[Met_help] [rt.rap.ucar.edu #68135] History for A question about the propertiese of MODE
John Halley Gotway via RT
met_help at ucar.edu
Fri Jul 11 09:51:14 MDT 2014
----------------------------------------------------------------
Initial Request
----------------------------------------------------------------
Dear Sir/Madam:
Is there a default lowest threshold value of total interest? When i look into the the *.obj file, although there are dozens of objects of forecasts and observations, the total interest are only calculated between a few objects. Until now, i didn't seen a value less than 0.4.
Thanks!
Kefeng
*******************************************
kefeng at nju.edu.cn
朱科锋
南京大学大气科学学院
邮编:210093
南京市汉口路22号
中国江苏
*******************************************
----------------------------------------------------------------
Complete Ticket History
----------------------------------------------------------------
Subject: A question about the propertiese of MODE
From: John Halley Gotway
Time: Thu Jul 10 10:01:14 2014
Kefeng,
There are 2 settings in the MODE configuration file that are affecting
what
you see in the *.obj files. During the matching step of MODE, it
considers
all possible combinations of simple forecast object and observation
object. So if there are 10 forecast objects and 15 observation
objects,
that's 10 x 15 = 150 possible object pair combinations to consider.
However, some of those comparisons would be silly when the objects are
so
far away they have little chance of being matched.
The "max_centroid_dist" in the MODE configuration file specifies how
close
the centroids for a pair of forecast and observation objects need to
be for
them to be considered for matching. We do this to make MODE run
faster and
skip over the silly comparisons. The maximum centroid distance is
specified in grid points and the default value is 800/grid_res, where
grid_res is set by the user at the top of the config file. When
grid_res =
4km, only forecast/observation pairs whose centroids are 800/4 = 200
grid
points away or less will be compared. The interest value for object
pairs
that are skipped because they are too far is set to "NA" and they are
not
written to the *.obj output file.
In the MODE config file, you'll see the "total_interest_threshold" and
"print_interest_threshold". The total_interest_threshold is very
important
since it determines which objects are matched. Object pairs whose
total
interest exceeds this threshold are matched, while those with interest
less
that it remain unmatched. The default is 0.7.
The print_interest_threshold is less important. It can be set to only
write out information to the *.obj output file for object pairs whose
total
interest meet this criteria. The default is 0, which means that
object
pair information is written out for all pairs that were computed (but
not
for the ones that were too far away).
I suspect that you're not seeing total interest value less than 0.4
because
those pairs of objects were too far away for a total interest value to
be
computed.
If you want to see them, you could set max_centroid_dist very high and
keep
the print_interest_threshold set to 0.
Hope that helps clarify.
Thanks,
John
On Wed, Jul 9, 2014 at 9:17 PM, kefeng at nju.edu.cn via RT
<met_help at ucar.edu>
wrote:
>
> Wed Jul 09 21:17:54 2014: Request 68135 was acted upon.
> Transaction: Ticket created by kefeng at nju.edu.cn
> Queue: met_help
> Subject: A question about the propertiese of MODE
> Owner: Nobody
> Requestors: kefeng at nju.edu.cn
> Status: new
> Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=68135 >
>
>
> Dear Sir/Madam:
>
>
> Is there a default lowest threshold value of total
interest?
> When i look into the the *.obj file, although there are dozens of
objects
> of forecasts and observations, the total interest are only
calculated
> between a few objects. Until now, i didn't seen a value less than
0.4.
>
> Thanks!
>
> Kefeng
>
>
>
>
>
>
> *******************************************
> kefeng at nju.edu.cn
> 朱科锋
> 南京大学大气科学学院
> 邮编:210093
> 南京市汉口路22号
> 中国江苏
> *******************************************
>
------------------------------------------------
Subject: A question about the propertiese of MODE
From: kefeng at nju.edu.cn
Time: Thu Jul 10 19:36:55 2014
Thanks! John.
Got it. I set the max_centroid_dist = 100. I think it may be the
reason.
*******************************************
kefeng at nju.edu.cn
朱科锋
南京大学大气科学学院
邮编:210093
南京市汉口路22号
中国江苏
*******************************************
From: John Halley Gotway via RT
Date: 2014-07-11 00:01
To: kefeng
Subject: Re: [rt.rap.ucar.edu #68135] A question about the propertiese
of MODE
Kefeng,
There are 2 settings in the MODE configuration file that are affecting
what
you see in the *.obj files. During the matching step of MODE, it
considers
all possible combinations of simple forecast object and observation
object. So if there are 10 forecast objects and 15 observation
objects,
that's 10 x 15 = 150 possible object pair combinations to consider.
However, some of those comparisons would be silly when the objects are
so
far away they have little chance of being matched.
The "max_centroid_dist" in the MODE configuration file specifies how
close
the centroids for a pair of forecast and observation objects need to
be for
them to be considered for matching. We do this to make MODE run
faster and
skip over the silly comparisons. The maximum centroid distance is
specified in grid points and the default value is 800/grid_res, where
grid_res is set by the user at the top of the config file. When
grid_res =
4km, only forecast/observation pairs whose centroids are 800/4 = 200
grid
points away or less will be compared. The interest value for object
pairs
that are skipped because they are too far is set to "NA" and they are
not
written to the *.obj output file.
In the MODE config file, you'll see the "total_interest_threshold" and
"print_interest_threshold". The total_interest_threshold is very
important
since it determines which objects are matched. Object pairs whose
total
interest exceeds this threshold are matched, while those with interest
less
that it remain unmatched. The default is 0.7.
The print_interest_threshold is less important. It can be set to only
write out information to the *.obj output file for object pairs whose
total
interest meet this criteria. The default is 0, which means that
object
pair information is written out for all pairs that were computed (but
not
for the ones that were too far away).
I suspect that you're not seeing total interest value less than 0.4
because
those pairs of objects were too far away for a total interest value to
be
computed.
If you want to see them, you could set max_centroid_dist very high and
keep
the print_interest_threshold set to 0.
Hope that helps clarify.
Thanks,
John
On Wed, Jul 9, 2014 at 9:17 PM, kefeng at nju.edu.cn via RT
<met_help at ucar.edu>
wrote:
>
> Wed Jul 09 21:17:54 2014: Request 68135 was acted upon.
> Transaction: Ticket created by kefeng at nju.edu.cn
> Queue: met_help
> Subject: A question about the propertiese of MODE
> Owner: Nobody
> Requestors: kefeng at nju.edu.cn
> Status: new
> Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=68135 >
>
>
> Dear Sir/Madam:
>
>
> Is there a default lowest threshold value of total
interest?
> When i look into the the *.obj file, although there are dozens of
objects
> of forecasts and observations, the total interest are only
calculated
> between a few objects. Until now, i didn't seen a value less than
0.4.
>
> Thanks!
>
> Kefeng
>
>
>
>
>
>
> *******************************************
> kefeng at nju.edu.cn
> 朱科锋
> 南京大学大气科学学院
> 邮编:210093
> 南京市汉口路22号
> 中国江苏
> *******************************************
>
------------------------------------------------
More information about the Met_help
mailing list