National Physical Laboratory

The leap second debate

In November 2015, the International Telecommunication Union's Radiocommunication sector (ITU-R) will hold a meeting to decide the fate of leap seconds. Leap seconds have long been a subject of debate, and some organisations have argued that our increasing dependence on precision timekeeping technology makes the leap second system costly and difficult to implement.

Technologies that produce or require precise timekeeping currently have to be adjusted whenever a leap second is added, a process that leaves the systems vulnerable to human error. But if leap seconds were removed, the length of a 'clock day' (86,400 SI seconds) would slowly drift away from the 'solar day' defined by the movement of the sun across the sky.

What is the difference between GMT and UTC, and where do leap seconds come in?

Greenwich Mean Time (GMT) is mean solar time measured on the Greenwich meridian, and is based on the Earth's rotation. A mean solar day is defined so that on average over a year, the Sun is directly overhead on the Greenwich meridian at noon, and that day is then broken down into hours, minutes and seconds. However, the rotation speed of the Earth fluctuates unpredictably as a result of changes in the Earth's core and atmosphere, and in the long term is slowing down due to friction caused by the ocean tides. This means that the length of the mean solar day is not consistent enough for use in precise timekeeping. The name Greenwich Mean Time was officially changed to Universal Time (UT) back in 1928, although the previous name has remained in widespread use. As understanding of the Earth's behaviour has improved, more precisely defined versions of UT have been introduced, and the nearest modern equivalent to GMT is a measure of time based on the Earth's rotation called UT1.

Time scales based on atomic clocks provide a much more precise method of timekeeping than the Earth's rotation, but at the moment the Earth's rotation is a little slower than atomic time, and a 'UT1 year' is around 300 milliseconds longer than a 'clock year'. The time scale that forms the basis of all precise timekeeping worldwide is called Coordinated Universal Time or UTC, and is a compromise between solar time (UT1) and atomic clock time. In order for the time kept by atomic clocks to keep in step with UT1, leap second adjustments are made to UTC when required to keep the difference between UT1 and UTC below 0.9 s, at a rate of one every few years. Leap seconds can be either added (positive) or removed (negative) as the unpredictable changes in the Earth's rotation can be greater than the long-term slowing down, although all leap seconds to date have been positive.

What would happen to the time system if leap seconds were removed?

All time signals are already linked to UTC and all countries base their civil time on UTC, with time zone adjustments. Losing leap seconds would mean that clock time and solar time would no longer be linked, and would slowly drift apart by a second every few years, adding up to around a minute over 100 years or perhaps 30 minutes over 1,000 years. UTC would no longer provide an approximation to UT1, and the UK's laws would have to be changed to refer to UTC explicitly, rather than GMT. The link between the Earth's rotation and the time of day would be lost, and the most likely outcome is that many countries would eventually (after several hundred years) decide to change their time zones to correct the accumulated difference between the solar day and clock time.

What are the arguments for removing leap seconds?

The leap second procedure was introduced in 1972, at a time when no one could have predicted our current need for high-precision timekeeping. The move away from celestial navigation has greatly reduced the need for civil timekeeping to be closely linked to the Earth's rotation. With civil time already offset by an hour or more from the local mean solar time in many places, due to time zones and, in some countries, 1-hour daylight saving time changes twice a year, it is argued that most people will not notice a small, very slow drift of civil time away from solar time if leap seconds are abandoned.

Complex timekeeping systems and software have to have leap seconds added to them manually. As the Earth's rotation is irregular, leap seconds do not follow a simple pattern (unlike leap years) so each leap second has to be programmed in separately after it is announced, with only six months' notice. This can be a difficult and costly process, and it means that human error could affect timekeeping systems (although there have not been any reports of this as yet).

Implementing a leap second can also cause problems for equipment that has to handle them correctly and reliably, for example in the time-stamping of financial transactions or in air traffic control systems. The difficulties are most severe in regions where leap seconds can occur during the working day, such as East Asia or the west coast of the USA.

What are the arguments for keeping leap seconds?

Losing leap seconds would mean losing the link between the Sun and timekeeping - however small the drift is - and eventually, after hundreds of years, the drift would be substantial enough to be noticeable. Astronomers and earth scientists making measurements based on the Earth's rotation would have to compensate for the loss of leap seconds in their systems, which would also be a manual and difficult process. There is also concern that ending leap seconds would have the effect of redefining the day to be 86,400 UTC seconds, and the gradual divergence between the UTC day and the mean solar day might have many social and legal consequences that have not yet been identified.

Some countries, including the UK, have argued that the leap seconds procedure has been in use for over 40 years and continues to work adequately well, and that the difficulties reportedly caused by leap seconds should be dealt with by technical solutions rather than by changing the fundamental basis of civil timekeeping. Systems such as GPS are already in place to provide precise, atomic clock-based timekeeping standards without the implementation of leap seconds. Those whose work needs a consistent time scale can use these systems without disrupting the current international time system.

Are there any other options?

It has been suggested that leap seconds could be replaced with larger steps in UTC occurring less frequently (for example, inserting a leap minute every few decades). However, a 2004 proposal from the USA to replace leap seconds with leap hours was rejected for a number of reasons, including concerns that a leap hour would create far more severe problems than a leap second. A change to any steps in UTC other than one-second steps is likely to be technically too difficult, too risky and too costly. Unlike summer time systems, where the underlying time scale (UTC) remains the same but an hour is added to or taken away from the offset, a leap hour (or leap minute) would change this reference time scale entirely. Any software or clocks (including atomic clocks) dependent on the time scale would have to be upgraded to deal with this change, and coordinating a noticeable worldwide time shift would be challenging for developers and disruptive to users.

Other proposals that would reduce the impact of leap seconds have included announcing them much further in advance, perhaps as long as 10 years before their insertion date, and scheduling their insertion for the end of a Saturday rather than a calendar month to avoid normal working hours. However, none of these ideas are being considered formally by the ITU.

An alternative method that some countries favour is to encourage the use of International Atomic Time (TAI; the leap second-free time scale that UTC is derived from) alongside UTC for applications that have difficulty implementing leap seconds. The difference between TAI and UTC is always an integer number of seconds, changing by 1 when a leap second is inserted into UTC, and this difference could be disseminated by some time and frequency transmissions and the internet.

Alternatively, if leap seconds are ended, civil time could remain aligned with mean solar time by countries changing their time zones by one hour whenever the difference between UTC and UT1 becomes sufficiently large (perhaps after several hundred years, when the offset has reached 30 minutes). In those countries that implement daylight saving time, the adjustment could simply involve not putting the clocks forward for summer time in the chosen year.

Where does the UK stand?

The British government opposes the proposal to end leap seconds in UTC. It has reviewed the evidence presented to the ITU of difficulties caused by leap seconds, and concluded that the reported problems are relatively minor and in many cases could be reduced or eliminated by improved procedures for handling leap seconds. It was also concerned that the consequences of redefining the day as 86400 UTC seconds have not been considered fully, and that breaking the link between civil timekeeping and the Earth's rotation would be unpopular and could undermine public trust in science. Consultations with Government agencies and professional bodies that have an interest in precise timekeeping found no support for the proposal but strong opposition from some quarters.

What will happen next?

The proposal to end leap seconds in UTC will be discussed and voted on at a meeting of the ITU-R World Radiocommunication Conference in November 2015. If the proposal is accepted, there will be no further leap seconds in UTC after an agreed date (possibly 2020) and UTC will then gradually drift away from UT1. If the proposal is rejected, leap seconds will continue, although there are likely to be changes to reduce the difficulties that they cause. For example, the maximum allowed offset between UTC and UT1 could be increased so that leap seconds can be announced several years in advance rather than six months.

Last Updated: 5 May 2015
Created: 27 Oct 2011

Registration

Please note that the information will not be divulged to third parties, or used without your permission

Login