Scientists Report Unprecedented Changes to Earth's Rotation Speed

On June 29, 2022, Earth completed its daily rotation 1.59 milliseconds ahead of schedule — the shortest day ever recorded since atomic clocks began measuring planetary spin in the 1950s [1]. That fraction of a second, imperceptible to any human, set off a chain of consequences that now reaches from geophysics laboratories to Wall Street trading floors, from GPS satellite constellations to the electrical grids that power entire nations.

Three years later, the planet's rotation continues to confound the scientists who track it. And a new study published in March 2026 has added a striking finding: while Earth has recently been spinning faster due to changes deep in its molten core, climate change is simultaneously slowing the planet's rotation at a rate not seen in 3.6 million years [2]. These two forces are pulling in opposite directions, creating what researchers describe as an unprecedented situation for global timekeeping.

The Measurements: What the Atomic Clocks Show

Since the 1960s, atomic clocks — which measure time by counting oscillations of cesium atoms — have provided a fixed reference against which Earth's variable rotation can be measured with extraordinary precision. The standard day is defined as exactly 86,400 seconds. But Earth rarely hits that mark exactly.

Source: IERS / timeanddate.com

Data as of Jul 24, 2025CSV

The data shows a clear shift. In 2005 and 2012, the longest deviations from a standard day were positive — meaning Earth was rotating slower than the atomic standard, with days running about 1.05 to 1.26 milliseconds long [1]. But starting around 2020, the pattern reversed. July 19, 2020 saw a day 1.47 milliseconds short, and the June 29, 2022 record of -1.59 milliseconds confirmed an acceleration trend that continued into 2024 and 2025 [1][3].

To maintain alignment between atomic time (International Atomic Time, or TAI) and the rotation of the planet, the International Earth Rotation and Reference Systems Service (IERS) has periodically inserted "leap seconds" into Coordinated Universal Time (UTC) since 1972. A total of 27 leap seconds have been added over the past five decades — all of them positive, adding a second to compensate for a planet that was gradually slowing down [4].

Source: IERS Bulletin C

Data as of Jan 1, 2025CSV

The plateau visible after 2017 — no leap second has been added for eight years — reflects the recent acceleration. No positive leap second has been needed because Earth is no longer falling behind atomic time. It may, in fact, be pulling ahead.

Two Forces, One Planet

Understanding what is happening requires separating two distinct physical processes that operate on different timescales and pull Earth's rotation in opposite directions.

The core acceleration. Earth's liquid outer core — a 2,200-kilometer-thick layer of molten iron and nickel — generates the planet's magnetic field through convective flows. These flows also transfer angular momentum between the core and the mantle, and fluctuations in core dynamics can speed up or slow down the solid Earth above. Duncan Agnew, a geophysicist at the Scripps Institution of Oceanography at UC San Diego, identified this core effect as the primary driver of the recent acceleration in a landmark 2024 paper published in Nature [5]. The core's eddies and flows act in ways that remain difficult to predict, but they appear responsible for the shorter days observed since 2020 [5][6].

The climate brake. Simultaneously, the melting of ice sheets in Greenland and Antarctica is redistributing enormous volumes of mass from the poles toward the equator via rising sea levels. The physics is analogous to a figure skater extending their arms during a spin — moving mass away from the axis of rotation slows the spin [6][7]. A 2026 study by Mostafa Kiani Shahvandi of the University of Vienna and Benedikt Soja of ETH Zurich, published in the Journal of Geophysical Research: Solid Earth, quantified this effect: climate-driven factors are currently lengthening the day at a rate of approximately 1.33 milliseconds per century [2][8].

Source: Kiani Shahvandi & Soja (2026), Journal of Geophysical Research

Data as of Mar 14, 2026CSV

That rate may sound trivial, but the researchers found it is without precedent in the past 3.6 million years. Using fossilized shells of benthic foraminifera — single-celled marine organisms whose chemistry records ancient sea levels — the team reconstructed day-length variations across geological time. "Only one time — around 2 million years ago — the rate of change in length of day" approached current levels, Kiani Shahvandi said, and even that period showed a slightly slower rate than what has been measured between 2000 and 2020 [2][8].

By the end of the 21st century, the climate effect on day length is projected to reach 2.62 milliseconds per century — exceeding even the Moon's tidal braking force of about 2.4 milliseconds per century, which has been the dominant influence on Earth's rotation for billions of years [2]. "The current rapid rise in day length can thus be attributed primarily to human influences," Soja stated [8].

The Negative Leap Second Problem

The interaction of these forces creates a practical problem. Agnew's 2024 analysis concluded that without climate change, the core-driven acceleration would have required the first-ever negative leap second — the subtraction of a second from UTC — as early as 2026 [5]. The ice melt effect has acted as a counterweight, pushing that date back by roughly three years, to around 2029 [5][9].

A negative leap second would work by suppressing the final second of a chosen day: 23:59:58 would be followed directly by 00:00:00 of the next day, skipping 23:59:59 entirely [10]. This has never been done. And that is what concerns engineers.

The 27 positive leap seconds added since 1972 have been tested in practice — and they have repeatedly caused failures. The most documented case occurred on June 30, 2012, when a positive leap second triggered bugs in the Linux kernel's timekeeping code and in Java applications that interfaced with the Network Time Protocol (NTP) [11][12]. Reddit, Mozilla, LinkedIn, FourSquare, Yelp, and Gawker all experienced outages [12]. Australia's Qantas and Virgin Australia airlines were hit particularly hard: the Amadeus Altea reservation system, one of the world's largest, went offline for over an hour, forcing staff to check in passengers manually and delaying more than 50 flights [12][13].

A similar, though smaller, round of disruptions followed the 2015 leap second. The CME Group and the London Stock Exchange Group issued technical notices warning of potential impacts on electronic trading and clearing systems [14]. One economic analysis estimates that a single leap second event exposes the global economy to between tens of millions and approximately $100 million in direct disruptions, with tail risks concentrated in finance, aviation, and internet platforms [14].

A negative leap second carries additional risk precisely because it has never been implemented. During a 2022 testing exercise, one participant's hardware treated a negative leap second as a positive one — the exact opposite of the intended adjustment — because the timekeeping protocol it used had no provision for negative values [10]. The timekeeping and GNSS communities "need to check their software and hardware for possible breakdowns due to positive or negative leap seconds," researchers Matsakis and McCarthy warned [10].

Who Gets Hurt: GPS, Trading, and the Grid

The systems most exposed to timing disruptions share a common feature: they depend on synchronization at the microsecond level or finer.

GPS and navigation. The Global Positioning System calculates location by measuring the time it takes signals to travel from satellites to receivers. GPS time does not use leap seconds — it diverged from UTC by design in 1980 — but the interface between GPS time and UTC-dependent systems creates vulnerability. A 2016 GPS timing malfunction that offset signals by 13 microseconds disrupted communication networks and power systems that relied on GPS as a timing source [15].

Financial trading. High-frequency trading firms execute transactions in microseconds, and regulatory frameworks in the United States and Europe require nanosecond-precision timestamps on trades. The Precision Time Protocol (PTP) used by these firms provides sub-microsecond accuracy, but it depends on a stable UTC reference [15][16]. A leap second error — positive or negative — that propagates through trading infrastructure could create mismatched timestamps, rejected trades, or briefly desynchronized order books.

Power grids. Electrical grids use GPS-synchronized phasor measurement units to maintain the precise frequency matching required for alternating current systems. Microsecond-level accuracy allows fault detection and location within roughly 300 meters [15]. Timing drift beyond that threshold degrades grid operators' ability to detect and isolate faults, with cascading implications for grid stability.

Is "Unprecedented" the Right Word?

The ETH Zurich study's finding that the current rate of day-length change is unprecedented in 3.6 million years has generated both attention and skepticism. The claim requires important context.

Atomic clocks have existed for roughly 70 years. Before that, scientists relied on astronomical observations — recorded solar and lunar eclipses — to infer changes in Earth's rotation. A landmark 2016 study by F.R. Stephenson, L.V. Morrison, and C.Y. Hohenkerk compiled eclipse records from 720 BC to AD 2015 and lunar occultation data from 1600 to 2015, producing a roughly 2,700-year reconstruction of rotational variation [17]. That record shows significant fluctuations, but the resolution is measured in centuries, not days.

The Kiani Shahvandi and Soja study extended the record to 3.6 million years using benthic foraminifera proxies — an indirect measure of sea level that is then converted to inferred day-length change through a physics model [2]. This is a different kind of measurement than an atomic clock reading, and the uncertainty bands widen substantially as you go further back in time.

Source: OpenAlex

Data as of Jan 1, 2026CSV

The surge in academic publications on Earth's rotation — peaking at over 15,600 papers in 2023 — reflects genuine scientific interest, but also the difficulty of separating signal from noise in a field where the measured changes are on the order of a millisecond. Several researchers have cautioned against overstating the immediate practical significance. A 2024 analysis posted to the arXiv preprint server found that the anomalous acceleration observed in 2021–2024 had slowed, and argued that fears about an imminent negative leap second "apparently have no grounds based on latest observational data" [18]. The Earth's rotation has always been variable; the question is whether the current variability exceeds the envelope of natural fluctuation, or whether the alarm reflects the limits of our measurement history.

The honest answer is: the rate of climate-driven change is almost certainly unprecedented on a multi-million-year timescale, based on the best available proxy data. The short-term acceleration driven by core dynamics is harder to evaluate because core fluctuations are inherently unpredictable and the modern observational record is short. If atomic-clock-precision instruments had existed 10,000 years ago, we might find comparable short-term wobbles. We do not know.

What Happens Next

The international timekeeping community has already set a deadline for itself. In November 2022, the General Conference on Weights and Measures (CGPM) voted to phase out leap seconds by or before 2035 [19]. Under this resolution, the maximum permitted difference between UT1 (astronomical time) and UTC (atomic time) will be increased beyond one second, and the International Bureau of Weights and Measures (BIPM) is developing a plan, in consultation with the International Telecommunication Union (ITU), to ensure UTC continuity for at least a century [19].

The question is whether the negative leap second problem will arrive before the 2035 fix. Agnew's estimate of ~2029 for a possible negative leap second remains the most cited projection [5]. If Earth's core dynamics continue to accelerate the planet's spin faster than ice melt can slow it, the IERS would need to make the call — typically announced six months in advance via its Bulletin C — to subtract a second from UTC.

The projected timeline depends heavily on the balance between two poorly constrained variables: the behavior of flows in the outer core (which geophysicists cannot predict more than a few years ahead) and the rate of ice sheet mass loss (which climate models project will accelerate but with wide uncertainty bands). If the core acceleration moderates — as some recent data suggests it may be doing [18] — the 2035 abolition of leap seconds could arrive in time to make the negative leap second moot.

If it does not, the world's timekeeping systems will face a test they have never been designed for: subtracting a second that many of them have no protocol to handle. The stakes are measured in milliseconds. The consequences are not.