open TELEMAC-MASCARET The mathematically superior suite of solvers

Dear Jacek & greetings to all,

I'd like to add a couple of point to this interesting discussion. Sometimes, when we have to deal problems presenting with low water depth and morphodynamics, we play a bit around with the slope effect. From a mathematical point of view, the addition of this effect is equivalent to add some diffusion to the solution, with a physical meaning behind.

As pointed Riadh, some refinements of the mesh could be necessary in the problematic zones, especially where the presence of strong bathymetry gradients can affect the validity of the hypothesis behind the shallow water models.

I'm 100% in agreement with you guys that some revision of the algorithms must be done to solve this kind of issues. We will see internally how we can attack this problem in an efficient way.

Best wishes,

Pablo

PD: Jacek, since I'm working with Telemac, a lot of feedback, solutions and clever ideas were discussed and proposed in the framework of the Consortium meeting. It's one of the few places (the other is the TUC) where smart people with the same interest can gather and discuss all together with the same topic in mind .

Hello,

The user has already been advised before by Sisyphe developers to increase slope effect and to manipulate thresholds controlling effects while drying/wetting, but it seems in vain or ineffective, but of course, errors are human...

Yes, the model must be refined and stabilised first, and only if we still have troubles we would turn back to philosophying again. It is hard to see things clearly in a model of such low resolution and so many aspects involved. And I know exactly this kind of physics should be computed in 3D, but due to internal requirements we cannot.

(But in the time between, please analyse that positive depths, and anyway: Please give us back the old FEM algorithm again!)

It is good to hear that there are people on the consortium meetings still willing to discuss technical/professional matters. In protocols I was allowed to see there's only management, personnel, money, overall work plans...

Best regards,
jaj

Dear all,

this post is intended only to keep you informed.

In the time between our suspicions concerning the Sisyphe's "positive depths" algorithm based on the model result (posts above) and a rough analyse of the code with help of available literature have been confirmed by the developer:

[...] There is nothing in the process that says that there should not be fluxes towards a dry point because when we solve the Exner equation, the water depth is just not looked at, we just do the advection of the sediment height. [...].

We have here exactly the unforeseen case: an island growing up from a sand bank in the middle of the steadily flowing river. A serious trouble, because taking some exceptions into the Exner equation -- treated as a distributive advection scheme -- does not have to be trivial (as in any advection scheme):

[...] The simple idea that would consist in cancelling the fluxes to dry points is thus not so simple to implement. This would be a new positive depths algorithm with an extra step [...]

So what remains is to wish the developer a lot of fun by making Sisyphe's most important algorithm fit and robust enough for real world settings in practical projects on river engineering...

Best regards, with fingers crossed,
jaj

Dear all,

As an occasional Sisyphe user, I have been reading this thread with interest. I also observed a similar phenomenon ("growing islands on tidal flats") in some river models, both with T2D-Sisyphe *and* DHI's MIKE21-MT (2D mud transport module).

Would be interesting to know if this limitation is something classic when dealing with the Exner equation (experience from other codes?)...

Best regards
PL

Dear PL,

sorry, my experience with morphodynamics is very limited and I am only a kind of users' advocate -- so I can provide only an overall philosophy: One distinguishes between directly coupled models in which one solves the hydrodynamic and morphodynamic equations simultaneously in just one equation system and the indirect coupling with separate solutions.

There are examples of the direct approach which is especially attractive in the case when the resulting equation system is hyperbolic with presently quite a lot of publications, dealing with more intensive bottom changes, like sand dam breaking, intensive scouring, moving dunes, etc. Note: It is more difficult in the case of general flow situations, which do not have to be hyperbolic at all...

By indirect coupling one solves hydrodynamics and morphodynamics completely separately, and then exchanges the necessary information between the different modules -- velocity (shear stresses) in one direction and bottom changes (including roughness at best) back. This approach is justified when the time scale of the morphodynamic changes is much larger than one the flow, so just like the slow bed load transport in lowland rivers, for example. This is the case of Sisyphe and Telemac and the situation for which the original Exner equation is formulated, actually!

The trouble of the indirect coupling is the treatment of the non-erodible bottom and drying-out areas (and not so severe, flooding of some previously dry land). This information must be exchanged between the modules in some proper way.

The case with the present "positive depths" algorithm in Sisyphe seems to be that while its introduction a few years ago resolved the previously notorious troubles with non-erodible bottom areas, it seems that the drying-out due to the bed load arriving at the expanding dry shore has not been taken into account. It is strange to me that it has not been reported before; maybe the case with a sand bank growing up to become an island in the middle of the river is somehow exceptional. Should it not appear everywhere along the shorelines? Maybe some other tricks in the code covered the ever-present fault up, like this diffusive "slope effect" treatment.

Anyway, let's hope that the developer will find out a solution. Anyway, we have a severe validation case ready to be deployed!

Best regards,
jaj

Hello,

I just include a short information for Sisyphe users that a prototype of the Exner equation "positive depths" solver for Sisyphe, which takes into account drying and wetting (yes! it has not foreseen drying and wetting!) based on manipulating fluxes to and from drying/wetting nodes (I presume) has been written and is going to be tested "soon". Thank you, Jean-Michel.

In the consequence previous Sisyphe results concerning drying and wetting processes due to bed load transport should be assumed to be falsified. Sorry.

Best regards,
Jacek

Hello all,

I have also been plagued with the same issue, this time in coastal morphodynamics. I can get shore/boundary nodes evolving up to a height o 2.5m until an error in fractions eventually steps in to halt the computation. So I have been monitoring this thread with great interest.

Following Jacek's latest post, I have noticed that JMH's fix has been added to the trunk -revision 7713- so everyone can test. I have already started to do so.

Best Regards,
Costas