Upsweep: the rising tone still echoing across the Pacific

Ever since NOAA gained access to the SOSUS hydroacoustic data in 1991, listening stations across the Pacific have been recording trains of pulses a few seconds long whose pitch sweeps upward. The source is strong enough to cross an entire basin, and is usually clearest in spring and autumn. It was named straight after its shape: Upsweep. More than three decades later its position has been roughly bracketed, but the specific source has still not been confirmed.

The widely circulated audio clip has been sped up about twenty times, so the electronic-siren sound is not what the signal is like underwater. The spectrogram is the record that matters: the frequency rises through each pulse, the pulses repeat in trains, and the signal level shifts with the season. The name does not imply that anything is "sending a message".

Locating a source from arrival times

An underwater sound reaches different hydrophones at different moments. Knowing where the stations are and how sound speed behaves in seawater, researchers estimate bearings and look for where they intersect. Upsweep has been placed near 54° south, 140° west, a remote part of the South Pacific close to suspected volcanic seismic activity.

The uncertainty is not small, because the ocean is not uniform. Temperature, salinity, depth and currents bend the propagation paths; a source that lasts a long time or has several emission points is harder to fix than a single sharp bang. "Near a volcano" is a geographical clue, not a photograph of the vent that is making the sound.

What does the seasonal rhythm tell us?

The intensity usually peaks in the southern spring and autumn. The source may genuinely change with the season, or the acoustic environment may change so that the same source simply sounds louder. Noise levels from storms, ice and shipping also vary, which affects how detectable the signal is.

NOAA records the source level as declining since 1991, although the autonomous equatorial hydrophone arrays still pick it up. A biological source would usually suggest a migration cycle; a geophysical one points to stress, hydrothermal activity or propagation conditions. The spectral shape and the location make a volcanic source the favoured explanation, yet no simultaneous measurement has been made close enough to lock down the mechanism.

The geophysical candidates

Fluid moving through fissures, oscillations in a magma chamber, hydrothermal venting or movement along a ridge can all produce structured sound. To choose between them you need to compare the fundamental frequency, the harmonics, the length of the train and the relationship with earthquakes. A good model has to account for the upward sweep, the seasonality and the declining trend all at once.

Many submarine volcanic phenomena have never been observed directly at close range. Instruments have to survive pressure, corrosion and strong currents, and the target area lies far from research shipping lanes. The cost of laying a dense array around a source that is only roughly bracketed is why this mystery persists - not for want of ideas, but because the decisive measurement is hard to make.

Upsweep is not Bloop

Bloop was a set of broadband sounds recorded in 1997 and matched to icequakes; Upsweep is a narrow, repeating, seasonal train that has been heard since at least 1991. Lumping every ocean sound together into one "call" destroys the fingerprints that matter. It is precisely the differences between their spectrograms that let NOAA classify sources as ice, biology, shipping or geology.

Upsweep still lacks a detailed identity, but it is not a total blank: the data points to one region, one type of sound and one family of plausible mechanisms. The mystery worth keeping is which geophysical process is playing that rising note beneath a stretch of ocean almost nobody ever visits.

From a story to a record that can be checked

A strong mystery does not require every detail to be unknown. The first task is to separate physical evidence, contemporary documents and measurements from illustrations, late recollections and literary retellings. A signal can be genuinely recorded without anyone knowing who sent it; a person can have really existed while the famous nickname was invented by a later generation.

Each kind of source answers a different kind of question. A receiver gives you frequency and timing, not intent. An inscription tells you a unit was stationed somewhere, not the fate of an entire army. A chronicle tells you which version was passed down, not that its author witnessed anything. Reading each source within its limits is what brings the unknown part into focus.

A gap in the record is not a blank slate

When the record is thin, a great many stories can be made to fit. But they are not equal: a hypothesis still has to match the last firm date, the geography, the technology available, the motives involved and whatever should have survived. A conclusion that needs fewer assumptions and explains more of the evidence deserves priority, even when absolute proof is out of reach.

"No evidence" also has to be given a scope. Nothing found in the archives that have been searched is not the same as never having existed; nothing detected again within one week is not the same as a source that never repeats. Stating that scope keeps you clear of both extremes: treating silence as proof of a cover-up, or treating a thin record as grounds for dismissing every account.

A timeline is the best defence against legend

Ordering the sources by date shows which details were there from the beginning and which were added decades later. A velvet mask turns to iron, a prime suspect turns into a stage monster, a technical signal turns into an apocalyptic message. A late detail is not automatically false, but it does have to explain how it came to know something the earliest sources never recorded.

It is also worth separating the date of the event, the date someone realised what it meant, and the date it was published. Data can sit unread for years before it is understood; an artefact can be excavated and then treated by the standards of its day. Those delays decide how much information modern techniques can still recover.

Independent confirmation comes in several forms

Two instruments running the same software can share the same fault; two chronicles can be copying one storyteller. Independence is strongest when different lines of evidence converge: a receiver far away, satellite imagery, an artefact with secure stratigraphy, a letter with an unbroken chain of custody. The fewer single links a conclusion rests on, the better it holds.

Reproduction matters just as much. A microwave oven that recreates perytons under stated conditions is causal evidence; a volcanic model that merely "could produce sound" is not enough for Upsweep unless it also reproduces the upward sweep and the seasonal rhythm. History rarely allows a rerun, so it has to cross-check texts, archaeology and probability instead.

Do not let an illustration become false evidence

There are no photographs of medieval events, and no direct images of many cosmic sources. The artwork in an article helps you picture the setting; it does not assert that a face, a gesture or a composition really looked that way. A caption should describe what the picture is explaining rather than present it as a scene from the site.

Images that travel beyond the article usually lose their caption. That is why illustrations should avoid invented text, over-confident symbols and sensational details with no source behind them. Atmosphere can come from light, empty space and traces; it does not need "evidence" planted in it that history never supplied.

A checklist for the reader

Ask what the earliest source is; whether its author was present; whether the raw data survives; whether an independent group confirmed it; when the famous detail first appears; what the hypothesis predicts; and what evidence would show it to be wrong. If nothing could refute it, it is a flexible story rather than an answer.

Finally, keep "cannot be ruled out" apart from "likely". Countless scenarios cannot be excluded absolutely, because the data is gone, yet history and science can still rank them. Talking in probabilities does not make an article duller; it tells the reader exactly where the solid ground ends.

The margin of error changes with the type of evidence

A recording can pin down frequency and timing quite precisely, yet still fail to identify the source if the sensor network is too sparse. A document can name a person and a place, but the author's intent, the intervening copies and the language of the period all introduce drift. Physical artefacts, which feel like the most objective evidence of all, can still be separated from their stratigraphy, over-restored, or dated from an object that happened to lie beside them. So it is a mistake to sweep every question mark into the single word "mystery". Each case needs its own map of uncertainty: what was measured directly, what was inferred, what a witness remembered, and what was only retold generations later.

When two pieces of evidence contradict each other, the sensible response is not to reach for the more dramatic one. You have to weigh resolution, the gap in time, the chain of custody, and whether the person recording it had an agenda of their own. A receiver in the right place can be worth more than ten people listening by ear; conversely, a clean stream of data with no context is just as easy to over-read. Being open about the level of uncertainty is what lets an account survive the next discovery. It also separates caution from evasion: a good researcher has to be able to say why one possibility is preferred over another, even when nothing can be settled for good.

A name can shape the whole story

Many famous phenomena are known through a catchy nickname. The nickname helps the public remember, but it also smuggles in an interpretation: "the trumpet", "the ghost", "the vanished legion" or "the man in the iron mask" all invite the mind to fill in pictures of its own. After a few hundred retellings, the label is easily mistaken for the original description. The remedy is to go back to the earliest wording, compare the translations and say plainly which name is modern. A single small difference - iron or velvet, vanished or redeployed, a one-off sound or a repeating train - can change the entire set of hypotheses.

The media adds a selection effect of its own: the coincidence is kept, while hundreds of ordinary occasions disappear from memory. Vivid details outlive dry numbers, so the later version tends to sound more complete and more frightening than the original record. That does not mean it is all imagination. It means the phenomenon and the history of the story about the phenomenon both need studying. Once you know where a detail first appeared, you can tell whether it is an independent fact, a new way of phrasing an old one, or decoration added to help the story travel.

What could move the case forward?

For acoustic or radio signals, progress usually comes from synchronised data at several stations, calibrated instruments and an environmental record alongside. Differences in arrival time give you a bearing; the spectrum and the repetition pattern help separate machinery, atmosphere, geology and astronomy. Raw data has to be archived so that another group can rerun the analysis instead of squinting at a processed plot. A single new detection need not solve an old mystery, but it can turn the vague question "what is it?" into specific tests of location, energy, periodicity and mechanism.

In history, progress can be slower: an archive is re-catalogued, private letters surface, organic material is dated, or isotopes reveal where somebody once lived. The technology is only useful when the question and the provenance of the sample are both clear. Testing an object with a broken custody history can produce a very precise number for the wrong subject. Conversely, a small administrative note placed correctly on the timeline can sometimes eliminate an entire major hypothesis. The value lies in connecting a new measurement to an old record, not in how modern the instrument looks.

Why is an answer sometimes still incomplete?

Some cases have a strong answer that still cannot be reconstructed minute by minute. We can identify a class of source, a social cause or a high-probability identity while the individual motives and the detailed sequence of actions are lost. Demanding that an answer cover every detail sets an impossible standard; but accepting a broad label without testing the exceptions is too easy. The balance is to grade the conclusion in layers: what is close to certain, what is well supported, what is still contested, and what there is simply no data to judge.

That unfinished quality is the honest part of research. A case can be closed on mechanism and still open on people, or the other way round. Once the main source is identified, the secondary questions do not become pointless: why did the community react as it did, why did the authorities of the day stay silent, and why does another version keep spreading? A mystery is therefore not only a hunt for a single answer. It is also a way of watching the limits of memory, of instruments, and of the power to keep records - the three things that decide which parts of the world survive for the people who come after.

The strangeness worth keeping

A natural explanation can be stranger than the legend: a dead star used as a clock, an oven in a staff kitchen imitating the cosmos, or a signal that crosses the entire Pacific. In history, what haunts is usually a deliberate silence - a name erased, an order never written down, people without power leaving none of their own words behind.

The best mysteries have clear edges. You know exactly what was recorded, what is merely plausible, and which measurement is still missing. The chill then comes not from exaggeration but from standing right at the edge of the record and seeing how much of the past, or of the universe, has moved out of reach.

References

• NOAA PMEL - Upsweep
• NOAA PMEL - ocean acoustics programme
• NOAA/NPS - ocean noise reference station network