Probably what should instead be listed
in the attribute is the actual length of the axis as defined.
I see. That would increase the chance of compatibility with other software.
Note
the "axis span" shown here - this is what is used in modulo
calculations.
yes? define
axis/t0=1-JAN-0001/calendar=noleap/edges/t=0:365/npoints=73\
/units=days/MODULO pntd
[ . . . ]
Axis span (to cell edges) = 365 (modulo length = axis
span)
L T TBOX
TBOXLO TSTEP (DAYS)
1> 03-JAN 12:00:00 5 01-JAN
00:00:00 2.5
2> 08-JAN 12:00:00 5 06-JAN
00:00:00 7.5
Aha. That makes perfect sense. As usual, Ferret assumes the box centered about each gridpoint. Thanks.
By the way, I've just noticed that Ferret doesn't include the point_spacing = "even" attribute or the cell-boundary variable (PNTD_bnds for the above example) in the output netCDF file for the above time axis.
Best regards,
Ryo