by Markus Kuhn
Leap seconds were introduced in 1971 to reconcile astronomical
time, which is based on the rotation of the Earth, and physical time,
which can be measured with amazing accuracy using atomic clocks.
Why do we have them?
Tidal friction within the Earth, caused by the gravitational pull
of both the Moon and the Sun, continuously slow down the daily
rotation of our planet. The modern-day definition of the unit of time,
one second, approximates the length of the day divided by
60×60×24 as it was in 1820. Since then, the length of a day has
increased by roughly 2.5 ms, and continues to increase
in the order of 1.7 ms per century.
With the invention of atomic clocks in the mid 1950s, a means of
measuring time became available that is far more stable and accurate
than the rotation of our planet about its own axis. This led to two
alternative definitions of time:
- International Atomic Time (TAI), a time scale that was
started in 1958 and as been defined uninterrupted by an international
network of atomic clocks since then;
- Universal Time 1 (UT1), an astronomical definition of time
based on the position of the Sun in the sky, a modern and refined
definition of astronomical time, replacing what used to be called
Greenwich Mean Time.
TAI and UT1 were approximately equal at the introduction of TAI in
1958. But UT1 fluctuates all the time and is slowing down increasingly
in the long term. As a result, TAI is now more than 32 seconds ahead
With two different definitions of time available, the question
arose, how exactly the civilian time zones according to which we set
our clocks should be defined. In 1971, a committee of the International
Telecommunication Union (ITU), which is in charge of setting up
international rules for the operation of radio time signals, proposed
a third definition of time, as a practical compromise:
- Coordinated Universal Time (UTC), a time scale defined such
- the difference between UTC and TAI is always an integral number of
- the difference between UTC and UT1 is never larger than 0.9
UTC is a time scale defined by atomic clocks, just like TAI, but
these atomic clocks are adjusted by inserting or deleting an
additional leap second whenever UTC and UT1 drift apart by
more than half a second. The ITU has put the responsibility of
deciding when exactly to insert or delete a new leap seconds into the
hands of the IERS.
From 1972 on, UTC quickly replaced GMT as the international
reference time for setting clocks, because it was easily available
from time-signal radio stations. UTC had the advantage of being
generated by atomic clocks, which are far cheaper and easier to
operate than the precision telescopes used earlier, without changing
the relationship to the traditional astronomical definition of time by
more than a second, an amount negligible for all but a few specialist
applications (such as pointing telescopes). Today, most civilian time
zones are defined in relation to UTC, from which they typically differ
by an integral number of hours (sometimes half hours). As a result,
civilian time zones have leap seconds simultaneously with UTC. If a
leap second is announced,
it is inserted as the 61st second (23:59:60 UTC) of the last “minute”
of the month June or December.
Leap seconds have to be inserted on average every 1–2 years during
this century. However, due to an unusual temporary acceleration of the
Earth, no leap second was needed during the 7-year gap between the end
of 1998 and the end of 2005.
An in-depth introduction into leap seconds, especially in the
context of recent proposals to get rid of them, is
Efforts to abolish leap seconds
Starting with a UTC Questionnaire
distributed by IERS in late
1999, several international scientific organizations (ITU, URSI, IAU, etc.) have initiated a
consultation process of whether the definition of UTC should be
modified. In particular, it has been proposed to remove leap seconds
from the international reference time (UTC) that is broadcast in radio
time signals, which is the basis of almost all regional civilian time
zones, thereby effectively decoupling civilian time keeping from the
rotation of the Earth.
The main arguments in favour of abandoning leap seconds are:
- leap seconds could cause disruptions where computers are tightly
synchronized with UTC;
- leap seconds are a rare anomaly to deal with, which is a worry in
particular with safety-critical real-time systems (e.g., new concepts
for air-traffic control entirely based on satellite navigation);
- exact astronomical time plays no significant role in most people’s
daily lives, and those who need to know UT1 exactly know already where
to look it up.
The main arguments against abandoning leap seconds are:
- there is so far a lack of credible reports about serious problems
caused by leap seconds;
- the assumption that UTC and UT1 differ by no more than a second is
hardwired into a huge number of deployed systems (e.g., antennas that
track satellites), which cost a lot to modify;
- system designers who worry about leap seconds simply should use
TAI instead of UTC and all we need is more easy access to TAI
- desktop computers and network servers can easily cope with leap
seconds and all we need are standardized guidelines on how to steer a
computer’s clock around them;
- we must not give up the >5000 years old human practice of defining
time through Earth’s rotation because of unfounded worries of some
air-traffic control engineers;
- abandoning leap seconds would break sundials.
Most of the discussion on this proposal since 2000 has taken place
mailing list (Steve
Allan’s 2000-2006 archive,
old rom.usno.navy.mil link).
A first meeting dedicated to the question of revising UTC and
abandoning the leap second was the ITU-R
SRG 7A Colloquium on the UTC timescale, which took place in
Torino, Italy, 28–29 May 2003.
One of the proposals discussed there was to discontinue leap
seconds in UTC in a few years time. At this point, the international
reference time scale that is today UTC would be renamed into
International Time (TI) and take over UTC’s role as the basis of local
civilian time zones.
I gave there a presentation
in which I discussed the history, current practice and available
options of handling leap seconds in networked computer systems. I
- Abandoning leap seconds in the international standard time would
not cause any problems for computer systems that are not involved in
controlling astronomical or aerospace systems, as long as
national/regional civilian time zones do the same.
- Not abandoning leap seconds would not cause significant problems
for the vast majority of computers either. Most problems and concerns
surrounding leap seconds and computers come merely from a lack of good
standardized guidelines for how to handle them and can be overcome
easily. A simple example of what such a guideline could look like is
my UTC with Smoothed Leap Seconds proposal.
It simply says that computer operating systems should slow down the
speed of the clock that normally provides UTC to applications by 0.1%
during the 1000 seconds before the end of any leap second inserted
into UTC. Those computers that synchronize their internal clock
tightly with UTC via a network time protocol usually have already
facilities to calibrate the clock speed, so this is technically not
very difficult to do.
- Much of the leap-second problem is caused by the primitive data
formats used to disseminate time, especially those of various national
LF time transmitters. They leave much to be desired and need to be
upgraded to provide better information about leap seconds, atomic time
and Earth-orientation parameters. Both TAI and UTC should be broadcast
There are also some
personal notes on the Torino meeting, which I posted to the
LEAPSECS mailing list.
The US proposal
In September 2004, the United States delegation to ITU-R working
party 7A submitted a proposed
revision of the UTC standard (ITU-R TF.460), which aims to abandon
leap seconds, for consideration by the ITU working
party in November 2005. Daniel Gambis, head of the IERS Earth
Orientation Center, used the occasion to suggested to all UTC users
that this would be a good time to send their comments on this proposal to their
ITU representatives. See Steve
Allen’s page for details.
- maintains the name UTC;
- increases the allowed tolerance between UTC and UT1 from 0.9
seconds to 1 hour;
- asks the IERS to keep UTC from deviating by more than 1 hour from
UT1 using leap hours instead of leap seconds;
- suggests that this change to the definition of UTC be implemented
by December 2007.
In my opinion, this proposal is not acceptable for a number of
- UTC is widely used in computer applications today specifically
because it lacks the very disruptive leap hours that occur in
many local time zones due to summer-time (US: daylight-saving time)
arrangements. Introducing leap hours into UTC would eliminate one of
UTC’s main advantages for such applications.
- A leap second is a fairly minor discontinuity, one that can be
corrected comfortably in computer system by a tiny change in the
software-clock speed for a few minutes. A leap hour, on the other
hand, is 3600 times more disruptive than a leap second! It is a very
major discontinuity in what is meant to be a uniform time scale.
- A fairly regular roughly annual event, such as a UTC leap second,
is an expected routine matter that every person involved with
precision timekeeping encounters and handles many times during their
lifetime. On the other hand, an extraordinary event, such as a UTC
leap hour, is a once-every-few-centuries matter that will hardly play
any role in any system designs and will therefore hit those
responsible for dealing with it largely unprepared.
- Introducing a leap hour into UTC will be perceived by computer
operators as a potentially catastrophic hazard to the correct
operation of millions of mission-critical computer systems. Our
experience with the comparatively minor concern regarding the 2-digit
year overflow from 1999 to 2000 (“Y2K bugs”) does not even begin to
describe the range of concerns, problems and costs that a leap hour in
UTC would create.
- Is it realistic to expect that the IERS or ITU-R – a small and
obscure service office for astronomers and United Nations
radio-standards agency – will have the political power (or will even
still exist) to cut an hour out of UTC when the first UTC leap hour is
due in the 27th century? After all, this will likely be an
unprecedented and extremely expensive and risky disruption. After all,
UTC is the form of time that one uses exactly to avoid the
annual summertime leaps of civilian time. Back in 1582, the pope still
had enough power to cut 10 days from the calendar, but even that took
half a millennium to be accepted everywhere. And that was before we
had computers and liability!
I cannot imagine that even the authors of this proposal can
seriously imagine that a UTC leap hour is a realistic and practically
implementable option. I can, therefore, only conclude that this
proposed revision of UTC was meant as a fake package. This proposal is
nothing but an attempt to replace UTC forever with a leap-free atomic
time. It only pretends – in the interest of diplomatic convenience –
to be nothing but a simple relaxation of the |UTC−UT1| tolerance.
While I do not believe that any modification to UTC is actually
necessary, I could in principle live with a permanent switch from
astronomic time to atomic time. However, any such proposal should
- be honest enough to actually say that its goal is to detach
civilian timekeeping entirely and forever from the rotation of Earth;
- no longer be called UTC;
- at least attempt to discuss how it proposes to future generations
to handle the unbounded and ever faster increasing gap between TAI and
UT1, which will eventually grow well above 1 day;
- be announced at least five years in advance, such that the many
systems that depend on the current |UTC−UT1| < 0.9 s limit can be
upgraded or taken out of service first, without unreasonable cost or
The 2004 US proposal failed on all of these points. It has not been
implemented and further leap seconds have since been inserted in 2005,
2008, and 2012. However, the supporters of the UTC-leap-hour idea have
not yet given up
late 2008 claimed increasing support among ITU member countries.
On the other hand, the UK, Canadian and Chinese governments appear to
have gone on record now by expressing their opposition. The next vote
on the issue was in January 2012 at
Radiocommunication Conference 2012 and preceeding
The RA-12 Plenary
of 19 January 2012 debate on the leap second decided to postpone
the issue until the next Radiocommunication Assembly and World
Radiocommunication Conference, scheduled for 2015. To be continued ...
Some of my postings on LEAPSEC
Atomic Time and the post-Gregorian calendar corrections is a
possible solution (the first that I have seen) on what to do when
local civilian times and the proposed leap-second-free International
Time start to diverge by more than a day after the year 5000. It picks
up on an idea first mentioned in What
to do if International Time hits the International Date Line?,
namely to reconcile the long-term deviation of Universal Time and
Atomic Time by dropping a 29 February whenever they diverge by more
than 12 hours. If the rotation of the Earth around the Sun is equally
stable as atomic time, this approach could potentially solve at the
same time the long-term deviation between the Gregorian Calendar and
the solar year. One question with this proposal is whether week-day
continuity should be violated at this missing leap year in civilian
local times in favour of week-day continuity in international (atomic)
leap second problems just a legend?
(Note: The quoted links have moved.)
- Geoff Brumfiel: Astronomers
leap to defence of extra seconds in time debate, Nature 423, 12
June 2003, p 671, doi:10.1038/423671a
- David Adam: What
time is it? Well, no one knows for sure, Guardian, 26 June 2003.
- Andreas Hirstein: Der Schaltsekunde schlägt die Stunde. Neue
Zürcher Zeitung, NZZ am Sonntag, 13 July 2003, pp 57-58.
- Jim Boulden: Time
to skip leap seconds?, CNN.com, 22 July 2003.
- David Adam: The
final countdown? The Age, Melbourne, 19 August 2003.
- Manfred Dworschak: Furcht vor
dem Datenschluckauf. Der Spiegel, Nr. 35, 25 August 2003, pp
- Adam Rogers: Why the
World’s Clocks Are Wrong. Wired Magazine, Issue 11.11, November
- Stephen Battersby: Time warp. New
Scientist, 22 November 2003. (letters)
- James Ryerson: Time
Gap, The. New York Times Magazine, 14 December 2003.
- Roger Highfield: Hang
on a second, what’s the real time? Daily Telegraph, 31 December
- Daniel Bächtold: Unser Kalender
braucht mehr Zeit. Tagesanzeiger, 27 February 2004.
- Karen Wright: Leap
Seconds. Discover Magazine, Vol. 25, No. 03, March 2004.
- Roxanne Khamsi: Leap
second to be added to 2005. Nature News, 6 July 2005,
- Al Sicherman: Hey, got a minute? Or maybe just a second. Star Tribune,
Minneapolis, 18 July 2005.
- Keith J. Winstein: Why
the U.S. Wants To End the Link Between Time and Sun. The Wall
Street Journal, 29 July 2005, page A1.
- Wendy M. Grossman: Time bandits. 5
- Frank D. Roylance: Clock-syncing effort buys some time. The Baltimore Sun, 8 August 2005.
- Christian Speicher: Hat der
Schaltsekunde die Stunde geschlagen? Neue Zürcher Zeitung, No.
232, p. 61, 5 October 2005.
more leap second? Heise Newsticker (English transl.), 9
Leap second talks are postponed. BBC News, 9 November 2005.
- Devin Powell: Calls to scrap the 'leap second' grow. New Scientist, 17 December 2008.
- David Rooney: A place of pilgrimage at centre of the time system. Commentary, The Times, 17 December 2008.
- Mark Henderson: How a split second could spell end of GMT. The Times, 18 December 2008.
- Every second counts. Comment, Leading Article, The Times, 18 December 2008.
- Chris Irvine: Scientists propose 'leap hour' to fix time system.
Telegraph.co.uk, 18 December 2008.
es der Schaltsekunde an den Kragen? Neue Züricher Zeitung, 11
- Leap seconds –
Their time has come. The Economist, 14 January 2012.
Radiocommunication Conference 2012
of Coordinated Universal Time (UTC) study in ITU-R
- Countries consider time out on the 'leap second'. The Examiner, 17 January 2012.
Gast: Die grauen
Herren von Genf. Die Zeit, 26 January 2012.
Thanks to Steve Allen for providing various URLs.