[Met_help] [rt.rap.ucar.edu #67206] History for MET-TC question about interp12=TRUE/False option

[Kathryn Newman via RT]

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

I have a couple questions about the interp12 option in the tc_pairs config
file.

When running tc_pairs with interp12 = TRUE, and the 12 hour interpolated
model listed with the other models, I get the attached .tcst file, which
appears to be correct.  However, when I run tc_stat, it only recognizes the
20130619_00 HWF2 forecast, and not the 20130617_12 forecast.  Therefore, I
end up with 13 verifiable forecasts instead of 14.  (In other words, if I
manually remove the first HWF2 entry, I get the exact same result as when
it is left in, even though no HWFI is present at that time.)

My questions are:
1) Is there a way to check if a .tcst file is correctly identifying "I" vs
"2" models?
2) How does tc_stat know to replace the I with the 2?

-Dave Zelinsky


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

Subject: MET-TC question about interp12=TRUE/False option
From: Kathryn Newman
Time: Thu May 22 09:11:41 2014

Hi Dave,

Great to see you are working with the system! I've looked over the
file you sent, and I see where you are running into an issue with this
option.

But first, to address your questions:
1. The “2”s that have been renamed as an “I” will show up appended at
the end of your .tcst file.  So in the example you sent – you can see
the 20130619_000000 forecasts have been added to the end of the file
with the ID HWFI (copied from the HWF2 entry).

2. tc_stat will only work on the “I” forecasts (if that is what you
have set as your amodel) because the “2”s have already been renamed to
“I” in the tc_pairs output.  Basically, tc_pairs does the work of
renaming the tracks so tc_stat only needs to worry about the model of
interest.  If you list both HWFI and HWF2 for tc_stat, it will treat
them as separate models (but the HWFI will already include the renamed
HWF2 tracks).

As far as your issue with 20130617_12 not being renamed: This case is
not getting flagged to make the substitution.  The current logic is to
search the track for a 12 hr gap (catching cases where the model is
delivered 12 hourly, or there is a missing forecast in the track).
However, for this case the HWF2 at 20130617_120000 appears to be the
first forecast, but HWFI is present at  20130617_180000,
20130618_000000, etc… so no gap is found in the track.  This case
suggests we need to modify our logic to be less strict.

To be sure we are making changes that fit the intended purpose of this
flag - do you always want the 12-hr interp to fill holes in the 6-hr
interp track only when the 6-hr is missing?

Cheers,
Kathryn

------------------------------------------------
Subject: MET-TC question about interp12=TRUE/False option
From: David Zelinsky - NOAA Affiliate
Time: Fri May 23 06:34:30 2014

Kathryn,

Ok, so any leftover "2"s are generally the cases where an I was
already
present.  Oddly, adding the 2s did add a case to HWRF, where there
probably
was a 12 hour gap, but with an I on both ends.  For the 6 hour models,
does
this mean that it won't catch the 2 if it is the first case, but it
will
catch it the rest of the time?  And if so, would that also impact 12
hour
models?

Regarding the 2 vs I treatment in general, the way that we do it at
NHC is
to use the "2" to fill in for the "I" anytime that the I is missing,
and
anytime that both are present.  The original motivation for this was
if a 2
is present, it probably means that the I wasn't there originally when
the
forecast was made.  Personally, I'm not a big fan of that rule, since
there
are other situations that could cause both to be present, such as
guidance
being pulled from the NCEP supercomputer earlier than normal.  In
generally
though, we do want to be able to use the 2 to fill in for the I for 6-
hr
models, when that model is missing.  Would it be possible to give the
user
the option to take the 2 in all cases, or only in cases where the I is
missing?

-Dave


On Thu, May 22, 2014 at 11:11 AM, Kathryn Newman via RT
<met_help at ucar.edu>wrote:

> Hi Dave,
>
> Great to see you are working with the system! I've looked over the
file
> you sent, and I see where you are running into an issue with this
option.
>
> But first, to address your questions:
> 1. The “2”s that have been renamed as an “I” will show up appended
at the
> end of your .tcst file.  So in the example you sent – you can see
the
> 20130619_000000 forecasts have been added to the end of the file
with the
> ID HWFI (copied from the HWF2 entry).
>
> 2. tc_stat will only work on the “I” forecasts (if that is what you
have
> set as your amodel) because the “2”s have already been renamed to
“I” in
> the tc_pairs output.  Basically, tc_pairs does the work of renaming
the
> tracks so tc_stat only needs to worry about the model of interest.
If you
> list both HWFI and HWF2 for tc_stat, it will treat them as separate
models
> (but the HWFI will already include the renamed HWF2 tracks).
>
> As far as your issue with 20130617_12 not being renamed: This case
is not
> getting flagged to make the substitution.  The current logic is to
search
> the track for a 12 hr gap (catching cases where the model is
delivered 12
> hourly, or there is a missing forecast in the track).  However, for
this
> case the HWF2 at 20130617_120000 appears to be the first forecast,
but HWFI
> is present at  20130617_180000, 20130618_000000, etc… so no gap is
found in
> the track.  This case suggests we need to modify our logic to be
less
> strict.
>
> To be sure we are making changes that fit the intended purpose of
this
> flag - do you always want the 12-hr interp to fill holes in the 6-hr
interp
> track only when the 6-hr is missing?
>
> Cheers,
> Kathryn
>

------------------------------------------------
Subject: MET-TC question about interp12=TRUE/False option
From: Kathryn Newman
Time: Fri May 30 17:01:18 2014

Dave,

Yes, for the 6-hr models it won't catch the first time, but it should
catch all other times (as long as there aren't multiple times missing
in
a row creating a greater than 12 hr gap).  Same is true for the 12-hr
interpolated models... it will "fill in" the missing times, but if you
have an extra missing time that creates a larger gap than 12 hrs, it
will get skipped.

Since it is apparent this is not doing what you need, we will work on
getting some logic that better fits your description below. I'll
follow-up with you to be sure we are getting the desired functionally.
We want to be sure this flag is designed to match how it would be used
in operations.

Thanks,
Kathryn

On 5/23/14 6:34 AM, David Zelinsky - NOAA Affiliate via RT wrote:
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=67206 >
>
> Kathryn,
>
> Ok, so any leftover "2"s are generally the cases where an I was
already
> present.  Oddly, adding the 2s did add a case to HWRF, where there
probably
> was a 12 hour gap, but with an I on both ends.  For the 6 hour
models, does
> this mean that it won't catch the 2 if it is the first case, but it
will
> catch it the rest of the time?  And if so, would that also impact 12
hour
> models?
>
> Regarding the 2 vs I treatment in general, the way that we do it at
NHC is
> to use the "2" to fill in for the "I" anytime that the I is missing,
and
> anytime that both are present.  The original motivation for this was
if a 2
> is present, it probably means that the I wasn't there originally
when the
> forecast was made.  Personally, I'm not a big fan of that rule,
since there
> are other situations that could cause both to be present, such as
guidance
> being pulled from the NCEP supercomputer earlier than normal.  In
generally
> though, we do want to be able to use the 2 to fill in for the I for
6-hr
> models, when that model is missing.  Would it be possible to give
the user
> the option to take the 2 in all cases, or only in cases where the I
is
> missing?
>
> -Dave
>
>
> On Thu, May 22, 2014 at 11:11 AM, Kathryn Newman via RT
> <met_help at ucar.edu>wrote:
>
>> Hi Dave,
>>
>> Great to see you are working with the system! I've looked over the
file
>> you sent, and I see where you are running into an issue with this
option.
>>
>> But first, to address your questions:
>> 1. The “2”s that have been renamed as an “I” will show up appended
at the
>> end of your .tcst file.  So in the example you sent – you can see
the
>> 20130619_000000 forecasts have been added to the end of the file
with the
>> ID HWFI (copied from the HWF2 entry).
>>
>> 2. tc_stat will only work on the “I” forecasts (if that is what you
have
>> set as your amodel) because the “2”s have already been renamed to
“I” in
>> the tc_pairs output.  Basically, tc_pairs does the work of renaming
the
>> tracks so tc_stat only needs to worry about the model of interest.
If you
>> list both HWFI and HWF2 for tc_stat, it will treat them as separate
models
>> (but the HWFI will already include the renamed HWF2 tracks).
>>
>> As far as your issue with 20130617_12 not being renamed: This case
is not
>> getting flagged to make the substitution.  The current logic is to
search
>> the track for a 12 hr gap (catching cases where the model is
delivered 12
>> hourly, or there is a missing forecast in the track).  However, for
this
>> case the HWF2 at 20130617_120000 appears to be the first forecast,
but HWFI
>> is present at  20130617_180000, 20130618_000000, etc… so no gap is
found in
>> the track.  This case suggests we need to modify our logic to be
less
>> strict.
>>
>> To be sure we are making changes that fit the intended purpose of
this
>> flag - do you always want the 12-hr interp to fill holes in the 6-
hr interp
>> track only when the 6-hr is missing?
>>
>> Cheers,
>> Kathryn
>>


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