[MITgcm-support] Control variables in adjoint sensitivity experiments

[Nora Loose]

Hi all,

I am working with the verification experiment tutorial_tracer_adjsens in order to figure out how to set up the control variables in a sensitivity experiment. 

I am mostly interested in using surface fluxes as control variables (e.g., wind fields and precipitation). I modified the following files in tutorial_tracer_adjsens/code_ad:

1. I added the package exf to packages.conf
2. I replaced ctrl_map_ini_genarr.F by a copy of pkg/ctrl/ctrl_map_ini_genarr.F, to which I added:

	integer igen_windx, igen_precip

	igen_windx=0
	igen_precip=0

DO iarr = 1, maxCtrlArr2D
if (xx_genarr2d_weight(iarr).NE.' ') then

  if (xx_genarr2d_file(iarr)(1:11).EQ.'xx_windx’) 
    &     igen_windx=iarr

  if (xx_genarr2d_file(iarr)(1:11).EQ.'xx_precip') 
    &     igen_precip=iarr

endif
ENDDO

if (igen_windx.GT.0)
    &      call ctrl_map_genarr2d(uwind,igen_windx,myThid)

  if (igen_precip.GT.0)
    &      call ctrl_map_genarr2d(precip,igen_precip,myThid)



The compilation fails and the TAF output file tells me:


file ad_input_code.for:
28709      &      call ctrl_map_genarr2d(precip,igen_precip,myThid)
                                               ^
*ERROR* : variable precip must be available
                                               ^
*ERROR* : identifier not defined



I get the same error for swdown, lwdown, atemp and aqh. However, TAF does not complain about the variable uwind (and vwind). 
I checked the file ad_input_code.for and the common blocks including the arrays precip, swdown, lwdown, atemp and aqh are all listed there (alongside with uwind and vwind).

Does anybody have a clue what went wrong? 


Thanks in advance for your help,

Nora