[Dart-dev] obs_diag update - important - please read
thoar at ucar.edu
Thu Feb 7 10:46:57 MST 2008
Perhaps the subject line should be 'obs_diag bugfix pending' ...
For the purpose of this email
'old' obs_diag is anything before revision 3180.
'new' obs_diag is 3180 and later ... all of which output the netCDF
Only the new version tracks the U, V wind components individually ...
they are (still) correct.
The new obs_diag has what I consider a 'bug' in that it tracks
statistics of the
scalar wind speed rather than the vector wind velocity. The old
designed to compare to NCEP and tracks the vector wind velocity.
The sticky point is the old obs_diag accurately called this the
and the new obs_diag calculates the speed and accurately calls it
Clearly ... comparisons between the old and the new are not appropriate.
Since the primary intent of this routine is to allow accurate
comparisons with NCEP,
I will be changing obs_diag to track the vector wind velocity and
create a netCDF variable
called 'WIND_VELOCITY' (I'm dropping the HORIZONTAL part because the
names are so long already).
There are some other enhancements in obs_diag that will allow a
to the plot_observation_locations.m script, the ability to ingest
that have not been assimilated, and the ability to define regions
that span the prime meridian.
What the heck ... I might even update the documentation. So stay
tuned - I have not completed
testing on all the platforms but feel this information needs to go
Please also note that the ONLY rejection mechanism in the new
obs_diag is based
on the DART QC flag. The old version could be used in a misleading
fashion by lowering the
value of 'rat_cri' (a namelist variable) so that any observations
whose increments were greater
than 'rat_cri' were summarily rejected by obs_diag - regardless of
whether they were assimilated or not.
Before the DART QC flag existed, this was the only way we had to
discriminate 'good' observations from
Tim Hoar, Associate Scientist
National Center for Atmospheric Research
thoar at ucar.edu
303 497 1708
More information about the Dart-dev