[MITgcm-devel] conversion term

[Jean-Michel Campin]

Hi,

I would like to have an accurate diagnostic of the conversion term,
Potential Energy -> Kinetic Energy: = Integral of -g.w.(rho_k+rho_k-1)/2
for the ocean or -(omega/g).Delta_PI.(Theta_k+Theta_k-1)/2 for the atmosphere.
I see the advantage of this diagnostic for the atmosphere, 
since it needs to be accounted for in the energy budget and is also the
only source of KE (right ?). But it might be interesting also for the ocean.

I have few questions:

1) I don't know if this is the correct expression.

2) we have a sort of diagnostic, in the monitor, at least for the atmosphere,
but it only samples this term every monitorFreq.
 * should we change the monitor (for this specific term) to allow to average 
   all iterations, and only print out every monitorFreq (this will give us 
   only a global integral).
or
 * should we add a 3.D time average diagnostic of this term.
   I can see the advantage of having the mean vertical profile of the 
   conversion term, but don't know how to interpret the local (3.D) term.

3) with this later option, is it a good idea to add a 3.D, time_ave 
   diagnostic of rho ?

4) atmosphere: if I don't use the ideal-gas equation in calc_phi_hyd 
   but account for specific humidity q, I should also account for q in the 
   conversion term (presently missing in the monitor diagnostic) ?

5) r* coordinate: the wVel that is in common bloc is in fact w* ; 
   the true vertical velocity is w = (r*-Rlow)/H d.eta/dt + w* ;
   the (main) reason for this is that we need w* to advect a tracer.
  I think the conversion term should be computed with w and not w*.
   it's not very clear for me what is the correct expression, but I am 
   almost sure that -g.w*.rho_bar is wrong, because the horizontal integral 
   of w* is not zero, meaning that this expression change i(but is not
   supposed to) when I add a constant value to the density field.
  It makes sense for me to change wVel so that wVel=w (instead of wVel=w*),
   because:
    a) w is directly computed from the continuity equation, and then w*
       is computed. => wVel=w would be simpler.
    b) for the vertical advection of momentum, we need first to go back to 
       w, and from w we compute a w* at the W location and and an other w* 
       at S location. Again, wVel=w would be simpler.
    c) for the diagnostics it's nice when the horizontal integral of
       wVel is zero. And for the time_ave diagnostics of vertical 
       advection (WTtave, WStave), they needs to be replaced anyway by
       what comes from the gad_ routines.
   For now, I will not change wVel, mainly because I don't know
    what is needed for the Non-hydrostatic z* (w* or w ?). 
    
Comments and suggestions ?

Jean-Michel