- Checking of the input should always be done at first. It is important that dimensions, model parameters, numerical parameters
and boundary conditions are given in a correct manner. Also, consult Chapter 5, if necessary. In any case,
check the
`PRINT`file. - If all input is correct and the model is supposed to converge then the common measure is always to reduce the time step
(see command
`COMPUTE`). Especially in the case of a transient or spin up a small time step may be necessary. Since this is the easiest way to overcome problems, it is always a good practice to start with this measure. Note that sometimes the maximum Courant number should be decrease as well (see command`TIMEI EXPL [cflhig]`). - Sometimes instabilities may be due to wiggles in the solution. A possible remedy is to choose an upwind discretization or even the first order upwinding. Another improvement might be due to choosing a flux limiter (e.g. minmod).
- Sometimes the solution of the Poisson pressure equation can not be found as the BiCGSTAB solver did not converged.
This may happen, for instance, when a considerable number of layers ( 30) is involved.
In such a case the choice for the ILU preconditioner might be a good alternative (see command
`NONHYDrostatic ... PRECond ILU`).