[Met_help] [rt.rap.ucar.edu #94555] History for MODE tool
John Halley Gotway via RT
met_help at ucar.edu
Tue Mar 17 18:06:29 MDT 2020
----------------------------------------------------------------
Initial Request
----------------------------------------------------------------
Good afternoon,
My name is Marcus Johnson, and I am currently employing the MODE tool for rotation and hail research at the University of Oklahoma. I've come across some inconsistencies looking at MODE object output and my own interest functions. My total interest function is only weighted using boundary and centroid distance, with each taking the form = (max_centroid_dist - x)/(max_centroid_dist). max_centroid_dist is defined as 39.0/grid_res (3 km), or a unitless 13. These functions are defined in the input file as follows:
centroid_dist = (
( 0.0, 1.0 )
( 0.5*max_centroid_dist, 0.5 )
( max_centroid_dist, 0.0 )
);
boundary_dist = (
( 0.0, 1.0 )
( 0.5*max_centroid_dist, 0.5 )
( max_centroid_dist, 0.0 )
);
Since confidence maps are equal to 1 for the preceding attributes, the total interest function should reflect a linear combination of the two interest functions equally weighted by 0.5. Some of the .txt output doesn't reflect this though. At least one forecast cluster object (i.e., objects after merging only in their respective fields has taken place) matches with an observation object with centroid distance of 26.22, which twice exceeds the max_centroid_dist. Some of the object pairs with greater centroid and boundary distance have a larger interest than other pairs. I have attached the sample *obj.txt output file and my input file for completeness.
Thank you for your help!
Marcus
----------------------------------------------------------------
Complete Ticket History
----------------------------------------------------------------
Subject: MODE tool
From: John Halley Gotway
Time: Fri Mar 13 11:51:06 2020
Hi Marcus,
I see you have questions about the details of MODE's computation of
interest values. Before diving in, I wanted to point out that
increasing
the verbosity level to 5 prints out more detailed information about
the
interest value computations. Unfortunately, these messages are
written to
standard output (cout), not to the log file. So please try rerunning
with "-v 5" and redirect the output to a file.
mode -v 5 ... >& run_mode.log
When it's done you can inspect that log file. We're planning to do
the
met-9.0 release today, but I'll take a quick look to see if I can
change
this to route this debug information to MET's log files instead.
Listed
below is an example of the output you'll see. Note that you'll see
this
output is printed for all object pair comparisons MODE does... so lots
of
data to sift through.
Please take a look at these details to see if they explain the
behavior
you're seeing... or more clearly identify a bug! If questions remain,
please let me know.
Thanks,
John Halley Gotway
Centroid Distance:
------------------
Value = 9.445
Interest = 0.411
Confidence = 0.019
Weight = 2.000
Term = 0.016
Boundary Distance:
------------------
Value = 0.000
Interest = 1.000
Confidence = 1.000
Weight = 4.000
Term = 4.000
Convex Hull Distance:
---------------------
Value = 0.000
Interest = 1.000
Confidence = 1.000
Weight = 0.000
Term = 0.000
Angle Difference:
-----------------
Value = 22.201
Interest = 1.000
Confidence = 0.576
Weight = 1.000
Term = 0.576
Aspect Ratio Difference:
------------------------
Value = 0.115
Interest = 0.976
Confidence = 1.000
Weight = 0.000
Term = 0.000
Area Ratio:
-----------
Value = 0.019
Interest = 0.024
Confidence = 1.000
Weight = 1.000
Term = 0.024
Intersection/Area Ratio:
------------------------
Value = 0.846
Interest = 1.000
Confidence = 1.000
Weight = 2.000
Term = 2.000
Curvature Ratio:
----------------
Value = 0.938
Interest = 1.000
Confidence = 1.000
Weight = 0.000
Term = 0.000
Complexity Ratio:
-----------------
Value = 0.146
Interest = 0.183
Confidence = 1.000
Weight = 0.000
Term = 0.000
Percentile (50th) Intensity Ratio:
--------------------
Value = 0.931
Interest = 1.000
Confidence = 1.000
Weight = 0.000
Term = 0.000
Total Interest = (sum of terms)/(sum of weights*confidence)
= 6.615/7.614
= 0.869
On Fri, Mar 13, 2020 at 11:17 AM Johnson, Marcus R. via RT <
met_help at ucar.edu> wrote:
>
> Fri Mar 13 11:17:03 2020: Request 94555 was acted upon.
> Transaction: Ticket created by marcus.johnson at ou.edu
> Queue: met_help
> Subject: MODE tool
> Owner: Nobody
> Requestors: marcus.johnson at ou.edu
> Status: new
> Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=94555 >
>
>
> Good afternoon,
>
> My name is Marcus Johnson, and I am currently employing the MODE
tool for
> rotation and hail research at the University of Oklahoma. I've come
across
> some inconsistencies looking at MODE object output and my own
interest
> functions. My total interest function is only weighted using
boundary and
> centroid distance, with each taking the form = (max_centroid_dist -
> x)/(max_centroid_dist). max_centroid_dist is defined as
39.0/grid_res (3
> km), or a unitless 13. These functions are defined in the input file
as
> follows:
>
> centroid_dist = (
> ( 0.0, 1.0 )
> ( 0.5*max_centroid_dist, 0.5 )
> ( max_centroid_dist, 0.0 )
> );
>
> boundary_dist = (
> ( 0.0, 1.0 )
> ( 0.5*max_centroid_dist, 0.5 )
> ( max_centroid_dist, 0.0 )
> );
>
>
> Since confidence maps are equal to 1 for the preceding attributes,
the
> total interest function should reflect a linear combination of the
two
> interest functions equally weighted by 0.5. Some of the .txt output
doesn't
> reflect this though. At least one forecast cluster object (i.e.,
objects
> after merging only in their respective fields has taken place)
matches with
> an observation object with centroid distance of 26.22, which twice
exceeds
> the max_centroid_dist. Some of the object pairs with greater
centroid and
> boundary distance have a larger interest than other pairs. I have
attached
> the sample *obj.txt output file and my input file for completeness.
>
> Thank you for your help!
> Marcus
>
>
------------------------------------------------
Subject: MODE tool
From: John Halley Gotway
Time: Fri Mar 13 12:45:46 2020
Marcus,
FYI, I just finished updating MODE for the met-9.0 release to write
these
messages to the standard log file. I also realized that there was no
indication of which objects were being compared. So I updated the log
messages to include that. Listed below is what these will look like
in
met-9.0. Does this look good to you?
Thanks,
John
DEBUG 5:
DEBUG 5:
--------------------------------------------------------------------------------
DEBUG 5:
DEBUG 5: Computing total interest for single forecast object 1 and
single
observation object 1 pair.
DEBUG 5:
DEBUG 5: Centroid Distance:
DEBUG 5: ------------------
DEBUG 5: Value = 7.32109
DEBUG 5: Interest = 0.568808
DEBUG 5: Confidence = 0.229323
DEBUG 5: Weight = 2
DEBUG 5: Term = 0.260882
DEBUG 5:
DEBUG 5: Boundary Distance:
DEBUG 5: ------------------
DEBUG 5: Value = 0
DEBUG 5: Interest = 1
DEBUG 5: Confidence = 1
DEBUG 5: Weight = 4
DEBUG 5: Term = 4
DEBUG 5:
DEBUG 5: Convex Hull Distance:
DEBUG 5: ---------------------
DEBUG 5: Value = 0
DEBUG 5: Interest = 1
DEBUG 5: Confidence = 1
DEBUG 5: Weight = 0
DEBUG 5: Term = 0
DEBUG 5:
DEBUG 5: Angle Difference:
DEBUG 5: -----------------
DEBUG 5: Value = 27.2613
DEBUG 5: Interest = 1
DEBUG 5: Confidence = 0.251052
DEBUG 5: Weight = 1
DEBUG 5: Term = 0.251052
DEBUG 5:
DEBUG 5: Aspect Ratio Difference:
DEBUG 5: ------------------------
DEBUG 5: Value = 0.340972
DEBUG 5: Interest = 0.629275
DEBUG 5: Confidence = 1
DEBUG 5: Weight = 0
DEBUG 5: Term = 0
DEBUG 5:
DEBUG 5: Area Ratio:
DEBUG 5: -----------
DEBUG 5: Value = 0.229323
DEBUG 5: Interest = 0.286654
DEBUG 5: Confidence = 1
DEBUG 5: Weight = 1
DEBUG 5: Term = 0.286654
DEBUG 5:
DEBUG 5: Intersection/Area Ratio:
DEBUG 5: ------------------------
DEBUG 5: Value = 1
DEBUG 5: Interest = 1
DEBUG 5: Confidence = 1
DEBUG 5: Weight = 2
DEBUG 5: Term = 2
DEBUG 5:
DEBUG 5: Curvature Ratio:
DEBUG 5: ----------------
DEBUG 5: Value = 0.94598
DEBUG 5: Interest = 1
DEBUG 5: Confidence = 1
DEBUG 5: Weight = 0
DEBUG 5: Term = 0
DEBUG 5:
DEBUG 5: Complexity Ratio:
DEBUG 5: -----------------
DEBUG 5: Value = 0.338813
DEBUG 5: Interest = 0.423516
DEBUG 5: Confidence = 1
DEBUG 5: Weight = 0
DEBUG 5: Term = 0
DEBUG 5:
DEBUG 5: Percentile (50th) Intensity Ratio:
DEBUG 5: --------------------
DEBUG 5: Value = 0.90625
DEBUG 5: Interest = 1
DEBUG 5: Confidence = 1
DEBUG 5: Weight = 0
DEBUG 5: Term = 0
DEBUG 5:
DEBUG 5: Total Interest = (sum of terms)/(sum of weights*confidence)
DEBUG 5:
DEBUG 5: = 6.79859/7.7097
DEBUG 5:
DEBUG 5: = 0.881823
DEBUG 5:
DEBUG 5:
On Fri, Mar 13, 2020 at 11:50 AM John Halley Gotway <johnhg at ucar.edu>
wrote:
> Hi Marcus,
>
> I see you have questions about the details of MODE's computation of
> interest values. Before diving in, I wanted to point out that
increasing
> the verbosity level to 5 prints out more detailed information about
the
> interest value computations. Unfortunately, these messages are
written to
> standard output (cout), not to the log file. So please try
rerunning
> with "-v 5" and redirect the output to a file.
>
> mode -v 5 ... >& run_mode.log
>
> When it's done you can inspect that log file. We're planning to do
the
> met-9.0 release today, but I'll take a quick look to see if I can
change
> this to route this debug information to MET's log files instead.
Listed
> below is an example of the output you'll see. Note that you'll see
this
> output is printed for all object pair comparisons MODE does... so
lots of
> data to sift through.
>
> Please take a look at these details to see if they explain the
behavior
> you're seeing... or more clearly identify a bug! If questions
remain,
> please let me know.
>
> Thanks,
> John Halley Gotway
>
> Centroid Distance:
> ------------------
> Value = 9.445
> Interest = 0.411
> Confidence = 0.019
> Weight = 2.000
> Term = 0.016
>
> Boundary Distance:
> ------------------
> Value = 0.000
> Interest = 1.000
> Confidence = 1.000
> Weight = 4.000
> Term = 4.000
>
> Convex Hull Distance:
> ---------------------
> Value = 0.000
> Interest = 1.000
> Confidence = 1.000
> Weight = 0.000
> Term = 0.000
>
> Angle Difference:
> -----------------
> Value = 22.201
> Interest = 1.000
> Confidence = 0.576
> Weight = 1.000
> Term = 0.576
>
> Aspect Ratio Difference:
> ------------------------
> Value = 0.115
> Interest = 0.976
> Confidence = 1.000
> Weight = 0.000
> Term = 0.000
>
> Area Ratio:
> -----------
> Value = 0.019
> Interest = 0.024
> Confidence = 1.000
> Weight = 1.000
> Term = 0.024
>
> Intersection/Area Ratio:
> ------------------------
> Value = 0.846
> Interest = 1.000
> Confidence = 1.000
> Weight = 2.000
> Term = 2.000
>
> Curvature Ratio:
> ----------------
> Value = 0.938
> Interest = 1.000
> Confidence = 1.000
> Weight = 0.000
> Term = 0.000
>
> Complexity Ratio:
> -----------------
> Value = 0.146
> Interest = 0.183
> Confidence = 1.000
> Weight = 0.000
> Term = 0.000
>
> Percentile (50th) Intensity Ratio:
> --------------------
> Value = 0.931
> Interest = 1.000
> Confidence = 1.000
> Weight = 0.000
> Term = 0.000
>
> Total Interest = (sum of terms)/(sum of weights*confidence)
>
> = 6.615/7.614
>
> = 0.869
>
>
> On Fri, Mar 13, 2020 at 11:17 AM Johnson, Marcus R. via RT <
> met_help at ucar.edu> wrote:
>
>>
>> Fri Mar 13 11:17:03 2020: Request 94555 was acted upon.
>> Transaction: Ticket created by marcus.johnson at ou.edu
>> Queue: met_help
>> Subject: MODE tool
>> Owner: Nobody
>> Requestors: marcus.johnson at ou.edu
>> Status: new
>> Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=94555 >
>>
>>
>> Good afternoon,
>>
>> My name is Marcus Johnson, and I am currently employing the MODE
tool for
>> rotation and hail research at the University of Oklahoma. I've come
across
>> some inconsistencies looking at MODE object output and my own
interest
>> functions. My total interest function is only weighted using
boundary and
>> centroid distance, with each taking the form = (max_centroid_dist -
>> x)/(max_centroid_dist). max_centroid_dist is defined as
39.0/grid_res (3
>> km), or a unitless 13. These functions are defined in the input
file as
>> follows:
>>
>> centroid_dist = (
>> ( 0.0, 1.0 )
>> ( 0.5*max_centroid_dist, 0.5 )
>> ( max_centroid_dist, 0.0 )
>> );
>>
>> boundary_dist = (
>> ( 0.0, 1.0 )
>> ( 0.5*max_centroid_dist, 0.5 )
>> ( max_centroid_dist, 0.0 )
>> );
>>
>>
>> Since confidence maps are equal to 1 for the preceding attributes,
the
>> total interest function should reflect a linear combination of the
two
>> interest functions equally weighted by 0.5. Some of the .txt output
doesn't
>> reflect this though. At least one forecast cluster object (i.e.,
objects
>> after merging only in their respective fields has taken place)
matches with
>> an observation object with centroid distance of 26.22, which twice
exceeds
>> the max_centroid_dist. Some of the object pairs with greater
centroid and
>> boundary distance have a larger interest than other pairs. I have
attached
>> the sample *obj.txt output file and my input file for completeness.
>>
>> Thank you for your help!
>> Marcus
>>
>>
------------------------------------------------
Subject: RE: [rt.rap.ucar.edu #94555] MODE tool
From: Johnson, Marcus R.
Time: Fri Mar 13 14:47:59 2020
John,
Thank you so much for your help. When looking at MET output (as you
suggested) I noticed that the confidence greatly affected the final
interest score, something I failed to take into account. So the
interests match up to what I was expecting.
I was able to locate relevant interests because of the .txt file that
also gets produced, but I think the following debug information you've
provided (listing object number) would be very helpful. It might also
be helpful delineating between original and merged (cluster) objects
as well.
Thanks!
Marcus
-----Original Message-----
From: John Halley Gotway via RT <met_help at ucar.edu>
Sent: Friday, March 13, 2020 1:46 PM
To: Johnson, Marcus R. <marcus.johnson at ou.edu>
Subject: Re: [rt.rap.ucar.edu #94555] MODE tool
Marcus,
FYI, I just finished updating MODE for the met-9.0 release to write
these messages to the standard log file. I also realized that there
was no indication of which objects were being compared. So I updated
the log messages to include that. Listed below is what these will
look like in met-9.0. Does this look good to you?
Thanks,
John
DEBUG 5:
DEBUG 5:
--------------------------------------------------------------------------------
DEBUG 5:
DEBUG 5: Computing total interest for single forecast object 1 and
single observation object 1 pair.
DEBUG 5:
DEBUG 5: Centroid Distance:
DEBUG 5: ------------------
DEBUG 5: Value = 7.32109
DEBUG 5: Interest = 0.568808
DEBUG 5: Confidence = 0.229323
DEBUG 5: Weight = 2
DEBUG 5: Term = 0.260882
DEBUG 5:
DEBUG 5: Boundary Distance:
DEBUG 5: ------------------
DEBUG 5: Value = 0
DEBUG 5: Interest = 1
DEBUG 5: Confidence = 1
DEBUG 5: Weight = 4
DEBUG 5: Term = 4
DEBUG 5:
DEBUG 5: Convex Hull Distance:
DEBUG 5: ---------------------
DEBUG 5: Value = 0
DEBUG 5: Interest = 1
DEBUG 5: Confidence = 1
DEBUG 5: Weight = 0
DEBUG 5: Term = 0
DEBUG 5:
DEBUG 5: Angle Difference:
DEBUG 5: -----------------
DEBUG 5: Value = 27.2613
DEBUG 5: Interest = 1
DEBUG 5: Confidence = 0.251052
DEBUG 5: Weight = 1
DEBUG 5: Term = 0.251052
DEBUG 5:
DEBUG 5: Aspect Ratio Difference:
DEBUG 5: ------------------------
DEBUG 5: Value = 0.340972
DEBUG 5: Interest = 0.629275
DEBUG 5: Confidence = 1
DEBUG 5: Weight = 0
DEBUG 5: Term = 0
DEBUG 5:
DEBUG 5: Area Ratio:
DEBUG 5: -----------
DEBUG 5: Value = 0.229323
DEBUG 5: Interest = 0.286654
DEBUG 5: Confidence = 1
DEBUG 5: Weight = 1
DEBUG 5: Term = 0.286654
DEBUG 5:
DEBUG 5: Intersection/Area Ratio:
DEBUG 5: ------------------------
DEBUG 5: Value = 1
DEBUG 5: Interest = 1
DEBUG 5: Confidence = 1
DEBUG 5: Weight = 2
DEBUG 5: Term = 2
DEBUG 5:
DEBUG 5: Curvature Ratio:
DEBUG 5: ----------------
DEBUG 5: Value = 0.94598
DEBUG 5: Interest = 1
DEBUG 5: Confidence = 1
DEBUG 5: Weight = 0
DEBUG 5: Term = 0
DEBUG 5:
DEBUG 5: Complexity Ratio:
DEBUG 5: -----------------
DEBUG 5: Value = 0.338813
DEBUG 5: Interest = 0.423516
DEBUG 5: Confidence = 1
DEBUG 5: Weight = 0
DEBUG 5: Term = 0
DEBUG 5:
DEBUG 5: Percentile (50th) Intensity Ratio:
DEBUG 5: --------------------
DEBUG 5: Value = 0.90625
DEBUG 5: Interest = 1
DEBUG 5: Confidence = 1
DEBUG 5: Weight = 0
DEBUG 5: Term = 0
DEBUG 5:
DEBUG 5: Total Interest = (sum of terms)/(sum of weights*confidence)
DEBUG 5:
DEBUG 5: = 6.79859/7.7097
DEBUG 5:
DEBUG 5: = 0.881823
DEBUG 5:
DEBUG 5:
On Fri, Mar 13, 2020 at 11:50 AM John Halley Gotway <johnhg at ucar.edu>
wrote:
> Hi Marcus,
>
> I see you have questions about the details of MODE's computation of
> interest values. Before diving in, I wanted to point out that
> increasing the verbosity level to 5 prints out more detailed
> information about the interest value computations. Unfortunately,
> these messages are written to standard output (cout), not to the log
> file. So please try rerunning with "-v 5" and redirect the output
to a file.
>
> mode -v 5 ... >& run_mode.log
>
> When it's done you can inspect that log file. We're planning to do
> the
> met-9.0 release today, but I'll take a quick look to see if I can
> change this to route this debug information to MET's log files
> instead. Listed below is an example of the output you'll see. Note
> that you'll see this output is printed for all object pair
comparisons
> MODE does... so lots of data to sift through.
>
> Please take a look at these details to see if they explain the
> behavior you're seeing... or more clearly identify a bug! If
> questions remain, please let me know.
>
> Thanks,
> John Halley Gotway
>
> Centroid Distance:
> ------------------
> Value = 9.445
> Interest = 0.411
> Confidence = 0.019
> Weight = 2.000
> Term = 0.016
>
> Boundary Distance:
> ------------------
> Value = 0.000
> Interest = 1.000
> Confidence = 1.000
> Weight = 4.000
> Term = 4.000
>
> Convex Hull Distance:
> ---------------------
> Value = 0.000
> Interest = 1.000
> Confidence = 1.000
> Weight = 0.000
> Term = 0.000
>
> Angle Difference:
> -----------------
> Value = 22.201
> Interest = 1.000
> Confidence = 0.576
> Weight = 1.000
> Term = 0.576
>
> Aspect Ratio Difference:
> ------------------------
> Value = 0.115
> Interest = 0.976
> Confidence = 1.000
> Weight = 0.000
> Term = 0.000
>
> Area Ratio:
> -----------
> Value = 0.019
> Interest = 0.024
> Confidence = 1.000
> Weight = 1.000
> Term = 0.024
>
> Intersection/Area Ratio:
> ------------------------
> Value = 0.846
> Interest = 1.000
> Confidence = 1.000
> Weight = 2.000
> Term = 2.000
>
> Curvature Ratio:
> ----------------
> Value = 0.938
> Interest = 1.000
> Confidence = 1.000
> Weight = 0.000
> Term = 0.000
>
> Complexity Ratio:
> -----------------
> Value = 0.146
> Interest = 0.183
> Confidence = 1.000
> Weight = 0.000
> Term = 0.000
>
> Percentile (50th) Intensity Ratio:
> --------------------
> Value = 0.931
> Interest = 1.000
> Confidence = 1.000
> Weight = 0.000
> Term = 0.000
>
> Total Interest = (sum of terms)/(sum of weights*confidence)
>
> = 6.615/7.614
>
> = 0.869
>
>
> On Fri, Mar 13, 2020 at 11:17 AM Johnson, Marcus R. via RT <
> met_help at ucar.edu> wrote:
>
>>
>> Fri Mar 13 11:17:03 2020: Request 94555 was acted upon.
>> Transaction: Ticket created by marcus.johnson at ou.edu
>> Queue: met_help
>> Subject: MODE tool
>> Owner: Nobody
>> Requestors: marcus.johnson at ou.edu
>> Status: new
>> Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=94555
>> >
>>
>>
>> Good afternoon,
>>
>> My name is Marcus Johnson, and I am currently employing the MODE
tool
>> for rotation and hail research at the University of Oklahoma. I've
>> come across some inconsistencies looking at MODE object output and
my
>> own interest functions. My total interest function is only weighted
>> using boundary and centroid distance, with each taking the form =
>> (max_centroid_dist - x)/(max_centroid_dist). max_centroid_dist is
>> defined as 39.0/grid_res (3 km), or a unitless 13. These functions
>> are defined in the input file as
>> follows:
>>
>> centroid_dist = (
>> ( 0.0, 1.0 )
>> ( 0.5*max_centroid_dist, 0.5 )
>> ( max_centroid_dist, 0.0 )
>> );
>>
>> boundary_dist = (
>> ( 0.0, 1.0 )
>> ( 0.5*max_centroid_dist, 0.5 )
>> ( max_centroid_dist, 0.0 )
>> );
>>
>>
>> Since confidence maps are equal to 1 for the preceding attributes,
>> the total interest function should reflect a linear combination of
>> the two interest functions equally weighted by 0.5. Some of the
.txt
>> output doesn't reflect this though. At least one forecast cluster
>> object (i.e., objects after merging only in their respective fields
>> has taken place) matches with an observation object with centroid
>> distance of 26.22, which twice exceeds the max_centroid_dist. Some
of
>> the object pairs with greater centroid and boundary distance have a
>> larger interest than other pairs. I have attached the sample
*obj.txt output file and my input file for completeness.
>>
>> Thank you for your help!
>> Marcus
>>
>>
------------------------------------------------
Subject: MODE tool
From: John Halley Gotway
Time: Fri Mar 13 15:29:36 2020
Marcus,
Yep, since I last emailed you, I updated those logs messages to be
formatted as shown below. In between each of these lines, you'll see
all
the details for the interest value computations. I omitted those
details
here to save space. These log message match the formatting of the
OBJECT_ID column for pair lines. So hopefully that'll make it easier
to
find the right spot in the log messages.
Glad you were able to make sense of the output!
Thanks,
John
...
DEBUG 5: Computing total interest for single pair F008_O009.
DEBUG 5: Computing total interest for single pair F008_O010.
DEBUG 5: Computing total interest for single pair F008_O011.
DEBUG 5: Computing total interest for single pair F008_O012.
DEBUG 5: Computing total interest for cluster pair CF001_CO001.
DEBUG 5: Computing total interest for cluster pair CF002_CO002.
DEBUG 5: Computing total interest for cluster pair CF003_CO003.
DEBUG 5: Computing total interest for cluster pair CF004_CO004.
...
On Fri, Mar 13, 2020 at 2:48 PM Johnson, Marcus R. via RT
<met_help at ucar.edu>
wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=94555 >
>
> John,
>
> Thank you so much for your help. When looking at MET output (as you
> suggested) I noticed that the confidence greatly affected the final
> interest score, something I failed to take into account. So the
interests
> match up to what I was expecting.
>
> I was able to locate relevant interests because of the .txt file
that also
> gets produced, but I think the following debug information you've
provided
> (listing object number) would be very helpful. It might also be
helpful
> delineating between original and merged (cluster) objects as well.
>
> Thanks!
> Marcus
>
> -----Original Message-----
> From: John Halley Gotway via RT <met_help at ucar.edu>
> Sent: Friday, March 13, 2020 1:46 PM
> To: Johnson, Marcus R. <marcus.johnson at ou.edu>
> Subject: Re: [rt.rap.ucar.edu #94555] MODE tool
>
> Marcus,
>
> FYI, I just finished updating MODE for the met-9.0 release to write
these
> messages to the standard log file. I also realized that there was
no
> indication of which objects were being compared. So I updated the
log
> messages to include that. Listed below is what these will look like
in
> met-9.0. Does this look good to you?
>
> Thanks,
> John
>
> DEBUG 5:
> DEBUG 5:
>
>
--------------------------------------------------------------------------------
>
> DEBUG 5:
> DEBUG 5: Computing total interest for single forecast object 1 and
single
> observation object 1 pair.
> DEBUG 5:
> DEBUG 5: Centroid Distance:
> DEBUG 5: ------------------
> DEBUG 5: Value = 7.32109
> DEBUG 5: Interest = 0.568808
> DEBUG 5: Confidence = 0.229323
> DEBUG 5: Weight = 2
> DEBUG 5: Term = 0.260882
> DEBUG 5:
> DEBUG 5: Boundary Distance:
> DEBUG 5: ------------------
> DEBUG 5: Value = 0
> DEBUG 5: Interest = 1
> DEBUG 5: Confidence = 1
> DEBUG 5: Weight = 4
> DEBUG 5: Term = 4
> DEBUG 5:
> DEBUG 5: Convex Hull Distance:
> DEBUG 5: ---------------------
> DEBUG 5: Value = 0
> DEBUG 5: Interest = 1
> DEBUG 5: Confidence = 1
> DEBUG 5: Weight = 0
> DEBUG 5: Term = 0
> DEBUG 5:
> DEBUG 5: Angle Difference:
> DEBUG 5: -----------------
> DEBUG 5: Value = 27.2613
> DEBUG 5: Interest = 1
> DEBUG 5: Confidence = 0.251052
> DEBUG 5: Weight = 1
> DEBUG 5: Term = 0.251052
> DEBUG 5:
> DEBUG 5: Aspect Ratio Difference:
> DEBUG 5: ------------------------
> DEBUG 5: Value = 0.340972
> DEBUG 5: Interest = 0.629275
> DEBUG 5: Confidence = 1
> DEBUG 5: Weight = 0
> DEBUG 5: Term = 0
> DEBUG 5:
> DEBUG 5: Area Ratio:
> DEBUG 5: -----------
> DEBUG 5: Value = 0.229323
> DEBUG 5: Interest = 0.286654
> DEBUG 5: Confidence = 1
> DEBUG 5: Weight = 1
> DEBUG 5: Term = 0.286654
> DEBUG 5:
> DEBUG 5: Intersection/Area Ratio:
> DEBUG 5: ------------------------
> DEBUG 5: Value = 1
> DEBUG 5: Interest = 1
> DEBUG 5: Confidence = 1
> DEBUG 5: Weight = 2
> DEBUG 5: Term = 2
> DEBUG 5:
> DEBUG 5: Curvature Ratio:
> DEBUG 5: ----------------
> DEBUG 5: Value = 0.94598
> DEBUG 5: Interest = 1
> DEBUG 5: Confidence = 1
> DEBUG 5: Weight = 0
> DEBUG 5: Term = 0
> DEBUG 5:
> DEBUG 5: Complexity Ratio:
> DEBUG 5: -----------------
> DEBUG 5: Value = 0.338813
> DEBUG 5: Interest = 0.423516
> DEBUG 5: Confidence = 1
> DEBUG 5: Weight = 0
> DEBUG 5: Term = 0
> DEBUG 5:
> DEBUG 5: Percentile (50th) Intensity Ratio:
> DEBUG 5: --------------------
> DEBUG 5: Value = 0.90625
> DEBUG 5: Interest = 1
> DEBUG 5: Confidence = 1
> DEBUG 5: Weight = 0
> DEBUG 5: Term = 0
> DEBUG 5:
> DEBUG 5: Total Interest = (sum of terms)/(sum of weights*confidence)
DEBUG
> 5:
> DEBUG 5: = 6.79859/7.7097
> DEBUG 5:
> DEBUG 5: = 0.881823
> DEBUG 5:
> DEBUG 5:
>
>
> On Fri, Mar 13, 2020 at 11:50 AM John Halley Gotway
<johnhg at ucar.edu>
> wrote:
>
> > Hi Marcus,
> >
> > I see you have questions about the details of MODE's computation
of
> > interest values. Before diving in, I wanted to point out that
> > increasing the verbosity level to 5 prints out more detailed
> > information about the interest value computations. Unfortunately,
> > these messages are written to standard output (cout), not to the
log
> > file. So please try rerunning with "-v 5" and redirect the output
to a
> file.
> >
> > mode -v 5 ... >& run_mode.log
> >
> > When it's done you can inspect that log file. We're planning to
do
> > the
> > met-9.0 release today, but I'll take a quick look to see if I can
> > change this to route this debug information to MET's log files
> > instead. Listed below is an example of the output you'll see.
Note
> > that you'll see this output is printed for all object pair
comparisons
> > MODE does... so lots of data to sift through.
> >
> > Please take a look at these details to see if they explain the
> > behavior you're seeing... or more clearly identify a bug! If
> > questions remain, please let me know.
> >
> > Thanks,
> > John Halley Gotway
> >
> > Centroid Distance:
> > ------------------
> > Value = 9.445
> > Interest = 0.411
> > Confidence = 0.019
> > Weight = 2.000
> > Term = 0.016
> >
> > Boundary Distance:
> > ------------------
> > Value = 0.000
> > Interest = 1.000
> > Confidence = 1.000
> > Weight = 4.000
> > Term = 4.000
> >
> > Convex Hull Distance:
> > ---------------------
> > Value = 0.000
> > Interest = 1.000
> > Confidence = 1.000
> > Weight = 0.000
> > Term = 0.000
> >
> > Angle Difference:
> > -----------------
> > Value = 22.201
> > Interest = 1.000
> > Confidence = 0.576
> > Weight = 1.000
> > Term = 0.576
> >
> > Aspect Ratio Difference:
> > ------------------------
> > Value = 0.115
> > Interest = 0.976
> > Confidence = 1.000
> > Weight = 0.000
> > Term = 0.000
> >
> > Area Ratio:
> > -----------
> > Value = 0.019
> > Interest = 0.024
> > Confidence = 1.000
> > Weight = 1.000
> > Term = 0.024
> >
> > Intersection/Area Ratio:
> > ------------------------
> > Value = 0.846
> > Interest = 1.000
> > Confidence = 1.000
> > Weight = 2.000
> > Term = 2.000
> >
> > Curvature Ratio:
> > ----------------
> > Value = 0.938
> > Interest = 1.000
> > Confidence = 1.000
> > Weight = 0.000
> > Term = 0.000
> >
> > Complexity Ratio:
> > -----------------
> > Value = 0.146
> > Interest = 0.183
> > Confidence = 1.000
> > Weight = 0.000
> > Term = 0.000
> >
> > Percentile (50th) Intensity Ratio:
> > --------------------
> > Value = 0.931
> > Interest = 1.000
> > Confidence = 1.000
> > Weight = 0.000
> > Term = 0.000
> >
> > Total Interest = (sum of terms)/(sum of weights*confidence)
> >
> > = 6.615/7.614
> >
> > = 0.869
> >
> >
> > On Fri, Mar 13, 2020 at 11:17 AM Johnson, Marcus R. via RT <
> > met_help at ucar.edu> wrote:
> >
> >>
> >> Fri Mar 13 11:17:03 2020: Request 94555 was acted upon.
> >> Transaction: Ticket created by marcus.johnson at ou.edu
> >> Queue: met_help
> >> Subject: MODE tool
> >> Owner: Nobody
> >> Requestors: marcus.johnson at ou.edu
> >> Status: new
> >> Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=94555
> >> >
> >>
> >>
> >> Good afternoon,
> >>
> >> My name is Marcus Johnson, and I am currently employing the MODE
tool
> >> for rotation and hail research at the University of Oklahoma.
I've
> >> come across some inconsistencies looking at MODE object output
and my
> >> own interest functions. My total interest function is only
weighted
> >> using boundary and centroid distance, with each taking the form =
> >> (max_centroid_dist - x)/(max_centroid_dist). max_centroid_dist is
> >> defined as 39.0/grid_res (3 km), or a unitless 13. These
functions
> >> are defined in the input file as
> >> follows:
> >>
> >> centroid_dist = (
> >> ( 0.0, 1.0 )
> >> ( 0.5*max_centroid_dist, 0.5 )
> >> ( max_centroid_dist, 0.0 )
> >> );
> >>
> >> boundary_dist = (
> >> ( 0.0, 1.0 )
> >> ( 0.5*max_centroid_dist, 0.5 )
> >> ( max_centroid_dist, 0.0 )
> >> );
> >>
> >>
> >> Since confidence maps are equal to 1 for the preceding
attributes,
> >> the total interest function should reflect a linear combination
of
> >> the two interest functions equally weighted by 0.5. Some of the
.txt
> >> output doesn't reflect this though. At least one forecast cluster
> >> object (i.e., objects after merging only in their respective
fields
> >> has taken place) matches with an observation object with centroid
> >> distance of 26.22, which twice exceeds the max_centroid_dist.
Some of
> >> the object pairs with greater centroid and boundary distance have
a
> >> larger interest than other pairs. I have attached the sample
*obj.txt
> output file and my input file for completeness.
> >>
> >> Thank you for your help!
> >> Marcus
> >>
> >>
>
>
>
>
------------------------------------------------
More information about the Met_help
mailing list