PAPA 1D reference configuration does not work anymore. 2 problems:

• Bad variable in namelist_cfg for PAPA 1D (ln_2m does'nt exist)
• Results are wrong.

When Laplacian isopycnal tracer diffusion (the default option for ORCA2) is selected, the addition of non-zero horizontal background diffusion (parameter 'ln_ahtb_0') causes the test 'tra_ldf_iso_ad' of the corresponding tangent-linear and adjoint subroutines to fail. Without additional background diffusion the test of this option for lateral diffusion passes.

Analysis of the corresponding code revealed that the lateral diffusion coefficients are scaled by the value of variable 'rldf' (in OPA_SRC/LDF/ldftra_substitute.h90) which remains undefined in NEMOTAM (for the non-linear model the value of 'rldf' is set to 1.0). Setting 'rldf' to 1.0 at the start of each of the subroutines 'tra_ldf_tan', 'tra_ldf_adj', and 'tra_ldf_adj_tst' in module 'traldf_tam' would provide the expected values of the diffusion coefficients.

The assignment of 'rldf' with a value of 1.0 results in a failure of the test of the OPATAM subroutines for some of the options for lateral diffusion (the tests of the code in modules 'tra_lap_tam' and 'tra_iso_tam' fails, the test of the code in module 'tra_bilap_tam' completes successfully). If without definition of 'rld' the lateral diffusion coefficients are set to zero, the test of the tangent-linear and adjoint code for lateral diffusion becomes ineffective, which seems to explain why the subroutine tests for the tangent-linear and adjoint code previously did not reveal any problem with the 'tra_lap_tam' and 'tra_iso_tam' code.

As pointed out in ticket #1482, the test of the tangent-linear and adjoint subroutines of module 'traldf_iso_tam' fails after fixing the computation of the lateral diffusion coefficients or by selecting non-zero additional horizontal tracer diffusion.

Analysis of the relevant code revealed that the adjoint subroutine lacks the setting to zero of variables 'aditad' and 'adjtad' at the end of the loop over the tracer index, which would be required for it to correspond to its tangent-linear counterpart. At the end of the innermost body of the final nested loop structure (loop over the tracer index and the three spatial indices), both 'zditad(ij,jj,jk) and 'zditad(ij,jj,jk)' could be set to zero in order to make the loop correspond to the respective section of the tangent-linear counterpart of the subroutine. Further, at the end of the body of the outermost level of this loop (loop over the tracer index), the two lines 140 and 141 of the tangent-linear code should be mirrored in the adjoint code.