assimilate TSLA into the restart fields of my ROMS model

[xiaozhu557]

I'm trying to use EnKF to assimilate TSLA(along-track Sea Level Anomaly) data into the restart fields of my ROMS model.
I have defined the

Code: Select all

PERFECT_RESTART

option while compling ROMS. And my assimilation is offline. My CPP options associated with momentum equations is

UV_U3HADVECTION,
UV_C4VADVECTION

.
Because the variable zeta has been saved three time steps information with the form as zeta(ocean_time, three, eta_rho, xi_rho) in the restart file. My question is which time step of zeta in the restart fields should be assimilated? In other word, once I get the increment fields of zeta from the EnKF, which time step of zeta shoud be added the increments?
Thanks very much.

[wilkin]

The point of having a PERFECT RESTART option in ROMS is to be able to be able to restart a run as if it were never interrupted. To do so it needs all levels that are carried for the 3td-order predictor-corrector time stepping.

Since you propose to change the ocean state with an offline analysis, perfect restart makes no sense.

Just use the output to the history file HISNAME at the end of the run as the prior conditions to your analysis. Create your analysis in a file the exact same format as that, and then start with that as the initial conditions file ININAME.

[xiaozhu557]

Wilkin, thanks for your kindly replying very much.

Since I found the re-calculated results of restarting with the HISFILE or non-perfect restart RSTFILE were different from the original results without restart. For example,
(1)I run the model from Jan 1st, 2000 to Dec. 31, 2000 directly without PERFECT RESTART option at the first time, and archived the HISFILE or RSTFILE on the Feb 1, 2000.
(2)I restart the model with HISFILE or RSTFILE as the initial conditions file ININAME. Then I will get the new results from Feb 1st, 2000 to Dec 31, 2000.

Then I found that both results are different a few days later, more different with more longer of the model time. Especially in Dec 2000, the basic patterns of ocean circulation, such as North Equatorial Current, North Equatorial Counter Current, are different obviously. Even if I have used the same input files, such as surface forcing, boundary conditions. Do you have any suggestions for this?

But the situation will be much better once I use the PERFECT RESTART option. This is the main reason of using PERFECT RESTART option for me.

When I do assimilate, I need to restart the ROMS model after changing the ocean state with assimilation on each time. So I would like to know which time stepping should be changed for the 3rd-order predictor-corrector. I have not digged out the difference method in detail. It is troubled to me.

Thanks for your suggestions. Hope for your replying.

[xiaozhu557]

The point of having a PERFECT RESTART option in ROMS is to be able to be able to restart a run as if it were never interrupted. To do so it needs all levels that are carried for the 3td-order predictor-corrector time stepping.

Since you propose to change the ocean state with an offline analysis, perfect restart makes no sense.

We just want to get the differences between with- and without the assimilation. Or we have no idea to distinguish the roles of between assimilation and numerical calculation of model restart. So I would like to use PERFECT RESTART option.

[lindandan]

The point of having a PERFECT RESTART option in ROMS is to be able to be able to restart a run as if it were never interrupted. To do so it needs all levels that are carried for the 3td-order predictor-corrector time stepping. .

Hello , I have a basic confusion. When I use the 4Dvar to assimilate SLA ,Whether need I add MDT to the SLA? I find the script of Arago don't do that. So I want to ensure what's right when use the roms-4Dar.
Thank you very much.

[jzavala]

Yes, ROMS SSH is the sum of the MDT, the SLA and tides (if you have them defined in ROMS).
Be careful to add the model MDT and not other product such as that provided by AVISO.
The model MDT can be obtained as the mean of a long (say 5 years or more) forward run.

Jvr

[lindandan]

Yes, ROMS SSH is the sum of the MDT, the SLA and tides (if you have them defined in ROMS).
Be careful to add the model MDT and not other product such as that provided by AVISO.
The model MDT can be obtained as the mean of a long (say 5 years or more) forward run.

Jvr

got it, thanks very much!