open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi,
I have developed a Telemac model to investigate into the potential impact of a proposed project may have on the river hydraulic and neighbouring water lots.

I conducted the the unsteady state model simulation with present day bed geometry and with future bed geometry. The river is 25 km long. I compared the results by plotting the difference between velocity fields. The resulting changes near the project site makes sense. However, I am noticing small changes (< 0.1 m/s) throughout the model. In addition, there are noticeable changes (> 0.3 m/s) in multiple areas several kilometers away from the project site. I have hard time to believe these large differences are real.

I came across the paper by Nheill (First improvements toward a reproducible Telemac-2D, 2016). Is issue with reproducibility the reason why I am noticing these differences?

Is your mesh the same between 2 geometry ?

Yes. I used the same mesh. I just updated the bed elevation in localized area to reflex project induced change.

Hi,


You use the unsteady flow, are you sure compare the same time? (I know obvious question, but I have made that mistake).

Is the solution you get stable in time? Maybe there are oscilations (between each timestep) that don't show when looking at larger time intervals (output file).

Hope that helps,
Patrick

Thanks Patrick.

yes, I am comparing the same time. The 'differences' are smaller when the flows were lower and increased as the flows increased. Both models were simulated with same parameters.

I've tried lower timestep and other parameters but still notice similar issue. hence I am wondering if this is simply due to the issue associated with how the codes were designed in handling multi-cores. And if so, wonder if it's something that the project team may be looking into addressing in the near future.

Thanks.

We often do this kind of comparative simulation and we don't have this kind of issue.
Rules are same mesh (same node coordinnates) and fixed time step.

Could you add more details and illustrate with the difference map if possible ?

I am simulating an estuary system. As such, I have time-varying river inflow on the upstream end and time-varying tides (TPXO) on the open ocean boundaries.

I am running the model on CLOUD computing with 28 cores. I've tried both GCC and Intel compilers but still having the same problem.

I will do some testing with simple cases (steady state), and maybe a single core to further investigate the problem.

Okay. I may have posted the question on the wrong form.

From my test steady state runs, the 2D simulation results for the comparison runs look fine. i.e., no project effect in the far-field.

I did the simulation in Telemac3D in both sigma layer and in z-layer. Z-layer model resulted in several unrealistic project effect regions. Sigma layer model had fewer unrealistic project effect regions than z-layer.