Steric signals and along-track altimetry

[stef]

Hi,

Is there a common practice of how to remove seasonal thermo/halosteric signals from along-track altimetry in data assimilation?

Do these have to be removed from gridded altimetry?

Do I assume correctly that for gridded altimetry, the domain-averaged steric signals should have no effect on the tracer increments (even if not removed), because the model is Boussinesq?

However, if that's true for gridded, then along-track altimetry would still be biased, in the senses that steric signals would be confused with local baroclinic dynamics. Is this true?

I'm asking because I think in the BoB,the steric seasonal mean sea-level fluctuation can be like 5 cm amplitude.

Thanks for your help!

[stef]

I'm asking because I think in the BoB,the steric seasonal mean sea-level fluctuation can be like 5 cm amplitude.

In [1] it seems to be much less, their Fig. 3 suggests roughly 4 cm range (2 cm amplitude) of the seasonal steric component. That's domain-averaged over the northern Bay of Bengal, computed from EN4 data between 1993-2018 in the layer upwards of 900m depth.

And the domain-averaged *total* sea level range about 10cm. I must have confused this in the earlier post.

[1] Shaila Akhter, Fangli Qiao, Kejian Wu, Xunqiang Yin, K M Azam Chowdhury, Nur Uddin Md Khaled Chowdhury,
Seasonal and long-term sea-level variations and their forcing factors in the northern Bay of Bengal: A statistical analysis of temperature, salinity, wind stress curl, and regional climate index data,
Dynamics of Atmospheres and Oceans, Volume 95, 2021

[stef]

Do I assume correctly that for gridded altimetry, the domain-averaged steric signals should have no effect on the tracer increments (even if not removed), because the model is Boussinesq?

After reading the manuals I guess this is incorrect, with a balance operator one would see both and adjustment of the surface, and an adjustment of dynamic/steric height via tracers. I couldn't test it so far using a balance operator (see below), but looking at the definition of the matrix it seems pretty clear.

I made an idealized test with RBL4DVAR without balance operator in a domain with

* uniform temperature/salinity
* no motion
* no forcing
* closed boundaries
* spatially uniform altimeter forcing, 1cm innovation
* allowing initial conditions adjustments only

Running RBL4DVAR simply increments the surface to 1cm above the prior, without significant changes in dynamic heigth, i.e. the tracers don't get adjusted, probably because the balance operator is not activated. I wanted to make an experiment with balance operator, to test if I get adjustemnt of dynamic/steric height, but it blows up and then I saw:

Hernan wrote in [1]

We don't activate that option in the tutorials. The balance operator is an advanced feature for experienced users that know their application well because the bathymetry affects the level-of-no-motion that is needed in the inverse problem.

I am using a realistic topography, I guess it's easier with a flat bottom.

[1] viewtopic.php?p=22595#p22595

[stef]

After reading the manuals I guess this is incorrect, with a balance operator one would see both and adjustment of the surface, and an adjustment of dynamic/steric height via tracers. I couldn't test it so far using a balance operator (see below), but looking at the definition of the matrix it seems pretty clear.

Okay but these dynamic/steric height changes only occur when using the local vertical integral, right? Using the full elliptic equation shouldn't yield any buoyancy changes. And only the full elliptic equation is fully consistent with the Boussinesq model, right?