Bug routing for case coupling to WRF

[jgipsl]

In ​rev 947 a bug for the river routing was introduced (already present at branche MergeHydro). In coupled mode with LMDZ the model stops during interpolation of routing fields. In offline mode, the model did not stop but the interpolation was not done correctly.

After investigations we found that the test for determining if global case were wrong :

IF ( domain_minlon <= -179.5 .AND. domain_maxlon >= 179.5 .AND. &
      &  domain_minlat <= -89.5  .AND. domain_maxlat >= 89.5 ) THEN
       ! Here we do the entire globe

Therefor, in commit ​rev 1028 we now use the variable global to determinate case. The variable global is calculated in grid_stuff called from the offline driver. In coupled case, the default value gloabl=TRUE is used. With this commit, the case ORCHIDEE offline and LMDZ-ORCHIDEE are correct.

NB! If coupling to WRF the bug is still there. Calculation of the variable global must be added.

[jpolcher]

This is a quick change and we can use the variable "global" in grid.f90.
This can be done quickly and checked.

[jgipsl]

Following commit is slightly related to this ticket with a correction related to the variable global. [4744]. 2 corrections have been done :

• Missing bcast(global) : shold not have any impact as the variable global currently is used only in the routing scheme which is done on master proc only.
• Error in argument to haversine_singlepointploy and haversine_reglatlontoploy (global versus globalg) : I have not seen any impact on this correction using RegLatLon grid.

Also, globalg is now intent(inout) because this is needed in further developements by Jan for unstructurated grids.

[jgipsl]

This ticket will be closed. The current version of grid module in the trunk has evolved a lot since the ticket was created. WRF-orchidee has not been tested in the trunk with current version.