Thanks all for clearing up my confusion on this issue, and for
contributing their own preferences.
Seth, I agree that midpoint labels on a climatological time
coordinate are more human-friendly than endpoint labels.
For largely the same reasons as you. Including that if someone uses
the midpoint labels on a plot, it seems more intuitive that the
data refer to the whole climatology than if endpoints are used.
However, others may think endpoints are just as friendly, or
that the perceived lacked of ambiguity (no even/odd year
questions) in endpoints outweighs the user-friendliness of midpoints.
Both endpoints and midpoints are defensible choices and
CF allows either. The CF guidance is that the choice
be "representative" of the climatology. Given that both
practices are already embedded in different datasets, I
am averse to CF becoming more restrictive and requiring
one or the other. I would support a consensus statement
that midpoint-ish values are "recommended not required"
yet I think gathering that consensus is unlikely.
Best,
c
Hi all,
I would argue that time coordinates should always go somewhere in the
middle of the interval, not at either end.
Certainly, putting the time coordinate at one end of the interval is the
easy and unambiguous choice when generating data files. But in my
experience, it tends to cause confusion and problems when *using* the
files, particularly with software that automatically generates labels
and when aggregating the data to longer timescales.
The basic problem is that a time coordinate at the end of the interval
can be ambiguous to the end user as to whether it's at the beginning or
the end, and if you're not the person who created the data, it can take
a fair amount of digging to figure out which it is. Worse, it's easy
for the user to make an incorrect assumption and not realize it.
Putting the time coordinate in the middle of the interval makes this
kind of confusion much less likely, because it's more in line with human
intuition about what interval the coordinate belongs to.
Using the midpoint can require more logic on the data creator's part to
get right (e.g., what do you do when the interval is an even number of
cycles long), but I think it's a much better approach in terms of usability.
Cheers,
--Seth
On 4/30/2015 2:13 AM, David Charles Hassell wrote:
Hello Chris, Charlie, Karl, ....
Personally, I prefer to use the first (or last) year, as this is
unambiguous. Ambiguities about how to define the mid-year could arise
depending on whether number of years in the climatology is odd or
even. The mid-year could, of course, be carefully defined, but the
definition would be arbitrary, I think.
(I used "year" here, but the same argument applies to "day")
All the best,
David
--
Charlie Zender, Earth System Sci. & Computer Sci.
University of California, Irvine 949-891-2429 )'(
_______________________________________________
CF-metadata mailing list
[email protected]
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
