Feature/msienkie/acft bc cleanup tails

[gmao-msienkie]

Modifications made to 'GEOSdas.csm' and 'fvssi' scripts to add the 'cleanup_tail=.true.' flag to the options for the two aircraft varBC methods.
The cleanup_tail process removes any entries from the aircraft temperature bias correction file that have not been updated for more than a year. This helps limit the size of the aircraft coefficient file.

When a particular tail number has no reports the bias correction coefficients and their background error are kept the same until that particular tail number resumes reports again - the background error is not inflated. Thus, this change will not be zero diff in the case where an aircraft reappears. If a tail has been missing for more than a year it is probably beneficial to discard the bias correction since the characteristics of the aircraft equipment may have changed or the equipment may have been replaced during the downtime.

[gmao-msienkie]

The cleanup_tail action is now optional, controlled via an environment variable.
setenv CLEANUP_TAIL 1 -> cleanup_tail=.true.
if the environment variable is not set, the default is not to set cleanup_tail, and the GSI by default sets that as false.

[rtodling]

Meta,

Do we really need to make this optional? That is to have an env variable controlling this? Perhaps you want it that way to get it to be backward compatible ...

[rtodling]

... I think I understand what is controlled by this env var; but should we make the default env var 1 instead of 0 ... ideally we don't have to tell OPS to add another env var. And 0 would only be needed in cases we are trying the reproduce a previous run, no?

[gmao-msienkie]

I remember you had said something about increasing the array size would be better than cleaning up the tails. I wanted to add an option to allow people to clean tails without manually editing the gsi.rc. I agree when you are trying to reproduce a prior run you wouldn't want to be changing the ceofficients in the cftbias file in case some tail came back after being missing for a year. I am in favor with taking out the year-old 'stale' coefficients so if you feel that the default should be '1' (remove old coefficients) that is fine with me.

[rtodling]

Ok - I will merge as is, and will change the default when I make fold in other (unrelated) changes of my own.

[rtodling]

Actually, I am looking again and I am realizing that the way you changed this, when the env var is set to zero the gsi.rc after being edited will have a line:

">>>AIRCFT_BIAS<<<"

like this in the namelist and the fortran code is going to choke! We need to define the if differently ...

[gmao-msienkie]

I tested this in a run. When I had the enviroment variable set to 1 the CLEANUP_TAIL was set to TRUE. When I commented out the definition, the first time around it had CLEANUP_TAIL as TRUE because I think the job inherited the environment variable or something. When I independently submitted a segment it set the undefined CLEANUP_TAIL as FALSE. I am trying to run a segment where I have the environment variable set to 0 now so I can check what it does.

My code looks like this - it continues to set the >>>AIRCFT_BIAS<<< as needed and then appends the value of cleanup_tail to the substitution string if the environment variable is set to true (1).

 case 2:
      set cftstring = "aircraft_t_bc=.true.,"
      if ( $CLEANUP_TAIL ) set cftstring = "$cftstring cleanup_tail=.true.,"
      echo "s/>>>AIRCFT_BIAS<<</$cftstring/g"  >> sed_file
      echo 'Setting aircraft_t_bc to true, using VV.VV^2 bias correction'
      breaksw

[rtodling]

Sorry - I missed it ... all fine!

[rtodling]

ok - sorry - all good