[CF-metadata] Gregorian calendar in CF

From: Jonathan Gregory <jonathan.gregory>
Date: Mon, 02 Sep 2002 09:13:54 +0100

In the default calendar, the changeover from Julian to Gregorian is 1582-10-15
in CF, as in udunits. Steve Hankin wrote

> The discontinuity in
> calendars might make sense if you are a geopolitical historian, but it is
> nothing
> but trouble for scientific data. In the spirit of "best practices" CF should say
> something like
>
> 1. when using the default (Gregorian) calendar the time origin ("...since
> <date>") should lie on the same side of the 1582-10-15 calandar change as do
> the time coordinate values
> 2. when creating a time/date axis that crosses the 1582-10-15 calandar change you
> should use a "proleptic Gregorian" calendar -- achieving a consistently
> Gregorian behavior

I would be in favour of adopting these as recommendations in CF. (1) can be
used provided the time coordinate values do not span the discontinuity, when
clearly it is impossible. (2) could be used for model data that uses the real-
world calendar. For real-world data it might not be appropriate, but there is
no proper solution in this case, since the calendar change-over did not happen
everywhere in the world at the same time. The proleptic Gregorian calendar is
already permitted in CF.

Jonathan
Received on Mon Sep 02 2002 - 02:13:54 BST