SAS or Multibeam for Wreck Surveys: When Resolution Earns Its Premium

What the Monitor survey actually demonstrates

Northrop Grumman, working with NOAA, has released new imagery of the USS Monitor captured with its µSAS – micro synthetic aperture sonar – sensor. The Civil War ironclad sank on 31 December 1862, was located in 1973, and two years later became the United States’ first National Marine Sanctuary. It sits in 73m of water. The September 2025 work, run under Northrop Grumman’s Technology for Conservation initiative, produced what the parties describe as the highest-resolution acoustic images of the wreck to date, plus 3D digital and physical models spanning the ship’s history. NOAA’s stated interest was practical: it had been looking for readily available technology able to assess the site in greater detail and to establish a baseline for ongoing monitoring.

Strip away the heritage story and what remains is a straightforward hydrographic procurement decision that survey managers face on obstruction, debris-field and archaeological jobs every year. You can image and model a seabed target with a conventional multibeam echosounder (MBES), or you can commit to a synthetic aperture sonar spread. One carries a materially higher cost and logistical footprint. The question is when the resolution difference is worth paying for, and when it is specification gold-plating.

Why this is a real cost-and-risk decision, not a technology beauty contest

The two techniques do not produce the same product, so comparing them on a single “resolution” figure is misleading. SAS coherently combines successive pings to synthesise a long array, which decouples along-track resolution from range. The practical consequence is roughly constant, centimetre-class imagery across the full swath, out to ranges where an MBES footprint would have grown to tens of centimetres or more. MBES resolution, by contrast, is set by beamwidth multiplied by slant range. A 0.5° beam at a 73m stand-off already subtends about 0.64m at nadir, and the footprint grows with altitude and off-nadir angle. To make an MBES rival SAS imagery you have to fly the sonar low on an ROV or AUV, which collapses swath width and area coverage rate.

That geometry is the whole commercial argument. Where a job needs cm-scale characterisation over a large area, SAS often wins on area coverage rate at that resolution, not just on picture quality. Where the deliverable is survey-grade depth over a modest footprint, a well-run MBES spread does it with less mobilisation risk and a simpler processing chain. Getting this wrong costs money in both directions: over-specifying SAS burns budget on a heritage-grade image nobody needed, while under-specifying MBES delivers a dataset that fails the client’s detection or accuracy requirement and forces a remobilisation.

What genuinely drives the choice

Detection requirement before resolution number

The first parameter is not resolution – it is the smallest object you are contractually required to detect and, separately, to classify. IHO S-44 Edition 6.1.0 frames this as feature detection: cubic features greater than 0.5m for Exclusive Order, greater than 1m for Special Order, and greater than 2m (or 10% of depth beyond 40m) for Order 1a. Detection and classification are different thresholds. You can detect a 1m cubic feature with MBES comfortably inside Special Order; deciding whether that feature is ordnance, a valve, a wire rope or geology needs imagery resolution well below the object size. That gap between “there is something here” and “here is what it is” is where SAS earns its keep.

Imagery versus bathymetry

SAS in its base form is a backscatter imaging technique. It gives you a photograph-like acoustic image, not, on its own, calibrated depth. Interferometric SAS adds vertical receiver separation to recover bathymetry, but the vertical accuracy budget of an interferometric SAS is a different engineering problem from its stunning along-track imagery. MBES remains the reference technique for the 3D geometry: georeferenced soundings with a well-characterised total vertical uncertainty. On the Monitor-type job, the honest architecture is both – SAS for structural detail and debris characterisation, MBES for the metric 3D model and the volumetric baseline NOAA needs for change detection.

Navigation and georeferencing

SAS is unforgiving on platform motion. The coherent aperture only forms if the along-track position of successive phase centres is known to a small fraction of a wavelength, which is why SAS platforms rely on micronavigation – typically displaced phase centre (DPCA) correlation between overlapping receiver elements, tightly coupled with an inertial navigation system aided by a Doppler velocity log. That gives excellent relative geometry within a survey line. It does not, by itself, give you absolute georeferencing. Tying a SAS mosaic to a client datum still depends on USBL or LBL aiding and on INS drift management, and a common disappointment is a beautiful image whose absolute position is weaker than the imagery resolution implies. If the deliverable is an obstruction chart position for a rig move or a cable route, absolute accuracy – not picture sharpness – is the acceptance criterion.

Geometry and grazing angle

SAS wants a low grazing angle and a clean line-of-sight to develop shadows that make objects interpretable. Tall, complex wreck structure casts long acoustic shadows and produces layover and multipath that no processing fully removes. A partially buried, low-relief debris field is close to the ideal SAS target; a large, intact, high-relief hull is not. MBES flown overhead handles vertical structure and overhangs more naturally, and the two datasets co-registered resolve far more than either alone.

Where survey teams get this decision wrong

1. Specifying resolution instead of a detection and classification requirement

Tender documents that call for “highest possible resolution” without stating the smallest object to be detected, the smallest to be classified, and the required positional uncertainty are asking to be over- or under-served. Write the requirement as an S-44 order plus an explicit target size and a classification standard. Resolution then falls out of the physics rather than out of a marketing sheet.

2. Treating SAS as a drop-in MBES replacement

SAS backscatter imagery and MBES bathymetry are not substitutes. Teams that swap one for the other discover late that they have imagery without metric depth, or depth without the detail to characterise a contact. On anything beyond a quick-look, budget for both sensors on the same platform or on paired passes, and specify co-registration tolerance in the deliverable.

3. Underestimating the absolute positioning problem

Centimetre imagery on a metre-uncertain position is a metre-uncertain product. The internal geometric fidelity of a SAS mosaic flatters the absolute georeferencing. If the client will use the data to position an obstruction, the aiding architecture – USBL update rate, LBL array geometry, INS grade and drift, sound-velocity control – governs the deliverable, and it has to be designed and budgeted before mobilisation.

4. Ignoring the area-coverage-rate crossover

The decision is frequently framed as “expensive SAS versus cheap MBES”, which is only true at low resolution over small areas. The moment the requirement is cm-scale characterisation over hectares, a low-flying MBES becomes slow and expensive because its swath shrinks with altitude. Run the numbers on area coverage rate at the required resolution, not at each sensor’s most flattering setting. The crossover often favours SAS sooner than intuition suggests.

5. Confusing a one-off heritage image with a monitoring baseline

A single spectacular dataset is not automatically a baseline. Change detection needs repeatability: a documented sound-velocity regime, known and recorded sensor geometry, a stable datum, and a processing recipe that can be reproduced on the next visit. If the client wants to compare surveys years apart, the metadata and repeatability discipline matter more than the peak resolution of the first pass.

How to decide, concretely

Work the decision in this order, and make each step a written acceptance criterion rather than an intent.

• Fix the detection and classification targets first. State the smallest cubic feature to be detected and the smallest to be classified, and map them to an S-44 Edition 6.1.0 order (Exclusive, Special, 1a). If the smallest object to be classified is well below 0.5m, you are in SAS territory; if the requirement is detection of features 1m and larger with metric depth, a properly configured MBES meets it.

• Separate imagery from bathymetry in the deliverable. Specify backscatter imagery resolution and bathymetric total vertical/horizontal uncertainty as distinct line items. Do not let one imply the other. For a wreck baseline, require both a SAS mosaic and an MBES surface with a stated co-registration tolerance.

• Design the positioning to match the imagery. Set an absolute horizontal uncertainty target and build the aiding to hit it: USBL for open-water AUV work, or an LBL array with adequate baseline geometry where absolute accuracy is the acceptance criterion. Record INS grade, DVL bottom-lock status and sound-velocity control as QC evidence, per the intent of IMCA survey guidance on positioning verification.

• Run an area-coverage-rate trade at equal resolution. Compute survey duration for each option at the required resolution and grazing geometry, not at nominal settings. Include mobilisation, transit, line turns, and the reacquisition passes complex structure will demand. Let the crossover, not the day rate alone, drive the platform choice.

• Specify repeatability if the client wants a monitoring baseline. Require documented sound-velocity profiles at defined intervals, a fixed and recorded sensor geometry, a reproducible processing workflow, and delivery of the raw data alongside the products so a future survey can be genuinely compared.

• Verify against ground truth where the stakes justify it. For obstruction and clearance work feeding an engineering decision, corroborate at least a sample of contacts with an independent sensor or a visual pass. High-resolution acoustics resolve geometry; they do not always resolve material or condition.

The Monitor case is a clean illustration of the technique’s ceiling: murky water, 73m depth, complex structure, and a client that needed characterisation detail beyond what a surface MBES delivers, plus a baseline for the future. That combination justifies the premium. A flat, low-relief debris field where the requirement is detection of metre-class objects with charted positions usually does not – an MBES spread meets Special Order and costs less to field. The discipline is refusing to answer the platform question until the detection, accuracy and coverage requirements are written down. Get those three right and the sensor choice stops being a matter of preference and becomes a matter of physics and budget.

---

Based on: Advanced sonar technology captures the USS Monitor in unprecedented detail