open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,
I'm doing some tests with T2D + Gaia, increasing progressively my model's complexity.
I'm now testing a 2 layers model in a channel with a hard narrowing in the middle (see resserrement.png attached).
I declared the layers in the GAIA steering file with the keywords NUMBER OF LAYERS FOR INITIAL STRATIFICATION =2 and LAYERS INITIAL THICKNESS=10,10
Then I defined their thickness using the Fortran pattern file user_bed_init.f.

In the GAIA steering file, I claimed for the evolution of the layers thickness with the keyword VARIABLES FOR GRAPHIC PRINTOUTS = '*ES'.
In the result file I get 3 LAYER THICKNESS variables, where:
LAYER THICKNESS 1 corresponds to my total thickness
LAYER THICKNESS 2 = 0 everywhere
LAYER THICKNESS 3 = 0 everywhere
The other results seem consistent.
Everything happens as if the computation was done with 2 layers, and at the end those layers are melted in just one.
Thanks in advance for any help (complete files in mesh4.zip attached)

Hi Benoît,

the initial stratification describes the measured / assumed stratification and is not the same than the initial vertical discretisation. From the initial stratification the vertical discretisation (initial number of layers) is derived by defining an active layer. Usually the number of initial layers is number of initial stratification + 1. (I am not sure if it is in all cases).

As you have 2 layers with different stratification the NUMBER OF LAYERS FOR INITIAL STRATIFICATION =2 and the LAYERS INITIAL THICKNESS=10;10 (semicolon separated). This gives an initial total thickness of 20m which is changed in your user_bed_init.f for the middle part of the model.
The active layer thickness is by default constant and will be extracted from the first (or more) layers you have described in your initial stratification. If you define it to 5 m the first initial stratification will cut in 2 layers each 5m. If your initial layer 1 is smaller than 5m the layer thickness 2 in the result file will be zero and also the 3rd layer will be changed.

Be careful because in the middle of your model the active layer is still constant (e.g. 5m) but the sediments were mixed from the initial stratification of both layers. So this is probably not what you want?!?

It is a little complicated but I hope I could describe it understandable.
Best regards,
Rebekka

Thank you Rebekka for your answer.
Mixing the layers is absolutely not what I wanted to do. My purpose was to see if T2D/Gaia could take into account a locally non erodable bottom by this mean.
I hadn't understood how the bedload was computed with a stratified bed.
If I understand, the active layer thickness is constant in time and space ?
So which criteria do you think I should take to choose it ?

Hi Benoît,

if your second layer presents in fact the rigid bed, you can do this by limiting the total height to it. The active layer will then become smaller than the constant value:

NUMBER OF LAYERS FOR INITIAL STRATIFICATION =1
LAYERS INITIAL THICKNESS=10
ACTIVE LAYER THICKNESS = 1 (we usually choose between 0.1 to 1 m for the active layer. Otherwise the mixing is very strong and you will hardly see any vertical distribution effects)

and in user_bed_init.f there must not be ESTRATUM(2,IPOIN) or RATIO_INIT(ICLA,2,IPOIN) because there is at this part in the code only 1 layer.
The special declaration for the last sediment class is not needed...

With this you will have an initial uniform distribution but a limited thickness at the middle of your flume and 2 layers.

We did some tests with a coarse layer instead of a rigid bed and found that it is better to use a rigid bed because of mixing processes with the active layer which always produce some trouble in case of sedimentation over the coarse layer.

Best regards,
Rebekka

Hi Rebekka,

I leave this notes for me, you if interested, and anyone who might benefit them... who knows ?

My first intent was a 1-layer model with a rigid bed bellow.
Then 1 tried the same but with a 2-layers model with a very coarse (750 mm) sediment bellow.

Here are the results after 50000s of 20 m3/s flow :
Mesh3a : 1-layer with rigid bed
Mesh4a : 2-layers with very coarse sediment bellow
I checked the sensibility to :
- Number of layers (1 or 2, given the same active layer thickness of 1 m)
- Active layer thickness in the 1-layers scheme (no stratification defined or 1 m active layer thickness)
- Active layer thickness in the 2-layers scheme (0.2, 0.5 or 1 m)
- Last but not least, Timestep. My initial timestep was 50 ms. I made a try with 20 ms in the 2-layers scheme with a 0.2 m thick active layer


I made the same observations as you : the results are not the same because of the mixing processes with the active layer. Only the rigid bed (1-layer sheme) allows to conserve the original "bump" (1st file attached).
The active layer thickness doesn't matter if between 0.2 and 1 in the 2-layers scheme (3rd file attached), but is important in the 1-layer scheme (2nd file attached) : I think it is because of the cobbing effect.
Finally (4th file attached), I wonder I should lower the time step to get stability in the bottom...