Sammanfattning
Two different advection schemes, Flux Corrected Transport (FCT) scheme (Zalesak, 1979) and incremental remapping (Dukowicz and Baumgardner, 2000) have been tested in a two-dimensional test program where an initial cylindrical tracer distribution was advected in a circular velocity field. Evaluation of the tests showed that incremental remapping preserved the shape of the tracer distribution better than FCT. However, incremental remapping is computationally efficient first when many tracers are used. In addition, it has been developed for only two-dimensions so far. Consequently, it was decided to implement FCT into the Rossby Centre Ocean model (RCO) (Meier et al., 1999).
Model simulations with RCO for the period November 1902 to December 1998 were done with FCT as advection scheme. This model simulation was compared to a model simulation with modified split-quick (Webb et al., 1997) advection scheme (Meier and Kauker, 2003). Modified split-quick isa third order scheme and is the current advection scheme in RCO.
The advection of salinity and temperature was compared between the two model simulations. FCT was applied without explicit diffusion whereas modified split-quick needs explicit diffusion to eliminate subgrid-scale noise caused by the non-monotonicity of the scheme. Consequently, the simulation utilizing FCT was less diffusive. This is positive because as little implicit diffusion as possible is desirable, especially for long integrations. At sharp gradients modified split-quick may result in under- or overshooting. For example at the outflow of Neva and Kernijoki rivers a Sharp salinity gradient causes salinity to become negative for modified split-quick. FCT was shown to handle sharp gradients very well, no negative values for the salinity occurs with FCT as advection scheme