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

net/publication/250081957

An Interpretation of Trends in Grain Size Measures

Article in Journal of Sedimentary Research · January 1981

DOI: 10.1306/212F7CF2-2B24-11D7-8648000102C1865D

CITATIONS READS

344 2,448

1 author:

Patrick Mclaren
SedTrend Analysis Limited
56 PUBLICATIONS 1,728 CITATIONS

SEE PROFILE

All content following this page was uploaded by Patrick Mclaren on 29 May 2015.

The user has requested enhancement of the downloaded file.

 THE EFFECTS OF SEDIMENT TRANSPORT ON GRAIN.SIZE DISTRIBUTIONS'
PATRICK MCLAREN
Geological Sumey of Canada
P aciftc G eo s c ie nce C e ntre
P.O. Box 6000
Sidney, B.C. V8L 482, Canada
AND
DONALD BOWLES
Box 179, R.R. I
Chelsea, Quebec fOX 1N0, Canada

AssrRAc'r: Changes in statistics (mean, sorting, and skewness) describing grain-size distributions have long been used to speculate
on the direction of sediment transport. We present a simple model whereby the distributions of sediment in transport are related
to their source by a sediment transfer function which defines the relative probability that a grain within each particular class interval
will be eroded and transported. A variety ofempirically derived transfer functions exhibit negatively skewed distributions (on a phi
scale). Thus, when a sediment is oe-ing eroded, the probability of any grain going into transport increases with diminishing grain
size throughout more than half of its size range. This causes the sediment in transport to be finer and more negatively skewed than
its source, whereas the remaining sediment (a lag) must become relatively coarser and more positively skewed.
Flume experiments show that the distributions oftransfer functions change from having a highly negative skewness to being nearly
symmetrical (although still negatively skewed) as the energy of the transporting process increases. We call the two extremes /ow-
energy and high-energy transferfunctions, respectively. In an expanded sediment-transport model, successive deposits in the direction
of transport are related by a combination of two transfer functions. If energy is decreasing and the transfer functions have low-
energy distributions, successive deposits w.ill become finer and more negatively skewed. If, however, energy is decreasing, but the
initial transfer function has a high-energy distribution, successive deposits will become coarser and more positively skewed.
The variance of the distributions of lags, sediment in transport, and successive deposits in the down-current direction must
eventually decrease (i.e., the sediments will become better sorted). We demonstrate that it is possible for variance first to increase,
but suggest that, in reality, an increasing variance in the direction oftransport will seldom be observed, particularly when grain-size
distributions are described in phi units.
This model describing changes in sediment distributions was tested in a variety of environments where the transport direction
was known. The results indicate that the model has real-world validity and can provide a method to predict the directions of
sediment transport.

INTRODUCTION functions involving many variables, including grain ori-

The environmental interpretation of grain-size distri-
entation and the structure of turbulence, which are in-
butions found in sedimentary deposits has been, and still
trinsically random (Gessler 1976). For example, there is
substantial disagreement as to the actual velocities re-
is, a fundamental goal of sedimentology. Ever since quired to erode and transport particles of a given size,
Udden's work in l9l4 it has been recognized that sedi- particularly when the bed material is poorly sorted (Singer
ment size fractions approximate a log-normal distribu-
and Anderson 1984). Experimental difficulties appear to
tion. In reality, however, most sediments do not strictly
be as true today as in 1950 when Einstein wrote, "The
follow log-normality and deviations from the Gaussian
forces acting on individual particles of a natural sediment
model have been given various enyironmental interpre-
mixture in a bed cannot very well be measured" (Einstein
tations (Folk and Ward 1957; Mason and Folk 1958;
1950, p. 35). Both Slingerland (1977) and Singer and
Friedman 1961,1979). Other workers have differentiated
Anderson (1984) provide good summaries of the com-
log-normal subpopulations defined by distinct breaks plexities encountered by workers who have sought to doc-
within the complete size range ofthe sample (Moss 1962;
Visher I 9 6 9; Glaister and Nelson 197 4; Middleton 19 7 6).
ument the processes of entrainment and deposition of
sediments.
Each subpopulation is interpreted as representative ofa
The second difrculty lies in the imprint of the source
specific U"ansport mode (traction, intermittent suspen-
sion, and suspension), the relative concentrations ofeach
sediment characteristics on the characteristics of the de-
posit. This was recognized as long ago as 1938 when
suggesting particular depositional environments.
Although Blatt et al. (1980) argue that it should not be Krumbein suggested the importance of progressive or
continuous changes in grain-size distributions from source
surprising for specific process populations to exist because
different transport mechanisms differ in the way they se- to final deposit (Krumbein 1938). Progressive changes
have been recognized by several workers (Stapor and Tan-
lect grains for movement or deposition, many other au-
thors are unable to achieve similar interpretations (Gar-
ner 1975; McCave 1978; Haner 1984) and have been
analyzed in a deductive model by Mclaren (1981). He
row 1982; Flemming 1982; Anderson et al. 1982). There
suggested that the mean, sorting, and skewness of grain-
are at least two reasons for such disagreements. First, the
processes of erosion and deposition encompass complex size frequency distributions follow trends that identify
the direction of transport and the sedimentary processes
of winnowing, selective deposition, and total deposition.
t Manuscript received 8 February 1984; revised l6 November 1984. Using a hypothetical sediment distribution and an as-

JounNer- or SeoIueNr.rrry Prrnoroov, Vol. 55, No. 4, Julv, 1985, p. 04574470

Copyright @ 1985, The Society ofEconomic Paleontologists and Mineralogists OO22-4472/85/0055-0457/$03.00
 THE EFFECTS OF SEDIMENT TRANSPORT ON GRAIN.SIZE DISTRIBUTIONS1
PATRICK MCLAREN
Geological Sumey of Canada
Pacifrc Geoscience C entre
P.O. Box 6000
Sidney, B.C. V8L 482, Canada
AND
DONALD BOWLES
Box 179, R.R. I
Chelsea, Quebec JOX 1N0, Canada

AssrRAcr: Changes in statistics (mean, sorting, and skewness) describing girain-size distributions have long been used to speculate
on the direction of sediment transport. We present a simple model whereby the distributions of sediment in transport are related
to their source by a sediment transfer function which defines the relative probability that a grain within each particular class interval
will be eroded and transported. A variety ofempirically derived transfer functions exhibit negatively skewed distributions (on a phi
scale). Thus, when a sediment is being eroded, the probability ofany grain going into transport increases with diminishing grain
size throughout more than half of its size range. This causes the sediment in transport to be finer and more negatively skewed than
its source, whereas the remaining sediment (a lag) must become relatively coarser and more positively skewed.
Flume experiments show that the distributions oftransfer functions change from having a highly negative skewness to being nearly
symmetrical (although still negatively skewed) as the energy of the transporting process increases. We call the two extremes /ow-
energy and high-energy transferfunctions, respectively. In an expanded sediment-transport model, successive deposits in the direction
of transport are related by a combination of two transfer functions. If energy is decreasing and the transfer functions have low-
energy distributions, successive deposits will become finer and more negatively skewed. If, however, energy is decreasing, but the
initial transfer function has a high-energy distribution, successive deposits will become coarser and more positively skewed.
The variance of the distributions of lags, sediment in transport, and successive deposits in the down-current direction must
eventually decrease (i.e., the sediments will become better sorted). We demonstrate that it is possible for variance first to increase,
but suggest that, in reality, an increasing variance in the direction oftransport will seldom be observed, particularly when grain-size
distributions are described in phi units.
This model describing changes in sediment distributions was tested in a variety of environments where the transport direction
was known. The results indicate that the model has real-world validity and can provide a method to predict the directions of
sediment transport.

INTRODUCTION functions involving many variables, including grain ori-

The environmental interpretation of grain-size distri-
entation and the structure of turbulence, which are in-
trinsically random (Gessler 1976). For example, there is
butions found in sedimentary deposits has been, and still
substantial disagreement as to the actual velocities re-
is, a fundamental goal of sedimentology. Ever since quired to erode and transport particles of a given size,
Udden's work in l9l4 it has been recognized that sedi- particularly when the bed material is poorly sorted (Singer
ment size fractions approximate a log-normal distribu-
and Anderson 1984). Experimental difficulties appear to
tion. In reality, however, most sediments do not strictly
be as true today as in 1950 when Einstein wrote, "The
follow log-normality and deviations from the Gaussian
forces acting on individual particles of a natural sediment
model have been given various environmental interpre-
mixture in a bed cannot very well be measured" (Einstein
tations (Folk and Ward 1957; Mason and Folk 1958;
1950, p. 35). Both Slingerland (1977) and Singer and
Friedman 1961,1979). Other workers have differentiated
Anderson (1984) provide good summaries of the com-
log-normal subpopulations defined by distinct breaks plexities encountered by workers who have sought to doc-
within the complete size range ofthe sample (Moss 1962;
Visher 1 9 6 9; Glaister and Nelson I 9 7 4; Middleton I 9 7 6).
ument the processes of entrainment and deposition of
sediments.
Each subpopulation is interpreted as representative ofa
The second difficulty lies in the imprint of the source
specific tfansport mode (traction, intermittent suspen-
sion, and suspension), the relative concentrations ofeach
sediment characteristics on the characteristics of the de-
posit. This was recognized as long ago as 1938 when
suggesting particular depositional environments.
Although Blatt et al. (1980) argue that it should not be Krumbein suggested the importance of progressive or
continuous changes in grain-size distributions from source
surprising for specific process populations to exist because
different transport mechanisms differ in the way they se- to final deposit (Krumbein 1938). Progressive changes
have been recognized by several workers (Stapor and Tan-
lect grains for movement or deposition, many other au-
thors are unable to achieve similar interpretations (Gar-
ner 1975; McCave 1978; Haner 1984) and have been
row 1982; Flemming 1982; Anderson et al. 1982). There analyzed in a deductive model by Mclaren (1981). He
suggested that the mean, sorting, and skewness of grain-
are at least two reasons for such disagreements. First, the
processes of erosion and deposition encompass complex size frequency distributions follow trends that identify
the direction of transport and the sedimentary processes
of winnowing, selective deposition, and total deposition.
I Manuscript received 8 February 1984; revised 16 November 1984. Using a hypothetical sediment distribution and an as-

Jounrlrr or Spotuen-renv Perxolocv, Vol. 55, No. 4, Jur-v, 1985, p. 04574470

Copyright 6 1985, The Society ofEconomic Paleontologists and Mineralogists OO22-4472/85/0055-0457/$03.00
458 PATRICK MCI'AREN AND DONALD BOWLES

sumption that lieht grains have a greater probability of and

being eroded and transported than heavy grains, the mod-
el demonstrated that 1) sediment in transport must be k':
finer, better sorted, and more negatiYely skewed than its
source sediment; 2) a lag must become coarser, better ) e(sJ(l - t(sJ)
sorted, and more positively skewed; and 3) successive
deposits may become finer or coarser, but the sorting must As well as using distribution functions to characterize
become better and the skewness more positive. sediment size, we use t(s) in a similar manner to describe
In this paper we attempt to scrutinize the validity of statistically all of the processes which tend to move sed-
these trends in a more rigorous examination than the iment from one location to another. This function is de-
strictly deductive approach used by Mclaren (1981). We scribed in the same manner as a size distribution function
present a more refined model that demonstrates how grain- (i.e., by weight proportion, grain frequency, etc.), and each
size distributions of sedimentary deposits change in the t(s1 ) gives the probability oftransferring grains ofsize "si'
direction oftransport. The paper does not, however, pro- from a source (e(s)) into transport (r(s)). We propose to
vide solutions or increase our understanding oftransport call each t(si ) a transfer cofficient and t(s) a sediment-
and depositional processes; rather,,it recognizes that the transfer function. The latter may be thought of as a func-
nature ofthese processes is probabilistic, and their results tion that incorporates all sedimentary and dynamic pro-
should be reflected in the relative changes of grain-size cesses that result in initial movement and transport of
distributions found in interrelated sedimentary deposits. particular grain sizes during a period of time.
The model is tested in a variety of environments, and we Implicit in the earlier model proposed by Mclaren
propose a technique to interpret the observed changes in ( 1 98 1) was the assumption that light grains have a gxeater
grain-size distributions for the purpose of establishing probability ofbeing transferred from a source into trans-
sediment-transport directions. port than heavy grains. This suggests that t(s) is a function
which increases (on a phi scale) or decreases (on a mil-
limeter or linear scale) monotonically; that is, smoothly
INITIAL SEDIMENT TRANSPORT MODEL
and continuously.2 Although this concept of t(s) is a sim-
Let us consider any grain-size distribution (g(s)) (Fig. plification, the model derived from this assumption has
1). If eroded, the sediment which is transported has a new proven to be effective in practice for determining sedi-
distribution (r(s)) which is derived from g(s) according to ment-transport paths and the relationships among inter-
a function t(s) so that related sedimentary deposits (see, for example, Mcl-aren
r(q) : kg(s,)t(st) et al. l98l; Mclaren 1981, 1982, 1983, 1984).
If we assume that t(s) is monotonically increasing, it
or can be shown that the sediment in transport (r(s)) is always
finer and more negatively skewed than its source (g(s))
t(s,) : tll and that the lag (d(s)) is always coarser and more posi-
K8(si)
tively skewed than its source (g(s)). The trend in sorting
where g(s,) and r(s,) define the proportion of the sediment depends on the initial sediment distribution of the source
in the Ih grain-size class interval for each of the sediment and therefore cannot be determined for any g(s). Sorting
distributions. k is a scaling factor that normalizes r(s) so must, however, eventually become better in both the sed-
that iment in transport and in the lag. The proof for these
N results is not essential to the text of this paper and is
r(s) : 1. shown in the Appendix.
i-l

Thus, EXPANDED SEDIMENT.TRANSPORT MODEL

I We have shown in the Appendix how grain-size dis-

k: ' tributions of sediment in transport and the remaining lag
" deposit must change relative to a common source sedi-
) e(sJt(s)
ment, assuming that the sediment-transfer function is
monotonically increasing. We now wish to (l) expand the
With the removal of r(s) from g(s), the remaining sed- simplified model shown in Figure l, and (2) examine the
iment (a lag) has a new distribution denoted by d(s) (Fie.
validity of the assumption (i.e., the nature of t(s) in light
l), where of empirical data).
d(sJ:k'g(s')(l -t(sJ)
of 2 The discussion on the use of phi versus metric continues (see
McManus 1982). We find that the log-normal transformation to phi
t'(s,):ffi, provides better statistical descripton (i.e., mean, sorting, and skewness)
for the identification ofsediment-transport directions than those based
where on a linear scale. The remainder ofthe paper uses the phi scale exclu-
t'(st):l-t(si) sively.
 SE D I ME NT TRANS P O RT AN D G RA I N - S I ZE D I STRI B UT IO N S 459

@-tb)*l-r(s)-l
I
t'($it-t(s)

FIc. l.-Diagrammatic summary of the initial sediment-transport

model discussed in the text and the Appendix.

Consider a unique sediment source such as an eroding

cliff of unconsolidated sediments with a grain-size dis-
tribution g(s) (Fig. 2). Eroded sediments are deposited in
a down-current direction, forming a beach whose grain-
FGI-,ot"r @
r-tjt"r
size distributions are dr(s), dr(s), dr(s), . . . , respectively.
The sediments in transport are denoted by rr(s), rz(s), . . . . I
Let the first transport function (to(s)) be equal to I for all
"s" so that r,(s) : g(s). This is analogous to a sudden
mass wasting event on the cliffface whereby the complete
@
x (s)
distribution (g(s)) is momentarily in transport. This dis-
tribution (r'(s)) is then acted upon by a process repre- Frc. 2. -Diagrammatic summary of the expanded sediment-transport
model. The three enclosed boxes are analogous to the initial model
sented by the function tt(s) which results in a new dis- shown in Figure l. See text for definition of terms.
tribution in transport, rr(s). The sediment remaining is
deposited as d'(s) which is related to rt(s) by the function
I - tr(s). Similarily, r2(s) is acted upon by t2(s) with the
result that dr(s) is deposited. dr(s) : kd,(s)t,(s)(l - t'(s))
Any three boxes forming an equivalent pattern to those I- tr(s)
considered in Figure I (e.g., r,(s), rr(s), and dt(s); Fig.2) kd,(s)X(s),
can be analyzed in the manner outlined in the previous
section and the Appendix using the assumption that t(s)
is monotonically increasing. Therefore, sediments in
transport (rr(s), r2(s), r:(s), . . . , etc.) must become pro-
gressively finer and more negatively skewed, and each
x(s):ffi tzl

d(s) can be considered a lag of its corresponding sediment X(s) is a function which combines the effects ofthe two
in transport (r(s)). Thus, for example, dr(s) is coarser and transfer functions t,(s) and t2(s). As such, X(s) may also
more positively skewed than rr(s). be considered a transfer function in that it provides the
We now wish to determine the relative changes in sed- statistical relationship between two sequential deposits.
iment distributions among the sequential deposits dt(s), Similar to t(s), this function incorporates all of the pro-
dr(s), dr(s), . . . , bearing in mind that r(s), t(s), and (fre- cesses responsible for sediment transport and deposition
quently) g(s) are not observable. resulting in a sequence of sedimentary deposits over the
kt us suppose that dt(s) is related to dr(s) by a function period of time represented by the samples. Therefore, the
X(s), so that relative change in the distributions between d2(s) and dt(s)
' d'(s) : kd'(s)X(s) is dependent on the shape of the function X(s) which can
be determined by examining empirically derived t(s)
functions.
where
k: N The Shape of t(s)
sZ,t d,(s)X(q) Transfer functions were calculated from data in Day
i:l
(1980), Emmett et al. (1980), Ghosh et al. (1979) and
or Gibbs and Neill (1972). These data sets all produced sim-
dls) ilarily shaped curves in spite of a wide range of grain-size
x(s): kd,(s) distributions; however, for illustrative purposes we will
use data from flume experiments described by Day (1980),
As illustrated in Figure 2, dt(s) can also be related to dt(s) which are the most complete for our purposes.
by These experiments were conducted in a 2.46-m-wide
 460 PATRICK MCI.AREN AND DONALD BOWLES

Tesln l.-Dlitribution of sediment in transport with increasingflow rates (m s-I) and the respective transferfunctions (datafrom Day 1980,
Series A)

S€diment in Tmnspo( (wt %)

B€d Tmsfer Fuctions (sdiment-tmsfer @msiedts)
Grain Siz€ Material :
o wt% k(q) d$/g(q) (from Eqution l)
(o.soti, ) to.or'i' (o.oorir s,) to.z+? r'l t^
" "-'1
'."\' 4.oo t.42 0.1I 0.t2 0.09 0.07 0.08 0.08 0.06 0.06
3.00 1.30 1.05 0.7 | 0.75 0.45 0.81 0.54 0.57 0.35
' 2.50 4.46 8.00 6.0r s.22 3.31 t.79 1.35 t.t7 0.74
2.00 11.00 25.93 19.26 15.45 10.16 2.36 1.75 1.40 0.92
1.50 5.68 13.57 9.94 7.92 5.22 2.39 1.7 5 1.39 0.92
r.25 5.75 10.55 8.58 6.99 5.18 1.83 1.49 t.2t 0.90
1.00 4.31 7.33 6.33 5.38 3.84 1.70 1.47 1.25 0.89
0.75 2.69 5.2t 4.38 3.93 2.99 1.93 1.63 t.46 l.l I
' 0.50 2.56 4.09 4.03 3.87 3. 18 1.60 1.57 1.51 1.24
" 0.25 2.21 3.21 3.46 3.47 3. l8 1.45 1.57
l8
t.57
r.25
1.44
t.32
0.00 4.98 4.73 5.87 6.24 6.s6 0.95 1.

-0.50 3. l0 2.35 3.r6 3.49 4.15 0.76 t.o2 t.t2 1.34

6.96 4.64 6.73 7.48 9.18 0.67 0.97 t.o7 r.32
-0.75 t.2t
10.10 5.19 8.87 9.76 t2.21 0.51 0.88 0.97
-1.25 1.28
9.88 4.30 7.49 9.53 t2.69 0.44 0.76 0.96
-1.75 0.65 0.97
10.75 3.82 6.95 10.46 0.3s
-2.25 0.78
6.87 1.05 2.83 s.39 0.15 0.41
-2.67 o.2L 0.45
3.35 0.17 0.70 1.50 0.05
-3.00 0.15
_3.25 2,27 0.09 0.18 0.33 0.04 0.08
- 43.25 0.36
Mean 1.04 0.51 0.16 -0.37 1.03 0.65 0.39 -0.08
-0.31
Sorting 1.84 1.18 1.46 1.58 1.59 1.09 1.29 1.44 r.53
Skewness 0.20 -0.85 -0.45 -0.1 9 0.241 -0.30 -0.38 -0.34 -0.08
A: Average of runs 3 and 5.
B: Average ofruns 6 and 8.
C: Average ofruns 7 and 9.
D: Average of runs 10 and 11.
NOTE: Each sediment in transport distribution was derived from two runs with closely similar mean velocities.
rAnomalous skewness value possibly the result of inadequate sampling of fines at high flow rates (Day, pers' comm., 1984).

recirculating flume with a sediment-return system for both creasing proportion ofeach grain size going into transport
suspended and bed-sediment loads. We will examine the as the flow rate increases (i.e., t(st)^ < t(si)r < t($)c <
data from the Series A experiments which utilized a bed . . .). Also, we wish to point out that, although each t(s)
material that ranged from - 3.25 Q to 4.0 Q, and was both curve (Fig. 3) is a probability function, it is not a prob-
poorly sorted (1.81 O) and bimodal (Table l). Series A ability density function in that it does not define the prob-
consisted of I I separate runs in which discharge, depth, ability of the occurrence of all possible events. Therefore,
water-surface slope, and mean velocity of the flow were the area under the curve does not necessarily equal 1.
controlled. The sediment in transport was sampled by When graphed (Fig. 3), it is clear that the calculated
means of collection baskets as it returned to the upstream t(s) functions are not monotonically increasing, as was
end of the flume. We will look at the average weight assumed in the proof of the initial transport model (Ap-
percent distributions of sediment in transport and the pendix). Rather, each t(s) is an asymmetrical curve that
resultant transfer functions, as calculated from Equation rises to a peak before falling back to zero. Our assumption
I at four mean flow velocities (Table 1). that fine grains are more easily transferred than coarse
The values of each kt(s, ) shown in Table 1 provide a grains encompasses a second assumption which is hidden;
measure of the relative probability of transport for each namely, that the transfer of any particular grain size is
particular grain size. However, the absolute probability independent ofother grain sizes. A variety ofprocesses,
(t(sr)) cannot be calculated directly with the given data, such as shielding in which fines are protected from move-
as the absolute weights of each sediment size rather than ment by larger clasts, or the decreasing ability of the erod-
the weight percentages are required to do this. By defi- ing process to carry additional fines with increasing load
nition, t(si ) must be less than I (because it is a proba- clearly invalidates this hidden assumption.
bility), and we assume that the probability of transporting This observation, that sediment transport is dependent
any specific grain size increases with increasing flow rate. not only on grain size, but also on the interaction among
Considering these two factors, the absolute values of t(sr ) the different grain sizes present, appears to have two sig-
have been estimated by assuming a different value for k nificant effects. First, as stated previously, the transfer
for each flow rate (Table 2). Although this process is functions do not increase monotonically throughout the
somewhat arbitrary, the shape of the distributions re- complete size distribution of the sediment source; and
mains the same, and the t(sr ) values now reflect an in- second, the position of the transfer function on the ab-
 SEDIMENT TRANSPORT AND GRAIN-SIZE DISTRIBUTIONS 46r

+.D TrsLn 2.-Scaled sedimenl-transfer coeftcient values of the transfer

v
HIGH ENERGY from Table I (shown in Fig. 3) with inteasing flow rates
functions
(mr')
z Tmnsfer Functions (derived from Table l)

'g UJ

lr
Grain Size
o
tr ta
wherck:0-ll wherek:0.17 *n"* F= o.zs
(0.56ms') (0.6lms') (0.66 m s,')
wncre ib: o.sl
(0.74 m s_')
l! 4.00 0.01 0.0r 0.02 0.03
LrJ
o 3.00 0.09 0.10 0.17 0.18
o 2.50 0.20 o.23 0.34 0.39
E
IIJ
I
UJ
2.00
1.50
o.26
0.27
0.30
0.30
0.41
0.41
0.48
0.48
Ir
o
z
1.25 0.20 0.25 0.36 0.47
= 1.00 0.19 o.25 0.37 0.47
E o.75 o.2l 0.28 0.43 0.58
F 0.50 0.18 0.27 0.44 0.65
F 0.25 0.16 0.27 0.46 o.76
zIIJ 0.00 0.1 I 0.20 0.37 0.69
-0.50 0.08 0.17 0.33 0.71
0.07 0.16 0.31 0.69
=
6 -4 -3 -2 -1 0-.1 -o.75
0.06 0.15 0.29 o.64
IJJ -1.25
o s(o) ,l -t.75 0.05 0.13 0.28 o.67
-2.25 0.06 0.19 0.51
tA V=O.56 m./sec tg V=O.66 m./sec
0.03 o.t2 0.41
-2.67
sk= -0.3O sk= -O.34 0.01 0.06 0.24
-3.00
-3.25 0.01 o.o2 0.08
ts i=0.61 m,/sec ts i=O.74m,zsec
sk = -O.38 sk= -O.O8
Frc. 3.-Weight percentage of sediment distribution of bed material
(histogram) and the resultant transfer functions (from Table 2) under same transport process may be represented by a high-
different flow regimes (from data in Day 1980). energy transfer function when acting on fine sediments,
and by a low-energy transfer function when acting on
distribution of the
scissa (Fig. 3) is dependent on the size coarse sediments.
source sediment. For example, we have derived transfer
functions associated with sediments of different distri- The Shape of X(s)
butions and ranges ofgrain sizes from the example shown
here; however, the transfer functions were of the same Using these empirically derived t(s) functions (Table
general shape, the difference being their relative position 2) we can now examine the possible forms of X(s), the
on the abscissa, which shifted according to the range of function relating any two sequential deposits in the di-
grain sizes in the bed material. rection of transport (Fig. 2). We can calculate X(s) from
In spite of the failure of the transfer functions to in- Equation 2 using various pairs of the derived transfer
crease monotonically over the complete distribution of functions as t,(s) and t2(s), and hence determine the rel-
g(s), each t(s) is a negatively skewed curve, indicating that ative changes in grain-size distributions between dz(s) and
the function does increase over more than half of the d,(s) by applying the theoretical results of the Appendix.
grain-size distribution present in the bed material (Fig. For illustrative purposes we have chosen three pairs of
3). The fact that in nearly all of the experiments, the transfer functions from Table 2 to demonstrate the form
resultant sediment in transport does become finer, better of X(s) under the following conditions, which are sum-
sorted, and more negatively skewed (Table l), as proved marized in Figure 4.
in the Appendix with a monotonically increasing func- ( I ) /, < t, (energy is increasing in the direction of trans-
tion, suggests that t(s) does fulfill the assumption to a port):The resultant X(s) may be calculated by letting tr :
degree sufficient to produce tlie predicted trends. to (Table 2), which is the lowest energy function, and tt
The curves illustrated in Figure 3 show that, as the flow : tr,, the highest energy function. As seen in Table 3 and
rate increases, the distribution oft(s) changes from a rel- Figure 5, the result (Xo.o) has a negatively skewed dis-
atively high negative skewness to nearly symmetrical. We tribution. A similar function is derived for any of the tt,
propose to call the negatively skewed distributions that t2 pairs provided tr ( tz. Because ofits negative skewness,
result from relatively low flow rates low-energyfunctions, the function is monotonically increasing over most of the
and the near-symmetrical distributions that result from sediment distribution, and, therefore, we may apply the
high-fl ow rates high- ener gy functions. Because these terms results of the Appendix which indicate that d2(s) will be
are relative to the grain-size distribution of the source finer and more negatively skewed than dt(s). This situa-
sediment, we can expect that the coarser a sediment is, tion may be generally unobservable in reality because
the less likely it is to be acted upon by a high-energy dz(s) is likely to be eroded and removed by the increasing
sediment-transfer function. Conversely, the finer the sed- energy regime.
iment distribution, the easier it becomes for a high-energy (2) tt > tt (energy decreasing in the direction oftrans-
transfer function to operate on it. In other words, the port, and t, is a low-energy function): X(s) may be illus-
462 PATRICK MCI.AREN AND DONALD BOWLES

a(t
1. tt. tz lunctons)
(eneEy increasing; t1 and t2 eilher high or low energy
x
zF.3
UJ Xa,e sk=-O.46
I
I 9
lr Xs,s Sk=-O.39
lr
UJA
o..
o
E
UJ
.'. sediment becoming liner and lr
more negatively skewed in o
z
lhe dircclion ot transporl.
E,

t1 > t2 (energy decreaslng; t1 is a lor energy functon) z

2, IJJ

d
=^
UJ
U'
rI s(o)
Fro. 5.-X(s) distributions (from Table 3) calculated from various
combinations of tr and tr.

.'. sedimenl becoming finer and In addition to the shapes of the transfer function X(s),
more negalively skewed in the model presented in Figure 2 suggests that two other
lhe direclion ol transporl. forms of transfer functions may occur. First, in the event
that g(s) is known, then any deposit (d-(s)) can be re-
lated to it by
3, tr't2 (enersv,::":?::3' t1 is a hish enersv lunclion; t2 is
d"(s) : kg(s)(t")(l - t*' ).
The above rules are not affected by this somewhat dif-
ferent form ofthe transfer function represented by (t")'
(l - t"*t ).
I
I Second, in the event thatt2: 0, then dr(s) is a final or
total deposit (Mclaren l98l) and can be related to dt(s)
by

kg'G)g')
d'(s): r '
.'. sediment becomlng coarser and - tr

:lffHllTg"*r"'" The transfer function represented by --l+ is always neg-

Frc.4.-Diagrammatic summary of the resultant X(s) functions re- I - lr
lating deposits in the direction oftransport given selected combinations atively skewed, regardless of whether tr is a high- or low-
oftr(s) and tr(s). energy function; thus, dt(s) will be finer and more nega-
tively skewed than dt(s).
trated by l€tting tr : ts and t2: t4 (from Table 2), the
result of which (Xn,,c.) is also a negatively skewed distri- APPLICATIONS
bution (Table 3, Fig. 5). Thus, we can expect dr(s) to be
finer and more negatively skewed than dr(s). We now wish to determine if the changes predicted by
(3) t, > t2 @nergy decreasing in the direction of trans- the above model can be observed in natural environments
port, but t, is a high-energy function and t2 may be high for which the sediment-transport direction is known. Our
or low ): We can illustrate the resultant X(s) by letting t, : purposes are to (1) establish a real-world credibility for
to and tz: k (from Table 2) which is plotted as Xp,. in the model, (2) provide a method to predict the pattern
Table 3 and Figure 5. In this situation, X(s) becomes of sediment transport from changes in grain-size distri-
positively skewed, indicating that now, most of the func- butions, and (3) obtain information concerning the rel-
tion is decreasing monotonically, the reverse of the as- ative energy of the transport regime as summarized in
sumption used in the Appendix. The results will, there- Figure 4.
fore, be opposite to those in the proof and dr(s) will be According to the models presented in Figures I and 2,
coarser and more positiyely skewed than dt(s). grain-size distributions will change in response to erosion,
 S E D I ME NT TRANS PO RT AND GRAI N- S I ZE D IS TRI B UT' I O NS 463

Tnsln 3.-X(s,) distributions derived from various combinations of two +

transfer functions (Equntion 2) lMed in Table 2. For example, X^o
indicates t, :t^ and t, -- tDfrofi#lle 2. The futtctions are plotted in

X^o Xrr Xo.

Gnin Size
O l^<tp tl>t^ tp>tc
trt
ut
4.00
3.00
0.01
0.08
0.01
0.10
0.03
0.18
o
f
2.50 0.15 0.24 0.42 F
2.00 0.18
0.19
0.32
0.31
0.54
0.54
z
o
1.50
1.25 0.13 0.27 0.57
1.00 o.r2 0.27 0.56
0.75 0.1I 0.3 r 0.79 =
0.50 0.08 0.30 l.04
0.25 0.05 0.31 L7t
0.00 0.04 o.22 l.40
0.03 0.19 1.64
-0.50
-0.75 o.o2 0.18 t.54
-t.25 0.02 o.l7 1.26
- 1.75 o.o2 0.14 t.46
Frc. 6.-Summary of the changes in grain size measures (d) that may
0.06 0.84
-2.25 occur in a given direction oftransport. For exarnple, ifin the transport
0.03 0.61
-2.67 direction sediments are becoming coarser, the skewness will become
-3.00 0.01 0.30
increasingly positive; conversely, ifsediments become finer, the skew-
-3.2s 0.01 0.09
ness will become more nega.tive. Sorting (variance) will become better
in either direction but, depending on which side ofthe "peak" transport
begins, it may first become poorer.
transport, and deposition in such a way that a lag will be
coarser and more positiYely skewed than its source (case which there is a one-eighth probability ofeither occurring
A), whereas sequential deposits will become either finer at random (p:0.125). Because of the uncertainty asso-
and more negatively skewed (case B) or coarser and more ciated with variance, which can become larger (more
positively skewed (case C) (Fig. 6). Variance or sorting poorly sorted) before becoming smaller, we choose only
must eventually become better in each of the three cases- to accept better sorting as the criterion in the two cases
It is noted that case A and case C produce identical trends suggesting the direction of transport. To determine if the
and that case A by itself does not provide a transport number of occurrences of a particular case exceeds the
direction. The differentiation between cases A and C will random probability of 0.125, we test the following two
depend on the geological interpretation ofthe environ- hypotheses:
ments being sampled.
Ho: p 0.125, and there is no preferred direction; and
Hr: p >= 0.125, and transport is occurring in a preferred
Method direction.
In reality, a perfbct sequential change in grain-size dis- Using the Z-score (Spiegel 196l) in a one-tailed test,
tributions in the down-current direction, as illustrated H, is accepted if
in Figure 2, is seldom achieved due to complicating fac-
tors such as variability in the "original source" (g(s) in x-No
Fig. 2), local and temporal variability in the transfer func- VNpq
tions, and a variety of sediment sampling difficulties (see
Mclaren 1981 for further discussion on sampling). of
Therefore, we adopt a statistical approach to determine > 2.33 (0.01 level of significance),
the transport direction by examining all possible pairs in
a sample suite. Given a sequence of n samples, there are where x : observed number of pairs representing a par-
n2-n ticular case in one of the two opposing directions; and
; directionally oriented pairs that may exhibit a N: total number of possible.unidirectional pairs. N:
n2-n
trend suggesting transport in one direction, and an equal : 2
n: number of samples in the sequence;
where
number of pairs in the opposite direction. When any two
samples zre compared with respect to their mean size, p:0.125; and q: 1.0 - p:0.875.
sorting, and skewness, eight possible trends exist; com-
pared to d,, d2 may be (1) finer (F), better sorted (B), and The Z-statistic is considered valid for N > 30 (i.e., a
large sample). Thus, for this application, a suite of 8 or
more negatively skewed (-); (2) coarser (C), more poorly
sorted (P) and more positively skewed (+); (3) C, B, -;
9 samples is the minimum required to evaluate ade-
(4) F, P, -; (5) C,P, -l (6) F, B, +; (7) C, B, *; or (8) quately a transport direction (i.".,ry: 36, the total
F, P, +. Of these trends, only two are indicative of trans-
port, namely F, B, - (case B), and C, B, + (case C), for possible pairs in one direction).
464 PATRICK McI.AREN AND DONALD BOWLES

Trnrn 4.-Grain-size data from the East Fork River, Wyoming (cal'
culatedfrom data in Emmett et al. 1980)

43.. Bedload Trap I-etion

Moment Measws (0)
(s Fig. 7)

137 75 I o43 0.70 o.97 0.63

2 075 0.68 0.97 0.62
3 137 r.32 1.20 0.20
708 808 4 220 0.70 1.04 0.92
602 301 0.86 l.l I 0.55
5
898
516 6 421 1.65 t.42 0.35
1155 516 0.92 t.22 0.40
1315 985 0.37
1396 8 602 0.93 1.04
9 708 r.20 l.l0 0.35
1475 tir 1077 10
ll
808
898
1.38
1.35
1.00
1.35
0.42
0.55
0 t2 98s 1.55 t.t7 -0.07
1766
{@ l3
t4
1,077
1,155
0.89
0.92
1.04
t.25
0.29
0.66
15 t,24r 0.84 l.l5 0.69
1830
. 2OOm, t6 1,315 1.74 1.25 -0.24
I7 r,396 0.94 1.37 0.78
18 1,481 r.26 l.3l 0.55
L9 t,662 1.44 1.06 0.36
t.o7
I 20
2r
I,695
1,7 66
1.26
0.95 1.28
0.26
0.65
Fro. 7.-Map of the study reach in the East Fork River, Wyoming, 22 1,830 1.58 1.99 0.49
showing the 22 bed-material sample locations (after Emmett et al. 1980).

The lake occupies a glacially excavated valley which is

Example 1: Fluvial TransPort 27 km long and 6 km wide and contains a flat-floored
This example uses grain-size distributions of bed-ma- basin below 100 m (Fig. 8). The Godley River forms a
terial samples from the north-flowing East Fork River, delta at the northern end, the foresets of which slope
Wyoming (data calculated from Emmett et al. 1980). This southwards to merge with the basin floor 6 km away. The
river rises in the Wind River Range and flows about 50 lake itself is dammed in the valley by moraine and out-
km before arriving at the 2-kmJong study reach (Fig. 7). wash deposits through which the Tekapo River flows.
Here the stream is about 30 m wide and meanders over An examination of grain-size trends among 25 grab
a flood plain which is confined by glacial outwash terraces samples (Table 6) taken along the axis ofthe lake suggests
ofsand and gravel. These terraces provide a continuous the following:
source of fresh sediment wherever the river impinges lat- (1) Of 300 possible pairs contained in the complete suite
erally against them. A series of 22 bed-material samples of samples, case B in the south direction is the only
averaged across the channel, were collected between 43 significant trend (Table 7). Therefore, for the lake as
m and 1,830 m upstream from a reference point marked a whole, sediment trends predict accurately the trans-
by a bedJoad trap (Fig. 7). The moment measures re- port direction. They also suggest that the transport
ported in Table 4 were calculated from the sand fraction processes produce low-energy transfer functions and
only. that energy is generally decreasing in the transport
The 22 samples (n) provide a possible 231 (N) "north- direction, both of which are reasonable in this lacus-
trending" pairs and, conversely, 23I "south-trending" trine setting.
pairs. Of the possible cases indicative of a transport di- (2) For the nine delta slope samples case B is again the
rection, only case C (coarser, better-sorted, and more pos- preferred trend (Table 7), which is also perfectly rea-
itively skewed) in the north direction is significant (Table sonable, given a decreasing energy regime with in-
5), demonstrating that grain-size distributions are chang- creasing depth down the delta foreslope.
ing in the downstream direction in a manner predicted
by the model. The occurrence of the case C trend also Tr;l,n 5.-Summary of the numbers of pairs of East Fork samples
indicates that the energy regime of the river is tending to (Table 4) producingtransport ftends. N, x, and Z are defined in text
decrease downstream, although the transfer functions must
have "high-energy" shapes with respect to the sand-size North Trcnd South Trcnd

distributions present in the river. Case B F N: 231 N: 231

x:29 x:38
"- z: 0.02 Z:1.82
Example 2 : Delta-Lacustrine Transport
Case C c N: 231 N: 231
This example uses grain-size data from Lake Tekapo, B x:71 x:33
a deep, glacier-fed lake in New Zealand (data provided
+ Z: 8.381 z: 0.82

by R. A. Pickrill, pers. comm.; Pickrill and Irwin 1983). ' Significant at the 0.01 level.
 SEDIMENT TRANSPORT AND GRAIN-SIZE DIST:RIBUTIONS 465

rpB Tesr-e 6.-Grain-size datafrom Lake Tekapo, New Zealand (datafrom

g R. A. Pickrill, pers. comm.)
.o
Moment Masms (d)
o1 Sample Sorting
2co
I 5.03 1.78 0.54 Delta Slope
2 6.20 t.67 -0.24
J 6.38 1.65 0.05
4 4.07 1.93 0.67
5 6.67 t.7 t -0.1 5
6 7.t9 t.44 -o.34
7 7.74 t.o7 -0.70
8 7.12 r.36 -o.22
9 7.65 1. l3 -0.64
l0 8.00 0.82 -0.96 Basin
ll 7.99 0.85 - l.l3
t2 7.80 0.98 -0.84
l3 7.71 t.o4 -0.63
l4 7.03 l.l0 -o.44
l5 7.79 0.99 -0.66
l6 7.95 0.84 -0.86
l7 8.r0 0.67 -1.44
l8 8.00 0.85 - 1.06 South Slope
l9 7.92 0.94 - 1.01
20 7.99 0.93 -0.99
2t 7.86 1.00 -0.87
22 7.98 0.9r -t.02
23 8.05 0.79 -t.44
24 8.15 0.67 -t.27
25 7.40 0.59 0.28

currents moving upslope have been observed in sim-

ilar lakes (e.g., Kootenay Lake, British Columbia; C.
H. Pharo, pers. comm.). It is difficult, however, to
imagine that the energy regime is decreasing with
shoaling water, the requirement for case B transport
(i.e., t, > tr). On the contrary, not only are higher
eneryies associated with shoaling water, but south-
ward currents will be concentrated into the outflow.
Mathematically, it was shown that sediments also
become finer, better sorted, and more negatively
skewed ifenergy increases in the direction oftransport
(i.e., t, < t2; Fig. 4), but it was argued that successive
deposits would be removed by erosion with an in-
creasing energy regime. This example tends to suggest
that such deposits can remai4 probably as a result
Frc. 8.-Lake Tekapo (New Zealand) and sample locations (after R. ofthe difficulty in resuspending fine silt and clay-sized
A. Pickrill, pers. comm.). particles once they have been deposited.
There is also an apparent southward case C trend which
(3) Within the basin, case B trends are significant in both is significant at the 0.05 level (Table 7). This is the result
directions, suggesting the occurrence oftransport in of sample 25 taken close to the outflow, which is coarser,
the low-energy regime. Two reasons which may be better sorted, and more positivety skewed than any of the
responsible for obscuring a preferred direction are other south slope samples (Table 6). Case C is unlikely
sediment input from other rivers entering both sides as it demands a decreasing energy regime; therefore, case
of the lake, such as the Mistake and Cass rivers (Fig. A is accepted as a logical interpretation whereby sample
8), and the breakdown of the thermocline in winter, 25 is a lag of those sediments presently located in deeper
resulting in weak and random bottom currents (Pick- water.
rill and Irwin 1983).
(4) Eight samples from the south slope show an apparent Example 3: Longshore Transport
case B trend in the south direction (Table 7), indi-
cating that sediment transport is occurring upslope Coburg Peninsula, a spit in the Strait of Juan de Fuca
from below 100 m to the Tekapo River outflow. Such near Victoria, British Columbia, originates at its south-
466 PATRICK MCI.AREN AND DONALD BOWLES

Ttsrn 7 .- Summary of the numbers of pairs of Inke Tekapo samples (Table 6) producing transport trends. N, x, and Z are defined in text

All 25 Samplcs Delta slopc Basin Soulh slope

ffiffiffiffi
Case B F N: 300 N: 300 N:36 N:36 N:28 N:28 N:28 N:28
B x:34 x:213 x:4 x:26 x:13 x:13 x:6 x:12
z: -0.6t z: 30.64t z: -0.25 Z: l0.83l z: 5.43t Z: 5.431 Z: 1.43 Z: 4.861
Case C C N: 300 N:300 N:36 N: 36 N:28 N:28 N:28 N:28
B x:5 x:2O x:l x:l x:0 x:0 x:0 x:7
+ z: -5.67 Z: -3.06 Z: -1.76 Z: -1.76 z: -2.O0 z: -2.0O z: -2.0O Z: -2.O0'z
' Significant at the 0.01 level.
'?Significant at the 0.05 level.

west end from a low, eroding bluffcomposed of till (Fig. to the northeast (Table 9) in the direction of spit gxowth,
9). The spit extends northeast about 2.5 km, where it illustrating not only that the model appears to have real-
terminates at a narrow tidal channel: The direction of spit world validity, but can also predict correctly the direction
growth and several other similar spits on both sides of of sediment transport.
the Strait ofJuan de Fuca confirm a longshore transport
direction which is predominantly eastwards. DISCUSSION AND CONCLUSIONS
A sequence of eight samples from the lower beach face
and two samples from the eroding till bluffreveal that all We have attempted to demonstrate that grain-size dis-
of the beach samples are coarser, better sorted, and more tributions change in the direction oftransport according
positively skewed than the till (Table 8). This suggests to the shape of the transfer function X(s). Because X(s)
case C transport, which is consistent with the till being is the result of two transfer functions, (tr(s) and tr(s)),
the dominant sediment source for the beach. The eight whose shapes can be determinedempirically, we candem-
beach-face samples also show a significant case C trend onstrate that sequential deposits may become either
coarser, better sorted, and more positively skewed (high-
energy t1(s)), or finer, better sorted, and more negatively
skewed (ow-energy tr(s)) with a decreasing energy regime.
It is interesting to note that sediments cannot become
coarser forever because, with coarsening, it becomes less
and less likely that the transport processes will maintain
high-energy characteristics with respect to the coarsening
sediment. As the deposits become coarser, the transfer
function describing the processes will take on the char-
acteristics of the low-energy function, and the sediments
will become finer again.
The model indicates that sediments can also become
finer in the direction of transport with an increasing en-
ergy regime. This somewhat surprising result appears ini-
tially to be of theoretical value only as intuition would
suggest that down-current deposits could not remain to
be observed. flowever, one example used in the deter-
mination of transport direction suggests that such de-
posits can remain, possibly as a result of cohesion in fine
sediments or a high-sediment-supply rate.

TaB-e 8.-Grain-size datafrom theoCoburg Peninsula' British Colum-

I r.36 1.19 0.08 Eroding Till Blufl

2 0.27 0.80 t.29 Beach Face
3 0.32 0.98 1.03
4 0.30 0.94 0.96
5 o.27 0.80 1.13
6 -0.04 0.80 r.70
7 -o.27 0.73 2.93
8 0.00 0.78 2.35
9 0.09 0.95 1.60
t Average of 2 samples.
Frc. 9.-Coburg Peninsula (Vancouver Island) and sample locations-
 SEDIMENT T'RANSPORT AND GRAIN-SIZE DISTRIBUTIONS 467

Txsrpg.-Summary of numbers of pairs of Coburg Peninsula samples Blerr, H., MIoor-rroN, G., lNo Munrav, R., 1980, Origin of Sedi-
Gable 8) producing fiansport trends. N, x' and Z are defined in text mentary Rocks: Englewood Cliffs, New Jersey, Prentice-Hall, 782 p.
Dev, T. J., 1980, A study of the transport of graded sediments: Report
NE Trcnd SW Trcnd No. IT 190, Hydraulics Research Station, Wallingford, England, I I p'
F N: 28 N: 28 ErNsrEw, H. A., 1950, The bedJoad function for sediment transpor-
Case B
B x:0 x:3 tation in open-channel flows: U.S. Dept. of Agriculture, Tech. Bull.
Z: -2.O0 Z: -0.29 No. 1026, p. l-70.
Euverr, W. W., Mvnrcr, R. M., eNo Mreor, R. H., 1980, Field data
Case C c N:28 N: 28 describing the movement and storage of sediment in the East Fork
B x: 14 x:6 River, Wyoming, Part I. River hydraulics and sediment transport,
+ z: 6.ool Z: 1.43 1979: United States Department of the Interior, Geological Survey,
Open File Report 80-l189, 43 p.
' Significant at the 0.01 level.
Fuur'rrNc, B. W., 1982, Sediment mixing: its natural occurrence and
textural expression: Abstract, jn Eleventh International Congress on
Sedirnentology: McMaster University, Hamilton, Canada, August 22-
We have also modified the earlier model proposed by 27, 1982, p.81.
Mclaren (1981) which indicated that sediments always For-r, R. L., lNo Weno, W. C., 1957, Brazos River bar: a study in the
become better sorted in the direction of transport. Al- significance ofgrain size parameters: Jour. Sed. Petrology, v. 27' p.
though not necessarily true, we'suggest that in reality, 3-26.
successive deposits seldom become more poorly sorted, FnreplaN, G. M., 1961, Distinction between dune, beach and river
sands from textural characteristics: Jour. Sed. Petrology, v. 31, p.
particularly when sediment distributions are described in 514-529.
phi units. The log-transformation tends to make grain- 1979, Address of the retiring President of the International
size distributions relatively symmetrical, a requirement Association of Sedimentologists: ditrerences in size distributions of
that ensures a decrease in variance when the transfer func- populations of particles among sands of various origins: Sedimen-
tology, v. 26,p.3-32.
tion is predominantly increasing or decreasing monoton-
G.lnnow, H. C., 1982, A theoretical model of source and transport
ically. Mclaren (1981) also suggested that sediments could effects on grain-size distributions: abstract, iz Eleventh International
become finer and more positively skewed, which this new Congress on Sedimentology: McMaster University, Hamilton, C-an-
analysis, using realistic transfer functions, demonstrates ada, August 22-27, 1982, p. 80.
is incorrect. When sediments become finer, the skewness Grssrrn, J., 1976, Beginning and ceasing of sediment motion, in Shen,
must become more negative (Fig. 6). H. W., ed., River Mechanics, vol. 1: Fort Collins, Colorado, H. W.
Shen, p. 7-l-7-22.
At present, we have made no attempt to explain the GHosn, J. K., Mazuuorn, B. S., eNp SrNcurre, S., 1979, Methods of
processes responsible for determining the shape of X(s). computation of suspended load from bed materials and flow param-
We have, however, demonstrated a method for the pre- eters: Indian Statistical Institute, Calcutta, Technical Report No. P
diction of sediment transport paths, regardless of the pro- and E/Flume/l / 1979, 34 p.
cess, that appears to give the correct directions in a wide Grrrs, C. J., .lNo NEII-1, C. R., 1972, Interim report on laboratory study
of basket-type bedload samplers: Research Council of Alberta, Re-
variety of environments, even though we are using only port No. REH/72/2,5 p.
the mean, softing, and skewness as sediment-grain-size Gllrsrrn, R. P., ,lND Nrmor, H. W., 1974, Grain-size distributions,
descriptors. It would clearly be preferable to base the an aid in facies identification: Bull. Canadian Pefoleum Geology, v.
prediction of a transport trend on the complete distri- 22,p.203-240.
bution of X(s), but at present there is a lack of suitable HlNnn, B. E., 1984, Santa Ana River: an example of a sandy braided
floodplain system showing sediment source area imprintation and
experimental data which would allow us to examine in selective sediment modification: Sedimentary Geology, v. 38, p. 247 -
detail all the shapes that X(s) may take. In the future we 261.
hope to utilize the complete distribution of X(s) in the Knur"rnrrN, W. C., 1938, Size-frequency distributions of sediments and
determination of a transport trend and to correlate its the normal phi curve: Jour. Sed. Petrology, v. 8, p. 8'S-90.
shape with known processes and depositional environ- MrsoN, C. C., rNo Fox, R. L., 1958, Differentiation of beach, dune
and aeolian flat environments by size analysis, Mustang Island, Texas:
ments.
Jour. Sed. Petrology, v. 28, p. 2ll-226.
McCnvr, I. N., 1978, Grain size trends and transport along beaches:
ACKNOWLEDGMENTS example from eastern England: Mar. Geol., v. 28, M43-M51.
McLereN, P., 198 I , An interpretation of trends in grain size measures:
The wrlters would like to thank E. Klovan, R. E. Thom- Jour. Sed. Petrology, v. 51, p. 6ll-424.
1982, Hydraulic control ofgrain-size distributions in a macro-
son, and R. W. Dalrymple for critically reading earlier tidal estuary-discussion: Sedimentology, v. 29, p. 437-439.
drafts and providing the basis for considerable improve- 1983, Coastal sediments of the Strait of Juan de Fuca: impli-
ments. T. J. Day also provided advice and loaned ap- cations for oil spills, in Current Research, Pt. A, Geol. Surv. Canada,
propriate data. The help of R. Currie for necessary com- Pap.83-lA, p.241-244.
puter programming and D. chisholm for typing the 1984, The Whytecliffoil spill, British Columbia: sediment trends
and oil movement on a beach,.iz Current Research, Pt. A, Geol. Surv.
manuscript is gratefully acknowledged. Canada, Pap. 84-lA, p. 81-85.
McL,lnEN, P., B.nnnIE, W. B., ,tNo SrvrErs, J. M., 1981, The coastal
morphology and sedimentology of Cape Hatt: implications for the
REFERENCES
Bafrn Island Oil Spill Project (BIOS), in Current Research, Ft' B,
Geol. Surv. Canada, Pap. 8l-lB, p. 153-162.
AroensoN, J. 8., WolrrsrcH, C., Wr.Icnr, R., nNo Colr, M. L., 1982, McMeNus, D., 1982, Phi and sediment size analysis: discussion: Jour.
Determination of depositional environnents of sand bodies using Sed. Petrology, v. 52, p. 101l-1014.
vertical grain-size progressions: Transactions-Gulf Coast Associa- MrnorrroN, G. V., 1976, Hydraulic interpretation of sand distributions:
tion of Geological Societies, v. 32, p. 565-57'l . Jour. Geology, v. 84, p. 405-426.
468 PATRICK MCI.AREN AND DONALD BOWLES

Moss, A. J., 1962, The physical nature of common sandy pebbly de-
posits, Part I: Am. Jour. Sci., v. 260, p,337-373.
PrcrnrrL, R. A., aNo InwrN, J., 1983, Sedimentation in a deep glacier-
fed lake-Lake Tekapo, New Zealand: Sedimentology, v' 30, p. 63- r1= H
r()
75.
SnrcEr, J. K., eNp ANoEnsoN, J. 8., 1984, Use of total grain-size dis-
tributions to define bed erosion and transport for poorly sorted sed- rn
iment undergoing simulated bioturbation: Marine Geology, v. 57, p.

+.,.(L)=
335-3s9.

L
Sr-rNcrnr-eNo, R. L., 1977 , The etrects of entrainment of the hydraulic
equivalence relationships oflight and heavy minerals in sands: Jour.
Sed. Petrology, v.47, p.753-7'10.
SrrEcer, M. R., 1961, Theory and Problems of Statistics: Schaum's
Outline Series: New York, McGraw-Hill Book Co., 359 p.
Sraror, F. W., aNo TexNen, W. F., I975, Hydrodynamic implications
ofbeach, beach ridge and dune grain size studies: Jour. Sed. Petrology,
v.45, p. 926-931. dn
Uoorr, A., 1914, Mechanical composition of clastic sediments: Geo-
J.
logical Society of America Bulletin, v.25, p. 655-744. Fro. I 0. - Diagram illustrating the end result of any function g(s) after
VrsnEn, G. S., 1969, Grain size distributions and depositional processes: rep€ated multiplication with a monotonically increasing or decreasing
Jour. Sed. Petrology, v. 39, p. 1074-1 106. function to derive r(s) and d(s), respectively.

APPENDIX
g(s,)
As a first step in the analysis, we will consider the situation as outlined
in Figure l, where: ft'(s.)
g(s,) + s(s,)L---- *.,,[.*etJ.
g(s) is a function of "s" (grain size in d units) describing any grain- 1".
size distribution; Because it was assumed that t'(sr) > t'(s,) > t'(s3), then
t(s) is a transfer function that is a monotonically increasing function
of"s"; t(s') < t(s,) V i < j and t(sr) <I V i; l{91'- o ", lroovi>1
(s) is the size distribution of sediment in transport that is derived Lt'(s')l
from g(s); and
t'(s) equals I - t(s) and is a monotonically decreasing transfer function;
d(s) is the first deposit derived from g(s); it may be considered a lag
remaining behind after (s) has been removed.
d"(s,)-fril+?d.*
Irt tlasn-oo.
(q) : kc(sJt(q) tAI Because
N
where l,
)i-l 4(s'):
I
k : :: (k is applied to normalize r(s))
N then
)i-l e(sJt(q)
4(sJ-0 as n-ofori> I
and
d(q) : k'g(q)t'(sJ 4(s')-1(Fig'10)'
Lt:-
I By the converse argument, consider r"(s) when i : N (i.e., the last
N class interval). Expanding from Equation A, as above, gives
)i-l e(s')t'(q)
r"(s") : c(s.)t(s")"
g(s,)t(s,). + g(sr)t(sr)". . . g(sN)t(sN)"
We will now apply r'(s) z times in order to generate d.(s), which may
be considered as the nth lag ofg(s) (Fig. l0). g(s.)
-
Then,
' Itr.,lT ltrt,tl"
d-(s): k'"g(s)t'(s)" IB1 stq)lffil + cGJLisJl + "' s(s")
where Again, because t(sr) < t(sr) < t(s3), then
*'.:;L. tcl l-(!!-l'-o as n-ovi<N,
Lt(s")l
)i-l e(s,)t'(q)"
and
Note that for n : 0, 6(sJ : g(q).
c(s")
Consider d.(s) when i: I (i.e., the first class interval). From Equa- r'(sx)-o+o+d(s.)
tion B,
d"(s') - k'"g(s')t'(s')". -l as n+@.
Substituting for k'" from Equation C gives Because
N
g(s,)t'(s,)"
4($: g(sr)t'(sr)" + g(sr)t'(sr)" + .. .
)i-l r"(s): 1,
 SEDIMENT TRANSPORT AND GRAIN-SIZE DISTRIBUTIONS 469

UJ
o
2
str

lr
o
1'a
1!
o

a
ts
z .^g-i-- ?F?$'rt'^i
=
O<--n
no.of times transfer function (t(s)) Fro. 12.-Diagram illustrating how skewness must approach -oo in
tr =
acts on sediment source (g(s)) the direction of transport. The value "a" is arbitrarily small and is
approaching zero with the result that Sk,- -l/a' -oo as a - 0.
Frc. 1l.-Diagrammatic illustration of the changes in the variance
ofr(s) and d(s) relative to g(s). Although the variance ofboth r(s) and
d(s) must ultimately approach zero, the location of n 0 and the vari- : As
ance of g(s) may result in o,2 ot ou2 increasing before their eventual n+@
decrease. In this illustration, 4,2 has increased at n l. : o..2-sa2-s"2

then
-0.
Similarly, as
r"(s)-6 as n+oofori<N
n-@,
oc.2-s,2-sr2
r"(s.)-l(Fig.l0). -0.
Mean Grain Size We have shown that the variance of the lag and the variance of
N sediment in transport will approach zero when t(s) is applied "n" times.
p, : > r(s)s,, where p. : the mean grain size ofr(s), the distribution At n = 0, there is neither sediment in transport nor a lag produced, and
i-l we are at the "starting point" (i.e., the sediment source, g(s)). We know
of sediment in transport (Fig. l). that "somewhere" between the end points of zero variance of the lag
and the zero variance of the sediment in transport, there must be a
point of maximum variance. If the maximum variance and n : 0 co-
Let
N
incide then both the lag and sediment in transport must become better
ll..: ) r"(s')s,,
sorted for all "n." Ifthey do not, then decreasing variance may not be
i-l
true for all "n."
which, by applying Equation A, becomes This is illustrated graphically in Figure I 1, which shows the changing
N magnitude of variance for a lag and for sediment in transport (the exact
rrq:k")g(sJt(sJ"s,. shape ofthe curve has been drawn arbitrarily). The value ofthe variance
i=l for g(s) determines the position of n 0. At n : :
l, the sediment in
transport has increased its variance, which will not begin to decrease
As : :
until n 3. In reality, the position ofn 1 is arbitrary and "n" itself
n-@ has little physical meaning. It could, for example, represent an individual
pr. r sN Gig. l0). erosion event such as one wave, or it could be the combined effects of
ten years ofwaves.
Because Because the phi scale results in approximately Gaussian or normal
&" ( S., distributions which are symmetrical,'we believe thal os2 and the max-
imum variance of o.2 or o62 will coincide closely with each other and
then that an increasing variance for r(s) or d(s) will be rarely observed.
P..>&"Vn.
Skewness
This proves that, if the transport function t(s) is increasing mono-
tonically, tbpn sediment in transport must be finer than its source. By
L€t
the converse argument:
Fo.< F" V n (Fig. l0). sk, :. +; i,r, - pJ3e(s,),
\os-l-._ i_t
Thus, the first deposit (or lag) must be coarser than its source.
where S\ is the skewness of the source sediment. Similarly,

Variance (Sorting) sk'" : ,+ i,r' -

(o,"2)rtz -
r,,")'r"(s')'
Let
N
where Slg" is the skewness of the nth sediment in transport.
o,'z:2 r(sJ(si - pJ'z
Asn-o, let
i-l

where r.2 is the variance of (s), the grain-size distribution in transport r.(x)-1-a,
(Fie. 1). I-et r"(s*-,)- a,
N
o"^2 : l<^ )i-l t(tJ"e(s,X*' - pJ'. r.(s,)- 0 for i: 1,2,3,...,N- 2,
470 PATRICK MCI..AREN AND DONALD BOWLES

where "a" is arbitrarily small (Fig. l2). Similarly, as

Then n-oo,
l4,-sN(l -a)+sN-ra - +o;
Sk*"

- s* + a(s*-, - sy) '. Skd" > S\ V n.

We have shown, therefore, that the skewness of sediment in transport

o."2 - (s* - p")'(l - a) * (s.-' - p")'a must become more negative than the source sediment, and the skewness
- a(sN-r - sNF. of a lag must become more positive.
Thus,
(sr - t,)3r,(sJ - a(s.-r - sN)3
Summary
and a(s.-t - s.)t Given a monotonically increasing transfer function (in d units), it has
Slq" - been shown that sediment in transport must become progxessively finer
[a(s*-t - s")2]3/2
and more negatively skewed than its source sediment. Conversely, the
_1 lag must become coarser and more positively skewed than its source
a sediment. The change in sorting (variance) cannot be determined in the
-oo as a -l 0. general case, although at some point in the transport path sorting must
.Sk <S\Vn. become better.

View publication stats