Seth,
I don't work with ancient or model data sets, so I can't speak to the
question of UTC relative to Gregorian calendars. It seems to me that
since we already assume UTC +- a time zone offset for the time stamp in
the units attribute, that my suggestion causes no new problem.
As I think about it, I have to say that I overstated the case regarding
encoding of leap seconds. If the span from your epoch time stamp to the
last time stamp in the period of record for your dataset does not cross
a leap second date, no leap second is encoded. The conditions necessary
for one or more leap seconds to be encoded are described below.
Here's how an elapsed time can encode one or more leap seconds. Let's
consider a set of UTC time stamps and how they compare with the GPS time
counter. I'm going to set my epoch date/time to be the GPS epoch of
1980-01-06 00.00.00.0, and look at UTC time stamps running over the span
from the epoch date to the present day. The first leap second after this
epoch occurred at 1981-06-30 00:00:00.0, and there have been 16 leap
seconds total over the span from the epoch to the present day. If I use
time handling functions that don't consider leap seconds to convert
between time stamps and elapsed times (counts of elapsed seconds since
my epoch), I will find that the elapsed times calculated from my UTC
time stamps will match the GPS time count until that first date when a
leap second is encountered. After that date, my calculated elapsed times
will be less than the GPS time count by 1 second. If I had data at 1
second granularity, I would find that there are actually two UTC time
stamps at the time the leap second is applied that produce the same
elapsed time value. (This means that a time variable using these values
would actually fail the requirements for a true coordinate variable,
since the values wouldn't be monotonic.) After each leap second date the
elapsed time calculated from UTC time stamps would fall another second
behind the GPS time count. For a UTC time stamp of 2015-04-27 00:00:0.0,
the UTC-based elapsed time value would be 16 seconds less than the GPS
time count.
Grace and peace,
Jim
On 4/24/15 3:54 PM, Seth McGinnis wrote:
Jim--
Your suggestion in the last paragraph won't work. UTC is only
meaningfully defined for the real world, so the epoch timestamp in the
units attribute can only be UTC if the calendar is Gregorian...
I'm a bit confused by the way you talk about elapsed times. It seems to
me that elapsed time should be completely independent of the calendar
and the time system. I view the time coordinate as the number of
seconds (minutes/hours/days) that a hypothetical idealized clock
attached to whatever's generating the data would count from the epoch
marker, regardless of what the calendar is doing. An elapsed time would
then be the difference between two time coordinates. The calendar and
time system tell you what dates correspond to the endpoints of that
interval, but they have no effect on its length. Could you elaborate on
how an elapsed time would encode a leap second?
Cheers,
--Seth
On 4/24/15 1:22 PM, Jim Biard wrote:
Jonathan,
I think the answer depends rather heavily on the way people obtained
their elapsed times that they stored in their time variables. If they
started with time stamps and the conversion to and from elapsed times
was done with software that doesn't consider leap seconds, then the
recovered time stamps will be correct, whether they represent UTC, GPS,
TAI or something else. The problems arise when your input is elapsed
times that don't contain leap seconds (such as GPS or TAI), or when a
different scheme is used for conversion from a time stamp to an elapsed
time than for conversion from an elapsed time to a time stamp.
Given that most of our time conversions from time stamps to elapsed
times haven't taken leap seconds into account up until now, the elapsed
times in these cases *are* encoding leap seconds if the time stamps were
UTC. They aren't being reduced to true elapsed time since an epoch.
By the way, modern *nix systems provide a means to have your system
account for leap seconds when converting to and from time stamps to
elapsed times, so *nix elapsed times on some systems could be true
elapsed times. (Wheee!)
I don't think calendars are the right place to encode this. We could add
a new "time_system" attribute where you would declare whether your time
stamps and elapsed times were based on UTC, GPS, TAI, etc. If we take
this route, we should require the elapsed times to encode leap seconds
if the time system is UTC, and state that the default time system is UTC.
We could also be more strict, and say the epoch time stamp in the units
attribute must always be in UTC. The question would then be reduced to
whether or not leap seconds were counted into the elapsed times stored
in the time variable. In this case, we could add a "leap_seconds"
attribute which would have a value of "UTC" if UTC leap seconds were
counted into the elapsed times, and "none" if not. This would also allow
for some other system of leap seconds to be used. (I don't know if there
are others.) For backward compatibility, considering history, the
default value for this attribute would probably be "UTC".
Grace and peace,
Jim
On 4/24/15 12:26 PM, Jonathan Gregory wrote:
Dear all
Thanks for these interesting contributions. In CF terms, it seems that UTC and
GPS time are different calendars, because of the leap seconds in UTC. A given
YMDhms date-time will sometimes be a different number of time-units since
reference-time in the two calendars. We ought to clarify the CF document, which
current assumes that the default calendar (the real-world Julian/Gregorian one)
is UTC for modern times. Since udunits and most software doesn't support leap
seconds, as Seth says, maybe we should amend the document to point out that
the default calendar is not exactly UTC because of this, and note that it has a
fixed offset to GPS time. Then we would need to introduce a new calendar of
utc, valid only for dates since UTC was defined. We could do it the other way
round, and define the default as UTC, but that would probably not be right for
(nearly) all existing data, and not convenient for (nearly) all software. Is
this correct?
Best wishes
Jonathan
----- Forwarded message from Seth McGinnis <[email protected]> -----
Date: Thu, 23 Apr 2015 13:06:21 -0600
From: Seth McGinnis <[email protected]>
To: [email protected]
Subject: Re: [CF-metadata] How to define time coordinate in GPS?
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.9; rv:31.0)
Gecko/20100101 Thunderbird/31.6.0
Julien,
I would say that you don't need to do anything. Your data is already in
GPS time.
Strictly speaking, you can't specify a different time referential,
because CF says to specify the time units following the recommendations
of udunits, and udunits assumes that all times are relative to UTC.
HOWEVER, udunits ignores leap seconds, and time under CF is represented
as a continuous count of constant-size units from some starting point.
Which I believe means that although both spec documents say "UTC", in
practice CF-netcdf is actually representing time as GPS/TAI time instead.
Moreover, as far as I can tell, practically all the software out there
that converts netcdf times to dates is unaware of leap seconds anyway,
and even if you wanted them you would have to do extra work to get them.
So I think if you set the units attribute as given in your example,
you'll be fine. If you want to be extra-careful, you could annotate the
time variable with a "note" attribute indicating that it's GPS time, not
UTC time, with no leap seconds.
There was a discussion on the mailing list about leap seconds last July:
http://mailman.cgd.ucar.edu/pipermail/cf-metadata/2014/057556.html
Cheers,
--Seth
On 4/23/15 8:19 AM, Julien Demaria wrote:
Dear Jonathan,
I?m also not an expert on this:
?GPS, Global Positioning System time, is the atomic time scale
implemented by the atomic clocks in the GPS ground control stations and
the GPS satellites themselves. GPS time was zero at 0h 6-Jan-1980 and
since it is not perturbed by leap seconds GPS is now ahead of UTC by 16
seconds.?
http://www.leapsecond.com/java/gpsclock.htm
a more detailed explanation:
https://confluence.qps.nl/display/KBE/UTC+to+GPS+Time+Correction
Thanks in advance,
Julien
Jonathan Gregory j.m.gregory at reading.ac.uk
Thu Apr 23 07:58:09 MDT 2015
Dear Julien
Could you explain what the difference is between GPS time and UTC (for a non-
expert such as me)?
Thanks
Jonathan
*De :*Julien Demaria
*Envoy? :* jeudi 23 avril 2015 14:51
*? :* '[email protected]'
*Objet :* How to define time coordinate in GPS?
Hi,
I need to define a time coordinate variable which use the GPS time
referential instead of UTC, but I did not found how to specify this.
For the moment my variable look like this :
int64 time_stamp(rows) ;
time_stamp:standard_name = "time" ;
time_stamp:units = "microseconds since
2000-01-01 00:00:00" ;
time_stamp:_FillValue = -1L ;
Thanks in advance,
Julien
_______________________________________________
CF-metadata mailing list
[email protected]
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
_______________________________________________
CF-metadata mailing list
[email protected]
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
----- End forwarded message -----
_______________________________________________
CF-metadata mailing list
[email protected]
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
--
CICS-NC <http://www.cicsnc.org/> Visit us on
Facebook <http://www.facebook.com/cicsnc> *Jim Biard*
*Research Scholar*
Cooperative Institute for Climate and Satellites NC <http://cicsnc.org/>
North Carolina State University <http://ncsu.edu/>
NOAA National Centers for Environmental Information <http://ncdc.noaa.gov/>
/formerly NOAA’s National Climatic Data Center/
151 Patton Ave, Asheville, NC 28801
e: [email protected] <mailto:[email protected]>
o: +1 828 271 4900
/We will be updating our social media soon. Follow our current Facebook
(NOAA National Climatic Data Center
<https://www.facebook.com/NOAANationalClimaticDataCenter> and NOAA
National Oceanographic Data Center <https://www.facebook.com/noaa.nodc>)
and Twitter (@NOAANCDC <https://twitter.com/NOAANCDC> and @NOAAOceanData
<https://twitter.com/NOAAOceanData>) accounts for the latest information./
_______________________________________________
CF-metadata mailing list
[email protected]
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
_______________________________________________
CF-metadata mailing list
[email protected]
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
--
CICS-NC <http://www.cicsnc.org/> Visit us on
Facebook <http://www.facebook.com/cicsnc> *Jim Biard*
*Research Scholar*
Cooperative Institute for Climate and Satellites NC <http://cicsnc.org/>
North Carolina State University <http://ncsu.edu/>
NOAA National Centers for Environmental Information <http://ncdc.noaa.gov/>
/formerly NOAA’s National Climatic Data Center/
151 Patton Ave, Asheville, NC 28801
e: [email protected] <mailto:[email protected]>
o: +1 828 271 4900
/We will be updating our social media soon. Follow our current Facebook
(NOAA National Climatic Data Center
<https://www.facebook.com/NOAANationalClimaticDataCenter> and NOAA
National Oceanographic Data Center <https://www.facebook.com/noaa.nodc>)
and Twitter (@NOAANCDC <https://twitter.com/NOAANCDC> and @NOAAOceanData
<https://twitter.com/NOAAOceanData>) accounts for the latest information./
_______________________________________________
CF-metadata mailing list
[email protected]
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
