There are several bugs in the update of tidal forcing in sub barotropic loop. Noteworthy is a missing argument in the call to upd_tide (refreshment of tidal potential) in dynspg_ts that implies:

• No resfreshment of tidal potential during barotropic loop (value frozen at central time step, despite a call every barotropic step).

• If ln_bt_fw=.false., same problem as above but there's a very significant time lag of 2 * nn_baro * rdt seconds in the past !

Users using tides are stongly advised to track changes related to this ticket.

Using the standard implementation of isoneutral diffusion, the fields udiff_heattr and vdiff_heattr (saved as somehdif) created in traldf_iso.F90 have the wrong sign. This problem is evident in trunk, 3.4.1 and 2012/dev_MERGE_2012, and has existed for some time, I believe. It gives, e.g. at 1 deg, northward heat transports in the Atlantic 0.07 PW too high.

The problem is apparent (i) Looking at traldf_iso.F90 the updating of the tracer tendency has

ztra = zbtr * ( zftu(ji,jj,jk) - zftu(ji-1,jj,jk) + zftv(ji,jj,jk) - zftv(ji,jj-1,jk) ) pta(ji,jj,jk,jn) = pta(ji,jj,jk,jn) + ztra

implying that zftu and zftv are opposite to the fluxes.

However these fluxes are *added* to give the diagnostic:

z2d(:,:) = 0._wp DO jk = 1, jpkm1

DO jj = 2, jpjm1

DO ji = fs_2, fs_jpim1 ! vector opt.

z2d(ji,jj) = z2d(ji,jj) + zftv(ji,jj,jk)

END DO

END DO

END DO z2d(:,:) = zztmp * z2d(:,:) CALL lbc_lnk( z2d, 'V', -1. ) CALL iom_put( "vdiff_heattr", z2d ) ! heat transport in i-direction

(ii) Note also (l193-196 of traldf_iso.F90) that

zftv(ji,jj,jk) = ( zabe2 * zdjt(ji,jj,jk) &

& + zcof2 * ( zdkt (ji,jj+1) + zdk1t(ji,jj) & & + zdk1t(ji,jj+1) + zdkt (ji,jj) ) ) * vmask(ji,jj,jk)

where (l179) zabe2 is +ve and (l145)

zdjt(ji,jj,jk) = ( ptb(ji ,jj+1,jk,jn) - ptb(ji,jj,jk,jn) ) * vmask(ji,jj,jk)

so the part of zftv (zabe2 * zdjt(ji,jj,jk) ) from horizontal gradients in T has the *same* sign as the T-gradient, so it must be the opposite of the flux.

and (iii) diagnosed isoneutral heat fluxes are southward across the Gulf Stream, which makes no sense.

in dynspg_ts:

zssh_frc(:,:) = zssh_frc(:,:) + ssh_iau(:,:)

should be:

zssh_frc(:,:) = zssh_frc(:,:) - ssh_iau(:,:)

since

ssha_e(:,:) = (  sshn_e(:,:) - rdtbt * ( zssh_frc(:,:) + zhdiv(:,:) )  ) * tmask(:,:,1)