[Dart-dev] DART observation-space rank histogram ready for use.

Tim Hoar thoar at ucar.edu
Fri Oct 15 12:35:18 MDT 2010


I have just committed changes to the obs_diag.f90 program that relate
to the creation of the rank histogram.

1) I added a namelist variable for obs_diag_nml:

    outliers_in_histogram = .false.

    .false. means that DART QC == 7 observations (those rejected by  
the outlier
    threshold) will NOT be considered during the creation of the rank  
histogram.

    .true. means they WILL be considered - and you should expect to  
see more mass
    on the first and last bins ...

2) The matlab script 'DART/diagnostics/matlab/plot_rank_histogram.m'  
will read
    the global attribute "outliers_in_histogram" and change the Y axis  
label
    to reflect whether or not the outliers were used. If they ARE  
used, the label
    will read: 'count - including outlier observations'; if they are  
not used,
    the label will read: 'count'.

Happy Ranking ...

Tim

On Oct 8, 2010, at 4:54 PM, Tim Hoar wrote:

> I extended obs_diag.f90 to also output rank histogram information if
> there are any 'prior ensemble member xxx' copies in the  
> obs_seq.final files.
> (They exist if you specify  
> input.nml:filter_nml:num_output_obs_members == <ens_size>.)
>
> After you create the obs_diag_output.nc file, you can run
>
> DART/diagnostics/matlab/plot_rank_histogram.m
>
> to view the results. It has the 'expected' help file - available  
> from the matlab
> command line.
>
> At present, ALL observations with DART QC 0,1,2,3, and 7 are used to  
> create the
> rank histogram. 7 is the outlier observation, so even these are  
> being 'counted'.
> The observation error -as specified in each observation- is used to  
> generate
> the sampling error signal for the ensemble members' expected  
> observations that
> define the bin edges for the rank histogram.
>
> obs_diag.f90 has also been updated to provide better input guidance  
> for the
> instance when the user has specified a time window that does not  
> intersect with
> the observations in the files specified.
>
> Please
>
> svn update DART/diagnostics
>
> rebuild your obs_diag, and let me know how I can improve it. I have  
> one idea already,
> so stay tuned for a minor tweak to the matlab script on Monday.
>
> Tim Hoar, Associate Scientist
> National Center for Atmospheric Research
> thoar at ucar.edu
> 303 497 1708
>

Tim Hoar, Associate Scientist
National Center for Atmospheric Research
thoar at ucar.edu
303 497 1708

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.ucar.edu/pipermail/dart-dev/attachments/20101015/6d7e4199/attachment.html 


More information about the Dart-dev mailing list