Can offshore AI send the underwater soundscape ashore in under a minute? One three-week trial did
A solar-buoy system streamed 96 kHz underwater audio from 10 kilometres offshore and ran an anomaly detector on the device for three weeks. It produced 350 hours of recordings without a reported shutdown—but the chapter gives no detection accuracy or false-alarm rate.
The short version
- What happened
- A single MARLIN unit operated on a solar buoy for three weeks off Sesimbra, Portugal, in September 2025, producing more than 350 hours of audio across 4,000 files without a reported data failure or shutdown.
- Who it affects
- Faster access to underwater sound could help conservation teams and offshore operators inspect possible wildlife or noise events sooner, but a human decision-maker still needs a validated detector and a clear response protocol.
- What remains unproven
- One unit at one Portuguese site for three weeks in September 2025
Keep in mind
One unit at one Portuguese site for three weeks in September 2025
One algorithm, with architecture outside the chapter
No labelled detection denominator, accuracy, recall or false-alarm rate
No comparator system, cost analysis or full power-budget result
Cellular performance may not transfer to remote offshore conditions
Industry-affiliated authors helped develop the system
Scholarly book chapter; journal peer-review status is not described
Source analysis · 1 linked source. Primary analysed source date: 28 October 2026.
Keep following the evidence
Join The Impact Brief founding list for selected AI reports, their effect on people and the evidence limits. Read a sample first.
Read the full report · 4 min The evidence, methods, implications and limitations
The buoy kept recording while an AI model listened at the edge
Marine researchers and offshore operators often retrieve underwater recordings long after an event has happened. MARLIN is designed to narrow that delay: a hydrophone records the soundscape, a low-power computer can run detection models beside the water, and compressed audio plus detections travel to a server for human review. The authors define near real time as data becoming available within 60 seconds of recording.
The reported field test used one unit on a solar-powered buoy about 10 kilometres from the nearest cellular tower off Sesimbra, Portugal. Over three weeks in September 2025, it recorded at 96 kHz, produced more than 350 hours of audio in 4,000 five-minute files and had no reported data failure or shutdown. A typical compressed file was about 35 MB and uploaded in roughly 52 seconds on the measured 5.39 Mbit/s uplink.[1]
Continuous data transfer was demonstrated; detection quality was not
A deep-learning anomaly detector, trained on 100 audio files from the first days of the deployment, ran on the edge device and sent possible sonar signals to a classification pipeline. The system also surfaced mooring noise, dolphin whistles, humpback calls, echosounders and radio signals. Researchers could change model parameters remotely while losing only seconds of audio during an update.
That is evidence that one prototype can keep a high-rate acoustic pipeline running in one near-shore setting. It is not evidence that its AI detections are accurate enough for an exclusion zone, wildlife intervention or security decision. The model architecture is outside the chapter, no labelled test denominator or error rate is reported, and buoy noise affected the recordings. Three authors work for RS Aqua Ltd, the system developer. Independent trials across seasons, networks and offshore-energy sites would change the assessment.[1]
The distinction that matters
What the field trial actually tested
The system joined three steps in one offshore pipeline; only continuity and transfer performance received concrete field measurements.
Record
A hydrophone captured five-minute 96 kHz audio files
More than 350 hours across 4,000 files
Analyse
A site-trained anomaly detector ran on the edge device
Signals were flagged, but accuracy was not reported
Transmit
Compressed files crossed a cellular link to shore
About 35 MB transferred in roughly 52 seconds
Continue the story
Read next
Continue exploring edge ai: compare the finding and its evidence limits.
Can embedded MRAM make edge AI start faster and use less energy? A Japanese chip reports one measured prototype
A Tohoku University–AISIN 16 nm chip stored its operating system and Tiny YOLOv3 weights in 32 MB of on-chip STT-MRAM, reporting a 65.7 ms boot and 50.7 mJ per frame. The conference result is a hardware demonstration, not a general efficiency benchmark or deployed-product result.
Read the next reportExplore the full source and evidence record
What this means for people
- Faster access to underwater sound could help conservation teams and offshore operators inspect possible wildlife or noise events sooner, but a human decision-maker still needs a validated detector and a clear response protocol.
Global context
Passive acoustic monitoring supports marine conservation and offshore-wind mitigation in several countries, but connectivity, protected species, noise rules and response procedures differ. A successful Portuguese cellular deployment cannot be assumed to work at a remote wind farm, in polar waters or under a different regulatory threshold.
What the evidence does not yet show
- One unit at one Portuguese site for three weeks in September 2025
- One algorithm, with architecture outside the chapter
- No labelled detection denominator, accuracy, recall or false-alarm rate
- No comparator system, cost analysis or full power-budget result
- Cellular performance may not transfer to remote offshore conditions
- Industry-affiliated authors helped develop the system
- Scholarly book chapter; journal peer-review status is not described
What to watch next
- Independent deployments across multiple offshore sites and seasons
- Labelled detection accuracy and false-alarm measurements
- Performance over satellite and long-range radio links
- Power consumption and maintenance burden over longer deployments
- Documented use in a real mitigation or conservation decision
Living evidence record
Impact record IAI-0J5O9OR
Evidence stage
Studied
Confidence
Supported
Reporting basis
Source analysis
Independent or research support
Not yet
Record status
Monitoring
Last checked
28 October 2026
Source trail
1 direct source across 1 source type.
People impact
Documented in this record.
Uncertainty
Limits and next checks are explicit.
Stages describe the evidence available—not whether a technology is good or bad. See the public method.
Single-source reporting disclosure
This record analyses one direct source. It can establish what Springer Nature / The Effects of Noise on Aquatic Life IV published or reported, but it is not independent corroboration of every performance claim or predicted outcome. The confidence label will change only when broader evidence is added.
Evidence trail
Sources used for this report
Links checked 28 October 2026
Research question, topics and regions
Whether an edge-computing marine acoustic system can continuously record, analyse and transmit broadband underwater audio in near real time
Topics: Marine monitoring · Edge AI · Offshore energy · Biodiversity · Environmental data · Field trials
Regions: United Kingdom · Portugal · Europe
This report is labelled source analysis. We summarise and analyse source material in our own words; company statements remain attributed claims until independently supported. Translated summaries preserve the meaning of the original source and link back to it. Read our editorial standards.
Reader commentary
What does this mean where you are?
Tell us what you have seen at work, in your community or in everyday life. A question is welcome too. No account is needed; we review notes before publishing them.
Explore commentary across the portal →Published reader notes
0No published reader notes yet. You can start the evidence-led discussion above.
Prefer a private correction or response? Contact the newsroom.