Why I got such a different tide solution using TPXO data

[stone]

Hi,all
I can't understand this weird result I got. I use the TPXO7.1 tide harmonics as the boundary condition(using the method as OSU Tidal Prediction Software Example from https://www.myroms.org/wiki/index.php/Tidal_Forcing),and I extract the bathymetry from the TPXO grid_tpxo7.1.nc(a little smoothing was done by smooth_bath.m in ROMS matlab dir).The difference after-before smoothing.

Although I know TPXO tide data has assimilated some obs),I would expect my result should agree with TPXO,since I use its OBC and Bathymethy.
However,they are very different as follows.
beyond the color range is set to Nan;I choose M2 S2 O1 K1 as tide forcing.


The cpp is as follows (which is 2D case)

here is the parameters:
5.0000E+02 visc2 Horizontal, harmonic mixing coefficient (m2/s)
3.0000E-03 rdrg2 Quadratic bottom drag coefficient.

and I also did it in 3D, using a similar bathymetry(not the above).The result is just slightly better,but far from good.

Anyone can tell me why this happens? Thanks

[stone]

Hi,
I learn from viewtopic.php?f=2&t=61&hilit=tangential that, maybe I need to add tide body forcing namely Potential Tide since my domain is large and deep.
Can any expert using Tide Potential assure me this is necessary for my application.

Thanks.

[kate]

The tidal potential term is likely to be a small correction. If you have something strange going on without it, the tidal potential will not fix it.

[fragoso]

Hi stone,

I see you´re using ANA_FSOBC and Chapman boundary conditions. Maybe there´s a conflict there. Have you checked this?

Regards,

Mauricio.
Prooceano - Brasil.

[stone]

Hi,Mauricio
ANA_FSOBC and Chapman boundary conditions do not conflict with each other,
The first is for subtidal, and I use add_fsobc to add tide in,then I use chapman condition.

[gouillon]

Hi Stone,

Although I do believe, like Kate, that in your case combining the remote tide coming from the open boundaries and the body forcing (local tidal potential representing the astronomical forcing in your domain) will not necessarily correct for all your differences (because of the low latitude), you might want to try to use both forcing and post the results here. I'll be interested to see the impact of the local tidal potential on your solution.

Indeed, I recently showed for the Gulf of Mexico (where tides are not that important), that there exists some non-linear interactions between the 2 forcing that might lead locally to large amplification (or de-amplification) of your tidal signal. See Gouillon et al, 2010, JGR-Oceans, doi:10.1029/2010JC006122.

Also, your figures show that at the North and East boundaries, the M2 forcing seems to be quite close to TPXO (difference close to zero for both tidal phases and amplitudes) but at the South you have quite some difference. Are you sure there is not a problem how you prescribe the tide at the South boundary (angle with the grid, etc.)? How does changing the bottom friction magnitude impact the solution?
How long have you run forward the model (are you sure the energy in the domain has reached equilibrium)?
Anyway, my 2 cents...

Also, How the other tidal constituents look like? Say O1 that is a less sensitive tidal constituent to model parameters...?

Flav

======================
Flavien Gouillon
OMP/SHOM/DO/HOM/REC
Physical Oceanography
www.coaps.fsu.edu/~gouillon

14 av Edouard Belin
31401 Toulouse France
Phone: 05 61 33 29 11
Fax : 05 61 33 29 16

[stone]

Hi,Flav
Thanks for your kind reply.

Are you sure there is not a problem how you prescribe the tide at the South boundary (angle with the grid, etc.)?

I think Roms internal code will do the vector rotation ( see set_tides.F).

How does changing the bottom friction magnitude impact the solution

I didn't change bottom friction in 2D case,since it is difficult to determine where it should be large(small) and how large(small).And I don't thick tuning a constant bottom friction coefficient will do much improvement in my case.I used a UV_logdrag in a 3D case(zob=1.5e-3),there is not much change.

How long have you run forward the model (are you sure the energy in the domain has reached equilibrium)?

I run the tide model for 50 days, and take the last 31days to analysis.

Say O1 that is a less sensitive tidal constituent to model parameters

O1 is like this



I need to add tidal body force to check.

Thanks,guys

[feroda]

What's your resolution of you model? Please double check the viscosity coefficient(500, in you case) is appropriate for you.

[stone]

What's your resolution of you model? Please double check the (500, in you case) is appropriate for you.

My horizontal resolution is about 9km, will viscosity coefficient(500) affect the tide simulation here? The tide wave length is large comparing with grid resolution, and a smaller vicosity should be proper. I used 50 for viscosity coefficient in a 3D case which contains basic current(like Kuroshio) and Tide ,the eddy energy seems to imply that 50 is good,while Tide is just like the above.

Thanks for all the suggestions.

[stone]

Hi,
Here,I just report the tide simulation with Potential tide added in.
It doen't appears much approvment as Kate said.

I am totally lost now.
Thanks to Kate and gouillon for help and suggestions.

[gouillon]

Hi Stone,

OK so as expected the tidal potential did not do much.

Clearly your solution for semidiurnal and diurnal tidal phases and tidal amplitudes at the boundaries do not match TPX (especially at the Southern boundary) and it should. It is puzzling to me.
So 2 things:
(1) you're not prescribing correctly the tidal forcing at the open boundaries and/or,
(2) you have energy accumulation in your domain and resonance phenomena (higher modes generated) that contaminate your signal at OBs.

You need to isolate which one is your problem.

One thing could be to look at you tidal amplitudes at the 2nd day or a bit later but earlier in your run (i.e., before there is too much energy piling up in the basin if it is the case). Do the first M2 oscillation amplitudes (after the initial adjustment which happens fast) correspond better to TPX? If yes then maybe the waves are not radiated away properly and thus, after some time, the tidal signal at the boundary is noisy due to these resonance phenomena...

Have you tried also to prescribe the tides with another product if possible (e.g., ADCIRC?)? Or even prescribe the tides analytically (a M2 tide only for a given amplitude) at the open boundaries (a good test to check that the Flather conditions are actually working well because you know exactly what you should have...)?

Good luck!

Flav

======================
Flavien Gouillon
OMP/SHOM/DO/HOM/REC
Physical Oceanography
http://www.coaps.fsu.edu/~gouillon

14 av Edouard Belin
31401 Toulouse France
Phone: 05 61 33 29 11
Fax : 05 61 33 29 16

[johnluick]

I think Gouillon's idea to run for M2 only is good, but perhaps I'd make a slight change (or do both). Rather than adding M2 "analytically", consider using a netcdf file. (Easy to create because you just modify the script you are currently using to create TPXO tides input.) If you do it analytically, the phase change along the open boundary may not make any sense physically which could lead your sea level to simply switch to a different style of exotic dancing. Also, this approach avoids the need to change any of your ROMS input files (.h and .in). Secondly, I would use S2 instead of M2 because then you don't have to worry about the nodal effect on phase and amp. Good luck.

[stone]

Hi,
I compared the open boundary ssh between modeled and prescibed,
looks like there is something wrong from the early beginning of the simulation



what I show is the 4 tide componients case , I have not done the M2(S2)tide only for our computer workstation came accross some problem.

[ezaron]

Hi Stone,

In addition to the other suggestions, above, you should know that the OTIS bathymetry is unlikely to yield a very good result here. There is a paper by Jan Sen that discusses the shelf resonance between China and Taiwan, and the importance of good bathymetry for modeling M2 correctly. Also, the energetics of the diurnal tide are largely influenced by the barotropic to baroclinic conversion in Luzon Strait. I think modeling this conversion (which requires good and steep bathymetry and reasonable stratification) is necessary to model K1 and O1 correctly. The TPXO solution is strongly constrained by altimeter data in this region. Free-running the OTIS model here would yield a very different solution from the TPXO assimilations.

You might consider running a single constituent at a time to dis-entangle some of the issues.

Regards,
-Ed