Geomorphology 68 (2005) 25 – 37

www.elsevier.com/locate/geomorph

Measuring flow velocity and sediment transport with an acoustic

Doppler current profiler
Ray Kostaschuka,*, Jim Bestb, Paul Villardc, Jeff Peakallb, Mark Franklinb
a
Department of Geography, University of Guelph, Guelph, Ontario, Canada N1G 2W1
b
Earth and Biosphere Institute, School of Earth Sciences, University of Leeds, Leeds, LS2 9JT, West Yorkshire, United Kingdom
c
Parish Geomorphic, 10 Mountainview Road South, Suite 207, Georgetown, Ontario, Canada L7G 4J9
Received 1 August 2003; received in revised form 27 May 2004; accepted 20 July 2004
Available online 16 December 2004

Abstract

An acoustic Doppler current profiler (aDcp) measures three-dimensional velocity profiles within the water column using the
Doppler shift principle, whilst the bottom tracking function and acoustic backscatter can be used to measure bed load velocity
and estimate suspended sediment concentration. The aDcp offers many advantages over traditional single-point current meters
and sediment samplers, including deployment from a moving launch, a single instrument for both velocity and sediment
transport measurements, profiles of three-dimensional velocity and suspended sediment and the ability to map an entire flow
field. Limitations of aDcps include a large sampling diameter close to the bed, coarse measurement of vertical velocity,
fragmented bottom track records, a poor understanding of the relation between bottom tracking and the mechanisms of sediment
transport, and sensitivity of the acoustic backscatter to particle size.
D 2004 Elsevier B.V. All rights reserved.

Keywords: Acoustic Doppler current profiler; Flow velocity; Bed load velocity; Sediment transport

1. Introduction Smith and McLean, 1977; Kostaschuk and Villard,

1996). Mechanical samplers, such as the Helley–
A wide variety of techniques have been used to Smith, have also long been the standard for bed load
measure fluid velocity and sediment transport in measurement (e.g., Helley and Smith, 1971; Klinge-
geophysical flows. Measurements of flow velocity man and Milhous, 1971; Ryan and Porth, 1999), while
have traditionally relied on single-point current meters mechanical samplers and optical/acoustic sensors are
deployed from an anchored boat or bed frame (e.g., frequently used for measurement of the suspended
load (e.g., Downing, 1983; Hay, 1983; Thorne et al.,
1991; Zedel and Hay, 1999; Agrawal and Pottsmith,
* Corresponding author. Fax: +519 837 2940. 2000). All of these instruments require support or
E-mail address: [email protected] (R. Kostaschuk). sensor cables and can thus be difficult and dangerous
0169-555X/$ - see front matter D 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.geomorph.2004.07.012
26 R. Kostaschuk et al. / Geomorphology 68 (2005) 25–37

to deploy in deep flows, and their exact position sound is moving relative to the receiver, the frequency
within the water column is often unknown. In of the sound at the receiver is shifted from the transmit
addition, most of these instruments are limited to frequency, as expressed by:
single-point sampling and must be positioned at
FD ¼ 2FS ðV =cÞ ð1Þ
different levels above the bed to obtain vertical
profiles. Combined use of these instruments is where V is the relative velocity between source and
possible, but it is usually difficult to obtain simulta- receiver, c is the speed of sound, F D is the change in
neous multipoint measurements of flow velocity and the received frequency at the receiver (i.e., the
sediment transport. Acoustic Doppler current profilers Doppler shift), and F s is the frequency of the
(aDcp) are now widely used for measuring fluid transmitted sound. The transducers of an aDcp
velocity in geophysical flows (e.g., Admiraal and generate pulses of sound at a given frequency along
Demissie, 1996; Wewetzer et al., 1999; Best et al., a narrow beam of sound, in which the majority of
2001; Kostaschuk et al., 2004) and recent research has energy is concentrated in a cone only a few degrees
focussed on their potential to estimate bed load (e.g., wide. As the sound travels through the water, it is
Rennie et al., 2002; Villard et al., 2005) and reflected in all directions by particulate matter (e.g.,
suspended load (e.g., Reichel and Nachtnebel, 1994; sediment, biological matter, bubbles) but some portion
Lane et al., 1997; Reichel, 1998; Alvarez and Jones, of the reflected energy travels back along the trans-
2002). The purpose of the present paper is to review ducer axis toward the transducer where the processing
and evaluate aDcps for measurements of flow electronics measure the backscattered frequency, and
velocity, bed load and suspended load in fluvial and hence the Doppler shift. The Doppler shift measured
lacustrine environments. Examples are provided from by a single transducer thus quantifies the velocity of
the Fraser Estuary and Lillooet Lake in western the water along the axis of the acoustic beam. If the
Canada. The paper will demonstrate that aDcps offer distance between the transducer and the reflecting
many advantages over traditional measurements of object is decreasing, frequency increases, whereas if
velocity and sediment transport, but will highlight the distance is increasing, frequency decreases.
some serious limitations that require both consider- The measurement location within the water column
ation and development in future research. from the aDcp transducer is a function of the time at
which the return signal is sampled. By measuring the
return signal at different times following the transmit
2. Principles of operation pulse, the aDcp measures the profile of water velocity
at many distances from the transducer. The profile of
There are currently several manufacturers of water velocity is thus divided into range cells, or bins,
commercial aDcps (e.g., SonTek, RDI, Nortek) and where each cell represents the average of the return
interested readers should consult individual manufac- signal for a given period of time. For example, a 1-m
turers for details on their products (see www.sontek. range cell corresponds to an averaging time during
com; www.rdinstruments.com; www.nortek-as.com). which the acoustic pulse moves 1 m. Bin size and
The present paper specifically considers field use of effective sampling depth are a function of the
SonTek 1500 and 500 kHz systems, but the general operating frequency of the ADcp and the speed of
principles of operation are the same for all aDcps. sound through the water column, which may be
SonTek also manufactures several frequencies of influenced by factors such as water temperature,
aDcps and a pulse-coherent aDcp that is useful for salinity and sediment concentration. For example,
shallow flows. Additional information on these the SonTek 1500 kHz aDcp used herein can sample
systems is available from SonTek. 0.25 m bins to a depth of ~20 m and the 500 kHz
aDcp samples 1 m bins to a depth of ~100 m.
2.1. Measuring velocity with an aDcp An aDcp generally uses three or four beams to
measure velocity. SonTek aDcps employ three beams,
An aDcp measures the velocity of water using the with the 1500 and 500 kHz transducers being set at
Doppler shift principle, which states that if a source of 258 from the vertical axis and being equally spaced in
 R. Kostaschuk et al. / Geomorphology 68 (2005) 25–37 27

the horizontal (1208), thereby producing different the water column, and therefore the bin above the
orientations relative to the flow (Fig. 1). Since the inflection point is used to define the lower limit of
relative orientation of the three transducers is known, uncontaminated velocity measurements. In general, a
three-dimensional velocity can be generated by digital echosounder should also be deployed with any
combining the three along-beam velocity profiles. aDcp to provide more detailed and reliable bed
The raw velocity data are given as easting, northing profiles.
and up/down components but the easting and northing The SonTek aDcp provides a measure of standard
components can be rotated to produce velocities deviation for each velocity profile that is a combina-
relative to the mean flow direction. The static tion of instrument noise and true variation in flow
diameter of the sampling area increases with depth velocity. Instrument noise is given as a velocity
to a maximum of 0.93 depth at the bed. The SonTek standard deviation, r:
aDcp employs an internal compass (resolutionF28) to  
define flow direction and a tilt sensor (resolutionF18) r ¼ 140c Fs bN 1=2 ð2Þ
to correct for launch pitch and roll.
Due to beam spread and divergence, the aDcp where b is the bin size and N is the number of
beams can encounter the bed at different depths over measurements over the sampling interval. In the
rough topography, producing contamination of veloc- present study, c is assumed to be 1500 m s1, b is
ity measurements in some of the lowest bins. An set at the minimum bin size for each aDcp and N is the
estimate of mean bed position from SonTek aDcp number of pings. The number of pings is a product of
data, for example, can be determined by a sharp the ping rate, which was set at the maximum rate of 9
increase in signal amplitude averaged over the three pings s1, and the sampling interval. A 5 s sampling
beams. The inflection point above the maximum interval was selected for both SonTek aDcps used in
amplitude represents the transition from the bed to the present research because this was the optimal

Fig. 1. Deployment of a SonTek aDcp from a launch illustrating beam geometry, cell sizes and sampling diameters (not to scale).
28 R. Kostaschuk et al. / Geomorphology 68 (2005) 25–37

interval for operation with the differential global during a single ping, whilst errors associated with the
positioning system (DGPS). This interval provides averaging process result from heterogeneous transport
low velocity error (0.08 m s1) and visual observation within the individual beams. Errors may also be
of individual velocity profiles showed that the 5 s associated with DGPS positioning and the aDcp
profiles did not exhibit excessive scatter. compass and tilt measurements.
Deploying an aDcp from a moving boat also An issue related to sediment-transport in the use
requires velocity corrections for boat motion, either of bottom-tracking for bed load velocity estimation is
from the aDcp bottom-tracking or from a DGPS. the depth of penetration of the acoustic pulse into the
Bottom-tracking utilizes a separate Doppler pulse to mobile bed layer (Villard et al., 2005): since there is
determine boat speed along the vessel track. The a sediment concentration gradient within a mobile
precision of corrections from a DGPS will vary with sand layer, differentiation between near-bed suspen-
the DGPS system used and as a function of location, sion, bed load and the immobile sand bed is difficult.
topography and weather conditions. The bottom-tracking pulse actually measures a
volume, which varies dynamically while the pulse
2.2. Estimating bed load with an aDcp reflects off the bed (Rennie et al., 2002). For the
1500 kHz SonTek aDcp used in this study, a pulse
The bottom tracking capability of an aDcp, which length of 0. 2 m minimises the volume to within
involves measurement of the Doppler shift of an about 0.09 m of the bed (Rennie et al., 2002). Even
independent acoustic echo from the substrate, can be this confined volume includes a significant amount
used to estimate bed load transport (for a more of sand transport in suspension however, since bed
detailed explanation of the theory see Rennie et al., load thickness is usually assumed to occur within 2
2002). However, while bottom tracking is considered median grain heights of the bed (Bagnold, 1973).
more accurate than DGPS for measurements of boat However, Villard et al. (2005) argue that the acoustic
velocity, it produces an absolute correction only over approach should be well-suited to the measurement
hard, immobile substrates. In cases where the bed is of bed load or, at least, near-bed sand transport,
mobile, the bottom tracking Doppler shift is a function because the intensity of the acoustic return from
of both the velocity of the boat and the mobile bed. particulate scatterers is directly proportional to the
The velocity of the mobile bed is therefore related to concentration and size of the individual particles
the difference between the dapparentT boat velocity (Lynch and Agrawal, 1991). This size sensitivity
with respect to the bed measured with aDcp bottom biases the acoustic return toward the larger sediment
tracking and the dactualT boat velocity measured with fractions, which generally also compose the bed load
a DGPS. component. Finally, it is worth noting that high
Rennie et al. (2002) propose a detailed method to suspended sand concentrations could result in the
estimate bed load velocity from dapparentT boat bottom-track pulse dseeingT a false bed and register-
velocity determined from aDcp bottom tracking. Their ing sediment transport velocities that reflect sand
method specifically considers the bed load sampling transport in suspension rather than sand transported
area of the aDcp and explores several sources of error. as bed load.
Potential sources of error include preferential meas- The aDcp bottom-track provides a direct estimate
urement of particular particle sizes and those related to of bed load velocity but not of bed load flux. Bed load
the instrument. For instance, Rennie et al. (2002) velocity u b can be used to estimate bed load transport
noted that reflection of the Doppler pulse is a function q b from:
of both instrument frequency and particle size, and
qb ¼ u db c b ð3Þ
suggested that the 1500 kHz aDcp used in their study
under-represented fine and medium sands. Rennie et where d b is the thickness of the bed load layer, and c b
al. (2002) also identified instrument-related errors due is the bed load concentration, estimated from c b=rq s,
to instrument noise and the averaging process. Instru- where r is volumetric concentration and q s is sedi-
ment noise refers to the inability of the aDcp to ment density. Villard et al. (2005) used two methods
determine the average Doppler shift within a beam to estimate d b and c b for a mobile sand bed. In the first
 R. Kostaschuk et al. / Geomorphology 68 (2005) 25–37 29

method, characteristic values were chosen based on Table 1

previous literature. Bed load thickness was set using Values of sound absorption for fresh water (a) and sensitivity to
particle size for SonTek aDcps (based on data from SonTek, 1997)
the definition of bed load being a thickness of 2
ADcp a (dB/m) Particle radius for Minimum detectable
median grain heights above the bed (Bagnold, 1973).
frequency peak sensitivity particle radius (Am)
Bed load concentration was set using characteristic (kHz) (Am)
values of r=0.4 and q s=2650 kg m3. The second
3000 2.4 80 4
method is that of van Rijn (1984), where d b and c b are 1500 0.6 160 8
calculated as a function of dimensionless particle and 750 0.15 320 15
transport stage parameters (details of the procedure 500 0.067 480 25
can be found in van Rijn, 1984 and Villard et al., 250 0.017 960 50
2005). Field examples for 500 and 1500 kHz systems are provided in the
present paper.
2.3. Estimating suspended load with an aDcp

An aDcp can also be used to provide information Signal strength decreases with range into the water
on the type and quantity of particulate matter that column due to geometric spreading and absorption of
generates the Doppler shift. The strength of an aDcp acoustic energy, as expressed by:
acoustic return is a function of the instrument and the
SSD ¼  20logð R=cos/Þ  2að R=cos/Þ ð5Þ
flow properties, and has been used to estimate the
concentration of suspended sediment in a wide range where SSD is signal strength decay or loss (dB) at
of flows (e.g., Reichel and Nachtnebel, 1994; Guyot distance from the transducer R, / is the transducer
et al., 1998; Reichel, 1998; Alvarez and Jones, 2002). beam angle.
Instrument response is a function of frequency, Acoustic frequencies have different sensitivities
transmit power, receiver sensitivity and distance to dependent on particle size, and, at a single frequency,
the measurement volume, whilst the size, type and backscatter strength is a function of particle size, type
concentration of suspended sediment are the most and concentration. For a constant type and size of
important flow properties. particle, scattering strength is thus directly propor-
The sonar equation (SonTek, 1997) illustrates how tional to concentration. Each acoustic frequency has a
aDcp factors affect signal strength: particle size for which it possesses peak sensitivity
and a minimum detectable particle radius (Table 1),
EL ¼ SL þ 10logPL  2  logR  2aR þ Sv þ RS with sensitivity declining quickly for smaller particles,
ð4Þ particularly for lower frequency aDcps. Peak sensi-
tivity occurs where the particle circumference is equal
where is EL the decho levelT or signal strength, SL is to the acoustic wavelength (SonTek, 1997). It is
the dsource levelT or transmitted acoustic power, PL is important to note that, in practise, it is not possible to
d pulse lengthT or length of the acoustic pulse, R is the distinguish between the effects of suspended sediment
range between transducer and measurement volume concentration and particle size on the acoustic back-
(m), a is the sound absorption coefficient (dB m1) scatter (Reichel, 1998), and calibration with known
(Table 1), S v is volume scattering strength (dB), and values of concentration is therefore required for
RS is dreceive sensitivityT which expresses the absolute aDcp-based estimates of concentration.
relationship between pressure at the transducer face However, a particular calibration is only valid if the
and the measured signal strength. Signal strength data size distribution does not change.
in SonTek aDcps, for example, are given as amplitude
(counts), which are converted to dB through multi-
plication by 0.43. The most important variables in the 3. Examples of field applications
sonar equation are the range and pulse length, and in
order to compare measurements from different cells, Examples of applications are provided from recent
the signal strength must be normalized for range. work in the Fraser River estuary and Lillooet Lake,
30 R. Kostaschuk et al. / Geomorphology 68 (2005) 25–37

British Columbia, Canada, where SonTek 1500 and field velocity profiles of natural turbidity currents.
500 kHz aDcps were deployed. Measurements in the Preliminary analyses of the data showed that under-
Fraser River used the 1500 kHz model with a Trimble flows descend along a distinct plunge line and form
DGPS tied to a navigation beacon, allowing b1 m well-defined currents as they progress lakeward. In
accuracy in position. A Leica real-time kinematic some cases, particularly during low river flow, the
DGPS was used with the 500 kHz system in Lillooet underflows transformed into interflows a short
Lake to provide position accuracies of b1 cm. distance offshore. During high river flows the
currents continue as underflows for many kilometers
3.1. Field sites offshore.

The Fraser River flows through several distribu- 3.2. Velocity

tary channels on the Fraser Delta into the Strait of
Georgia, a mesotidal marine basin, about 20 km Fig. 2 shows horizontal and vertical flow velocities
south of Vancouver (Kostaschuk et al., 1998). The measured with the SonTek 1500 kHz aDcp at low tide
1500 kHz aDcp was deployed in this study in two in the Main Channel of the Fraser Estuary, with an
estuarine distributary channels—the Main Channel averaging time of 5 s. Horizontal velocities are highest
and Canoe Pass (see Kostaschuk et al., 1998, for over the dune crest and lowest over the troughs and
details). The Main Channel is up to 1 km wide and vertical velocities are positive (directed towards the
20 m deep, captures about 84% of the total river surface) above the upstream-facing stoss side and
discharge and has a bed of primarily medium sand crest areas and negative (directed towards the bed)
(McLaren and Ren, 1995). A salt-wedge intrusion over the steeper, downstream-facing leeside and
actively migrates along the channel and large, low- trough. The vertical velocity record is dcoarserT than
angle, symmetric dunes up to 4 m in height and over the horizontal velocity record, however, and some
100 m in length develop during high river flows in vertical velocities are quite high and there are several
June (Kostaschuk et al., 1989). Canoe Pass is a rapid transitions in direction (up–down) in adjacent
smaller distributary that branches from the Main bins. This suggests that care should be taken in
Channel and discharges onto Roberts Bank in the interpreting localised vertical velocities from aDcp
Strait of Georgia. The channel is approximately 350 data, especially in profiles from a moving boat that
m wide and the deepest sections are less than 10 m may not capture a true temporal average.
deep at high tide, with the bed sediment comprising In contrast to the localised variations noted above,
medium to coarse sand (Villard, 1995). While the zones of high and low horizontal velocity and positive
flows are tidally influenced, there is little evidence of and negative vertical velocity clearly extend to the
salt-wedge intrusion during the freshet and flows are water surface over the dunes. The velocity patterns
consistently seaward (Villard, 1995). Large, steep, over these dunes reflect topographic forcing by the
asymmetric dunes reach heights of 3 m and wave- dunes and are similar to patterns described in many
lengths of 50 m (Kostaschuk, 2000). other flume and field studies. In particular, the fixed
Lillooet Lake is situated in a glacial valley in the dune maps from flume studies by Bennett and Best
Coast Mountains about 300 km north of Vancouver, (1995) and Best and Kostaschuk (2002) show nearly
and is ~30 km in length and 3 km in width. Lillooet identical regions of high and low horizontal velocity
River has deposited a rapidly prograding delta in the and positive and negative vertical velocity. These
northern arm of the lake that is characterized by regions do not extend to the flow surface in the flume
steep foreset beds on the upper slope that pass studies, however, probably due to sidewall effects in
offshore into gently sloping bottomsets (Gilbert, laboratory studies (Best and Kostaschuk, 2002). The
1975). Lillooet River drains a glacierized catchment normalized contour plots of Kostaschuk and Villard
and cold, sediment-laden water enters Lillooet Lake (1999), that were based on electromagnetic current
as density underflows and interflows. The measure- meter data ensembled from several Fraser dunes,
ments obtained in Lillooet Lake in 2001 and 2002 confirm the patterns of horizontal velocity measured
with a 500 kHz aDcp are the first-ever whole flow by the aDcp.
 R. Kostaschuk et al. / Geomorphology 68 (2005) 25–37 31

Fig. 2. Horizontal (U) and vertical (V) velocities measured with the 1500 kHz aDcp deployed from a moving launch in a dune field in the Main
Channel of the Fraser Estuary on June 15 1999. Mean reach depth was 12.6 m and mean velocity was 2.36 m s1. Mean dune height was 1.92
m, mean length was 56 m, and mean lower lee slope was 118. Flow is from right to left. The black zones near the bed are contaminated data.

3.3. Bed load horizontal and vertical velocity were calculated as

the average of the two lowest uncontaminated bins.
Fig. 3 shows an at-a-point time series of horizon- Bed load velocity is poorly correlated with mean
tal flow velocity and bed load velocity (averaging horizontal velocity (R=0.08) but better correlated with
time=5 s) measured with a SonTek 1500 kHz aDcp near-bed horizontal velocity (R=0.20). Similarly,
from a launch anchored over the crest of a large dune correlations are slightly better for near-bed vertical
in Canoe Pass. Periods of poor bottom-tracking velocity (R=0.23) compared to mean vertical velocity
return were present during this, and all other, time (R=0.19). Although these preliminary results are
series, probably due to high suspended sand concen- intuitively as expected (e.g., Best and Kostaschuk,
trations, and resulted in a fragmented record of bed 2002), with near-bed vertical velocity driving bed load
load velocity. The bottom track record was more transport, all of the correlations are statistically
continuous at lower flow velocities, however, and the insignificant at the 95% level and it is thus not
highest bed load velocities (Fig. 3) are likely biased possible to make definitive process links. Indeed, such
by sand transport in suspension, rather than reflecting links may be difficult to establish because of the
true dbed loadT. probable bias of bottom track measurements from
Correlation analysis (exponential model) was used sand in suspension, plus the broad near-bed sampling
to determine if the aDcp data of Fig. 3 could be used diameter of aDcps which makes it difficult to quantify
to examine relations between bed load and fluid processes that may be acting on a smaller spatial scale.
velocities. Fluid velocity was calculated as the mean More research is clearly required to determine if aDcp
depth-averaged horizontal and vertical velocity of data can be used to examine bed load dynamics and
each velocity profile, whilst the mean near-bed processes in geophysical flows.
32 R. Kostaschuk et al. / Geomorphology 68 (2005) 25–37

Fig. 3. Time series of horizontal velocity (U) and bed load velocity measured with the 1500 kHz aDcp from a launch anchored over a dune crest
in Canoe Pass on June 3, 2000.

Measurements of bed load velocity using a Villard et al. (2005) also compared estimates of bed
SonTek 1500 kHz aDcp have been compared with load transport from aDcp bed load velocity measure-
measurements using traditional bed load samplers in ments and Eq. (3) with traditional Helley–Smith bed
sand and gravel reaches of the Fraser River and load samples in Canoe Pass. Villard et al. (2005)
Estuary (Rennie et al., 2002; Rennie and Villard, found good agreement (R=0.69) between Helley–
2003; Rennie and Millar, 2004; Villard et al., 2005) Smith data and aDcp estimates indicating both
demonstrated for gravel-beds that bed load velocity instruments are measuring similar fractions of near-
was significantly correlated with sampler measure- bed transport. However, Villard et al. (2005) also note
ments in all cases (0.62bRb0.94), although the that it is difficult to assess the effect of near-bed
results were influenced by particle size. Rennie et suspension on bottom tracking results.
al. (2002) also concluded that, because of the
intermittent nature of gravel transport, 5 s aDcp 3.4. Suspended load
averages of bed load velocity were extremely
variable and a sampling interval of around 25 A SonTek 500 kHz aDcp was deployed in
min was required for reliable results. Lillooet Lake on August 22nd 2001 during a
 R. Kostaschuk et al. / Geomorphology 68 (2005) 25–37 33

severe rainstorm that produced a significant flood in travelling lakeward as an interflow. The plots also
the Lillooet River and an associated turbidity shows a strong return flow in the upper water
current in the lake. Fig. 4 shows a dsnapshotT of column near the delta, likely generated by compen-
horizontal velocity and signal amplitude (EL) taken sation flow of lake water caused by displacement of
from a moving launch running from near the delta fluid by the density current, that is directed back
to offshore. Signal amplitude decreases from the towards the delta front. Although there is some
surface with depth due to beam attenuation, but agreement between the amplitude and velocity
increases abruptly in the turbidity current due to records, especially in the higher amplitude present
strong reflectivity of the suspended sediment. The in the basal density current, the poorer agreement
amplitude record indicates a continuous, near-bed within the water column probably reflects the fact
dcloudT of suspended sediment from the river mouth that this transect depicts just a snapshot of the
to about 180 m offshore. The cloud has lifted from velocity, whereas the sediment concentration may
the bed at around 200 m and is continuous at a record a more general stratification of sediment
depth of ~35 m to beyond 300 m offshore. within the water column. The velocity record (Fig.
The record of flow velocity for this transect (Fig. 4) shows periods of high and low velocity fluid
4) is more dpatchyT than the amplitude record. Flow captured from the moving launch, whereas the more
direction records (not shown) show that river water continuous amplitude record reflects sediment
extends from the lake surface to the bed to about within these periodic events, together with sus-
30 m offshore, where it plunges beneath the pended sediment entrained into, and settling within,
ambient lake water and continues offshore as a the ambient flow.
strong underflow to about 200 m: this region of Fig. 5 illustrates profiles of aDcp backscatter and
higher velocity flow near the bed can be clearly suspended sediment concentration and median grain
seen in Fig. 4. from the delta front to ~180 m size D 50 measured from at-a-point pump samples
offshore. The velocity record at 200–300 m off- taken in 21 m water depth at 1 m intervals during
shore shows patches of higher velocity fluid the August 22 event. Grain size was determined

Fig. 4. Fig. 2. Horizontal velocity (U) and signal amplitude measured with the 500 kHz aDcp deployed from a moving launch in Lillooet Lake
on August 22, 2001.
34 R. Kostaschuk et al. / Geomorphology 68 (2005) 25–37

Fig. 5. Profiles of aDcp backscatter (B) measured with the 500 kHz aDcp and suspended sediment concentration (C) and median grain size D 50
measured from pump samples taken on August 22, 2001 in Lillooet Lake.

with a Malvern particle-size analyser. Concentration, that all D 50 values are b30 Am, which are much
backscatter and D 50 all decrease with height above smaller than 960 Am diameter for peak sensitivity
bed in a similar fashion. It is interesting to note for a 500 kHz aDcp (Table 1), and often smaller

Fig. 6. Preliminary linear regression calibrations for aDcp backscatter (B) and suspended sediment concentration (C) based on data in Fig. 5.
Samples labelled 1–10 m refer to height above the bed. See text for discussion.
 R. Kostaschuk et al. / Geomorphology 68 (2005) 25–37 35

than the 25 Am predicted minimum detectable provides velocity and suspended sediment profiles
radius (Table 1). All of the samples are composed that can be used to describe whole depth flow and
of sediment at least 10% N25 Am however and suspension structures and patterns. These measure-
samples within the turbidity current are at least ments are simply not possible with single-point
20% N25 Am. Although these particle sizes were current meters and samplers that require continuous
not ideal for 500 kHz aDcp backscatter measure- re-positioning of the instruments at different ele-
ments, the results in Figs. 4 and 5 clearly show that vations above the bed and re-anchoring the boat at
there are sufficient concentrations of sizes larger different locations to capture spatial variations in
than the minimum detectable particle. flow.
Calibration with measured sediment concentration The most serious limitation of an aDcp is in
is required to allow direct prediction of suspended obtaining measurements near the bed. Beam geom-
sediment concentration from aDcp backscatter and etry results in a large sampling diameter close to
several empirical models have been used, including the bed that does not adequately capture variations
linear and exponential (e.g., Alvarez and Jones, in velocity, particularly vertical velocity, or sedi-
2002). Fig. 6 shows linear regression models for ment transport over complex topography. For
the data of Fig. 5 and the regression results are better example, this problem is particularly acute in the
for the linear model (R 2=0.90) compared to the lee of sand dunes where the aDcp beams may
exponential model (R 2=0.79) for all of the data encounter the bed at different depths, resulting in
(n=23). However, the relation between backscatter contamination of measurements and loss of data in
and concentration is very different for samples taken this region. Single-point current meters and sedi-
within the turbidity current (BN70 dB in Fig. 6; ment samplers will probably provide more reliable
n=10) compared to those from the flow above the measurements close to the bed and on lee sides,
current (Bb70 dB in Fig. 6), probably due to although accurate positioning in the lee can be
particle-size effects (see grain size changes in Fig. difficult, if not impossible.
5 above ~11 m). The regression is also stronger Although aDcps show potential for estimating bed
(R 2=0.93) for the samples within the turbidity load and suspended load transport both quickly and
current (labelled 1–10 m in Fig. 6) compared to accurately, several key issues remain to be explored.
the regression for all of the data. These regression These include testing aDcps over a wider range of bed
models must be considered preliminary, however, material and flow conditions for bed load, a more
and more data are required for reliable calibration thorough examination of the effect of near-bed
curves. suspension on bottom tracking, a better understanding
of instrument errors, and fuller testing of the
sensitivity of aDcp backscatter to suspended sediment
4. Discussion size and type under a variety of field conditions.

There are three significant advantages in using

an aDcp, compared to traditional single-point Acknowledgements
current meters and sediment samplers, to measure
flow velocity and sediment transport in rivers and Thanks to Mike Church for his ongoing support
lakes: (1) Firstly, an aDcp is a single, non-intrusive and advice and to Arjoon Ramnarine for piloting the
instrument that can be used for simultaneous flow launch. Financial support was provided through a
and sediment transport measurements. (2) Secondly, NATO Collaborative Research Grant to Kostaschuk
an aDcp relies on acoustic profiles and positioning and Best, a Natural Environment Research Council
from a DGPS so it can be used from a moving (UK) Grant to Best, Peakall and Kostaschuk (NER/B/
launch. This eliminates the need to anchor a vessel S/2000/00189), Natural Sciences and Engineering
in often dangerous navigation channels, a procedure Research Council of Canada (NSERC) Grants to
that is complicated by sensor and support cables or Kostaschuk and Church and an NSERC Postdoctoral
positioning of bed frames. (3) Thirdly, an aDcp Fellowship to Villard.
36 R. Kostaschuk et al. / Geomorphology 68 (2005) 25–37

References estuary: Fraser River, British Columbia. Can. J. Earth Sci. 26,
1440 – 1452.
Admiraal, D., Demissie, M., 1996. Velocity and discharge measure- Kostaschuk, R.A., Luternauer, J.L., Church, M.A., 1998. Sedimen-
ments at selected locations on the Mississippi River during the tary processes in the estuary. In: Clague, J.J., Luternauer, J.L.,
great flood of 1993 using an acoustic Doppler current profiler. Mosher, D.C. (Eds.), Geology and Natural Hazards of the Fraser
Water Int. 21 (3), 144 – 151. River Delta, British Columbia, Geological Survey of Canada
Agrawal, Y.C., Pottsmith, H.C., 2000. Instruments for particle size Bulletin, vol. 525, pp. 41 – 56.
and settling velocity observations in sediment transport. Mar. Kostaschuk, R.A., Villard, P.V., Best, J.L., 2004. Measuring
Geol. 168, 89 – 114. velocity and shear stress over dunes with an Acoustic Doppler
Alvarez, L.G., Jones, S.E., 2002. Factors influencing suspended Profiler. J. Hydraul. Eng. 130, 932–936.
sediment flux in the Upper Gulf of California. Estuar. Coast. Lane, A., Prandle, D., Harrison, A.J., Jones, P.D., Jarvis, C.J., 1997.
Shelf Sci. 54, 747 – 759. Measuring fluxes in tidal estuaries: sensitivity to instrumenta-
Bagnold, R.A., 1973. The nature of saltation and bed-load tion and associated data analyses. Estuar. Coast. Shelf Sci. 45,
transport in water. Proc. R. Soc. Lond., A Math. Phys. Sci. 433 – 451.
332, 473 – 504. Lynch, J.F., Agrawal, Y.C., 1991. A model-dependent method for
Bennett, S.J., Best, J.L., 1995. Mean flow and turbulence structure inverting vertical profiles of scattering to obtain particle size
over fixed, two-dimensional dunes: implications for sediment spectra in boundary layers. Mar. Geol. 99, 387 – 401.
transport and bedform stability. Sedimentology 42, 491 – 513. McLaren, P., Ren, P., 1995. Sediment transport and its environ-
Best, J.L., Kostaschuk, R.A., 2002. An experimental study of mental implications in the lower Fraser River and Fraser Delta.
turbulent flow over a low-angle dune. J. Geophys. Res. 107, Canada Department of Environment, Environmental Conserva-
3135 – 3153. tion, DOE FCAP 1995-3.
Best, J.L., Kostaschuk, R.A., Villard, P.V., 2001. Quantitative Reichel, G., 1998. Suspended sediment monitoring: use of acoustic
visualization of flow fields associated with alluvial sand dunes: Doppler current profiler. In: Herschy, R.W, Fairbridge, R.W.
results from the laboratory and field using ultrasonic and (Eds.), Encyclopedia of Hydrology and Water Resources.
acoustic Doppler anemometry. J. Vis. 4, 373 – 381. Kluwer Academic Publishers, Dordrecht, The Netherlands,
Downing, J.P., 1983. An optical instrument for monitoring pp. 638 – 644.
suspended particulates in ocean and laboratory. Proceedings of Reichel, G., Nachtnebel, H.P., 1994. Suspended sediment monitor-
Oceans ’83. ASCE, San Francisco, CA, pp. 199 – 202. ing in a fluvial environment: advantages and limitations
Gilbert, R., 1975. Sedimentation in Lillooet Lake, British Columbia. applying an acoustic Doppler current profiler. Water Res. 28
Can. J. Earth Sci. 12 (10), 1697 – 1711. (4), 751 – 761.
Guyot, J.L., Filizola, N., Guimaraes, V., 1998. Amazon Rennie, C.D., Millar, R.G., 2004. Measurement of bedload transport
suspended yield measurements using an acoustic Doppler velocity spatial distribution. Earth Surface Processes and Land-
current profiler (ADCP): first results. Hydrology in the Humid forms 29, 1173–1193.
Tropic Environment, IAHS Publication vol. 253. International Rennie, C.D., Villard, P.V., 2003. Bedload measurement in both
Association of Hydrological Sciences, Wallingford, Kingston, sand and gravel using an aDcp. 16th Canadian Hydrotechnical
Jamaica, pp. 109 – 115. Conference October 22–24, 2003. Canadian Society for Civil
Hay, A.E., 1983. On the remote acoustic detection of suspended Engineers, Burlington, ON. 10 pp.
sediment at long wavelengths. J. Geophys. Res. 88 (C12), Rennie, C.D., Millar, R.G., Church, M.A., 2002. Measurement of
7525 – 7542. bedload velocity using an acoustic Doppler current profiler.
Helley, E.J., Smith W., 1971. Development and calibration of a J. Hydraul. Eng. 128, 473 – 483.
pressure difference bedload sampler. USGS Water Resources Ryan, S.E., Porth, L.S., 1999. A field comparison of three
Division Open-file Report. 18 pp. pressure-difference bedload samplers. Geomorphology 30,
Klingeman, P.C., Milhous, R.T., 1971. Oak creek vortex bed-load 307 – 322.
sampler. EOS, Trans.-Am. Geophys. Union 52 (5), 434. Smith, J.D., McLean, S.R., 1977. Spatially-averaged flow over a
Kostaschuk, R.A., 2000. A field study of turbulence and sediment wavy surface. J. Geophys. Res. 82, 1735 – 1746.
dynamics over subaqueous dunes with flow separation. Sed- SonTek, 1997. SonTek Doppler current meters—using signal
imentology 47, 519 – 531. strength to monitor suspended sediment concentration. SonTek
Kostaschuk, R.A., Villard, P.V., 1996. Flow and sediment transport Application Notes, San Diego. 7pp.
over large subaqueous dunes: Fraser River, Canada. Sedimen- Thorne, P.D., Vincent, C.E., Hardcastle, P.J., Rehman, S.,
tology 43, 849 – 863. Pearson, N., 1991. Measuring suspended sediment con-
Kostaschuk, R.A., Villard, P.V., 1999. Turbulent sand suspension centrations using acoustic backscatter devices. Mar. Geol. 98,
over dunes. In: Smith, N.D., Rogers, J. (Eds.), Fluvial 7 – 16.
Sedimentology VI. Special Publication Number 28 of the van Rijn, L.C., 1984. Sediment transport: Part I–III. J. Hydraul.
International Association of Sedimentologists. Blackwell Sci- Eng. 110, 1431 – 1754.
ence, Oxford, pp. 3 – 14. Villard, P.V., 1995. Spatial and Temporal Behaviour of Turbulent
Kostaschuk, R.A., Church, M.A., Luternauer, J.L., 1989. Bed- Sand Suspension over Dunes: Fraser River, British Columbia.
forms, bed material and bedload transport in a salt-wedge Unpublished MSc thesis, University of Guelph.
 R. Kostaschuk et al. / Geomorphology 68 (2005) 25–37 37

Villard, P.V., Church, M.A., Kostaschuk, R.A., 2005. Estimating environments: examples from the Tay Estuary, Scotland.
bed load in sand-bed channels using bottom tracking from an Geomorphology 29 (1), 21 – 31.
acoustic Doppler profiler. In: Blum, M.D., Marriott, S.B. (Eds.), Zedel, L., Hay, A.E., 1999. A coherent Doppler profiler for high-
Fluvial sedimentology VII. International Association of Sed- resolution particle velocimetry in the ocean: laboratory measure-
imentologists, Blackwell. In press. ments for turbulence and particle flux. J. Atmos. Ocean.
Wewetzer, S., Duck, R.W., Anderson, J.M., 1999. Acoustic Technol. 16, 1102 – 1117.
Doppler current profiler measurements in coastal and estuarine