- reference state need to be hydrostatically balanced
- reference state function of height only
(- reference state satisfy the governing equations for an atmosphere at rest)
- use ref state of idealized stuff ?

attention dans start_em

c'est l'avant-dernier level qui compte

problem
- souci de temperature negative parce que base_lapse trop eleve
- repartition verticale pas bonne, meilleure avec exponentielle


fin de compute_eta (plus rajouter des -2)

         !  Add those 2 levels back into the middle, just above the 8 levels
         !  that semi define a boundary layer.  After we open up the levels,
         !  then we just linearly interpolate in znw.  So now levels 1-8 are
         !  specified as the fixed boundary layer levels given in this routine.
         !  The top levels, 12 through kte are those computed.  The middle
         !  levels 9, 10, and 11 are equi-spaced in znw, and are each 1/2 the
         !  the znw thickness of levels 11 through 12.

         DO k = kte-2 , 9 , -1
            znw(k+2) = znw(k)
         END DO

         znw( 9) = 0.75 * znw( 8) + 0.25 * znw(12)
         znw(10) = 0.50 * znw( 8) + 0.50 * znw(12)
         znw(11) = 0.25 * znw( 8) + 0.75 * znw(12)