Custom Query (964 matches)

The routine step3d_uv.F assigns the value of local variable AK(i,N(ng)) twice. This reassigment is redundant in the nonlinear, tangent linear, and representer model and doesn't affect the results. However, this constitutes a bug in the adjoint model since reassigments are added twice.

There is also few typos associated with 4DVAR adjustement of surface forcing fields in routines ad_wrt_his.F, tl_wrt_his.F, and cost_grad.F. Many thanks to Gregoire Broquet for reporting these typos.

The routine step2d_LF_AM3.h has an unnecessary specification of I_RANGE and J_RANGE when updating rzeta. This is not a bug. Similar cleaning is done in the TLM, RPM, and ADM versions of this routine.

Corrected a bug in nemuro.h when computing zooplankton egestion. Many thanks to Al Hermann for reporting this bug.

please relocate the command CALL SWEXITMPI within Waves/SWAN/Src/swanmain.F from line 874 to line 914.

Please look at the swanmain file in my branch for correct location. There was a potential conflict with swan exiting mpi before it collected the data from all the nodes.

wvelocity.F, near the bottom we have:

          wvel(i,j,0)=cff1*(vert(i,j,1)-
     &                      (vert(i,j,2)-vert(i,j,1))*
     &                      2.0_r8*(z_r(i,j,1)-z_w(i,j,0)))+
     &                cff2*vert(i,j,1)-
     &                cff3*vert(i,j,2)

cff1, cff2, and cff3 are coefficients for the interpolation (just numbers). vert has dimensions of m/s. So this is:

( m/s - ( m/s - m/s)* 2 * (m))   +   m/s  +   m/s

This does not look dimensionally correct. Same issue for wvel(i,j,N).