id,summary,reporter,owner,description,type,status,priority,milestone,component,version,severity,resolution,keywords,cc 115,ldf coefficient,rblod,ctlod,"Flyspray 79: When the lateral mixing coefficient on the dynamics is a function of x,y (key_dynldf_c2d) the mixing coefficient is made proportional to grid spacing (if laplacian) or the cube of grid spacing (if bilaplacian). Logically, the same should be done for the mixing coefficient on tracers when key_ldftra_c2d is active. This is not what is done in ldftra_c2d.h90: for the biharmonic it is OK but the laplacian is left constant. This in not a bug but rather an inconsistency; however it lingers in the code (maybe we mentioned it a long time ago before flyspray was in operation?) I attach the DRAKKAR version of the routine. Flysrapy 75: 2 things to do: # make the coefficients ahm & aht proportional to Dx/Dx3 for Laplacian/Bi-Laplacian operator in routines ldf[tra/dyn]_c[2/3]d.h90 # in doing that, we must avoid numerical problems in the case of strong anisotropy (in north side), take the MAX(e1[tuv],e2[tuv]) of the scale factor in both direction as below for the Laplacian in ldftra_c2d.h90 ! proportionnal to the maximum grid spacing zdx_max = MAXVAL( e1t(:,:) ) IF( lk_mpp ) CALL mpp_max( zdx_max ) ! max over the global domain za00 = aht0 / ( zdx_max ) DO jj = 1, jpj DO ji = 1, jpi zeumax = MAX( e1u(ji,jj), e2u(ji,jj) ) zevmax = MAX( e1v(ji,jj), e2v(ji,jj) ) zetmax = MAX( e1t(ji,jj), e2t(ji,jj) ) ahtu(ji,jj) = za00 * zeumax ! set ahtu = ahtv at u- and v-points, ahtv(ji,jj) = za00 * zevmax ! and ahtw at w-point (idem T-point) ahtw(ji,jj) = za00 * zetmax ! END DO END DO ",Enhancement,closed,normal,,OCE,v2,,fixed,,