The GLS scheme has been developed in v3.1, i.e. with an implicit computation of the bottom friction. Since v3.2 the bottom friction is explicit, therefore there is no need of wbot(u and v) saved in restart. Furthermore, the bottom friction is now proportional to the before velocity, not the after velocity (as stated in dynzdf_imp.F90) or to now velocity (as stated in zdfbfr.F90).

In addition, rn_hbro defined in zdfbfr.F90 is only used in zdfgls.F90 and it is not a namelist parameter. It should be defined in zdfgls and renamed hbro

In my own version I have applied a bug fix from Sebastien Masson (personal comm.) concerning zdfddm. I see that in the nemo3beta, the bug is still there. So, as a reminder, this concerns the following lines:

    153          ! Update avt and avs
    154          ! ------------------
    155          ! Constant eddy coefficient: reset to the background value
    156 !CDIR NOVERRCHK
    157          DO jj = 1, jpj
    158 !CDIR NOVERRCHK
    159             DO ji = 1, jpi
    160                zinr = 1./rrau(ji,jj,jk)
    161                ! salt fingering
    162                zrr = rrau(ji,jj,jk)/hsbfr
    163                zrr = zrr * zrr
    164                zavfs = avts / ( 1 + zrr*zrr*zrr ) * zmsks(ji,jj) *zmskf(ji,jj)
    165                zavft = 0.7 * zavfs * zinr
    166                ! diffusive layering
    167                zavdt = 1.3635e-6 * EXP(4.6*EXP(-0.54*(zinr-1.) ) )   &
    168                                  * zmsks(ji,jj) * zmskd1(ji,jj)
    169                zavds = zavdt * zmsks(ji,jj)   &
    170                      * ( (1.85 * rrau(ji,jj,jk) - 0.85 ) * zmskd3(ji,jj)   &
    171                        +  0.15 * rrau(ji,jj,jk)          * zmskd2(ji,jj)  )
    172                ! add to the eddy viscosity coef. previously computed
    173                avs (ji,jj,jk) = avt(ji,jj,jk) + zavfs + zavds
    174                avt (ji,jj,jk) = avt(ji,jj,jk) + zavft + zavdt
    175             END DO
    176          END DO

Cheers,

Jean-Marc

Amongst some other constants, xlsn (volumetric latent heat fusion of snow) is only declared in OPA_SRC/DOM/phycst.F90 if key_lim3 is NOT defined. However, in coupled mode (key_coupled) the variable is referenced in OPA_SRC/SBC/sbccpl.F90, thus preventing the code from compiling.