I'm a beginner in 4dvar and would like to do minimalistic forecast cycle
impact/sensitiviy runs. Apart from subsetting the different observation types,
I'd like to compare the effect of using the balance operator.
Now I have a question regarding the matlab preprocessing scripts to
computed unbalanced standard deviation. I don't have matlab
so I can't run the code, but in d_std_unbalanced.m it says:
% Get basic state to use in the balance operator. Read in selected record
e.g to compute the thermal expansion coefficients for "unbalancing" zeta
the instantaneous timestep from the original history file is taken, i.e.
the one to compute the anomaly from the mean.
Shouldn't the "basic" state in this case be instead the average fields?
In Weaver et. al. 2006, they take the previous outer loop (k-1).
These are some relevant functions:
In d_std_unbalanced.m, the call
[K] = balance_4dvar(A);
is made with
A.HisTimeRec = rec; % needed in balance_4dvar
In balance_4dvar.m, it says
% A.HisTimeRec History NetCDF time record to use in the
% computation of thermal expansion and
% saline contraction coefficients (integer)
and the call
[K]=ini_balance(A.Gname,A.Hname,A.HisTimeRec);
sets the expansion coefficients.