Error in Qsr when using key_vvl

[gm]

For optimisation purposes, the coefficient etot3 used to compute the divergence of the downward solar irradiance (traqsr.F90) is computed one for all in case of 2 bands penetration or in RGB with constant chlorophyll. This is OK in fixed in time model level, but not in key_vvl case.

In this vvl case, etot3 should be systematically recomputed at each time step to have the exact solution. Since the change in the depth of model level is quite small (a multiplicative factor of 1+ssh/H is used), it is a not so bad approximation to assume a constant etot3 in the 2 cases mentioned above. Nevertheless, in the initialisation phase of traqsr (tra_qsr_init) the depth and scale factor used (fsdepw, fse3t) should be replaced by the reference one (fsdepw_0, fse3t_0) in order to ensure the model restartability in those cases.

Note that such an approximation become rather strong for shallow seas and thus coastal studies. A specific vvl case can be added in which etot3 is computed at each model time step...

to summarise :

• short term : simply change fsdepw & fse3t into fsdepw_0 & fse3t_0 in tra_qsr_init ONLY, to ensure restartability

• medium term : modify tra_qsr to recompute systematically etot3 in vvl case

[nicolasmartin]

Milestone 2010 Stream 0: Coastal ocean ready deleted