coastline value of u-, v-tau in presence of sea-ice

[mathiot]

Context

ocean surface stresses are unmask along coastlines. It seems to be needed for zdftke at least. In case of sea ice, the coastline value (on land) in some places is very different to the adjacent value (on water), especially in area with concentration close to 1. It is because the sea ice concentration interpolated at u- and v-points along coastlines is computed using sic land value (0 I think) in the averaging.

Analysis

Fix

I suggest to do a weighted average using tmask(:,:,1) to compute the sic at u- and v- points in iceupdate (trunk) and sbclim, sbclim_2 (nemo_3.6).
It means replace :

zat_u  = ( at_i(ji,jj) + at_i(ji+1,jj) ) * 0.5_wp   ! ice area at u and V-points
zat_v  = ( at_i(ji,jj) + at_i(ji,jj+1) ) * 0.5_wp

by:

! ice area at u and v-points
! land values on u-, v- points along coastline set to adjacent t-point ocean value 
zat_u  = ( at_i(ji,jj) * tmask(ji,jj) + at_i(ji+1,jj  ) * tmask(ji+1,jj  ) ) / MAX(1.0_wp,tmask(ji,jj)+tmask(ji+1,jj  ))
zat_v  = ( at_i(ji,jj) * tmask(ji,jj) + at_i(ji  ,jj+1) * tmask(ji  ,jj+1) ) / MAX(1.0_wp,tmask(ji,jj)+tmask(ji  ,jj+1))

[mathiot]

In 9774:

Error: Failed to load processor CommitTicketReference

No macro or processor named 'CommitTicketReference' found

[mathiot]

In 9778:

Error: Failed to load processor CommitTicketReference

No macro or processor named 'CommitTicketReference' found

[mathiot]

There was a bug in the suggested fix committed this morning (r9774 and r9778). Thank you Clement. tmask is obviously 3d not 2d. Cray compiler catches it as a warning not an error and assume it is the first level. So compilation and execution went fine. It is now fixed in r9782 (trunk) and r9783 (v3.6).