open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi,
I have a project where we try to simulate wind setup in a bay during storm events. However, we have a problem with that the water surface fluctuate/oscillate in the inner part of the bay (and this is not seen in real water level observations in the bay).

Does anyone have any ideas how to resolve this?

I'm attaching our setup as well as figures showing the bay and the how surface level varies.

Thank you for your help!

Hi

2 ideas you could explore:
- 3D effects in reality which connot be represented with a depth average model...
- Effect of wind outside the bay which is probably enclosed in the boundary condition for the water level but not take in account for the velocity part? Then the wind you add on the model generate an internal resonance?

Hope this could give you some tips

Regards

Hello,

Could you check if these oscillations are a seiche phenomenon ? That is if the period is about the time for a long wave with celerity sqrt(gh) to cross the domain twice ? In this case it would be however natural, or triggered by your boundary conditions at the open boundary, or even your initial conditions. Otherwise in 2D computation with wind we generally find the correct free surface slope, but the currents are wrong because we do not get the upwelling effect which is only 3D.

With best regards,

Jean-Michel Hervouet

We have tried 3D - same problem
Without wind - same problem
Changed turbulence model - same problem
Changed time step - same problem
Much larger domain - same problem
Smooth bottom at the boundary - same problem
The period is too short to be seiche and we also have measurements from inside the bay that does not show our oscillations

Our idea is that something in the model that cannot handle rapid variations of the water level at the boundary?

Hello Charlotta,

If I understand well the axis in your pictures, the bay would be about 25 km long, and I see 43 oscillations in 3 days, which is a period of about 6000 s, it cannot have numerical explanations. A long wave traveling through and back in the bay would do 50 km, and to do this in 6000 s we need a celerity of about 8.3, if this celerity is sqrt(gh) we get an average depth of 7 m, which is not too unrealistic, so it could be that you trigger a resonance in your bay with the boundary conditions.

A solution would be to try Thompson's boundary conditions, another would be to try relaxation, it consists in choosing a parameter alfa and giving as boundary condition for the free surface :

prescribed free surface = free surface at the previous time step
+ alfa(data - free surface at the previous time step)

with alfa less than 1 so that there is no sudden jump. This can be done also with Thompson by changing the hardcoded parameter that is commented in subroutine thomps.f (TETA at line 269). Depending on your time step a small parameter is certainly necessary to delay the free surface by a sufficient fraction of 6000 s so as to break the resonance.

Another interesting thing to know is the sampling period of the data, to know if the oscillations could have been filtered (such seiches do exist sometimes, in lake Léman for example). However such a resonance triggered by boundary conditions has already been reported once.

With best regards,

Jean-Michel Hervouet

Thank's for the analysis! I will try the relaxation and I come back with the results when it's ready.

Thank's again,
Charlotta

The good news are that we are cooperating with DHI in this project and they have exactly the same problem with MIKE 3 FM. The bad news are that the relaxation is not very helpful since the time step of the boundary data is only 1 min. I will try with 1 hour averages as well.
It seems that there are some oscillations in the indata and when the amplitude of these increses the amplitude of the water level within the bay also increases (but much more). We are planning to try to do some filtering to decrease the oscillation of certain periods in the indata.

To be contiuned...

/Charlotta

Hi,

Can I ask if there were any solutions to this issue?

We are running a flood inundation model of a storm surge at Immingham Port, UK. Within the first hour of simulation, oscillations start forming in the dock, which is separated from the sea level in the estuary by the dock gates, and therefore doesn't have any external forcing at the start (supposedly!). These oscillations cause water to spill out of the dock and on to land before the storm surge arrives.

Reducing the time interval reduces the amplitude of the oscillations and the amount of premature flooding, but even at 0.25 seconds the oscillations and flooding occur.

Any suggestions beyond those given above would be very gratefully received?

Many thanks,

Robert Mawdsley

Hello,

The best thing was to use the relaxation (and maximize it) that JMH suggested. In the measurements there was a weak oscillation and we found that it could have physical explanation (seiche) but the model amplified it very much and that was the problem. It seems like this bay is quite tricky to simulate.

Hope this will help.

Charlotta

Hi. How shallow is the water off the measuring station? Is there a considerable stretch of shallow water in the direction these long waves approach the station (see the streamlines)? What is the bed composition in the inner bay?

If the bed resistance in the inner bay is considerably higher than the outer part (much shallower for a sufficiently long distance; seagrass, boulders), the increased bed roughness will tend to damp the inertial oscillations. This is why salt marshes are good for coastal protection, in general. This foreshore mechanism could be at play with the early waves that are not so high.

Is this perhaps relevant to your location of interest? If so, you could try and include the relevant zoning of the bed roughness in your model that simulates this damping.