 2 with
> > non-missing
> > > >> data.
> > > >>
> > > >> Jonathan and Kathryn, do you agree that that's a reasonable
way of
> > > >> populating an output column that would otherwise be set to
"NA" for
> > > >> missing
> > > >> data?
> > > >>
> > > >> From a coding perspective, this should be really easy... just
a few
> > > lines
> > > >> of code in the right spot.
> > > >>
> > > >> Thanks,
> > > >> John
> > > >>
> > > >> On Tue, Oct 16, 2018 at 4:42 PM David Ahijevych via RT <
> > > met_help at ucar.edu
> > > >> >
> > > >> wrote:
> > > >>
> > > >> >
> > > >> > <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357 >
> > > >> >
> > > >> > Hi John,
> > > >> > I'd like to verify the max radii of 34/50/64kt winds in any
> > quadrant.
> > > >> >
> > > >> > The document
> > > >> >
https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt
> > > uses
> > > >> the
> > > >> > same wording to describe full circle radius as quadrant
radius,
> so I
> > > >> > interpreted that as the maximum radius over the full
circle.
> > > >> >
> > > >> > I suppose it could be the mean, but I interpreted the full-
circle
> > wind
> > > >> > radii to the maximum of the 4 quadrants.
> > > >> >
> > > >> > Unfortunately, the -column_union option doesn't really
help.  I
> > would
> > > >> like
> > > >> > separate error statistics for each of the 3 different wind
radii.
> > > >> >
> > > >> > Dave
> > > >> >
> > > >> >
> > > >> > On Tue, Oct 16, 2018 at 1:19 PM John Halley Gotway via RT <
> > > >> > met_help at ucar.edu>
> > > >> > wrote:
> > > >> >
> > > >> > > Dave,
> > > >> > >
> > > >> > > I looked into this a bit more and did some testing
similar to
> what
> > > >> you've
> > > >> > > done.
> > > >> > >
> > > >> > > Looking at the ATCF documentation, the WINDCODE is either
set to
> > AAA
> > > >> *or*
> > > >> > > NEQ (for NE Quadrant).  I asked around the Jonathan Vigh
(cc'ed
> > > here)
> > > >> and
> > > >> > > Kathryn Newman, and they agree that the ATCF data should
contain
> > > >> either
> > > >> > the
> > > >> > > full circle winds or the quadrant info, not both.
Technically,
> > the
> > > >> > > starting quadrant can be specified as any of the 4, but
in
> > practice,
> > > >> I've
> > > >> > > only ever seen NEQ.
> > > >> > >
> > > >> > > So there's currently no way of running tc_pairs to
populate both
> > the
> > > >> > > all-winds columns (AAL_WIND* and BAL_WIND*) and the wind
> quadrant
> > > >> columns
> > > >> > > for the same track.
> > > >> > >
> > > >> > > If tc_pairs were enhanced to derive the AAL_WIND* and
BAL_WIND*
> > > >> columns
> > > >> > > from the 4 quadrants, how would you do it?  Should it be
the max
> > of
> > > >> the 4
> > > >> > > quadrants... the mean of them... or something else?
> > > >> > >
> > > >> > > And let me step back and ask what exactly are you trying
to do?
> > We
> > > >> did
> > > >> > > make a minor tweak to met-8.0 based on a request from
someone at
> > > >> NHC.  In
> > > >> > > tc_stat (and stat_analysis), the new "-column_union"
option
> tells
> > > >> those
> > > >> > > tools to run the summary job on the union of multiple
input
> > columns.
> > > >> > > Specifically, he wanted to the see wind errors summarized
across
> > > all 4
> > > >> > > quadrants.
> > > >> > >
> > > >> > > For example, running the following job results in 15
output
> > lines...
> > > >> > ADECK
> > > >> > > Wind - BDECK wind for full-circle and the 4 quadrants:
> > > >> > >    met-8.0/bin/tc_stat -lookin tc_pairs -job summary
-column
> WIND
> > > >> > >
> > > >> > > But adding "-column_union true" results in 1 output line
> > > representing
> > > >> the
> > > >> > > full distribution of all the 15 error types from above:
> > > >> > >       met-8.0/bin/tc_stat -lookin tc_pairs -job summary
-column
> > WIND
> > > >> > > -column_union true
> > > >> > >
> > > >> > > Does that help?
> > > >> > >
> > > >> > > Thanks,
> > > >> > > John
> > > >> > >
> > > >> > >
> > > >> > > On Mon, Oct 15, 2018 at 10:09 AM David Ahijevych via RT <
> > > >> > met_help at ucar.edu
> > > >> > > >
> > > >> > > wrote:
> > > >> > >
> > > >> > > >
> > > >> > > > <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357
> >
> > > >> > > >
> > > >> > > > Hi John,
> > > >> > > >
> > > >> > > > I thought the solution might be to derive AAA (full
circle)
> wind
> > > >> radii
> > > >> > > and
> > > >> > > > add those lines to the atcf file. But tc_pairs will
ignore the
> > AAA
> > > >> > > > full-circle wind radii lines when the NEQ lines are
present.
> If
> > I
> > > >> > > manually
> > > >> > > > remove the NEQ lines, then tc_pairs appears to work--it
> doesn't
> > > >> throw
> > > >> > > away
> > > >> > > > the full-circle wind radii.
> > > >> > > >
> > > >> > > > tc_pairs should probably allow both types of lines, and
> process
> > > them
> > > >> > > both,
> > > >> > > > because tc_pairs doesn't derive the full circle radii
on its
> > own.
> > > >> > > >
> > > >> > > > The best solution would be to derive AAA full-circle
from the
> > NEQ
> > > >> > > > individual quadrants, as you mentioned earlier.
> > > >> > > >
> > > >> > > > But barring that, processing both full-circle and
individual
> > > >> quadrant
> > > >> > > wind
> > > >> > > > radii makes sense.
> > > >> > > >
> > > >> > > > Perhaps this was done to replicate the old NHC software
> exactly,
> > > >> but it
> > > >> > > > makes tc_pairs less useful. It makes it hard to do
> verification
> > on
> > > >> > > > hurricane size.
> > > >> > > >
> > > >> > > > Dave
> > > >> > > >
> > > >> > > >
> > > >> > > >
> > > >> > > >
> > > >> > > >
> > > >> > > > On Thu, Oct 11, 2018 at 3:44 PM David Ahijevych <
> > > ahijevyc at ucar.edu>
> > > >> > > wrote:
> > > >> > > >
> > > >> > > > > Hi John
> > > >> > > > >
> > > >> > > > > I see.  Thanks for the quick response.   It makes
total
> sense
> > to
> > > >> > derive
> > > >> > > > > the full circle radius from the quadrants, but I also
> > understand
> > > >> your
> > > >> > > > > mandate to replicate NHC vx tools exactly.
> > > >> > > > >
> > > >> > > > > Given your helpful explanation, I think I can work
around
> this
> > > by
> > > >> > > > deriving
> > > >> > > > > my own full circle ATCF lines.
> > > >> > > > >
> > > >> > > > > Dave
> > > >> > > > >
> > > >> > > > >
> > > >> > > > > On Thu, Oct 11, 2018 at 3:15 PM John Halley Gotway
via RT <
> > > >> > > > > met_help at ucar.edu> wrote:
> > > >> > > > >
> > > >> > > > >> Dave,
> > > >> > > > >>
> > > >> > > > >> I see you have a question about wind quadrants in
MET.
> > MET-TC
> > > is
> > > >> > not
> > > >> > > > >> actually computing these values itself.  Instead,
it's just
> > > >> passing
> > > >> > > the
> > > >> > > > >> values through from the input ATCF files to the
output
> .tcst
> > > >> output
> > > >> > > > file.
> > > >> > > > >>
> > > >> > > > >> Take a look at the ATCF format:
> > > >> > > > >>
> > > >>
https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt
> > > >> > > > >>
> > > >> > > > >> While the "full circle" is listed as an option, it's
> > typically
> > > >> not
> > > >> > > > >> actually
> > > >> > > > >> specified in the ATCF files we see.  Therefore, it
shows up
> > as
> > > >> NA in
> > > >> > > the
> > > >> > > > >> .tcst output.
> > > >> > > > >>
> > > >> > > > >> Presumably, tc-pairs could be enhanced to use the
quadrants
> > > info
> > > >> and
> > > >> > > > >> compute something intelligible for the full circle
and
> write
> > > >> that to
> > > >> > > the
> > > >> > > > >> output.  But it doesn't currently do that because
the NHC
> vx
> > > >> tools
> > > >> > > > didn't
> > > >> > > > >> do that.
> > > >> > > > >>
> > > >> > > > >> Thanks,
> > > >> > > > >> John
> > > >> > > > >>
> > > >> > > > >> On Thu, Oct 11, 2018 at 11:57 AM David Ahijevych via
RT <
> > > >> > > > >> met_help at ucar.edu>
> > > >> > > > >> wrote:
> > > >> > > > >>
> > > >> > > > >> >
> > > >> > > > >> > Thu Oct 11 11:57:31 2018: Request 87357 was acted
upon.
> > > >> > > > >> > Transaction: Ticket created by ahijevyc
> > > >> > > > >> >        Queue: met_help
> > > >> > > > >> >      Subject: met-tc wind radius
> > > >> > > > >> >        Owner: Nobody
> > > >> > > > >> >   Requestors: ahijevyc at ucar.edu
> > > >> > > > >> >       Status: new
> > > >> > > > >> >  Ticket <URL:
> > > >> > > https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357
> > > >> > > > >
> > > >> > > > >> >
> > > >> > > > >> >
> > > >> > > > >> > This is the first time I've used MET to verify TC
wind
> > > radius.
> > > >> > I'm
> > > >> > > > >> using
> > > >> > > > >> > met-7.0
> > > >> > > > >> >
> > > >> > > > >> > AAL_WIND_34 is the maximum radius from all 4
quadrants,
> > > right?
> > > >> > > > >> >
> > > >> > > > >> > It seems like the wind radius from all 4 quadrants
> > > >> (AAL_WIND_34)
> > > >> > is
> > > >> > > > >> always
> > > >> > > > >> > "NA", even though the wind radii for individual
quadrants
> > > >> > > > (ANE_WIND_34,
> > > >> > > > >> > ASE_WIND_34, ASW_WIND_34, ANW_WIND_34) are valid.
> > > >> > > > >> >
> > > >> > > > >> > Even if one of the quadrants has a value of "NA"
it
> should
> > be
> > > >> > > treated
> > > >> > > > >> as a
> > > >> > > > >> > zero when calculating the maximum, AAL_WIND_34.
> > > >> > > > >> >
> > > >> > > > >> > For example,
> > > >> > > > >> > ANE_WIND_34 = NA
> > > >> > > > >> > ASE_WIND_34 = NA
> > > >> > > > >> > ASW_WIND_34 = NA
> > > >> > > > >> > ANW_WIND_34 = 55
> > > >> > > > >> > AAL_WIND_34 = 55
> > > >> > > > >> >
> > > >> > > > >> >
> > > >> > > > >> > This is probably an issue with 50 and 64 knots
too, but I
> > > >> haven't
> > > >> > > > >> checked.
> > > >> > > > >> > Also the bdeck (or best track) counterparts to the
adeck.
> > > >> > > > >> >
> > > >> > > > >> > I attached an example of this behavior from
tc_pairs.
> > > >> > > > >> >
> > > >> > > > >> > Dave
> > > >> > > > >> >
> > > >> > > > >> >
> > > >> > > > >>
> > > >> > > > >>
> > > >> > > >
> > > >> > > >
> > > >> > >
> > > >> > >
> > > >> >
> > > >> >
> > > >>
> > > >>
> > >
> > >
> >
> >
>
>

------------------------------------------------
Subject: met-tc wind radius
From: jvigh
Time: Wed Oct 17 18:07:24 2018

Hi Dave,
    Just to give you an update, I've sent a detailed query to NHC and
the developers of ATCF to try to understand the full intent and
definition of the wind radii parameters in the ATCF. I've also asked
if
NHC is interested in having the full circle max wind radii be verified
as a separate quantity, distinct from the four-quadrant values. If and
when we get feedback, we'll digest this and then figure out if any
changes are warranted in the behavior of tcpairs.

Best,
   Jonathan



On 10/17/18 9:02 AM, John Halley Gotway wrote:
> Dave,
>
> OK, so the behavior you'd like to see is this...
>
> MET can only process the *ONE* wind code anyway.  AAA for full
circle
> or NEQ (typically) for quadrants.  So when the quadrant wind code is
> used, populate the AAL_WIND_## and BAL_WIND_## columns (where ## is
> 34, 50, and 64) with the maximum of the 4 quadrants from the ADeck
and
> BDeck, respectively.
>
> And compute it as the maximum of the *non-missing* quadrant values. 
> So if 2 quadrants have missing data, report the max of the 2 with
> non-missing data.
>
> Jonathan and Kathryn, do you agree that that's a reasonable way of
> populating an output column that would otherwise be set to "NA" for
> missing data?
>
> From a coding perspective, this should be really easy... just a few
> lines of code in the right spot.
>
> Thanks,
> John
>
> On Tue, Oct 16, 2018 at 4:42 PM David Ahijevych via RT
> <met_help at ucar.edu <mailto:met_help at ucar.edu>> wrote:
>
>
>     <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357 >
>
>     Hi John,
>     I'd like to verify the max radii of 34/50/64kt winds in any
quadrant.
>
>     The document
>
https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt
>     uses the
>     same wording to describe full circle radius as quadrant radius,
so I
>     interpreted that as the maximum radius over the full circle.
>
>     I suppose it could be the mean, but I interpreted the full-
circle wind
>     radii to the maximum of the 4 quadrants.
>
>     Unfortunately, the -column_union option doesn't really help. I
>     would like
>     separate error statistics for each of the 3 different wind
radii.
>
>     Dave
>
>
>     On Tue, Oct 16, 2018 at 1:19 PM John Halley Gotway via RT
>     <met_help at ucar.edu <mailto:met_help at ucar.edu>>
>     wrote:
>
>     > Dave,
>     >
>     > I looked into this a bit more and did some testing similar to
>     what you've
>     > done.
>     >
>     > Looking at the ATCF documentation, the WINDCODE is either set
to
>     AAA *or*
>     > NEQ (for NE Quadrant).  I asked around the Jonathan Vigh
(cc'ed
>     here) and
>     > Kathryn Newman, and they agree that the ATCF data should
contain
>     either the
>     > full circle winds or the quadrant info, not both. Technically,
the
>     > starting quadrant can be specified as any of the 4, but in
>     practice, I've
>     > only ever seen NEQ.
>     >
>     > So there's currently no way of running tc_pairs to populate
both the
>     > all-winds columns (AAL_WIND* and BAL_WIND*) and the wind
>     quadrant columns
>     > for the same track.
>     >
>     > If tc_pairs were enhanced to derive the AAL_WIND* and
BAL_WIND*
>     columns
>     > from the 4 quadrants, how would you do it?  Should it be the
max
>     of the 4
>     > quadrants... the mean of them... or something else?
>     >
>     > And let me step back and ask what exactly are you trying to
do? 
>     We did
>     > make a minor tweak to met-8.0 based on a request from someone
at
>     NHC.  In
>     > tc_stat (and stat_analysis), the new "-column_union" option
>     tells those
>     > tools to run the summary job on the union of multiple input
columns.
>     > Specifically, he wanted to the see wind errors summarized
across
>     all 4
>     > quadrants.
>     >
>     > For example, running the following job results in 15 output
>     lines... ADECK
>     > Wind - BDECK wind for full-circle and the 4 quadrants:
>     >    met-8.0/bin/tc_stat -lookin tc_pairs -job summary -column
WIND
>     >
>     > But adding "-column_union true" results in 1 output line
>     representing the
>     > full distribution of all the 15 error types from above:
>     >       met-8.0/bin/tc_stat -lookin tc_pairs -job summary
-column WIND
>     > -column_union true
>     >
>     > Does that help?
>     >
>     > Thanks,
>     > John
>     >
>     >
>     > On Mon, Oct 15, 2018 at 10:09 AM David Ahijevych via RT
>     <met_help at ucar.edu <mailto:met_help at ucar.edu>
>     > >
>     > wrote:
>     >
>     > >
>     > > <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357 >
>     > >
>     > > Hi John,
>     > >
>     > > I thought the solution might be to derive AAA (full circle)
>     wind radii
>     > and
>     > > add those lines to the atcf file. But tc_pairs will ignore
the AAA
>     > > full-circle wind radii lines when the NEQ lines are present.
If I
>     > manually
>     > > remove the NEQ lines, then tc_pairs appears to work--it
>     doesn't throw
>     > away
>     > > the full-circle wind radii.
>     > >
>     > > tc_pairs should probably allow both types of lines, and
>     process them
>     > both,
>     > > because tc_pairs doesn't derive the full circle radii on its
own.
>     > >
>     > > The best solution would be to derive AAA full-circle from
the NEQ
>     > > individual quadrants, as you mentioned earlier.
>     > >
>     > > But barring that, processing both full-circle and individual
>     quadrant
>     > wind
>     > > radii makes sense.
>     > >
>     > > Perhaps this was done to replicate the old NHC software
>     exactly, but it
>     > > makes tc_pairs less useful. It makes it hard to do
verification on
>     > > hurricane size.
>     > >
>     > > Dave
>     > >
>     > >
>     > >
>     > >
>     > >
>     > > On Thu, Oct 11, 2018 at 3:44 PM David Ahijevych
>     <ahijevyc at ucar.edu <mailto:ahijevyc at ucar.edu>>
>     > wrote:
>     > >
>     > > > Hi John
>     > > >
>     > > > I see.  Thanks for the quick response.   It makes total
>     sense to derive
>     > > > the full circle radius from the quadrants, but I also
>     understand your
>     > > > mandate to replicate NHC vx tools exactly.
>     > > >
>     > > > Given your helpful explanation, I think I can work around
>     this by
>     > > deriving
>     > > > my own full circle ATCF lines.
>     > > >
>     > > > Dave
>     > > >
>     > > >
>     > > > On Thu, Oct 11, 2018 at 3:15 PM John Halley Gotway via RT
<
>     > > > met_help at ucar.edu <mailto:met_help at ucar.edu>> wrote:
>     > > >
>     > > >> Dave,
>     > > >>
>     > > >> I see you have a question about wind quadrants in MET. 
>     MET-TC is not
>     > > >> actually computing these values itself. Instead, it's
just
>     passing
>     > the
>     > > >> values through from the input ATCF files to the output
>     .tcst output
>     > > file.
>     > > >>
>     > > >> Take a look at the ATCF format:
>     > > >>
>
https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt
>     > > >>
>     > > >> While the "full circle" is listed as an option, it's
>     typically not
>     > > >> actually
>     > > >> specified in the ATCF files we see. Therefore, it shows
up
>     as NA in
>     > the
>     > > >> .tcst output.
>     > > >>
>     > > >> Presumably, tc-pairs could be enhanced to use the
quadrants
>     info and
>     > > >> compute something intelligible for the full circle and
>     write that to
>     > the
>     > > >> output.  But it doesn't currently do that because the NHC
>     vx tools
>     > > didn't
>     > > >> do that.
>     > > >>
>     > > >> Thanks,
>     > > >> John
>     > > >>
>     > > >> On Thu, Oct 11, 2018 at 11:57 AM David Ahijevych via RT <
>     > > >> met_help at ucar.edu <mailto:met_help at ucar.edu>>
>     > > >> wrote:
>     > > >>
>     > > >> >
>     > > >> > Thu Oct 11 11:57:31 2018: Request 87357 was acted upon.
>     > > >> > Transaction: Ticket created by ahijevyc
>     > > >> >        Queue: met_help
>     > > >> >      Subject: met-tc wind radius
>     > > >> >        Owner: Nobody
>     > > >> >   Requestors: ahijevyc at ucar.edu
<mailto:ahijevyc at ucar.edu>
>     > > >> >       Status: new
>     > > >> >  Ticket <URL:
>     > https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357
>     > > >
>     > > >> >
>     > > >> >
>     > > >> > This is the first time I've used MET to verify TC wind
>     radius.  I'm
>     > > >> using
>     > > >> > met-7.0
>     > > >> >
>     > > >> > AAL_WIND_34 is the maximum radius from all 4 quadrants,
>     right?
>     > > >> >
>     > > >> > It seems like the wind radius from all 4 quadrants
>     (AAL_WIND_34) is
>     > > >> always
>     > > >> > "NA", even though the wind radii for individual
quadrants
>     > > (ANE_WIND_34,
>     > > >> > ASE_WIND_34, ASW_WIND_34, ANW_WIND_34) are valid.
>     > > >> >
>     > > >> > Even if one of the quadrants has a value of "NA" it
should be
>     > treated
>     > > >> as a
>     > > >> > zero when calculating the maximum, AAL_WIND_34.
>     > > >> >
>     > > >> > For example,
>     > > >> > ANE_WIND_34 = NA
>     > > >> > ASE_WIND_34 = NA
>     > > >> > ASW_WIND_34 = NA
>     > > >> > ANW_WIND_34 = 55
>     > > >> > AAL_WIND_34 = 55
>     > > >> >
>     > > >> >
>     > > >> > This is probably an issue with 50 and 64 knots too, but
I
>     haven't
>     > > >> checked.
>     > > >> > Also the bdeck (or best track) counterparts to the
adeck.
>     > > >> >
>     > > >> > I attached an example of this behavior from tc_pairs.
>     > > >> >
>     > > >> > Dave
>     > > >> >
>     > > >> >
>     > > >>
>     > > >>
>     > >
>     > >
>     >
>     >
>


------------------------------------------------
Subject: met-tc wind radius
From: David Ahijevych
Time: Thu Oct 18 08:35:03 2018

Thanks for the update.

On Wed, Oct 17, 2018 at 6:07 PM jvigh via RT <met_help at ucar.edu>
wrote:

> Hi Dave,
>     Just to give you an update, I've sent a detailed query to NHC
and
> the developers of ATCF to try to understand the full intent and
> definition of the wind radii parameters in the ATCF. I've also asked
if
> NHC is interested in having the full circle max wind radii be
verified
> as a separate quantity, distinct from the four-quadrant values. If
and
> when we get feedback, we'll digest this and then figure out if any
> changes are warranted in the behavior of tcpairs.
>
> Best,
>    Jonathan
>
>
>
> On 10/17/18 9:02 AM, John Halley Gotway wrote:
> > Dave,
> >
> > OK, so the behavior you'd like to see is this...
> >
> > MET can only process the *ONE* wind code anyway.  AAA for full
circle
> > or NEQ (typically) for quadrants.  So when the quadrant wind code
is
> > used, populate the AAL_WIND_## and BAL_WIND_## columns (where ##
is
> > 34, 50, and 64) with the maximum of the 4 quadrants from the ADeck
and
> > BDeck, respectively.
> >
> > And compute it as the maximum of the *non-missing* quadrant
values.
> > So if 2 quadrants have missing data, report the max of the 2 with
> > non-missing data.
> >
> > Jonathan and Kathryn, do you agree that that's a reasonable way of
> > populating an output column that would otherwise be set to "NA"
for
> > missing data?
> >
> > From a coding perspective, this should be really easy... just a
few
> > lines of code in the right spot.
> >
> > Thanks,
> > John
> >
> > On Tue, Oct 16, 2018 at 4:42 PM David Ahijevych via RT
> > <met_help at ucar.edu <mailto:met_help at ucar.edu>> wrote:
> >
> >
> >     <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357
>
> >
> >     Hi John,
> >     I'd like to verify the max radii of 34/50/64kt winds in any
quadrant.
> >
> >     The document
> >
https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt
> >     uses the
> >     same wording to describe full circle radius as quadrant
radius, so I
> >     interpreted that as the maximum radius over the full circle.
> >
> >     I suppose it could be the mean, but I interpreted the full-
circle
> wind
> >     radii to the maximum of the 4 quadrants.
> >
> >     Unfortunately, the -column_union option doesn't really help. I
> >     would like
> >     separate error statistics for each of the 3 different wind
radii.
> >
> >     Dave
> >
> >
> >     On Tue, Oct 16, 2018 at 1:19 PM John Halley Gotway via RT
> >     <met_help at ucar.edu <mailto:met_help at ucar.edu>>
> >     wrote:
> >
> >     > Dave,
> >     >
> >     > I looked into this a bit more and did some testing similar
to
> >     what you've
> >     > done.
> >     >
> >     > Looking at the ATCF documentation, the WINDCODE is either
set to
> >     AAA *or*
> >     > NEQ (for NE Quadrant).  I asked around the Jonathan Vigh
(cc'ed
> >     here) and
> >     > Kathryn Newman, and they agree that the ATCF data should
contain
> >     either the
> >     > full circle winds or the quadrant info, not both.
Technically, the
> >     > starting quadrant can be specified as any of the 4, but in
> >     practice, I've
> >     > only ever seen NEQ.
> >     >
> >     > So there's currently no way of running tc_pairs to populate
both
> the
> >     > all-winds columns (AAL_WIND* and BAL_WIND*) and the wind
> >     quadrant columns
> >     > for the same track.
> >     >
> >     > If tc_pairs were enhanced to derive the AAL_WIND* and
BAL_WIND*
> >     columns
> >     > from the 4 quadrants, how would you do it?  Should it be the
max
> >     of the 4
> >     > quadrants... the mean of them... or something else?
> >     >
> >     > And let me step back and ask what exactly are you trying to
do?
> >     We did
> >     > make a minor tweak to met-8.0 based on a request from
someone at
> >     NHC.  In
> >     > tc_stat (and stat_analysis), the new "-column_union" option
> >     tells those
> >     > tools to run the summary job on the union of multiple input
> columns.
> >     > Specifically, he wanted to the see wind errors summarized
across
> >     all 4
> >     > quadrants.
> >     >
> >     > For example, running the following job results in 15 output
> >     lines... ADECK
> >     > Wind - BDECK wind for full-circle and the 4 quadrants:
> >     >    met-8.0/bin/tc_stat -lookin tc_pairs -job summary -column
WIND
> >     >
> >     > But adding "-column_union true" results in 1 output line
> >     representing the
> >     > full distribution of all the 15 error types from above:
> >     >       met-8.0/bin/tc_stat -lookin tc_pairs -job summary
-column
> WIND
> >     > -column_union true
> >     >
> >     > Does that help?
> >     >
> >     > Thanks,
> >     > John
> >     >
> >     >
> >     > On Mon, Oct 15, 2018 at 10:09 AM David Ahijevych via RT
> >     <met_help at ucar.edu <mailto:met_help at ucar.edu>
> >     > >
> >     > wrote:
> >     >
> >     > >
> >     > > <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357 >
> >     > >
> >     > > Hi John,
> >     > >
> >     > > I thought the solution might be to derive AAA (full
circle)
> >     wind radii
> >     > and
> >     > > add those lines to the atcf file. But tc_pairs will ignore
the
> AAA
> >     > > full-circle wind radii lines when the NEQ lines are
present. If I
> >     > manually
> >     > > remove the NEQ lines, then tc_pairs appears to work--it
> >     doesn't throw
> >     > away
> >     > > the full-circle wind radii.
> >     > >
> >     > > tc_pairs should probably allow both types of lines, and
> >     process them
> >     > both,
> >     > > because tc_pairs doesn't derive the full circle radii on
its own.
> >     > >
> >     > > The best solution would be to derive AAA full-circle from
the NEQ
> >     > > individual quadrants, as you mentioned earlier.
> >     > >
> >     > > But barring that, processing both full-circle and
individual
> >     quadrant
> >     > wind
> >     > > radii makes sense.
> >     > >
> >     > > Perhaps this was done to replicate the old NHC software
> >     exactly, but it
> >     > > makes tc_pairs less useful. It makes it hard to do
verification
> on
> >     > > hurricane size.
> >     > >
> >     > > Dave
> >     > >
> >     > >
> >     > >
> >     > >
> >     > >
> >     > > On Thu, Oct 11, 2018 at 3:44 PM David Ahijevych
> >     <ahijevyc at ucar.edu <mailto:ahijevyc at ucar.edu>>
> >     > wrote:
> >     > >
> >     > > > Hi John
> >     > > >
> >     > > > I see.  Thanks for the quick response.   It makes total
> >     sense to derive
> >     > > > the full circle radius from the quadrants, but I also
> >     understand your
> >     > > > mandate to replicate NHC vx tools exactly.
> >     > > >
> >     > > > Given your helpful explanation, I think I can work
around
> >     this by
> >     > > deriving
> >     > > > my own full circle ATCF lines.
> >     > > >
> >     > > > Dave
> >     > > >
> >     > > >
> >     > > > On Thu, Oct 11, 2018 at 3:15 PM John Halley Gotway via
RT <
> >     > > > met_help at ucar.edu <mailto:met_help at ucar.edu>> wrote:
> >     > > >
> >     > > >> Dave,
> >     > > >>
> >     > > >> I see you have a question about wind quadrants in MET.
> >     MET-TC is not
> >     > > >> actually computing these values itself. Instead, it's
just
> >     passing
> >     > the
> >     > > >> values through from the input ATCF files to the output
> >     .tcst output
> >     > > file.
> >     > > >>
> >     > > >> Take a look at the ATCF format:
> >     > > >>
> >
https://www.nrlmry.navy.mil/atcf_web/docs/database/new/abdeck.txt
> >     > > >>
> >     > > >> While the "full circle" is listed as an option, it's
> >     typically not
> >     > > >> actually
> >     > > >> specified in the ATCF files we see. Therefore, it shows
up
> >     as NA in
> >     > the
> >     > > >> .tcst output.
> >     > > >>
> >     > > >> Presumably, tc-pairs could be enhanced to use the
quadrants
> >     info and
> >     > > >> compute something intelligible for the full circle and
> >     write that to
> >     > the
> >     > > >> output.  But it doesn't currently do that because the
NHC
> >     vx tools
> >     > > didn't
> >     > > >> do that.
> >     > > >>
> >     > > >> Thanks,
> >     > > >> John
> >     > > >>
> >     > > >> On Thu, Oct 11, 2018 at 11:57 AM David Ahijevych via RT
<
> >     > > >> met_help at ucar.edu <mailto:met_help at ucar.edu>>
> >     > > >> wrote:
> >     > > >>
> >     > > >> >
> >     > > >> > Thu Oct 11 11:57:31 2018: Request 87357 was acted
upon.
> >     > > >> > Transaction: Ticket created by ahijevyc
> >     > > >> >        Queue: met_help
> >     > > >> >      Subject: met-tc wind radius
> >     > > >> >        Owner: Nobody
> >     > > >> >   Requestors: ahijevyc at ucar.edu
<mailto:ahijevyc at ucar.edu>
> >     > > >> >       Status: new
> >     > > >> >  Ticket <URL:
> >     > https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=87357
> >     > > >
> >     > > >> >
> >     > > >> >
> >     > > >> > This is the first time I've used MET to verify TC
wind
> >     radius.  I'm
> >     > > >> using
> >     > > >> > met-7.0
> >     > > >> >
> >     > > >> > AAL_WIND_34 is the maximum radius from all 4
quadrants,
> >     right?
> >     > > >> >
> >     > > >> > It seems like the wind radius from all 4 quadrants
> >     (AAL_WIND_34) is
> >     > > >> always
> >     > > >> > "NA", even though the wind radii for individual
quadrants
> >     > > (ANE_WIND_34,
> >     > > >> > ASE_WIND_34, ASW_WIND_34, ANW_WIND_34) are valid.
> >     > > >> >
> >     > > >> > Even if one of the quadrants has a value of "NA" it
should
> be
> >     > treated
> >     > > >> as a
> >     > > >> > zero when calculating the maximum, AAL_WIND_34.
> >     > > >> >
> >     > > >> > For example,
> >     > > >> > ANE_WIND_34 = NA
> >     > > >> > ASE_WIND_34 = NA
> >     > > >> > ASW_WIND_34 = NA
> >     > > >> > ANW_WIND_34 = 55
> >     > > >> > AAL_WIND_34 = 55
> >     > > >> >
> >     > > >> >
> >     > > >> > This is probably an issue with 50 and 64 knots too,
but I
> >     haven't
> >     > > >> checked.
> >     > > >> > Also the bdeck (or best track) counterparts to the
adeck.
> >     > > >> >
> >     > > >> > I attached an example of this behavior from tc_pairs.
> >     > > >> >
> >     > > >> > Dave
> >     > > >> >
> >     > > >> >
> >     > > >>
> >     > > >>
> >     > >
> >     > >
> >     >
> >     >
> >
>
>
>

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


More information about the Met_help mailing list