I am in process of implementing incremental strong constraint ROMS-4DVAR (I4D-Var) for the operational forecast of Indian Ocean. I have successfully tested the WC13 configuration in our cluster. Now I am trying to setup ROMS-4DVAR for the Indian Ocean with 1/12 deg resolution and 40 vertical levels.
In this regards, I have downloaded MetOffice Levitus et al profiles data (Temp and Salt) from http://hadobs.metoffice.com/en4/download-en4-2-1.html and prepared ROMS observations for Aug 2016 with Matlab script d_ts_metoffice.m provided with ROMS. I also prepared monthly standard deviation files for initial condition, surface forcing and boundary condition with d_std.m, d_std_frc.m and d_std_bry.m respectively. I made sure that the ocean_time for std files and obs_time/survey_time for obs file match with that of the model initial file.
Then I moved to Normalization directory and computed the error covariance normalization factors using the randomization approach.
Now I am trying to run I4DVAR for 15 days with the initial file on 15-AUG-2016. But here I am surprised to see that the model is not assimilating the observations at all!
It is running like ROMS without any assimilation. The output it gives are io_fwd_000.nc, io_rst.nc and io_mod.nc only. The ROMS log file (4dvar.out.dat, attached) shows that it doesn't even read the observations. Though it is reading those STD and NRM files.Is the format of ROMS observation file correct (attached a ncdump'ed file of obs, io_ts_super_obs.nc.in)? Or the procedure I followed is wrong somewhere? I have attached necessary scripts and outputs for your references. Any kind of help or advice would be highly appreciated. Thanks in advance.
Biswamoy Paul
