Industries / Technology / Telecommunications / Fiber optics
Sensing with fiber
Sensing with fiber uses the fiber itself as the measuring instrument: changes in the light travelling through it reveal rotation, temperature, strain or vibration along its length.
7
Milestones achieved, from 1976 to 2023.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1976
A fiber coil senses rotation
Victor Vali and Richard Shorthill of the University of Utah Research Institute sent light both ways around a coil of fiber; when the coil turns, the two beams fall slightly out of step, the Sagnac effect, so the coil works as a gyroscope with no moving parts. Physics Today (October 1981) says they built "the first fiber-optic gyroscope," but noise was far above predictions, and not until 1979 did groups at Telefunken in Ulm and Thomson-CSF in Orsay get it low enough to detect the Earth's rotation. Navigation needs about a thousandth of that rate, 0.01 degree per hour. Exail says the two Pléiades satellites, launched in 2011, were the first to fly its Astrix 200 fiber gyroscope, built with Airbus Defence and Space.
Lessonthe noise came from light scattered back inside the fiber and at the interfaces between parts, and years of work on it came before the gyroscope was fit to navigate with.
Credit
Vali and Shorthill; Telefunken; Thomson-CSF; Exail with Airbus
1978
Gratings written by light
K. O. Hill, Y. Fujii, D. C. Johnson and B. S. Kawasaki at the Communications Research Centre Canada, in Ottawa, found that laser light left a permanent pattern in a fiber's core. The pattern, a fiber Bragg grating, is a built-in filter that reflects one narrow color. In 1989 G. Meltz, W. W. Morey and W. H. Glenn of United Technologies Research Center wrote gratings from the side, shining an ultraviolet interference pattern through the fiber's outer glass. Gratings are now used in communications and in sensor systems.
Lessonthe effect waited 11 years for a practical way to make gratings.
Credit
Hill and colleagues (IEEE Milestone, dedicated 11 August 2023); Meltz, Morey and Glenn
1985
Temperature along a whole fiber
In distributed temperature sensing (DTS), a laser pulse goes down the fiber and a faint echo comes back at shifted colors, called Raman scattering; one of those colors grows stronger as the glass warms, and the echo's travel time gives the position. J. P. Dakin and D. J. Pratt of Plessey, with G. W. Bibby and J. N. Ross of the Central Electricity Generating Board, reported temperature profiles using a semiconductor laser and detector in June 1985. In November 1985 A. H. Hartog and A. P. Leach of York VSOP and M. P. Gold of the University of Southampton reported 1 K resolution over 1 km of fiber at 7.5 m spacing. DTS is now used to rate power cables, to watch oil and gas wells and to detect fires in tunnels.
Credit
Dakin, Pratt, Bibby and Ross; Hartog, Leach and Gold
1990
Strain mapped with Brillouin light
Brillouin scattering is light bouncing off sound waves in the glass, and its color shifts when the fiber is stretched. Tsuneo Horiguchi, Toshio Kurashima and Mitsuhiro Tateda at NTT sent pump and probe light in from opposite ends of a fiber, a method called BOTDA, and mapped strain to better than 2 parts in 100,000 with 100 m resolution. Their January 1990 conference paper measured fibers before and after they were installed in a field conduit, because stress weakens telecom fiber.
Lessonthe method began as a check on telecom cables themselves.
Credit
NTT Transmission Systems Laboratories
2017
Spare telecom fiber records earthquakes
Distributed acoustic sensing (DAS) reads the faint echo from every few meters of a fiber, which turns one cable into thousands of vibration sensors. A Berkeley Lab team led by Jonathan Ajo-Franklin, with Nate Lindsey of UC Berkeley and Eileen Martin of Stanford, used it on arrays in California and Alaska, some on "dark fiber," unused telecom cable, and reported in December 2017 that DAS was about as sensitive to earthquakes as standard seismometers, though noisier.
Credit
Lindsey, Martin, Dreger, Freifeld, Cole, James, Biondi and Ajo-Franklin
2018
Undersea data cables hear earthquakes
Giuseppe Marra's team, from the UK, Italy and Malta, used ultrastable lasers on existing land and undersea links 75 to 535 km long and detected earthquakes 25 to 18,500 km away, without interrupting service or changing the cables (Science, 2018). In February 2021 Zhongwen Zhan of Caltech and Google engineers reported about 20 moderate to large earthquakes, including a magnitude 7.7 off Jamaica, and ocean swells, from the polarization of ordinary data traffic on Google's Curie cable from Los Angeles to Valparaíso, read from the cable's own transceivers with no added equipment.
Credit
Marra and co-authors; Zhan, Caltech and Google
2023
DAS data in one place
PubDAS, described by Zack Spica and 14 co-authors in Seismological Research Letters in March 2023, opened with eight DAS datasets, up to about 90 TB, from cities, mines and the sea floor; the authors call it "the first large-scale open-source repository" of DAS data from multiple experiments. In July 2024 DASCore, an open Python library that reads many DAS file types, was published in Seismica.
Lessonshared data alone did not make methods comparable; in 2026 the authors of SeisBench DAS still found "no standard for benchmark datasets.".
Credit
Spica and co-authors; Chambers and co-authors (DASCore)
Watch for
Sensing across an ocean
Undersea cables have amplifiers every few tens of kilometers, and ordinary DAS cannot see past the first one. A preprint led by Nokia Bell Labs reports continuous DAS over a 4,400 km California–Hawaii cable with about 100 repeaters, at 88,000 points 50 m apart at normal telecom power, and a recording of the magnitude 4.9 Susanville, California, earthquake of 30 November 2025, about 1,250 km from the landing station. A peer-reviewed paper by University of Washington and Nokia authors, on Oregon's research cables, reports usable data past three repeaters, more than 300 km, and at least 12 earthquakes recorded.
Achieved whenthe 4,400 km result passes peer review, or an outside group recovers catalogued earthquakes from data taken beyond the first repeater on a cable longer than 1,000 km.
Claimed by
Mikael Mazur, Nicolas Fontaine and co-authors, March 2026
A cable feels a tsunami
A team led by Diane Rivet of Géoazur, Université Côte d'Azur, reports that a 450 km DAS array on telecom cables off Chile recorded the tsunami from the 2025 magnitude 8.8 Kamchatka earthquake, with waves about 2 cm high straining the cable measurably, and estimates that a warning could "potentially provide 3 to 15 minutes of lead time" before the waves reach the coast.
Achieved whena peer-reviewed paper checks the DAS wave heights against independent tide gauges or ocean buoys, or a warning service uses DAS in its forecasts.
Claimed by
Diane Rivet, Yuqing Xie, Anthony Sladen, Martijn van den Ende, Alister Trabattoni, Jean-Paul Ampuero and Sergio Barrientos, in a conference abstract updated June 2026
Open
01
A benchmark for DAS denoisers
A public benchmark of labeled earthquake windows and noise-only windows from several PubDAS datasets, in the SeisBench DAS format, with one scoring script that any denoising method can be run through.
The briefCredit · unclaimed02
Earthquakes counted past repeaters
An outside, span-by-span count of the earthquakes in the public November 2025 to January 2026 multi-span DAS data from Oregon's Ocean Observatories Initiative cables, with every catalogued earthquake above a stated size marked as detected or missed on each span.
The briefCredit · unclaimed
Open 01 · the brief
A benchmark for DAS denoisers
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public benchmark of labeled earthquake windows and noise-only windows from several PubDAS datasets, in the SeisBench DAS format, with one scoring script that any denoising method can be run through.
Why it is still open
The SeisBench DAS authors wrote in September 2026 that "there exists no standard for benchmark datasets" and that methods lack comparability; the only DAS dataset they integrated is MLSubDAS, and no denoising benchmark is included.
Built with what exists
PubDAS data; DASCore to read the files; SeisBench DAS and xdas for the format and for running earthquake pickers on DAS; an f-k filter as the baseline, which the SeisBench DAS authors call "highly effective at noise suppression"; a public earthquake catalog for labels.
First step
Download one PubDAS dataset, read it with DASCore, write 50 windows in the SeisBench DAS format, and publish the conversion script.
Done looks like
At least 3 PubDAS datasets from different settings, each with at least 100 earthquake windows and 100 noise windows, at least 3 denoisers scored on two fixed measures (signal-to-noise gain, and change in picker recall and timing error), one command to reproduce the table, and a release with a DOI or a merge into SeisBench.
Open 02 · the brief
Earthquakes counted past repeaters
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An outside, span-by-span count of the earthquakes in the public November 2025 to January 2026 multi-span DAS data from Oregon's Ocean Observatories Initiative cables, with every catalogued earthquake above a stated size marked as detected or missed on each span.
Why it is still open
The team's own paper reports at least 12 earthquakes recorded with clear P and S waves, and its abstract gives no count of the catalogued earthquakes that were missed or of detections span by span. The data repository explains how to read the files but lists no earthquake catalog, and no analysis by anyone outside the team was found.
Built with what exists
The Zenodo datasets 10.5281/zenodo.19930402 (Nokia multi-span) and 10.5281/zenodo.19930457 (a conventional DAS unit on the first span); the reading notebooks in uwfiberlab/OOI_DAS_2025; DASCore or xdas; SeisBench DAS pickers; a public regional earthquake catalog.
First step
Download one day of multi-span data around a known regional earthquake, run the repository's notebook, and plot strain against distance to look for arrivals past the first repeater.
Done looks like
Every catalogued earthquake above a stated magnitude and within a stated distance in the recording window, marked detected or missed on each span, with detection rates per span, a side-by-side comparison with the conventional DAS on the first span, and code and results under a DOI.