GrainSize, SedimentTransport Regime, and Channel Slope in Alluvial Rivers Author(s): W.Brian Dade and PeterF.

Friend Source: The Journal of Geology, Vol. 106, No. 6 (November 1998), pp. 661-676 Published by: The University of Chicago Press Stable URL: http://www.jstor.org/stable/10.1086/516052 . Accessed: 12/08/2011 16:59
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to The Journal of Geology.

http://www.jstor.org

Grain-Size, Sediment-Transport Regime, and Channel Slope in Alluvial Rivers 1

W. Brian Dade and Peter F. Friend 2
Institute of Theoretical Geophysics, Department of Earth Sciences and Department of Applied Mathematics and Theoretical Physics, University of Cambridge, Cambridge CB2 3EQ, UK

ABSTRACT The general relationship between channel morphology and the grain size of sediment in the channel bed is an important but poorly known aspect of alluvial rivers. An analysis of an equation for total sediment ux in the limits of suspension-, bedload-, and mixed-modes of transport indicates distinct, steady-state regimes of channel morphology. Such regimes are readily seen in published data for modern alluvial rivers by way of a conventional Shields plot or a plot of channel slope as a function of relative grain size d / h and the ratio w s / u , where d and w s are, respectively, * mean diameter and fall speed of bed sediments, and h and u are, respectively, mean depth and friction velocity of * the ow. With slope and mode of transport in an alluvial river constrained by grain size and channel depth alone, estimates of discharge and sediment ux follow directly. Introduction of the sediment ux relationship into conventional diffusion models for the evolution of an alluvial system provides nominal estimates of the response time for channel adjustment to some external changes. For some major modern rivers, this time of response along the entire length of channel is in the range 10 3 10 5 yr, underscoring the potential for complicated, long-time interaction of large alluvial systems with, for example, climatic variability.

Introduction The form of an alluvial river channel reects the range of environmental factors that determine the erosion, transportation, and deposition of unconsolidated debris by the river itself. An understanding of the interaction of these factors is important for the analysis of the response of modern rivers to human perturbation and to the interpretation of ancient uvial deposits. At present, however, the relationships between the hydraulic and sedimentary controls of the form of an alluvial channel are expressed in terms of empirical laws that may or may not be readily applied to new settings. An example of such an approach is the classication of rivers with plan forms characterized by single-thread (meandering and straight) or multiple-thread (braided) channels (Leopold and Wolman 1957). Most of these schemes proceed, implicitly or explicitly, from the assumption that a self-forming, alluvial river achieves a channel pattern with specic, uni1 Manuscript received December 31, 1997; accepted July 7, 1998. 2 Department of Earth Sciences, University of Cambridge, Cambridge CB2 3EQ, UK.

versal properties. The processes that result in such a pattern can include the adjustment of channel morphology to a state in which sediment ux through a given reach of a river, averaged over a suitable interval of time, is continuous. Under such conditions neither net erosion nor net deposition takes place, and the river is said to be graded in grade (Mackin 1948; Hack 1960), or in regime (Lacey 1930). This concept can be traced (in the English-language literature) back to the ideas of G. K. Gilbert, which are summarized by Baker and Pyne (1978), Chorley and Beckinsale (1980), and Leopold (1980). Despite its pedigree, the notion of the graded stream has not met with universal acceptance. One possibility is that real alluvial systems never perfectly achieve in time the dynamic equilibrium represented by the ideally graded channel. Important efforts to understand the evolution of an alluvial system include the consideration of the mathematical behavior of the system of equations that describe the spatial distribution of momentum and the continuity of water and sediment transport in a stream ow (Ribberink and van der Sande 1985).

[The Journal of Geology, 1998, volume 106, p. 661675] 1998 by The University of Chicago. All rights reserved. 0022-1376/98/10602-0004$01.00

661

662

W. B. DADE AND P. F. FRIEND

More recently, C. Paola and colleagues have made important advances in their consideration of a general system of equations that describe the transport of mass and momentum transport in a gravel-bedded stream as it relates to the evolution of alluvial basin ll (Paola et al. 1992), downstream distribution of grain size (Paola and Seal 1995), and the inference of channel slope in ancient braided streams (Paola and Mohrig 1996). In this paper we reconsider the general hypothesis that within a graded stream the total sediment ux q T (or, equivalently, sediment discharge per width of channel) is constant in time and does not vary in the downstream (x ) direction. This latter condition is given mathematically as q T / x 0. (1)

Table 1. Summary of Notation

Symbol A Units Denition

In this regard our results follow the ideas of Parker (1978, 1979) for self-formed, gravel-bed rivers. In our consideration of the ramications of eq. 1, however, we make the important distinction between alluvial systems characterized by regimes of predominantly bedload, mixed-load, or suspended-sediment transport. By considering mass ux per channel width, we avoid complications associated with converging channels and increasing discharge in the downstream direction. These complications represent an important aspect of alluvial-river form, of course, but are secondary to the main advance from previous studies. We nd that sedimentological control is reected in an approximately constant value of the Shields parameter for each of the three different transport regimes. A key assumption underlying this result is that the grain size of material making up a river bed is strongly correlated with (and thus in equilibrium with) the caliber of the sediment load in transport through a graded reach. In the following analysis we rst summarize a basis for distinguishing the different modes of sediment transport, and then propose a general equation for sediment transport that accommodates the different transport regimes and is based on energetic constraints of stream ow. We then derive predictive relationships that emerge from this equation for conditions of continuous sediment ux in the different sediment-transport regimes. These relationships are used to reanalyze existing data for sedimentological controls of channel slope and depth. Last, we consider several, specic, applications of our ndings and summarize the key ndings of our analysis. A summary of the notation used in this paper appears in table 1.

depth-averaged concentration of suspended sediment L2 area of drainage basin AD B mobility parameter C volumetric concentration of sediment D grain size parameter, (gRd 3 / 2) 1/3 K L 2 T 1 diffusion coefcient L L length of channel in drainage basin P Rouse parameter, w s / u* R relative excess density of sediment particles, ( p )/ Re ow Reynolds number, uh / Re grain Reynolds number, u d / * Q* L 3 T 1 total water discharge, uhw T basin response time TR CD quadratic drag coefcient d L median grain size of channel bed material g LT 2 acceleration due to gravity h L channel depth q L 2 T 1 water ux, Q / w qb L 2 T 1 bedload sediment ux L 2 T 1 suspended sediment ux qs qT L 2 T 1 total sediment ux s channel slope u L T 1 average ow speed friction velocity u L T 1 * w L channel width 1 ws LT settling velocity of individual, sedimentary particles x L downstream spatial coordinate z L vertical spatial coordinate grouping of coefcients , c D and s b fraction of total sediment ux in bedload b bedload transport efciency or friction factor s suspended-load transport efciency factor L bed elevation above arbitrary datum von Karmans constant L 2 T 1 kinematic viscosity of transporting uid density M L 3 Shields parameter sh / Rd 1 2 0 M L T bed shear stress, ghs relative depth, h / z b dimensionless sediment transport L 2 T 1 specic stream power, qs subscripts on C, u and b bedload i index for b, m or s m mixed load o stationary bed s suspended load Note M mass; L length; T time.

Analysis
Modes of Sediment Transport. Sediment transport associated with local hydraulic conditions occurs in two distinct modes (e.g., Middleton and Southard 1984). Sedimentary particles that roll, slide, skip, and hop along the bottom make up the bedload. Such material is transported primarily

Journal of Geology

CHANNEL SLOPE IN ALLUVIAL RIVERS

663

near the bed and is in contact with the channel bed for a considerable fraction of the time. In contrast, particles supported by the turbulence of the mean ow reside primarily in the water column and spend little time in contact with the bed. Such material makes up the suspended load. Sediment for which the transport mode is intermediate in character contributes to a third transport regime, that of mixed load. An additional mode of transport, wash load, is recognized by some workers and comprises extremely small particles that pass through a river system relatively independently of local ow conditions. We suggest that the removal of such material from an alluvial system would have little affect on channel morphology. Accordingly, we do not consider wash load in this analysis of total transport q T. The distinctions between suspended-, mixed-, and bedload transport are arbitrary, and in an individual river all three modes can occur to varying degrees. An objective basis for the distinction of the different transport modes can nevertheless be given as follows. In a channel ow that is neither eroding nor depositing sediment, the vertical distribution of material in the ow represents a balance between the effects of the downward settling and upward turbulent diffusion of individual particles. In such a setting, the volumetric concentration C (z ) at elevation z above the bed relative to a near-bed reference concentration C b at z b is given approximately by: C (z )/ C b (z / z b )P, (2)

and sediment concentration, respectively, of the bedload layer. From eq. 2 the relative, depth-averaged concentration of sediment A in a ow of depth h is given in terms of P and the relative depth h / z b by the expression A (h z b ) 1 (
1P

(z / z b )P dz

zb

(4)

1)/{(1 P )( 1)}.

From this result we note that the ratio of the depthintegrated mass of sediment in suspension transport to that in bedload transport is given by ( 1P 1)/(1 P ), and thus the fraction b of the total sediment load which travels in the bedload is given by b (1 P )/(1P P ). (5)

In gure 1, values of b calculated from eq. 5 are shown as a function of the ratio w s / u* and relative depth . From these calculations we note that for conditions in which values of the ratio w s / u are * much less than unity, b becomes small and material in transport is predominantly in turbulent suspension. Under such conditions the material in transport is well mixed throughout the ow and so the degree of partitioning between bedload and suspended load is related to the relative depth . For conditions in which w s / u* is much greater than unity, on the other hand, b approaches unity and so the mobile material travels predominantly in the

where the Rouse parameter P w s / ku , w s is the * average settling velocity of the material available for transport and 0.4 is von Karmans constant. The friction velocity u is a characteristic velocity * and is related to the bed shear stress o of a channel ow of water with density by the equation o u2 *. (3)

Equation 2 is a simplied form of a more complete expression given by Middleton and Southard (1984), and details concerning the its derivation and physical signicance can be found there. An analysis similar to that which follows but using the more complicated expression given by Middleton and Southard, which accommodates a vanishing sediment concentration at the free surface of the channel ow, requires numerical calculations yet adds little to the key result. We assume the reference parameters z b and C b to correspond to the thickness

Figure 1. Fraction b of total load travelling as bedload as a function of w s / u and relative depth . Solid lines * indicate calculations from eq. 5. Dashed lines indicate criteria given in eq. 6 for the distinction of ows characterised by predominantly suspended-load, mixed-load or bedload transport.

664

W. B. DADE AND P. F. FRIEND

bedload. Intermediate conditions for which w s / u * 1 correspond to mixed-load transport in which the uxes of suspended and bedload material are approximately equal in magnitude. For the sake of deniteness in this analysis, we adopt the following criteria for the distinction of bedload, mixed-load, and suspended-load transport: w s / u* 0.3 predominantly suspended-load transport 0.3 w s / u* 3 predominantly mixed-load transport w s / u* 3 predominantly bedload transport. (6)

w s (gRd ) 1/2.

(9)

We note that the choice of the inertial drag coefcient approaching unity as the ratio w s d / becomes very large is one of mathematical expedience, which also describes well the settling behavior of prolate- and irregularly-shaped particles. In the course of our analysis we consider the settling behaviors given by eqs. 8 and 9 for the end-member cases of suspended- and bedload transport. An Equation for Sediment Transport. In steady and uniform channel ow, there is an approximate balance between the downslope weight of the ow and the boundary shear stress o (cf. eq. 3). This balance is expressed as u 2 c D u 2 ghs, * (10)

These criteria are indicated in gure 1. From this plot we see that the dominance of suspension transport corresponds to conditions for which less than approximately 1020% of the total sediment load is bedload. The dominance of bedload transport corresponds to conditions for which more than approximately 8090% of the total sediment load is bedload. In applying these criteria to observations from rivers, we adopt an analytical expression for the settling velocity w s of water-borne transport of debris given in terms of the kinematic viscosity of water v, the acceleration due to gravity g, and the average size d and the relative excess density R of individual grains in transport. This expression is given by w s ( / d ){(81 D 3 ) 9}, (7)

where the dimensionless grain size parameter D (gRd 3 / 2 ) 1/3 and R ( s )/ , where s is the density of individual grains submerged in water with density . Equation 7 reects a force balance between the submerged weight of a particle and the drag force associated with its settling motion described in terms of the quadratic drag coefcient 24 / w s d 1. Equation 7 is valid for all grain sizes to within a factor of about 2. We note in particular that for water-borne debris with the density of quartz and a grain size much smaller than about 1 mm, this expression yields values of w s which correspond to Stokes equation w s gRd 2 /18. (8)

where c D is a dimensionless drag coefcient, which is typically in the range 10 3 10 2, and h and s are respectively the depth and slope of the channel. In such a ow there is also a balance between the rate of production of turbulent kinetic energy (TKE) associated with shear in the mean ow and the rate of dissipation of TKE due to viscous forces. In the case of a sediment-transporting ow, energy is also lost owing to the mobilization and transport of debris. This balance can provide a basis for a sediment-transport equation (Bagnold 1966). We propose the following equation that describes this relationship in terms of the mean ow speed u and channel depth h, the critical shear stress for the onset of sediment motion u 2 cr, the ux q s uhC b A * and fall speed w s of suspended sediment, and the ux of bedload material q b u b z b C b in the nearbed transport layer:
2 1 1 u (u 2 * u *cr ) s gRq s w s / u b gRq b. (11) II III I

For sediment with similar density but for which d 1 mm, on the other hand, eq. 7 yields the inertial settling law

Term I in eq. 11 represents the rate of TKE production through the interaction of the mean ow and the shear stress in excess of that required to mobilize sediment making up the channel bed. Term II represents the rate at which energy is consumed in maintaining the suspended load, and term III represents the rate at which energy is consumed in maintaining the bedload. The coefcient s represents a dimensionless efciency parameter that must be determined empirically, and typically takes on a value in the range 10 2 10 1 (Bagnold 1966). The value of s reects the small fraction of TKE production available for the maintenance of the suspended load. The majority of TKE goes into a cascade of turbulent eddies and is ultimately lost

Journal of Geology

CHANNEL SLOPE IN ALLUVIAL RIVERS

665

to viscous dissipation. Because this constraint applies only to the suspended load, the coefcient s is applied only to Term II. The coefcient b, on the other hand, is an empirical friction term that relates the weight of the bedload to the downstreamdirected shear stress of the transporting ow. It is expected to be of the order of unity. The balance indicated in eq. 11 reects the assumption that sediment transport in a natural river is driven by the channel ow and represents work performed by the ow. The dichotomy between suspended- and bedload transport is related to the dynamics that control the respective transport modes. Given unlimited supply of ne material, sediment transport in a suspended-load river is limited by the TKE available throughout the ow. Sediment transport in a bedload-river, on the other hand, is limited solely by the stress conditions in the immediate vicinity of the bed. We will revisit this point in the discussion. Upon substitution of eq. 10 and the dimensionless Shields parameter u 2 * / gRd, eq. 11 can be rearranged to give
1 1 q s P b q b qs ( cr )/ R.

suspended-load transport for which q T q s and B / P. Equation 13 thus approaches the limiting form given by q T qs ( cr )/ PR. (15)

Upon substitution of the denitions of P and , eq. 15 is equivalently given as q T ( s qs / R ) (u / w s ) ( cr )/ . (16)

In conditions of high-stage transport for which cr, eq. 16 reduces to Bagnolds formula for suspended-load transport rate (Bagnold 1966). Substitution of eq. 8 and subsequent rearrangement indicate that for the transport of ne-grained material eq. 16 can be expressed as q T / (18 s / c D )( cr ), (17)

and, under conditions of high-stage transport, equivalently given by q T / (18 s / c D ) 2. (18)

(12)

In eq. 12, q uh is the specic discharge of water, or discharge per unit of width of the channel, P is the Rouse parameter dened in the previous section and the dimensionless coefcient 1 1 c 1/2 D s conveniently combines the parameters , c D, and s. Upon substitution of the denition for total sediment ux q T q s q b and subsequent rearrangement, eq. 12 yields an expression for total transport ux given by q T Bqs ( cr )/ R, where
1 1 B { 1 P b ( 1 P b )} .

This transport equation can also be expressed in terms of the dimensionless transport parameter q T /(gRd 3 ) 1/2 and the grain-size parameter D and given by (18 s / c D ) 2 D 3/2. (19)

(13)

(14)

Equation 13 represents a general equation for total sediment ux and is analogous to the expression derived by Bagnold (1966). From his extensive study of sediment transport we assume the general validity of eq. 13, and now consider its asymptotic behavior for conditions in which the relative depth is large and in which the Rouse parameter P becomes, respectively, very large (corresponding to predominantly bedload transport), very small (corresponding to predominantly suspended transport), and intermediate in value. For example, under conditions for which P is much less than unity, eqs. 13 and 14 correspond to

Equations 18 and 19 are different but equivalent dimensionless forms of a sediment transport equation comparable to expressions found in the sedimentological and engineering literature. Equation 19, for example, is similar to many of the empirical formulae listed by Sleath (1984) for suspended- and total-load transport of ne-grained material. For nominal values of c D 0.005 and s 0.03, the prefactor in parentheses is expected to take on a value of the order 10 2. Under conditions for which P is much greater than unity q T q b, the ratio b approaches unity and thus B b. Upon substitution of the relationship qs uhs (ghs ) 3/2 / gc D 1/2, eq. 13 yields a general expression for the dimensionless bedload transport rate given by
3/2 ( b R / c 1/2 cr 1/2 ). D )(

(20)

For c D 0.005, R 1.6, and b of order unity, the prefactor ( b R / c 1/2 D ) in eq. 20 takes on a value of order 10. This result thus corresponds well to the em-

666

W. B. DADE AND P. F. FRIEND

ller pirical law reported by Meyer-Peter and Mu (1948) for bedload transport and given by 8.5( 0.047) 3/2. (21)

In the case of mixed-load transport, the parameters P, , and B are all near unity so that q T qs (25)

Under conditions of mixed-load transport for which P approaches unity, the parameter B in eq. 1 1 13 approaches the value {1 ln()}/{ b ln()}. For the nominal parameter values c D 0.005, s 0.03 and b 1, this result indicates that B is also of order unity if is large.
Sedimentological Constraints on the Graded Stream. We consider the constraints imposed by

eq. 1 on eqs. 17 and 20 for a graded stream in which neither deposition nor erosion occurs. In the case of high-stage transport of suspended sediment, for example, eq. 17 indicates that for q T 0 (either in time or space), the Shields parameter must approach a constant value s. Noting that the ow Reynolds number Re q / and assuming that the drag coefcient c D is constant, this limiting value is given by s (c D ReC s /18 s ) 1/2 (22)

for the transport of ne-grained material in suspension. Taking again the values c D 0.005 and s 0.03 and the nominal values Re 10 7 and c s 10 3, s is estimated to be of the order of 10. The value of this estimate is not necessarily universal but is related to, among other things, the Reynolds number of a channel ow and the capacity-limited concentration of suspended sediment carried by a stream in approximate grade. The condition s 10 does not reect limitations imposed by the supply of ne material in an alluvial basin, and so probably represents a maximal value. In the case of predominantly bedload transport, the condition q T 0 applied to eq. 20 requires that D / (D / )( cr /3)/( cr ), (23)

if c D and b are constants. If grain size is also insensitive to transport conditions then D / 0, and eq. 23 indicates that the Shields parameter approaches a constant value b given by b cr /3. (24)

Sediment transport does not occur, of course, for values of less than the critical value cr, The limit to b, then, is cr which is typically taken to be in the range 0.030.06 for gravel-sized material.

if, as is expected in this case, cr. Under such conditions there are no explicit constraints on the Shields parameter . We note, however, that in the case of coarse debris w s (gRd ) 1/2 (c.f. eq. 9), so that P 1 1/2. Thus under conditions of mixed-load transport of coarse sand and gravel for which P 1, the Shields parameter is also expected to take on a value m of the order of unity, and to reect transitional behavior between b and s. Summary of the Analysis. We have proposed a general equation for sediment transport analogous to Bagnolds formula for total sediment load. The constraint that sediment ux is steady and non-divergent in an alluvial channel corresponds to a state of dynamic equilibrium in which neither net deposition nor erosion occurs. This constraint for a graded channel requires that the Shields parameter approach a constant value for each of the regimes of predominantly suspended-, mixed-, and bedload transport. The parameter sh / Rd, where s and h are, respectively, channel slope and depth, and R and d are, respectively, relative excess density and average grain size of the individual particles in transport. To the degree to which the material making up the bed of an alluvial river in grade is correlated with the debris in transport through a given reach, this result indicates a direct relationship between the sediment characteristics and the depth-slope product of a channel. The different transport regimes are distinguished by the criteria given in eq. 6. In general, we predict that for predominantly suspended-load transport of ne-grained material for which the ratio w s / u* is very small, the Shields parameter approaches the value s given by eq. 22. The value of s is seen from eq. 22 to be dependent on the Reynolds number of the ow and the equilibrium concentration of sediment in suspended transport. It is expected to be of the order of 10 for most natural rivers. In the case of predominantly bedload transport for which w s / u becomes much greater than unity, * the Shields parameter is predicted to take on a value b on the order of the critical value cr 0.04 for the mobilization of coarse sand and gravel. This relationship, given in eq. 24, reects a relative insensitivity of grain size to changes in transport conditions. Equivalently, it reects a condition of invariant specic stream power qs. Under

Journal of Geology

CHANNEL SLOPE IN ALLUVIAL RIVERS

667

conditions of mixed-load transport for which w s / u* is near unity, the Shields parameter is expected to take on a value m, which is also near unity. Comparison of the Analysis with Data from Alluvial Rivers
Summary of the Data. We compare our analysis with extensive observations of Leopold and Wolman (1957), Schumm (1968), Chitale (1970), and Andrews (1984). Additional data for sand-bed rivers from Church and Rood (1983) are also included. In each of these sources the data appear in concise summary tables. The data comprise over 120 cases of channel slope, depth, width, discharge and grain size of the bed material of alluvial rivers. A wide range of transport conditions for perennial streams is represented. The values of mean annual discharge are in the range 10 1 10 5 m 3 s 1, and the range of Reynolds numbers of the channel ows are in the range 10 4 10 8. Estimates of channel slope range from 10 5 to 10 2, and estimates of the average grain size of bed material range from 10 m to 30 cm. Among the largest rivers considered are the Mississippi and the Indus Rivers each with a mean annual discharge in excess of 10 4 m 3 s 1. Bed materials in each of these rivers are composed of medium sand. Among the smallest ows considered are the well-studied Watts Branch near Rockville, Maryland, with a mean annual discharge of 10 1 m 3 s 1, and many small streams throughout the continental US. In general these smaller streams have beds of gravel and coarse sand. The data represent either single or multiple observations of channel properties. In cases of multiple reports for a single river, each entry usually represents a distinct reach in a relatively large river. Average values are used, however, where multiple entries were recorded for a single reach (as in the data listed by Church and Rood 1983). In general the discharges used in our calculations correspond to mean annual values or, in a few cases, values of bank-full discharge. Estimates of slope correspond to the values measured in the eld unless only map estimates were given. Estimates of the median grain size of channel material reported in both data sets reect a range of techniques with different levels of accuracy and precision. Estimates of bed shear stress are made from eq. 10 with no correction for the drag associated with bedforms. There has been no screening of the data other than to eliminate a few redundant cases between the data sets. Sedimentological Constraints on the Longitudinal Prole of Graded Channels. Figure 2 is known as a

Figure 2. Shields parameter as a function of particle Reynolds number Re and mode of sediment transport * in alluvial rivers. The data, the sources of which are described in the text, are represented by the symbols and have been partitioned into regimes of channel ow characterized by predominantly suspended-load, mixed-load and bedload transport. Solid lines indicate regimes characterized by no sediment motion (NM), ripples and dunes (R/D) and upper-stage plane bed (UP) after Allen (1982).

Shields plot and is a conventional presentation for sediment-transport phenomena (cf. Middleton and Southard 1984). It plots the Shields parameter sh / Rd as a function of the particle Reynolds number Re u d / for natural alluvial rivers. * * The data have been partitioned into classes distinguished by the criteria given by eq. 6, which correspond to rivers dominated by suspended-, mixed-, and bedload transport. Indicated in this plot are the ow regimes that correspond approximately to no sediment motion (NM), undulatory bedforms (i.e., ripples and dunes, R/D) and upper-stage plane beds (UP) in laboratory ows (Allen 1982). The considerable scatter in the data reects the wide-ranging sources of error in the observations associated the different techniques and measures of discharge, channel geometry and grain size employed in the original studies. The data nevertheless yield welldened mean estimates (1 standard error) of the Shields parameter s 12 2, m 1.9 0.2 and b 0.04 0.003 for alluvial rivers dominated respectively by suspended-, mixed-, and bedload transport. This is in good agreement with the predictions from our analysis that s 10, m 1, and b 0.04 for graded alluvial systems. Discussion Given the numerous sources of scatter in the existing data (gure 2 and mentioned above), we con-

668

W. B. DADE AND P. F. FRIEND

sider the good agreement between observation and prediction for alluvial streams remarkable. The implication is that the values of s, m and b that emerge from our analysis and are conrmed by observation represent quasi-universal values. This is intriguing. The description of bedload rivers in terms of a constant value of b that is universally related to cr for coarse debris has been recognized before (Kellerhals 1967; Parker 1979), but we know of no physical reason why the values of m and s, in particular, should be strictly universal for sandy rivers as they appear to be in gure 2. There are several complications, moreover, which we have not considered in our analysis. These include the effects of increasing channel width and total discharge in the downstream direction or, more technically, the constraint imposed on the average bed shear stress in an active channel with stable (noneroding) banks (Parker 1978, 1979). These complications do not appear to undermine the general predictions of our simple analysis, however. Additional implications and applications of our analysis include consideration of the importance of limitations due to ow competence and capacity for uvial sediment transport, consideration of the importance of laboratory analogues for natural, alluvial systems, and application of our analysis to the interpretation of spatial pattern and temporal evolution of ancient and modern rivers. To pursue some of these points we propose that alluvial rivers are in a state of quasi-equilibrium, although we examine this hypothesis as well. Competence vs. Capacity in Graded Rivers. The concepts of competence and capacity are often invoked by engineers and geomorphologists to describe the ability of a river to transport sediment. Competence is a measure of the largest grain size that can be mobilized by a channel ow, while capacity is a measure of the sediment discharge of a stream in the absence of supply limitations. Equations 18 and 22 indicate that the limiting value of s for rivers dominated by suspended sediment transport is determined by the limiting value of sediment discharge in such rivers. The value of s is much greater than the value required for the mobilization of the ne-grained sediment. Competence, therefore, appears to have little relevance, and the graded prole of a ne-grained river is determined rather by the capacity of the channel ow. In contrast, the Shields parameter for streams dominated by bedload transport is near the critical value required for the mobilization of average sediment in the bed. The graded prole of a bedload river is thus closely related to the mobility of the debris making up the channel bed, and thus to the

competence of the overlying ow. This distinction represents a fundamental difference between rivers dominated by suspension transport of ne-grained material and rivers dominated by bedload transport of coarse debris, and justies the partitioning of sediment-transport terms in eq. 11.
Constraints on the Interpretation of Laboratory Analogs. Experiments that involve laboratory-scale

alluvial channels have long been an important component of the study of the evolution and form of river channels (e.g., Leopold and Wolman 1957; Wolman and Brush 1961; Schumm and Kahn 1972). In light of the analysis given here, however, we point out an important limitation of this approach. In gure 3a is a Shields plot with the ume data of Leopold and Wolman (1957). The data for natural alluvial rivers shown in gure 2 are included for comparison. The experimental systems characterized by mixed-mode transport do not achieve the same equilibrium values of the Shields parameter m as do the prototype rivers. In fact the form of all laboratory-scale channels appear to be near critical conditions for sediment transport. The longitudinal proles of such systems are thus interpreted to be competence-limited, as are the natural alluvial systems dominated by bedload transport. This difference between experimental and natural channels is due in part to the relatively small Reynolds numbers of laboratory-scale ows (gure 3b ). Channel ows for which Re is less than about 10 4 do not exhibit fully developed turbulence (Schlichting 1979). Another potentially important factor is the physical limitation to the ratio d / h in small-scale ows. In short, well-developed turbulent suspension of the material in transport is simply not to be expected in the ume studies of Leopold and Wolman (1957). These and similar experiments appear to provide at best a basis for the study of natural rivers dominated by bedload transport and whose equilibrium form is limited by the competence of the channel ow. A useful laboratory analogy for natural, ne-grained rivers dominated by suspended-load transport, on the other hand, seems problematic. It may be that some aspects of channel form are independent of the mode of sediment transport. To our knowledge this has not been established, however, and as we have shown here, it is certainly not the case that the longitudinal prole of a natural, alluvial channel is insensitive to the mode of sediment transport. Constraints on Downstream Fining. Gradual downstream ning of material in gravel-bed rivers is well studied and is understood to reect the interaction of abrasion and size-selective transport of individual clasts (e.g., Paola and Seal 1995). The

Journal of Geology

CHANNEL SLOPE IN ALLUVIAL RIVERS

669

Figure 3. a. Shields parameter as a function of particle Reynolds number Re and mode of sediment transport * in laboratory experiments and alluvial rivers. The data from laboratory experiments were reported by Leopold and Wolman (1957). River data same as in gure 2 for comparison. b. Shields parameter as a function of ow Reynolds number Re and mode of sediment transport. Laboratory and river data as in 3a.

analysis given here, however, provides a complementary explanation. We note that nominal ow depth h can be equivalently expressed as q / u which, upon substitution of eq. 10 and some rearrangement, yields the relationship h (c D q 2 / gs ) 1/3. Thus the nominally constant values of s, m, and b can be rearranged to yield the equivalent relationship for specic stream power qs given by
1/2 ( 3 (gR 3 d 3 ) 1/2, i /cD)

(26)

where the subscript i indicates the appropriate parameter value for systems which are dominated by suspension (i s ), mixed-load (m ), and bedload (b ) transport. Equation 26 suggests that in a graded channel the stream power is proportional to grain size d raised to the power 3/2. The value of the coefcient of proportionality is dependent on the value of i and c D.

Shown in gure 4a are the same data as in gure 2, but now as a plot of stream power as a function of the grain size parameter (gR 3 d 3 / c D ) 1/2 and the ratio w s / u . The solid lines indicate the proportional* ity (gR 3 d 3 / c D ) 1/2 for suspension-, mixed-, and bedload-dominated regimes. In general the longitudinal proles of natural rivers are concave owing to, among other things, some combination of the effects of subsidence and a downstream increase in mass ux (Sinha and Parker 1996). In addition, natural rivers are dissipativeas the energy of a stream is lost to heat in the downstream direction, stream power and the capacity to carry material also diminishes. The relationships in gure 4a suggest that the natural process of debouchment, in which diminishes in the downstream direction, is systematically associated with a downstream decrease in grain size if the channel is in approximate grade. This is seen as well in the single case of downstream ning in the bedload-dominated reach of the Allt Dubhaig of Scotland, as shown in gure 4b. The pattern of downstream ning in this locality is clearly associated with a decrease in specic stream power in such a way that the Shields parameter remains approximately constant about the critical value for gravel. The explanation for downstream ning given above is related to an ultimate cause for the pattern. The proximal mechanisms by which a pattern of downstream ning develops is not readily evident from the analysis but is likely to include a process of selective sorting similar to that considered by Paola and Seal (1995). In any event, we suggest that a downstream decrease in grain size in an individual river will be predictable and gradual as long as the stream remains in the same transport regime. In many rivers, however, downstream changes in channel properties include abrupt transitions in grain size and slope (Yatsu 1955; Sambrook Smith and Ferguson 1995). The nature of this step-like change in channel characteristics has remained an unresolved question regarding natural, alluvial rivers (Parker 1996). Our analysis provides a basis for the interpretation of this phenomenon. Abrupt Downstream Changes in Channel Slope. In gure 5 is a plot of channel slope s as a function of the ratio of grain size to ow depth d / h. The data for natural rivers in gure 2 are distinguished as before regarding channels dominated by suspended-, mixed-, and bedload transport. The different classes of transport, earlier distinguished by different values of the parameter sh / Rd appear as distinct, linear trends in gure 5, where s i Rd / h. We propose that an abrupt transition from gravel

670 Figure 4. a. Specic stream power as a function of the grains size parameter (gR 3 d 3 / c D ) 1/2 and mode of sediment transport. Data as in gure 2. b. Specic stream power and the Shields parameter as functions of grain size in the uppermost 3-km reach of the Allt Dubhaig, Scotland. Data represented by symbols from Ferguson and Ashworth (1991). The solid lines indicate the relationship d 3/2. The dashed line in (b) indicates the critical value of the Shields parameter 0.04.

W. B. DADE AND P. F. FRIEND

Figure 5. Channel slope s as a function of relative grain size d / h and mode of sediment transport. Data as in gure 2. The solid lines indicate the relationship s d / h. U D indicates the character of the abrupt upstreamdownstream transition from gravel to sand at about 3 km on the Allt Dubhaig, Scotland, as reported by Sambrook Smith and Ferguson (1995).

to sand that is accompanied by an abrupt change in the slope in many individual rivers reects a change in the mode of sediment transport from predominantly bedload to incipient suspension during the natural course of downstream debouchment and ning. This view is consistent with the observations reported by Sambrook Smith and Ferguson (1995) for the Allt Dubhaig, indicated in gure 5. A river in approximate grade accommodates a change in transport mode during the course of downstream ning with a change in channel form that corresponds to a jump from b to m. To the degree to which the ux of water in an alluvial channel remains continuous in the downstream direction, the jump from b to m requires a quantum, interdependent adjustment of both slope and the grain size of material making up the bed. Alluvial Response to Perturbation. In each of the topics discussed above, we implicitly assumed that a river is allowed to achieve in time a state of undisturbed quasi-equilibrium. Externally imposed

Journal of Geology

CHANNEL SLOPE IN ALLUVIAL RIVERS

671

changes to river channels, natural or otherwise, often occur, however. By externally imposed change, we refer, for example, to climatically determined change in discharge, tectonic change of slope, and the introduction of anomalously-sized sediment due to local bank failure or the emergence of a new source of debris. An understanding of the effects of such changes presents a difcult challenge. The analysis provides a quantitative basis for the prediction of the response of an alluvial channel to externally imposed disturbances. Channel response is most easily interpreted with reference to gure 4a. Given the availability of all grain sizes, for example, a change of stream power in a graded stream will elicit a complementary change in bed grain-size d to the effect that a new equilibrium prole is achieved from among the quasi-universal states indicated by the data for the respective regimes of suspension-, mixed-, and bedload-dominated regimes. Conversely, an externally imposed change in grain size will elicit appropriate changes in slope and discharge to the effect that a new, equilibrium stream power is achieved. The direction of the changes is not necessarily unique. For instance, a change in discharge is likely to be associated with a change in depth h. This is likely to be associated with complementary changes in s and d if a constant value of i is maintained. Thus an increase in discharge of a bedload stream, say, could be associated with the emplacement of coarser sediments on the channel bed or a reduction in the channel gradient through scour so that ultimately remains constant at b. If the increase in discharge is sufcient to elicit a change in regime from bedload to predominantly mixed-load or suspensionload transport, however, then channel slope and grain size will co-evolve to meet the new m or s constraint. An important limitation to this analysis arises due to the effects of a critical condition for sediment mobility. If material is introduced to a channel for which the existing conditions correspond to a Shields parameter for the new material which is less than the critical value cr for mobility, then that sediment cannot be reworked by the existing mean ow. The introduced debris remains in the channel as a stranded deposit until one or more extreme events redistribute it downstream (cf. Friend 1993). The Response Time of an Alluvial System. The response of a channel to changes in discharge, regional slope or sediment supply is, of course, not necessarily instantaneous. An estimate of the time required for an alluvial system to achieve the dy-

namic equilibrium of a graded prole is derived as follows. The sediment continuity equation states that the rate of change of bed elevation above an arbitrary datum equals the sum of the effects of local subsidence S (x ) and the divergence in sediment ux (e.g., Paola et al. 1992). In mathematical terms this relationship is given by
1 q T / x S (x ), / t C o

(27)

where Co is the volumetric concentration of solids in the stationary bed. Note that the condition specied by eq. 1 corresponds to a state in which the river neither aggrades nor degrades and there is negligible tectonic movement. Substitution of eq. 13 into eq. 27 yields an expression for suspension and mixed load transport for which cr given by
1 / t C 0 (Bqs )/ x S (x ).

(28)

Equation 28 and similar expressions provide a basis for the analysis of the evolution of the sedimentary ll of an alluvial basin in terms of a diffusion equation (e.g., Paola et al. 1992). Upon substituting into eq. 28 the relationship s / x and assuming that specic discharge q is constant in time and space and that there is no subsidence, one obtains the expression / t K 2 / x 2, (29)

where K (Bq / R ). Equation 29 is a form of the linear diffusion equation for which K is the effective diffusion coefcient. In the application of eq. 29 to alluvial systems, boundary conditions are usually given in terms of a source ux condition and a fareld value of 0 at base level. The characteristic time T R required for the evolution of a system described by eq. 29 to the quasi-steady, graded channel with overall length L is given by T R L 2 / K. (30)

The response time of an alluvial system dened in this way is thus seen to be proportional to the square of its linear dimension, and inversely proportional to the specic discharge q and the relative mobility of sediment embodied in the parameter B given by eq. 14. We note that our choice of K in eqs. 29 and 30 reects one possible scheme of linearization of the behavior of an alluvial system. It does not, for example, accommodate downstream increases in specic discharge q. In addition, this pa-

672

W. B. DADE AND P. F. FRIEND

Table 2. Response Times of Selected Rivers River Mississippi Brahmaputra Indus Savannah (SE USA) North Platte, midwest USA) Cheyenne (midwest USA) Q (m 3 s 1) 4.3 10 4 2.5 10 4 7.1 10 3 850 65 3.6 AD (km 2) 3.2 10 6 10 6 10 6 2.7 10 4 8.3 10 4 1.9 10 4 L (km) 6710 2900 3100 530 1000 420 q (m 2 s 1) 28 2.6 7.2 8.0 .4 .13 h (m) 18.3 1.5 4.9 5.2 4.9 0.24 d (cm) .04 .03 .018 .08 .018 .005 ws/u * 0.55 1.65 0.31 1.33 0.26 0.034 TR (ka) 65 85 21 2.4 74 5.5

Note. Values for Q, A D, L, q, h and d from Leopold and Wolman (1957), Chitale (1970) and Leopold (1994). Channel length L of Savannah, North Platte and Cheyenne Rivers estimated from the empirical relationship L 1.16A 0.6 D (where L and A D are expressed in km and km 2, respectively) modied from the expression given by Leopold et al. (1964). T R calculated from eq. 30 with s 0.03.

rameterization corresponds only to rivers dominated by suspended- or mixed-load transport. Finally, in the following exercise we assign values to K based on current properties of apparently nearly graded rivers. Accordingly, calculations based on these properties overestimate the response time T R. The values of T R given below are meant only to be comparative. Listed in table 2 are nominal response times calculated from eq. 30 and estimated properties of selected rivers. These values of T R range from thousands of years to tens-of-thousands of years. Note that these times correspond to the approximate interval required for the evolution of a channel over its entire length given present hydrological and sedimentological conditions, and not over a reach of local extent. These calculations indicate, even given our caveat regarding the likely overestimation of T R, that some of the largest alluvial systems are only now emerging from the effects of the last glaciation that ended approximately 10 ka ago. This result leads us to suggest that some of the scatter in the data shown in gure 25 reects the fact that many individual rivers are only now approaching the dynamic equilibrium associated with Holocene conditions. This result may also explain in part why specic sediment yields from glaciated terrains are in general considerably greater than from unglaciated basins of similar area and relatively low stream discharge (Church et al. 1989). Put simply, glaciated basins of low stream discharge are still adjusting to the hydrological conditions only recently imposed by the Holocene climate. Another ramication of the magnitude of these calculated response times is that the full extent of the adjustment of alluvial rivers to human-induced environmental change will take considerable time to observe. The predictive relationships given here for the ultimate equilibrium forms expected of perturbed rivers may help in addressing this problem. Mass Flux in Ancient Rivers. As a nal topic of

discussion we address the application of our analysis to the interpretation of ancient alluvial deposits. To do so we assume that from exposure(s) a eld geologist can estimate for a paleochannel a representative grain size d of bed material, channel width w and depth h. In modern alluvial rivers there is a clear relationship between grain size and the ratio w S / u*, as is indicated by the data shown in gure 6. Using this empirical relationship one can use grain size alone to ascertain whether the ancient system is likely to represent a channel dominated by suspension-, mixed-, or bedload transport. Having made this distinction, one can estimate paleoslope from gure 5 and stream power from gure 4a. With estimates of s, , and w one then can calculate total discharge Q, and calculations for sediment transport q T follow from eqs. 18 or 25. We give two specic, hypothetical examples of this analysis. There are important caveats, however, attached to the application of our analysis of modern, graded streams to the ancient. The very existence of the deposits indicates that the ancient

Figure 6. The ratio w s / u as a function of grain size d * in alluvial rivers. Data sources are summarized in the text. Solid line indicates an approximate relationship between ws / u and d. *

Journal of Geology

CHANNEL SLOPE IN ALLUVIAL RIVERS

673

channel ow was alluviating and thus not in equilibrium per eq. 1. The estimates of total discharge and sediment yield given below correspond rather to nominal values for the analogous channel in grade. In addition, the estimates of sediment ux given here do not include the wash load. Case A: d 0.08 cm; h 5 m; w 100 m From gure 6 we estimate that w s / u*, was of the order of unity, which implies that the ancient river was dominated by mixed-load transport. We estimate from gure 4a that stream power 2 10 3 m 2 s 1, and from gure 5 that paleoslope was 3 10 4. Together with the estimate for channel width w, these results imply that specic discharge q 7 m 2 s 1 and total discharge Q 700 m 3 s 1. From eq. 25 we estimate that the sediment ux q r , and so total sediment yield was approximately 0.2 m 3 s 1, or equivalently about 17 ton a 1. The Savannah River (SE USA) is the modern setting on which this example is based. Case B: d 5 cm; h 0.5 m; w 50 mFrom gure 6 we estimate that w s / u* was of the order of 10, which implies that the ancient river was dominated by bedload transport. We estimate from gure 4a that stream power 7 10 4 m 2 s 1, and from gure 5 that paleoslope was 2 10 3. Together with the estimate for channel width w, these results imply that specic discharge q 0.4 m 2 s 1 and total discharge Q 20 m 3 s 1. During mean ow conditions there would have been only small amounts of sediment transport through this bedload-dominated system. The Wind River (Wyoming, USA) is the modern alluvial environment on which this example is based. Summary and Conclusions We have established quantitative relationships between the calibre of the bed sediment and the longitudinal prole of a stream channel in approximate grade. We rst established criteria to distinguish rivers characterized by suspended-sediment transport, mixed-load transport, and bedload transport. The ratio of the representative settling velocity w s of the material making up a channel bed and the friction velocity u of the channel ow is the key * parameter in this approach. In eqs. 25 we examined the quantitative relationship between w s / u* and the vertical distribution of the debris in transport in a stream ow. The calculated fraction of total load that travels as bedload, shown in gure 1 as a function of the ratio w s / u , exhibits a step-like * decrease for ow conditions that correspond to w s / u* 1. This is the basis of our exact, although arbitrary, criteria given in eq. 6 for the distinction be-

tween suspended-load, mixed-load, and bedload rivers. We then proposed a general expression for sediment transport analogous to that derived by Bagnold (1966). The basis for this expression is an energetic balance for the total sediment ux, partitioned into the suspended load and bedload, and the incorporation of semi-empirical efciency factors. It yields expressions that correspond well to existing formulae for suspended-load and bedload transport for the limits in which the value of w s / u is much less than unity and much greater * than unity, respectively. In each of the limiting regimes of suspendedload, mixed-load, and bedload transport, our analysis indicates that the condition that sediment ux be continuous in a graded channel requires that a distinctive value of the Shields parameter be maintained in that channel. When data from over 120 published cases were plotted on a Shields plot (gure 2), the constant values of the Shields parameter for the different regimes are clear, in spite of considerable scatter. This suggests that the respective values for the Shields parameter for each of the three transport modes may be universal. This important result provides a powerful basis for the interpretation of river sedimentation patterns. It deserves closer scrutiny. Several points of discussion emerged from this analysis. 1) In an alluvial river characterized by suspended-sediment transport, sediment ux and the longitudinal prole of the channel reects constraint imposed by the capacity of the ow to transport ne material, whereas in a river dominated by bedload transport, sediment ux and channel prole reect constraints imposed by the competence of the ow. 2) Laboratory-scale, alluvial systems exhibit longitudinal proles characteristic of bedload channels. 3) A gradual pattern of ning in channel deposits can be related to the gradual reduction of specic stream power in the downstream direction. 4) An abrupt downstream change in grain size and channel prole can be related to a change in the mode of sediment transport. 5) The change in channel prole associated with an externally imposed change in discharge (e.g., climatic), slope (tectonic), or grain size (local supply) can be interpreted in terms of the complementary changes required of the equilibrium state within the existing transport mode, or non-unique changes associated with a change in transport mode. 6) The quantitative description of the evolution of an alluvial system can be reduced to a form of

674

W. B. DADE AND P. F. FRIEND

the diffusion equation. Consequently the system exhibits a characteristic response time to approach the equilibrium or graded state considered here. Nominal values of this response time are in the range of thousands to tens-of-thousands of years. Some large alluvial basins or basins with relatively low discharge may only now be approaching a form that reects Holocene conditions. 7) If the cross-sectional geometry and the grain size of the ll of an ancient alluvial channel can be measured from its deposits, nominal estimates of

slope, total discharge, and sediment yield of the system can be made.
ACKNOWLEDGMENTS

C. Paola, R. Slingerland, and P. Talling offered helpful comments which served to clarify our thinking and to improve the text. M. Church and T. Hoey generously provided data. W. B. Dade gratefully acknowledges support from the Leverhulme Foundation and E. F. Dade.

REFERENCES CITED

Allen, J. R. L., 1982, Sedimentary Structures (Developments in Sedimentology 30): Amsterdam, Elsevier, 1256 p. Andrews, E. D. 1984, Bed-material entrainment and hydraulic geometry of gravel-bed rivers in Colorado: Geol. Soc. America Bull., v. 95, p. 371378. Bagnold, R. A., 1966, An approach to the sediment transport problem from general physics: U.S. Geol. Survey Prof. Paper 422-I, 37 p. Baker, V. R., and Pyne, S., 1978, G. K. Gilbert and modern geomorphology: Am. Jour. Sci., v. 278, p. 97123. Chitale, S. V., 1970, River channel patterns: Jour. Hydraul. Div. ASCE, v. 96 (HY1), p. 201221. Chorley, R. J., and Beckinsale, R. P., 1980, G. K. Gilberts geomorphology, in Yochelson, E. L., ed., The scientic ideas of G. K. Gilbert: Geol. Soc. America Spec. Paper 183, p. 129142. Church, M.; Kellerhals, R.; and Day, T. J., 1989, Regional clastic yield in British Columbia: Can. Jour. Earth Sci., v. 26, p. 3145. , and Rood, K., 1983, Catalogue of Alluvial River Channel Regime Data: Vancouver: Dept. Geography, University of British Columbia, 99 p. Ferguson, R., and Ashworth, P., 1991, Slope-induced changes in channel character along a gravel-bed stream: The Allt Dubhaig, Scotland: Earth Surf. Landforms, v. 16, p. 6582. Friend, P. F., 1993, Control of river morphology by the grain-size of sediment supplied: Sed. Geol., v. 85, p. 171177. Hack, J. T., 1960, Interpretation of erosion topography in humid temperate regions: Am. Jour. Sci., v. 258A, p. 8097. Kellerhals, R., 1967, Stable channels with gravel-paved beds: Am Soc. Civil Eng. Jour. Waterways and Harbors Div., v. 93, p. 6384. Lacey, G., 1930, Stable channels in alluvium: Proc. ICE, v. 229, p. 259270. Leopold, L. B., 1980, Techniques and Interpretation; The sediment studies of G. K. Gilbert, in Yochelson, E. L., ed., The scientic ideas of G. K. Gilbert: Geol. Soc. America Spec. Paper 183, p. 125128. , 1994, A View of the River: London, Harvard University Press, 298 p.

, and Wolman, M. G., 1957, River channel patterns: Braided, meandering, and straight: U.S. Geol. Survey Prof. Paper 282-B, 85 p. ; ; and Miller, J., 1964, Fluvial Processes in Geomorphology: New York, Dover, 522 p. Mackin, J. H., 1948, Concept of the graded river: Geol. Soc. America Bull., v. 59, p. 463512. Meyer-Peter, E., and Mu ller, R., 1948, Formula for bedload transport: Proc. Int. Assoc. Hydraul. Struct. Res., Stockholm, p. 3964. Middleton, G. V., and Southard, J. B., 1984, Mechanics of Sediment Movement. Short Course No. 3: Tulsa, OK, Soc. Econ. Paleont. and Mineral., 401 p. Paola, C.; Heller, P. L.; and Angevine, C. L., 1992, The large-scale dynamics of grain-size variation in alluvial basins, 1: Theory: Basin Res., v. 4, p. 7390. , and Mohrig, D., 1996, Paleohydraulics revisited: Paleoslope estimation in coarse-grained braided rivers: Basin Res., v. 8, p. 243254. Paola, C., and Seal, R., 1995, Grain size patchiness as a cause of selective deposition and downstream ning: Water Resources Res., v. 31, p. 13951407. Parker, G., 1978, Self-formed straight rivers with equilibrium banks and mobile bed. Part 2. The gravel river: Jour. Fluid Mech., v. 89, p. 127146. , 1979, Hydraulic geometry of active gravel rivers: Jour. Hydraul. Div. ASCE, v. 105, p. 11851201. , 1996, Some speculations on the relation between channel morphology and channel-scale ow structures, in Ashworth, P. J.; Bennett, S. J.; et al. eds., Coherent Flow Structures in Open Channels: New York, Wiley, p. 423458. Ribberink, J. S., and van der Sande, J. T. M., 1985, Aggradation in rivers due to overloadinganalytical approaches: Jour. Hydraul. Res., v. 23, p. 273283. Sambrook Smith, G. H., and Ferguson, R. I., 1995, The gravel-sand transition along river channels: Jour. Sed. Res., v. A65, p. 423430. Schlichting, H., 1979, Boundary-Layer Theory: New York, McGraw-Hill, 817 p. Schumm, S. A., 1968, River adjustment to altered hydrologic regimenMurrumbidgee River and paleochannels, Australia: U.S. Geol. Survey Prof. Paper, 598, 65 p.

Journal of Geology

CHANNEL SLOPE IN ALLUVIAL RIVERS

675

, and Kahn, H. R., 1972, Experimental study of channel patterns: Geol. Soc. America Bull., v. 83, p. 17551770. Sinha, S. K., and Parker, G., 1996, Causes of concavity in longitudinal proles of rivers; Water Resources Res., v. 32, p. 14171428. Sleath, J. F. A., 1984, Sea Bed Mechanics: New York, Wiley, 450 p.

Wolman, M. G., and Brush, L. M., 1961, Factors controlling the size and shape of stream channels in coarse noncohesive sands: U.S. Geol. Survey Prof. Paper 282G, 27 p. Yatsu, E., 1955, On the longitudinal prole of the graded river: Trans. Am. Geophys. Union, v. 36, p. 655663.