Hi @ChrisBarker-NOAA : thanks ... but my question was slightly different. 
Suppose, for instance, that the data records the frequency of some event (e.g. 
`frequency_of_lightning_flashes_per_unit_area` which has units of `m-2 s-1`) 
and the user wants to know the number of events in the elapsed time interval. 
Taking the two times as `0` and `2 minutes since 2016-12-31 23:59:00`, your 
answers imply that the correct approach is to convert the time to UTC 
time-stamps to compute the elapsed time of 121 seconds and use that in his 
calculation rather than using the numerical time coordinate directly. 

This is rather unwieldy, and likely to be ignored. It would also mean, if the 
user went through the process, he would get a different answer from that 
obtained by the data provider who has declared his intention to ignore leap 
seconds. 

As you suggest, this calendar would have to be restricted specifically for time 
values computed directly from the UTC time stamps. Someone who is taking 
measurements at fixed intervals from a starting point, and would need to use a 
different calendar. There also appears to be an assumption that the user is 
storing raw observations, as any calculation the user does which involves time 
is likely to be affected by the neglect of the leap second. Do you think the 
idea is to have the use of this calendar restricted in this way?

To me, this looks as though it would be better encoded with a new standard 
name, e.g. `time_stamp`, which should make it  clearer to users that a direct 
delta of the variable values is not the same as a time delta.

-- 
You are receiving this because you commented.
Reply to this email directly or view it on GitHub:
https://github.com/cf-convention/cf-conventions/issues/148#issuecomment-437272623

Reply via email to