I got my hands on ROMS in the beginning of this year and I try to set up a simplified scenario of an ice shelf close to a continental slope. The objective is to study the (eddy-based) overturning across a pronounced front, which is setup by downwelling associated to along slope winds.
We apply periodic boundary conditions along the slope (EW) whereas SST and SSS as well as the northern boundary will be relaxed to climatologies.
I hope, the model will reach some kind of steady state, which is reproducing an annual cycle of the overturning across the front. However, all the nudging plus surface fluxes from ice shelf melting will change the tracer budget and I fear the model might drift away before I reach equillibrium.
Is there any possibility in ROMS to conserve the tracer budget (e.q. by subtracting the induced changes at the boundaries evenly distributed in each grid cell of the entire model domain)? Is there a smarter solution to my problem? Is the situation hopeless?
Basically the same question about momentum: I try to resolve eddies (horz. res. 1.5km) and hence want to have none or very low diffusion/ viscosity. How to find a viscosity that does prevent the model to drown in numerical noise but does not kill the eddies? Which advection scheme suits best? Can you suggest some literature about this kind of approaches? Of course I will get some trouble with kinetic energy at the open boundaries as well (currently I'm doubting that EWperiodic will do a good job, but do not know what else to use), but maybe that's for another post
Thanks for any kind of help!