THE EFFECTS OF SEDIMENT TRANSPORT ON GRAIN-SIZE DISTRIBUTIONS I PATRICK McLAREN

Geological Survey of Canada Pac~lic Geoscience Centre

P.O. Box 6000 Sidney, B.C. V8L 4B2, Canada
AND DONALD BOWLES

Box 179, R.R. 1 Chelsea, Quebec JOX 1NO, Canada

ABSTaACr: 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 of empirically derived transfer functions exhibit negatively skewed distributions (on a phi scale). Thus, when a sediment is being 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 of transfer 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 lowenergy 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 lowenergy 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 of transport 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

The environmental interpretation of grain-size distributions found in sedimentary deposits has been, and still is, a fundamental goal of sedimentology. Ever since Udden's work in 1914 it has been recognized that sediment size fractions approximate a log-normal distribution. In reality, however, most sediments do not strictly follow log-normality and deviations from the Gaussian model have been given various environmental interpretations (Folk and Ward 1957; Mason and Folk 1958; Friedman 1961, 1979). Other workers have differentiated log-normal subpopulations defined by distinct breaks within the complete size range of the sample (Moss 1962; Visher 1969; Glaister and Nelson 1974; Middleton 1976). Each subpopulation is interpreted as representative of a specific transport mode (traction, intermittent suspension, and suspension), the relative concentrations of each suggesting particular depositional environments. Although Blatt et al. (1980) argue that it should not be surprising for specific process populations to exist because different transport mechanisms differ in the way they select grains for movement or deposition, many other authors are unable to achieve similar interpretations (Garrow 1982; H e m m i n g 1982; Anderson et al. 1982). There are at least two reasons for such disagreements. First, the processes of erosion and deposition encompass complex
t Manuscript received 8 February 1984; revised 16 November 1984.

functions involving many variables, including grain orientation and the structure of turbulence, which are intrinsically random (Gessler 1976). For example, there is substantial disagreement as to the actual velocities required to erode and transport particles of a given size, particularly when the bed material is poorly sorted (Singer and Anderson 1984). Experimental diiticulties appear to be as true today as in 1950 when Einstein wrote, "The forces acting on individual particles of a natural sediment mixture in a bed cannot very well be measured" (Einstein 1950, p. 35). Both Slingerland (1977) and Singer and Anderson (1984) provide good summaries of the complexities encountered by workers who have sought to document the processes of entrainment and deposition of sediments. The second difficulty lies in the imprint of the source sediment characteristics on the characteristics of the deposit. This was recognized as long ago as 1938 when Krurnbein suggested the importance of progressive or continuous changes in grain-size distributions from source to final deposit (Krumbein 1938). Progressive changes have been recognized by several workers (Stapor and Tanner 1975; McCave 1978; Haner 1984) and have been analyzed in a deductive model by McLaren (1981). He suggested that the mean, sorting, and skewness of grainsize frequency distributions follow trends that identify the direction of transport and the sedimentary processes of winnowing, selective deposition, and total deposition. Using a hypothetical sediment distribution and an as-

JOURNAL OF SEDIMENTARYPETROLOGY,VOL. 55, NO. 4, JULY, 1985, P. 0457--0470 Copyright 1985, The Society of Economic Paleontologists and Mineralogists 0022-4472/85/0055-0457/$03.00

458

PATRICK M c L A R E N AND DONALD B O W L E S

and k' = 1
N

sumption that light grains have a greater probability o f being eroded and transported than heavy grains, the model demonstrated that I) sediment in transport must be finer, better sorted, and more negatively skewed than its source sediment; 2) a lag must become coarser, better sorted, and more positively skewed; and 3) successive deposits m a y become finer or coarser, but the sorting must become better and the skewness more positive. In this paper we attempt to scrutinize the validity o f these trends in a more rigorous examination than the strictly deductive approach used by McLaren (1981). We present a more refined model that demonstrates how grainsize distributions o f sedimentary deposits change in the direction o f transport. The paper does not, however, provide solutions or increase our understanding o f transport and depositional processes; rather, it recognizes that the nature o f these processes is probabilistic, and their results should be reflected in the relative changes o f grain-size distributions found in interrelated sedimentary deposits. The model is tested in a variety o f environments, and we propose a technique to interpret the observed changes in grain-size distributions for the purpose of establishing sediment-transport directions. INITIAL SEDIMENTTRANSPORT MODEL Let us consider any grain-size distribution (g(s)) (Fig. 1). If eroded, the sediment which is transported has a new distribution (r(s)) which is derived from g(s) according to a function t(s) so that r(si) = kg(si)t(si) or t(sl) r(sO kg(s3 '

g(si)(l
i~l

t(s0)

[11

where g(s3 and r(sl) define the proportion o f the sediment in the t~ grain-size class interval for each o f the sediment distributions, k is a scaling factor that normalizes r(s) so that
N

Z r(si) = 1.
i=|

As well as using distribution functions to characterize sediment size, we use t(s) in a similar manner to describe statistically all o f the processes which tend to m o v e sediment from one location to another. This function is described in the same m a n n e r as a size distribution function (i.e., by weight proportion, grain frequency, etc.), and each t(si) gives the probability o f transferring grains o f size "si" from a source (g(s)) into transport (r(s)). We propose to call each t(si) a transfer coefficient and t(s) a sedimenttransfer function. The latter m a y be thought o f as a function that incorporates all sedimentary and dynamic processes that result in initial m o v e m e n t and transport of particular grain sizes during a period o f time. Implicit in the earlier model proposed by McLaren (l 9 8 l) was the assumption that light grains have a greater probability o f being transferred from a source into transport than heavy grains. This suggests that t(s) is a function which increases (on a phi scale) or decreases (on a millimeter or linear scale) monotonically; that is, smoothly and continuously. ~ Although this concept oft(s) is a simplification, the model derived from this assumption has proven to be effective in practice for determining sediment-transport paths and the relationships a m o n g interrelated sedimentary deposits (see, for example, McLaren et al. 1981; McLaren 1981, 1982, 1983, 1984). If we assume that t(s) is monotonically increasing, it can be shown that the sediment in transport (r(s)) is always finer and more negatively skewed than its source (g(s)) and that the lag (d(s)) is always coarser and more positively skewed than its source (g(s)). The trend in sorting depends on the initial sediment distribution o f the source and therefore cannot be determined for any g(s). Sorting must, however, eventually become better in both the sediment in transport and in the lag. The p r o o f for these results is not essential to the text o f this paper and is shown in the Appendix. EXPANDED SEDIMENT-TRANSPORTMODEL

Thus, k1
N

Z g(si)t(Si)
i=l

With the removal o f r(s) from g(s), the remaining sediment (a lag) has a new distribution denoted by d(s) (Fig. 1), where d(si) = k'g(si)(1 - t(si)) or d(si) t'(si) = k'g(si) ' where t'(s3 = 1 - t(s3

We have shown in the Appendix how grain-size distributions o f sediment in transport and the remaining lag deposit must change relative to a c o m m o n source sediment, assuming that the sediment-transfer function is monotonically increasing. We n o w wish to (1) expand the simplified model shown in Figure l, and (2) examine the validity o f the assumption (i.e., the nature o f t(s) in light of empirical data). 2 The discussion on the use of phi versus metric continues (see McManus 1982). We find that the log-normal transformation to phi provides better statistical descriptors (i.e., mean, sorting, and skewness) for the identification of sediment-transport directions than those based on a linear scale. The remainder of the paper uses the phi scale exclusively.

S E D I M E N T T R A N S P O R T A N D GRA I N - S I Z E D I S T R I B U T I O N S

459

~-~-~--t (s)--~-~-] I t'(s)il-t(s)

FIG. l.--Diagrammatic summary of the initial sediment-transport model discussed in the text and the Appendix.

WAVE OR CURRENT DIRECTION

Consider a unique sediment source such as an eroding cliff o f unconsolidated sediments with a grain-size distribution g(s) (Fig. 2). Eroded sediments are deposited in a down-current direction, forming a beach whose grainsize distributions are all(S), d2(s), d3(s) . . . . . respectively. The sediments in transport are denoted by r~(s), r2(s) . . . . . Let the first transport function (t0(s)) be equal to 1 for all " s " so that r~(s) = g(s). This is analogous to a sudden mass wasting event on the eliffface whereby the complete distribution (g(s)) is momentarily in transport. This distribution (r~(s)) is then acted upon by a process represented by the function tt(s) which results in a new distribution in transport, rz(s ). The sediment remaining is deposited as dj(s) which is related to r~(s) by the function 1 - h(s). Similarily, rz(s) is acted upon by tz(s) with the result that d2(s) is deposited. Any three boxes forming an equivalent pattern to those considered in Figure 1 (e.g., r~(s), r2(s), and dr(s); Fig. 2) can be analyzed in the m a n n e r outlined in the previous section and the Appendix using the assumption that t(s) is monotonically increasing. Therefore, sediments in transport (r~(s), r:(s), r~(s) . . . . . etc.) must become progressively finer and more negatively skewed, and each d(s) can be considered a lag of its corresponding sediment in transport (r(s)). Thus, for example, d3(s) is coarser and more positively skewed than r3(s). We now wish to determine the relative changes in sediment distributions a m o n g the sequential deposits dj(s), dz(s), d3(s) . . . . . bearing in mind that r(s), t(s), and (frequently) g(s) are not observable. Let us suppose that dz(s) is related to d~(s) by a function X(s), so that d2(s) = kd,(s)X(s) where k
N

~-~-~ -t(s)~iliiiii~ii~ii!i ~t2(sl~t~(s) )

!~i~:~g:~!!ii!!!~i
x (s) ~ J
FIG. 2.--Diagrammatic summary of the expanded sediment-transport m o d e l The three enclosed boxes are analogous to the initial model shown in Figure 1. See text for definition o f terms.

i i~~~i! !~~~~~~~~~~i ii~~i~J ii i~~~~~~i~ii i~ ~ i i i ~i i i i i i i i i i i~ i i i~~ i i i i i i i ~ i

dz(s) =

kdt(s)h(s)(1 - t:(s))
1 h(s)

= kd,(s)X(s), where X(s) t,(s)(1 t2(s))

1 -

tl(S)

[2]

X(S) is a function which combines the effects of the two transfer functions h(s) and t2(s). As such, X(s) m a y also be considered a transfer function in that it provides the statistical relationship between two sequential deposits. Similar to t(s), this function incorporates all of the processes responsible for sediment transport and deposition resulting in a sequence of sedimentary deposits over the period of time represented by the samples. Therefore, the relative change in the distributions between dz(s) and dl(s) is dependent on the shape of the function X(s) which can be determined by examining empirically derived t(s) functions.

The Shape oft(s)

d,(si)X(si) Transfer functions were calculated from data in Day (1980), E m m e t t et al. (1980), G h o s h et al. (1979) and Gibbs and Neill (1972). These data sets all produced similafily shaped curves in spite o f a wide range o f grain-size distributions; however, for illustrative purposes we will use data from flume experiments described by Day (1980), which are the most complete for our purposes. These experiments were conducted in a 2.46-m-wide

i=l or
X(s) -

dds)
kdt(s)

"

As illustrated in Figure 2, d2(s) can also be related to d~(s) by

460

PA T R I C K M c L A R E N A N D D O N A L D B O W L E S

TABLE l.--Distribution o f sediment in transport with increasing flow rates (m s J) and the respective transfer functions (data from Day 1980,

Se~es A)
Bed Material wl %
g(,%~

Sedimentin Transport(wt %) rA
(0.56 m s - 9

Grain Size 4'

s~

re
(0.61 m s-~)

re
(0.66 m s -~)

ro
(0.74 m s-~) t^

TransferFunctions(sediment-transfercoe~icients) kt(si) ffi r(~/g(~ (fromEquation 1)

tn tc to

4.00 3.00 2.50 2.00 1.50 1.25 1.00 0.75 0.50 0.25 0.00 -0.50 -0.75 - 1.25 - 1.75 -2.25 -2.67 -3.00 - 3.25 <3.25 Mean Sorting Skewness

1.42 1.30 4.46 11.00 5.68 5.75 4.31 2.69 2.56 2.21 4.98 3.10 6.96 10.10 9.88 10.75 6.87 3.35 2.27 0.36 -0.31 1.84 0.20

0.11 1.05 8.00 25.93 13.57 10.55 7.33 5.21 4.09 3.21 4.73 2.35 4.64 5.19 4.30

0.12 0.71 6.01 19.26 9.94 8.58 6.33 4.38 4.03 3.46 5.87 3.16 6.73 8.87 7.49 3.82 1.05 0.17 0.09 0.51 1.46 -0.45

0.09 0.75 5.22 15.45 7.92 6.99 5.38 3.93 3.87 3.47 6.24 3.49 7.48 9.76 9.53 6.95 2.83 0.70 0.18 0.16 1.58 -0.19

0.07 0.45 3.31 10.16 5.22 5.18 3.84 2.99 3.18 3.18 6.56 4.15 9.18 12.21 12.69 10.46 5.39 1.50 0.33 -0.37 1.59 0.24 t

0.08 0.81 1.79 2.36 2.39 1.83 1.70 1.93 1.60 1.45 0.95 0.76 0.67 0.51 0.44

0.08 0.54 1.35 1.75 1.75 1.49 1.47 1.63 1.57 1.57 1.18 1.02 0.97 0.88 0.76 0.35 0.15 0.05 0.04 0.65 1.29 -0.38

0.06 0.57 1.17 1.40 1.39 1.21 1.25 1.46 1.51 1.57 1.25 1.12 1.07 0.97 0.96 0.65 0.41 0.21 0.08 0.39 1.44 -0.34

0.06 0.35 0.74 0.92 0.92 0.90 0.89 1.11 1.24 1.44 1.32 1.34 1.32 1.21 1.28 0.97 0.78 0.45 0.15 -0.08 1.53 -0.08

1.04 1.18 -0.85

1.03 1.09 -0.30

A: Average of rims 3 and 5. B: Average of runs 6 and 8. C: Average of runs 7 and 9. D: Average o f r u n s 10and 11. NOTE: Each sediment m transport distribution was derived from two runs with closely similar mean velocities. t Anomalous 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 suspended and bed-sediment loads. We will examine the data from the Series A experiments which utilized a bed material that ranged from - 3 . 2 5 ~ to 4.0 ~, and was both poorly sorted (1.81 ~) and bimodal (Table 1). Series A consisted of 11 separate runs in which discharge, depth, water-surface slope, and mean velocity of the flow were controlled. The sediment in transport was sampled by means of collection baskets as it returned to the upstream end of the flume. We will look at the average weight percent distributions of sediment in transport and the resultant transfer functions, as calculated from Equation 1 at four mean flow velocities (Table l). The values of each kt(si) shown in Table 1 provide a measure of the relative probability of transport for each particular grain size. However, the absolute probability (t(si)) cannot be calculated directly with the given data, as the absolute weights of each sediment size rather than the weight percentages are required to do this. By definition, t(si) must be less than 1 (because it is a probability), and we assume that the probability of transporting any specific grain size increases with increasing flow rate. Considering these two factors, the absolute values of t(sl) have been estimated by assuming a different value for k for each flow rate (Table 2). Although this process is somewhat arbitrary, the shape of the distributions remains the same, and the t(si) values now reflect an in-

creasing proportion of each grain size going into transport as the flow rate increases (i.e., t(si )A < t(si)~ < t(si )c < ..). Also, we wish to point out that, although each t(s) curve (Fig. 3) is a probability function, it is not a probability density function in that it does not define the probability o f the occurrence of all possible events. Therefore, the area under the curve does not necessarily equal 1. When graphed (Fig. 3), it is clear that the calculated t(s) functions are not monotomcally increasing, as was assumed in the proof of the initial transport model (Appendix). Rather, each t(s) is an asymmetrical curve that rises to a peak before falling back to zero. Our assumption that fine grains are more easily transferred than coarse grains encompasses a second assumption which is hidden; namely, that the transfer of any particular grain size is independent of other grain sizes. A variety of processes, such as shielding in which fines are protected from movement by larger clasts, or the decreasing ability of the eroding process to carry additional fines with increasing load clearly invalidates this hidden assumption. This observation, that sediment transport is dependent not only on grain size, but also on the interaction among the different grain sizes present, appears to have two significant effects. First, as stated previously, the transfer functions do not increase monotonically throughout the complete size distribution of the sediment source; and second, the position of the transfer function on the ab-

SEDIMENT TRANSPORT AND GRAIN-SIZE DISTRIBUTIONS

461

TABLE 2.--Scaled sediment-transfer coefficient values of the transfer functions from Table 1 (shown in Fig. 3) with increasing flow rates (m s-')

z lid

Transfer Functions (derived from Table 1) Grain Size tA wherek=0.il (0.56 m s -L) ta wllerek=0.17 (0.61 m s-~) tc wherek=0.29 (0.66 m s ') tD
wherek=0.53

o
tl. tU 0

(0.74 m s-')

tlJ Lk B z
tlz

_o
w

uJ

:E
~ uJ -4
-a tA -2 -1

os(o)~

~=0.56 .n/sec sk = - 0.30

tc
tD

9 = 0 . 6 6 m/sec sk = -0.34 ~=0.74m/sec sk = -0.08

t B 7=0.61rn/sec
sk = -0.38

4.00 3.00 2.50 2.00 1.50 1.25 1.00 0.75 0.50 0.25 0.00 -0.50 -0.75 - 1.25 - 1.75 - 2.25 - 2.67 -3.00 - 3.25

0.01 0.09 0.20 0.26 0.27 0.20 0.19 0.21 0.18 0.16 0.11 0.08 0.07 0.06 0.05

0.01 0.10 0.23 0.30 0.30 0.25 0.25 0.28 0.27 0.27 0.20 0.17 0.16 0.15 0.13 0.06 0.03 0.01 0.01

0.02 0.17 0.34 0.41 0.41 0.36 0.37 0.43 0.44 0.46 0.37 0.33 0.31 0.29 0.28 0.19 0.12 0.06 0.02

0.03 0.18 0.39 0.48 0.48 0.47 0.47 0.58 0.65 0.76 0.69 0.71 0.69 0.64 0.67 0.5 ! 0.4 l 0.24 0.08

FIG. 3.-- Weight percentage of sediment distribution of bed material (histogram) and the resultant transfer functions (from Table 2) under different flow regimes (from data in Day 1980).

scissa (Fig. 3) is dependent on the size distribution of the source sediment. For example, we have derived transfer functions associated with sediments of different distriThe Shape of X(s) butions and ranges o f grain 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 digrain sizes in the bed material. rection o f 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 o f the derived transfer crease monotonically over the complete distribution of functions as tl(S) and t2(s), and hence determine the relg(s), each t(s) is a negatively skewed curve, indicating that ative changes in grain-size distributions between d2(s) and the function does increase over more than half of the dl(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 sumsorted, and more negatively skewed (Table 1), as proved marized in Figure 4. in the Appendix with a monotonically increasing func(1) t~ < t2 (energy is increasing in the direction of transtion, suggests that t(s) does fulfill the assumption to a port): The resultant X(s) may be calculated by letting tl = degree sufficient to produce the predicted trends. tA (Table 2), which is the lowest energy function, and t 2 The curves illustrated in Figure 3 show that, as the flow = tD, the highest energy function. As seen in Table 3 and rate increases, the distribution oft(s) changes from a rel- Figure 5, the result (XA,D) has a negatively skewed disatively high negative skewness to nearly symmetrical. We tribution. A similar function is derived for any of the h, propose to call the negatively skewed distributions that t2 pairs provided tl < t2. Because of its negative skewness, result from relatively low flow rates low-energy functions, the function is monotonically increasing over most o f the and the near-symmetrical distributions that result from sediment distribution, and, therefore, we may apply the high-flow rates high-energy functions. Because these terms results of the Appendix which indicate that dz(s) will be are relative to the grain-size distribution of the source finer and more negatively skewed than dl(S). This situasediment, 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 d2(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) tl > t2 (energy decreasing in the direction of transtransfer function to operate on it. In other words, the port, and tl is a low-energy function): X(s) may be illus-

same transport process may be represented by a highenergy transfer function when acting on fine sediments, and by a low-energy transfer function when acting on coarse sediments.

462
,

PA TRICK M c L A R E N AND DONALD BO WLES

t l < t 2 (energytncreasing; t I and t2 either high or low energy

~u~eons)

mm m

I - .3 z UJ o 1.1. LI.

XA, D s k = - 0 . 4 6 ~

.2
o

cc
.*, sediment becoming finer and more negaUvely skewed in the direction of transport.

cn z <
I-IZ UJ :S

IJJ IJ.

t l > t 20mergy decrming; t I is a low energy function)

11

c3 uJ
mm m

0 -4

-3

-2

-1

s(O)
FIG. 5.--X(s) distributions (from Table 3) calculated from various combinations of tl and t2.

sediment becoming finer and more negalively skewed In the direclion of transport.

In addition to the shapes of the transfer function X(s), the model presented in Figure 2 suggests that two other forms of transfer functions may occur. First, in the event that g(s) is known, then any deposit (d~(s)) can be related to it by ~ ( s ) = kg(sX~ )(1 - t~+, ). The above rules are not affected by this somewhat different form of the transfer function represented by (t,). (1 - tn.t). Second, in the event that t: = 0, then dz(s) is a final or total deposit (McLaren 1981) and can be related to dl(s) by d2(s) kdl(s)(tt) 1 tt

t l > t 2 (energyclecrming; t 1 isa high energy function; t 2 is

either high or low)

m m

sediment becoming coarser and more pmiUvely skewed in the direction of transport.

FIG. 4.--Diagrammatic summary of the resultant X(s) functions relating deposits in the direction of transport given selected combinations

of tl(s) and t2(s). trated by letting tl = t B and t2 = tA (from Table 2), the result of which (XB.A) is also a negatively skewed distribution (Table 3, Fig. 5). Thus, we can expect d2(s) t o be finer and more negatively skewed than d~(s). (3) tj > t2 (energy decreasing in the direction o f transport, but tl is a high-energy function and t 2 m a y be high or low): We can illustrate the resultant X(s) by letting tt = tD and t2 = tc (from Table 2) which is plotted as XD. in Table 3 and Figure 5. In this situation, X(s) becomes positively skewed, indicating that now, most of the function is decreasing monotonically, the reverse of the assumption used in the Appendix. The results will, therefore, be opposite to those in the proof and d2(s) will be coarser and more positively skewed than d~(s).

tl is always neg1 - tt atively skewed, regardless of whether t~ is a high- or lowenergy function; thus, d2(s) will be finer and more negatively skewed than d~(s).
APPLICATIONS

The transfer function represented by

We now wish to determine if the changes predicted by the above model can be observed in natural environments for which the sediment-transport direction is known. Our purposes are to (1) establish a real-world credibility for the model, (2) provide a method to predict the pattern of sediment transport from changes in grain-size distributions, and (3) obtain information concerning the relative energy of the transport regime as summarized in Figure 4. According to the models presented in Figures 1 and 2, grain-size distributions will change in response to erosion,

SEDIMENT

TRANSPORT AND GRAIN-SIZE

DISTRIBUTIONS

463

TABLE 3.--X(s) distributions derived from various combinations o f two

transfer functions (Equation 2) listed in Table 2. For example, XAD indicates t~ = tA and t2 = tDffOm Table 2. The functions are plotted in Figure 5
Grain Size 4-' X~o tA < ID Xa~ tn > t^ Xixc tt, > tc

\SKEWNESS

4.00 3.00 2.50 2.00 1.50 1.25 1.00 0.75 0.50 0.25 0.00 -0.50 -0.75 - 1.25 - 1.75 - 2.25 -2.67 -3.00 -3.25

0.01 0.08 0.15 0.18 0.19 0.13 0.12 0.11 0.08 0.05 0.04 0.03 0.02 0.02 0.02

0.01 0.10 0.24 0.32 0.31 0.27 0.27 0.31 0.30 0.31 0.22 0.19 0.18 0.17 0.14 0.06 0.03 0.01 0.01

0.03 0.18 0.42 0.54 0.54 0.57 0.56 0.79 1.04 1.71 1.40 1.64 1.54 1.26 1.46 0.84 0.61 0.30 0.09

~ T R A N S P O R T DIRECTION

CASE B TRANSPORT DIRECTION~ FOR FINING SEDIMENTS"Y

\ \ '~

FIG. 6.--Summary of the changes in grain size measures (0) that may occur in a given direction of transport. For example, if in the transport direction sediments are becoming coarser, the skewness will become increasingly positive; conversely, if sediments become finer, the skewness ~ become more negative. Sorting (variance) will become better in either direction but, depending on which side of the "peak" transport begins, it may first become poorer.

transport, and deposition in such a way that a lag will be coarser and more positively skewed than its source (case A), whereas sequential deposits will become either finer and more negatively skewed (case B) or coarser and more positively skewed (case C) (Fig. 6). Variance or sorting must eventually become better in each of the three cases. It is noted that case A and case C produce identical trends and that case A by itself does not provide a transport direction. The differentiation between cases A and C will depend on the geological interpretation o f the environments being sampled. Method In reality, a perfect sequential change in grain-size distributions in the down-current direction, as illustrated in Figure 2, is seldom achieved due to complicating factors such as variability in the "original source" (g(s) in Fig. 2), local and temporal variability in the transfer functions, and a variety of sediment sampling difficulties (see McLaren 1981 for further discussion on sampling). Therefore, we adopt a statistical approach to determine the transport direction by examining all possible pairs in a sample suite. Given a sequence o f n samples, there are - directionally oriented pairs that may exhibit a 2 trend suggesting transport in one direction, and an equal number of pairs in the opposite direction. When any two samples are compared with respect to their mean size, sorting, and skewness, eight possible trends exist; compared to d~, d2 may be (1) finer (F), better sorted (B), and more negatively skewed ( - ) ; (2) coarser (C), more poorly sorted (P) and more positively skewed (+); (3) C, B, - ; (4) F, P, - ; (5) C, P, - ; (6) F, B, +; (7) C, B, +; or (8) F, P, +. O f these trends, only two are indicative of transport, namely F, B, - (case B), and C, B, + (case C), for
n 2 -- n

which there is a one-eighth probability of either occurring at random (p = 0.125). Because of the uncertainty assod a t e d with variance, which can become larger (more poorly sorted) before becoming smaller, we choose only to accept better sorting as the criterion in the two cases suggesting the direction of transport. To determine if the number of occurrences o f a particular case exceeds the random probability of 0.125, we test the following two hypotheses: Ho: p -< 0.125, and there is no preferred direction; and H i : P > 0.125, and transport is occurring in a preferred direction. Using the Z-score (Spiegel 1961) in a one-tailed test, Ht is accepted if Z or

x - Np __ > 1.645 (0.05 level of significance)

x/tim

> 2.33 (0.01 level of significance), where x = observed number of pairs representing a particular case in one of the two opposing directions; and N = total number of possible unidirectional pairs. N = n 2 -- n - where n = number of samples in the sequence; 2 p = 0.125; and q = 1.0 - p = 0.875. The Z-statistic is considered valid for N -> 30 (i.e., a large sample). Thus, for this application, a suite of 8 or 9 samples is the m i n i m u m required to evaluate ade92 - 9 quately a transport direction (i.e., - - 36, the total 2 possible pairs in one direction).

464

PA T R I C K M c L A R E N AND DONALD BO W L E S

\
= ======~~oad Trap

TABLE4.-- Grain-size data from the East Fork River, Wyoming (calculated from data in Emmett et aL 1980)
Location (see Fig. 7) Moment Measures (0) Me.an Sorting Skewne~

Sample

' 1 3

~ 9

8 ~

8 985

ing
River

1479~

12"41 1077

(_~:~;~:~ 1766 1662 ~1830

___
I 20Oral

f ....

FIG. 7.--Map of the study reach in the East Fork River, Wyoming, showing the 22 bed-material sample locations (after Emmett et al. 1980).

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22

043 075 137 220 301 421 516 602 708 808 898 985 1,077 1,155 1,241 1,315 1,396 1,481 1,662 1,695 1,766 1,830

0.70 0.68 1.32 0.70 0.86 1.65 0.92 0.93 1.20 1.38 1.35 1.55 0.89 0.92 0.84 1.74 0.94 1.26 1.44 1.26 0.95 1.58

0.97 0.97 1.20 1.04 1.11 1.42 1.22 1.04 1.10 1.00 1.35 1.17 1.04 1.25 1.15 1.25 1.37 1.31 1.06 1.07 1.28 1.99

0.63 0.62 0.20 0.92 0.55 0.35 0.40 0.37 0.35 0.42 0.55 -0.07 0.29 0.66 0.69 -0.24 0.78 0.55 0.36 0.26 0.65 0.49

E x a m p l e 1: Fluvial Transport

T h i s e x a m p l e uses grain-size distributions o f b e d - m a terial samples from the north-flowing East F o r k River, W y o m i n g (data calculated from E m m e t t et al. 1980). This river rises in the W i n d R i v e r Range a n d flows about 50 krn before arriving at the 2-kin-long study reach (Fig. 7). Here the stream is a b o u t 30 m wide a n d m e a n d e r s o v e r a flood plain which is confined by glacial outwash terraces o f sand a n d gravel. These terraces p r o v i d e a continuous source o f fresh s e d i m e n t wherever the river impinges laterally against them. A series o f 22 b e d - m a t e r i a l samples averaged across the channel, were collected between 43 m a n d 1,830 m u p s t r e a m from a reference p o i n t m a r k e d by a b e d - l o a d trap (Fig. 7). The m o m e n t measures reported in Table 4 were calculated from the s a n d fraction only. The 22 s a m p l e s (n) p r o v i d e a possible 231 (NO " n o r t h trending" pairs and, conversely, 231 "south-trending'" pairs. O f the possible cases indicative o f a t r a n s p o r t direction, only case C (coarser, better-sorted, a n d m o r e positively skewed) in the north direction is significant (Table 5), d e m o n s t r a t i n g that grain-size distributions are changing in the d o w n s t r e a m direction m a m a n n e r predicted by the model. The occurrence o f the case C t r e n d also indicates that the energy regime o f the river is tending to decrease downstream, although the transfer functions must h a v e "high-energy" shapes with respect to the sand-size distributions present in the river.
E x a m p l e 2: Delta-Lacustrine Transport

The lake occupies a glacially e x c a v a t e d valley which is 27 k m long a n d 6 k m wide a n d contains a flat-floored basin below 100 m (Fig. 8). T h e G o d l e y R i v e r forms a delta at the northern end, the foresets o f which slope southwards to merge with the basin floor 6 k m away. The lake itself is d a m m e d in the valley b y m o r a i n e a n d outwash deposits through which the T e k a p o R i v e r flows. A n e x a m i n a t i o n o f grain-size trends a m o n g 25 grab samples (Table 6) taken along the axis o f the lake suggests the following: (1) O f 300 possible pairs c o n t a i n e d in the c o m p l e t e suite o f samples, case B in the south direction is the only significant trend (Table 7). Therefore, for the lake as a whole, s e d i m e n t trends predict accurately the transp o r t direction. T h e y also suggest that the transport processes produce low-energy transfer functions a n d that energy is generally decreasing in the transport direction, b o t h o f which are reasonable in this lacustrine setting. (2) F o r the nine delta slope s a m p l e s case B is again the preferred t r e n d (Table 7), which is also perfectly reasonable, given a decreasing energy regime with increasing d e p t h d o w n the delta foreslope. TABLE 5.--Summary of the numbers of pairs of East Fork samples
(Table 4) producing transport trends. N, x, and Z are defined in text
North Trend South Trend

Case B Case C

F
B

N = 231
x = 29

C
B +

Z = 0.02 N = 231
x = 71 Z = 8.38'

T h i s e x a m p l e uses grain-size d a t a from Lake Tekapo, a deep, glacier-fed lake in N e w Z e a l a n d (data p r o v i d e d by R. A. Pickrill, pers. comm.; Pickrill a n d Irwin 1983).

N = 231 x = 38 Z = 1.82 N = 231 x = 33

Z = 0.82

t Significant at the 0.01 level.

SEDIMENT TRANSPORT AND GRAIN-SIZE DISTRIBUTIONS

465

TABLE6.--Grain-size datafrom Lake Tekapo, New Zealand (datafrom

R. A. Pickrill, pers. comm.)
Moment Measures () Sample Mean Sorting Skewness

1 2 3
4

5.03 6.20 6.38

4.07

1.78 1.67 1.65

1.93

0.54 -0.24 0.05

0.67

Delta Slope

5 6 7 8 9
10 11 12

6.67 7.19 7.74 7.12 7.65

8.00 7.99 7.80

!.71 1.44 1.07 1.36 1.13

0.82 0.85 0.98

-0.15 -0.34 -0.70 -0.22 -0.64

- 0. 96 - 1. 13 - 0. 84 Basin

13 14 15 16 17 18 19 20
21

7.71 7.03 7.79 7.95 8.10 8.00 7.92 7.99

7.86

1.04
1.10

22 23 24 25

7.98 8.05 8.15 7.40

0.99 0.84 0.67 0.85 0.94 0.93 1.00 0.9 i 0.79 0.67 0.59

-0.63 -0.44 -0.66 -0.86 -1.44 - 1.06 - 1.01 -0.99

- 0. 87

South Slope

- 1.02 - 1.44 - 1.27 0.28

FxG, 8.--Lake

Tekapo (New Zealand) and sample locations (aiter R.

A. PickriU,pets.comm.).

currents m o v i n g upslope have been observed in similar 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., tl > h ) . On the contrary, not only are higher energies associated with shoaling water, but southward currents will be concentrated into the outflow. Mathematically, it was shown that sediments also become finer, better sorted, and more negatively skewed if energy increases in the direction o f transport (i.e., h < t2; Fig. 4), but it was argued that successive deposits would be r e m o v e d by erosion with an increasing energy regime. This example tends to suggest that such deposits can remain, probably as a result o f the difficulty in resuspending fine silt and clay-sized particles once they have been deposited. There is also an apparent southward case C trend which is significant at the 0.05 level (Table 7). This is the result o f sample 25 taken close to the outflow, which is coarser, better sorted, and m o r e positively skewed than any of the other south slope samples (Table 6). Case C is unlikely as it demands a decreasing energy regime; therefore, case A is accepted as a logical interpretation whereby sample 25 is a lag o f those sediments presently located in deeper water.
E x a m p l e 3: Longshore Transport

(3) Within the basin, ease B trends are significant in both directions, suggesting the occurrence o f transport in the low-energy regime. Two reasons which m a y be responsible for obscuring a preferred direction are sediment input from other rivers entering both sides o f the lake, such as the Mistake and Cass rivers (Fig. 8), and the breakdown o f the thermocline in winter, resulting in weak and r a n d o m b o t t o m currents (Pickrill and Irwin 1983). (4) Eight samples from the south slope show an apparent case B trend in the south direction (Table 7), indicating that sediment transport is occurring upslope from below 100 m to the Tekapo River outflow. Such

Coburg Peninsula, a spit in the Strait o f Juan de Fuca near Victoria, British Columbia, originates at its south-

466

PA T R I C K M c L A R E N A N D D O N A L D BO W L E S

TABLE7.--Summary of the numbers of pairs of Lake Tekapo samples (Table 6) producing transport trends. N, x, and Z are defined ~n text
All 25 Samples N Trend S Trend N Trend Delta Slope S Trend N Trend Basin S Trend N Trend South Slol~ S Trend

Case B

F
B

N = 300
x=34

N = 300
x=213

N = 36
x=4

N = 36
x=26

N = 28
x= 13

N = 28
x= 13

N = 28
x=6

Case C

C
B

Z=-0.61 N = 300
x=5

Z=30.64 ~ N = 300
x=20

Z=-0.25 N = 36
x= 1

Z = 10.83~ N = 36
x= 1

Z=5.431 N = 28
x=0

Z=5.43 ~ N = 28
x=0

Z= 1.43 N = 28
x=0

N = 28 x = 12 Z=4.86 t N = 28
x=7

Z--5.67

Z--3.06

Z~-1.76

Z~-1.76

Z=-2.00

Z=-2.00

Z~-2.00

Z=-2.002

Significantat the 0.01 level. 2Significantat the 0.05 level.

west end from a low, eroding bluff composed of till (Fig. 9). The spit extends northeast about 2.5 kin, where it terminates at a narrow tidal channel. The direction of spit growth a n d several other similar spits on both sides of the Strait o f Juan de Fuca confirm a longshore transport direction which is p r e d o m i n a n t l y eastwards. A sequence of eight samples from the lower beach face and two samples from the eroding till bluff reveal that all of the beach samples are coarser, better sorted, and more positively skewed than the till (Table 8). This suggests case C transport, which is consistent with the till being the d o m i n a n t sediment source for the beach. The eight beach-face samples also show a significant ease C trend

to the northeast (Table 9) in the direction of spit growth, illustrating n o t only that the m o d e l appears to have realworld validity, but can also predict correctly the direction o f sediment transport. DISCUSSION A N D CONCLUSIONS

Esquimalt

w.couvER ( \

//8 / / 7 Coburg //6 Pe.,nsu,.

4

Till ~4 2

~Q~'

'"'7
/
Head Albert

o,:

(
~ L~:~'7~" ~

We have attempted to demonstrate that grain-size dist f i b u t i o n s change in the direction o f transport according to the shape of the transfer function X(s). Because X(s) is the result of two transfer functions, (h(s) a n d t2(s)), whose shapes can be d e t e r m i n e d empirically, we can demonstrate that sequential deposits m a y become either coarser, better sorted, a n d more positively skewed (highenergy h(s)), or finer, better sorted, and more negatively skewed flow-energy t~(s)) with a decreasing energy regime. It is interesting to note that sediments c a n n o t become coarser forever because, with coarsening, it becomes less a n d less likely that the transport processes will m a i n t a i n high-energy characteristics with respect to the coarsening sediment. As the deposits become coarser, the transfer function describing the processes will take on the characteristics o f the low-energy function, a n d the sediments will become finer again. The model indicates that sediments can also become finer in the direction of transport with a n increasing energy regime. This somewhat surprising result appears initially to be o f theoretical value only as i n t u i t i o n would suggest that down-current deposits could n o t r e m a i n to be observed. However, one example used in the determ i n a t i o n of transport direction suggests that such deposits can remain, possibly as a result o f cohesion in fine sediments or a high-sediment-supply rate.
TABLE 8.--Grain-size data from the Coburg Peninsula, British Columbia
Sample M~.ln Sorting Skcwtem

I 2 3 4

1.36 0.27 0.32 0.30

1.19 0.80 0.98 0.94

0.08 1.29 1.03 0.96

Eroding Till BlulP Beach Face

FIo, 9.-Coburg Peninsula (VancouverIsland) and sample locations.

5 0.27 0.80 6 -0.04 0.80 7 -0.27 0.73 8 0.00 0.78 9 0.09 0.95 1 Average of 2 samples.

1.13 1.70 2.93 2.35 1.60

SEDIMENT TRANSPORT AND GRAIN-SIZE DISTRIBUTIONS

TABLE 9.--Summary of numbers of pairs of Coburg Peninsula samples (Table 8) producing transport trends. N, x, and Z are defined in text
NE Trend SW Trend

467

Case B

F B C B +

N = 28 x=0 Z = -2.00 N = 28 x = 14 Z=6.00 ~

N = 28 x=3 Z = -0.29 N = 28 x=6 Z = 1.43

Case C

' Significant at the 0.01 level.

We have also modified the earlier model proposed by McLaren (1981) which indicated that sediments always become better sorted in the direction of transport. Although not necessarily true, we suggest that in reality, successive deposits seldom become more poorly sorted, particularly when sediment distributions are described in phi units. The log-transformation tends to make grainsize distributions relatively symmetrical, a requirement that ensures a decrease in variance when the transfer function is predominantly increasing or decreasing monotonically. McLaren (1981) also suggested that sediments could become finer and more positively skewed, which this new analysis, using realistic transfer functions, demonstrates is incorrect. When sediments become finer, the skewness must become more negative (Fig. 6). At present, we have made no attempt to explain the processes responsible for determining the shape of X(s). We have, however, demonstrated a method for the prediction of sediment transport paths, regardless of the process, that appears to give the correct directions in a wide variety of environments, even though we are using only the mean, sorting, and skewness as sediment-grain-size descriptors. It would clearly be preferable to base the prediction of a transport trend on the complete distribution of X(s), but at present there is a lack of suitable experimental data which would allow us to examine in detail all the shapes that X(s) may take. In the future we hope to utilize the complete distribution of X(s) in the determination of a transport trend and to correlate its shape with known processes and depositional environments.
ACKNOWLEDGMENTS

The writers would like to thank E. Klovan, R. E. T h o m son, and R. W. Dalrymple for critically reading earlier drafts and providing the basis for considerable improvements. T. J. Day also provided advice and loaned appropriate data. The help of R. Currie for necessary computer programming and D. Chisholm for typing the manuscript is gratefully acknowledged.

REFERENCES ANDERSON,J. B., WOLFTEICH, C., WRIGHT, R., AND COLE) M. L., 1982, Determination of depositional environments of sand bodies using vertical grain-size progressions: Transactions--Gulf Coast Association of Geological Societies, v. 32, p. 565-577.

BLATT, H., M1DDLETON, G., AND MURRAY, R., 1980, Origin of Sedimentary Rocks: Englewood Cliffs, New Jersey, Prentice-Hall, 782 p. DAY, T. J., J980, A study of the transport of graded sediments: Report No. IT 190, Hydraulics Research Station, Wallingford, England, 11 p. EINSaaZIN, H. A., 1950, The bed-load function for sediment transportation in open-channel flows: U.S. Dept. of Agriculture, Tech. Bull. No. 1026, p. 1-70. EMMETr, W. W., MVmCK, R. M., AND MEADE, R. H., 1980, Field data describing the movement and storage of sediment in the FEast Fork River, Wyoming, Part I. River hydraulics and sediment transport, 1979: United States Department of the Interior, Geological Survey, Open File Report 80-1189, 43 p. FLEMMING, B. W., 1982, Sediment mixing." its natural occurrence and textural expression: Abstract, in Eleventh International Congress on Sedimentology: McMaster University, Hamilton, Canada, August 2227, 1982, p. 81. FOLK, R. L., AND WARD~ W. C., 1957, Brazos River bar: a study in the significance of grain size parameters: Jour. Sed. Petrology, v. 27, p. 3-26. FRIEDMAN, G. M., 1961, Distinction between dune, beach and river sands from textural characteristics: Jour. Sed. Petrology, v. 31, p. 514-529. , 1979, Address of the retiring President of the International Association of Sedimentologists: differences in size distributions of populations o f particles among sands of various origins: Sedimentology, v. 26, p. 3-32. G~a~a~ow, H. C., 1982, A theoretical model o f source and transport effects on grain-size distributions: abstract, in Eleventh International Congress on Sedimentology: McMaster University, Hamilton, Canada, August 22-27, 1982, p. 80. (~ESSLER,J., 1976, Beginmng and ceasing of sediment motion, m Shen, H. W., ed., River Mechanics, vol. I: Fort Collins, Colorado, H. W. Shen, p. 7-1-7-22. GHOSH, J. K+, M.AZU'MDER,B. S., AND SENGUPTA,S., 1979, Methods of computation of suspended load from bed materials and flow parameters: Indian Statistical Institute, Calcutta, Technical Report No. P and E/Flume/I/1979, 34 p. GIBBS, C. J., AND NEILL, C. R., 1972, Interim report on laboratory study o f basket-type bed-load samplers: Research Council of Alberta, Report No. REH/72/2, 5 p. GLAISTEm, R. P., AND NELSON, H. W., 1974, Grain-size distributions, an aid in facies identification: Bull. Canadian Petroleum Geology, v. 22, p. 203-240. HANER, B. E., 1984, Santa Aria River: an example of a sandy braided floodplain system showing sediment source area imprintation and selective sediment modification: Sedimentary C~ology, v. 38, p. 247261. KaUMaEIN, W. C., 1938, Size-frequency distributions of sediments and the normal phi curve: Jour. SOd. Petrology, v. 8, p. 84-90. MASON, C. C., AND FOLK, R. L., 1958, Differentiation of beach, dune and aeolian flat environments by size analysis, Mustang Island, Texas: Jour. Sed. Petrology, v. 28, p. 211-226. McCAvE, I. N., 1978, Grain size trends and transport along beaches: example from eastern England: Mar. Geol., v. 28, M43-M51. MCLAREN, P., 1981, An interpretation of trends in grain size measures: Jour. Sed. Petrology, v. 51, p. 611-624. ---, 1982, Hydraulic control of grain-size distributions in a macrotidal estuary--discussion: Sedimentology, v. 29, p. 437-439. , 1983, Coastal sediments of the Strait o f Juan de Fuca: implications for oil spiUs, in Current Research, Pt. A, Geol. Surv. Canada, Pap. 83-1A, p. 241-244. ---, 1984, The Whytecliffofl spill, British Columbia: sediment trends and off movement on a beach,in Current Research, Pt. A, Geol. Surv. Canada, Pap. 84-1A, p. 81-85. McLAREN, P., BARRIE, W. B., AND SEMPELS, J. M., 1981, The coastal morphology and sedimentology of Cape Hatt: implications for the Baffm Island Off Spill Project (BIOS), in C u r r ~ t Research, Pt. B, Genl. Surv. Canada, Pap. 81-1B, p. 153-162. McMANus, D., 1982, Phi and sediment size analysis: discussion: Jour. Sed. Petrology, v. 52, p. 1011-1014. MXDDLE'roN,G. V., 1976, Hydraulic interpretation of sand distributions: Jour. Geology, v. 84, p. 405-426.

46 8

PA T R I C K M c L A R E N A N D D O N A L D B O W L E S

Moss, A. J., 1962, The physical nature of common sandy pebbly deposits, Part I: Am. Jour. Sci., v. 260, p. 337-373. PICKRILL, R. A., AND IRWIN, J., 1983, Sedimentation in a deep glacierfed lake--Lake Tekapo, New Zealand: Sedimentology, v. 30, p. 6375. SINGER, J. K., AND ANDERSON, J. B., 1984, Use of total grain-size distributions to define bed erosion and transport for poorly sorted sediment undergoing simulated bioturbation: Marine Geology, v. 57, p. 335-359. SUNGERLAND, R. L., 1977, The effects of entrainment o f t b e hydraulic equivalence relationships of light and heavy minerals in sands: Jour. Sed. Petrology, v. 47, p. 753-770. SPmGEL, M. R., 1961, Theory and Problems of Statistics: Sehaum's Outline Series: New York, M ~ . raw-Hill Book Co., 359 p. STAPOR, F. W., AND TANNER, W. F., 1975, Hydrodynamic implications of beach, beach ridge and dune grain size studies: Jour. Sed. Petrology, v. 45, p. 926-931. UDDEN, J. A., 1914, Mechanical composition of elastic sediments: Geological Society of America Bulletin, v. 25, p. 655-744. VISHER,(3. S., 1969, Grain size distributions and depositional processes: Jour. Sed. Petrology, v. 39, p. 1074-1106. APPENDIX As a first step in the analysis, we will consider the situation as outlined in Figure 1, where: g(s) is a function of "s" (grain size in units) describing any gramsize distribution; t(s) is a transfer function that is a monotonically increasing function o f " s ' ; t(si) < t(sj) V i < j and t(s~) -< 1 V i; r(s) is the size distribution of sediment in transport that is derived
f r o m g(s);

xl
[ I~J

rn

g [s)

X
[ tel

In= id(S,)d.
g(s,)

FIG. 10.-- Diagram illustrating the end result of any function g(s) after repeated multiplication with a monotonically increasing or decreasing function to derive r(s) and d(s), respectively.

g,s ,t j +
0 as n

. .[t(sO~

. .[ (s,)T
+...

Because it was assumed that t'(Sl) > t'(s2) > t'(s3), then

[t'(s0-p
t'(s,)J
~(s,)

aovi>

t'(s) equals I - t(s) and is a monotonically decreasing transfer function; d(s) is the first deposit derived from S(s); it may be considered ala$ remaining behind after r(s) has been removed. Let r(s3 = kg(~t(,~) where
I

and
g(s,) S(s,) + 0 + 0 + . . .
~ 1 asn~co.

[AI

Because

eL(sO
i* 1

= 1,

(k is applied to normalize r(s))

then d~(sO~O and as n ~ c o f o r i > 1

~ g(si)t(si)
i-[

and d ( ~ = k'g(s~)t'(~ k'

i-i

d.(s,) ~ 1 ( F i g . 10).

By the converse argument, consider r . ( ~ when i = N (i.e., the last class interval). Expanding from Equation A, as above, gives r.(~) =
=
.

We will now apply t'(s) n times in order m generate ~(s), which may be considered as the ~ lag of g(s) (Fig. I0). Then,

g(s0t(s,)"

+ g(s2)t(s2)" . g ( s ~ ) t ( ~ ) "

d.(s3 = k'~g(s0YCs0 n
where
1

[B]

,lt(s,)T It(so?
s(s~)
.

A g a i n , b e c a u s e t(s,) < t ( s 9 < t(s~), t h e n

k'o
i-I

[C] and

t~J

as

n~ooVi<N,

~ g(s0t'(s~) ~ Note that for n = 0, do(s3 = g(~. Consider d~(si) when i = 1 (i.e., the first class interval). From Equation B, d.(s,) = k'ng(s0t'(s~p. Substituting for k'. from Equation C gives
S(sat'(s,) s(s,)t'(s,)" + g(s~)t'(s~p + . . .

~(~)
0 + 0 +...
1 as n-pp.

g(sN)

Because
N

i-I

SEDIMENT

TRANSPORT

AND GRAIN-SIZE

DISTRIBUTIONS

469

!
/i

2
I

i ~

1-8

/ i / i
/ -~ ' /

I i
. _I I L ~ , ~'-'--'] I

I i

%.
~. ~.

I SEDIMENT IN
I~ TRANISPORT i
I I

O0-~-- n

n--~00

rl

rn

N-1

no.of times transfer function (t(s)) n = acts on sediment source (g(s)}

Fro. 11.--Diagrammatic illustration of the changes in the variance of r(s) and d(s) relative to g(s). Although the variance of both r(s) and d(s) must ultimately approach zero, the location o f n ~ O and the variance of g(s) may result in a2 or (r~ increasing before their eventual 2 decrease. In this illustration, ~2 has increased at n ~ 1. then r.(sJ-O and ro(s~) I (Fig. 10). as n~fori<N

FIG. 12.--Diagram illustrating how skewness must approach - ~ in the direction of transport. The value "a" is arbitrarily small and is approaching zero with the result that S k ~ - l / a ~ - o o as a ~ 0.
As
n~Go
aT, 2 ~ ~0. Si,i2 - - SN2

Similarly, as ad.2 ~ S~2 -- S,2

0.

Mean Grain Size

Vr = ~ r(~st, where ~ = the mean grain size ofr(s), the distribution
i-I

of sediment in transport (Fig. 1). Let U.. = ~ r=(s,)s.

i-,

which, by applying Equation A, becomes ~. = ~ ~ g(st)t(sO~s~.

i-[

As
n ~

~, - sN (Fig. 10).
]~cause <

SN,

then ~ >~Vn. This proves that, if the transport function t(s) is increasing monotonically, then sediment in transport must be finer than its source. By the converse argument:

We have shown that the variance of the lag and the variance of sediment in transport will approach zero when t(s) is applied " n " times. At n = 0, there is neither sediment in transport nor a lag produced, and we are at the "starting point" (i.e., the sediment source, $(s)). We know 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 coincide then both the lag and sediment in transport must become better sorted for all "n." If they do not, then decreasing variance may not be true for all "n." This is illustrated graphically in Figure 11, which shows the changing magnitude of variance for a lag and for sediment in transport (the exact shape of the curve has been drawn arbitrarily). The value of the variance for g(s) determines the position of n = 0. At n = 1, the sediment in tnmsport has increased its variance, which will not besin to decrease until n = 3. In reality, the position o f n = 1 is arbitrary and " n " itself has little physical meaning. It could, for example, represent an individual erosion event such as one wave, or it could be the combined effects of ten years of waves. Because the phi scale results in approximately Gaussian or normal distributions which are symmetrical, we believe that ~ and the maximum variance of a2 or a2 will coincide closely with each other and that an increasing variance for r(s) or d(s) will be rarely observed.

Skewness Let
Sk, =

t% < ~s V n (Fig. 10).

Thus, the first deposit (or lag) must be coarser than its source.

1
i-I

where Slq is the skewness of the source sediment. Similarly,

Variance (SortingJ Let

N

Sk~. = (~,?p,2 ,-, (s~ - ~)3r(s0, where Sk,. is the skewness of the nth sediment in transport. As n ~ oc, let r,(sN) ~ 1 - a,
r.(~_,) - a,

i-|

where a2 is the variance of r(s), the grain-size distribution in transport and

i-i

r.(s0~0

for

i=1,2,3 .....

N-2,

470 where " a " is arbitrarily small (Fig. 12). Then

PA T R I C K M c L A R E N A N D D O N A L D BO W L E S
Similarly, as
S~ ~ +oo;

g . - s ~ ( l - a) + s ~ _ ~ a
sN + a(SN--L -- S~)

.'. Sk~ > SksV n. We have shown, therefore, that the skewness of sediment in transport m u s t become more negative than the source sediment, and the skewness of a lag m u s t become more positive.

and
o~.2 ~ ( s , - u ~ f ( l - a) + (sN_~ - w f a

a(sN_, - sn)2. Thus, (ss - v~.)3r~(sO ~ a(s~_, - sN)3

Summary
and Slq. a(sn_~ - sNY [a(sN_, - s.)2] ~'~
i
a

~-~

as a ~ O . .'. Sk~ < S k , V n.

Given a monotonically increasing transfer function (m units), it has been shown that sediment in transport m u s t become progressively finer and more negatively skewed than its source sediment. Conversely, the lag m u s t become coarser and more positively skewed than its soure. sediment. The change in sorting (variance) cannot be determined in the general case, although at some point in the transport path sorting must become better.