[CF-metadata] representing model times in climate models

From: Jonathan Gregory <j.m.gregory>
Date: Fri, 30 Mar 2007 09:22:06 +0100

Dear Jon

If the model has a seasonal cycle, it must have a calendar, so the appropriate
unit is of the form "time-unit since reference-time", with the calendar
specified using the calendar attribute. Indeed, many models use 360-day years,
which is one of the possibilites supported by the calendar attribute. The use
of the reference time ties the times to the seasonal cycle. The year itself
might be arbitrary, as you say - it often is, in control runs.

"year" is not a good unit to use, because in udunits it does not mean a
calendar year, but exactly 365.242 days, which is almost certainly not what
would be intended. It is better to use days or shorter units.

For models with no seasonal cycle, CF permits calendar="none".

Cheers

Jonathan
Received on Fri Mar 30 2007 - 02:22:06 BST