3.6Computational Timestep
The selection of the timestep is critically important for the success of a model. The run time is directly proportional to the number of timesteps required to calculate model behaviour for the required time period, while the computations may become unstable and meaningless if the timestep is greater than a limiting value. This is known as the Courant stability criterion.
3.6.12D Domains
For the 2D scheme, the Courant Number generally needs to be less than 10 and is typically around 5 for most real-world applications (Syme 1991). The computation timestep in the .tcf file (see Timestep) should be set in accordance with this criterion as given in the equation below.
2 D Square Grid (1)
As a rule, the timestep is typically half the cell size. For steep models with high Froude numbers and supercritical flow, smaller timesteps may be required. It is strongly advised to not simply reduce the timestep if the model is unstable, but rather to establish why it is unstable and, in most instances, adjust the model topography, initial conditions or boundary conditions to remove the instability.
If the model is operating at high Courant numbers (>10), sensitivity testing with smaller timesteps to demonstrate no measurable change in results should be carried out.
The occurrence of high mass errors is also an indicator of using too high a timestep (see Section 6.3.3).
Dostları ilə paylaş: |
