Build-namelist calculation of CLM dtime assumes NCPL_BASE_PERIOD=ndays

[billsacks]

Bill Sacks < sacks@ucar.edu > - 2017-01-16 10:21:20 -0700
Bugzilla Id: 2404
Bugzilla CC: lipscomb@lanl.gov, mvertens@ucar.edu, rfisher@ucar.edu,

From Bill Lipscomb:

In file …/clm/bld/CLMBuildNamelist.pm, l. 1917, the base period is hardwired to 3600 * 24. This means (if I understand correctly how dtime is computed) that if LND_NCPL is set to a large value (consistent with sub-daily atm coupling), I get a much shorter CLM time step than I want.

--

I agree with his assessment. Mariana is putting in a fix for a similar problem for mosart and rtm, but these are using some new shared python code. A similar fix is needed in CLM's perl-based build-namelist. It amounts to about 10 lines of code that need to be added to determine the basedt based on NCPL_BASE_PERIOD (I just don't remember what those 10 lines of code look like).

This is unlikely to affect users much: Bill Lipscomb was doing something unusual by changing NCPL_BASE_PERIOD. However, my understanding is that this will bite anyone who does change NCPL_BASE_PERIOD.

[billsacks]

Erik Kluzek < erik@ucar.edu > - 2017-01-17 17:38:46 -0700

This will be fixed after the science freeze.