Skip to content

Common questions

How to select a good time step (dt)?

Choosing a time step is a good example of the simulation quality vs. performance bias. Smaller time step might better reproduce some physical phenomena, however, it will also require more computations. On the other hand, bigger time step might hinder the convergence of the PDE solver (your fluid flow dynamics) and produce unstable simulation. For this reason, we are often interested in a maximal acceptable time step. Luckily, this number can be infered based on the convergence condition by Courant–Friedrichs–Lewy (CFL) and it is exactly what DYNAMICO is doing.

In the out_execution log, after your first simulation, look for this line:

$ grep "Max. time step assuming" out_execution -A 1
Max. time step assuming c=340 m/s and Courant number=3 (ARK2.3) :
  418.024494240445 #(1)!
  1. This maximal time step is in seconds!

In this particular case, we should go with a time step of at most 418 seconds!

How to calculate LAM resolution?

Take LAM radius and divide it by the nbp icosahedral division factor. For example, this Antarctica configuation:

radius = 3360 #(1)!
nbp = 32
  1. Units of the radius are kilometers.

corresponds to a 105 km grid size (radius [km] / nbp = 3360 [km] / 32 = 105 [km])

How to select dissipation parameters?

These three parameters should be adjusted in accordance to the LAM resolution. Selecting them is a matter of trail and error, so that's why you can use proposed values:

resolution [km] tau_graddiv tau_gradrot tau_divgrad