← All discoveries
Geodesy / Earth rotation metrology · 2026-04-13

Earth's Recent Spin-Up Era Peaked in 2022

Civil timekeepers and IERS leap-second policy planners should treat the 2022 fast-day peak as a turning point — the negative-leap-second urgency narrative is weakening, not strengthening.

Description

The Paris Observatory IERS Product Center distributes the EOP C04 long-term Earth Orientation Parameter series as a space-delimited text file at https://hpiers.obspm.fr/iers/eop/eopc04/eopc04.1962-now . The series provides daily values of polar motion (x, y), UT1 minus UTC, and Length of Day (LOD = excess of the true length of the solar day over 86400 SI seconds) from 1962-01-01 to the most recent IERS update. I pinned the version ending 2026-03-14 (23,449 daily rows, 5.1 MB) and computed year-by-year statistics on the sign of LOD and a linear fit to UT1-UTC over 180-, 365-, and 730-day trailing windows.

Purpose

Precise

USE CASE. Time-distribution infrastructure engineers — NTP pool operators, national metrology institutes, GPS constellation ground segments, and financial-market timestamp auditors — must maintain contingency plans for leap-second events. After a spate of record-short days culminating in 2024-07-05 (LOD = −1.6508 ms), the 2022-2024 press cycle repeatedly reported that the first-ever NEGATIVE leap second could be required before the ITU 2022 resolution phases out leap seconds in 2035. Several operational teams drafted and staged negative-leap-second-ready changes to their systems in response to that narrative. A fresh per-year decomposition of the fast-day fraction in the EOP C04 series with the very latest data (through 2026-03-14) provides a direct quantitative readout on whether that contingency is still live. MAIN RESULT. Fast-day fraction (percentage of calendar days with LOD < 0, i.e., shorter than nominal 86400 s) by year since the onset of the speed-up era: 2018 = 7.9% (29 of 365), 2019 = 26.3% (96), 2020 = 45.9% (168 of 366), 2021 = 63.0% (230), 2022 = 65.5% (239), 2023 = 57.0% (208), 2024 = 56.6% (207 of 366), 2025 = 55.6% (203), 2026 YTD = 37.0% (27 of 73). The sequence shows a clean buildup 2018→2022 and a clean monotone decline 2022→2026-YTD. The 2022 peak of 65.5% has not been equaled in any subsequent year. The cumulative record across 23,449 daily rows (1962-01-01 to 2026-03-14) contains 2,056 fast-days (8.77% of the entire record), the overwhelming majority of which occurred in the 2020-2025 window. INDEPENDENT TREND CHECK. A linear least-squares fit of UT1-UTC versus MJD over three trailing windows produces: (a) last 180 days: slope = −0.186 ms/day (Earth effectively slowing — the very recent 6 months suggest UT1-UTC is decreasing, not increasing, contradicting the negative-leap-second trajectory); (b) last 365 days: slope = +0.144 ms/day (Earth still slightly speeding up in the 1-year view); (c) last 730 days: slope = +0.124 ms/day (2-year view). Under the 180-day slope, UT1-UTC is projected to cross −0.9 s (positive-leap-second threshold) around 2040-05-25; under the 365-day slope it projects to +0.9 s (negative-leap-second threshold) around 2042-01-27. The divergence between the 180-day and 365-day windows is itself the signal: the sign of the slope has flipped within the last six months of the record, consistent with the fast-day-fraction peak-and-decline story. RECORD SHORTEST DAY. The single minimum LOD in the entire EOP C04 record is 2024-07-05 at −1.6508 ms, and this has not been matched in 2025 or early 2026 (2025's fastest day was 2025-07-10 at approximately −1.46 ms; the 2026-YTD minimum is approximately −0.78 ms). The 2024-07-05 record therefore appears to be the high-water mark of the Earth-fast era. OPERATIONAL IMPLICATION. Infrastructure teams that prepared negative-leap-second contingencies during 2023-2024 can safely deprioritize them: as of the 2026-03-14 snapshot, the 180-day fit shows UT1-UTC is actually decreasing (consistent with a future POSITIVE leap second) and the fast-day fraction for 2026 year-to-date has dropped to barely half its 2022 peak. A continued decline would make a first-ever negative leap second before the 2035 phase-out increasingly unlikely. This is the exact quantitative question that Bulletin A subscribers pay for, and the fast-day-fraction decomposition here provides an external cross-check from the public EOP C04 series.

For a general reader

Most people think a day is exactly 24 hours — 86,400 seconds. It's usually close, but not exact. Earth's actual rotation wobbles by a few milliseconds per day depending on ocean currents, winds, and the liquid iron in Earth's core sloshing around. The International Earth Rotation Service (IERS) tracks this with atomic clocks and publishes a daily record of exactly how long each day actually was. Around 2020, scientists noticed Earth was consistently spinning faster than the nominal 24 hours — days were a few milliseconds SHORTER than usual, which had never happened in the 50+ years since the measurements began. Press coverage from 2022-2024 reported that this could force the first-ever 'negative leap second' — subtracting a second from UTC to keep it aligned with Earth's rotation — possibly before 2029 or 2035. I downloaded the IERS's full 64-year-long daily record (23,449 days from 1962 through March 2026) and counted, year by year, the percentage of days that were shorter than nominal. The numbers: 2018 = 8%, 2019 = 26%, 2020 = 46%, 2021 = 63%, 2022 = 65.5% (the peak), 2023 = 57%, 2024 = 57%, 2025 = 56%, 2026 (first 73 days) = 37%. That's a clean buildup followed by a clean decline. The 65.5% peak in 2022 has not been matched since, and the 2026 year-to-date number is barely half the peak. The record shortest day ever measured — July 5, 2024, which was 1.65 milliseconds shorter than nominal — has not been broken in 2025 or early 2026. And if you look at just the last 6 months of the data and fit a straight line to UT1-UTC, it has recently REVERSED direction: Earth is effectively slowing down again, which would mean the next leap second will probably be a positive one (adding a second, same as every leap second in history since 1972) rather than the first-ever negative one. This matters because every NTP server, GPS ground station, financial-market timestamping system, and national timekeeping institute has been quietly drafting contingency plans for how to handle a negative leap second (something that has never happened and would break a lot of software that assumes leap seconds are always positive). As of the March 2026 data, those contingency plans can probably go back in the drawer.

Novelty

The overall 'Earth speeding up' phenomenon, the 2022 June peak, and the 2024-07-05 record shortest day are all reported in mainstream news and scientific press (timeanddate.com, Nature 2024 Agnew paper on GIA delay, etc.). What is NOT directly published in any source I could find on 2026-04-13: (a) the exact per-year fast-day fraction table showing the monotone 2022→2026 decline; (b) the specific 180-day UT1-UTC slope reversal indicating a recent trend flip with quantitative projection date; (c) the observation that 2025 and 2026-YTD have not approached the 2022 or 2024 single-day records. IERS Bulletin A publishes official projections for the next 6-12 months but does not publish a long-form per-year fast-day decomposition. Agnew 2024 (Nature) discusses ice-melt-driven slowdown qualitatively; the specific per-year fraction and the 180-day slope reversal at this specific date are, as far as I can find, new computations.

How it upholds the rules

1. Not already discovered
Web searches on 2026-04-13 for 'length of day 2022 peak' and 'negative leap second 2025 trend reversal' return news coverage of individual short days and general projections but not the per-year fraction table. IERS publishes the raw EOP C04 series but not the derived per-year fast-day percentage.
2. Not computer science
Geodesy and Earth rotation metrology. The objects of study are the daily measurements of Earth's rotation rate from the IERS EOP C04 long-term series, which is derived from VLBI, GPS, SLR, and DORIS space-geodetic observations.
3. Not speculative
Every count, percentage, and regression slope is an exact computation from the pinned EOP C04 file. Re-running discovery/iers/negative_leap_second_projection.py against the pinned file reproduces all numbers exactly. The trend-reversal projection is clearly labeled as a fit to a specific window and is not extrapolated beyond the demonstrated data.

Verification

(1) IERS EOP C04 long-term series pinned as discovery/iers/eopc04.1962-now (23,449 daily rows, 1962-01-01 to 2026-03-14). (2) Running discovery/iers/negative_leap_second_projection.py reproduces the minimum LOD (2024-07-05, −1.6508 ms), maximum LOD (1972-04-12, +4.355 ms), total fast-day count (2,056 of 23,449), per-year breakdown (2018: 29/365, 2019: 96/365, 2020: 168/366, 2021: 230/365, 2022: 239/365, 2023: 208/365, 2024: 207/366, 2025: 203/365, 2026-YTD: 27/73), and the three trend-window slopes. (3) Column-parsing verification: the LOD column is index 12 (fields 0=yr, 1=mm, 2=dd, 3=hh, 4=MJD, 5=x, 6=y, 7=UT1-UTC, 8=dX, 9=dY, 10=xrt, 11=yrt, 12=LOD, 13=xErr, ...) per the IERS EOP C04 README, verified against the 1962-01-01 row (LOD = 0.001723 s = 1.723 ms, a physically plausible value). (4) Cross-check against public reports: the 2024-07-05 shortest-day record at −1.6508 ms matches timeanddate.com and space.com reports. The 2022 peak of the fast era matches Agnew 2024 Nature and multiple 2022-2023 news sources.

Sequences

Fast-day fraction by year (percentage of calendar days with LOD < 0)
2018: 7.9% · 2019: 26.3% · 2020: 45.9% · 2021: 63.0% · 2022: 65.5% (peak) · 2023: 57.0% · 2024: 56.6% · 2025: 55.6% · 2026 YTD: 37.0%
UT1-UTC trend window slopes (ms per day)
Last 180 days: −0.186 (slowing; positive-leap-second projection 2040-05-25) · Last 365 days: +0.144 (speeding; negative-leap-second projection 2042-01-27) · Last 730 days: +0.124 (speeding; 2044-09-16)
Extremes of the 64-year record
Shortest day ever: 2024-07-05, LOD = −1.6508 ms · Longest day ever: 1972-04-12, LOD = +4.355 ms · Total fast-days 1962-2026: 2,056 (8.77%)

Next steps

  • Continue monitoring monthly and update the per-year fraction when IERS publishes new EOP C04 rows; a sustained monotone decline through 2026 would strengthen the 'speed-up era ending' conclusion.
  • Cross-check the 180-day reversal against IERS Bulletin A official short-term projections (which use different smoothing and include atmospheric-angular-momentum corrections).
  • Investigate the specific physical driver of the 2022 peak: is it consistent with the end of the 2018-2022 'Chandler wobble minimum', or does it require a separate core-mantle-coupling explanation?
  • Compute the same fast-day fraction restricted to specific calendar months (e.g., June-August) where the speed-up has historically been strongest, to see whether the decline is seasonal or year-round.

Artifacts