[Mitgcm-support] Note: December 7, 2000

[mitgcm-support at dev.mitgcm.org]

HyperNews notification of new message.  See: 
  http://escher.JPL.NASA.GOV:2000/HyperNews/get/forums/wkmtg/75.html

     =======================================================================     
     
NOTES FROM THE ASSIMILATION MEETING : 7 December 2000    

1) K-filter
------------
   - Calibration of the BT and BC filters are proceeding.  The BT
     filter will be redone this weekend.  In calibrating for the BC
     filter, the model seems to show that the first baroclinic mode
     may not account for as much sea level variability as it did with
     MOM.  This may be due to the difference in forcing function;
     momentum is input as body force in the top most layer, and MOM
     had a thicker first layer (50m) than our model (10m). 

2) Adjoint
-----------
   - The 1-year run has not begun.  Instead, additional sensitivity
     experiments are being conducted with the 3-month assimilation. 

   - Timing of TAMC at SIO and MIT has been conducted.  Timing on both
     machines are comparable with that on alhena.  Compilation takes
     approximately 5-minutes.  Detlef had reported it taking 45
     minutes at MIT.  The difference might be with Detlef's code
     having a lot of dependencies associated with various packages and
     the more complete cost function than what we are using.  The
     timing results have been reported to Detlef and Carl.  The
     communication time between JPL and SIO are OK, but too slow
     between JPL and MIT.

   - TAF still has a problem and does not work as does TAMC.  The
     problem was reported to Ralf. 

   - The ECCO code had a bug in SST relaxation.  The relaxation
     coefficient had the wrong sign.  Now the ECCO code with KPP runs
     for 90-days.  More testing is continued to see if correction of
     this solves the ECCO code bombing. 

3) Data comparison
-------------------
  - A thread[1] has been initiated on the forum for comparisons of
    assimilated estimates and observations.  

  - 20% of the TAO moorings are more than 20km away from their nominal
    location.  Comparisons would need to be performed at actual
    locations and not take nominal locations for granted.  Some
    of this "movement" is due to moorings set at slightly
    different locations and others due to moorings breaking loose.  

  - Comparison between bottom pressure data and the BT K-filter
    (kf019) show improvements in most locations, except Drake Passage.
    Only data from GLOUP are compared.  There is another set that was
    used in the MOM assimilation (from Ray Peterson).  

  - Comparison with TAO T show improvements with the K-filter
    (kf017).  However, there is little improvement in velocity.  

  --> Both kf017 and kf019 are baseline assimilation using NCEP wind
  covariance as model process noise.  There is no attempt of matching
  the resulting simulation error with prior error estimates or to tune
  the performance of the filter.  Calibration is ongoing.  (item 1
  above.)

4) Action Items 
----------------
  - Larry Eversole is still working to get our Ames account. 
    
=======================================================================     

-----------------------------------------------------------------------
[1] http://escher.JPL.NASA.GOV:2000/HyperNews/get/forums/assim/106.html