[Met_help] Mode Analysis Question

[John Halley Gotway]

Bob,

Using MODE-Analysis, I believe that all you're getting is an average distance between the centroids of the matched objects.  Those values are contained in the "CENTROID_DIST" column.  The X,Y
locations of those centroids are contained in the output file, so one could compute the X and Y displacements, but MODE-Analysis isn't actually doing that.

Try running the following type of job:

   mode_analysis -lookin "your mode output files" \
   -summary -column centroid_dist -pair -cluster \
   -interest_min 0.7

This job will look through all the MODE output and pick out the lines that correspond to pairs of matched cluster objects.  From those lines, it'll extract the CENTROID_DIST column and give you
summary info about those values.  All the values in these lines correspond to matched clusters of objects - otherwise there wouldn't be an output line.  Next, try increasing the "-interest_min" value
from 0.7 to 0.8 and higher.  As you increase the minimum interest value, that'll select better and better matches.

You could also look at additional columns from those lines, like: BOUNDARY_DIST, AREA_RATIO, INTERSECTION_AREA, ...

Just let us know if you need further clarification.  Also, Randy wanted me to mention that the config file GUI should now be fixed.  Please give it a try and let us know if it's working with your
version of Internet Explorer:
   http://verif.rap.ucar.edu/cgi/metgui/base.cgi

Thanks,
John Halley Gotway
met_help at ucar.edu

Craig, Robert J Civ USAF AFWA 16 WS/WXN wrote:
> We are using mode to get a feel for what errors we get in forecast cloud object positioning compared to ground truth position.  Running mode analysis we get a average centroid x and y displacements.  I am assuming that this is the average of all objects.  If this is true then we will tend to get noisy results since some matched objects aren't matched well (low interest) compared to others.  Is there a way to filter the matched objects x and y displacements so that the average is only for the better object matches?
> 
> 
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> Met_help mailing list
> Met_help at mailman.ucar.edu
> http://mailman.ucar.edu/mailman/listinfo/met_help