Why Gravity, and Why Near the Bottom

Gravity tells you about density, and density tells you about what’s under the seafloor: mineral deposits, hydrothermal fluids, tectonic structure, or even human-made things like shipwrecks and pipelines. Satellites and ships can measure gravity over the ocean, but from that far away you only resolve features on the order of 10–20 km. Getting close to the seafloor is how you see the fine-scale structure.

The problem is that getting close has historically meant trading away something important:

• Stationary measurements are precise but sparse. Systems deployed from Alvin or ROVs to seafloor benchmarks get excellent repeatability, but only at the handful of points where you can stop and wait.
• Crewed and towed vehicles move, but they’re noisy. Towed systems inherit the motion of the ship above them, have to stay high off the bottom for safety, and tie up the ship for the whole survey.
• Prior AUV work didn’t go deep enough. The closest comparable system, aboard Japan’s AUV Urashima, worked well, but operated at around 1,550 meters and was rated to 4,200. An earlier attempt with Sentry ran into thermal instability that was never validated as fixed at sea.

We wanted continuous, dense, near-bottom gravity at near full-ocean depth, from a vehicle that could run for more than a day without a ship babysitting it.

The System

The gravimeter is the DGS AUV-1, which Daniel Aliod, Nigel Brady, and Sam Bailey at Dynamic Gravity Systems designed and built from their AT1M shipboard gravimeter line. It’s a self-leveling, gimbaled gravity meter squeezed into a roughly 9-inch titanium housing rated to 6,000 meters, bolted into an open bay on Sentry’s underside.

Sentry turns out to be a great platform for this. Its unusual shape keeps pitch and roll very small: on our first dive, pitch never exceeded 4.25 degrees, and 99% of readings stayed within 2.71 degrees. Other AUVs can see 20 to 40 degrees during normal operation, which is a problem for a rotation-sensitive instrument.

Getting the gravimeter onto Sentry was largely Justin Fujii’s work. Besides serving as our Expedition Leader, Justin built the housing and the integration with the vehicle, including a secondary jacket plumbed into Sentry’s cooling system. While Sentry sits on deck, chilled seawater runs over the housing, which keeps the sensor from overheating and softens the thermal shock at launch.

I handled the software side of the deployments. Sentry runs an open-source software stack, which I extended so we could control the gravimeter over the acoustic link: the gimbal stays clamped until Sentry reaches operating depth, and operators can check the sensor’s status and reboot it remotely mid-dive. Sentry’s INS and pressure sensor record the acceleration and depth data needed to correct the gravity signal after the fact.

Six Dives

We ran six dives (Sentry Dives 791–796) targeting the Jurassic Quiet Zone in the western Pacific, some of the oldest seafloor on Earth. The magnetic signal there is weak enough that near-bottom surveys are really the only way to get useful data. Altogether, that’s 186.8 hours in the water with 162.5 hours of surveying, dives up to 37.9 hours long, and a slow 0.7 m/s survey speed to stretch endurance. Each dive also carried three magnetometers, so gravity and magnetics share identical tracks and navigation.

The dives themselves were a pretty honest picture of fieldwork:

• Dive 791: no data, on purpose. We left the gimbal locked to measure Sentry’s in-situ pitch, then shimmed the gravimeter to compensate.
• Dive 792: no data, not on purpose. The motion of launching Sentry off the crane tripped an internal safety lock. Afterward, I changed the software so the clamps stay engaged through launch and descent, and added the remote status checks and resets over the acoustic link.
• Dives 793–795: full data. 793 needed no intervention. On 794 and 795, the acoustic link showed the sensor needed a reset, which we did during descent before any survey data was affected. This is exactly the fix from 792 paying off.
• Dive 796: full data, short dive. The dive ended early for unrelated reasons, but gravity data was unaffected.

Making Sense of the Data

A gravimeter on a moving vehicle measures gravity plus the vehicle’s own acceleration, and it can’t tell the two apart. So most of the work is subtracting Sentry’s motion out using its navigation data. Jyun-Nai Wu, Hannah Mark, Allisa Dalpe, and Zac Berkowitz worked through the processing and analysis, and a few things made it harder than expected:

• Clocks drift. The gravimeter and navigation clocks weren’t synchronized, and the cold at depth made the sensor’s clock run about 0.2 s/day slow. We recovered both the fixed delay and the drift by cross-correlating the sensor output against the navigation-derived corrections.
• Temperature matters a lot. The heater couldn’t quite keep up with cold deep water, so the sensor dropped about 5 °C on descent. We corrected for that using calibrated temperature and internal-pressure models.
• Depth-derived acceleration is noisy. Vertical acceleration comes from double-differentiating depth, which amplifies high-frequency noise. A Kalman smoother handled this much better than a simple Gaussian filter at short filter lengths.

After all that, we get noise levels of 0.232 mGal at 21-meter resolution and 0.072 mGal at 84-meter resolution. For comparison with satellites, that’s resolving features at better than 100 meters rather than 10+ kilometers. We also compared Dive 793 against surface gravimeter data, satellite-derived gravity, and gravity forward-modeled from bathymetry, and the AUV signal reproduces the same long-wavelength trends and major anomalies. That’s a good sign the processing chain works, though since we didn’t fly repeat or crossing lines at survey altitude, it isn’t yet an absolute accuracy number.

Why It Matters

This is, as far as we know, the first AUV-mounted gravimeter capable of routine, dense, near-bottom gravity mapping of the deep ocean floor. Because it lives on Sentry, which is a national facility vehicle, the goal isn’t a one-off demonstration: it’s a capability that the broader scientific community can request, just like multibeam or magnetics.

What’s Next

None of the problems we hit were flaws in the sensor itself; they’re engineering and operations issues we now know how to address:

• Better thermal management, so the sensor holds temperature on descent and the clock stops drifting.
• Resolving the remaining mechanical clamping issues.
• Repeat and crossing survey lines, to put a real absolute-accuracy number on the system.
• Eventually, routine facility operation of near-bottom gravimetry aboard Sentry.