Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

WATER RESOURCES RESEARCH, VOL. 45, W09421, doi:10.

1029/2008WR007604, 2009

Exner equation: A continuum approximation of a discrete


granular system
S. E. Coleman1 and V. I. Nikora2
Received 24 November 2008; revised 14 May 2009; accepted 29 May 2009; published 29 September 2009.
[1] In contrast to using a standard control volume approach, general statements of
sediment mass balance are derived herein from spatial averaging of the sub-particle-scale
differential equation of solid mass conservation. The general form of the Exner equation
for sediment continuity that is obtained enables analyses in terms of size fractions and
also in terms of individual or successive layers, where layer interfaces (e.g., for the bed
surface and for bed and suspended loads) are defined on the basis of isosurfaces of
sediment concentration (volume fraction) or other sediment properties (e.g., densities or
transport rates) within regions of constant concentration. The presented expressions
highlight the averaged nature of variables and also the effects of the scales of
consideration on definition and interpretation of the macroscopic (mixture-scale) sediment
and layer properties (e.g., averaged densities, volume concentrations or fractions,
velocities, transport modes and rates, interfaces, and bed layers). For appropriate
simplifications, the general form of the Exner equation is shown to reduce to give more
specific conventional expressions, revealing the assumptions implicit in these equations.
Citation: Coleman, S. E., and V. I. Nikora (2009), Exner equation: A continuum approximation of a discrete granular system, Water
Resour. Res., 45, W09421, doi:10.1029/2008WR007604.

1. Introduction Jordaan, 2000; Das et al., 2004, Blom and Parker, 2004;
[2] Exner [1925, 1931] proposed that the rate of change Cui and Parker, 2005; Jerolmack and Mohrig, 2005; Zhang
of the local elevation h of the sediment bed of a river can be et al., 2006; Kubatko and Westerink, 2007; Sumer, 2007], to
related to spatial variation of the local depth-averaged analytical models for boundary instability [e.g., Kennedy,
overhead streamwise water velocity U as 1963; Smith, 1970; Richards, 1980; Engelund and Fredsøe,
1982; McLean, 1990; Coleman and Fenton, 2000; Colombini,
2004], to estimates of sediment transport based on the
@h @U movement of sediment waves [e.g., Simons et al., 1965;
þa ¼ 0 or ð1Þ Engel and Lau, 1980; Graf, 1984; Van den Berg, 1987; Ten
@t @x
Brinke et al., 1999; Hoekstra et al., 2004].
[4] Derivations to date of various forms of the Exner
@h @qs equation have implicitly assumed that the sediment-water
þb ¼0 ð2Þ mixture is a continuum, with analyses based on the mass
@t @x
balance for a control volume VCV that includes the bed
where t is time, x is distance in the bed-parallel flow surface (e.g., Figure 1a, where f is an average bed porosity,
direction, a and b are coefficients, a downstream increase in h is the average bed surface level for VCV, and z = h is the
U induces erosion, and the volume rate qs of sediment water surface). These derivations have varying underlying
transport per unit width is taken to increase with fluid intuitive assumptions in regard to scales of consideration
velocity. (streamwise, vertical, and lateral), and respective definitions
[3] The second of these equations is conventionally of bed and basement surfaces, modes and directions of
referred to as the Exner equation and has been presented sediment transport, bed layers and their interactions, particle
and used in various forms since the thirties as a foundation for concentrations, and porosity [e.g., Smith, 1970; Yalin, 1992;
analyses of Earth surface morphodynamics [e.g., Raudkivi, Parker et al., 2000; Blom and Parker, 2004; Paola and
1976; Parker et al., 2000; Dietrich et al., 2003; Paola and Voller, 2005; Parker, 2008].
Voller, 2005; Kubatko and Westerink, 2007; Parker, 2008]. [5] In contrast to intuitively formulating conservation
Example application areas range from numerical modeling equations for sediment-water mixtures as continua, deriving
of erodible boundary development [e.g., Batuca and these equations from spatial averaging of the equivalent
conservation equations for local (point) phase properties can
potentially reveal unexpected terms in the mixture-scale
1
Department of Civil and Environmental Engineering, University of equations. Form-induced (or dispersive) stresses and explicit
Auckland, Auckland, New Zealand. expressions for form and surface friction drag as revealed by
2
School of Engineering, University of Aberdeen, Aberdeen, UK. time and space averaging of the Navier-Stokes equations are
examples of such additional terms [e.g., Nikora et al., 2007].
Copyright 2009 by the American Geophysical Union.
0043-1397/09/2008WR007604
Insight into the natures of intuitively formulated terms at

W09421 1 of 8
W09421 COLEMAN AND NIKORA: EXNER EQUATION W09421

Figure 1. Subaqueous sediments: (a) sectional view of sediment bed and flow with conventional control
volume VCV and (b) plan view of thin bed-parallel averaging volume Vo, with the shown differential
equation of solid mass conservation applying at the subparticle scale.

mixture scales can also be gained from spatial averaging of where rs is the sediment-particle solid density, and usi =
the governing equations for local (point) phase properties; (us, vs, ws) is the instantaneous local velocity in direction
e.g., the Coleman and Nikora [2008] discussion of the xi = (x, y, z) (Figure 1). Spatial averaging of this equation
Shields entrainment function and its dependence on scales across a fluid-sediment mixture (e.g., Figure 1b) requires
of consideration and flow and sediment properties. use of two averaging theorems [e.g., Whitaker, 1999]
[6] Recognizing the potential benefits given by spatial   ZZ
averaging of local (point) descriptions, a general form of the @qs 1 @fs hqs i 1
¼  qs uIi ni dS and ð4Þ
macroscopic (mixture-scale) Exner equation of sediment @t fs @t Vs
continuity is explicitly derived in Section 2 from spatial SI

averaging of the sub-particle-scale differential equation for


sediment mass conservation. As shown in Sections 2 and 3,   ZZ
this approach can clarify the details of the Exner equation @qs 1 @fs hqs i 1
¼ þ qs ni dS ð5Þ
and its application, including definitions of terms, and the @xi fs @xi Vs
SI
significance of the scales of consideration. Although mass
conservation is discussed herein in terms of subaqueous RRR
where hqsi = (1/Vs) qs dV is the intrinsic spatial average
sediments and surfaces, e.g., gravel or sand beds, bedrock Vs
streams, and inundated floodplains and alluvial fans, the of qs ; qs is any variable (e.g., scalar or vector component)
analyses presented are also applicable to aeolian surfaces defined in the sediment phase but not at points occupied by
such as desert dunes and soil-mantled hillslopes [e.g., fluid (where it can also be assumed to be zero); fs = Vs/Vo =
Dietrich et al., 2003]. 1  f is the volume fraction (concentration) of sediment
solids in Vo, where the volume of the total averaging domain
2. Spatial-Averaging Approach for Sediment Vo can be varied for the analysis of different scales; f is the
Mass Conservation volume fraction of fluid (porosity); SI is extent of water-
sediment interface within the thin (in the bed-normal
[7] The sediment-water mixture being analyzed consists of direction) averaging volume Vo; uIi is the instantaneous
two distinct and interacting phases separated by solid-fluid local (point) velocity of the solid-fluid interface in direction
interfaces (Figure 1b). Spatially averaged equations for xi; and ni is the ith component of the unit vector normal to
instantaneous sediment mass conservation are derived below the element dS of the interfacial surface and directed
along with double-averaged (in time and space) equivalents. outward from the sediment phase into the fluid. Averaging
2.1. Spatially Averaged Mass Conservation (3) across the thin-slab volume Vo of Figure 1b and using
[8] As for any continuous substance, the local (point) the theorems of (4) and (5) gives (for usi = uIi on the
sub-particle-scale differential equation for conservation of interface SI)
sediment mass (Figure 1b) is given by
@fs hrs i @fs hrs usi i
þ ¼ 0 or ð6Þ
@rs @ ðrs usi Þ @t @xi
þ ¼0 ð3Þ
@t @xi

2 of 8
W09421 COLEMAN AND NIKORA: EXNER EQUATION W09421

only; fsm = Vm/Vo and Vm is the part of volume Vo where ft =


@fs @fs husi i Ts/To 6¼ 0 within To (also as in F. Ballio, personal
þ ¼0 ð7Þ
@t @xi communication, 2009); and the theorems of (8) and (9)
for constant sediment solid density rs. are written for fs not varying with time or changing very
[9] Equations (6) and (7) are macroscopic (mixture-scale) slowly (i.e., at a time scale much larger than To). Reason-
statements of sediment continuity that indicate the rate of ably assuming that ft does not (spatially) correlate with the
change of mass (or solid volume Vs for (7)) within Vo to be time-averaged parameters, i.e., hft qs i = hftihqs i [Nikora et
equal to the sum of across-boundary fluxes. Both (6) and (7) al., 2007], (10) can be written as
apply at any location within a two-phase mixture, where
equations similar or equivalent to (6) and (7) are used in @fst hrs i @fst hrs usi i
þ ¼0 ð11Þ
multiphase flow studies [e.g., Smith, 1970; Drew, 1983; @t @xi
Zhang and Prosperetti, 1997; Ancey, 2007; Dufek and
Bergantz, 2007]. The averaged natures of fs, hrsi, husii,
and hrsusii are apparent in (6) and (7), where values of these representing an analog of (6) and (7) for double-averaged
variables can all vary with the size of the spatial-averaging variables, with the mean volume concentration of sediments
volume Vo. Equation (6) includes the effect of correlation fst = fsmhfti in lieu of fs. Note that (11) includes the
between spatial fluctuations in sediment velocity (f usi = usi  potential effects of correlations between spatial (e.g., uf si )

husii) and solid density res = rs  hrsi, where in general, res can and temporal (e.g., u0si) fluctuations in sediment velocity and
be nonzero for a spatially varying sediment density. solid density for sediments with varying rs, where usi =
usi + u0si = husi i + uf 0 0
es +
si + u si, rs = rs + r s = hrs i + r
2.2. Double-Averaged (in Time and Space) Mass 0
rs, hhrs ihusi ii = hrs ihusi i for well-behaved flow variables,
Conservation
and thus
[10] Double-averaged (in time and space) versions of (6)
and (7) may be required where rapid fluctuations occur in
the variables, or for consistency with equations adopted for @fst hrs i @fst hrs usi i @fst hrs i @fst hrs ihusi i
associated fluid flows. Application of the double-averaging þ ¼ þ
@t @xi @t @xi
theorems (adapted from Nikora et al. [2007])  
@fst hres uf
si i @fst r0s u0si
þ þ ¼0 ð12Þ
@xi @xi
* + D E
ZZ s
@qs 1 @fsm ft qs 1 1
ft ¼  qs uIi ni dS 2.3. Continuum Interfaces and Vertically Integrated
@t fsm @t fsm Vo
SI Mass Conservation
ð8Þ [11] The Exner equation can be viewed as a vertically
integrated equation of mass conservation. Desired integra-
and D E tion limits can be defined using varying isosurfaces z = hn
* + ZZ
@qs 1 @fsm ft qs 1 1
s (x, y, t) for sediment properties, on which
ft ¼ þ qs ni dS ð9Þ
@xi fsm @xi fsm Vo
SI
Gn ðx; y; z; t Þ ¼ g  g n ¼ 0 ð13Þ

to average (3) in time and space gives (for usi = uIi on the where g n is the value of the variable g that is used to
interface SI) characterize the nth surface. As will be seen, the following
equations explicitly require sediment concentration to be
* D E constant on each surface. Other sediment properties, e.g.,
s +
@rs @ ðrs usi Þ @fsm hft rs i @fsm ft ðrs usi Þ density hrs i or double-averaged mass transport rate per unit
þ ¼ þ ¼0 normal area Fsi = fsthrs usi i, can be used to define g for
@t @xi @t @xi
s surfaces within regions of constant concentration. A
ð10Þ condition of fst ! 0 can be adopted to define a clearwater
flow bed surface, for example. Alternatively, a reference
value of fst or Fsi can be used to distinguish between bed
where an overbar indicates time averaging; ‘‘s’’ together and suspended sediment loads, or Fsi tending to zero can be
with an averaging operator indicates ‘‘superficial’’ aver- used to define the base level of sediment in motion.
aging in time or space over the total averaging domains of Importantly, such isosurfaces are not necessarily tied to
quantities To or Vo (rather than the phase-specific alternatives physical objects, but can move relative to the sediment
Ts or Vs used in the ‘‘intrinsic’’ averages of (4) and (5), as by particles and are thereby ‘‘permeable.’’ Figure 2 illustrates
Nikora et al. [2007]); qs = ft q; hqsis = hft qis = fsmhft qi; this principle for a sediment bed surface defined by fst ! 0
ft = Ts/To = the time analog of fs = Vs/Vo; To is total at times t1 and t2.
averaging time interval including periods when the spatial [12] When averaging over temporal fluctuations in sedi-
points are intermittently occupied by fluid and sediment ment properties, e.g., because of individual particle move-
phases (e.g., by moving bed particles); Ts is averaging time ments that are much smaller than mean transport scales, the
interval equal to the sum of time periods when a spatial starting point for analyses is the double-averaged descrip-
point under consideration is occupied by the sediment phase tion of (11), where volume concentrations fst are assumed

3 of 8
W09421 COLEMAN AND NIKORA: EXNER EQUATION W09421

Figure 2. Example movement relative to individual sediment particles of a sediment bed surface
defined by fst ! 0.

to vary at time scales much larger than the temporal surface in direction xi; hn is a double-averaged level (i.e., a
averaging period To. Alternatively, the equations for spa- moving average); the material derivative DFn/Dt = @Fn/@t +
tially averaged instantaneous variables should be used with uni @Fn/@xi = 0 for the surface z = hn (x, y, t) defined by Fn =
fst replaced by fs, as indicated by comparison of (11) and hn  z = 0, with (8) and (9) giving
(6) above. Integrating (11) between respective lower and
upper interfaces of z = h1 and z = h2 (Figure 3a) gives * + "    #
s
Z @Fn @Fn @ Fn @ Fn
@ h2
@gsx @gsy þ uni ¼ fst þ huni i
ðfst hrs iÞdz þ þ @t @xi @t @xi
s
@t h1 @x @y 
    @fst @f
  @ F2 þ Fn þ huni i st ¼ 0 or ð15Þ
 fst2 hrs usi i2  hrs i2 hu2i i @t @xi
@x
 i
  @ F1    
þ fst1 hrs usi i1  hrs i1 hu1i i ¼0 ð14Þ @ Fn @ Fn
@xi þ huni i ¼0 ð16Þ
@t @xi
where the first term is the time rate of change of sediment
massRin a layer between h1 and h2 (per unit area); gsx,y (x, y, Note that fst = fshfti remains unchanged on the moving
h
t) = h 2 (fsthrs usx;y i) dz is the double-averaged solid mass interface (refer to the discussion above regarding interface
1
flux in a layer between h1 and h2 (per unit width) in the definition); uni are uncorrelated with gradients in Fn;
respective directions x and y (including potential fluctuation @hFn i/@t = @hn/@t; and @hFn i/@xi = (@hn/@x, @hn/@ y , 1).
si i and hrs usi i as highlighted in
correlation components hres uf 0 0
In deriving (14), use is made of the Leibniz rule for integrals
(12)); huni i is the double-averaged velocity of the nth

Figure 3. Schematic vertically integrated averaging volumes: (a) for general interfaces z = hn (where
DGsx is the change across Vo in solid mass flux in direction x, GsF2 is the solid mass flux normal to F2,
and Axy is the plan area of the averaging volume Vo); and (b) extending from the water surface to a level
z = h1 below any scour potential within the bed (with the bed surface at z = hbs).
4 of 8
W09421 COLEMAN AND NIKORA: EXNER EQUATION W09421

Figure 4. Schematic variations of sediment concentration fst for patch-scale and dune-scale averaging
volumes Vo. Also shown are potential definitions of the bed surface z = hbs based on these distributions.

with limits of integration (hn) depending on the differential implicit in these equations. In the following, some useful
variable (xi or t); e.g., for t it is written as (similarly for xi) definitions and applications of this equation are discussed,
with (14) briefly compared with alternative formulations
Z h2 Z h2
@f ð x; y; z; t Þ @ @h2 @h based on control volume analyses.
dz ¼ f ð x; y; z; t Þdz  f2 þ f1 1
h1 @t @t h1 @t @t 3.1. Transport and Standard Forms of the Exner
ð17Þ Equation
[14] The derivation leading to (14) explicitly shows that
where fn is the value of the variable f on the Rsurface z = hn; the double-averaged sediment mass transport rate (per unit
h
f, hn and their derivatives are continuous; and h 2 (@f/@z)dz = area) is given by the product of volume concentration, solid
1
f2  f1 by definition. The last two terms of (14) are solid density, and sediment velocity (relative to the boundary).
mass fluxes normal to the upper and lower interfaces F2 and For constant rs, for example, the solid volume flux in a
F1 (per unit plan area) that arise from relative movements layer between h1 and h2 in direction x (per unit width) is
between the ‘‘permeable’’ interfaces and sediment given by the product
R h of volume concentration and velocity
particles, where the averaged normal solid mass flux as qsx = gsx/rs = h 2 (fsthusx i)dz.
1
per unit area of Fn is given by fstn(hrs usi in  hrs in huni i)nni; [15] The Exner equation is conventionally applied from
nni = (@hFn i/@xi)/k@hFn i/@xik is the averaged unit (down- the water surface (z = h2 = hw) to a level z = h1 below any
ward directed) vector normal to Fn; and the averaged vector potential scour or movement within the bed (husi i1 = hu1i i =
magnitude k@hFn i/@xik defines the ratio of the area of Fn to 0), with the internal bed surface (z = hbs, e.g., of a constant
its area in plan. For the interested reader, a more detailed value of fst) delineating between transport as bed (qsb) and
version of the derivation of (14) from (11) is given in an suspended (qss) sediment (Figure 3b). For constant rs and
electronic supplement to this paper (Figures S1 and S2).1 no sediment flux through the water surface (husi i2 = hu2i i),
(14) becomes
3. Exner Equation Z h2
@ @qsx @qsy
[13] Equation (14) is a general form of the Exner equation ðfst Þdz þ þ ¼ 0 or ð18Þ
@t h1 @x @y
that states that the rate of buildup (per unit plan area) of
sediment mass between the surfaces h1 and h2 is equal to the
sum of the solid mass fluxes across the boundaries; that is, " Z h2
mass is conserved. Equation (14) is derived on the basis of @hbs 1 @
¼ ðfst Þdz
integration of spatially averaged conservation equations for @t fb @t hbs
local (point) phase properties, and it includes the effects of #
@qsbx @qsby @qssx @qssy
correlations between fluctuations in sediment properties þ þ þ þ ð19Þ
within the averaging volume Vo. As shown below, (14) @x @y @x @y
can be readily reduced to give more specific conventional Rh
versions of the Exner equation, revealing the assumptions where @t@ h 2 (fst) dz is the rate of change (per unit plan area)
bs

1
of suspended sediment volume, sediment concentration in
Auxiliary materials are available in the HTML. doi:10.1029/ the bed is taken to be constant (fb) and equal to that in the
2008WR007604.

5 of 8
W09421 COLEMAN AND NIKORA: EXNER EQUATION W09421
Rh
bed load layer, qsbx = h bs (fbhusx i)dz, and @h1/@t = 0. hrs usi i = hrs ihusi i excluding any potential effects of pos-
1
Equation (19) represents the standard form of the Exner sible correlations in fluctuations within Vo, unlike the
equation. Various forms of this equation are quoted present analyses. As for the Paola and Voller [2005] analy-
depending upon additional assumptions [e.g., Smith, 1970; ses, (14) can be applied over successive bed layers to incor-
Yalin, 1992; Kubatko and Westerink, 2007; Lanzoni, 2008; porate the respective effects of intralayer changes (e.g.,
Garcı́a, 2008; Parker, 2008], with (2) recovered for no compaction), interfacial movements (e.g., basement uplift
suspended sediment and one-dimensional (1D) bed load or subsidence, bed surface erosion and deposition, etc.), and
transport in direction x. horizontal and interlayer fluxes (including soil creep, bed,
suspended and wash loads, and transfers between these). The
3.2. Layers and Layer Interfaces
sources and sinks of sediment mass production and destruc-
[16] In terms of the structure of a sediment bed, ‘‘active- tion (e.g., precipitation or dissolution due to geochemical
layer’’ approaches to mass conservation assume a well- processes) specified in (PV.3) and (PV.13) are not apparent in
mixed active layer of sediments that exchanges mass with (14), but can be readily incorporated in the right-hand-side of
the moving bed load. Beneath this layer is assumed to be an this equation. Equation (14) beneficially highlights the aver-
inactive structured substrate that only influences the sedi- aged (in space and time) nature of variables, and also includes
ment balance as the bed aggrades or degrades. Many active- the effects of variations in properties (e.g., fst, hrs i, hrs usi i,
layer-based formulations of the Exner equation have been and size fractions as discussed below) within the integrated
developed with varying additional assumptions [e.g., Blom control volume.
and Parker, 2004].
[17] Along with such assumed layering, the transport of 3.3. Size Fractions
sediment is commonly apportioned into bed and suspended [20] By considering the properties of a kth size fraction of
loads. By their natures, these modes of transport are defined a sediment mixture (e.g., double-averaged fkst, hrks i and
by their proximity to, or frequency of contact with, the bed hrks uksi i), (6), (11) and (14) can be used to describe the
surface. transport and balance of this fraction [e.g., Parker et al.,
[18] Equation (14) can be readily applied to analyze 2000], where fkst = (Vsk/Vs)fst can be adopted for the volume
individual or successive layers of sediment. For the bed concentration of the kth size fraction. The effects of abra-
surface (of a threshold value of fst) given by z = h2 = hbs sion [e.g., Parker, 2008] can also be incorporated as source/
and no sediment flux through a stationary lower interface of sink terms in the equation for a given size fraction. The
z = h1 within the bed, the bed load balance transport and balance of the total sediment mixture can then
simply be determined by summing the fractional balances
 
and transport.
Z hbs
@ @qsbx @qsby   @ F2 3.4. Scales of Consideration and Sediment and Layer
ðfst Þdz þ þ  fst2 husi i2 hu2i i ¼0
@t h1 @x @y @xi Properties
ð20Þ [21] The time-averaging period To and the size of the
averaging volume (e.g., VCV of Figure 1a, or Vo of Figure 3)
is obtained from (14) for constant rs, with the first term of define temporal and spatial scales of consideration for
(20) given by fb @hbs/@t for a constant bed sediment application of the Exner equation (e.g., discussed by Parker
concentration of fb. Sediment is supplied to or entrained et al. [2000] and Paola and Voller [2005]). The conven-
from the bed surface [e.g., Parker et al., 2000; Blom and tional 1D Exner equation of (2) for riverbeds, with sediment
Parker, 2004; Elhakeem and Imran, 2007; Garcı́a, 2008; transport as bed load qsbx (x, t), is thereby limited to a
Parker, 2008] through the relative bed surface and sediment minimum scale of consideration of the channel width, as
velocities of the last term of (20). The equivalent equation for standard 1D hydrodynamic models. The equivalent two-
for the suspended load balance (of z = h1 = hbs and z = h2 = dimensional equation of qsb (x, y, t) = (qsbx, qsby) allows
hw, with no flux through the water surface) contains an consideration of smaller scales. In terms of streamwise
equal and opposite term for loss to or gain from bed scales, the Exner equation is applied at sub-bed form
sediments [e.g., Garcı́a, 2008; Parker, 2008]. As discussed [e.g., Exner, 1925], multidune [e.g., Parker et al., 2000],
in Section 2.3, appropriate definitions of sediment layers or even larger (e.g., basin) scales. The multidune approach
(e.g., the bed surface, active and inactive sediments) and notably utilizes larger control volumes in order to sample a
transport modes (e.g., bed and suspended loads) can be statistically significant range of bed elevations, e.g., over the
made in terms of isosurfaces of sediment concentration fst, passage of a number of bed forms [Blom and Parker, 2004;
and also density hrs i or double-averaged mass transport rate Elhakeem and Imran, 2007]. The spatial scales of consid-
per unit normal area Fsi = fsthrs usi i, within regions of eration for the fully 3D equations of (6), (11) and (14) are
constant fst. determined by the domain of the averaging volume Vo,
[19] Paola and Voller [2005] present general forms of the where this is typically thin in the bed-normal direction and
Exner equation as their equations (PV.3) and (PV.13). The can vary in plan from the scale of a single grain, to bed
present spatial-averaging-derived general equation (14) is form, channel width or reach scales.
equivalent to (PV.3) per unit plan area and also (PV.13), [22] The analysis scales that are important, and thereby
where the interfacial fluxes of (PV.13) remain expressed in the size of the averaging volume to be adopted in applica-
terms of relative interface sediment movements (as per tion of the Exner equation, depend on the particular phe-
PV.3) for (14). Note that the Paola and Voller [2005] nomena or problems being studied. The choices of these
equations are in terms of mass per mixture analysis scales inherently govern all definitions and interpretations
volume Vo (i.e., their a, which is (fsthrs i) here), with of macroscopic (mixture-scale) sediment and layer proper-

6 of 8
W09421 COLEMAN AND NIKORA: EXNER EQUATION W09421

ties, e.g., sediment densities (where varying) and volume differential equation of mass conservation. These general
fractions, sediment transport modes and rates, and interfaces expressions include the effects of fluctuations in sediment
and (bed) layers. properties (e.g., density, velocity, and concentration or
[23] To illustrate the potential effects of varying Vo, volume fraction) within analysis volumes, and also enable
example patch-scale and dune-scale averaging volumes calculations in terms of size fractions. The obtained expres-
are shown in Figure 4, along with corresponding schematic sions highlight the averaged nature of variables, with
distributions of fst (z). At the patch scale, fst = (Vs/Vo)hfti is double-averaged sediment mass transport rate (per unit area)
a sediment volume concentration, where fs = 1  f and explicitly shown to be given by the product of volume
f is the local average sediment porosity. In contrast, fst = concentration, solid density, and sediment velocity. The
(Vs/Vo)hfti at the dune scale principally represents the general form of the Exner equation of (14) enables analyses
volume fraction of Vo that is occupied by a bed form. in terms of individual or successive layers, including bed
Averaged husi i for the bed sediments furthermore give and suspended loads, where layer interfaces (e.g., the bed
physically different patch velocities and dune celerities at surface) are shown to be defined on the basis of isosurfaces
the respective scales,
R h and the corresponding averaged trans- of sediment concentration, or other sediment properties
port rates gsi = h 2 (fsthrs usi i)dz represent the different (e.g., densities or transport rates) within regions of constant
1
mechanisms of transport as patches or as dune bed forms. concentration. This general equation can thereby incorpo-
The impact of averaging scales on definitions of bed surface rate the respective effects on the mass balance of intralayer
and bed and suspended loads are also highlighted in Figure 4. changes, interfacial (e.g., basement and bed surface) move-
It is notable that the bed surface level z = hbs is defined in ments, and horizontal and interlayer fluxes. For appropriate
terms of different characteristic values of fst for the two simplifications, the general form of the Exner equation is
analysis scales of Figure 4. In each case, this double- shown to reduce to give more specific conventional expres-
averaged (shown in relation to (14)) level is constant across sions, revealing the assumptions implicit in these equations.
Vo, and it is apparent that the determined bed surface level The effects of the scales of consideration, which can be
can be dependent on the plan dimensions of the averaging varied over wide ranges, on defining and interpreting
volume, e.g., influencing numerical diffusion in morphody- macroscopic (mixture-scale) sediment and layer properties
namic modeling. The definition of the bed surface in a (e.g., averaged densities, volume concentrations or frac-
hydrodynamic sense is noted by Nikora et al. [2002] to tions, velocities, transport modes and rates, interfaces and
similarly vary with averaging volume size. It is also bed layers) are highlighted.
apparent from Figure 4 that when averaging over larger
extents, e.g., for the dune of Figure 4b, adopting the value [25] Acknowledgments. The writers acknowledge useful and con-
of fst defining z = hbs to also distinguish between bed and structive comments made by the Associate Editor, Stephen McLean, Chris
Paola, Francesco Ballio, and an anonymous reviewer that have acted to
suspended sediment loads will result in some suspended strengthen the paper.
sediments immediately above the troughs being considered
as bed load and bed sediments near the crest being consid-
ered as suspended load. This blurring of the physical References
distinction between bed and suspended loads can be miti- Ancey, C. (2007), Plasticity and geophysical flows: A review, J. Non-
gated somewhat if a separate value of fst is adopted to Newtonian Fluid Mech., 142, 4 – 35, doi:10.1016/j.jnnfm.2006.05.005.
Batuca, D. G., and J. M. Jordaan (2000), Silting and Desilting of Reservoirs,
distinguish between bed and suspended loads, e.g., with all 353 pp., A. A. Balkema, Rotterdam, Netherlands.
sediments below crests considered as bed load for a selected Blom, A., and G. Parker (2004), Vertical sorting and the morphodynamics
value of fst defining the bed form crests. For a flat bed, of bed form – dominated rivers: A modeling framework, J. Geophys.
different values of fst can similarly potentially be used to Res., 109, F02007, doi:10.1029/2003JF000069.
Coleman, S. E., and J. D. Fenton (2000), Potential-flow instability theory
separately define an immobile bed surface and the top of the and alluvial stream bed forms, J. Fluid Mech., 418, 101 – 117,
bed load layer. It is clear from this example that interpre- doi:10.1017/S0022112000001099.
tation of any predictions of the Exner equation requires a Coleman, S. E., and V. I. Nikora (2008), A unifying framework for particle
good understanding of both the averaging nature of the entrainment, Water Resour. Res., 44, W04415, doi:10.1029/
equation, particularly when considering highly variable 2007WR006363.
Colombini, M. (2004), Revisiting the linear theory of sand dune formation,
topography, and the significance of the analysis scales in J. Fluid Mech., 502, 1 – 16, doi:10.1017/S0022112003007201.
this regard. Cui, Y., and G. Parker (2005), Numerical model of sediment pulses and
sediment-supply disturbances in mountain rivers, J. Hydraul. Eng.,
131(8), 646 – 656.
4. Conclusions Das, H. S., J. Imran, C. Pirmez, and D. Mohrig (2004), Numerical modeling
[24] The Exner equation of sediment mass conservation is of flow and bed evolution in meandering submarine channels, J. Geophys.
Res., 109, C10009, doi:10.1029/2002JC001518.
the foundation of morphodynamic analyses. As recently Dietrich, W. E., D. G. Bellugi, L. S. Sklar, J. D. Stock, A. M. Heimsath, and
highlighted by Paola and Voller [2005] and Parker [2008], J. J. Roering (2003), Geomorphic transport laws for predicting landscape
this equation has been stated in various ad hoc formulations form and dynamics, in Prediction in Geomorphology, edited by P. R.
for different situations, giving rise to the need for a general Wilcock and R. M. Iverson, Geophys. Monogr. Ser., 135, pp. 103 – 132,
AGU, Washington, D.C.
expression that both provides a universal description of the Drew, D. A. (1983), Mathematical modeling of two-phase flow, Annu. Rev.
sediment mass balance and also enables interpretation of the Fluid Mech., 15, 261 – 291, doi:10.1146/annurev.fl.15.010183.001401.
assumptions and limitations implicit in formulations of Dufek, J., and G. W. Bergantz (2007), Suspended load and bed-load transport
reduced, combined or improvised terms. In contrast to of particle-laden gravity currents: The role of particle – bed interaction,
conventional ‘‘mixture-scale’’ control volume approaches, Theor. Comput. Fluid Dyn., 21, 119 – 145, doi:10.1007/s00162-007-
0041-6.
general statements of sediment mass balance are derived Elhakeem, M., and J. Imran (2007), Density functions for entrainment and
herein from spatial averaging of the sub-particle-scale deposition rates of uniform sediment, J. Hydraul. Eng., 133(8), 917 – 926.

7 of 8
W09421 COLEMAN AND NIKORA: EXNER EQUATION W09421

Engel, P., and Y. L. Lau (1980), Computation of bed load using bathymetric Parker, G. (2008), Transport of gravel and sediment mixtures, in Sedimentation
data, J. Hydraul. Eng., 106(3), 369 – 380. Engineering: Processes, Measurements, Modeling, and Practice, Manuals
Engelund, F., and J. Fredsøe (1982), Sediment ripples and dunes, Annu. and Rep. on Eng. Practice 110, edited by M. H. Garcı́a, ASCE, Reston, Va.
Rev. Fluid Mech., 14, 13 – 37, doi:10.1146/annurev.fl.14.010182.000305. Parker, G., C. Paola, and S. Leclair (2000), Probabilistic Exner sediment
Exner, F. M. (1925), Über die Wechselwirkung zwischen Wasser und continuity equation for mixtures with no active layer, J. Hydraul. Eng.,
Geschiebe in Flüssen (in German), Sitz. Acad. Wiss. Wien Math. Natur- 126(11), 818 – 826.
wiss. Abt. 2a, 134, 165 – 203. Raudkivi, A. J. (1976), Loose Boundary Hydraulics, 2nd ed., Pergamon
Exner, F. M. (1931), Zur Dynamik der Bewegungsformen auf der Erdober- Press, Oxford, U. K.
fläche (in German), Ergeb. Kosmisch. Phys., 1, 374 – 445. Richards, K. J. (1980), The formation of ripples and dunes on an erodible
Garcı́a, M. H. (2008), Sediment transport and morphodynamics, in bed, J. Fluid Mech., 99(3), 597 – 618, doi:10.1017/S002211208000078X.
Sedimentation Engineering: Processes, Measurements, Modeling, and Simons, D. B., E. V. Richardson, and C. F. Nordin (1965), Bedload Equa-
Practice, Manuals Rep. Eng. Pract., 110, edited by M. H. Garcı́a, chap. 2, tion for Ripples and Dunes, U.S. Geol. Surv. Prof. Pap. 462-H, 9 pp.
pp. 21 – 164, Am. Soc. Civ. Eng., Reston, Va. Smith, J. D. (1970), Stability of a sand bed subjected to a shear flow of low
Graf, W. H. (1984), Hydraulics of Sediment Transport, Water Resour. Publ., Froude number, J. Geophys. Res., 75(30), 5928 – 5940, doi:10.1029/
Littleton, Co. JC075i030p05928.
Hoekstra, P., P. Bell, P. van Santen, N. Roode, F. Levoy, and R. Whitehouse Sumer, B. M. (2007), Mathematical modelling of scour: A review, J. Hydraul.
(2004), Bedform migration and bedload transport on an intertidal shoal, Res., 45(6), 723 – 735.
Cont. Shelf Res., 24, 1249 – 1269, doi:10.1016/j.csr.2004.03.006. Ten Brinke, W. B. M., A. W. E. Wilbers, and C. Wesseling (1999), Dune
Jerolmack, D. J., and D. Mohrig (2005), A unified model for subaqueous growth, decay and migration rates during a large-magnitude flood at a
bed form dynamics, Water Resour. Res., 41, W12421, doi:10.1029/ sand and mixed sand-gravel bed in the Dutch Rhine river system, in
2005WR004329. Fluvial Sedimentology VI, Spec. Publ. Int. Assoc. Sedimentol., 28, edited
Kennedy, J. F. (1963), The mechanics of dunes and antidunes in erodible-bed by N. D. Smith and J. Rogers, pp. 15 – 32, Wiley, New York.
channels, J. Fluid Mech., 16, 521 – 544, doi:10.1017/S0022112063000975. Van den Berg, J. H. (1987), Bedform migration and bed load transport in
Kubatko, E. J., and J. J. Westerink (2007), Exact discontinuous solutions of some rivers and tidal environments, Sedimentology, 34, 681 – 698,
Exner’s bed evolution model: Simple theory for sediment bores, doi:10.1111/j.1365-3091.1987.tb00794.x.
J. Hydraul. Eng., 133(3), 305 – 311. Whitaker, S. (1999), The Method of Volume Averaging, Kluwer Acad,
Lanzoni, S. (2008), Mathematical modelling of bedload transport over Dordrecht, Netherlands.
partially dry areas, Acta Geophys., 56(3), 734 – 752, doi:10.2478/ Yalin, M. S. (1992), River Mechanics, Pergamon Press, New York, N. Y.
s11600-008-0033-y. Zhang, D. Z., and A. Prosperetti (1997), Momentum and energy equations
McLean, S. R. (1990), The stability of ripples and dunes, Earth Sci. Rev., for disperse two-phase flows and their closure for dilute suspensions, Int.
29, 131 – 144. J. Multiphase Flow, 23(3), 425 – 453, doi:10.1016/S0301-9322(96)
Nikora, V., K. Koll, S. McLean, A. Dittrich, and J. Aberle (2002), Zero- 00080-8.
plane displacement for rough-bed open-channel flows, in Proceedings of Zhang, H., H. Nakagawa, Y. Muto, Y. Baba, and T. Ishigaki (2006),
the International Conference on Fluvial Hydraulics River Flow 2002, Numerical simulation of flow and local scour around hydraulic struc-
vol. 1, edited by D. Bousmar and Y. Zech, pp. 83 – 92, A.A. Balkema, tures, in River Flow 2006, edited by R. M. L. Ferreira et al., pp. 1683 –
Rotterdam, Netherlands. 1693, Taylor and Francis Group, London.
Nikora, V. I., I. McEwan, S. R. McLean, S. E. Coleman, D. Pokrajac, and
R. Walters (2007), Double-averaging concept for rough-bed open-channel
and overland flows: Theoretical background, J. Hydraul. Eng., 133(8), 

873 – 883. S. E. Coleman, Department of Civil and Environmental Engineering,
Paola, C., and V. R. Voller (2005), A generalized Exner equation for sedi- University of Auckland, Private Bag 92019, Auckland, New Zealand.
ment mass balance, J. Geophys. Res., 110, F04014, doi:10.1029/ (s.coleman@auckland.ac.nz)
2004JF000274. V. I. Nikora, School of Engineering, University of Aberdeen, Aberdeen,
AB24 3UE, UK. (v.nikora@abdn.ac.uk)

8 of 8

You might also like