12/28/2017
The nine daily netcdf fields were created on discover by running the grads scripts currently in 
/gpfsm/dhome/jacarton/merra2/execs/daily/1980-2017 which produce a daily average from the
24 hourly analysis fields.  Then the cdo routine import_binary, e.g.:
cdo -b F32 -f nc4 import_binary lwgab.ctl /gpfsm/dnb32/jacarton/lwgab-1980-2017.nc
was run to create nc4 files and these were sftp'd to:
/aosc/indian/carton/merra2/daily_data/1980-2017.  Finally, each field was compared to the
corresponding field in /aosc/indian/carton/merra2/daily_data/1980-2015 to ensure that the downloads 
are correct. This procedure showed that there was an erroneous  1dy shift in a single older file:
/1980-2015/merra2_precsnoocn_1980-2015.nc

12/28/2017
I checked the min and max of three fields: precsonocn (0 to 0.003), rainocn (0 to 0.01), 
and qv2m (e-7 to 0.03).  All seem within reasonable limits so there was no need to filter.