time-step size

General scientific issues regarding ROMS

Moderators: arango, robertson

Post Reply

time-step size

#1 Post by redhat007 » Tue Oct 14, 2014 10:33 am

In refinement grids, is it possible that the donor and receiver grids be not an integer factor (RefineScale) of the time-step size( while are in grid-size) ? For example is it possible that for RefineScale=5, time steps be 50 and 40 sec in coarse and fine grids, respectively, not 50 and 10 sec?

Posts: 157
Joined: Mon Apr 28, 2003 5:12 pm
Location: NOAA

Re: time-step size

#2 Post by lanerolle » Wed Oct 15, 2014 2:09 pm

Even if this was possible, using a different time step (other than once scaled by the ratio of the parent:nested grid size) would cause the nested grids to have different CFL numbers, diffusion numbers (K*dt/dx^2, K*dt/dy^2) than the parent grid and hence alter the numerical properties of the solutions in the nested grids because the truncation errors of the numerical schemes employed to solve the equations are dependent on these dimensionless numbers. So the error properties of the parent and nested grids would be less consistent (with each other) and this may lead to stability and/or accuracy issues. Just my thoughts.

Post Reply