Morris & Fan. Reservoir Sedimentation Handbook
Подождите немного. Документ загружается.
FLUVIAL MORPHOLOGY AND SEDIMENT SAMPLING 8.17
FIGURE 8.12 Intake boom for use with a pumped sampler which maintains the sampler intake
at a fixed proportional depth (e.g., 60 percent of total depth) in small streams (after Eads an
d
Thomas, 1983).
Proportional depth sampling can be achieved in small streams by using a sampling
boom anchored to the streambed and fitted with a float at the downstream end as shown
in Fig. 8.12. Anchoring the boom on the bottom instead of suspending it from above
facilitates the passage of floating debris. The intake nozzle is oriented parallel to the
boom, facing downstream, and is set to sample at 60 percent of the stream depth. A test
system was found to produce sample concentrations similar to simultaneously measured
depth-integrated samples taken by a DH-48 sampler (Eads and Thomas, 1983).
8.3.4 Nozzle Orientation
Sampling efficiency is the ratio of the sampled sediment concentration to the stream
concentration at the sampling point. Essentially 100 percent sampling efficiency is
achieved by using a sampling nozzle pointed in the upstream direction with an intake
velocity equal to the ambient flow velocity (isokinetic sampling), but an upstream-facing
nozzle is easily clogged and is not recommended for automatic samplers (Edwards and
Glysson, 1988).
is
achieved by using a sampling nozzle pointed in the upstream direction with an intake
velocity equal to the ambient flow velocity (isokinetic sampling), but an upstream-facing
nozzle is easily clogged and is not recommended for automatic samplers (Edwards and
Glysson, 1988).
A nozzle set at an angle to the flow will produce flow curvature, and because
sediment particles will not curve as quickly as the water entering the nozzle, decreased
sampling efficiency results (Fig. 8.9d). Winterstein (1986) performed flume tests to
determine the effect on sampling efficiency of orientation and flow velocity in a 6.35-mm
inside-diameter nozzle for 0.06- and 0.20-mm-diameter silica sand, using an upstream-
facing nozzle as the reference. It was shown that the size distribution of suspended
sediment samples collected by a nozzle set at an angle to the flow will be biased toward
smaller sediment sizes. A small sampling error is produced by upstream-oriented nozzles
or by a directly downstream-oriented nozzle, but a high degree of undersampling occurs
for nozzles oriented at 90°. Fines will be only slightly undersampled, but undersampling
error increases as a function of particle size and sampling efficiency may be lower than
50 percent for 0.2-mm sand compared to isokinetic sampling. Winterstein concluded that
the high sampling efficiency of the downstream orientation occurred because sand grains
were thrown in front of the nozzle by the turbulent wake shed by the nozzle, but
cautioned that this high degree of efficiency may not be achieved in the field where the
nozzle may not be oriented exactly downstream to the flow. Edwards and Glysson (1988)
recommend a downstream orientation for sampling from pumped samplers.
A nozzle set at an angle to the flow will produce flow curvature, and because
sediment particles will not curve as quickly as the water entering the nozzle, decreased
sampling efficiency results (Fig. 8.9d). Winterstein (1986) performed flume tests to
determine the effect on sampling efficiency of orientation and flow velocity in a 6.35-mm
inside-diameter nozzle for 0.06- and 0.20-mm-diameter silica sand, using an upstream-
facing nozzle as the reference. It was shown that the size distribution of suspended
sediment samples collected by a nozzle set at an angle to the flow will be biased toward
smaller sediment sizes. A small sampling error is produced by upstream-oriented nozzles
or by a directly downstream-oriented nozzle, but a high degree of undersampling occurs
for nozzles oriented at 90°. Fines will be only slightly undersampled, but undersampling
error increases as a function of particle size and sampling efficiency may be lower than
50 percent for 0.2-mm sand compared to isokinetic sampling. Winterstein concluded that
the high sampling efficiency of the downstream orientation occurred because sand grains
were thrown in front of the nozzle by the turbulent wake shed by the nozzle, but
cautioned that this high degree of efficiency may not be achieved in the field where the
nozzle may not be oriented exactly downstream to the flow. Edwards and Glysson (1988)
recommend a downstream orientation for sampling from pumped samplers.
The intakes for commercial samplers are often supplied with strainers designed to
exclude debris. Strainers should normally be removed for collecting sediment data, since
they can create a zone of reduced turbulence at the intake which may allow sedimentation
and undersampling of larger grains.
The intakes for commercial samplers are often supplied with strainers designed to
exclude debris. Strainers should normally be removed for collecting sediment data, since
they can create a zone of reduced turbulence at the intake which may allow sedimentation
and undersampling of larger grains.
FLUVIAL MORPHOLOGY AND SEDIMENT SAMPLING 8.18
8.3.5 Cross-Section Coefficient
It is necessary to determine the relationship between the pumped sample concentration
and the mean sediment concentration in the stream by manual sampling over a large
range of flows. Pumped samples should be taken at the start and end of the more time-
consuming manual sampling, and the cross-section sample is compared to the average of
the pumped samples. Single-point sampling can accurately represent the mean
concentration of fine sediments (silts and clays) that exhibit an essentially uniform
concentration distribution. However, the measurement of sand load presents a difficult
problem and can produce very poor correlation between the point samples and the mean
concentration. One possible approach is to separate both pumped and manual samples
into fine and coarse fractions, and to establish relationships for each individually.
8.4 CONTINUOUS TURBIDITY MEASUREMENT
The sediment concentration in streams can vary by orders of magnitude during a flood
event and may be poorly correlated to changes in discharge (see Figs 7.5 and 7.17). The
use of automated pumped samplers increases the frequency of sediment sampling,
allowing more complete sampling of runoff events, especially in flashy streams and at
remote sites. However, the larger number of samples increases laboratory costs, yet does
not produce the coverage needed to accurately track concentration on a continuous basis.
Ideally, a gage parameter would be available which varies directly as a function of
suspended sediment, and which could be continuously monitored and logged. A perfect
gage parameter for suspended sediment concentration does not exist, but turbidity has
gained acceptance as the imperfect parameter of choice for suspended sediment
monitoring.
8.4.1 Application
Turbidity is generally a better predictor of suspended fluvial sediment concentration than
is discharge, and turbidity readings logged at short intervals can continuously track the
change in concentration in the rising and falling limbs of runoff events.
Turbidity is well-suited to measurement of changes in suspended sediment
concentration in streams where the sediment composition is relatively stable as a function
of time and discharge, but its application is more complicated in streams where sediment
characteristics (and their optical properties) change significantly. Best results are
achieved with the combination of a turbidimeter and pumped sampler. The suspended
solids concentration is continuously tracked by using turbidity, and pumped samples are
collected to provide the calibration between turbidity and solids throughout the event.
This results in fewer samples to be analyzed and better coverage of events. Commercial
pumped samplers now offer both stage and turbidity probes, plus extensive programming
and data logging capability, greatly facilitating this type of data collection.
The most important advantage offered by turbidimeters is continuous measurement.
The number of data points that can be logged is virtually unlimited, enabling the
turbidimeter to continuously log events at short time intervals and define the variation in
concentration much more clearly than pumped samples alone. Instantaneous instream
variations in turbidity can be smoothed by internal routines that time-average multiple
data points over periods of several minutes. Lewis (1996) indicated that a 10-min logging
interval was adequate for a small flashy northern California stream. By monitoring dis-
charge and turbidity simultaneously, the suspended sediment load can be computed from
the logged datafiles by applying the correction relationships developed for the station.
FLUVIAL MORPHOLOGY AND SEDIMENT SAMPLING 8.19
Turbidimeters cannot replace conventional sampling techniques because the
relationship between point turbidity and the mean suspended sediment concentration in
the cross section is not constant; it requires calibration and periodic checking.
Turbidimeters also cannot operate unattended for long periods; they require attention at
regular intervals to checked for clogged lines, to clean optical surfaces, and to collect
calibration samples from pumped samplers operating in parallel with the turbidimeter. A
1- or 2-week interval between maintenance visits might be used initially and adjusted
according to conditions. In systems equipped to communicate with a base station by
cellular telephone or radio, problems can be signaled using the microprocessor at the
gage station, or the raw data can be periodically sent to the base station and processed to
detect malfunction conditions. This type of reporting system can be used to minimize
visits by technicians, spreading routine maintenance visits further apart while responding
immediately to malfunctions, thereby minimizing station downtime.
8.4.2 Types of Turbidimeters
Turbidity is a measure of light scattering by the suspended particles in the water.
Turbidity values are only approximately related to suspended solids concentration, being
influenced by factors such as particle size, shape, mineral composition, organic matter,
and color. Turbidity was originally measured with the Jackson candle turbidimeter, an
instrument consisting of a tall calibrated glass cylinder with a standard candle beneath,
and turbidity was related to the water depth at which the image of the flame was no
longer distinguishable. From this was derived the Jackson turbidity unit (JTU), which
may be encountered in older literature but is no longer in general use.
Three types of turbidimeters have been used to monitor stream sediment at gage
stations. Attenuance turbidimeters measure the attenuation of a light beam passing
through a fixed distance. Dual beams are frequently employed in field turbidimeters to
compensate for lamp aging and biofouling. Turbidimeters measuring changes in
transmitted light are relatively insensitive to low scattering intensities, and are subject to
multiple-scattering interference at high concentration.
The measurement of scattered light at an angle to the light beam helps overcome these
problems and reduces sensitivity to variation in particle size. Instruments measuring
scattered light are termed nephelometers and are the accepted standard for turbidity
measurement as specified by Standard Methods (APHA, 1985). Their readings are
expressed as nephelometric turbidity units (NTUs), as opposed to the term formazin
turbidity units (FTUs) which is reserved for instruments measuring transmitted or
absorbed light (Hach et al., 1990). The different turbidity units are not strictly
interchangeable.
Low-power submersible turbidity probes based on both principles are available from
commercial suppliers and may be integrated with pumped samplers. The Hach surface
scatter turbidimeter uses a somewhat different approach in that the light source is
directed onto the surface of a water sample continuously pumped into an enclosed
housing, and reflection is measured from the surface. Biofouling is essentially eliminated
since optical surfaces are not in contact with the water, but this equipment requires a
separate pump and line power.
8.4.3 Relationship between Turbidity and Suspended Solids
Suspended particles differ in their optical characteristics, which affect their absolute
levels of attenuance and scatter as well as the ratio of attenuance to scatter. Gippel (1995)
calibrated both attenuance and nephelometric turbidimeters to the same formazin
standard and subsequently measured natural turbidity. The attenuance turbidimeter
FLUVIAL MORPHOLOGY AND SEDIMENT SAMPLING 8.20
typically measured 2.5 times higher turbidity values than a nephelometric turbidimeter
for the silt-clay suspensions analyzed. This occurs because the formazin standard has a
much higher scattering efficiency than most natural particles. It was also found that
dissolved organic color is unlikely to alter turbidity readings by more than 10 percent.
The term specific turbidity refers to the turbidity measured in formazin units divided
by the mass particle concentration in milligrams per liter. Gippel (1995) also examined
the effect of particle size on specific turbidity, and found that particle size variation can
cause turbidity to vary by a factor of 4 for the same suspended solids concentration.
Differences in specific turbidity as a function of turbidimeter type and particle size are
summarized in Fig. 8.13a. Foster et al. (1992) examined the response of two attenuance
turbidimeter probes to various concentrations of river silts and clays using both native
water and dispersant, and found that the specific turbidity decreased more than tenfold
over the range of sediment diameters tested. Larger particles produce less turbidity
response per unit of mass.
For a turbidimeter that has been calibrated to give a linear response to a formazin
standard, the turbidity of a mixture having a fixed grain size composition will vary
linearly according to the function
T = aKC
S
+ b (8.2)
where T = turbidity, K = specific turbidity, and C
S
= suspended solids concentration. For
nephelometric turbidimeters a 1 and is a function of dissolved organic color. For
attenuance turbidimeters a = 1. The constant b is a function of organic color and has a
value of b = 0 for nephelometric turbidimeters. For attenuance turbidimeters b = 0 only
when there is no color at zero suspended solids concentration (Gippel, 1995). The
experimental relationship between turbidity and suspended solids concentration
development by Lewis (1995) in Fig. 8.13b shows both the linearity of the relationship
when grain size is constant and the greater specific turbidity associated with the finer
particles. The larger sand-size particles produce a relatively small effect on turbidity.
When particle size varies as a function of concentration, the turbidity response will
have a nonlinear form :
T = a C
s
C
+ b (8.3)
FIGURE 8.13 (a) Variation in nephelometric and attenuance turbidity as a function of grain
size and concentration (after Gippel, 1995). (b) Linear variation in turbidity as a function o
f
concentration for coarse and fine sediment showing the larger specific turbidity response
generated by fine sediment (Lewis. 1996).
FLUVIAL MORPHOLOGY AND SEDIMENT SAMPLING 8.21
where a = a characteristic coefficient. When particle size increases as a function of
suspended solids concentration, c<1 and indicates decreasing specific turbidity. When
particle size decreases as a function of concentration, c> 1 (Gippel, 1995).
The relationship between turbidity and suspended solids concentration should be con-
structed for each stream and for each distinct hydrologic condition (e.g., rising limb,
falling limb, runoff from thunderstorms, snowmelt, and high versus low discharge).
Numerous close correlations between turbidity and suspended solids concentration have
been reported in the literature, which suggests that either the particle size variations are
not usually great, or that particle size variations are often associated with systematic
variations in suspended solids concentration. The best correlations will occur in
environments where sediment properties are likely to remain relatively constant, such as
small watersheds, and where there is a wide variation in concentration (Gippel, 1995).
Some variance in specific turbidity can be tolerated, since the improved data obtained
from a continuous estimate of suspended solids concentration overcomes the problem of
infrequent sampling, which is the greatest source of error in estimating sediment loads
(Olive and Rieger, 1988).
8.4.4 Limitations of Turbidity Data
Although turbidimeters offer numerous advantages as a complement to traditional
sampling techniques, turbidity measurement also has a number of limitations which must
be understood and addressed if accurate data are to be obtained.
Turbidity is a point sampling method and incorporates the limitations associated with
the measurement of mean cross section suspended-sediment concentration by
sampling a single point. Depth-integrated samples are required to calibrate the point-
sampling system. The sampling point must be carefully located to be representative
of the entire stream according to criteria previously discussed for point sampling
using pumped samplers.
Because small particles and organics have a much higher specific turbidity than
larger sand-size particles, turbidity measurements will tend to track the concentration
of the finer fraction of the load while being rather insensitive to the coarse fraction.
This makes it difficult to interpret turbidity data from streams with a significant
amount of coarse material. It may be helpful to divide calibration samples into fine
and coarse fractions to better model the suspended solids-turbidity-discharge
relationship in a particular stream.
Optical components may be subject to biofouling after a period of days or weeks,
and the rate of biofouling can be expected to increase as a function of higher water
temperature and nutrient content. In smaller streams experiencing a wide range of
stage, the turbidimeter may be placed above the low-flow water level to reduce
biofouling of optical surfaces. Since little sediment is discharged during low-flow
periods, and suspended solids concentration is low, the missing data is of little
consequence to load computations. Periodic inspection and cleaning of all optical
surfaces is required.
Logged data should be searched for anomalous conditions, such as large turbidity
spikes with little change in discharge. These may be caused by discharge of
pollutants from upstream factories or agricultural operations. If a pump is delivering
water to a turbidimeter, a change in stream stage without a corresponding change in
turbidity may reflect clogging or pump failure. Whereas this would he detected in a
pumped sampler by empty sample bottles, the turbidimeter will continue to send
signals regardless of what it is measuring. Turbidity trends over periods of days or
weeks may reflect fouling of optical surfaces.
Turbidity measurement does not provide grain size data.
FLUVIAL MORPHOLOGY AND SEDIMENT SAMPLING 8.22
Despite the limitations associated with turbidity measurement, continuous monitoring
with some loss of precision can give a more accurate estimate of total sediment load and
its timewise variation, compared to a smaller number of pumped samples which fail to
accurately track large changes in sediment concentration. The use of a pumped sampler in
conjunction with a turbidimeter, as described in the next section, can greatly enhance the
reliability of turbidity data in streams where the concentration-turbidity relationship
changes during runoff events.
8.5 SAMPLING STRATEGIES FOR PUMPED
SAMPLERS AND TURBIDITY
Pumped samplers are frequently activated by a stage sensor which initiates a timer that
withdraws samplers at fixed intervals during events. The resulting samples are used to
construct a rating relationship or to estimate sediment load by interpolation. However,
with samplers now containing microprocessors, more sophisticated statistically based
sampling plans can be implemented which produce unbiased estimates of both sediment
load and variance, and overcome the bias that accompanies the use of traditional
nonstatistical sampling schemes. Pumped samplers can also be used to calibrate
turbidimeters.
8.5.1 Statistically Based Strategies for Pumped Samplers
When sediment load is to be estimated from pumped samples alone, it is desirable to use
stratified or variable-probability sampling strategies that reduce variance by taking
advantage of the population structure. Sampling density is increased at high flows by
using an onboard microprocessor and stage sensor to select sampling times. Thomas
(1985) initially introduced the SALT (selection at list time) sampling scheme, but more
recently has proposed time-stratified and flow-stratified sampling schemes which
produce lower variance than the SALT sample plan.
Time-stratified sampling (Thomas and Lewis, 1993) partitions a hydrograph into a
series of time intervals called strata. The duration of each time stratum is predetermined
by the operator and varies as a function of hydrograph stage and direction. The sampling
strata are programmed into the sampler's computer. At the beginning of a time stratum,
the stream stage and direction (rising or falling) are sampled, the corresponding stratum
length and sample size are retrieved from memory, the sampling times within the strata
are determined, and the sampler is operated at these times. At the end of each time
stratum the procedure is repeated.
Flow-stratified sampling (Thomas and Lewis, 1995) stratifies a hydrograph by
discharge rather than time. The total flow range is stratified into classes based on stage
height and direction, and each class is randomly sampled during the time it is occupied.
Under this method noncontiguous samples from the same flow stratum are placed
together.
Comparisons among the SALT, time-stratified, and flow-stratified sampling plans
using data from a station measuring runoff from 10.4 km
2
of the North Fork Mad River in
northern California revealed that time-stratified sampling generally produced the lowest
variance for estimating total load from individual storms, but for estimating total load
over a longer period with multiple storm peaks the flow-stratified sampling plan
produced the lowest variance (Thomas and Lewis, 1995).
FLUVIAL MORPHOLOGY AND SEDIMENT SAMPLING 8.23
8.5.2 Sampling Strategies for Turbidimeters
Lewis (1996) described the use of turbidity-controlled sampling for suspended load
estimation on a flashy mountainous stream in northern California. Turbidity was
monitored continuously and logged at 10-min intervals. A programmable data logger
signaled a pumping sampler to collect a suspended sediment concentration sample at
specific turbidity thresholds, thereby providing the samples needed to recalibrate the
suspended solids versus turbidity relationship. This recalibration during discharge
events produced an accurate estimate of the suspended solids load, even though
the sand fraction increased as a function of suspended sediment concentration and
constituted more than half the sediment load at high concentrations. Turbidity
thresholds were uniformly spaced after a log or power transformation of the turbidity
scale. Simulations revealed that, for this stream, sampling thresholds that were
uniformly spaced on the basis of the square root of turbidity gave better load estimates
than cube root or a logarithmic scale, probably because of increased emphasis on
high turbidities. To avoid sampling of ephemeral turbidity spikes caused by debris,
threshold values were exceeded for two sampling intervals before activation of the
sampler. A hydrograph reversal threshold was also established. In the studied stream,
75 percent of the suspended sediment was delivered after the discharge peak so denser
sampling thresholds were established for falling turbidities than for rising. The
relationship between turbidity and suspended solids for two events is given in Fig.
8.14, showing that hysteresis is important in some storms and absent in others at this
site. The particular sampling protocol at any other stream should be selected and
adjusted on the basis of continuous turbidity measurements at the site. By personal
communication, Lewis indicated that relations between suspended solids and turbidity
are generally linear (N. California streams). As compared to reliance on pumped
samples alone and a probability-sampling protocol such as SALT, a turbidity-controlled
protocol will provide a load estimate of comparable reliability for larger storms with
only
1
/
6
as many pumped samples, and turbidity monitoring also gives much more
detailed information.
FIGURE 8.14 Suspended-sediment concentration versus turbidity for two events on
the North Fork Mad River, California, illustrating the presence of hysteresis in one
event and its absence in another. Periodic recalibration of the concentration-turbidity
relationship using pumped samples produced an accurate load estimate (Lewis, 1996).
FLUVIAL MORPHOLOGY AND SEDIMENT SAMPLING 8.24
8.6 DURATION OF MONITORING
How long must a fluvial sediment station be monitored to obtain useful data? The
recommended duration of a monitoring program depends on the program objective. Three
situations will be considered: (I) construction of rating relationship, (2) determining mean
sediment yield, and (3) detection of trends. Construction of a sediment rating relationship
can be performed for a short record, if that record contains data representative of the
critical discharge classes responsible for most of the sediment load. However, extremely
short monitoring periods covering only a year or two which do not include major runoff
events should be used with extreme caution. The problems resulting from a short
sediment record have been discussed in Sec. 7.3 with respect to the Randolph Jennings
Reservoir. Nevertheless, accurate short-term data are far superior to long-term monitoring
which fails to adequately monitor high-discharge events, as illustrated in the Cachí case
study (Chap. 19). The Cachí case study illustrates how a flow-duration curve can be
constructed and applied to the rating curve to estimate the range of discharges that are
most critical from the standpoint of sediment yield (see Fig. 19.9). Although several years
of sediment monitoring are normally recommended for development of a rating
relationship, at Cachí it was possible to prepare an apparently accurate rating relationship
by using continuous turbidity data from a relatively small number of runoff events.
However, short datasets (including Cachí) typically do not include large episodic runoff
events, and the problem of estimating discharge during these events remains unaddressed.
In mountains or arid areas where episodic events can contribute extreme sediment loads,
this is an important consideration.
The second objective, determination of the mean sediment yield, may be achieved by
using a 5- to 10-year data-collection program. Evaluation of the Canadian sediment
sampling
,
program (Day, 1988) suggested that the historical emphasis on long-term data
collection, up to 30 years in duration, is not necessary for many engineering purposes. At
stations operating continuously for 5 to 10 years, the mean annual sediment discharge
may be determined directly without extrapolation to a longer-range discharge record as
long as sampled conditions are reasonably representative of long-term conditions.
Rooseboom (1992) suggested that a minimum of 6 years of data collection is adequate to
determine mean sediment load, but that a much longer record is required to observe
trends. Summer et al. (1992) found that 10 years of sediment data collection was
adequate to define load parameters for engineering purposes in the Danube River in
Austria. Inadequate sampling programs typically underestimate sediment load. The
coefficient of variance of the annual loads can be used to determine the standard error of
the resulting estimate of mean annual yield using Eq. (7.1); also see Fig. 7.16 and the
discussion of uncertainty in sediment yield contained in Sec. 7.3.
The third objective, detection of trends, requires long-term monitoring extending over
a period of decades. While long-term trends can be dramatic in some cases, as illustrated
in Figs. 7.8 and 7.13, in other cases trends may be difficult to identify or interpret because
of factors such as flow diversion, upstream dam construction, and the episodic nature of
sediment yield. Difficulties of this nature experienced in the Canadian sediment monitor-
ing program were described by Day (1988) and are briefly summarized in Sec. 7.3.3.
8.7 BED LOAD SAMPLING
8.7.1 Bed Load Transport
The zone close to the bed of the river is not sampled by suspended load samplers and is
termed the unsampled zone or the zone of bed load transport (Fig. 8.7). Transport in the
FLUVIAL MORPHOLOGY AND SEDIMENT SAMPLING 8.25
unsampled zone includes the rolling and jumping material in the bed load, plus some
portion of the suspended load, and the amount of material transported in this zone can
represent a significant fraction of the total load in sand-bed rivers. Bed load is difficult to
measure, and at most gaging stations it is estimated or computed rather than measured
directly. Because sediments in arid zones tend to be coarse-grained, bed load generally
constitutes a larger part of the total load in streams draining desert areas than those in
humid areas.
One common method to estimate bed load is to compute it as a percentage of the
suspended sediment load using Table 7.4 or a local relationship determined by sampling.
In rivers having a movable bed of coarse sediment, the bed load discharge can be
estimated from a bed load formula, but this method itself is subject to considerable un-
certainty. When bed load transport is supply-limited, bed load equations cannot be used.
The potential importance of bed load transport was illustrated by Long (1989),
referring to unpublished work by Lin Binwen who in 1982 computed the sediment
balance along a reach of the Yellow River between the Shaanxian gaging station, just
below Sanmenxia dam, and the Tongguan gaging station about 120 km upstream (see
Fig. 24.1 for location map). In this reach the river transports primarily silts but has a
sandy bed. During the period March–June 1976, when Sanmenxia Reservoir was trapping
sediment, the volume of material deposited in the reservoir was computed from the
difference in suspended sediment discharge measured at the upstream and downstream
stations. Sediment accumulation was also measured using the range method in the
reservoir. Sediment grain size data were collected simultaneously, making it possible to
compare the two methods not only in terms of total load but also as a function of grain
size. Suspended solids monitoring under-estimated deposition by 16 million tons over the
4-month period, and 80 percent of the unmeasured load consisted of sand-size particles
larger than 0.05 mm diameter.
8.7.2 Bed Load Sampling
Whereas suspended load transport rate is computed as the product of concentration and
discharge, the bed load sampler must directly measure the load of particles moving along
the bed. One of the major difficulties to be overcome in bed load measurement is the high
temporal and spatial variability of the movement of bed material. Numerous researchers
have documented highly irregular rates of bed load transport under conditions of constant
discharge (Fig. 8.15). At a given discharge, the short-interval bed load transport rate
typically varies from near zero to 4 times the mean rate (Hubbell, 1987). Transport rate
also varies laterally, and multiple sampling traverses across a stream are required to
determine the transport rate. Sampler efficiency is defined as the ratio of the mass of bed
load collected during any sampling interval to the true mass of bed load that would have
passed through the sampler entrance width during the same period had the sampler not
been present. Sampler efficiency is determined by calibration in a hydraulic flume, and
sampled transport rates are divided by efficiency to determine the true transport rate.
Sampler efficiency varies as a function of grain size and transport rate, further
complicating bed load measurement.
Early portable bed load samplers interfered with the flow field, resulting in a low
sampling efficiency, and sediment tended to accumulate at the sampler entrance. These
problems are minimized by a sampler which has an entrance nozzle followed by an
expanding section that generates higher velocities in the throat and clears the entrance of
entrained sediment. A permeable bag is attached to the downstream end of the expanding
nozzle to hold the collected sample. The Helly-Smith type sampler used in the United
States (Fig. 8.10d) is based on this principle and is available in two sizes. Bed load
FLUVIAL MORPHOLOGY AND SEDIMENT SAMPLING 8.26
FIGURE 8.15 Timewise variability in the rate of bed load transport (Edwards and Glysson,
1988).
sam
plers of varying designs are used in different countries, and the resulting loads are not
directly comparable. Threefold differences in measured bed load transport rates in the
same river but using different samplers were illustrated in the comparison of U.S. and
Chinese samplers reported by Childers et al. (1989).
8.7.3 Continuous Bed Load Measurement
Continuous bed load sampling is undertaken with a permanent (and costly) structure to
sample the entire bed load transported across the sampling section. Designs include a slot
in the streambed that allows the sediment to fall into a box where the accumulation is
continuously weighted (Laronne et al., 1992), or sediment may be continuously removed
from the stream by a vortex tube or by conveyer for weighing and sizing (Tacconi and
Billi, 1987). Continuous sediment discharge has also been measured by implanting
acoustic sensors in the streambed which register strikes by moving bed material. These
stations are costly, and in 1992 Laronne et al. reported that continuous bed load data
were available from only six streams worldwide. However, detailed data from these