RTTOV12 support initial commit and tested against RTTOV11 version for clear-sky test

[david-rundle]

I've started this pull request to kick off a discussion about what's left to do regarding RTTOV support and to advertise the fact that the branch is here if any interested in the COSP community would like to implement specific features that they would like to use.

There isn't recent reference data for RTTOV in COSP v2 so I had to use some old data from v1.4 from when I did the RTTOV integration. The calls to write out rttov_tbs in the netcdf file seem to be missing (by design?) also. I have left them out for now but can presumably add them back as a result of discussion here. There were some small differences (max <0.5K typically <0.1K) suggesting that the implementation is probably correct but with limited test data it's difficult to say conclusively.

[alejandrobodas]

Hi @david-rundle Thanks for this pull request. Currently working from home without access to an RTTOV library. I have built the standard COSP2 driver, without RTTOV support, and run cosp2_test. The first time I run it, I forgot to set Ltbrttov to .false. and I got the following non-critical errors:

ERROR: COSP input variable: cospgridIN%o3 contains values out of range ERROR: COSP input variable: cospgridIN%emis_sfc contains values out of range ERROR: COSP input variable: cospIN%u_sfc contains values out of range ERROR: COSP input variable: cospIN%v_sfc contains values out of range ERROR: COSP input variable: cospIN%lat contains values out of range

I guess this is caused by a lack of initialisation of these variables to sensible physical values when RTTOV is called when COSP is built without RTTOV support. They go away when Ltbrttov is set to .false.

I have looked at the changes and I don't see any major problems. I'd prefer to have the output in the standard netcdf files, and remove all the write statements. This will allow regressions tests in future developments.