[Met_help] Re: mode analysis tool question

[John Halley Gotway]

Bill,

I've found that using the configuration file is a bit overkill for running the MODE-Analysis tool.  Instead, you can just specify your filtering parameters on the command line.  If you'd like to do 
more complex filtering, like conditioning on sizes of objects or locations or time intervals, the command line may get cumbersome, and you may want to use the config file.  For simple queries though, 
the command line options should suffice.

And the "-fcst" (as opposed to "-obs") command line argument should definitely work.

For example, on Ebony go to: /d1/johnhg/MET_releases/METv1.1beta6.

 From that directory, run the following two commands:
bin/mode_analysis -lookin out/mode/mode_APCP_12_SFC_vs_APCP_12_SFC_120000L_20050807_120000V_120000A_obj.txt -summary -column AREA -fcst -simple -single -dump_row ~/fcst_simple_single.txt
bin/mode_analysis -lookin out/mode/mode_APCP_12_SFC_vs_APCP_12_SFC_120000L_20050807_120000V_120000A_obj.txt -summary -column AREA -obs  -simple -single -dump_row ~/obs_simple_single.txt

The first looks at the "AREA" column for the "fcst, simple, single" MODE output lines (i.e. lines where the OBJECT_ID column looks like FNNN, where NNN is a number starting at 001).
The second looks at the "AREA" column for the "obs, simple, single" MODE output lines (i.e. lines where the OBJECT_ID column looks like ONNN, where NNN is a number starting at 001).

You can look at the files specified using the "-dump_row" command line argument to make sure that MODE-Analysis processed the correct sets of MODE lines.

To summarize additional columns, just add another "-column NAME" command line argument, where NAME is the column header of interest.

For more examples, let me refer you to the draft version of the online tutorial for METv1.1: http://www.dtcenter.org/met/users/support/online_tutorial/index.php

Take a look at the section on the MODE-Analysis tool.  In particular, take a look at the "MODE-Analysis: Configuration" page for an explanation of the -fcst vs -obs, -single vs -pair, -simple vs 
-composite, and -matched vs -unmatched toggle options.  I think that's the most confusing part of the MODE-Analysis tool!

And while you're at it, please let me know if you have any comments or suggestions for the tutorial :)

Thanks,
John

wgallus at iastate.edu wrote:
> John,
> 
> I'm trying to follow the PDF file with instructions about mode_analysis, 
> but have a question.  I found your test_mode_analysis script, and am 
> trying to play with it.  I want it to summarize some columns (variables) 
> for just the forecasted objects.   I see your test script pointed to a 
> configuration file that was all 0s, and thus my first attempt produced 
> all zeroes.  The PDF file makes it seem like I should be able to put a
> -fcst
> entry in my call to mode_analysis, but it complains when I do that.
> 
> Am I correct that one way to do this would be to put a 1 instead of a 0 
> in the configuraiton file?  The problem there, is, that your script also 
> then tries to get some stats just for objects, and it calls the same 
> configuration file.  It would seem to me then that I'd have to have a 
> dozen different configuration files that only differ by one or two 
> entries.    That is why I'd like to see if I could tell it to just do 
> forecast objects or just do observed ones from a command line call.
> 
> I tried putting -fcst 1   but it didn't like that.   I'm afraid this 
> part of the PDF document is not very clear, or else I just get easily 
> confused.
> 
> Bill
> 
> ------------------------------------------------
> |   William A. Gallus, Jr.                       |
> |   Professor of Meteorology                     |
> |   Professor-in-charge, Meteorology Program     |
> |   3025 Agronomy                                |
> |   Iowa State University                        |
> |   515-294-2270                                 |
>  ------------------------------------------------