[Met_help] MET MODE Analysis - No Centroid X and Y Data

Tue Oct 27 08:10:09 MDT 2009

Bob,

It's all in that one "mode_*_obj.txt" file.

Let's step through an example.  Please take a look in the directory "METv2.0/out/mode" at the PostScript file named:
mode_APCP_12_SFC_vs_APCP_12_SFC_120000L_20050807_120000V_120000A.ps

Also, open up the corresponding stats file named:
mode_APCP_24_SFC_vs_APCP_24_SFC_240000L_20050808_000000V_240000A_obj.txt

In this example, there are 7 simple forecast objects and 4 simple observed objects.  Take a look in the column labeled "OBJECT_ID" in the stats file.  The first 7 lines, named "F001...F007" contain
information about the 7 simple forecast objects.  The next 4 lines, named "O001...O004" contain information about the 4 simple observed objects.

The next 10 lines, with "F00A_O00B" in the OBJECT_ID column, contain information about the pairs of simple objects, where A goes from 1 to 7 and B goes from 1 to 4.

The next 2 lines, named "CF001...CF002", contain information about the "cluster" forecast objects.  And the next 2 lines, named "CO001...CO002", contain information about the "cluster" observation
objects.

And the last 2 lines, named "CF00A_CO00B", contain information about the pairs of cluster objects.

Are you interested in the differences between pairs of SIMPLE objects or pairs of CLUSTER objects?

For pairs of simple objects, you'll need to parse the lines named "F00A_O00B" in the OBJECT_ID column.  Let's look at the simple object pair "F006_O004" - those are both green objects in the
PostScript plot.  Look at the line with "F006_O004" for the CENTROID_DIST value. For "F006_O004", that distance is 7.35.  Next, you'll parse that string "F006_O004" down into the simple objects being
compared "F006" and "O004".  You'll look in the OBJECT_ID column for those lines and find that "F006" has a centroid (x, y) of (72.71, 39.69).  And that "O004" has a centroid (x, y) of (65.39
40.33).  If you do the math, the distance between those two centroids does work out to be 7.35 grid squares.  So you could use those centroid (x, y) values to compute the x-offset and y-offset.  In
this case, the forecast object was offset 7.32 grid square in the x-direction and -0.64 grid squares in the y-direction.

For pairs of cluster objects, you'd do the exact same thing, but using the "CF00A_CO00B" pair lines and the "CF00A" and "CO00B" single object lines.  In this example, "F006" belongs to forecast
cluster object "CF002" and "O004" belongs to observed cluster object "CO002".  Looking at the line "CF002_CO002", the CENTROID_DIST is 6.46.  And looking at the lines for "CF002" and "CO002", their
centroids are (72.71, 39.69) and (66.70, 42.04).  So for this pair of cluster objects, the x-offset is 6.01 and the y-offset is -2.35.

Hope that helps.

John Halley Gotway

Craig, Robert J Civ USAF AFWA 2 WXG/WEA wrote:
> John, thanks for the information.  I am using IDL and can write a
> program to extract the data.  What file do I need to read to get the
> pair object IDs which I can then use to find the corresponding objects
> in the obj.txt file?  
> 
> Thanks
> 
> Bob Craig
> 
> -----Original Message-----
> From: John Halley Gotway [mailto:johnhg at ucar.edu] 
> Sent: Friday, October 16, 2009 9:10 AM
> To: Craig, Robert J Civ USAF AFWA 2 WXG/WEA
> Cc: met_help at ucar.edu
> Subject: Re: [Met_help] MET MODE Analysis - No Centroid X and Y Data
> 
> Robert,
> 
> It sounds like you're interested not just in the distance between the
> centroids of objects, but how that distance is broken down into it's x
> and y components.  That's a very reasonable question to
> ask.  Unfortunately though, the MODE-Analysis tool won't give you the
> answer directly.  Here's what's going on...
> 
> The "centroid_dist" column contains the distance between the centroids
> of a pair of objects.  So it only applies to "pair" lines in the MODE
> output.
> The "centroid_x" and "centroid_y" columns contain the centroid (x, y)
> for a single object. So it only applies to "single" lines in the MODE
> output.
> 
> In your MODE-Analysis command line, you used "-simple -pair" to select
> out only the MODE lines for pairs of simple objects.  Since the
> "centroid_x" and "centroid_y" columns don't apply to "pair"
> lines, there's no data to summarize.  It'd probably be better if were to
> dump out "NA" instead of "0" in the summary when the column doesn't
> apply.
> 
> It still is possible to extract the information you want from the MODE
> output files, but you'll have to do a little more work to get it.  For
> each simple pair line in the MODE output, you could look
> at the "object_id" column to extract out the simple forecast and simple
> observation that make up this pair.  Then you could find the simple
> single line for the forecast object and the one for the
> observed object and extract their "centroid_x" and "centroid_y"
> values... and then take the difference.
> 
> I've done this sort of thing in the past using an R script.  If you
> happen to have R installed on your machine, I could probably put
> together a script for you to get at this info.  Would that work for
> you?
> 
> John
> 
> Craig, Robert J Civ USAF AFWA 2 WXG/WEA wrote:
>> I am running MET MODE on cloud forecasts using a gridded cloud
> analysis.
>> I am running mode_analysis on several weeks worth of cloud data using
>> the following command line:
>>
>>  
>>
>> bin/mode_analysis -lookin out/mode/t04 -summary -simple -pair -column
>> CENTROID_DIST -column CENTROID_X -column CENTROID_Y -column area_ratio
>> -out bob_test -dump_row dump_file
>>
>>  
>>
>>  
>>
>> In the output file I have data for the centroid distance but only 0s
> for
>> the centroid_x and y values.  I want to be able to tell what the
> typical
>> direction is of the centroid displacement.  Am I using the write
> command
>> line arguments?
>>
>>  
>>
>> Thanks
>>
>>  
>>
>> Bob Craig
>>
>> 402-294-3186
>>
>>  
>>
>>  
>>
>>  
>>
>>  
>>
>>
>>
>>
>>
> ------------------------------------------------------------------------
>> _______________________________________________
>> Met_help mailing list
>> Met_help at mailman.ucar.edu
>> http://mailman.ucar.edu/mailman/listinfo/met_help