open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,

A tentative correction is enclosed, a call to plane_bottom added in telemac3d.f. This will be in the next release. Your case now works a few thousands of time steps without any NaN.

On my way I found possible slight improvements :

* remove the keyword MINIMUM VALUE FOR DEPTH (no use here)
* change the accuracy for propagation to 1.E-8 (should have an effect on mass conservation)
* add the keyword:

TREATMENT OF FLUXES AT THE BOUNDARIES : 2 (for your unique boundary)

* check the influence of IMPLICITATION FOR DEPTH (if 1. works, see the difference on mass conservation)

* I suspect that TREATMENT ON TIDAL FLATS FOR TRACERS = 1 spoils the mass conservation (cancelling a tracer on a tidal flat with depth = 0 may however change the mass of tracer if there are nearby points with water, just by linear interpolation).

* lastly your line T3DBDF = (/7252:7326:1/) and the following are considered a syntax error by the Nag compiler, so it is probably not Fortran 90.

* in CONDIM use MESH%KNOLG under condition of NCSIZE.GT.1 (and if not just do N=J) so that the program works also in scalar mode.

That's all, have a good week-end,

Jean-Michel Hervouet

Thanks for your help! I'll apply these improvements in my model.

Have a good weekend!
Qilong