- Email: [email protected]
A robust in situ settling velocity box for coastal seas
101
Journal of Sea Research 36 (1/2): 101-107 (1996)
A R O B U S T I N S I T U S E T T L I N G V E L O C I T Y BOX F O R C O A S T A L S E A S
P.B. MURRAY 1, I.N. McCAVE 1, T.R.E. OWEN 1, M. MASON 1 and M.O. GREEN 2 1Bullard Laboratories, Department of Earth Sciences, University of Cambridge, Cambridge, CB3 0EZ, UK 2NIWA Ecosystems, PO Box 11-115, Hamilton, New Zealand
ABSTRACT A new instrument for measuring the in situ settling velocity able to withstand the harsh conditions common in coastal waters is introduced. This settling box traps a parcel of fluid and measures the decaying concentration with an array of optical backscatter probes; output from these four sensors is then combined to produce a single settling velocity distribution. Preliminary deployments have been undertaken in the river Elbe and have produced encouraging results. Once the box is closed, an initial period of turbulent fluctuations is followed by settling and a smooth decrease in concentration. Settling velocity distributions recorded at transitions from flood/ebb tides are distinct from those at ebb/ flood transitions; median settling velocity was found to increase with sediment concentration.
Keywords: Settling velocity, cohesive, flocculation, coastal seas, optical backscatter, particle size, calibration
1. INTRODUCTION The settling velocity of suspended particles is an important parameter necessary for modelling the suspended sediment distribution in coastal waters. Samples of sediment may be resuspended in the laboratory and the settling velocity recorded. This technique is suitable for sand sized particles, but not for finer material. Silts and clays are cohesive and the individual particles bind together in larger structures or flocs. Flocculation occurs in response to the physical, chemical and biological conditions at a particular time in an environment. Sampling the material destroys the delicate floc structure. Accurate determination of the settling velocity thus requires in situ measurements. The earliest successful device for measuring the in situ settling velocity was the Owen tube (Owen, 1976). This takes a horizontal sample of fluid, which is brought to the surface and rotated 90 degrees, whereupon the tube acts as a settling column and samples are withdrawn at timed intervals. Subsequent instrumentation has followed two principles. Firstly, tracking of individual flocs in a small volume of fluid, using film or photography (Fennessy & Dyer, 1994; Van Leussen & Cornelisse, 1993). Such instruments provide relatively precise measurements, but sample only a small volume of fluid, appear to bias
towards the largest flocs which the eye is drawn towards, and are delicate, rendering them difficult to operate outside estuaries. We decided to follow the alternative path of automating the Owen tube with optical sensing and building an instrument that would be confidently employed in coastal waters as well as estuaries. A robust settling box was designed that uses an array of sensors for measuring backscattered infrared light. This is related to sediment concentration by calibration experiments. The design follows similar instruments used by Kineke et al. (1989) and McCave & Gross (1991), and will commonly be attached to our bottom mounted instrument frame, the Quadrapod. This frame is deployed at sea for up to 60 days during which time, the settling box will close for pre-programmed durations, recording the settling backscatter almost continuously. Deployment in locations inaccessible to other instruments and remote operation, often during shelf storms, will enhance our knowledge considerably. 2. THE SETTLING VELOCITY BOX The cylindrical settling box (Fig. la) (30 cm long and 25 cm diameter) traps a parcel of fluid between two end plates or lids and then measures the decaying backscatter by an array of Miniature Optical Back-
102
P.B. MURRAY, I.N. MCCAVE, T.R.E. OWEN, M. MASON & M.O. GREEN
b
TOP
SAMF>LE
,
I~/FRA -- RED BAEI'(SCATTEI
I I?
N II
1
I
'"°
....
BOTTOM SEALING
,.
LID
(L =i G N i ' =
~¢,
Fig. 1. (a) the Settling Velocity Box and (b) a Miniature Optical Backscatter Sensor (MOBS). scatter Sensors. The device is physically strong, with collars to reinforce the cylinder ends and a side plate between the two collars; the plate is then available for attaching the mechanics. Mechanical strength comes from having a simple method of performing the single mechanical operation, lid closure. An electric screwdriver motor winds a crank along a threaded rod, thereby moving the arms that simultaneously close the top and bottom lids. The lids and cylinder ends are tapered, with a neoprene lining on the lids ensuring a good seal, important to avoid contamination of the sample by the external flow. The motor-driven lids close, compressing the neoprene lining until no further movement is possible, whereupon the current overloads and the motor stalls. Closure is slow (approximately 10 seconds) in order to minimize turbulence generation. The lids open to an angle of 45 ° and in the opposite direction to one another so as to maximize flushing, by either waves or a tidal current. A 15 dm 3 parcei of fluid is trapped by the box. Sediment particles fall out of suspension, and the decaying backscatter is measured by four MOBS positioned at 25, 50, 100 and 200 mm from the top of the cylinder. The MOBS are flush mounted so as not to interfere with the settling, and can be readily removed for calibration.
The settling box was designed to be mounted on and driven by the Quadrapod. However, the box can also be operated autonomously in a dedicated frame which is then supported by a cable. In this configuration, the MOBS electronics is located beside the box, an electronic cable brings power from the surface and another allows box control and returns data on-line. 3. SEDIMENT CONCENTRATION SENSORS Optical backscatter instruments have been used successfully to measure suspended sediments in a variety of environments (Downing et al., 1981). Sensors are small, particularly important when attempting to resolve the steep concentration gradient immediately above the seabed, and can measure high concentrations at frequencies sufficient to resolve turbulent as well as wave fluctuations. Our experiments demanded an instrument that could be deployed at sea for typically 60 days, with a consistent response over a wide range of sediment types and concentrations. Several sensors should be capable of operating alongside one another. Optimizing the considerations, led to the development of a new Miniature ~ t i c a l Backscatter Sensor (the MOBS) at the Bullard Laboratories, University of
IN SITU SETTLING VELOCITY BOX
103
mm and 50 mm below the top lid) overlap and that the highest beam 'sees' the surface lid causing the beam to be asymmetric (Smith, 1994). A second generation of MOBS have been designed with a reduced beam width thereby ensuring that all sensors measure a unique and symmetrical volume of fluid. 4. CALIBRATION
) PMDPELL~R
c~,l=
Fig. 2. The turbidity tank.
Cambridge. This versatile sensor (Fig. l b) uses an infrared light emitting diode with a spectral peak at 850 nm, to illuminate a volume of fluid immediately in front of the sensor, and a matched photo diode to measure the amount of light backscattered by the suspended particles. Emitter and receiver are separated behind identical lenses, and are inclined at about 20" to one another. This separation also enables automatic compensation for fouling on the windows. The geometry confines the vertical spreading of the beam ensuring a deeper measurement field even in high concentrations, and enhances the sensitivity (dynamic range is 1:20000). The emitter operates in a pulsed mode to maximize the signal to noise ratio, with automatic gain subtraction of background light and detector offset to minimize temperature effects. Pulsed emission further allows up to 16 sensors to operate in close proximity without problems of interference. The MOBS are small (35 mm x 17 mm) and sample at 1Hz. They are made in strings of four, and operate cyclically to avoid interference. Modelling of the vertical beam widths of the sensors when located in the settling box has demonstrated that in clear water, the top two beams (at 25
The in situ measurements of backscatter due to suspended material that are measured by the MOBS must be converted into sediment concentration. A calibration is necessary at each site using representative material, because backscatter is a function of particle size, shape and density (as well as concentration). Previous calibrations have been undertaken in the turbidity tank shown in Fig. 2. Material from the seabed is added to the tank in a known quantity. The propeller mixes the material, generating a homogeneous suspension, which is then recorded by the MOBS. The process is repeated for a variety of concentrations. Such a simple method has successfully been employed for calibrating MOBS that are used to record tidal and wave driven suspended sediment fluctuations. In the settling box, however, as the concentration decreases with time, the particle size also diminishes because the larger particles settle first. Small grains are more effective at scattering light. A calibration based on the bulk suspended material will tend to overestimate concentrations in the tail of the curve where the particles are small. A time-dependent calibration that makes allowance for the decreasing particle size is required. A standard technique for deriving the weight frequency settling velocity distribution of sediments comprising silts and muds is pipette analysis (see Galehouse, 1971 for a full account). A dispersed, homogeneous, suspension of known concentration is allowed to settle, while samples are withdrawn by pipette at a set level below the surface (settling distance = h) and at particular times (settling time = t). The time taken to fall the distance gives the settling velocity, Ws=h/to Samples are dried and weighed and the decrease in concentration (as a proportion of the starting concentration) gives the weight frequency for that range in settling velocities. By performing several pipette experiments with different starting concentrations and simultaneously measuring the MOBS voltage, it is possible to draw plots of the concentration against the reciprocal of settling velocity (used instead of time because it accounts for the differing fall distances) and superimpose lines of equal voltage on the graph (see Fig. 3). If particle size has no effect then voltage lines will be parallel to the x-axis and there will be a unique concentration for each voltage. Alternatively, the voltage lines will slope away from the x-axis with time, and there will be a range of concentrations for each voltage dependent on how long into the settling the
104
P.B. MURRAY, I.N. MCCAVE, T.R.E. OWEN, M. MASON & M.O. GREEN
400
..... i
i
~
.....
T
.............
]
/
100
I
0
I ....
10
~
20
'"'~"
30
,
I
40
50
1/Ws (1/(mm/s)) Fig. 3. The time-varying calibration scheme. Decay curves (full lines) are drawn through the measured concentrations (crosses) of four settling experiments.Simultaneous records of MOBS output enable the addition of voltage contours (dotted lines). measurement is made, and therefore the size of particles remaining in suspension. The turbidity tank was used for settling the mud (Fig. 2); the lid was not used for fear of disrupting the near-surface settling. A pipette analysis was undertaken, modified to extract the greatest accuracy from the facilities available. The first modification was to extract the samples and position the MOBS as low in the tank as possible. By having a long settling distance, the settling time is expanded. This helps to minimize the relative importance of turbulence that takes several tens of seconds to decay after the pump has stopped. Also, a slower concentration decay is better for sampling than a steep decay because of the inaccuracies in averaging over a rapidly changing variable. The second modification was to pump the withdrawal of samples through a glass tube, and not use a pipette. Pumping the samples ensures a constant and controllable extraction rate. The glass tube is in residence throughout the experiment, avoiding disruption of settling by insertion or extraction of a pipette. Finally, the large volume of the tank enables large samples to be extracted without significant alteration of the water level and thus the settling distance. The starting concentration (pump on) was sampled and then during settling, at 3, 6, 12, 24 min (and further times increasing by a multiple of 2 up to 12 h) after stopping the pump. Settling velocities between 491 mm.s 1 (600 mm per 1.223s) and 0.014 mm.s -1 (500 mm per 12h) were recorded. Samples of 250 cm 3 were extracted at the first three settling times, and 500 cm 3 at later times.
ments that make in situ measurements of settling velocity and particle size. The exercise was conducted in the mouth of the river Elbe, where the University of Hamburg have a pontoon moored in about 20 m of water, near to Brunsbfittel. This facility was made available for the experiment, together with a GKSS research vessel, RV 'Ludwig Prandtl'. Participants from Europe and North America, comprising fifteen groups were divided between the pontoon and the ship. Measurements were made over four days, between 0800 and 2000 hrs for three days and 0800 and 1200 hrs on the final day. The ideal measurement strategy was to take measurements at 9 m depth and every hour on the hour. The settling box was suspended from a cable and lowered by winch into the water. A transmissometer was attached to the frame of the box in order to monitor the external sediment concentration. Data from both transmissometer and settling box were transmitted to the surface and displayed on a computer in real-time. A lead weight was attached to the underside of the box in an attempt to keep it vertical in the fast flowing tidal current that exceeded 1.5 m.s"1. This proved insufficient, and the box was pulled from the vertical for much of the tide. It was decided to deploy only when the wire angle was less than 10"; this limited data gathering to about 2 h either side of slack water. Nine settling periods were measured, all close to times of slack water. Four captures were taken of the ebb/flood transition and five of the flood/ebb. A typical closure is shown in Fig. 4. Each MOBS exhibits initial rapid fluctuations in backscatter, evidence of trapped turbulence. This dies out after about three minutes and gives way to a fairly smooth decrease in backscatter with time. The uppermost sensor yields the most variable profile because this is the region of lowest concentration and therefore greatest sensitivity to perturbation. Fortunately the most important sensor early in the experiment is the lowest one. A jump indi-
_•
w ...... T ............ ~
A trial deployment of the settling box was undertaken as part of an intercalibration programme of instru-
7
r ]
+
'
I
+
i
m
.
[
5. THE ELBE EXPERIMENT
i
800
0
...... 7-----
1200
2400
I :3600
- - - 1 4800
6000
Time after box closure (s) Fig. 4. Output from the four MOBS during a box closure.
IN SITU SE'r-rLING VELOCITY BOX
I
1000
400
~' ;
I
'
'
'
Y. 1(~0
'
'
'
r= = 0.935
200
300
Sediment concentration (mg/I) Fig. 5. Calibration curve for MOBS 0. cating decreased backscatter evident in all sensors after about 4000 and 4500 seconds probably indicates tidal movement of the box. Earlier variations may be naturally occurring, for instance by differential settling whereby particles that are falling at different rates coalesce. The sensor that is located highest in the box (25 mm from the surface lid) shows the steepest rate of decay in backscatter; decay rate decreases for each lower sensor as the settling distance increases. The coarse grains pass the highest sensor rapidly and their progression is recorded quickly without resolving much detail. Each lower sensor doubles the settling distance, effectively expanding the x-axis by 2 and loosing the very slow-settling grains. For example, after 1 h of measurements, the sensor at 25 mm has recorded grains with settling velocities between 25 mm.s "1 and 6.94x10 "3 mm.s "1, while settling speeds between 200 mm.s I and 0.056 mm.s 1 are recorded by the sensor at 200 mm. In this way, detail concerning the fastest settling grains is acquired by the lowest sensor, with extension to slower settling grains provided by ever higher sensors. The full range of settling velocities is obtained by using all four sensors. 5.1. CALIBRATION A local estuary bed sample of mud was used for calibrating the Elbe experiment. Under the zero-order premise that no flocculation was better than an unknown and probably unrepresentative amount of flocculation, it was decided to calibrate with the mud in a dispersed state. The sample was washed to remove any salts, the turbidity tank was filled with distilled and deionized water, and a dispersant was added (1% Calgon solution). Four settling experiments were performed, each with a different starting concentration. The rate of settling was slower in the turbidity tank than was observed in the settling box. Little overlap in
300
l
105
'
'
'
I 10'
'
'
'
'
I
I
'
I
I
':i o
8
0| 10o
I
10z
10 3
10'
Time (s) Fig. 6 The decrease in concentration with time after box closure. The main plot shows the time-series plotted on a logarithmic timescale and the inset shows the first 600 s only on a linear timescale. the MOBS voltages for the different settling curves prevented the construction of equal voltage lines necessary for the time-varying calibration. Instead, a calibration of the four values of concentration against backscatter was done at each of the first three extraction times (3, 6 and 9 min). For each MOBS, the exponent decreased with time, that is, a given concentration was associated with a higher voltage later in the settling process, when the particles were smaller. Particle size is therefore having an important effect. These calibrations were not considered sufficiently rigorous, based on four points only, to be applied generally to the data, and it was decided to pursue a time-invariant calibration while recognizing that this is a first approximation. The concentrations and voltages derived from a single settling experiment were used for a global calibration (Fig. 5). The effects of varying particle size are thus not differentiated in the calibration. Settling was presumed to begin once turbulence had died away. Plots of concentration against the logarithm of time highlight the early turbulent period during which sediment is maintained in suspension (Fig. 6-main). Looking primarily at the curve of MOBS 3 which determines the early settling behaviour shows that concentration begins to decrease after three minutes (Fig. 6-inset). The first three minutes of each timeseries are therefore ignored and timed settling commences thereafter. This problem is not specific to our box, just readily identified because of the continuous recording MOBS. All other settling columns will exhibit this phenomena, to amounts dependent on the column dimensions. Ignoring the importance of trapped turbulence will lead to underestimations of settling velocity. This may well have occurred in results from Owen tubes.
106
P.B. MURRAY I.N. MCCAVE, T.R.E. OWEN, M. MASON & M.O. GREEN
100
EC3
i - ....
i I Ebt~lood Rood/ebb
i
i
i
. i . . . . . i. . . . .
~..
.' 0"l~w
80-
/
-~
0.08
S
=2"2E-4xC+~00~/j
600.06-
E
,o
j
o
4013_
/
> 0.04c/)
.E
20I
I
I
i
I
i
I
i
I l l
1
Z:: 0.02-
Ebb/flood Flood/ebb
-I
[. . . . .
•
F 0.0
Settling velocity (mm/s)
[]
×
I ....... 1 O0
T--200
I 300
4OO
Sediment concentration (mg/I) Fig. 7. The weight frequency (bar chart) and cumulative weight frequency (line chart) settling velocity distributions Fig. 8 The dependence of settling velocity on sediment confrom two closures of the settling box; one at a flood/ebb slack centration; large values in both parameters are associated water and one at an ebb/flood. with ebb/flood tidal transitions. Having derived concentration time-series for each sensor, we calculated the weight frequency distribu- situ measurements of optical backscatter. Converting tion between eleven logarithmic settling groups, from the voltages output from the MOBS into sediment 10-3 in power intervals of 0.5, up to 10 ~'5. Summation concentrations and then settling velocities has proven gives the cumulative values. Only those sensors that rather more difficult. have adequately resolved the settling velocities assoA time-variable calibration is proposed that ciated with a group are used, We therefore use the accounts for a systematic change in particle size. upper sensor to obtain the slow settling end of the After demonstrating the importance of particle size, spectrum and the lower sensor for the fast end. The this method was abandoned due to insufficient calimiddle ones are used in various combinations and bration data and a time-invariant calibration was enable a smooth overlap to be made. employed globally. Our present strategy is to pursue a time-dependent calibration, but now allowing the sed5.2. RESULTS AND DISCUSSION iment to flocculate and thus settle more rapidly. In this way it is hoped to reproduce the rate of concentration The settling velocity distributions divide into two decrease exhibited by the data from the Elbe. groups and typical examples are shown in Fig. 7. The next deployment of the settling box will be on Both the frequency bar graph and the cumulative fre- the Quadrapod and will include the following modifiquency line graph show that during the transition from cations : a flood-ebb tide, the suspended sediment is domi- --1. The MOBS will have been specifically designed for the box and will have a narrower vertical nated by particles with a slow settling velocity range. beam angle. The measurement volume of each In contrast, during the ebb-flood transition, the sediment exhibits a more even distribution of settling sensor will be unique and will not be impinged velocities about a median velocity that is significantly upon by the box lid. larger. These differences between the tidal slack --2. The mud for calibration can be obtained from in situ suspended sediment samplers, and reconwaters were repeated for each of the nine successful stituted for calibration in the proportions found in closures of the box, and may reflect the location of these syringe samplers. Twelve syringe samthe turbidity maxima as it is advected by the tide. plers are housed by the Quadrapod; samplers The Elbe data are summarized in a plot of median will be extracted at the same elevation as the settling velocity against sediment concentration (Fig. box, and at times coincident with box closure. 8). A positive relationship between the two variables is apparent. The association of the highest concentra- Already proven in an estuary, the settling box will protions and settling velocities with ebb-flood slack water vide some very interesting data when used as part of the Quadrapod system. is evident. 6. CONCLUSIONS The trial deployment of the settling box was very successful, proving it as a new robust tool for making in
Acknowledgements.--We would like to thank the University of Hamburg for the use of the pontoon and GKSS for use of RV 'Ludwig Prandtl'. We are grateful to D. Eisma and K. Dyer for organizing the experiment.
IN SITU SETTLING VELOCITY BOX
7. REFERENCES Downing, J.P., R.W. Sternbsrg & C.R.B. Lister, 1981. New instrumentation for the investigation of sediment suspension processes in the shallow madne environment.--Mar. Geol. 42: 19-34. Fennessy, M.J. & K.R. Dyer, 1994. INSSEV: An instrument to measure the size and settling velocity of flocs in situ.mMar. Geol. 117: 107-117. Galehouse, J.S., 1971. Sedimentation analysis. In: R.E. Carver. Procedures in Sedimentary Petrology. WileyInterscience, New York, U.S.A., 653 pp. Kineke, G.C., R.W. Sternberg & R. Johnson, 1989. A new instrument for measuring settling velocities in situ. Mar. Geol. 90: 149-158.
107
McCave, I.N. & T.F. Gross, 1991. In-situ measurements of particle settling velocity in the deep s e a . ~ a r . Geol. 99:403-411. Owen, M.W., 1976. Determination of the settling velocities of cohesive muds. Hydraulics Research Station, Wallingford, Rep. IT 161: 1-8. Smith, R.J., 1994. Investigation of an Optical Backscatter Device. Part II project for Department of Physics, University of Cambridge, unpublished. Van Leussen, W. & J.M. Cornelisse, 1993. The determination of the sizes and settling velocities of estuarine flocs by an underwater video system.~eth. J. Sea Res. 31: 231-241.
Journal of Sea Research 36 (1/2): 101-107 (1996)
A R O B U S T I N S I T U S E T T L I N G V E L O C I T Y BOX F O R C O A S T A L S E A S
P.B. MURRAY 1, I.N. McCAVE 1, T.R.E. OWEN 1, M. MASON 1 and M.O. GREEN 2 1Bullard Laboratories, Department of Earth Sciences, University of Cambridge, Cambridge, CB3 0EZ, UK 2NIWA Ecosystems, PO Box 11-115, Hamilton, New Zealand
ABSTRACT A new instrument for measuring the in situ settling velocity able to withstand the harsh conditions common in coastal waters is introduced. This settling box traps a parcel of fluid and measures the decaying concentration with an array of optical backscatter probes; output from these four sensors is then combined to produce a single settling velocity distribution. Preliminary deployments have been undertaken in the river Elbe and have produced encouraging results. Once the box is closed, an initial period of turbulent fluctuations is followed by settling and a smooth decrease in concentration. Settling velocity distributions recorded at transitions from flood/ebb tides are distinct from those at ebb/ flood transitions; median settling velocity was found to increase with sediment concentration.
Keywords: Settling velocity, cohesive, flocculation, coastal seas, optical backscatter, particle size, calibration
1. INTRODUCTION The settling velocity of suspended particles is an important parameter necessary for modelling the suspended sediment distribution in coastal waters. Samples of sediment may be resuspended in the laboratory and the settling velocity recorded. This technique is suitable for sand sized particles, but not for finer material. Silts and clays are cohesive and the individual particles bind together in larger structures or flocs. Flocculation occurs in response to the physical, chemical and biological conditions at a particular time in an environment. Sampling the material destroys the delicate floc structure. Accurate determination of the settling velocity thus requires in situ measurements. The earliest successful device for measuring the in situ settling velocity was the Owen tube (Owen, 1976). This takes a horizontal sample of fluid, which is brought to the surface and rotated 90 degrees, whereupon the tube acts as a settling column and samples are withdrawn at timed intervals. Subsequent instrumentation has followed two principles. Firstly, tracking of individual flocs in a small volume of fluid, using film or photography (Fennessy & Dyer, 1994; Van Leussen & Cornelisse, 1993). Such instruments provide relatively precise measurements, but sample only a small volume of fluid, appear to bias
towards the largest flocs which the eye is drawn towards, and are delicate, rendering them difficult to operate outside estuaries. We decided to follow the alternative path of automating the Owen tube with optical sensing and building an instrument that would be confidently employed in coastal waters as well as estuaries. A robust settling box was designed that uses an array of sensors for measuring backscattered infrared light. This is related to sediment concentration by calibration experiments. The design follows similar instruments used by Kineke et al. (1989) and McCave & Gross (1991), and will commonly be attached to our bottom mounted instrument frame, the Quadrapod. This frame is deployed at sea for up to 60 days during which time, the settling box will close for pre-programmed durations, recording the settling backscatter almost continuously. Deployment in locations inaccessible to other instruments and remote operation, often during shelf storms, will enhance our knowledge considerably. 2. THE SETTLING VELOCITY BOX The cylindrical settling box (Fig. la) (30 cm long and 25 cm diameter) traps a parcel of fluid between two end plates or lids and then measures the decaying backscatter by an array of Miniature Optical Back-
102
P.B. MURRAY, I.N. MCCAVE, T.R.E. OWEN, M. MASON & M.O. GREEN
b
TOP
SAMF>LE
,
I~/FRA -- RED BAEI'(SCATTEI
I I?
N II
1
I
'"°
....
BOTTOM SEALING
,.
LID
(L =i G N i ' =
~¢,
Fig. 1. (a) the Settling Velocity Box and (b) a Miniature Optical Backscatter Sensor (MOBS). scatter Sensors. The device is physically strong, with collars to reinforce the cylinder ends and a side plate between the two collars; the plate is then available for attaching the mechanics. Mechanical strength comes from having a simple method of performing the single mechanical operation, lid closure. An electric screwdriver motor winds a crank along a threaded rod, thereby moving the arms that simultaneously close the top and bottom lids. The lids and cylinder ends are tapered, with a neoprene lining on the lids ensuring a good seal, important to avoid contamination of the sample by the external flow. The motor-driven lids close, compressing the neoprene lining until no further movement is possible, whereupon the current overloads and the motor stalls. Closure is slow (approximately 10 seconds) in order to minimize turbulence generation. The lids open to an angle of 45 ° and in the opposite direction to one another so as to maximize flushing, by either waves or a tidal current. A 15 dm 3 parcei of fluid is trapped by the box. Sediment particles fall out of suspension, and the decaying backscatter is measured by four MOBS positioned at 25, 50, 100 and 200 mm from the top of the cylinder. The MOBS are flush mounted so as not to interfere with the settling, and can be readily removed for calibration.
The settling box was designed to be mounted on and driven by the Quadrapod. However, the box can also be operated autonomously in a dedicated frame which is then supported by a cable. In this configuration, the MOBS electronics is located beside the box, an electronic cable brings power from the surface and another allows box control and returns data on-line. 3. SEDIMENT CONCENTRATION SENSORS Optical backscatter instruments have been used successfully to measure suspended sediments in a variety of environments (Downing et al., 1981). Sensors are small, particularly important when attempting to resolve the steep concentration gradient immediately above the seabed, and can measure high concentrations at frequencies sufficient to resolve turbulent as well as wave fluctuations. Our experiments demanded an instrument that could be deployed at sea for typically 60 days, with a consistent response over a wide range of sediment types and concentrations. Several sensors should be capable of operating alongside one another. Optimizing the considerations, led to the development of a new Miniature ~ t i c a l Backscatter Sensor (the MOBS) at the Bullard Laboratories, University of
IN SITU SETTLING VELOCITY BOX
103
mm and 50 mm below the top lid) overlap and that the highest beam 'sees' the surface lid causing the beam to be asymmetric (Smith, 1994). A second generation of MOBS have been designed with a reduced beam width thereby ensuring that all sensors measure a unique and symmetrical volume of fluid. 4. CALIBRATION
) PMDPELL~R
c~,l=
Fig. 2. The turbidity tank.
Cambridge. This versatile sensor (Fig. l b) uses an infrared light emitting diode with a spectral peak at 850 nm, to illuminate a volume of fluid immediately in front of the sensor, and a matched photo diode to measure the amount of light backscattered by the suspended particles. Emitter and receiver are separated behind identical lenses, and are inclined at about 20" to one another. This separation also enables automatic compensation for fouling on the windows. The geometry confines the vertical spreading of the beam ensuring a deeper measurement field even in high concentrations, and enhances the sensitivity (dynamic range is 1:20000). The emitter operates in a pulsed mode to maximize the signal to noise ratio, with automatic gain subtraction of background light and detector offset to minimize temperature effects. Pulsed emission further allows up to 16 sensors to operate in close proximity without problems of interference. The MOBS are small (35 mm x 17 mm) and sample at 1Hz. They are made in strings of four, and operate cyclically to avoid interference. Modelling of the vertical beam widths of the sensors when located in the settling box has demonstrated that in clear water, the top two beams (at 25
The in situ measurements of backscatter due to suspended material that are measured by the MOBS must be converted into sediment concentration. A calibration is necessary at each site using representative material, because backscatter is a function of particle size, shape and density (as well as concentration). Previous calibrations have been undertaken in the turbidity tank shown in Fig. 2. Material from the seabed is added to the tank in a known quantity. The propeller mixes the material, generating a homogeneous suspension, which is then recorded by the MOBS. The process is repeated for a variety of concentrations. Such a simple method has successfully been employed for calibrating MOBS that are used to record tidal and wave driven suspended sediment fluctuations. In the settling box, however, as the concentration decreases with time, the particle size also diminishes because the larger particles settle first. Small grains are more effective at scattering light. A calibration based on the bulk suspended material will tend to overestimate concentrations in the tail of the curve where the particles are small. A time-dependent calibration that makes allowance for the decreasing particle size is required. A standard technique for deriving the weight frequency settling velocity distribution of sediments comprising silts and muds is pipette analysis (see Galehouse, 1971 for a full account). A dispersed, homogeneous, suspension of known concentration is allowed to settle, while samples are withdrawn by pipette at a set level below the surface (settling distance = h) and at particular times (settling time = t). The time taken to fall the distance gives the settling velocity, Ws=h/to Samples are dried and weighed and the decrease in concentration (as a proportion of the starting concentration) gives the weight frequency for that range in settling velocities. By performing several pipette experiments with different starting concentrations and simultaneously measuring the MOBS voltage, it is possible to draw plots of the concentration against the reciprocal of settling velocity (used instead of time because it accounts for the differing fall distances) and superimpose lines of equal voltage on the graph (see Fig. 3). If particle size has no effect then voltage lines will be parallel to the x-axis and there will be a unique concentration for each voltage. Alternatively, the voltage lines will slope away from the x-axis with time, and there will be a range of concentrations for each voltage dependent on how long into the settling the
104
P.B. MURRAY, I.N. MCCAVE, T.R.E. OWEN, M. MASON & M.O. GREEN
400
..... i
i
~
.....
T
.............
]
/
100
I
0
I ....
10
~
20
'"'~"
30
,
I
40
50
1/Ws (1/(mm/s)) Fig. 3. The time-varying calibration scheme. Decay curves (full lines) are drawn through the measured concentrations (crosses) of four settling experiments.Simultaneous records of MOBS output enable the addition of voltage contours (dotted lines). measurement is made, and therefore the size of particles remaining in suspension. The turbidity tank was used for settling the mud (Fig. 2); the lid was not used for fear of disrupting the near-surface settling. A pipette analysis was undertaken, modified to extract the greatest accuracy from the facilities available. The first modification was to extract the samples and position the MOBS as low in the tank as possible. By having a long settling distance, the settling time is expanded. This helps to minimize the relative importance of turbulence that takes several tens of seconds to decay after the pump has stopped. Also, a slower concentration decay is better for sampling than a steep decay because of the inaccuracies in averaging over a rapidly changing variable. The second modification was to pump the withdrawal of samples through a glass tube, and not use a pipette. Pumping the samples ensures a constant and controllable extraction rate. The glass tube is in residence throughout the experiment, avoiding disruption of settling by insertion or extraction of a pipette. Finally, the large volume of the tank enables large samples to be extracted without significant alteration of the water level and thus the settling distance. The starting concentration (pump on) was sampled and then during settling, at 3, 6, 12, 24 min (and further times increasing by a multiple of 2 up to 12 h) after stopping the pump. Settling velocities between 491 mm.s 1 (600 mm per 1.223s) and 0.014 mm.s -1 (500 mm per 12h) were recorded. Samples of 250 cm 3 were extracted at the first three settling times, and 500 cm 3 at later times.
ments that make in situ measurements of settling velocity and particle size. The exercise was conducted in the mouth of the river Elbe, where the University of Hamburg have a pontoon moored in about 20 m of water, near to Brunsbfittel. This facility was made available for the experiment, together with a GKSS research vessel, RV 'Ludwig Prandtl'. Participants from Europe and North America, comprising fifteen groups were divided between the pontoon and the ship. Measurements were made over four days, between 0800 and 2000 hrs for three days and 0800 and 1200 hrs on the final day. The ideal measurement strategy was to take measurements at 9 m depth and every hour on the hour. The settling box was suspended from a cable and lowered by winch into the water. A transmissometer was attached to the frame of the box in order to monitor the external sediment concentration. Data from both transmissometer and settling box were transmitted to the surface and displayed on a computer in real-time. A lead weight was attached to the underside of the box in an attempt to keep it vertical in the fast flowing tidal current that exceeded 1.5 m.s"1. This proved insufficient, and the box was pulled from the vertical for much of the tide. It was decided to deploy only when the wire angle was less than 10"; this limited data gathering to about 2 h either side of slack water. Nine settling periods were measured, all close to times of slack water. Four captures were taken of the ebb/flood transition and five of the flood/ebb. A typical closure is shown in Fig. 4. Each MOBS exhibits initial rapid fluctuations in backscatter, evidence of trapped turbulence. This dies out after about three minutes and gives way to a fairly smooth decrease in backscatter with time. The uppermost sensor yields the most variable profile because this is the region of lowest concentration and therefore greatest sensitivity to perturbation. Fortunately the most important sensor early in the experiment is the lowest one. A jump indi-
_•
w ...... T ............ ~
A trial deployment of the settling box was undertaken as part of an intercalibration programme of instru-
7
r ]
+
'
I
+
i
m
.
[
5. THE ELBE EXPERIMENT
i
800
0
...... 7-----
1200
2400
I :3600
- - - 1 4800
6000
Time after box closure (s) Fig. 4. Output from the four MOBS during a box closure.
IN SITU SE'r-rLING VELOCITY BOX
I
1000
400
~' ;
I
'
'
'
Y. 1(~0
'
'
'
r= = 0.935
200
300
Sediment concentration (mg/I) Fig. 5. Calibration curve for MOBS 0. cating decreased backscatter evident in all sensors after about 4000 and 4500 seconds probably indicates tidal movement of the box. Earlier variations may be naturally occurring, for instance by differential settling whereby particles that are falling at different rates coalesce. The sensor that is located highest in the box (25 mm from the surface lid) shows the steepest rate of decay in backscatter; decay rate decreases for each lower sensor as the settling distance increases. The coarse grains pass the highest sensor rapidly and their progression is recorded quickly without resolving much detail. Each lower sensor doubles the settling distance, effectively expanding the x-axis by 2 and loosing the very slow-settling grains. For example, after 1 h of measurements, the sensor at 25 mm has recorded grains with settling velocities between 25 mm.s "1 and 6.94x10 "3 mm.s "1, while settling speeds between 200 mm.s I and 0.056 mm.s 1 are recorded by the sensor at 200 mm. In this way, detail concerning the fastest settling grains is acquired by the lowest sensor, with extension to slower settling grains provided by ever higher sensors. The full range of settling velocities is obtained by using all four sensors. 5.1. CALIBRATION A local estuary bed sample of mud was used for calibrating the Elbe experiment. Under the zero-order premise that no flocculation was better than an unknown and probably unrepresentative amount of flocculation, it was decided to calibrate with the mud in a dispersed state. The sample was washed to remove any salts, the turbidity tank was filled with distilled and deionized water, and a dispersant was added (1% Calgon solution). Four settling experiments were performed, each with a different starting concentration. The rate of settling was slower in the turbidity tank than was observed in the settling box. Little overlap in
300
l
105
'
'
'
I 10'
'
'
'
'
I
I
'
I
I
':i o
8
0| 10o
I
10z
10 3
10'
Time (s) Fig. 6 The decrease in concentration with time after box closure. The main plot shows the time-series plotted on a logarithmic timescale and the inset shows the first 600 s only on a linear timescale. the MOBS voltages for the different settling curves prevented the construction of equal voltage lines necessary for the time-varying calibration. Instead, a calibration of the four values of concentration against backscatter was done at each of the first three extraction times (3, 6 and 9 min). For each MOBS, the exponent decreased with time, that is, a given concentration was associated with a higher voltage later in the settling process, when the particles were smaller. Particle size is therefore having an important effect. These calibrations were not considered sufficiently rigorous, based on four points only, to be applied generally to the data, and it was decided to pursue a time-invariant calibration while recognizing that this is a first approximation. The concentrations and voltages derived from a single settling experiment were used for a global calibration (Fig. 5). The effects of varying particle size are thus not differentiated in the calibration. Settling was presumed to begin once turbulence had died away. Plots of concentration against the logarithm of time highlight the early turbulent period during which sediment is maintained in suspension (Fig. 6-main). Looking primarily at the curve of MOBS 3 which determines the early settling behaviour shows that concentration begins to decrease after three minutes (Fig. 6-inset). The first three minutes of each timeseries are therefore ignored and timed settling commences thereafter. This problem is not specific to our box, just readily identified because of the continuous recording MOBS. All other settling columns will exhibit this phenomena, to amounts dependent on the column dimensions. Ignoring the importance of trapped turbulence will lead to underestimations of settling velocity. This may well have occurred in results from Owen tubes.
106
P.B. MURRAY I.N. MCCAVE, T.R.E. OWEN, M. MASON & M.O. GREEN
100
EC3
i - ....
i I Ebt~lood Rood/ebb
i
i
i
. i . . . . . i. . . . .
~..
.' 0"l~w
80-
/
-~
0.08
S
=2"2E-4xC+~00~/j
600.06-
E
,o
j
o
4013_
/
> 0.04c/)
.E
20I
I
I
i
I
i
I
i
I l l
1
Z:: 0.02-
Ebb/flood Flood/ebb
-I
[. . . . .
•
F 0.0
Settling velocity (mm/s)
[]
×
I ....... 1 O0
T--200
I 300
4OO
Sediment concentration (mg/I) Fig. 7. The weight frequency (bar chart) and cumulative weight frequency (line chart) settling velocity distributions Fig. 8 The dependence of settling velocity on sediment confrom two closures of the settling box; one at a flood/ebb slack centration; large values in both parameters are associated water and one at an ebb/flood. with ebb/flood tidal transitions. Having derived concentration time-series for each sensor, we calculated the weight frequency distribu- situ measurements of optical backscatter. Converting tion between eleven logarithmic settling groups, from the voltages output from the MOBS into sediment 10-3 in power intervals of 0.5, up to 10 ~'5. Summation concentrations and then settling velocities has proven gives the cumulative values. Only those sensors that rather more difficult. have adequately resolved the settling velocities assoA time-variable calibration is proposed that ciated with a group are used, We therefore use the accounts for a systematic change in particle size. upper sensor to obtain the slow settling end of the After demonstrating the importance of particle size, spectrum and the lower sensor for the fast end. The this method was abandoned due to insufficient calimiddle ones are used in various combinations and bration data and a time-invariant calibration was enable a smooth overlap to be made. employed globally. Our present strategy is to pursue a time-dependent calibration, but now allowing the sed5.2. RESULTS AND DISCUSSION iment to flocculate and thus settle more rapidly. In this way it is hoped to reproduce the rate of concentration The settling velocity distributions divide into two decrease exhibited by the data from the Elbe. groups and typical examples are shown in Fig. 7. The next deployment of the settling box will be on Both the frequency bar graph and the cumulative fre- the Quadrapod and will include the following modifiquency line graph show that during the transition from cations : a flood-ebb tide, the suspended sediment is domi- --1. The MOBS will have been specifically designed for the box and will have a narrower vertical nated by particles with a slow settling velocity range. beam angle. The measurement volume of each In contrast, during the ebb-flood transition, the sediment exhibits a more even distribution of settling sensor will be unique and will not be impinged velocities about a median velocity that is significantly upon by the box lid. larger. These differences between the tidal slack --2. The mud for calibration can be obtained from in situ suspended sediment samplers, and reconwaters were repeated for each of the nine successful stituted for calibration in the proportions found in closures of the box, and may reflect the location of these syringe samplers. Twelve syringe samthe turbidity maxima as it is advected by the tide. plers are housed by the Quadrapod; samplers The Elbe data are summarized in a plot of median will be extracted at the same elevation as the settling velocity against sediment concentration (Fig. box, and at times coincident with box closure. 8). A positive relationship between the two variables is apparent. The association of the highest concentra- Already proven in an estuary, the settling box will protions and settling velocities with ebb-flood slack water vide some very interesting data when used as part of the Quadrapod system. is evident. 6. CONCLUSIONS The trial deployment of the settling box was very successful, proving it as a new robust tool for making in
Acknowledgements.--We would like to thank the University of Hamburg for the use of the pontoon and GKSS for use of RV 'Ludwig Prandtl'. We are grateful to D. Eisma and K. Dyer for organizing the experiment.
IN SITU SETTLING VELOCITY BOX
7. REFERENCES Downing, J.P., R.W. Sternbsrg & C.R.B. Lister, 1981. New instrumentation for the investigation of sediment suspension processes in the shallow madne environment.--Mar. Geol. 42: 19-34. Fennessy, M.J. & K.R. Dyer, 1994. INSSEV: An instrument to measure the size and settling velocity of flocs in situ.mMar. Geol. 117: 107-117. Galehouse, J.S., 1971. Sedimentation analysis. In: R.E. Carver. Procedures in Sedimentary Petrology. WileyInterscience, New York, U.S.A., 653 pp. Kineke, G.C., R.W. Sternberg & R. Johnson, 1989. A new instrument for measuring settling velocities in situ. Mar. Geol. 90: 149-158.
107
McCave, I.N. & T.F. Gross, 1991. In-situ measurements of particle settling velocity in the deep s e a . ~ a r . Geol. 99:403-411. Owen, M.W., 1976. Determination of the settling velocities of cohesive muds. Hydraulics Research Station, Wallingford, Rep. IT 161: 1-8. Smith, R.J., 1994. Investigation of an Optical Backscatter Device. Part II project for Department of Physics, University of Cambridge, unpublished. Van Leussen, W. & J.M. Cornelisse, 1993. The determination of the sizes and settling velocities of estuarine flocs by an underwater video system.~eth. J. Sea Res. 31: 231-241.