[Met_help] [rt.rap.ucar.edu #60134] History for Using Mode Tool to centroid distance between forecast and observed area of precipitation

John Halley Gotway via RT met_help at ucar.edu
Mon Feb 4 09:09:17 MST 2013


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

Hello,

I am attempting to analyze model performance, specifically in terms of
comparing the precipitation forecast versus observed areas of precipitation
using the Mode Tool. I am interested in knowing the distance between the
centroids of the forecast area of interest and the observed area of
interest, but I don't see a way to output a non-normalized number for that.
Is there an output flag for this, or a way that I can obtain this distance
from the output generated by the Mode Tool?

Thanks,

--
Zachary Ebenstein
Assistant in Research
University of Michigan
Dept. Atmospheric, Oceanic and Space Sciences


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

Subject: Re: [rt.rap.ucar.edu #60134] Using Mode Tool to centroid distance between forecast and observed area of precipitation
From: John Halley Gotway
Time: Fri Feb 01 08:59:59 2013

Zachary,

I'll be happy to help you with this question, but first I need you to
fill out the free online registration for MET.  We are only able to
provide direct support to users who have filled out the
registration form in order to be sure that we remain in compliance
with the international export control laws of the United States.

Just go to the MET downloads page, enter your email address, and hit
enter:

    http://www.dtcenter.org/met/users/downloads/index.php

Please write me back to let me know that you've completed the
registration, and we can work on your issue.

Thanks,
John Halley Gotway
met_help at ucar.edu

On 02/01/2013 05:48 AM, Zachary Ebenstein via RT wrote:
>
> Fri Feb 01 05:48:15 2013: Request 60134 was acted upon.
> Transaction: Ticket created by ebenszac at umich.edu
>         Queue: met_help
>       Subject: Using Mode Tool to centroid distance between forecast
and observed area of precipitation
>         Owner: Nobody
>    Requestors: ebenszac at umich.edu
>        Status: new
>   Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=60134 >
>
>
> Hello,
>
> I am attempting to analyze model performance, specifically in terms
of
> comparing the precipitation forecast versus observed areas of
precipitation
> using the Mode Tool. I am interested in knowing the distance between
the
> centroids of the forecast area of interest and the observed area of
> interest, but I don't see a way to output a non-normalized number
for that.
> Is there an output flag for this, or a way that I can obtain this
distance
> from the output generated by the Mode Tool?
>
> Thanks,
>
> --
> Zachary Ebenstein
> Assistant in Research
> University of Michigan
> Dept. Atmospheric, Oceanic and Space Sciences
>

------------------------------------------------
Subject: Using Mode Tool to centroid distance between forecast and observed area of precipitation
From: Zachary Ebenstein
Time: Fri Feb 01 09:08:08 2013

John,

I have filled out the registration form.

Zac

On Fri, Feb 1, 2013 at 10:59 AM, John Halley Gotway via RT <
met_help at ucar.edu> wrote:

> Zachary,
>
> I'll be happy to help you with this question, but first I need you
to fill
> out the free online registration for MET.  We are only able to
provide
> direct support to users who have filled out the
> registration form in order to be sure that we remain in compliance
with
> the international export control laws of the United States.
>
> Just go to the MET downloads page, enter your email address, and hit
enter:
>
>     http://www.dtcenter.org/met/users/downloads/index.php
>
> Please write me back to let me know that you've completed the
> registration, and we can work on your issue.
>
> Thanks,
> John Halley Gotway
> met_help at ucar.edu
>
> On 02/01/2013 05:48 AM, Zachary Ebenstein via RT wrote:
> >
> > Fri Feb 01 05:48:15 2013: Request 60134 was acted upon.
> > Transaction: Ticket created by ebenszac at umich.edu
> >         Queue: met_help
> >       Subject: Using Mode Tool to centroid distance between
forecast and
> observed area of precipitation
> >         Owner: Nobody
> >    Requestors: ebenszac at umich.edu
> >        Status: new
> >   Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=60134 >
> >
> >
> > Hello,
> >
> > I am attempting to analyze model performance, specifically in
terms of
> > comparing the precipitation forecast versus observed areas of
> precipitation
> > using the Mode Tool. I am interested in knowing the distance
between the
> > centroids of the forecast area of interest and the observed area
of
> > interest, but I don't see a way to output a non-normalized number
for
> that.
> > Is there an output flag for this, or a way that I can obtain this
> distance
> > from the output generated by the Mode Tool?
> >
> > Thanks,
> >
> > --
> > Zachary Ebenstein
> > Assistant in Research
> > University of Michigan
> > Dept. Atmospheric, Oceanic and Space Sciences
> >
>
>


--
Zachary Ebenstein
University of Michigan
College of Engineering
Dept. Atmospheric, Oceanic and Space Sciences
Climate Physics (B.S.E), Dec. 2011

------------------------------------------------
Subject: Re: [rt.rap.ucar.edu #60134] Using Mode Tool to centroid distance between forecast and observed area of precipitation
From: John Halley Gotway
Time: Fri Feb 01 09:56:50 2013

Zac,

Great, thanks for filling that out.

Your question is about the distance between the centroids of the
forecast and observation objects in MODE.  That distance is listed in
the column named CENTROID_DIST in the ascii output of MODE.
Using the output of the MODE test scripts as an example, please look
in:
    METv4.0/out/mode/mode_120000L_20050807_120000V_120000A_obj.txt

In line 13 of that file the OBJECT_ID column contains F001_O001,
indicating that this is a comparison of forecast object number 1 to
observation object number 1.  The CENTROID_DIST column contains the
value 14.146 indicating that their centroids are that many grid units
away from one another.

All of the computations in MODE are done in grid space, rather than
earth coordinates.  Therefore, all of the measures of distance and
area are in grid units.  For most grids, that's fine, but for
grids that get very distorted, it can be problematic when the true
area of each grid box varies a lot.  We did at one point consider
doing computations in earth coordinates, rather than grid
coordinates, but realized that it'd add another level of complexity to
an already complex method, and slow down an already slow process.  So
we decided not to.

So there really is not way for MODE to dump out distances or areas in
earth units directly.

However, we are dumping out the lat/lon of the centroid locations for
this very reason.  You could use those lat/lon's to compute true
distances if you'd like.

In the above example, you could retrieve the centroid lat,lon for F001
and O001 from lines 2 and 8, respectively.  They are (33.224, -82.998)
and (33.214, -83.988).  And then use whatever method you'd
like to compute the distance between those points.

I do admit though that having to look across multiple lines is a bit
cumbersome.  If you happen to be familiar with R, there is an Rscript
included in the MET tarball that reads in MODE ASCII output
and does additional analysis on it.  You may be able to adapt it to
dump out the additional information you're after.  Listed below is an
example of running it on our example data file.

Hope that helps.

Thanks,
John Halley Gotway
met_help at ucar.edu

[johnhg at rambler]% Rscript METv4.0/scripts/Rscripts/mode_summary.R
METv4.0/out/mode/mode_120000L_20050807_120000V_120000A_obj.txt
Read 22 lines from MODE file:
METv4.0/out/mode/mode_120000L_20050807_120000V_120000A_obj.txt
Total Number of Files Processed = 1
Total Number of Single Objects = 11
Number of Single Fcst Objects = 6
Number of Matched Single Fcst Objects = 2
Number of Unmatched Single Fcst Objects = 4
Number of Single Obs Objects = 5
Number of Matched Single Obs Objects = 2
Number of Unmatched Single Obs Objects = 3
Total Number of Cluster Objects = 2
Total Number of Fcst Cluster Objects = 1
Total Number of Obs Cluster Objects = 1
Total Area of Objects = 1476
Area of Single Fcst Objects = 882
Area of Matched Single Fcst Objects = 539
Area of Unmatched Single Fcst Objects = 343
Area of Single Obs Objects = 594
Area of Matched Single Obs Objects = 579
Area of Unmatched Single Obs Objects = 15
Median of Max Interest by Fcst Object = 0.556115
Median of Max Interest by Obs Object = 0.47733
Median of Max Interest by Fcst+Obs Objects = 0.50583
Cluster Pair Centroid Distance:
    Min. 1st Qu.  Median    Mean 3rd Qu.    Max.
   6.719   6.719   6.719   6.719   6.719   6.719
Cluster Pair Centroid X-Offsets:
    Min. 1st Qu.  Median    Mean 3rd Qu.    Max.
  0.3018  0.3018  0.3018  0.3018  0.3018  0.3018
Cluster Pair Centroid Y-Offsets:
    Min. 1st Qu.  Median    Mean 3rd Qu.    Max.
   6.712   6.712   6.712   6.712   6.712   6.712
Cluster Pair Centroid X,Y Offsets and Distance [1]:
(0.3018, 6.7119) 6.7187



On 02/01/2013 09:08 AM, Zachary Ebenstein via RT wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=60134 >
>
> John,
>
> I have filled out the registration form.
>
> Zac
>
> On Fri, Feb 1, 2013 at 10:59 AM, John Halley Gotway via RT <
> met_help at ucar.edu> wrote:
>
>> Zachary,
>>
>> I'll be happy to help you with this question, but first I need you
to fill
>> out the free online registration for MET.  We are only able to
provide
>> direct support to users who have filled out the
>> registration form in order to be sure that we remain in compliance
with
>> the international export control laws of the United States.
>>
>> Just go to the MET downloads page, enter your email address, and
hit enter:
>>
>>      http://www.dtcenter.org/met/users/downloads/index.php
>>
>> Please write me back to let me know that you've completed the
>> registration, and we can work on your issue.
>>
>> Thanks,
>> John Halley Gotway
>> met_help at ucar.edu
>>
>> On 02/01/2013 05:48 AM, Zachary Ebenstein via RT wrote:
>>>
>>> Fri Feb 01 05:48:15 2013: Request 60134 was acted upon.
>>> Transaction: Ticket created by ebenszac at umich.edu
>>>          Queue: met_help
>>>        Subject: Using Mode Tool to centroid distance between
forecast and
>> observed area of precipitation
>>>          Owner: Nobody
>>>     Requestors: ebenszac at umich.edu
>>>         Status: new
>>>    Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=60134 >
>>>
>>>
>>> Hello,
>>>
>>> I am attempting to analyze model performance, specifically in
terms of
>>> comparing the precipitation forecast versus observed areas of
>> precipitation
>>> using the Mode Tool. I am interested in knowing the distance
between the
>>> centroids of the forecast area of interest and the observed area
of
>>> interest, but I don't see a way to output a non-normalized number
for
>> that.
>>> Is there an output flag for this, or a way that I can obtain this
>> distance
>>> from the output generated by the Mode Tool?
>>>
>>> Thanks,
>>>
>>> --
>>> Zachary Ebenstein
>>> Assistant in Research
>>> University of Michigan
>>> Dept. Atmospheric, Oceanic and Space Sciences
>>>
>>
>>
>
>

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


More information about the Met_help mailing list