Only part of the array e_i (and others) stored in restart file

[anders]

The array e_i is allocated with third dimension as 1:nlay_i+1 (NEMO/LIM_SRC_3/ice.F90) but only 1:nlay_i is stored in the restart file (NEMO/LIM_SRC_3/limrst.F90). This leads to uninitialised memory will be used in LIM3 after a restart (e,g, in NEMO/LIM_SRC_3/limtrp.F90 where third dimensions is left unspecified in summations: zei = SUM( e_i(ji,jj,:,jl) )), potentially leading to strange results or numerical exceptions. In my case there are values in the range 1E280, which soon gives NaN.

The mismatch between allocated size and stored size in restart files applies to the arrays s_i, sxe, sye, sxxe, syye, sxye as well, but I have not confirmed that these ones are actually causing problems.

This bug has been investigated with a slightly modified version of revision 4908 of trunk but relevant parts of the code seem unchanged in current trunk (r5160).

[clem]

Concerning e_i, we only use e_i(1:nlay_i) in limtrp and (I think) in the other routines. The bug you are mentioning was corrected some time ago (especially in the trunk). However, I'll investigate the other arrays sxe, sxxe etc to check if there is an issue here.

[clem]

Ok, a number of variables are allocated with a dimension 1:nlay_i+1 (found in both ice.F90 and thd_ice.F90), namely t_i, s_i, e_i, sxe, sye, sxxe, syye, sxye, t_i_1d, s_i_1d, q_i_1d. This is only useful for q_i_1d ( i.e. e_i used in the thermodynamics ). Nevertheless, it should not impact your simulation since the value at nlay_i+1 is never used/recorded (in the LIM3 version of the trunk).

For more clarity, one can remove useless nlay+1 dimension.

[clem]

not a bug but fixed anyway in ticket #1492