Bottom Tracking vs Water Profiling: Which ADCP Mode?

Why the two depth ranges differ, where each mode fails, and how five questions settle the choice.

Every argument about ADCP modes starts with a quiet misunderstanding: that bottom tracking and water profiling are two settings of the same measurement. They are not. One measures how fast the platform moves over the bed; the other measures how fast water moves past the instrument. A survey usually needs both, and the failures in reports almost always come from running one while believing you are running the other.

This article separates them by what each one references, explains why the depth specifications differ so much, and puts the failure modes on one page.

Two modes, two different questions

Water profiling is the measurement the instrument exists for. Each beam transmits a pulse, listens to the echo from suspended particles at successive delay times, and converts Doppler shift into a velocity for each depth cell. The result is a profile relative to the transducer. That phrase matters: if the transducer moves, the motion is added to every cell in the profile, and nothing inside the instrument can tell you how much of the observed velocity is water and how much is the boat.

Bottom tracking closes that gap. The same beams also receive a strong return from the bed. Because the bed is a solid boundary running at the speed of the ground, the instrument can lock onto it, measure its own velocity relative to it, and report the range to it along each beam. In a moving-boat discharge measurement, that bottom-track velocity is the reference speed the whole computation rests on: water velocity minus platform velocity, integrated over the cross-section.

The practical consequence is simple: profiling tells you what the water is doing, bottom tracking tells you how fast the instrument is moving. Run the first without the second on a moving platform and you have velocity data with an unknown offset added to it.

Why bottom-track depth range exceeds profile range

On paper the two depth figures look like a contradiction. A 600 kHz unit profiles to 65 or 75 m, yet its bottom tracking reaches 120 m. The 300 kHz unit profiles to 120 or 150 m and tracks bottom to 300 m. The 1200 kHz unit profiles to 22 or 30 m and tracks to 45 m.

The two numbers describe different targets.

Bottom tracking works against a discrete, hard reflector. The bed returns a strong, compact echo, and the instrument only has to detect it and correlate successive pings to get a velocity. Profiling works against volume scattering: the echo from suspended sediment in a single cell is orders of magnitude weaker, and each cell needs usable signal-to-noise on its own. The same transmit power therefore reaches further as a bottom detector than as a profiler.

That is why the specification you quote depends on the question. If someone asks how deep the instrument can see the bed, the answer is the bottom-track range. If someone asks how much of the column can be measured, the answer is the profile range. Using 120 m as a profiling claim for a 600 kHz unit is wrong by 45 m, and that error ends up in tender documents.

ModelProfile range (high precision / max)Bottom-track depth rangeDepth-sounding beamMinimum cell
ADCP-600K (Tingseek)65 m / 75 m0.2–120 m0.2–160 m0.08–4 m
ADCP-300K (Tingseek)120 m / 150 m2–300 m—1–8 m
ADCP-1200K (Tingseek)22 m / 30 m0.2–45 m—0.04–2 m
ADCP-R9 (Tingseek)0.06–35 mNot specified500 kHz sounding beam0.02–2 m

ADCP-600K 600 kHz acoustic Doppler current profiler with five-beam transducer head

ADCP-600K: profiles to 65–75 m, tracks bottom from 0.2 m out to 120 m. The gap between those two numbers is the whole point of this article.

The second asymmetry is at the shallow end, and it decides more surveys than the deep end does. Bottom tracking cannot start inside the transducer's near field, so the minimum tracking depth follows the frequency. At 600 kHz bottom tracking begins at 0.2 m; at 300 kHz it begins at 2 m. In a channel 6 m deep, that 1.8 m difference is not a rounding error, it is the difference between resolving the surface layer and losing it.

What bottom tracking actually gives you

Three products come out of bottom tracking, each failing in its own way.

The first is platform velocity over ground. Four Janus beams at a nominal 30 degrees measure relative velocity along each beam; beam geometry combines them into three components in the instrument frame, then heading and tilt rotate them into earth coordinates. That velocity is independent of GNSS, which is why bottom tracking keeps working under bridges, inside locks and in narrow canyons where a fix degrades.

The second is depth. Each beam reports a slant range, and with beam angle, instrument depth and attitude that range converts to water depth. Our 600 kHz unit also carries a vertical sounding beam with a 0.2–160 m range, which is the better check: when slanted-beam depths and the vertical beam disagree, the usual cause is a sloping bed or a poor attitude solution.

The third is the data quality flag. Bottom tracking reports whether it is locked and how consistent the four beam solutions are. Log that flag per ensemble, rather than trusting the computed discharge afterwards.

Accuracy is quoted per velocity, not per mode: ±0.25% of the measured value plus 0.2 cm/s. In practice the dominant error is not the Doppler measurement but the reference geometry: an assumed instead of surveyed beam angle, a default sound speed, a heading offset of a couple of degrees, or a lever arm nobody measured.

Where bottom tracking stops working

Six conditions cover almost every loss of bottom lock.

Soft mud and fluid mud. The bed still reflects, but the echo is weak and diffuse, and the correlation peak that the tracker needs becomes broad and unstable. The unit may report a lock while the range wanders by a metre or more. Lowering the frequency and lengthening the pulse helps; past that, the honest fix is to stop relying on bottom tracking for the reference.

Steep or broken bed. Four beams aimed at 30 degrees hit the bed at different ranges when the bottom slopes, and the per-beam depths stop describing one plane. Keep transects along the contour rather than across it, and trust the attitude data over an assumed flat bed.

Water deeper than the bottom-track range. Beyond 120 m a 600 kHz unit loses the bed entirely, and beyond 300 m a 300 kHz unit does. Note that this limit is independent of the profiling limit: at 90 m of depth a 600 kHz unit still tracks bottom but profiles only part of the column, so the reference stays valid while the velocity data runs out.

High sediment load. Suspended sediment raises absorption and scattering at once. The bed echo is attenuated on the way back while the column returns more energy, so the tracker works against a brighter background. The sediment discussion in our forum covers the settings that recover most of it.

Bubbles and wake. Entrained air under the transducer and the vessel's own wake scatter far more strongly than the bed. The tracker locks onto the bubble layer or drops out for the length of the wake. Moving the transducer clear of the wake path, or running the transect the other way, beats changing any setting.

A moving bed. This one deserves its own section, because the instrument reports a perfectly healthy lock while the answer is wrong.

Water profiling when there is no usable bottom

Fixed deployments are the easy case. On a seabed frame, mooring or pile the transducer does not move, so a water profile is already absolute and bottom tracking is optional. The bed may still serve as a range check and as a zero-velocity reference during a loiter, but nothing depends on it.

Moving platforms without a trackable bed are the hard case. The reference has to come from elsewhere: GNSS, ideally with RTK, or a USBL system on a vehicle. Three error sources then matter more than the ADCP specifications:

  • Heading. A one-degree heading error at 2 m/s gives about 3.5 cm/s of cross-track velocity, larger than the instrument's own accuracy. Dual-antenna heading or a surveyed gyro calibration is worth the effort.
  • Latency and alignment. GNSS positions and ADCP ensembles are timestamped separately, and the offset injects a velocity error proportional to acceleration. Log both on a common clock and compute the offset instead of assuming zero.
  • Lever arm. The offset between GNSS antenna and transducer turns pitch and roll into velocity error. Survey it, do not guess it.

If none of those references is available, the remaining option is to profile from a stationary platform. That is not a compromise: for shear, stratification and near-bed backflow it is the better measurement, because the instrument is not moving through the structure you are trying to resolve.

The moving-bed problem

In sand-bed rivers the bed itself travels. Bedload moves at a speed comparable to the lower water column, and bottom tracking locks onto whatever the beams hit: sand grains in motion. The instrument cannot tell that its reference is drifting, and it reports a valid lock with normal-looking numbers.

The error is not subtle in consequence. Discharge is computed from water velocity relative to the bed, integrated across the section using the platform's motion. If the reference velocity is biased, the bias enters discharge in direct proportion, and a survey can be wrong by several percent while every quality flag says the data are good.

Two checks catch it, and both are cheap.

The station-keeping test. Hold the boat stationary over the bed for a minute or two, on a bow anchor or a fixed pile. A correct bottom track reads close to zero in both components. A steady non-zero reading of a few centimetres per second, consistent in direction, is the bed moving, not the boat.

The cross-check against ground truth. Compare bottom-track speed with RTK GNSS speed over the same transect. Where the two diverge systematically and the GNSS solution is healthy, the difference is bed velocity. Record it and report it: it is also an estimate of bedload transport rate.

A 40 m river, and how five questions settle it

Take a tidal river 40 m deep, a moving-boat discharge measurement, and a requirement to resolve the shear layer near slack water at 0.5 m cells.

With a 600 kHz unit, 0.5 m cells give 80 cells across the column, about 78 usable after blanking; bottom tracking starts at 0.2 m and reaches 120 m, so the reference never limits the survey. With a 300 kHz unit, the minimum cell is 1 m, so the column yields 40 cells, bottom tracking starts at 2 m and reaches 300 m, and the shear layer is averaged into two cells instead of one. Here the 600 kHz unit is right, and the 300 kHz unit's extra range buys nothing.

At 90 m the answer flips. A 600 kHz unit still tracks bottom at 90 m, so the reference stays valid, but its profile stops at 75 m and the upper column is missing. The 300 kHz unit profiles the whole 40-cell column at 1 m cells and tracks bottom with margin to 300 m. What changed was not the bed but the ratio between profile range and water depth.

ADCP-300K 300 kHz acoustic Doppler current profiler for deep river and coastal sections

ADCP-300K: profiles to 120–150 m and tracks bottom from 2 m out to 300 m, which is what makes a 90 m section workable at 1 m cells.

Five questions settle the mode before any hardware choice:

QuestionAnswer that points to bottom trackingAnswer that points to water profiling only
Is the platform moving?Moving boat, vessel, USVSeabed frame, mooring, pile
Do you need discharge or just velocity structure?Discharge: reference speed requiredStructure: reference is the instrument
Is there a GNSS/RTK reference available?Not needed, or degradedAvailable and well calibrated
Is the bed trackable?Sand, gravel, rock, cohesive bedFluid mud, very deep water, heavy bubble load
Is the bed moving?Station-keeping test reads zeroTest reads a steady offset: use GNSS reference

Question one decides whether a reference is needed at all. Question three decides where it comes from. Question four decides whether bottom tracking can supply it. Question five decides whether the number it supplies is trustworthy. Skip any of them and the rest of the survey inherits the assumption.

FAQ

Can I run bottom tracking and water profiling at the same time?

Yes. Most profilers interleave bottom-track and profile pings in one ensemble, which is why a moving-boat survey normally logs both. The cost is ping rate and power: bottom-track pings cut the number of profile pings per second, so low-power deployments trade temporal resolution for a reference they may not need.

Is bottom-track depth the same as charted water depth?

Not directly. Bottom tracking reports slant range along each beam, converted using beam angle, instrument depth and attitude. On a sloping bed the four beams disagree, and the vertical sounding beam is the more trustworthy depth. On flat, soft bottom the two agree within a few decimetres and the distinction stops mattering.

Does bottom tracking work in 200 m of water?

It depends on the frequency: a 600 kHz unit stops at 120 m and a 300 kHz unit reaches 300 m, so a 200 m section already rules one of them out. Past the bottom-track range the reference has to come from GNSS or another positioning system, while the profile still works, because the two limits have different causes.

What happens if bottom tracking drops out mid-transect?

Log the validity flag and let processing fall back to GNSS over ground for that segment, rather than discarding the transect. A segment with a flagged substitution is defensible in a report; a segment with a silent hold-last-value is not. This is also why the reference source belongs in the metadata of every discharge measurement.

Bottom Tracking vs Water Profiling: Which ADCP Mode?
China Sonar November 16, 2026
Archive
600 kHz vs 300 kHz ADCP: Which Frequency Fits Your Survey?
Absorption, cell size and bottom tracking, compared against the published figures of RDI, Nortek and SonTek.