Dear all,
I have set up my roms to model the California Current System. So far I've been able to run the non-linear model, the tangent linear model and the adjoint model. Actually I want to calculate the Singular Vector Decomposition, but I don't know how to perform such calculation.
Can someone please tell me how to do that, i.e. wich cpp routine have to be defined in cppdefs.h, and how are the input file required for such calculation.
Thank you very much,
Cesare Mosca
Singular Vector Calculation
You need to define OPT_PERTURBATION in cppdefs (and FORWARD_MIXING if you want to read vertical mixing coefficients from the forward model history files).
As input files you really need only the forward model trajectory (FWDNAME=ocean_his.nc), but ROMS requires the following files anyway: the external forcing, the tangent, and adjoint initial condition file (ITLNAME and IADNAME). You can just set the last 2 to zeros because they are not really needed for the SVD calculation.
The other necessary files are the same as in your forward model run (boundary conditions, climatology, etc).
In the input file, these are the other key things to specify:
NEV (start with a small number to get an idea on how long it takes to converge)
NCV
MaxIterGST
Ritz_tol
As output, the singular vectors are in ocean_tlm.nc, in the following order (assume NEV=2):
mode 2, initial condition
mode 2, final state (initial condition evolved through the tangent linear)
mode 1, initial condition
mode 1, final state
So, if you are interested in knowing the optimal initial condition perturbation (for the particular integration time you chose), you have to look at record 3 and 1 for this example.
Milena Veneziani
As input files you really need only the forward model trajectory (FWDNAME=ocean_his.nc), but ROMS requires the following files anyway: the external forcing, the tangent, and adjoint initial condition file (ITLNAME and IADNAME). You can just set the last 2 to zeros because they are not really needed for the SVD calculation.
The other necessary files are the same as in your forward model run (boundary conditions, climatology, etc).
In the input file, these are the other key things to specify:
NEV (start with a small number to get an idea on how long it takes to converge)
NCV
MaxIterGST
Ritz_tol
As output, the singular vectors are in ocean_tlm.nc, in the following order (assume NEV=2):
mode 2, initial condition
mode 2, final state (initial condition evolved through the tangent linear)
mode 1, initial condition
mode 1, final state
So, if you are interested in knowing the optimal initial condition perturbation (for the particular integration time you chose), you have to look at record 3 and 1 for this example.
Milena Veneziani