Time as dimension variable

• To: netcdfgroup@xxxxxxxxxxxxxxxx
• Subject: Time as dimension variable
• From: raymond@xxxxxxxxxxxxxxx (D. J. Raymond)
• Date: Tue, 4 Feb 1992 18:13:21 MST

Hi,

     Dan Packman raises an important issue in that a single precision
variable doesn't give enough accuracy for time in some circumstances.
A more elegant solution than having two time variables would be to have
double precision dimension variables.  I think that IEEE double precision
has enough bits to keep millisecond time for 10000 years -- it takes
49 bits to do this.  If dimension variables can be only single or
double precision, I would choose double precision.  This would not
waste much space in situations where such precision wasn't needed, so
the overhead would be small.

Dave Raymond
raymond@xxxxxxxxxxxxxxx