Vertical mass fluxes issues

[ymamoutos]

Greetings to all,

here at the University of the Aegean we have been using ROMS to simulate processes taking place in deep secluded sub-basins during and following dense-water formation episodes.
We try to close the mass budgets during and following deep-water formation episodes.
In order to understand the problem we are facing, one has to consider a deep sub-basin, secluded from lateral advection below a certain depth, let's say 600 m.
Now, during a deep-water formation episode, we estimate the convective mass flux bringing "new" dense water in the basin below 600 m by multiplying w with rho and integrating over the horizontal area A(600m) within the 500 m closed isobath lineating the "upper lip" of the basin.
We believe that (in a parallel to Reynolds averaging way) the volume flux estimated in this way (just by integrating w over the surface A(600m) should be zero (as the deeper-then-600 m part of the basin cannot change volume) but the "convective" mass flux should be positive (downwards) during the deep-water formation episode, thus transfering a positive mass anomaly to the deep basin.
Following the deep-water formation episode, the integral of w should remain zero, for the same as before reason. During that period, we expect that the convective mass flux should be minimized and the vertical upward (negative) mass flux should be dominated by turbulent diffusion of mass, which we also estimate from the model using AKt * drho/dz. However, we find that convective mixing, despite changing sign to negative (upwards) during the stagnation period (following the deep-water formation episode), still dominates by far the vertical mass flux process, being many orders of magnitude larger than turbulent diffusive mixing.
So, we have two questions:
(a) How is it possible that volume does not seem to be conserved within the deep basin? Even during deep-water formation, narrow "chimneys" of positive (downward) w should be balanced by broader areas of small upward w, thus the horizontal integral of w should be zero at all times.... Could this problem be related to the hydrostatic approximation not being valid during formation periods? But even during stagnation periods, the mean w (horizontal integral of w divided by the surface area at 600 m of the basin) is not zero, thus this should not be related to the non-hydrostatic nature of the flow during convection.
and
(b) Why is convective mass flux dominating over turbulent diffusive mass flux, even during stagnation periods? We suspect that it might be a scale problem, i.e. that we should include a large part of the convective flux into the turbulent one... But could (a) and (b) be related?
In order to provide some evidence, we attach three figures of (1) mean vertical velocity, (2) mean convective mass flux, and (3) mean turbulent mass flux versus time - the formation period ends somewhere mid April). Again, by saying "mean" we mean horizontal surface integral over the 500 m deep surface area of the basin, divided by the same area.
Plotting the time series of the mean vertically and horizontally integrated density below 600 m we have certified that the basin indeed gains mass during formation and loses mass during stagnation.
Any help or suggestion will be highly appreciated,
Thank you very much in advance,
Yiannis