Baroclinic vs. barotropic mode
Some description of the Baroclinic and Barotropic equations is given in the SELFE paper. A little elaboration will convey how these are implemented and used in practice.
The baroclinic and barotropic modes are toggled in the model by means of the parameter ibcc in param.in. They differ in that in Baroclinic mode, the contribution of temperature- and salinity-induced density differences is included in the momentum equations. In contrast, in Barotropic mode salt and temperature are (at most) considered passive tracers.
In barotropic mode:
- You can choose whether to calculate salt and temperature transport by toggling the itransport parameter in the param.in file.
- This choice affects performance (see below) and also data requirements for salt and temperature. The model (****is/isn't smart) about ...
- The contribution of density to horizontal pressure gradients is dropped.
- The turbulence closure is still calculated in 3D, but the suppression of turbulence due to stratification is not included.
Barotropic mode calculation can be a lot faster than baroclinic. A key speedup is achieved by setting itransport=0 and neglecting salt and temperature entirely, which is particularly common in the case that subcycling of transport is a performance limiter.
Boundary and well-posedness issues
Sequential barotropic-baroclinic analysis
One common usage pattern is to use a sequence of barotropic-baroclinic model runs. The barotropic simulation is used in a preliminary sweep in order to calculate 3D boundary conditions for the subsequent baroclinic analysis. Typically this is done with transport off, maximizing speed for the barotropic step.