Using CFL time adaptation method

kevinl

Hi,

I am trying to use the CFL time adaptation method.

I am setting a transient run of 50 years with a 1 year time step.

If I don't use the CFL condition (md.timestepping.time_adapt=0) I do have 50 time steps of a year. However, if I set time_adapt=1 (keeping the default value for cfl_coefficient=0.5) I get only one time step as if dt was changed to 50 yrs.

How can I properly use the CFL time adaptation method?

Thanks

basiledefleurian

Hi,

Most probably, the issue is that your simulation have a coarse enough mesh and small velocities which leads to a computed time step larger than 50 years.

To check if that is the case just run a stress balance run and check the value of the time step:

dt=cfl_step(md,md.results.StressbalanceSolution.Vx,md.results.StressbalanceSolution.Vy)

Basile

kevinl

Following on the same subject, when I use md.timestepping.time_adapt=1, my time step goes from 1 year to 0.1 and goes as low as 0.0001.

However, cfl_step(md,md.results.TransientSolution(i).Vx,md.results.TransientSolution(i).Vy) give me values higher than 70 over the full run.

Any ida of why my time steps get so small?

mathieumorlighem

It is probably due to vz, the vertical velocity, that is not taken into account in cfl_step. Your vertical velocities must be a bit high