[MITgcm-devel] problem with 59c

[Dimitris Menemenlis]

> Not sure, but if I remember well, advcheme=7 needs an overlap of
> 4, but needs to be double on CS-grid, so it should be set to 8
> to be strictly correct.

Ouch!  Why does it need to be double for CS-grid?  Is this a corner-only issue?

> What value you are using for Olx,Oly right now ?

I am using Olx=Oly=5
I arrived at above value by running the 32x6x32 test domain with progressively 
larger values until results were indistinguishable to machine precision and also 
that there was nothing obviously different at edges and corners.

A less diffusive advection scheme is really important for Arctic solution and 
also for Equator.  With advscheme=33, spurious vertical diffusivity from the 
advection scheme dominates.  advscheme=7 gives much improved ... although not 
perfect results.  It will be interesting to try and see how prather behaves when 
ready.  Holloway claims that Prather does miracles for Arctic circulation.

D.

P.S. useRealFreshWaterFlux is now onboard on the CS510 domain! I dug minimum 
depth on faces 3 and 6 to 20 m and limit HEFF_MAX to 10 m.