We at SSESCO have decided to implement our netCDF time storage as
a string basetime (ala udunits) "UNITS since 1994-01-14 9:06:01.1 -6:00"
and a double float offset from that which can be any units set in the bastime.
Our visualization software expects the time/date to come in as a
unix-time format (seconds since 1970) so we wrote some conversion
routines.
Liam Gumley writes:
> I agree that using Julian Day, and number of milliseconds since midnight
of
> the Julian day is a good way to store time data.
>
> However, does it introduce some possible confusion? The following is
from
> "Numerical Recipes in C", 2nd Ed., p. 12.
>
> "Astronomers number each 24-hour period, starting and ending at noon,
with
> a unique integer, the Julian Day Number."
>
> That is, Julian Day Number 2440000 starts at 1200 GMT on 23 May 1968,
not
> at 0000 GMT.
>...
This could be resolved by adding a basetime string to the variable
idicating whether the basetime was 1-1-1 12:00:00 -5:00 or 1-1-1 0:00:00 -5:00
..
what are the limitations to the 'udunits' approach that we have used?
--bcl blincoln@ssesco.com ---------------------- SSESCO: (612) 342-0003