The assessment of shear stress and bed stability in stream ecology


Arved C. Schwendel, Ecology Group, Institute of Natural Resources, Massey University, Private Bag 11222, Palmerston North 4442, New Zealand. E-mail:


1. Substratum stability and shear stress exerted by flowing water can have a strong influence on the structure of benthic communities. Bed stability can be characterised in a variety of ways, e.g. flow competence, threshold of particle entrainment, measures of erosion and deposition, particle transport distance, abrasion and bedload transport rate. This paper reviews methods for the quantification of bed stability and shear stress in streams and rivers that are relevant for the examination of the relationships between stream biota and bed stability.

2. The most suitable method for a research project depends mainly on the objectives. The targeted group of biota, spatial and temporal scale of investigation, as well as hydraulic conditions and substratum characteristics at the study site(s) determine the choice of a technique for the assessment of bed stability.

3. Indirect measurement of shear stress can be more accurate than calculations based on the DuBoys equation. However, the latter is preferred for reach-wide applications within the limits imposed by hydraulic conditions. The entrainment of the substratum is most effectively assessed using a combination of shear stress and competence equations, but the latter require careful parameterisation. At the patch-scale, direct measurement of entrainment force is a valid alternative.

4. Morphometric budgeting is the most comprehensive and least invasive technique for the assessment of rates of erosion and deposition. The transport of substratum particles is efficiently monitored with in situ marked or active tracer particles which allow for rapid and non-invasive identification and high recovery rate. As the assessment of bedload transport rate by formulae can be inaccurate, direct measurement is preferred. However, bedload traps interfere with the substratum and continuity of measurement with samplers is limited. Thus developments in the sector of acoustic and piezoelectric devices offer a potential alternative.

5. The abrasive forces by suspended sediments on stream biota are effectively evaluated with artificial blocks that are fixed on the stream bed. Descriptive surveys that assess bed stability offer an alternative to direct measurement and calculations. They are straightforward and non-invasive but can be observer-biased. If single methods do not provide useful links with biological data this may be improved by the application of a multivariate approach.

6. Many of the methods assessed have not yet been applied in research on benthic communities, but these hydraulic and geomorphologic techniques offer considerable potential for the assessment of bed stability in stream ecology.


Floods are an important controlling force on lotic ecosystems (Death, 2008) and influence the composition of benthic communities (Resh et al., 1988; Reice, Wissmar & Naiman, 1990; Lake, 2000). Most stream ecologists agree that discharges exceeding some threshold act as a disturbance to benthic communities, although determining those values can be problematic (Poff, 1992; Death & Winterbourn, 1994).

Under low water velocity and shear stress sediment is not entrained and the impact on benthic organisms is limited to shear force (drag and lift) exerted by flowing water. This alone may cause the patchy distribution of benthic organisms and can lead to downstream displacement of macrophytes (Biggs et al., 2001), periphyton (Biggs, Smith & Duncan, 1999; Suren & Duncan, 1999) and invertebrates (Lancaster & Hildrew, 1993a; Bond & Downes, 2000, 2003). As velocity and shear stress increase, phase-I bedload transport occurs when fine sediments may be winnowed (washed out) and rolled over a mostly stable coarser bed. This can lead to an additional impact on stream biota by abrasion (Downes et al., 1998; Bond & Downes, 2003). At a critical flow velocity, the movement of larger particles is initiated (phase-II bedload transport). This usually involves disruption of any armour layer (see Appendix S1 for definitions) at the bed surface and can result in patchy areas of scour and deposition (Powell, 1998; Matthaei, Peacock & Townsend, 1999b). In more extreme events, the whole bed may be mobilised, altering the habitat structure dramatically. This can lead to displacement of plants and invertebrates (Giberson & Caissie, 1998; Matthaei, Arbuckle & Townsend, 2000; Bond & Downes, 2003) and mortality of invertebrates crushed by rolling stones. Thus floods which induce bedload transport are often associated with the most dramatic changes in the composition, density and biomass of benthic invertebrate communities (Holomuzki & Biggs, 2000; Death, 2008) and periphyton (Biggs et al., 1999).

To examine the relationship between benthic biota and bed stability it is essential to quantify the latter accurately (Gordon, McMahon & Finlayson, 1992). There have been numerous attempts to do so, but most of the methods developed for stream hydraulics and fluvial geomorphology have yet to be adopted by stream ecologists. Furthermore, recent technological advances (e.g. acoustic and electronic sensors, active tracer particles and topographic survey methods) offer considerable potential for improving the measurement of bed movement for the study of stability–biota relationships.

This review presents methods that are used to assess different aspects of bed stability at different spatial and temporal scales, including: (1) shear stress, (2) entrainment, (3) erosion and deposition, (4) bedload transport and (5) abrasion. The techniques are evaluated not only for their potential to predict shear force and sediment movement per se, but also for their ability to explain biota-substratum stability relations.

Characteristics of bed stability

Shear stress

When stream flow lacks sufficient energy to move bedload (non-competent discharges), or where the bed is armoured or substratum particles are locked together (imbricated), the shear stress exerted on benthic biota by increased flows may be sufficient to alter the composition of benthic communities (Lancaster & Hildrew, 1993a; Bond & Downes, 2000, 2003). Shear forces exerted on organisms depend on their morphometry as well as kinematic viscosity and fluid velocity. Hence, the measurement of the latter can be used to determine shear stress. However, measurement of the velocity that affects small benthic organisms is difficult because of the steep velocity gradient in the boundary layer. Consequently indirect methods, like exposure to the flow of particles of known weight and/or size, are employed to estimate the shear stress exerted at the channel bottom.

DuBoys equation.  In stream ecology it is common to use the DuBoys eqn (1) to gain an estimate of the mean boundary shear stress τo at the reach level (e.g. Statzner, Gore & Resh, 1988; Matthaei et al., 1996, 1999b; Duncan, Suren & Brown, 1999).


The friction slope Sf (see Appendix S2 for symbol annotation) differs from the bed slope Sb and the water surface slope Sw, because flow resistance is responsible for energy losses (Robert, 1990). Sf can be calculated using a backwater calculation if flow data and channel geometry are available. However, the observed differences between Sf and Sw are often slight, especially under conditions of high discharge (Powell & Ashworth, 1995; Milan et al., 2001). Thus, the more easily measured Sw is an acceptable first-order approximation for Sf (Baker & Ritter, 1975; Lorang & Hauer, 2003). When the width-depth ratio of the channel is high (>16.9 according to Giberson & Caissie, 1998), which is common in coarse bedload transporting streams, mean flow depth h may be substituted for the hydraulic radius R (Baker & Ritter, 1975; Powell & Ashworth, 1995; Downes, Glaister & Lake, 1997; Giberson & Caissie, 1998). The use of local bed slope and depth instead of R and Sf in (1) might be preferable for the estimation of stream stability at the patch-scale, although actual shear stress is underestimated (Lorang & Hauer, 2003). Furthermore, it should be remembered that fluid density ρf is usually higher than the 1000 kg m−3 typically used because of suspended material, particularly during floods (Giberson & Caissie, 1998).

The DuBoys equation is strictly applicable only under uniform flow conditions (implying even bed topography and regular channel geometry) in wide channels (W/h > 20) (Gordon et al., 1992; Gore, 1996). Three-dimensional flow effects (Milan et al., 2001), bedform structures (e.g. pebble clusters and imbrication) (Carson & Griffiths, 1987) and the exposure to the thalweg (main thread of maximum velocity flow) are not accounted for. The values derived are high compared with local shear stress calculated from velocity profiles (Robert, 1990), but tend to underestimate the effective shear force (Carson & Griffiths, 1987).

The theoretical assessment of mean boundary shear stress is mostly based on the DuBoys equation. The choice of the parameters determines scale and accuracy of the calculation (Table 1). As the flow in natural rivers (especially shallow high gradient boulder- and gravel-bed rivers) is usually not uniform, the explanatory power of equations assuming the latter is limited (Campbell & Sidle, 1985). This may be enhanced by the inclusion of parameters like flow resistance, channel geometry and the energy slope (Lorang & Hauer, 2003). Thus shear stress estimations based on the DuBoys formula apply best under conditions of increased relative depth [R/D84 > 4 (Hey, 1979)], e.g. during high discharges, when flow is approximately uniform (Bhowmik, 1982; Milan et al., 2001). However, mean boundary shear stress from the DuBoys equation has been linked with the distribution of benthic invertebrates in several studies under various discharges (Statzner et al., 1988; Matthaei et al., 1996). The equation provides a useful tool for reach-wide investigations of shear stress biota relationships.

Table 1.   Methods for the assessment of shear stress (for annotations see Appendix S2)
MethodScaleConstraintsInterference with substratum Accuracy/relation to biological data
Du Boys equationReachUniform flow, W/h > 20Low (measurement of parameters)Overestimation of local shear stress (Robert, 1990) but underestimation of mean shear stress (Carson & Griffiths, 1987), recommended to assess the spatial distribution of invertebrates (Statzner et al., 1988)
DuBoys equation (using h and Sb)PatchUniform flowLow (measurement of parameters)Underestimation of local shear stress (Lorang & Hauer, 2003)
FST-hemispheresPatch to reach, short-termUsually normal flow conditionsLowRelated to invertebrate distribution (Dittrich & Schmedtje, 1995; Merigoux & Doledec, 2004), negative linear relationship with invertebrate taxon richness (Merigoux & Doledec, 2004) and with mussel density (Hardison & Layzer, 2001)
Point near-bed flow velocityPatchh/D84 > 3LowRelated to invertebrate distribution (Effenberger et al., 2006)
Depth averaged near-bed flow velocityPatchSimple flow geometry, logarithmic velocity profileLowThree times more accurate than point measurement (Wilcock, 1996)
Velocity profilePatchSimple flow geometry, logarithmic velocity profileLowProfiles least accurate compared with point and depth-averaged velocity, but no knowledge of bed roughness necessary (Wilcock, 1996)

FST-hemispheres.  Calibrated FliesswasserStammtisch (FST) hemispheres of different densities offer a measure of actual near-bed shear stress at a particular point in time (Statzner & Muller, 1989; Statzner, Kohmann & Hildrew, 1991). Despite some debate about the usefulness of FST-hemispheres for assessment of near-bed shear stress (Frutiger & Schib, 1993; Statzner, 1993; Dittrich & Schmedtje, 1995) they performed consistently well as indicators of ecologically relevant near-bed shear forces in hydraulically rough stream beds (Lancaster & Hildrew, 1993b; Scarsbrook & Townsend, 1993; Dittrich & Schmedtje, 1995; Hardison & Layzer, 2001; Merigoux & Doledec, 2004). However, Frutiger & Schib (1993) reported that only 50% of their benthic invertebrate taxa showed a relation between abundance and FST data. Statistical models based on FST measurements allow long-term characterisation of shear stress variability (Lamouroux et al., 1992) that can be linked with variation in the density of invertebrate taxa in different hydraulic microhabitats (Doledec et al., 2007). FST-hemispheres are a useful tool for investigating the spatial distribution of stream biota at base flow (Table 1). However, at higher discharges application is limited because of interference from bedload (impacts from saltating particles) and safety reasons (but see Gore et al., 1994).

Near-bed flow velocity.  Local shear stress can be estimated from measurements of flow velocity (e.g. single near-bed, vertical profile). Often a semi-logarithmic relationship between depth and velocity is assumed which is violated in reaches with high relative roughness [e.g. h/D84 < 3 (Bray, 1980)]. Wiberg & Smith (1991) found that local shear stress calculated from depth averaged velocity derived from a profile was accurate for h/D84 > 1. In comparison, single point near-bed measurement allows a calculation of shear stress for the widest range of conditions, but is not as accurate as the depth averaged method. Estimations of boundary shear stress based on the relation of v and ln(1 − h) in velocity profiles (e.g. Bhowmik, 1982) are the least accurate and apply in the most restricted flow conditions but require no estimate of bed roughness (Wilcock, 1996).

Effenberger et al. (2006) found a strong relationship between point measurements of near-bed flow velocity and the spatial distribution of invertebrates. Death & Winterbourn (1994) also found a strong positive correlation between the variability of near-bed flow velocity and the movement of marked stones.

Locally, indirect measurement of shear stress can provide more accurate results than the DuBoys approach. It may also provide an indication of the impact of shear stress on stream biota (Table 1) although the small spatial and temporal extent of the measurements limits the use for larger reaches and/or long-term studies.

Substratum entrainment

Relationship between substratum grain size and tractive force.  The proximal equality between mean boundary shear stress, calculated by the DuBoys equation, and the maximum diameter of entrained particles (rounded, non-cohesive, >0.05 m) (Lane, 1955) has been widely exploited to define critical particle size for entrainment (Newbury, 1984; Death & Winterbourn, 1994; Muotka & Virtanen, 1995; Giberson & Caissie, 1998). Even for non-rounded particles comparable relationships have been developed (Newbury, 1984). Although this relationship can provide a good indication of habitat stability amongst sites within a stream (Giberson & Caissie, 1998), it can overestimate particle movement in steep or narrow rivers (W/h < 16.5) as well as underestimate it in wide and shallow channels (W/h > 36.9) (Hallisey & Belt, 1996). This approach is subject to the same constraints as the DuBoys equation and does not account for potential equal mobility because of hiding and protrusion of particles. Thus the applicability of this concept is constrained to rivers with a high relative depth (h >> D50) [c. 6–7 (Newbury, 1984) and >10 (Duncan et al., 1999)] and bed slopes <0.01, conditions which are more likely to be met in lowland rivers.

Not surprisingly, therefore, several authors found no significant relationship with other measures of bed stability when they applied this approach in steep and shallow streams (Death & Winterbourn, 1994; Duncan et al., 1999). In contrast Cobb, Galloway & Flannagan (1992), Scarsbrook & Townsend (1993) and Muotka & Virtanen (1995) found a link between critical tractive force and the distribution of invertebrates and bryophytes. However, the relationship between tractive force and critical particle diameter cannot predict entrainment of the substratum consistently and applies in a limited range of rivers with gentle slope and high relative depth.

Shields equation.  The Shields eqn (2) (Shields, 1936) relates boundary shear stress to particle entrainment. It estimates the critical shear stress for a substratum grain size Di at the point of incipient motion.


The Shields coefficient θcrit is a non-dimensional variable dependent on particle shape, substratum particle size distribution, exposure and other packing factors (Lorang & Hauer, 2003). It reaches a constant value for non-cohesive materials larger than 6 mm (Lorang & Hauer, 2003) for hydraulically rough beds (boundary Re > 100). θcrit varies coarsely between 0.02 and 0.08, but more extreme values have been reported (Ashworth & Ferguson, 1989; Buffington & Montgomery, 1997; Shvidchenko, Pender & Hoey, 2001). Increasing channel slope [related to relative flow depth (h/D50)], decreasing relative size (Di/D50) and substratum heterogeneity (size distribution) increases the Shields coefficient systematically (Bathurst, Graf & Cao, 1987; Buffington & Montgomery, 1997; Shvidchenko et al., 2001). Furthermore, the definition of incipient motion (e.g. reference- or visual observation-based), grain shape, orientation, hiding effects (e.g. sheltering of smaller particles by larger), as well as discharge and bank vegetation influence θcrit (Andrews, 1984). Values for θcrit derived from visual-based studies (typically around 0.045) are recommended for analyses of incipient motion in discrete bed surface patches. In contrast, the usually higher reference-based θcrit may provide a better estimate of entrainment on a reach-average level because of its derivation from bedload transport measures and thus the integration of differential bed patch mobility (Buffington & Montgomery, 1997). Compared with the original Shields coefficient of 0.06, in gravel bed streams with a heterogeneous substratum, a lower θcrit is expected, for instance down to 0.02 in high gradient rivers (Sw > 0.002), where DMax/D50 > 22 (Lorang & Hauer, 2003) and the effects of form roughness and form drag resistance are considerable. A value of 0.045 for θcrit has been used in many studies and is widely accepted for beds with coarse particles and high boundary Reynolds numbers (Miller, McCave & Komar, 1977; Yalin & Karahan, 1979; Komar, 1989; Duncan et al., 1999).

There have been several attempts to improve the Shields equation and to widen its range of use (e.g. Komar, 1987; Thompson & Croke, 2008). Formulae such as eqn (3) incorporate the effects of hiding and heterogeneous beds in the Shields equation (Komar, 1989):


Duncan et al. (1999) also applied corrections to allow for small relative depths (h/D < 2.5) and high water surface slopes. Thompson & Croke (2008) incorporated the effects of bed form, microtopography and bed packing into the Shields equation. Lorang & Hauer (2003) found that critical shear stress calculated with a modified Shields equation overestimated the actual value for large cobble- and boulder-bed rivers by as much as an order of magnitude.

Andrews (1983) (cf. Parker, Klingeman & McLean, 1982) proposed the following relationship to calculate θcrit for 0.3 < Disurface/D50subsurface < 4.2:


This highlights the fact that critical shear stress is influenced more by relative grain size than absolute grain size (Ferguson, 1994; Shvidchenko et al., 2001). With the typical ratio of D50surface/D50subsurface = 2.5 for gravel bed rivers (Parker et al., 1982) θcrit can be estimated. However, in other studies the value for the first factor in eqn (4) lies between 0.019 and 0.087, whilst the exponent ranges from −0.32 to −1.25 (Buffington & Montgomery, 1997) and the values differ between riffles and pools (Sear, 1996).

A comparison between mean boundary shear stress (eqn 1) and critical shear stress for a particular grain size has been used to indicate zones of entrainment (Milan et al., 2001), calculate the critical size of substratum particles moved (Duncan et al., 1999) and define critical depth (Fuller et al., 2002). Predictions of entrainment were well correlated with measurements of morphological change in most areas of a gravel bed stream (Milan et al., 2001) and entrainment of in situ tagged particles (Biggs et al., 2001). Bed stability measurements derived from a combination of eqns (1 & 3) showed a strong relationship with the composition of bryophyte communities (Duncan et al., 1999) and periphyton biomass (Biggs et al., 2001; Table 2).

Table 2.   Methods for the assessment of critical shear stress and flow competence (for annotations see Appendix S2)
MethodScaleConstraintsInterference with substratumAccuracy/relation to biological data
Critical shear stress
τcrit ≈ D (Lane, 1955)Reachh ≫ D50, Sw < 0.01, uniform flow, unarmoured bedLow (measurement of D)Weak relationship with other measurers of bed stability or bryophyte cover (Death & Winterbourn, 1994; Duncan et al., 1999), linked to bryophyte (Muotka & Virtanen, 1995) and invertebrate distribution (Cobb et al., 1992), negative linear to number of invertebrates (Death & Winterbourn, 1995)
τcrit = θcrit (γ− γf) DiPatchUniform flow, uniform bed, low h/Di, low SLow (measurement of parameters)Depending on choice of θcrit
τcrit = θcrit (γ− γf) D50c DidPatchUniform flow, low h/Di, low SLow (measurement of parameters)Depending on choice of c, d, θcrit
Combination of Shields equation and DuBoys equation [+ corrections (Duncan et al., 1999)]ReachUniform flow, unarmoured bedLow (measurement of D, R, S)Related to actual entrainment (Milan et al., 2001), negative linear relationship with bryophyte cover (Duncan et al., 1999), related to periphyton biomass (Biggs et al., 1999)
τcrit = a DbPatchSite specificLow (measurement of D)Depending on parameters a, b
Spring balancePatchSubjectivity of particle choiceHigh(Downes et al., 1997)

Given the difficulties of selecting the most suitable parameters for empirical equations or the Shields coefficient, the calculation of the critical shear stress for entrainment is not straightforward, especially when a wide range of streams is being examined. However, for reach-scale investigations of the relationship between biota and bed stability a combination of the DuBoys formula and an advanced Shields equation (e.g. Duncan et al., 1999) may be useful.

Empirical equations of critical shear stress.  Several studies have produced empirical entrainment equations of the type τcrit = a Db (Thompson & Croke, 2008), where a and b range from 26.6 to 110 and from 0.38 to 1.21 respectively. The large range in parameter values is because of the difference in substratum assemblage between sites and differing methods used to define parameters (Lorang & Hauer, 2003). These empirical entrainment equations are thus too stream-specific to allow a general application of this approach.

Spring balance. Downes et al. (1997) used spring balances to measure the force necessary to initiate motion of particles in streams. This cannot be related directly to the critical shear stress but high forces will generally equate with high shear stresses as long as selective entrainment occurs (Downes et al., 1997). This is a labour intensive methodology for reach-scale studies and the choice of particles can be subjective, but it will reflect actual shear stress to entrain particles better than indirect measurements.

Erosion and deposition

Scour chains and other buried devices.  In both ecology and hydrology the deployment of metal scour chains is a common method for measuring scour and deposition of bed materials (Laronne & Duncan, 1992; Laronne et al., 1992; Palmer, Bely & Berg, 1992; Matthaei et al., 1999b; Matthaei, Guggelberger & Huber, 2003; Effenberger et al., 2006). It allows quantification of the height of fill and the depth of scour with an accuracy ranging from <D25 to D84 (Laronne et al., 1994; Matthaei et al., 2003) on a patch-scale systematic grid. Installation is relatively rapid [33 chains per person per day (Matthaei et al., 1999b)] and causes little damage to sediment structure. Effenberger et al. (2006) observed no long-term effects on the invertebrate community. The chains proved to be resistant to dislocation and can be relocated after floods using coloured ropes or magnetic tracers. However, the assessment of temporal variation of scour and fill during bed moving events is limited and relocation is required after each event that is likely to result in substratum movement (Laronne et al., 1994). As (phase-I) bedload transport occurs in patches in gravel bed rivers the suggested resolution of measurement is higher than one observation per square metre (Matthaei et al., 1999b; Laronne, Garcia & Reid, 2001).

Scour chains were employed for the identification of stable bed patches which can serve as local refugia for benthic organisms during floods (Matthaei et al., 1999b). Measures of scour and fill using scour chains have been related to density and vertical distribution of invertebrates (Palmer et al., 1992; Effenberger et al., 2006) as well as to the spatial distribution of benthic algae (Matthaei et al., 2003; Table 3).

Table 3.   Methods for the assessment of erosion and deposition
MethodScaleConstraintsInterference with substratum Accuracy/relation to biological data
Scour chainsPatch/reach, event-basedSubstratum < bouldersIntermediate during installationRelated to distribution of algae (Matthaei et al., 2003) and invertebrate taxa (Palmer et al., 1992; Effenberger et al., 2006)
Scour platesPatch, event-basedSubstratum < bouldersHighRelated to vertical invertebrate distribution (Palmer et al., 1992)
Dyed sand columns/painted gravelPatch, event-basedSubstratum sizeHighRelated to vertical invertebrate distribution (Palmer et al., 1992)
Pressure pillowsPatch, continuousSubstratum < bouldersHigh during installation(Kurashige, 2002)
Morphometric budgetingReach, event-basedGravel/cobble substratumLowAccuracy depends on surface roughness (Brasington et al., 2000)

Alternatively, metal scour plates, buried at fixed depths can serve as measurement of scour depth and in sandy streams columns of dyed sand inserted in the top layer of the bed can replace scour chains (Palmer et al., 1992). Wilcock (1997) measured the depth of entrainment with buried painted gravels. But both installation and retrieval require a disturbance of the substratum. Hence these methods are not appropriate for studies targeting benthic biota or for armoured and imbricated streambeds. Pressure pillows inserted into the surface of an artificial stream bed were used by Kurashige (2002) to measure sedimentation rates continuously but the construction was susceptible to damage during high bedload discharges.

Morphometric sediment budget models.  Movement of the substratum is reflected in changes of the morphology of the channel (Leopold, 1992). These changes can be assessed with repeated airborne surveys using digital photogrammetry or laser altimetry (e.g. Lane, 2001; Westaway, Lane & Hicks, 2001) or ground surveys employing tacheometry or photogrammetry (e.g. Ferguson & Ashworth, 1992; Lane, Chandler & Richards, 1994; Fuller et al., 2002; Heritage & Milan, 2004).

Ground surveys have been conducted with a theodolite-electronic distance measurement (EDM) system (Chappell et al., 2003; Fuller, Large & Milan, 2003b; Fuller et al., 2005) but more recently also with Real-Time Kinematic differential-GPS (RTK-dGPS; Brasington, Rumsby & McVey, 2000; Fuller & Hutchinson, 2007). The difference in altitude of cross-sections or digital elevation models (DEM) between surveys is used to determine areas of quantified deposition or erosion (Brasington et al., 2000; Brewer & Passmore, 2002). The calculation with DEM is preferable because sediment budgets derived from planform and cross-section measurement underestimate the magnitude of volumetric change compared with DEM subtraction, nor do they permit identification of the spatial pattern of volumetric change (Fuller et al., 2003a). Altitude measurements with RTK-dGPS or a theodolite-EDM system are, within the limits imposed by surface roughness (e.g. D50) highly accurate and more than 2000 points with high spatial resolution can be obtained per day (Brasington et al., 2000). The use of a GPS system is, however, limited at closed canopy sites and in deep valleys where satellite reception is critical.

Brasington, Langham & Rumsby (2003) indicate that ground surveys are much more precise than remote survey methods (especially at submerged zones; cf. Westaway, Lane & Hicks, 2000) and thus preferable for morphometric budgeting. However, for very wide river beds or reaches of more than a few hundred metres in length, the use of photogrammetry should be considered (Lane, Westaway & Hicks, 2003).

Morphometric budgeting has the advantage over scour chains to be less invasive and the ability to monitor an entire reach. However, scour chains may integrate effects of scour-fill compensation during single events. Both techniques present a lower bound estimate of the sediment flux because they do not account for substratum that is transported completely through the reach (Fuller et al., 2003a). According to Martin & Church (1995) the morphometric approach provides information of a quality comparable with or superior to that of direct measurements of transport, yet requires less field effort. Its application is restricted to gravel- and cobble-bed rivers. To the best of our knowledge, these measures have not been used in connection with biological data.

Bed load transport

Bedload is the sediment component that moves downstream by rolling or saltation. In rivers and streams where hydraulic conditions are generally unsteady (Lisle et al., 2000) and spatial substratum grain size variability is high (Dollar, 2002), transport rate is highly variable in space and time (Gomez, 1991; Batalla, 1997; Ferguson, 2003; Vericat & Batalla, 2007). Bedload discharge also depends on the supply of sediments within the catchment and lateral and longitudinal connectivity of the river (Dietrich et al., 1989; Hooke, 2003; Fryirs et al., 2007). The transport of substratum can be expressed as volumetric change in sediment budgets, transport rate at a point, cross-sectional discharge or distance travelled by individual particles. Techniques for measuring bedload transport are ideally non-intrusive, flexible and representative for different types of transport (Ergenzinger & de Jong, 2003). To date most stream ecologists have only been interested in qualitative measures of bed stability. At the single particle-scale, qualitative assessment might be sufficient, but for whole reaches bedload transport occurs on a continuous graduation. For stream ecologists, quantitative measures of bedload transport can act as a superior indicator for the level of bed stability, particularly if only partial mobilisation of the bed occurs.

Tracer particles.  Tracers are well suited for the stochastic and spatially variable nature of bedload transport because they reflect the movement of individual particles of known characteristics (Wilcock, 1997). Marked or tagged natural particles and artificial tracers are used to assess step length of movement (e.g. Habersack, 2001), proportion of the bed surface entrained (e.g. Laronne & Duncan, 1992), transport behaviour (e.g. Gottesfeld & Tunnicliffe, 2003) and transport rate (e.g. Ergenzinger & Conrady, 1982), or as an indicator of bed stability (e.g. Death & Winterbourn, 1994). Further they could facilitate the measurement of recolonisation periods of individual particles.

Stones coated with ordinary paint or fluorescent dye placed on the riverbed are often employed by ecologists and hydrologists (Death & Winterbourn, 1994; Townsend, Scarsbrook & Doledec, 1997; Ferguson & Wathen, 1998; Death, 2002; Ergenzinger & de Jong, 2003; Death & Zimmermann, 2005), but they have the disadvantage of a low recovery rate because of burial (Table 4). To overcome this, metal bars (Laronne et al., 1992; Schmidt & Ergenzinger, 1992) or magnets (Hassan, Church & Schick, 1991; Laronne & Duncan, 1992; Bunte, 1996; Ferguson & Wathen, 1998) can be inserted into the particles and they are detected using a metal detector or a magnetometer respectively. Magnetic tracers usually have a larger detection range (McEwan, Habersack & Heald, 2001) than metal tracers. An easier but less durable alternative to the insertion of metal is the wrapping of stones with aluminium foil (Sear et al., 2003). The transport rate and transport behaviour of particles marked with magnets or stones containing magnetic minerals can be monitored with a bar equipped with electromagnetic coils across the stream (Ergenzinger, 1985; Carling et al., 1998; Froehlich, 2003) or with a longitudinal line of ‘Bed Movement Detectors’ (Gottesfeld & Tunnicliffe, 2003). The overpassing of a magnetic particle induces an electric signal which is stored with high temporal resolution. The calculation of bedload discharge is possible using dispersion models (Sear et al., 2000b).

Table 4.   Methods for reach-scale tracking of tracer particles
MethodConstraintsDetection depthRecovery rate (%)Relation to biological data and comments
Tracking of initially unembedded particles
Painted tracer (visual)Armour layer, burialSurface15–60Negative with periphyton biomass (Death & Zimmermann, 2005), negative linear with invertebrate species number and species richness (Death & Winterbourn, 1995; Death, 2002; Death & Zimmermann, 2005), quadratic with invertebrate taxon number (Townsend et al., 1997)
Metal tracer (passive)Armour layer, particle size0.5–1 m50–90 
Stones wrapped in aluminium foil(passive)Armour layer0.25 m  
Magnetic tracer (passive)Armour layer, particle size0.5–1 m, usually higher than with metal tracer50–90 
Transmitters (active)Armour layer, particle size, battery, low conductivityShallow waterUp to 100Life span: a few weeks to 10 months (size from 0.01 to 0.08 m respectively)
Radioactive tracer (passive)Armour layer, environmental issues c. 5 
Different lithology (visual)Armour, burialSurface5–30 
Artificial tracer (visual/passive)Armour layer, representativeness of substratumVariablec. 35 
DUMPLING (active)Size (0.3 m), weight (37 kg) 100 
Tracking of initially embedded particles
Chiselled stones (visual)Particle choiceLowLow 
Dyed quick concrete mix (visual)Particle choiceSurface Distribution of invertebrates (Barquin & Death, 2006)

Marking of tracer particles has been further advanced via insertion of radio transmitters into a particle. A signal is transmitted either continuously, at a programmed interval or when the particle is turned 180o (Ergenzinger, Schmidt & Busskamp, 1989; Schmidt & Ergenzinger, 1992; Busskamp & Hasholt, 1996; Habersack, 2001). The tagged stones can be tracked from the banks with a set of antennae but application is restricted to shallow water and low conductivity (Ergenzinger & de Jong, 2003). Battery capacity (size) is a trade-off between life span and lower size boundary of particles (Habersack, 2003). These tags enable the monitoring of step length and transport behaviour as well as initiation of motion.

Radioactive tracers (e.g. 137Cs) are an alternative to tags because they do not change density or centre of gravity (e.g. Bartnik, Madeyski & Michalik, 1992). However, they are no longer widely applied because of environmental issues (Ergenzinger & de Jong, 2003). The employment of tracers of differing lithology from the natural substratum (Mosley, 1978; Kondolf & Matthews, 1986) provides an effective and easy measure for event-based distribution of transport length, although recovery rate is low.

For the in situ marking of substratum particles Downes et al. (1998) and Matthaei, Peacock & Townsend (1999a) used chisels and drills with long drill bit extensions, but relocation is difficult and embeddedness may be disturbed during the marking process. Thus this method is more suitable for the qualitative measurement of entrainment. Barquin & Death (2006) used dyed quick curing concrete mix to mark embedded stones.

Artificial stones provide an alternative to natural particles and also provide the opportunity to examine the influence of shape on transport length (Schmidt & Ergenzinger, 1992). The use of cast aluminium forms avoids the insertion of metal bars in pebbles (Sear et al., 2003). The collection of complex information about particle transport is also possible with artificial boulders like the DUMPLING (Ergenzinger & de Jong, 2003), although its size and weight restricts its application to bouldery streams.

The measurement of bedload transport with tracers provides comparable results to direct measures but requires less effort and avoids large-scale intervention in the stream bed. For low transport rates, tracers are likely to be more accurate (Wilcock, 1997). However, the dominating influence of bed structure and channel morphology on the distribution of tracer stones and the weak relationship with stream power (Kondolf & Matthews, 1986; Hassan, Church & Ashworth, 1992) suggests that short-term studies with tracers are not sufficient to compute rates of bedload transport. In contrast, shorter-term studies are more suitable for investigating the movement of surface particles because the transport rate of tracer particles decreases because of vertical mixing (burial) and storage in less active zones of the system (e.g. floodplain and bars) (Ferguson et al., 2002). If particles have to be removed from the stream for marking, bed structures and imbrication are destroyed and tracer particles placed on the bed surface may not represent the size characteristics of the substratum (Downes et al., 1998; Biggs et al., 1999). Longer-term studies can account for this, but they do not provide information about the frequency and magnitude of single disturbance events. The subjective choice and the shape of particles, as well as their number, may bias the results of tracer experiments (Schmidt & Gintz, 1995; Duncan et al., 1999; Warburton & Demir, 2000; Ferguson & Hoey, 2002).

Nevertheless, a stability index derived from tracer experiments showed a strong negative relationship with invertebrate diversity and periphyton biomass (Death & Winterbourn, 1995; Death, 2002; Death & Zimmermann, 2005; Table 4). In situ marked stones were also used to identify stable stones that can serve as refugia during floods (Matthaei et al., 2000). They relate the shear forces to the local substratum and consequently provide a better estimate of bed stability than unembedded tracers (Downes et al., 1998; Matthaei et al., 1999a). In combination with a non-invasive detection technique, in situ marked particles may be highly appropriate for ecological studies. Along with the objectives of a study, selection of an optimal tracer technique should consider representation of the substratum, tracer recoverability, longevity, durability, possibility of explicit identification of particles as well as labour and cost efficiency (Sear et al., 2000b).

Bedload transport sampler and traps.  The rate of bedload transport can be assessed with samplers and traps at various scales (Table 5). The most common handheld bedload transport samplers are of the pressure-difference type (Helley-Smith-, Vyzkum Ustav Vodohospodarsky (VUV)- and Arnhem sampler) with orifices up to 0.05 m2 (Leopold, 1992; Hoey, Cudden & Shvidchenko, 2001; Hardardottir & Snorrason, 2003). Their sampling efficiency usually varies between 30% and 70%, but can be up to 100% (Helley-Smith sampler) (Gomez, 1991). A common constraint of these samplers is that the opening area needs calibration for hydraulic and substratum conditions (Gomez, 1991) but, much more critically, the sampling scheme should be sufficient to account for the cross-sectional substratum variability of the reach and the temporal variability in bedload transport (Ergenzinger & de Jong, 2003). This requires adjustment of the sampling period and may result in large sampling efforts in wide rivers. Therefore, predictions of bedload transport based on sampler measurements are often not very accurate (uncertainty of ± 50%) (Wilcock, 2001). In conditions encountered in mountain streams (e.g. local high flow velocities and high surface roughness) bedload transport samplers are less applicable (Mizuyama, Fujita & Nonaka, 2003). Here portable net traps fixed to platforms on the stream bed may be used, delivering similar results to pit traps (Wilcock, 2001; Bunte & Abt, 2003; Bunte et al., 2004). Bedload samplers are not frequently employed by stream ecologists perhaps because of the mentioned constraints and inaccuracy. However, for small-scale, event-based studies they constitute a potentially valid option for direct measurement of bedload transport rate.

Table 5.   Methods for the assessment of bedload transport
MethodScaleConstraintsInterference with substratumAccuracy/relation to biological data
  1. ADCP, acoustic Doppler current profiling.

Pressure-difference samplerPatch, short-termOrifice area (up to 0.05 m2), upscaling to stream widthLowSampling efficiency usually 30–70%, can reach up to 100%, small volume
Birkbeck slot samplerPatch/reachSlot width, upscaling to stream widthHigh for installationContinuous during smaller floods
Sediment trapCross-section, continuous High for installationSampling efficiency up to 100%
Acoustic sensorsPatch/reachCalibrationLow – high for installationComparable accuracy as bedload traps (Downing et al., 2003)
ADCPPatch/reachSandy substratum, high suspended loadNone 
Electronic momentum sensorPatchCalibrationLowMeasures a combination of particle size and speed (Richardson et al., 2003)
Piezoelectric sensorsReach, long-termCalibrationLow (installation)Limited accuracy for single events (Rickenmann & McArdell, 2007)
Bedload transport formulaeReachCalibration site specificLow (measurement of parameters)Inaccurate for general application

Slot traps of various dimensions, inserted into the river bed, are used in many parts of the world (Salehi, Lagace & Pesant, 1997; Martin-Vide et al., 1999; Hassan & Church, 2001; Sear et al., 2003; Bond, 2004). They range from small sized pit traps, without continuous measurement, to Birkbeck samplers and large, stream-wide constructions for continuous monitoring. The latter is achieved with the employment of a weighing device (pressure cushion or load cell) below the sampling box or outside the channel (vortex tube, pump or conveyor belt) (Gomez, 1991; Sear et al., 2000a; Ergenzinger & de Jong, 2003; Sear, 2003). Load cell systems are more reliable than pressure cushion devices because they are less susceptible to damage (e.g. puncture of pressure pillows; Lewis, 1991). Smaller pit traps may fill rapidly during large events but are generally more accurate than handheld bedload transport samplers (Wilcock, 2001). Sampling efficiency for pit traps is up to 100%, decreasing with increasing fill (Laronne et al., 2003). In particular at base flow, bedload transport traps may also sample suspended sediments (Batalla, 1997). The installation and maintenance of a bedload trap is expensive and involves a serious disturbance of the stream bed and biota. For this reason, bedload traps have not been used for investigations of benthic biota but for long-term projects they offer a useful tool for the assessment of ecologically relevant bedload discharge. As an alternative, monitoring of sediment volume accumulated in natural traps (basins), reservoirs or retention and diversion devices provides an opportunity to assess bedload transport rate, but calibration to exclude suspended sediments is difficult (Gomez, 1991).

Acoustic sensors.  Acoustic sensors can be used to assess bedload transport intensity and the onset and cessation of movement (Ergenzinger & de Jong, 2003). In addition, estimates of transport rate using acoustic energy and estimates of transported particle size using the emitted frequency can be obtained (Bogen & Moen, 2003; Downing et al., 2003; Froehlich, 2003; Mizuyama et al., 2003). Hydrophones must be calibrated against actual bedload samples at each site. The sensor consists of a plate fixed horizontally on the bed (Bogen & Moen, 2003), a vertical pressure plate (Downing et al., 2003) or horizontal steel pipes across the stream bed (Froehlich, 2003; Mizuyama et al., 2003). Calibration limits the application at numerous sites, but the accuracy can be similar to a bedload trap. Acoustic Doppler current profiling (ADCP) allows the combined measurement of multi dimensional flow and velocity of bedload and suspended load (Rennie & Millar, 2004). Limitations of this technique include problems with the differentiation between near-bed suspension, bedload and fine grained bottom sediments as well as varying sensitivity to different particle sizes (Kostaschuk et al., 2005).

Other sensors. Richardson, Benson & Carling (2003) presented an electronic sensor that allows detection of the momentum of impacting particles in bedrock channels. It provides a relative measure of bedload transport but needs to be calibrated. The latter can create some difficulties because the sensor measures a combination of grain mass and speed.

The piezoelectric bedload impact sensor employed by Rickenmann & McArdell (2007) can measure impacts of transported grains larger than 10–30 mm. These sensors are placed in an array over the whole stream width in a concrete bar. The measure is a reliable and continuous indicator of total bedload transport, but it needs to be calibrated and has limited accuracy for single events or small bedload volumes. Further it provides no information about the grain size distribution of the overpassing sediments (Rickenmann & McArdell, 2007).

Bedload transport formulae.  Bedload transport formulae (e.g. Schoklitsch-type eqn (5)) are generally based on four principal approaches: shear stress, stream discharge, stream power and a stochastic function for sediment transport (Gomez & Church, 1989).


In this example, bedload discharge qb depends on excess water discharge and a sediment coefficient X′. Most bedload transport formulae originate from physical principles but their precision has been improved by the use of empirical datasets from flumes and streams. The formulae are consistent in that they employ in most instances the same hydraulic parameters (energy gradient, flow velocity, depth and discharge) which are in part intercorrelated (Gomez & Church, 1989; Martin & Church, 2000). Most formulae are well suited and parameterised for the dataset of their development, but fail when applied to other conditions (Knighton, 2008). They are based on limited basic assumptions which vary between streams and even within streams (e.g. selective entrainment). Characteristics like armouring, exposure to flow, equal mobility, variable sediment supply and pulsing cannot be fully accounted for, although some approaches try to incorporate these points (Parker, 1990; Duan & Scott, 2007; Thompson & Croke, 2008). Furthermore, the spatial variability within a stream is ignored because of the one-dimensional nature of the formulae (Hoey et al., 2001; Ferguson, 2003; Martin & Ham, 2005). The result of comparative studies with bedload samplers/traps (Gomez & Church, 1989; Batalla, 1997; Martin-Vide et al., 1999; Habersack & Laronne, 2002; Barry, 2004) and morphologic budgeting (Martin & Ham, 2005) show clearly that bedload transport formulae perform inconsistently (but see Bartnik et al., 1992). Thus, bedload transport formulae need to be carefully selected according to the conditions for which they were developed, for instance turbulent and shallow mountain streams require other types of models than gravel-bed rivers (Biggs et al., 2001; Mizuyama et al., 2003; Ancey et al., 2008). Additionally, empirical parameters and the entrainment threshold have to be determined to suit a new dataset, which is a difficult task (Wilcock, 2001; Habersack & Laronne, 2002). Thus the application of direct measurements of bedload transport is preferable to the use of bedload transport formulae (Gomez, 1991; Laronne et al., 1992).

Abrasion by suspended sediments

Abrasion is an often neglected form of disturbance which can affect benthic flora and fauna. At normal flows the stream biota may be subjected to constant in situ abrasion by small suspended particles, which may represent a significant disturbance at higher discharges (Biggs, 1996; Peterson, 1996). It is unclear if sandblasting affects invertebrates (Rosenberg & Wiens, 1978; Culp, Wrona & Davies, 1986; Bond & Downes, 2003) but the effect on benthic algae is clearly recognised (Biggs et al., 1999; Webb et al., 2006).

The exposure of natural or artificial tracers to abrasion is an obvious opportunity for quantification (Table 6). The use of natural rocks that are cut in cubes or artificial blocks improves the visual monitoring of abrasion because the loss of edges and corners is simply detected. Furthermore, impact marks on the cube faces are subject to easy distinction and aide the interpretation of bedload moving events (Brewer, Leeks & Lewin, 1992). Blocks that are of the same lithology as the river sediments have the advantage that they provide a better estimation of the actual abrasion in the channel. However, for the quantification of the impact on biota a measure of relative abrasion is sufficient. Thus ecologists prefer to use artificial tracers, like autoclaved lightweight aerated concrete blocks (Webb et al., 2006). The latter have standardised material properties and abrade consistently proportional to the physical work performed on their surface. Moreover the abrasion rate is high enough to allow short deployment times (e.g. 2 months) which minimises mass loss by dissolution. Abrasion blocks need to be protected from the impact of bedload transport to gain a pure measure of abrasion by suspended particles. There is also the choice between blocks fixed on the stream bed or on bedrock, semi-mobile tethered blocks as well as loose tracer particles of known weight and size (Stott & Sawyer, 2000). For measurements relevant to stream invertebrates or periphyton it is preferable to place the blocks on the stream bed. Although fixed or tethered blocks may split and get lost or buried by sediments, the recovery rate can be high (Brewer et al., 1992). These methods do not allow distinction between effects of sandblasting, overpassing bed materials and the physical impingement of fast flowing water. However, the practical consequences for ecologists are small because, in the field, biota are usually exposed to a combination of these effects (Webb et al., 2006).

Table 6.   Methods for the assessment of abrasion by suspended sediments
MethodScaleConstraintsInterference with substratum Accuracy/relation to biological data
Stone blocksPatchAbsoluteLowActual abrasion of sediment
Artificial blocksPatch/monthsDissolution, high bedload TransportLowOnly relative measurement
Abrasion coefficientsReach/patchCalibrationNoneUnderestimation of actual abrasion (Lewin & Brewer, 2002)

Abrasion coefficients derived from laboratory experiments are an easy alternative to field measurements but they generally underestimate the actual abrasion in rivers (Lewin & Brewer, 2002). Sklar & Dietrich (2004) presented a model to predict bedrock abrasion by saltating particles but it has not yet been applied in context with stream biota.

Descriptive surveys of substratum stability and multivariate approaches

Pfankuch Stability Index.  The Pfankuch Stability Index is a qualitative measure that describes the probability of occurrence of substratum-moving discharges (Pfankuch, 1975). It consists of 15 variables representing properties of the upper and lower banks and the stream bed. Despite its subjectivity it shows a strong positive relation with the entrainment of painted stones (Townsend et al., 1997), but not when the painted stones are used as an indicator of tractive force over time (Death & Winterbourn, 1994). If just the stream bottom component of the Pfankuch Index is employed, the relationship with other stability measures is considerably higher (Death & Winterbourn, 1994; McIntosh, 2000) and the assessment of stability at finer spatial scales might be possible (Winterbourn & Collier, 1987).

Descriptive approaches for the assessment of stream bed stability provide an easily applicable tool which has been widely exploited for investigations of biota in streams. Their major problem is the propensity to be observer-biased (Duncan et al., 1999). Additionally, large temporal variation in scores can occur between surveys of the same reach by the same observer (A. C. Schwendel, unpubl. data). Nevertheless, relations between bed stability assessed with the stream bed component of the Pfankuch Index and biological data have been established (Table 7).

Table 7.   Descriptive surveys for the estimation of bed stability on a reach-scale
MethodConstraintsInterference with substratum Accuracy/relation to biological data
Pfankuch IndexSubjectivity of perceptionNoneRelated to other measures of bed stability, negative linear relationship with invertebrate taxon number (Townsend et al., 1997)
Pfankuch Index bottom componentSubjectivity of perceptionNonePositively related to other measures of bed stability (Death & Winterbourn, 1994), negative linear relationship with bryophyte cover (Suren, 1996; Duncan et al., 1999), negative linear relation to invertebrate species richness, number and density (Death & Winterbourn, 1995; Death, 2002)

Multivariate approaches.  Approaches that combine more than one measure of bed stability can have a stronger relationship with biological data because they can incorporate different aspects of substratum stability. Death & Winterbourn (1994) showed that a multivariate instability score consisting of hydraulic parameters (patch-scale), the movement of painted stones, water temperature and the bottom component of the Pfankuch Index (reach-scale) had a stronger positive linear relationship with invertebrate species richness than with any of the constituent single variables.


The composition of benthic communities is a function of habitat and biotic interactions. Habitat stability in rivers is primarily determined by the forces of flowing water exerted on biota and substratum. Hence measurement of shear stress and substratum stability can indicate the distribution of benthic stream organisms, but they differ in precision and the aspect of bed stability they describe. Clearly there is no single technique suitable for all applications. Thus the selection of an appropriate method is subject to: (1) targeted fauna (mobility and range of activity), (2) spatial (3) temporal scale of investigation (flood event-based or long-term), (4) hydraulic (5) substratum conditions and (6) research question of the study (e.g. range of flow and aspect of bed stability).

Most of the methods presented have been developed for research into stream hydraulics and fluvial geomorphology. Despite recent technological advances and development of new techniques only a few of them have been applied in ecological studies. Given the importance of bed stability for the biota of many streams and rivers and the multitude of ways to characterise that stability, we would like to encourage stream ecologists to consider also the potential of alternative techniques highlighted in this review for examining the links between stream stability and biota.


The authors thank Alan Hildrew, Bernhard Statzner and an anonymous reviewer for their helpful comments on earlier versions of this manuscript.