In addition, MIKE 21 FM uses an adaptive time step, which will therefore never be strictly the same as when using a constant time step in the MIKE 21 SG version.Īnother not so obvious difference between the two models is the way the bathymetry is handled, and this is probably where the most obvious reason lies for simulation differences between the two types of models. Thereby, we will, in general, be able to achieve a rather significant speed improvement when comparing to simulations with the rectangular MIKE 21 SG model. However, the enlarged simulation time for FM with a reduced time-step is typically handled through running the FM simulations using more advanced parallelization techniques (MPI) or with GPUs (a big step forward compared to running only on CPUs). Consequently, we will typically experience that it is necessary to reduce the time-step in FM when converting from one model type to another. The semi-implicit solution method in MIKE 21 SG allows larger Courant numbers (1-1.5 or sometimes even larger) and still offers a stable solution. The primary reason for this is again related to the numerical methods, where MIKE 21 SG uses a semi-implicit solution method and MIKE 21 FM uses an explicit numerical solution approach, where the latter has a much more strict stability criterion as CFL-numbers (Froude-number) must be less than 1 and preferably less than 0.8 or 0.9 to keep the solution stable. One of the first issues typically encountered is the requirement of using a different time step in MIKE 21 FM in order to keep the model stable as compared to simulations set up in MIKE 21 SG. Different numerical solutions can basically not be compared and it is therefore actually quite expected that the results will differ between the two types of models – even though the model schematisation is replicated such that the rectangular grid in the MIKE 21 SG model and the mesh in MIKE21 FM are identical. MIKE 21 SG is a grid-based Finite Difference method whereas the FM-model uses a cell-centred Finite Volume method.

Why is it so? Which version is more correct? Answer:įirst of all, the main thing to put in focus here is that the two simulations will use two different numerical solution methods when comparing the rectangular single (or ‘classic’) grid (SG) to the Flexible Mesh (FM) version of MIKE 21. I applied similar parameters, however the results differ. MIKE FLOOD: I run MIKE FLOOD successively with MIKE 21 rectangular grid and flexible mesh versions. Please do not hesitate to e-mail us if you have questions and/or answers yourself:

Mike 21 flood