The ocean blocks a simple satellite positioning answer
Did you know that a ship can know its satellite position accurately while the seabed instruments beneath it remain impossible to locate with ordinary satellite reception? Conductive seawater strongly attenuates the radio-frequency signals used for standard GNSS. GNSS–acoustic geodesy instead joins satellite observations on a surface platform with acoustic ranging to transponders on the seafloor. A coordinate is inferred from many observations rather than received directly underwater.
This distinction determines how to read a report. A measured acoustic travel time is not a measured eastward movement. The travel time must be interpreted with sound speed, transponder turnaround delay, platform position and geometry. A difference between inferred coordinates at two dates may indicate tectonic displacement, but it may also include changes in seawater structure, measurement setup or analysis assumptions.
Treat every calculation below as a teaching example, not a record of a field campaign. The purpose is to trace how a technical observation becomes a modelled position, and then a possible physical interpretation. Responsible translation keeps observation, assumption, fitted value and inference separate.
Three instruments, three different locations
The GNSS antenna above the water, acoustic transducer carried by the surface platform, and seafloor acoustic transponder are three distinct physical objects. GNSS locates the antenna phase centre. The acoustic measurement refers to the submerged transducer. A seabed reply comes from its assigned transponder. Copying a vessel position into a transducer-position column without the offset correction changes the geometry.
The offset between antenna and transducer is the lever arm, usually surveyed in a defined body-fixed reference frame. Roll, pitch and heading must rotate that vector into the navigation frame. A rigid platform still changes the transducer’s earth-fixed location when it tilts. Precision in the antenna coordinates cannot erase error in lever-arm measurement or attitude timing.
When reading a figure, identify its coordinate origin. Some authors label the vessel centre, others the antenna or transducer. Ask whether the positions refer to the same reference point before subtraction. A simple drawing showing two instruments is often more important for comprehension than a table with impressive decimal precision.
The difference between acoustic time and acoustic distance
A surface transducer sends an interrogation and receives a reply. The measured two-way travel time includes acoustic propagation and, depending on instrumentation, a fixed or measured transponder turnaround interval. Subtract the relevant electronic delays before using water propagation physics. The travelled distance is not the one-way range multiplied by one: a pulse has gone to the seabed and returned.
Under a deliberately simplified constant-speed straight-ray model, one-way range R equals c times corrected two-way propagation time t divided by two. Here c is in metres per second, t in seconds and R in metres. The formula is a powerful unit check, but should not silently replace a depth-dependent sound-speed and ray-tracing model used in professional geodesy.
Example: total logged interval 2.400 seconds, turnaround 0.050 second, and assumed c of 1,500 metres per second. Corrected t is 2.350 seconds. R equals 1500 times 2.35 divided by two, or 1,762.5 metres. If someone forgets the turnaround, the calculated range becomes 1,800 metres: a 37.5-metre error that has nothing to do with movement of the seafloor.
Sound speed cannot be guessed once for every ocean
Sound travels through seawater at speeds influenced by temperature, salinity and pressure. Those properties vary with depth and time and can also change horizontally across a survey. A CTD profile helps describe the water column; repeated environmental observations may be needed when an internal wave, front or seasonal change affects propagation. A single surface temperature is not a full acoustic model.
Straight rays are only an approximation. When speed changes with depth, paths bend and the integrated travel time depends on the ray geometry. An analysis may ray-trace through a supplied profile, fit effective sound-speed corrections, or constrain horizontal gradients. Each choice creates a different set of assumptions and error trade-offs.
A good summary does not write ‘sound speed was corrected, so all ocean effects were removed’. It states how the water model was obtained, what spatial and temporal variation was permitted, and which residual uncertainty remains. The ocean is a moving part of the instrument system.
Range error versus coordinate error
For the simple relation R = ct/2, holding t fixed gives a range sensitivity of t/2 metres for each one-metre-per-second error in c. If corrected t is 2.350 seconds, a speed bias of 1 metre per second gives roughly 1.175 metres of equivalent range bias. That calculation is useful, but it is not a claim that the horizontal seafloor position must be wrong by exactly 1.175 metres.
Coordinates are fitted from multiple slant ranges from different places. Common errors may cancel partly in the array-centre solution, while geometric asymmetry can convert them into systematic horizontal shifts. Sound-speed gradients matter especially when one side of the array is sampled under different conditions from another.
Always distinguish uncertainty of an individual slant range, uncertainty of one transponder coordinate and uncertainty of a multi-transponder array centre. They are not different labels for the same number. A very sensitive raw range can support a more stable horizontal centre under a well-designed, sufficiently constrained observation pattern.
Why an array often helps
Multiple transponders make a seafloor site more than a single range endpoint. The geometry of a transponder array can let processing estimate a common horizontal centre while suppressing some shared water-column disturbances. Symmetric distributions of surface observations improve separation between a true horizontal shift and a sound-speed model error under suitable assumptions.
That advantage does not come free. Transponder placement, array rigidity, seafloor stability, acoustic reliability and survey duration matter. Three or more stations are common in some precision designs, but a number in the methods table does not guarantee a particular accuracy. Shallow-water single-transponder methods can also work when their observation geometry and modelling are appropriate.
Translate statements about precision as conditional. ‘The array approach can reduce horizontal error caused by common sound-speed variation’ is defensible. ‘Three stations eliminate ocean uncertainty’ is not. Published performance should be tied to the actual dataset and its independently evaluated uncertainty.
Why a balanced vessel track matters
A surface vessel that observes a transponder only from one direction provides weak information about some position components. A vessel that obtains good ranges from different bearings can better constrain the three-dimensional solution. A circular-looking planned track is not automatically balanced if acoustic dropouts cause most usable observations to come from one side.
Sound-speed gradients complicate geometry: a common speed bias and a horizontal position change may affect travel-time residuals in partially similar ways. Observing both sides of the site can help distinguish them, but rapidly changing conditions and incomplete coverage may leave ambiguity.
Read the actual usable-ping locations rather than just the cruise line. Ask for counts by sector, time span and transponder, and for uncertainty ellipses. A small numerical residual with an unbalanced design can still hide a biased coordinate.
Attitude corrections have centimetre-scale consequences
To get the transducer’s earth-fixed position from the GNSS antenna, one rotates the surveyed lever arm using the measured orientation of the surface platform. Roll and pitch may change over seconds; heading conventions and magnetic versus true directions must also be consistent. Attitude errors therefore translate into position errors even when GNSS itself is precise.
For a simple perpendicular lever arm of 1.5 metres, an orientation error of one degree is approximately 1.5 times 0.01745, or 0.026 metre under the small-angle approximation. That is about 2.6 centimetres, already comparable with the movements some studies investigate. Real transforms use full three-dimensional rotations and correctly aligned timestamps.
Do not interpret that example as a universal survey error. Its direction and magnitude depend on lever-arm orientation, roll/pitch/yaw and covariance. It shows why a report should record the body frame, orientation sensors, offsets and calibration rather than quoting GNSS precision alone.
Clock synchronization and turnaround time
An acoustic travel time can be contaminated by device latency, reply-delay calibration or a mismatch between clocks. Different subsystems may log GNSS, inertial attitude and acoustic arrivals at distinct rates. Processing needs the transducer position at the relevant transmission and reception epochs, not merely the nearest position printed in a log.
A one-millisecond uncorrected error in two-way travel time equals 0.001 second. At constant c = 1,500 metres per second it represents about 0.75 metre of equivalent range. This illustrates sensitivity without suggesting that modern instruments necessarily have millisecond errors. Practical systems seek much better timing accuracy with documented calibration.
Language must distinguish an electronic delay from additional sound path length. The former is an instrument parameter; the latter is a property of propagation and geometry. A time series with an abrupt jump after firmware maintenance should be checked for altered delay conventions before being called earthquake deformation.
Geometry of one acoustic range
A single ideal range constrains a three-dimensional point to a sphere around the acoustic transducer. It does not identify one point on the seabed uniquely. Many ranges from changing surface positions intersect these constraints in a fitted solution. Geometric diversity and a well-characterized depth help separate east, north and up coordinates.
For a constant-speed thought experiment, range is the square root of the sum of squared differences in x, y and z. This is Euclidean geometry, not a complete model of curved acoustic rays. A processing algorithm predicts travel times for an assumed transponder location and adjusts that location to fit many observed values.
An inverse model may fit other parameters simultaneously, including sound-speed corrections or instrumental offsets. A small numerical minimum does not prove a unique correct physical solution when different parameter combinations produce nearly equivalent predictions. Read convergence together with residual patterns and parameter correlations.
Residual plots as diagnostic instruments
Travel-time residuals compare measured and predicted values, usually in seconds or milliseconds. Equivalent range residuals are sometimes given in metres. A low average residual can conceal positive and negative errors that cancel. Root-mean-square residual, median, range and time trends each reveal a different feature of fit quality.
Plot residual against observation time, bearing, transponder ID, slant range and vessel position. Drift over time may be consistent with changing sound-speed conditions or clocks. Azimuthal patterns may suggest horizontal gradients or lever-arm modelling. One consistently troublesome transponder may have a distinct delay, response or installation issue.
These are alternative hypotheses, not automatic verdicts. A QC section is strongest when it tests the candidates and shows sensitivity to rejected points. Simply saying ‘all residuals are small’ without units, counts or independent validation is weak evidence.
Vertical and horizontal displacement have different limits
Horizontal array-centre positioning can be especially robust when symmetric geometries suppress common water-column errors. Vertical estimates may be more sensitive to sound-speed structure, datum interpretation and instrument motion. Different coordinate directions can therefore have unequal uncertainty even in the same survey.
If a table gives an east shift of 3 centimetres with 1-centimetre credible uncertainty but an up shift of 8 centimetres with 12-centimetre uncertainty, the horizontal result is more strongly resolved under those stated assumptions. Do not turn both point estimates into equally certain motions by listing them without error bars.
Some reports intentionally prioritize horizontal deformation along offshore plate boundaries. An educational explanation should preserve that design choice instead of implying that absence of a precise vertical result means the instrument failed.
Reference frame is part of the measurement
A GNSS-A location depends on the terrestrial reference frame, realization and coordinate epoch. Coordinates can change with plate motion and frame updates even if a particular local benchmark is stable relative to its neighbourhood. Latitude–longitude alone does not tell readers enough to compare two high-precision geodetic results.
Vertical datums introduce another dimension. Ellipsoidal height, geoid-related height, chart datum and instantaneous sea level answer different questions. Combining a bathymetric surface with geodetic positions requires consistent definitions and carefully assessed transformations.
Before reporting a displacement, confirm that both survey solutions are in the same frame and epoch convention. A coordinate subtraction across inconsistent references is arithmetic but not a valid measurement of deformation.
Comparing two campaigns honestly
Position change is usually obtained by subtracting two solutions separated in time. Its uncertainty depends on the errors of both and whether those errors are correlated. Repeatability within a day does not automatically quantify long-term stability across seasons, instruments, vessel configurations and software versions.
Illustrative east coordinates of 120.00 and 120.04 metres differ by four centimetres. If each campaign has independent one-standard-deviation uncertainty of three centimetres, the standard uncertainty of their difference is the square root of nine plus nine square centimetres, about 4.24 centimetres. Under that illustration a four-centimetre difference is not strongly resolved.
The covariance assumption matters: shared biases may cancel, and unmodelled systematic changes may enlarge uncertainty. Report both the point estimate and its evidential limits. Extra decimal places are not a substitute for calibration or a trustworthy error model.
From geodetic movement to geological cause
GNSS-A can help investigate offshore earthquakes, slow slip and deformation along plate boundaries. But a positional shift is not a unique explanation of its cause. Local sediment disturbance, array movement, instrument replacement and seasonal modelling errors must be considered alongside tectonic interpretations.
If observations are made only every few months, a displacement step is known to occur within an interval, not necessarily on a precisely known day. If the interval contains an earthquake, that provides a candidate event; attributing all observed displacement to it still requires careful timing and model checks.
Plain language can say a change is consistent with offshore deformation during a specified interval. Stronger claims about fault-slip magnitude, direction or exact timing require independent structural models and additional evidence. That distinction protects readers from misleading certainty.
Comparing method families without conflating them
GNSS-A estimates positions of identifiable seafloor transponder sites. Multibeam sonar estimates seafloor depth and shape. Bottom pressure systems measure pressure influenced by water loading and other processes. Satellite altimetry addresses sea-surface elevation. These measurements complement one another but are not interchangeable.
For example, a change in local bathymetry might reflect sediment migration or mapping uncertainty rather than rigid movement of a GNSS-A transponder. A bottom pressure change could reflect a tide or storm. A transponder coordinate shift could be genuine even without a visibly changed bathymetric image.
An interdisciplinary report should state which instrument measured what, then explain how the observations were transformed to a shared reference frame and epoch. Agreement across independent methods can increase confidence; apparent disagreement may expose datum and correction issues.
Acoustic multipath and outlier rejection
Acoustic pulses can reflect from the surface or seafloor and reach a receiver on different paths. Low signal-to-noise ratio, mistaken arrival picking, interference and missed replies can degrade the travel-time observations. Good processing distinguishes valid direct arrivals from delayed multipath and questionable detections.
Outlier thresholds need to be transparent. If a disproportionate number of western observations are rejected, the remaining geometry may become biased even while the formal residual statistics improve. Ask for attempted interrogations, received replies, accepted pings and reasons for exclusion.
Quality control should also test whether removing or restoring suspect observations materially changes the solved position. A highly precise-looking estimate obtained by deleting inconvenient data is not necessarily an accurate one.
Worked range and unit exercises
Example one: total measured TWTT 1.80 seconds, turnaround 0.10 second, constant c 1,500 metres per second. Corrected time is 1.70 seconds; simple one-way range equals 1.70 times 1,500 divided by two = 1,275 metres. The division by two is necessary because timing includes the outward and return paths.
Example two: a two-way timing offset of 0.004 second gives a simple range change of 1,500 times 0.004 divided by two = 3.0 metres under the same assumptions. This is a sensitivity estimate, not a prediction of final coordinate error.
Check yourself: corrected travel time 3.0 seconds and c 1,480 metres per second. Equivalent range is 2,220 metres. A result of 4,440 metres usually means someone treated the round trip as one-way propagation.
Worked geometry exercises
Suppose a seabed transponder lies exactly 1,000 metres below the acoustic transducer, giving a straight-line range of 1,000 metres. Move the transducer 600 metres horizontally while keeping vertical separation fixed. The new range is the square root of 1,000 squared plus 600 squared, approximately 1,166.2 metres.
The changing observation angle adds geometric information. If all observations are made from the same azimuth, it becomes harder to distinguish some position errors from propagation biases. A balanced collection of bearings provides more informative constraints.
Try this independent transfer: 1,200 metres vertical separation and 500 metres horizontal offset produce a straight-line range of exactly 1,300 metres. This Pythagorean calculation is useful as a baseline, while real GNSS-A software may use curved acoustic paths.
Worked orientation error and correction
A 2.0-metre lever arm with a 0.5-degree perpendicular attitude error produces an approximate lateral error of 2 times 0.008727 = 0.01745 metre, or 1.75 centimetres. The angle must be converted to radians for this small-angle multiplication.
At a 3.0-metre offset and one degree, the same simple estimate is 3 times 0.01745 = 0.0524 metre. This illustrates why instrument surveys and timing matter when reporting centimetre-level positions.
Question: would the full three-dimensional displacement always equal this product? No. Real rotations act on vector components, and the error direction depends on platform geometry. The simple product is a sensitivity bound for a specified perpendicular configuration, not a replacement for the navigation transform.
Error patterns and repairs
Overclaim one: ‘we recorded a two-centimetre travel time’. Repair: travel time has temporal units; position changes have distance units. Overclaim two: ‘the sound-speed correction eliminates every ocean effect’. Repair: a correction follows a model and retains residual uncertainty.
Overclaim three: ‘the GNSS antenna moved, so the seabed moved’. Repair: platform motion is not seabed deformation; many acoustic ranges and reference corrections connect the two. Overclaim four: ‘the map shows fault slip exactly when the vessel surveyed’. Repair: a survey locates a position at that epoch and may only constrain a change between campaigns.
Overclaim five: ‘the fit is precise because the least-squares program converged’. Repair: assess residuals, calibration, correlations and alternate models. Good translation makes the evidence chain readable without turning numerical optimization into proof.
Frequently asked questions
Can GNSS-A map the entire seabed? No. It locates designated transponders or an array centre; extending those measurements to fault deformation requires additional models. Can a single acoustic range uniquely locate a three-dimensional point? No. One range leaves many geometrically possible locations.
Is a horizontal coordinate necessarily more reliable than a vertical one? Not universally, but some symmetric-array GNSS-A designs suppress common sound-speed errors more effectively horizontally. Does using multiple transponders automatically guarantee centimetre accuracy? No; timing, environment, survey geometry and instrument health still matter.
What should a reader ask first? What was measured, what was fitted, which coordinate frame was used, and how uncertain is the change? Those four questions prevent most of the common interpretation failures.
An evidence-first interpretation standard
The finished explanation should name the GNSS antenna geometry, transducer offsets and attitude corrections, transponder arrangement, corrected two-way travel times, sound-speed model, positioning inversion, coordinate reference and uncertainty budget. Missing details should be identified rather than invented.
Related reading includes the ocean CTD profile guide, the multibeam datum and uncertainty guide and the sea-level altimetry guide. Each teaches a different part of the ocean measurement problem.
The safest conclusion is that acoustic and GNSS observations were jointly modelled to estimate seabed positions, and their epoch-to-epoch differences support a physical interpretation only to the extent allowed by geometry, calibration and uncertainty. That sentence preserves more scientific meaning than an unexplained claim that a fault moved a precise number of centimetres.
When a sound-speed profile must be updated
An oceanographic profile is not merely a formality attached to an acoustic survey. Temperature and salinity layering can shift over hours, and surface heating or internal motions can change sound travel along different rays. If the surface platform surveys a broad array over several days, a single profile at the start may not describe later conditions. The report should identify which profiles were used at which times and why interpolation was adequate.
An analyst may fit effective sound-speed perturbations from the travel-time data, but doing so creates relationships with geometric unknowns. A fitted speed change can absorb a small position error; conversely, a fitted transponder shift can mask an unmodelled ocean gradient. Evaluating both parameter sensitivity and independent CTD evidence is more reliable than treating whichever model gives the smallest residual as unquestionable truth.
Transfer question: if profiles before and after a survey disagree, can their arithmetic midpoint automatically be called the correct profile for all pings? No. Time evolution, spatial gradients and the functional relationship between speed and travel time all matter. A midpoint can be a modelling choice only when its assumptions and uncertainty are justified.
Array centre, array shape and site stability
An array centre is a derived coordinate, not necessarily a metal marker placed at the geometric middle of the seafloor. The centre can be defined from estimated transponder positions and, in some methods, rigid-body constraints. If a transponder moves relative to its neighbours, the assumption that the entire cluster behaves as a rigid unit may no longer be valid.
Compare relative inter-transponder distances across campaigns before attributing every change in centre position to regional tectonics. A damaged mount, slumping sediment or an inadvertently relocated transponder could alter the fitted centre. Redundant sites and consistent relative geometry can provide evidence that the array behaves as intended.
Suppose one member shows an unusual shift while all others remain mutually consistent. A reader should not average away that observation automatically, nor call it a regional earthquake signal. The next step is to inspect that instrument, the range residuals, the seafloor setting and alternate solutions with and without the questionable member.
Travel-time ray curvature and path length
A sound ray can bend when the water-column sound speed changes with depth. The straight-line distance between transducer and transponder remains a geometric quantity, while the acoustic travel time follows an actual propagation path. Multiplying the measured time by one average speed can therefore introduce a model-dependent bias, especially for long oblique rays through strongly stratified water.
Ray tracing generally uses a speed profile to predict both path and travel time. The details depend on the approximation, boundary conditions and environmental inputs. Analysts may compare a straight-ray approximation with a more complete model and quantify the difference. The decision should be made using the survey precision required, not by assuming that the more complicated algorithm is always better in the absence of valid inputs.
Worked reasoning: two rays reaching the same transponder from different bearings may have different travel-time sensitivities to a horizontal sound-speed gradient. A residual difference between those bearings can indicate environmental structure. It is not automatically evidence that the transponder moved between successive acoustic pings.
Identifiability and parameter trade-offs
An inverse problem is identifiable when the data constrain the unknown parameters sufficiently to distinguish plausible solutions. In GNSS-A, some combinations of transponder depth, average sound speed, turnaround delay and geometry can produce similar travel-time predictions. Survey design and independent calibration limit these trade-offs.
The condition of a normal matrix or a posterior covariance can reveal poorly constrained parameter directions, but such numerical summaries depend on the assumed errors and model. When a report says a model is stable, ask whether stability was tested against different initial guesses, different sound-speed priors and reduced datasets. A low residual alone is insufficient.
A useful explanation might say that acoustic observations strongly constrain the array’s horizontal centre while vertical position and a sound-speed parameter remain correlated. That is more informative than describing all coordinates as equally accurate. It also tells readers which additional measurement could most improve the next survey.
Precision, accuracy and reproducibility
Precision concerns the spread of estimates under specified repeated conditions; accuracy concerns their closeness to the true quantity, which can be difficult to establish for an inaccessible seafloor site. Reproducibility across different ships, seasons or processing methods tests a broader range of biases than repeated solutions from the same few pings.
A reported uncertainty of one centimetre may refer to internal formal precision of the least-squares fit, not to total positioning accuracy. Look for external comparisons, calibration checks, sensitivity studies and residual spatial structure. If systematic uncertainty dominates, merely increasing the number of highly correlated pings cannot reduce it by the square root of the sample count.
Transfer problem: two independent processing teams obtain positions that differ by eight centimetres, but each quotes a two-centimetre formal error. Is it defensible to retain only the smaller error? No. The discrepancy is evidence for unaccounted differences or model assumptions that should be investigated before a consolidated uncertainty is stated.
Position time series, rates and steps
GNSS-A monitoring can yield a sequence of position estimates, often separated by weeks or months. A velocity estimated from such a time series is a fitted change per unit time, such as millimetres per year; an event-related step is a finite displacement. A rate should not be translated as a one-time jump, and a step should not be presented as a continuous trend.
Consider illustrative eastward positions separated by two years and differing by 40 millimetres. Dividing gives an average 20 millimetres per year over that interval. However, if an unobserved earthquake occurred in between, interpreting all change as a steady velocity could be misleading. The time series and independent event record determine which model is appropriate.
A report comparing epochs must state the observation dates, data gaps, reference frame and fit model. An apparently straight line across sparse points can conceal transients. Clear reporting preserves which movement was observed directly and which was interpolated or estimated.
Different error correlations across epochs
Measurement errors can be independent, correlated within a survey, shared across surveys or biased differently at each epoch. A common reference-frame offset may cancel when differences are formed; a sensor replacement that alters transducer geometry may introduce a new bias. The covariance of a displacement depends on these relationships, not just two individually printed standard deviations.
For independent errors with standard deviations σ1 and σ2, the variance of a difference is σ1 squared plus σ2 squared. With covariance C between the errors, it becomes σ1 squared plus σ2 squared minus twice C. A positive shared covariance can reduce difference uncertainty, but an unknown systematic shift can increase practical uncertainty beyond that simple model.
Transfer exercise: independent 2-centimetre and 3-centimetre errors give a combined standard uncertainty of the square root of 4 plus 9, about 3.61 centimetres. If the campaign estimates share a bias, that calculation is no longer complete. State the independence assumption before quoting the result.
Checking a coordinate table without specialist software
A reader can perform a useful first audit using only column headings and units. Find the survey date, site code, number of transponders, coordinate reference, east/north/up values, a measure of uncertainty, and data-quality counts. Confirm that all repeated observations refer to the same physical array and that any instrument replacements are documented.
Next subtract only columns with matching frames, epochs and definitions. Convert millimetres and centimetres to metres consistently. Check whether the reported displacement magnitude equals the square root of the squared horizontal components, and whether ‘total displacement’ includes the vertical component. If not, ask what definition the author intended.
One example: 0.03 metre east and 0.04 metre north correspond to a horizontal magnitude of 0.05 metre, using the Pythagorean theorem. That vector length says nothing by itself about measurement significance; covariance and confidence limits still decide whether the direction and magnitude are resolved.
A statement of no detected movement can be meaningful
Sometimes an accurately operated survey finds no statistically resolved change. That is not a failure of science or a reason to manufacture a dramatic story. A non-detection can constrain how large a movement could have been without producing a distinguishable signal under the stated uncertainty.
A credible non-detection report provides the measured difference, detection capability and monitoring interval. It distinguishes lack of evidence for movement from proof of absolutely zero movement. It also avoids combining unrelated uncertainties merely to make a narrow interval.
Readers benefit from a concrete explanation: the instrument array was observed on two occasions, but the position difference was too small relative to measurement and model uncertainty to confidently infer motion. That finding can still inform future survey design and scientific hypotheses.
Seafloor installation and physical integrity
A transponder used as a geodetic marker must remain coupled to the seafloor location it is intended to represent. Sediment settling, trawling impacts, cable tension, anchor movement or local mass wasting could alter the instrument position independently of regional plate motion. The relevance of each mechanism depends on the hardware and seabed environment.
Survey documentation should record deployment, recovery if applicable, seafloor conditions and any changes to instrument status. An unexpected isolated motion of one marker may justify checking installation rather than forcing it into a regional deformation model. That is particularly important when the inferred displacement is small.
Translating instrument status into public language should not claim seabed damage without evidence. ‘A possible local instrument or foundation shift requires evaluation’ is a disciplined conclusion; ‘the fault ruptured beneath the transponder’ needs a much stronger chain of evidence.
A better way to present a GNSS-A results figure
An accessible figure set can place raw travel times and quality flags first, followed by representative residual plots, survey-track geometry, array centre and uncertainty ellipse, and finally a time series of comparable positions. Each panel answers a different question. Showing only the last displacement arrow hides the assumptions that made it possible.
Use a common coordinate orientation and indicate whether arrows are exaggerated for visibility. Label reference frame and time interval on the figure itself. When error ellipses are present, state their statistical confidence and whether they include systematic terms. A tiny plotted ellipse with no definition is not a meaningful guarantee.
Captions should distinguish direct observation from fitted model. The phrase ‘observed position’ is common shorthand, but an educational explanation can say ‘position estimated from acoustic ranges and GNSS’. That small correction prevents an important conceptual misunderstanding.
Additional independent transfer tests
Test one: if an acoustic instrument reports two-way corrected time 2.0 seconds and an assumed constant speed of 1,500 metres per second, the straight-path range is 1,500 metres. Test two: a timing offset of 0.002 second corresponds to 1.5 metres of simple range bias. These examples share a physical formula but test distinct inputs.
Test three: a horizontal displacement has east component 6 centimetres and north component 8 centimetres. Magnitude is 10 centimetres, but a reader still needs confidence intervals to know whether that is significant. Test four: the same array appears stationary in one reference frame and moving in another. Check frame motion and epoch transformations before claiming a contradiction.
Test five: one epoch uses a new acoustic transducer mounted at a different lever arm. Can the old offset be reused? No. Test six: a model fits almost perfectly after removing half the pings from one sector. Is that necessarily an improved geodetic result? No. Inspect whether the missing geometry has weakened the solution despite lower residuals.
Original scientific and methodological sources
GNATSS GNSS-A Experimental Design
GNATSS Preprocessing and Lever Arms
Xie et al. 2023 Earth and Space Science, Shallow-Water Seafloor Geodesy
Yokota et al. 2022 Earth and Space Science, Sound-Speed Structure and GNSS-A