LIM3 important bug fix (crash)

[clem]

Dear all,

we just found bugs in LIM3 with the capacity of crashing the code after several months/years.
The bugs seem to affect preferentially higher resolution simulations but problems are also encountered for ORCA2 after some years.
 
The reasons are numbered:

1) Ice with a concentration lower than 1.e-10 is systematically forced to melt because it creates issues when divisions by concentration are encountered in the code. But divisions by thickness (in thermodynamics) also caused serious issues. Therefore we also melt ice with thickness and volume lower than 1.e-10. This bug fix does not alter the heat, salt and mass conservations of the ice-ocean system.  

2) A minimum of ice salinity was not strictly defined before thermodynamics. It could lead to erroneous enthalpy/temperature and therefore to abnormal growth rates

3) Ice age was totally wrong. Therefore, the code was trying to write huge values in output (like 1.e36) and failed.

4) the heat diffusion scheme can sometimes result in ice temperature very slightly higher or equal to the melting point. In this case, ice is supposed to melt but it was not taken into account explicitly.  

We have tested the bug fixes on several years of a very high (2-3km) regional simulation as well as on ORCA1 and we haven't encountered any more issues.

Commits are underway

Clem & Martin

[clem]

fixed in rev 5202