VERY IMPORTANT: Corrected usage of DIURNAL_SRFLUX in 4D-Var

[arango]

Corrected a bug in the 4D-Var algorithms when DIURNAL_SRFLUX, FORWARD_FLUXES, and PRIOR_BULK_FLUXES are activated.

Recall that the DIURNAL_SRFLUX option is used to modulate the diurnal cycle to input daily-averaged shortwave radiation flux data. In the adjoint-based algorithms, the FORWARD_FLUXES option is used to process surface fluxes for the tangent linear, representers, and adjoint models from the nonlinear model background trajectory stored in the quicksave (QCK) NetCDF file. See ​src:ticket:869 for more information. The PRIOR_BULK_FLUXES option is used to impose the initial background trajectory (prior) surface forcing fields stored in QCK in subsequent outer-loops (Nouter>1) and the analysis phase, as also explained in ​src:ticket:869.

The routines set_data.F, tl_set_data.F, rp_set_data.F, and ad_set_data.F were corrected to avoid modulating the shortwave radiation diurnal cycle when the above options are activated.

In tl_set_data.F, rp_set_data.F, and ad_set_data.F we now have:

#   if defined DIURNAL_SRFLUX && !defined FORWARD_FLUXES
!
!  Modulate the averaged shortwave radiation flux by the local diurnal
!  cycle.
!
      CALL ana_srflux (ng, tile, iTLM)
#   endif

while in set_data.F we get:

#   ifdef DIURNAL_SRFLUX
!
!  Modulate the averaged shortwave radiation flux by the local diurnal
!  cycle.
!
      IF (Lprocess) THEN
        CALL ana_srflux (ng, tile, iNLM)
      END IF
#   endif

These changes are necessary to avoid increasing the shortwave radiation values in other 4D-Var outer-loops and the analysis step.

Many thanks to Julia Levin and Gordon Zhang for bringing this problem to our attention.