You are on page 1of 12

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/252215503

Sediment trapping characteristics of a pit trap and the Helley-Smith sampler


in a cobble gravel bed river

Article in Water Resources Research · August 2002


DOI: 10.1029/2000WR000052

CITATIONS READS
71 233

2 authors, including:

Shannon Sterling
Dalhousie University
37 PUBLICATIONS 1,788 CITATIONS

SEE PROFILE

All content following this page was uploaded by Shannon Sterling on 02 May 2019.

The user has requested enhancement of the downloaded file.


WATER RESOURCES RESEARCH, VOL. 38, NO. 8, 1144, 10.1029/2000WR000052, 2002

Sediment trapping characteristics of a pit trap and the


Helley-Smith sampler in a cobble gravel bed river
Shannon M. Sterling1
Northwest Hydraulic Consultants, North Vancouver, British Columbia, Canada

Michael Church
Department of Geography, University of British Columbia, Vancouver, British Columbia, Canada

Received 26 October 2000; revised 13 March 2002; accepted 28 March 2002; published 14 August 2002.

[1] Although designed for granule gravels, the Helley-Smith (HS) bed load sampler is
often used in streams with much coarser beds. To compare magnitudes and grain size
distributions of sediment samples collected by a pit trap and a standard HS sampler, we
conducted 22 ‘‘sampling events’’ over a wide range of flows during a snowmelt freshet in a
stream with a coarse gravel bed. Each sampling event consisted of three simultaneous
measurements: one made with a pit trap installed in the bed, one with a HS sampler placed
on the bed directly beside the pit trap (HS-I), and one with a HS sampler placed on the
downstream rim of the pit trap (HS-P). We summed the catches of the pit trap and HS-P to
estimate ‘‘actual’’ bed load transport. The pit trap and HS-I catches were then compared to
the summed sample. The pit trap yielded a remarkably consistent, positively skewed,
sigmoidal distribution of catch efficiency for all 22 measurements, with near 100%
efficiency for material larger than 2.8 mm. The HS-I sampler was more variable in its catch
and trapping efficiency, exhibiting low trapping efficiency for midrange material (0.71 to
16 mm) but high efficiency for finer material. The results cast doubt on the accuracy of bed
load data sets collected by Helley-Smith samplers in coarse gravel channels. INDEX
TERMS: 1815 Hydrology: Erosion and sedimentation; 1894 Hydrology: Instruments and techniques;
KEYWORDS: bed load sampling, gravel bed rivers, Helley-Smith sampler, pit trap, sediment transport

1. Introduction and might reasonably be defined as ‘‘bed load.’’ The HS


[2] Measurement remains a major obstacle to understand- sampler is designed to intercept all of this material but, in
ing bed load transport in gravel bed rivers. Development addition, it may intercept material moving in suspension
and deployment of bed load samplers remains a difficult art. nearer the bed than the height of its entrance. This material
The most widely adopted design is that of Helley and Smith has been operationally classified as bed load. Prior to this
[1971], a pressure-difference bag sampler intended to trap study, we observed in simultaneous deployments in a coarse
coarse sand and fine gravels (2 – 10 mm in size) over fine gravel bed channel that a HS sampler caught material with a
gravel beds. It is supposed to trap all material traveling different size distribution (GSD) than a pit trap, raising
below the usual lower sampling limit of suspended sediment questions about the trapping performance of the two sam-
samplers. For lack of alternatives, this sampler has often plers. To elaborate this observation, we conducted compa-
been deployed on coarse gravel beds, including cobble- rative measurements of sediment trapping in a pit trap and
gravels. Some examples (among many) are Snake River, in a HS sampler. As our measurements were conducted at
with median bed surface particle diameter, D50s = 54 mm full scale in the field, we have no accurate reference
and Clearwater River, D50s = 74 mm [Jones and Seitz, estimate of the mean bed load transport rate. Our intent,
1980], Sagehen Creek, a gravel bed stream with D50s = rather, is to compare the sediment trapping performance of
58 mm [Andrews, 1994], and St. Louis Creek, a cobble- two commonly used sampling devices. Accordingly, the
boulder bed channel [Ryan and Porth, 1999]. Important purpose of the study is to test for systematic differences in
questions remain concerning the performance of Helley- catch and GSD between pit traps and HS samplers in a
Smith (HS) samplers in coarse gravel bed rivers, for which coarse gravel bed stream.
it was not designed.
[3] The purpose of a pit sampler is to intercept all 2. Methods
material that would be in contact with the bed at the slot 2.1. Study Site
position: such material might be rolling, sliding, or saltating [4] Measurements were made on a lateral bar in Harris
1
Creek, a cobble-gravel bedded stream located east of
Also at Division of Earth and Ocean Sciences, Nicholas School of the Vernon, British Columbia (Figure 1). At the study site, the
Environment and Earth Sciences, Duke University, Durham, North
Carolina, USA. drainage area is approximately 220 km2 [Church et al.,
1991]. The lateral bar has grown about 25 m downstream
Copyright 2002 by the American Geophysical Union. from a 30-degree bend in the channel caused by a bedrock
0043-1397/02/2000WR000052 obstruction on the right bank.
19 - 1
19 - 2 STERLING AND CHURCH: SEDIMENT TRAPPING IN A GRAVEL BED RIVER

Figure 1. Contour map of the study reach, showing gauge and trap locations: inset, location map.
Photos of the study site and channel bed are presented by Church et al. [1991].

[5] Compared with the streamward bar edge and chan- to about 500 mm diameter, though boulders (larger than
nel thalweg, the surface of the lateral bar has more 256 mm) are sparse. D50 of the surface material varies
exposed sand and granule gravel and has a sandy tail between 45 mm and 75 mm in the vicinity of the traps
zone. The largest stones in the channel and bar range up (Figure 2), but the subsurface material, which is supposed
STERLING AND CHURCH: SEDIMENT TRAPPING IN A GRAVEL BED RIVER 19 - 3

Figure 2. Grain size distribution for bed and surface material in the bar and partial size distributions for
suspended sediments trapped during this study (large volume samples were caught in an open container
laid horizontally in the flow). Size distribution of suspended sediments was not determined below 63 mm.

to represent the transported population in the long term, [8] Our HS samplers follow the design of Helley and
has D50  20 mm. Smith [1971] exactly. The sampler collects material while it
[6] A spring snowmelt flood of about 19 m3/s dominates rests atop the streambed; a pressure drop is induced at the
the hydrological regime at the study site [Church et al., inlet by an expanded downstream section, so that inlet
1991] (estimated by area-ratio adjustment of the mean velocities exceed ambient velocities. The intake is con-
annual flood determined at WSC gauge 08LC005, located structed of 6.3 mm (0.25 inch) steel or brass plate with a
6 km downstream). In the nival regime, flows vary rela- 76.2 mm (3.0 inch) square entrance leading to a 0.21 mm
tively slowly, so sampling is possible under quasi-steady mesh sample bag that is attached at the rear, and a support-
conditions. For this study, we established and rated a gauge ing tubular frame and tail assembly.
on the site. [9] Mobile sediment is collected in what we call a
‘sampling event’, consisting of three simultaneous measure-
ments with (Figure 4): (1) a pit trap; (2) a HS sampler
2.2. Measurement Protocol (HS-I) that was placed on the bed approximately 30 cm
[7] In 1988 12 identical traps were installed in the channel from the pit trap on the thalweg side with its entrance
bed with the upper rim level with the bar surface (trap details aligned with the upstream edge of the pit trap opening; and
in Figure 3a). After installation, there was degradation at (3) another HS sampler (HS-P) that was placed on the
traps 1A/B, 4A/B and 5A/B; for this study we used traps 3B, downstream rim of the pit trap. The purpose of HS-I is to
3C and 3D as they remained flush with the surface and provide a comparison with the pit trap; the purpose of HS-P
represented a sequence of traps that would allow access for is to catch material traveling in the path of the pit trap that
sampling over the full range of flows. In these simple pit escaped capture by the trap. The HS samplers were hand-
traps, material accumulates in an inner container that placed onto the bed; the same operators conducted all
requires periodic emptying, providing a composite ‘‘bed sampling events.
load’’ sample for all flows since the last emptying. Our pit [10] Twenty-two sampling events were observed over a
traps were designed to have a reasonable chance to intercept period of 6 days on the falling limb of the spring freshet
the largest material apt to be in motion (opening diameter, (Figure 5) in 1991. The discharge rapidly decreased during
d = 29 cm), and to be manageable in high flows (container the first two days of the sampling program from 18.4 m3/s
height, h = 38 cm). The pit trap opening is round so that it is to 4.18 m3/s (approximately 90% to 40% bank-full). Sus-
not sensitive to effects of varying direction of the oncoming pended sediment concentration remained remarkably low
current. There were no obstructions, such as large boulders through the peak of the freshet; the bed always remained
or cobbles, upstream of traps 3B/C/D. visible. With falling stage, trap location of the sampling
19 - 4 STERLING AND CHURCH: SEDIMENT TRAPPING IN A GRAVEL BED RIVER

Figure 3. Pit trap design used in this study: (a) elevation; (b) internal circulation (drawn from a similar
diagram of Tritton [1988, Figure 12.7].

events moved from 3D (one event), to 3C (7 events) to 3B dation and the trap rim became elevated above the bed
(14 events). Grain size of the bed surface material is slightly surface by a maximum of 2.5 cm.
larger at trap 3B, nearest the thalweg, than at trap 3D. [11] Our sampling events had unusually long sampling
During the study the pit traps were subject to local degra- times, which averaged short-term fluctuations in transport
STERLING AND CHURCH: SEDIMENT TRAPPING IN A GRAVEL BED RIVER 19 - 5

this reason that we used the heavily weighted frame/fin


assembly rather than an unstabilized hand sampler). When
HS-I was approximately one quarter full we ended the
sampling event so that the hydraulic efficiency of the HS
samplers would not be reduced. For four sampling events
(18, 20, 21, and 22) the filling rates were slow enough that
the samplers were left in the river overnight; flows on these
occasions were far less than those able to move the heavy
samplers. Each sample was sieved to half psi intervals down
to – 4 psi (0.063 mm), with organic material removed (we
adopt psi = log2D = - phi as a convenient unit of measure
for dominantly gravelly material). When comparing total
weight of samples (not separated by grain size), we trun-
cated samples at 0.25 mm, conservatively estimated to
eliminate the expected portion of the sample transported
mainly in suspension (see Figure 2).
[12] To estimate ‘‘actual bed load transport’’ rate, collec-
tion rates for the pit trap and the HS-P sampler are summed.
The pit and HS-I collection ‘‘efficiency’’ are estimated by
dividing the individual sampler collection rates by the
‘‘actual’’ rate. When material of a certain size category
was trapped by the HS-I sampler and not by the pit trap or
the HS-P sampler, the HS-I sampler is assigned 100%
efficiency for the relevant size category for that sampling
event.

2.3. Theoretical Efficiency of a Pit Trap


[13] While the bed load trapping efficiency of pit traps is
considered to be relatively high [cf. Hubbell, 1987], it may
be affected by water circulation within the trap relaunching
Figure 4. Sampling event layout. previously collected particles, and by saltating sediment
overpassing the trap. We investigated these conditions to
rate; accordingly, we are confident that there was adequate produce an estimate of the grain size below which the
chance to sample the range of grain sizes in motion. efficiency of the pit trap would be affected. This is
Sampling duration ranged from 20 minutes to 26 hours, important since the pit trap is effectively our reference
depending upon the filling rate of the HS sampler (it is for measurement.

Figure 5. Sampling events superimposed on the freshet hydrograph: 1991 season. Gauge was located
at the study site (see Figure 1 for gauge location).
19 - 6 STERLING AND CHURCH: SEDIMENT TRAPPING IN A GRAVEL BED RIVER

[14] The effect of circulation within the trap depends on have occurred in the traps we used for our study. From
whether there is one circulation cell, one circulation cell and equation (1), at trap 1B, the circulation might, then,
a dead zone, or more than one circulation cell. To determine relaunch grains of up to 3 mm size. At trap 3B, grains as
this condition we consider the aspect ratio a = h/d, wherein h large as 16 mm might theoretically be relaunched. We
is trap depth and d is trap diameter (slot length). In a review expect that these estimates are high.
of trap circulation in limnology [Hawley, 1988], values in [17] We also use velocity data to examine the saltation/
the range 3 < a < 5 are associated with the appearance of a suspension transition expected at the highest flows, and thus
dead water zone in the bottom of the trap; a < 3 would the lower limit of sizes apt to be caught in the pit trap.
indicate one circulation cell. Our trap has a = 1.3, implying a Results at 5 and 12 cm (measured simultaneously using a
single cell of water circulating within the trap (Figure 3b), stack of meters) yield shear velocity (v*) estimates of
with circulation velocity approximately equal to the near- 0.137 ± 0.040 m/s (2 standard error range; n = 7) at trap
bed flow velocity, v, immediately above the trap. 1B and 0.138 ± 0.134 m/s (n = 5) at trap 3B. Using the
[15] A grain not apt to be relaunched from within the trap convention for grain suspension, ws/v* < 1.0 (although a
must have FG > FD, FG being the submerged particle weight range of values to 1.25 has been cited [Komar, 1988]; ws
and FD the upwardly directed fluid drag force on the rising being the grain settling velocity), we infer that grains apt to
side of the circulation cell. FG = g(rsr)V, wherein g is the overpass the traps in suspension would be, on average, less
acceleration of gravity, rs is sediment density, r is the density than 1 mm in diameter and, at the upper limit of the
of water, and V is particle volume. FD = CDrv2A/2, wherein indicated distribution of shear velocities, less than 3.5 mm
CD is the drag coefficient, v, as above, is the near-bed flow in diameter. A maximum feasible size for suspended grains
velocity, and thus the velocity of the circulating cell, and A is appears, on this criterion, to be in the range 3 to 4 mm.
the projected area of the grain. Equating these forces under Grain size of suspended sediment samples collected in the
the assumption that we can approximate the grain by a study reach (unpublished) support the conclusion; the
sphere of equivalent volume and nominal diameter Dn, we maximum grain size of suspended sediment from the stream
obtain (Figure 2) was always less than 0.71 mm.
[18] Another check is available based on work on slot
Dn ¼ CD v 2 =21:6 ð1Þ length by Hubbell [1964] and by Poreh et al. [1970]. In
order to avoid overpassing of a pit trap by saltating
for rs = 2650 kg/m3. For CD = 1 (a value that we have found sediment, Hubbell suggested that the opening should be
from experiment to be reasonable for settling, natural sand at least 100D (on the basis of an estimate of particle step
grains), we find, for example, that at v = 0.2 m/s, a grain length made by Einstein [1944]), where D is the character-
larger than 2.0 mm is unlikely to be relaunched. This size istic size of the material to be trapped. As the opening
might in fact be increased by turbulent velocity fluctuations, diameter of our traps is 29 cm, Hubbell’s criterion suggests
and it probably is decreased by the limited probability to that our traps should retain all material up to 3 mm in
relaunch the particle entirely out of the circulation cell diameter if it is saltating. Poreh et al. [1970] reported that
within the trap into the passing flow. Settling experiments 100% efficiency was approached for the sizes they tested
using previously collected bed load samples support the when d  40D, a considerably less conservative criterion
relation; our measurements of the settling velocity of natural than that of Hubbell. Following Poreh et al. [1970], our
grains taken from the samples follow the relation closely for 29 cm maximum opening yields a 7.2 mm critical grain
D > 1 mm. size. However, the flow conditions we have investigated
[16] We use velocity data to apply equation (1) to fall substantially outside the range studied by Poreh et al.
estimate the largest circulation velocities that might be [1970], such that their results may not hold exactly in our
expected in the pit traps during the sampling events, and situation.
thus the largest particle size that might be relaunched. At [19] Poreh et al. [1970] further related trap efficiency for
the highest observed flows (15 – 18 m3/s), we measured a given slot to grain Froude number and grain size for grains
flow velocities using small axial-screw current meters at between 1.88 and 4.50 mm. For the smallest grains that we
5 cm above the streambed immediately beside traps 1B and theoretically expect to trap, given above as 3 mm and
3B (Figure 1), the two traps where the strongest currents are 16 mm, grain Froude number (based on mean flow veloc-
expected to occur. Five centimeters is about the top level of ities over the traps from the current meter stack, estimated to
the D75 of local bed material and was as close to the bed as be 0.8 m/s at trap 1B and 1.2 m/s at trap 3B) varies between
we could practically obtain readings. Individual values were 13.1 and 5.56, which is somewhat below the range inves-
established as the mean of between 14 and 29 successive tigated by Poreh et al. [1970]. From their equations we find
30-second measurements. Grand means at 5 cm were that our trap opening scales to non-dimensional opening
0.255 ± 0.145 m/s (2 standard deviation range; n = 5) at values greater than 9.0 for all trap and proposed limit grain
trap 1B, and 0.582 ± 0.195 m/s (n = 6) at trap 3B. The size combinations, well above the critical value of about 3.0
difference in means between the traps appears reasonable, for 100% trapping efficiency, so particles between 3 and
as the latter trap is much nearer the thalweg. Over the range 16 mm would certainly fall into the trap. Backsolving from
of flows measured, the systematic change in velocity was the limit scaled slot size of 3.0 (using equation (10) of
small and velocity actually declined as flow increased, Poreh et al. [1970]), we find that our 16 mm grains would
presumably as water depth increased. Since the velocity theoretically enter a 13 cm slot, while a 4 mm grain would
lines are somewhat depressed over the trap opening enter a 10 cm slot. Because the pit trap opening is circular
(Figure 3b), we suppose that these values reasonably the slot length ranges from 0 to 29 cm, but 13 cm chords
represent the largest velocities that might be expected to encompass 90% of the trap width. We conclude that the
STERLING AND CHURCH: SEDIMENT TRAPPING IN A GRAVEL BED RIVER 19 - 7

Table 1. Location and Duration of Sampling Events and Material Collection Ratesa
Standardized Collection Rate, Standardized collection rate Grain Size at
Event g m1 s1 Truncated at 0.25 mm, g m1 s1 Which 90% Pit
Duration, Discharge,b Trap Efficiency
Event Trap min m3 s1 Pit HS-P HS-I Pit HS-P HS-I Is Crossed, mm

1 3D 125 18.4 0.64 0.25 0.39 0.60 0.16 0.27 0.50 – 0.71
2 3C 53 17.7 3.40 0.85 1.75 3.36 0.65 1.48 0.71 – 1.0
3 3C 51 16.5 1.35 0.47 3.68 1.33 0.33 3.55 0.71 – 1.0
4 3C 31 15.3 0.74 0.21 1.05 0.72 0.13 0.83 2.0 – 2.8
5 3C 30 15.3 0.32 0.31 0.87 0.31 0.18 0.67 1.4 – 2.0
6 3C 32 14.8 1.04 0.062 0.51 1.02 0.026 0.36 0.35 – 0.50
7 3C 59 13.8 0.33 0.15 0.61 0.32 0.078 0.46 0.70 – 1.0
8 3C 208 12.8 0.15 0.067 0.12 0.14 0.021 0.069 0.70 – 1.0
9 3B 30 9.25 1.50 0.65 0.47 1.49 0.57 0.40 2.8 – 4.0
10 3B 57 8.93 2.41 0.23 1.63 2.41 0.19 1.55 2.0 – 2.8
11 3B 54 8.61 1.54 0.42 1.38 1.54 0.38 1.32 4.0 – 5.7
12 3B 55 8.30 0.39 0.32 0.33 0.39 0.27 0.28 4.0 – 5.7
13 3B 57 8.01 0.38 0.11 0.36 0.38 0.085 0.31 4.0 – 5.7
14 3B 85 7.72 0.14 0.078 0.16 0.14 0.053 0.13 8.0 – 11
15 3B 83 7.45 0.28 0.038 0.051 0.28 0.023 0.033 4.0 – 5.7
16 3B 173 7.19 0.37 0.072 0.053 0.36 0.058 0.040 2.0 – 2.8
17 3B 163 6.93 0.22 0.033 0.087 0.22 0.022 0.071 1.0 – 1.4
18 3B 701 6.93 0.18 0.013 0.022 0.18 0.0063 0.013 0.50 – 0.71
19 3B 503 6.69 0.12 0.014 0.029 0.12 0.0087 0.021 0.71 – 1.0
20 3B 907 6.98 0.10 0.020 0.035 0.10 0.015 0.029 2.0 – 2.8
21 3B 1473 5.15 0.0070 0.0045 0.031 0.0065 0.0019 0.028 1.0 – 1.4
22 3B 1560 4.18 0.0006 0.0017 0.0021 0.0003 0.0003 0.0006 2.8 – 4.0
a
Standardized collection rate means rate per unit width of the sampler/trap entrance. Entries in bold indicate ones in which the rate for HS-I exceeds that
of the combined pit plus HS-P. The catch of HS-I in event 21 is highly anomalous (the sampler, when recovered, contained incredible quantities of large
material), and this sample has been removed from summary calculations.
b
Measured at the beginning of the event.

traps approached 100% efficiency for all sizes not apt to be summary results median statistics are preferred because
relaunched. the small number of samples and individual eccentric results
[20] An independent indication of the limit size retained create some variability in means).
in our traps is provided in an analysis by Hassan and [23] The difference between the samplers’ collection rates
Church [2001] of fractional transport ratios for the Harris is striking when separated by grain size (Figure 7). The
Creek pit traps. They identified the grain size below which distribution of median sampler collection rates for the 21
fractional bed load transport evidently declined sharply, events over grain size resemble normal curves; the curve for
indicating sediment moving into suspension. At the highest the pit trap is shifted approximately 2 size classes to the left
flows observed, their median result was about 4 mm, and (coarser) relative to the HS-I curve. The pit trap has a higher
the extreme value indicated by an envelope under their capture rate than the HS-I in the 5.5 psi (45.3 mm) to 0.5 psi
results indicated a value of about 16 mm, in agreement with (1.41 mm) range. For the size ranges 0 psi (1 mm) to 4 psi
our calculations. (0.062 mm) HS-I has a higher capture rate. Figure 7 also
[21] In summary we estimate that, during the highest flows demonstrates that the modal size of trapped material (in the
experienced within the sampling period, the upper limit of range 0.5 to 2 mm), in fact the size of virtually all the
grain sizes apt to be relaunched from the pit trap is 3 – 16 mm; material apparently transported is smaller than the median
and the maximum size of material in suspension is 3– 4 mm. size of the bed material and much smaller than the median
For mean flow conditions during the sampling period, we size of the bed surface material. This phenomenon has
estimate that the effective opening width is 90% of the total previously been noted by Dietrich and Whiting [1989],
opening for a 16 mm grain, and 94% for a 4 mm grain. From Leopold [1992], and Lisle [1995].
these indications, pit trap efficiency should begin to decline [24] Next, we compare pit and HS-I with the summed
at approximately 3 – 4 mm (from high efficiency at larger ‘‘actual bed load transport’’. The pit trap has a median
grain sizes to lower efficiency at smaller grain sizes). 83% efficiency (standard deviation ±12%), while the
median efficiency of HS-I is 56% (standard deviation
±49%). Ryan and Porth [1999] compared annual collection
3. Observations and Discussion rates of bed load collected in a cobble-bed stream by several
[22] Our observations are summarized in Table 1. First samplers, including HS, and accumulation in a weir pond.
we compare pit and HS-I collection rates directly, and find Predicted rates of bed load transport from the HS catch
that collection rates of HS-I are substantially lower than summed to 55% of the weir pond bed load collection. While
those of the pit trap (Table 1 and Figure 6). The median HS- the similarity between this statistic and ours is striking,
I catch rate is 47% of the pit trap for 21 sampling events uncertainties introduced into the former result by integration
(standard deviation ±69%; an outlier, HS-I sample 21, is over large spans of time and flow prevent straightforward
removed from all analyses; in this and the following comparison.
19 - 8 STERLING AND CHURCH: SEDIMENT TRAPPING IN A GRAVEL BED RIVER

Figure 6. Comparison of gross collection rates for HS-I and pit trap. Samples are truncated at 0.25 mm
to eliminate the expected portion of the sample transported in suspension.

Figure 7. Median of collection rates recorded over all 22 events for individual 0.5 psi grain size classes.
The results represent material retained on the indicated screen; hence that is larger than the psi reference
number. Millimeter size scale is given at the top of the diagram.
STERLING AND CHURCH: SEDIMENT TRAPPING IN A GRAVEL BED RIVER 19 - 9

and 2 psi (4 mm), and then declines to less than 30%


between 2 psi (4 mm) and 2.0 psi (0.25 mm). The
consistent sigmoidal curve allows for the determination of
the grain size at which the pit trap efficiency was reduced in
our experiments, using the statistic ‘‘grain size where 90%
efficiency is crossed’’ (Table 1). Pit trap efficiency declines
between – 1 and +2 psi (0.5 – 4 mm). Comparing these
values with the a priori estimates, we find the upper
envelope of pit trap efficiency decline (4 mm) matches
closely the lower predicted value.
[26] The pattern of estimated efficiency by grain size
class of HS-I is opposite that of the pit trap (Figure 8a). HS-
I efficiency is 100% or greater for the small grain sizes (4
to 1 psi) (0.062 to 0.5 mm), then decreases substantially
for material larger than 1 psi in a pattern that is not
consistent for the 21 analyzed events. In our measurements,
there is the appearance that the HS sampler over-registered
sizes below 0.35 mm (1.5 psi; Figure 8a); essentially all of
this material likely was suspended at the time of trapping.
The HS sampler performed poorly at catching material
larger than 0.71 mm, much of which is likely saltating,
rolling, or sliding.
[27] The variability of HS-I is high at site 3C between 3.5
and 0 psi, the same range as that of substantially low
efficiency (Figures 8b and 8c). The change in variability
may be a consequence of sporadic transport of the larger
grain sizes on the bar top (trap 3C site), or of bed condition
(for sampler placement). The largest spread in the pit trap
efficiencies lies between 0.5 and 2 psi, the range of
rapidly declining efficiency of the trap; however, substantial
variability persists up to 1.0 psi at site 3B, where lower
flows and generally more sporadic movement of gravels
were experienced.
[28] In striking contrast to our results, Emmett [1980]
reported that the HS sampler had 100% trapping effi-
ciency for grains ranging from 0.5 to 16 mm. This
difference is revealing considering that Emmett [1980]
tested the HS sampler in a sand/granule gravel bed,
conforming with the bed material for which the sampler
was designed. We propose that a major cause of the
reduced efficiency of the HS sampler is the elevated
bottom lip above the bed in coarse-grained gravel bed
channels. Against cobbles, the stiff, flat lower lip of the
sampler does not touch the rough bed in most places and
therefore there is a high probability for there to be gaps so
that smaller rolling particles pass under the sampler. This
condition would be made worse by the discrepancy
between bed surface grain size and the much smaller size
of the grains in motion. In contrast, a sand/granule gravel
Figure 8. Estimated efficiency by grain size: (a) median bed would generally provide favorable conditions for
efficiency for all events; (b) efficiency ranges at traps 3C/3D seating the stiff sampler firmly on the bed.
(bar top). Results from the two traps are combined because [29] Other causes for reduced HS efficiency in trapping
only one measurement was made at trap 3D. (c) Efficiency coarse grains may include its lower lip thickness, its
ranges at trap 3B (bar edge near thalweg). Efficiency is narrower width, misalignment with the current, and the
calculated for the pit trap and for the HS-I sampler relative opening size. As described by Novak [1957] the 6.3 mm
to the summed catch of the pit trap plus the HS-P sampler. height of the bottom edge of the HS sampler reduced its
efficiency, particularly for grain sizes equal to or smaller
than the height of the step; Pitlick [1988] and, more
[25] The estimated efficiency of the pit trap over the recently, Ryan and Porth [1999] showed that a HS sampler
range of grain sizes follows a consistent sigmoidal curve with a 1.5 mm thick lower lip catches nearly twice the
for all 22 sampling events (Figure 8a). It has effectively amount of material as does the original 6.3 mm thick
100% efficiency for grain sizes between 5.5 psi (45.3 mm) design. Given that the HS sampled a narrower bed width
19 - 10 STERLING AND CHURCH: SEDIMENT TRAPPING IN A GRAVEL BED RIVER

than the pit trap, and given the sporadic movement of the different GSDs returned. Inasmuch as the HS sampler
larger material, there is a considerably reduced probability systematically caught finer material than the pit trap, we
for larger sizes (often represented by one or a few clasts) to are left with the conclusion that HS sampler catch system-
enter the HS sampler than the pit trap during the sample atically underestimates the grain size and transport rate of
period. Gaudet et al. [1994] found a rapid loss of catching bed load in coarse-grained gravel bed rivers, unless the
efficiency occurred when the sampler is misaligned with the sampler is deployed on a specially constructed sill [e.g.,
current. A possible source of error is that the particle sizes in Johnson et al., 1977; Kuhnle, 1992]. Inasmuch as a sub-
transport might be larger than the HS opening size. Such stantial proportion of all available bed load measurements in
material is present in the bed but does not appear to have gravel rivers have been obtained using HS samplers
moved during our sampling events; the largest material deployed on the natural streambed, we expect that there is
trapped by the pit (a single particle in the 44– 64 mm range) a systematic bias in the literature of GSD statistics and,
remained smaller than the HS-I opening (73  73 mm). possibly, bed load transport rates (depending on how the
[30] Why HS-I trapped more fine sediment (by up to 20% grain size of the reported bed load sample was truncated).
in the finest sizes) than the reference ‘‘actual rate’’ is not Given the emerging importance of specifying GSD for the
clear. The HS sampler systematically traps low flying mobile population in studies of transport over gravel, and
suspended sediment because of its 7.6 cm height. The HS the importance of bed load transport rates in general, this is
may also trap more fine sediment because of over-register- a serious problem.
ing from the high hydraulic efficiency of the sampler, which [33] We emphasize that we do not know the true bed load
is about 1.5 [Druffel et al., 1976]. However, the results transport rate at our sampling devices. We do know, how-
mean that HS-I captured more fines than the nearby HS-P. ever, that it was a good deal more than usually indicated by
Limitations in the sampling arrangement that may have the HS sampler, a result in conformity with other studies of
influenced the result include (1) the fact that the two HS performance of the original design in gravel channels. In
samplers were essentially measuring two different streams comparison, it appears as if the pit trap caught most of the
of sediment load due to spatial displacement; (2) turbulence bed load. We also emphasize that we do not know whether
or systematic velocity variations generated by the presence or not the efficiencies quoted in this paper are biased, since
of the pit trap may have altered the trapping performance of we used the pit trap itself as the source of the major part of
HS-P in comparison with HS-I; and (3) the sometimes the reference figures.
elevated rim of the pit trap may have reduced the efficiency [34] The appearance that the efficiency of the pit trap is
of the pit trap/HS-P combination in comparison with HS-I. limited principally by the probability for particles to be
However, in rebuttal it can be pointed out that the real relaunched raises an interesting possibility for the definition
variation in sediment streams appears to be low judged by of what constitutes ‘‘bed load’’. While the idea of bed load
the consistency of results among the three trap locations, is conceptually clear, the practical definition remains con-
and the pit trap rim might be expected to have least effect on strained by the propensity for certain sizes, usually within
the finest grains, most of which were probably suspended, the sand range, to be intermittently suspended. Since the
while the HS-I sampler entrance appears to have had step possibility to relaunch material from the trap is related to
problems of its own. near-bottom stream velocity, which also determines the
propensity for material to be launched off the streambed,
it appears as if it would be possible to define bed load
4. Conclusions consistently in terms of what is trapped within a pit with
[31] Key findings in this study are that the HS sampler some specified aspect ratio.
and the pit trap sample different components of load over a [35] As the result of our study, we recommend that HS
gravel bed and that the HS sampler catches substantially samplers not be used in coarse-grained gravel bed rivers
less material than the pit trap. Thus, the HS sampler and the unless an artificial sill is available to firmly seat the
pit trap each intercept moving particles in a manner that device. Other studies [Pitlick, 1988; Ryan and Porth,
prefers particular grain sizes, from which we infer that they 1999] suggest that thin-walled designs may perform in a
are biased toward a particular transport process. The pit trap less systematically biased way than our original device.
appears to reliably trap rolling and sliding material, and Nevertheless, they remain subject to most of the sources of
grains making short saltation hops, while it may miss error we have discussed. A pit trap appears to be more
material taking long saltation steps. The HS sampler, in reliable than the HS sampler in that its errors are consis-
contrast, appears to efficiently sample low-flying sand and, tent, and thus may be corrected. This would not easily be
more variably, saltating, rolling and sliding material. It accomplished for the HS sampler because of the high
appears not to intercept much of the coarsest material in variability in the sampler errors. Furthermore, pit traps can
sporadic motion. Thus the pit more consistently catches be instrumented to record actual transport rates by con-
more of the true bed load in transport; this empirical result is tinuously weighing the accumulating sample [Reid et al.,
supported by three independent checks of expected grain 1980]. Depending on design, however, there may be limits
size at which efficiency declines. (In contrast to our find- to the flow depth in which pits and related slot-type
ings, however, Ryan and Porth [1999] reported reasonable samplers can be operated.
agreement between GSDs trapped by HS samplers and the [36] We finally note that pit traps and HS samplers trap
GSD derived from the material trapped in a weir pond, on different components of the total sediment load and,
an annually integrated basis.) therefore, the transport rates they measure in gravel bed
[32] The fact that the two devices systematically trap streams are not directly comparable in any case. Depend-
different components of the sediment load leads to the ing upon the question asked, either device may be an
STERLING AND CHURCH: SEDIMENT TRAPPING IN A GRAVEL BED RIVER 19 - 11

appropriate tool if it is deployed in a manner consistent Hubbell, D. W., Apparatus and techniques for measuring bedload, U.S.
Geol. Surv. Water Supply Pap., 1748, 74 pp., 1964.
with its design. Hubbell, D. W., Bed load sampling and analysis, in Sediment Transport in
Gravel-bed Rivers, edited by C. R. Thorne, J. C. Bathurst, and R. D. Hey,
[37] Acknowledgments. This work was supported by the Natural pp. 89 – 106, John Wiley, New York, 1987.
Sciences and Engineering Research Council of Canada. We thank Ken Johnson, C. W., R. L. Engleman, J. P. Smith, and C. L. Hanson, Helley-
Christison, Michel Lapointe and H. Rohde for helpful comments and ideas, Smith bed load samplers, J. Hydraul. Div. Am. Soc. Civ. Eng., 103,
Marwan Hassan for supervision of the field work, Scott Babakaiff, Brett 1217 – 1221, 1977.
Eaton and Hou Tsizhui for field assistance, and Paul Jance for preparing the Jones, M. L., and H. R. Seitz, Sediment transport in the Snake and Clear-
diagrams. The journal’s referees provided constructive critiques incorporat- water Rivers in the vicinity of Lewiston, Idaho, U.S. Geol. Surv. Water
ing much additional experience that helped us to improve the report Resour. Invest. Open File Rep., 80 – 690, 179 pp., 1980.
substantially. Komar, P. D., Sediment transport by floods, in Flood Geomorphology,
edited by V. R. Baker, R. C. Kochel, and P. C. Patton, pp. 97 – 111,
Wiley-Interscience, New York, 1988.
References Kuhnle, R. A., Fractional transport rates of bedload on Goodwin Creek, in
Andrews, E. D., Marginal bed load transport in a gravel bed stream, Sage- Dynamics of Gravel-bed Rivers, edited by P. Billi et al., pp. 141 – 155,
hen Creek, California, Water Resour. Res., 30, 2241 – 2250, 1994. John Wiley, New York, 1992.
Church, M., J. F. Wolcott, and K. Fletcher, A test of equal mobility in Leopold, L. B., The sediment size that determines channel morphology, in
fluvial sediment transport: Behavior of the sand fraction, Water Resour. Dynamics of Gravel-bed Rivers, edited by P. Billi et al., pp. 297 – 311,
Res., 27, 2941 – 2951, 1991. John Wiley, New York, 1992.
Dietrich, W. E., and P. Whiting, Boundary shear stress and sediment trans- Lisle, T. E., Particle size variations between bed load and bed material in
port in river meanders of sand and gravels, in River Meandering, Water natural gravel bed channels, Water Resour. Res., 31, 1107 – 1118, 1995.
Resour. Monogr., vol. 12, edited by S. Ikeda and G. Parker, pp. 1 – 50, Novak, P.. Bed load meters development of a new type and determination
AGU, Washington, D.C., 1989. of their efficiency with the aid of scale models, paper presented at 7th
Druffel, L., W. W. Emmett, V. R. Schneider, and J. V. Skinner, Laboratory General Meeting, Int. Assoc. for Hydraul. Res., Lisbon, 1957.
hydraulic calibration of the Helley-Smith bed load sediment sampler, Pitlick, J., Variability of bed load measurements, Water Resour. Res., 24,
U.S. Geol. Surv. Open File Rep., 76 – 752, 63 pp., 1976. 173 – 177, 1988.
Einstein, H. A., Bedload transportation in Mountain Creek, Tech. Bull. U.S. Poreh, M., A. Sagiv, and I. Seginer, Sediment sampling efficiency of slots,
Dep. Agric., 55, 1944. Proc. Am. Soc. Civ. Eng. Hydraul. Div., 96, 2065 – 2078, 1970.
Emmett, W. W., A field calibration of the sediment-trapping characteristics Reid, I., J. T. Layman, and L. E. Frostick, The continuous measurement of
of the Helley-Smith bedload sampler, U.S. Geol. Surv. Prof. Pap., 1139, bedload discharge, J. Hydraul., 18, 243 – 249, 1980.
44 pp., 1980. Ryan, S. E., and L. S. Porth, A field comparison of three pressure-difference
Gaudet, J. M., A. G. Roy, and J. L. Best, Effect of orientation and size of bedload samplers, Geomorphology, 30, 307 – 322, 1999.
Helley-Smith sampler on its efficiency, J. Hydraul. Eng., 120, 758 – 766, Tritton, D. J., Physical Fluid Dynamics, Oxford University Press, New
1994. York, 1988.
Hassan, M., and M. Church, Sensitivity of bedload transport in Harris
Creek: Seasonal and spatial variation over a gravel bar, Water Resour.
Res., 37, 813 – 825, 2001. 

Hawley, N., Flow in cylindrical sediment traps, J. Great Lakes Res., 14, M. Church, Department of Geography, University of British Columbia,
76 – 88, 1988. Vancouver, British Columbia, Canada V6T 1Z2. (mchurch@geog.ubc.ca)
Helley, E. J., and W. Smith, Development and calibration of a pressure- S. M. Sterling, Northwest Hydraulic Consultants, 2-40 Gostick Place,
difference bedload sampler, U.S. Geol. Surv. Open File Rep., 18 pp., North Vancouver, British Columbia, Canada V7M 3G2. (shannon@duke.
1971. edu)

View publication stats

You might also like