[MITgcm-support] Virtual Carbon Flux Query

[Lauderdale, Jonathan]

Hello MITgcm fans,

I'd like some help clarifying a few points about handling concentration/dilution of tracers at the surface of a model with a rigid lid:

I have run some thought experiments that produce strong freshwater relaxation, using a coarse resolution model with rigid lid and interactive atmospheric CO2 box. When I calculate changes in carbon budgets (atmosphere, saturated, disequilibrium, soft tissue and carbonate pools) they do not integrate to zero unless the virtual fluxes are included. Furthermore these virtual fluxes are of the same order of magnitude as the other C anomalies. 

Currently I use the ALLOW_OLD_VIRTUALFLUX method for DIC and ALK in the dic package (dic_surfforcing.F). It seems the default method now is to let the ptracers package handle virtual fluxes in ptracers_forcing_surf.F by setting PTRACERS_EvPrN (to zero?) in data.ptracers and providing a uniform surface value in PTRACERS_ref? If I were to change, would I do this for all tracers or just DIC and ALK as above? Having said that, the old method includes the relaxation fluxes in the calculation whereas it appears that the ptracers package doesn't (and I have read about potential issues with the add2EmP flag on this list).

I am a bit confused about justifying these losses/gains of carbon by the model - is it ok to simply state that virtual fluxes are a necessity to account for concentration/dilution of tracers in the surface layer in this type of model and therefore maintain global conservation? Am I correct in thinking that if all else is equal, a rigid lid model with virtual fluxes should produce the same carbon partitioning as in a model where surface volume can change because the former changes the surface concentration explicitly by fluxing tracer through the surface while the latter handles it implicitly by changing the surface volume?

Maybe there is a more suitable configuration that I could use? Would I have to go for a nonlinear free surface and rstar coordinates to avoid virtual fluxes entirely or could I achieve something similar using the implicitFreeSurface/realFreshWaterFlux/exactConserve options?

Thanks very much for your help!

Jonathan
___________________________________________________________________________
Dr. Jonathan M. Lauderdale
Postdoctoral Research Associate
School of Environmental Sciences: Oceans and Ecosystems
University of Liverpool
4 Brownlow Street, 
Liverpool, L69 3GP, U.K.
Tel: +44 (0)151 794 4086
Email: <mailto:jonathan.lauderdale at liverpool.ac.uk>

UK Polar Network: <http://www.polarnetwork.org/about>
Association of Polar Early Career Scientists <http://www.apecs.is/>
___________________________________________________________________________