runtime when simulating trace

[smultsch]

Hello,

I am using a really basic setup of CMF to simulate stream flow in a small number of river segments. The simulation of solely water fluxes is quit fast regarding performance. But when adding a tracer the simulation time increased by a factor of >200. I cannot figure out the reason for the large difference in performance when using CMF with and without a tracer (keeping in ind that the tracer flux is only conservative transport).

Would be great discuss this!

Cheers,
Sebastian

[philippkraft]

A first observation: The largest numerical error is mostly at the solute storage of the 2nd reach in the first hour of the modelling. The performance problem is not related to the number of RHS evaluations

[philippkraft]

The problem seems to lie in the absolute error tolerance (to be used when a value approaches 0) - a18ef52 helps a bit, expecially with greater relative tolerances.

[smultsch]

Ok, but I don't no what to do next. Should I replace the files and conduct the test again?

[philippkraft]

No - the problem is still unsolved.