open TELEMAC-MASCARET The mathematically superior suite of solvers

We are currently in the process of understanting the potential of applying TELEMAC-2D for large areas with complex inland river-floodplain interaction.

As such, we read that TELEMAC has been applied by the Canadian Hydraulic Center on a complex 2000km² area here and some reports mention the application of up to 200Million element-meshes and give information about their pre-processing time here.

We are now firming up the idea to apply TELEMAC-2D for a multi channel river systems with extensive floodplain areas like they may be found in e.g. the Pantanal or the Okavango Delta or other wetlands of similar type. We estimate about 75.000km² domain size, and expect maybe around 10 Mill. elements.

The domain is subject to complex hydraulic and hydrologic processes:
- interaction between channel and floodplain flows in densely vegetated fields
- channels are located on an elevated inland delta and are spread longitudinally and laterally on vast floodplains
- high evaporation (and to some extent infiltration) is taking the water off the domain in the dry season
- rainfall causing widespread flooding across the whole domain in the wet season
- time period to be simulated: about 2-5 years

To substantiate our approach, I am looking for further experiences that others in this forum here may have in modelling similar conditions. Especially I am interested in:

- any studies or reports that can be shared where similar processes were depicted
- any rough idea of the computational requirements to enable run times of maybe 1-2 weeks for a simulation period of 2 years (considering the outlined domain size/elements and extensive wetting-drying processes and that we need multiple runs for calibration and scenarios)
- if no example regarding these points is available, we would be extremely grateful for any personal opinion on that matter

Thank you very much in advance!

Hello,

I do not know what study Jean-Michel Hervouet thought, but by googling TELEMAC and CHC NRC, I found a few reports which may give you some ideas:

www.yumpu.com/en/document/read/22583620/...-model-of-the-lower-

ijc.org/sites/default/files/report_Pembina_CHC_June_2010.pdf

ijc.org/sites/default/files/IJC_Rainy-Ri...-II-Report_final.pdf

Moreover, CPU times required to run a 2 year period of time depends on the number of elements of course, but also the time step, the size of the triangular elements, the velocity in the area... But the use of cluster should be recommended.

Hope this helps,

Chi-Tuan

Thank you very much for your reply and the links, Chi-Tuan!

The case studies are interesting, but we expect at least about 10-times the complexity of the described hydraulic problems (both in terms of number of channels and domain size).
We assume a cluster is needed here, yes, but nevertheless, it would be great if someone could share opinions on more specific requirements given we'd aim for 1-2 weeks of run time for a 2-year scenario.

To give some more details for being able to answer that question:
- we expect around 10 Mill cells
- the size of the elements should be in the range of 20m² to about 15000m²
- velocity should range from 0 (stagnant water in the largest cells) to maybe around 2m/s max in the smallest cells in the channels)
- the time step...basically we'd aim for the highest the model can handle under the described conditions...if we use the Courant criterion, we probably arrive somewhere below 10s?

Thanks for sharing your valuable experience!