Impact–entrainment relationship in a saltating cloud

The problem of impact–entrainment relationship is one of the central issues in understanding saltation, a primary aeolian transport mode. By using particle dynamic analyser measurement technology the movement of saltating particles at the very near‐surface level (1 mm above the bed) was detected. The impacting and entrained particles in the same impact–entrainment process were identified and the speeds, angle with respect to the horizontal, and energy of the impacting and entrained sand cloud were analysed. It was revealed that both the speed and angle of impacting and entrained particles vary widely. The probability distribution of the speed of impacting and entrained particles in the saltating cloud is best described by a Weibull distribution function. The mean impact speed is generally greater than the mean lift‐off speed except for the 0·1–0·2 mm sand whose entrainment is significantly influenced by air drag. Both the impact and lift‐off angles range from 0° to 180°. The mean lift‐off angles range from 39° to 94° while the mean impact angles range from 40° to 78°, much greater than those previously reported. The greater mean lift‐off and especially the mean impact angles are attributed to mid‐air collisions at the very low height, which are difficult to detect by conventional high‐speed photography and are generally ignored in the existing theoretical simulation models. The proportion of backward‐impacting particles also evidences the mid‐air collisions. The impact energy is generally greater than the entrainment energy except for the 0·1–0·2 mm sand. There exists a reasonably good correlation of the mean speed, angle and energy between the impacting and entrained cloud in the impact–entrainment process. The results presented in this paper deserve to be considered in modelling saltation. Copyright © 2002 John Wiley & Sons, Ltd.     

Development of the saltation system under controlled environmental conditions

The transport of sand by the wind occurs predominantly by the process of saltation. Following the entrainment of sand by an above threshold wind, the saltation system is regulated by the mutual interaction of the atmospheric boundary‐layer, the sand cloud and the sand bed. Despite existing data on the spatial and temporal development of the sand transport system, very little is known about the development of the saltation system towards equilibrium. Results are presented from wind‐tunnel experiments that were designed to address the simultaneous spatial and temporal development of the saltation system, with and without artificial sand feed. The development of the saltation system was monitored over a streamwise length of 8 m during a period of 3600 s. Mass flux data were measured simultaneously at 1 m intervals by the downwind deployment of seven Aarhus sand traps. Wind velocity data were collected throughout the experiments. The downwind spatial development of the saltation system is manifested by an overshoot in mass flux and friction velocity prior to declining towards a quasi‐equilibrium. Mass flux overshoots at approximately 4 m downwind, in remarkable agreement with existing data of a comparable scale. Friction velocity overshoots at approximately 6 m downwind, a result not previously witnessed in saltation studies. The overshoot of mass flux prior to the overshoot in friction velocity is a spatial manifestation of the time lag between the entrainment of grains and the deceleration of the wind by the grains in transport. Temporally, the development of the saltation system is controlled by the availability of entrainable grains from the sand bed. Through time the saltation system develops from a transport‐limited to a supply‐limited system. The depletion of the sand bed through time limits the appropriateness of the assumption of ‘equilibrium’ for the universal prediction of mass flux. Copyright © 2002 John Wiley & Sons, Ltd.     

Equilibration of saltation

A two‐dimensional numerical model of the saltation process was developed on a parallel computer in order to investigate the temporal behaviour of transport rate as well as its downwind distribution. Results show that the effects of unsteady flow on the transportation of particulates (sediment) have to be considered in two spatial dimensions (x, y). Transport rate Q(x, t) appears in the transport equation for mass M(x, t): where A = ΔxW denotes unit area composed of unit streamwise length Δx and width W. S(x, t) (units kg m⁻² s⁻¹) stands for the balance over the splash process. A transport equation for transport rate itself is suggested with U _c (x, t) a mean particle velocity at location x as the characteristic velocity of the grain cloud. For a steadily blowing wind over a 50 m long sediment bed it was found that downwind changes in Q cease after roughly 10–40 m, depending on the strength of the wind. The onset of stationarity (∂/∂t=0) was found to be a function of the friction velocity and location. The local equilibrium between transport rate and wind was obtained at different times for different downstream locations. Two time scales were found. One fast response (in the order of 1) to incipient wind and a longer time for equilibrium to be reached throughout the simulation length. Transport rate also has different equilibrium values at different locations. A series of numerical experiments was conducted to determine a propagation speed of the grain cloud. It was found that this velocity relates linearly to friction velocity. Copyright © 2000 John Wiley & Sons, Ltd.     

Experimental study of aeolian sand ripples in a wind tunnel

The topographic parameters and propagation velocity of aeolian sand ripples reflect complex erosion, transport, and deposition processes of sand on the land surface. In this study, three Nikon cameras located in the windward (0–1 m), middle (4.5–5.5 m), and downwind (9–10 m) zones of a 10 m long sand bed are used to continuously record changes in sand ripples. Based on the data extracted from these images, this study reaches the following conclusions. (1) The initial formation and full development times of sand ripples over a flatbed decrease with wind velocity. (2) The wavelengths of full development sand ripples are approximately twice the wavelengths of initially formed sand ripples. Both wavelengths increase linearly with friction velocity. During the developing stage of sand ripples, the wavelength increases linearly with time. (3) The propagation velocity of full development sand ripples is approximately 0.6 times that of the initially formed sand ripples. The propagation velocity of both initial and full development of sand ripples increase as power functions with respect to friction velocity. During the developing stage of sand ripples, the propagation velocity decreases with time following a power law. These results provide new information for understanding the formation and evolution of aeolian sand ripples and help improve numerical simulations. Copyright © 2017 John Wiley & Sons, Ltd.     

A wind tunnel investigation of the influences of fetch length on the flux profile of a sand cloud blowing over a gravel surface

Wind tunnel tests were conducted to examine the fetch effect of a gravel surface on the ?ux pro?le of the sand cloud blowing over it using typical dune sand. The results suggest that the ?ux pro?le of blown sand over a gravel surface differs from that over a sandy surface and is characterized by a peak ?ux at a height above the surface while that over a sandy surface decreases exponentially with height. The ?ux pro?le of a sand cloud over a gravel surface can be expressed by a Gaussian peak function: q = a + b exp (?0·5((h ? c)/d)²), where q is the sand transport rate at height h, and a, b, c and d are regression coef?cients. The signi?cance of the coef?cients in the function could be de?ned in accordance with the fetch length of the gravel surface and wind velocity. Coef?cient c represents the peak ?ux height and increases with both wind velocity and fetch length, implying that the peak ?ux height is related to the bounce height of the particles in the blowing sand cloud. Coef?cient d shows a tendency to increase with both wind velocity and fetch length. The sum of a and b, representing the peak ?ux, increases with wind velocity but decreases with fetch length. The average saltation height derived from the cumulative percentage curve shows a tendency to increase with both the fetch length and wind velocity. For any fetch length of a gravel surface the sand transport equation is expressed as Q = C(1 ? U_t/U)(ρ/g)U³, where Q is the sand transport rate, U is the wind velocity, U_t is the threshold velocity measured at the same height as U, g is the gravitational acceleration, ρ is the air density, C is a proportionality coef?cient that decreases with the fetch length of the gravel surface. At a given wind velocity, the sand transport rate over a gravel surface is only 52–68 per cent of that over a sandy surface. The ?ux rate in true creep over a gravel surface increases with wind velocity but decreases with the fetch length, whereas the creep proportion (the ratio of creep ?ux to the sand transport rate) decreases with both the wind velocity and fetch length. Two‐variable (including fetch length and wind velocity) equations were developed to predict the peak ?ux height, average saltation height and transport rate. Copyright © 2004 John Wiley & Sons, Ltd.     

Bagnold's kink: A physical feature of a wind velocity profile modified by blown sand?

Bagnold (1941) made detailed measurements of the wind profile modified by blown sand. He noted that each velocity profile appeared to be kinked and suggested that the position of the kink corresponded to the height at which an average or a characteristic trajectory extracted the bulk of its momentum from the wind. However, Anderson and Haff (1988) have shown that in aeolian sand transport the grain cloud is made up from a distribution of trajectory paths and that it was an over-simplification to attempt to describe the complex behaviour of the grain cloud in terms of a single representative trajectory. These recent developments leave the nature and even the existence of the kink observed by Bagnold (1941) open to question. This paper shows that a kink is indeed a physical feature of the modified wind velocity profile and that it is caused, as Bagnold (1941) suggested, by a maximum, occurring at some height above the surface, in the momentum extracted by the grains from the wind. Further comments are made on both the form of the modified wind velocity profile and the fluid shear stress profile in the grain layer. In particular these comments, based upon a theoretical analysis, suggest an experimental means to measure the mean surface fluid shear stress and to gain, from accurately measured wind profiles, a greater insight into the grain cloud which caused the modification.     

The effects of surface moisture on aeolian sediment transport threshold and mass flux on a beach

This paper presents results from a study designed to explore the effects of beach surface moisture and fetch effects on the threshold of movement, intensity of sand transport by wind and mass flux. The experiment was carried out over a period of five weeks at Greenwich Dunes, Prince Edward Island, Canada in May and June 2002. Moisture content was measured with a Delta‐T moisture probe over a 50 m by 25 m grid established on the beach. Measurements of wind speed and direction were made with arrays of cup anemometers and a two‐dimensional sonic anemometer. Transport intensity was measured at a height of 2–4 cm above the bed using omnidirectional saltation probes which count the impact of saltating grains on a piezoelectric crystal. Anemometers and saltation probes were sampled at 1 Hz. Sand transport was measured with vertical integrating sand traps over periods of 10–20 minutes. Results show that where there is a considerable supply of dry sand the saltation system responds very rapidly (1–2 s) to fluctuations in wind speed, i.e. to wind gusts. Where sand supply from the surface is limited by moisture, mean transport rates are much lower and this reflects in both a reduction in the instantaneous transport rate and in a transport system that becomes increasingly intermittent. Threshold wind speed is significantly correlated with an increase in surface moisture content near the upwind end of the beach fetch, but the relationship is not significant at the downwind end where sediment transport is initiated primarily by saltation impact from upwind. Mass flux increases with increasing fetch length and the relationship is described best by a power function. Further work is necessary to develop a theoretical function to predict the increase in transport with fetch distance as well as the critical fetch distance. Copyright © 2007 John Wiley & Sons, Ltd.     

Wind tunnel tests of the dynamic processes that control wind erosion of a sand bed

Aeolian sand transport is a complicated process that is affected by many factors (e.g. wind velocity, sand particle size, surface microtopography). Under different experimental conditions, erosion processes will therefore produce different results. In this study, we conducted a series of wind tunnel experiments across a range of wind velocities capable of entraining sand particles (8.0, 10.0, 12.0, and 14.0 m s^-1) to study the dynamic changes of the shear velocity, aerodynamic roughness length, and sand transport. We found that the shear velocity and aerodynamic roughness length are not constant; rather, they change dynamically over time, and the rules that describe their changes depend on the free-stream air velocity. For wind tunnel experiments without feeding sand into the airflow, the sand bed elevation decreases with increasing erosion time, and this change significantly affected the values of shear velocity and aerodynamic roughness length. A Gaussian distribution function described the relationships between the sand transport rate (q_T) and the duration of wind erosion (T). It is therefore necessary for modelers to consider both deflation of the bed and the time scale used when calculating sand transport or erosion rates. © 2018 John Wiley & Sons, Ltd.     

A stochastic model for initial movement of sand grains by wind

A stochastic model for entrainment of sand grains by wind is presented through analysis of the forces exerted on a single spherical grain, coupled with fluctuations of wind velocity and the change in grain position on the surface. The structure of the stochastic model is consistent with experimental data in the literature. The probability of initial motion increases first, and then decreases, with grain size. It reaches a maximum at diameters of about 0·9 mm. Some sand grains are still in motion at less than the conventional threshold velocity, even at very low velocities. The probability of sand grain movement reaches unity at twice the conventional threshold velocity. Considerable discrepancies amongst conventional threshold formulae may result from the different probabilities of initial movement implied in these formulae. Copyright © 2008 John Wiley & Sons, Ltd.     

Wind tunnel investigation of horizontal and vertical sand fluxes of ascending and descending sand particles in aeolian sand transport

The horizontal and vertical sand mass fluxes in aeolian sand transport are investigated in a wind tunnel by PTV (particle tracking velocimetry). According to the particle velocity and volume fraction of each individual particle from PTV images, the total horizontal sand mass flux, the horizontal mass fluxes of ascending and descending sand particles, and upward and downward vertical sand fluxes are analyzed. The results show that the horizontal mass fluxes of ascending and descending sand particles generally decrease with the increase of height and can be described by an exponential function above about 0.03 m height. At the same friction velocity, the decay heights of the total horizontal sand mass flux and the horizontal mass fluxes of ascending and descending sand particles are very similar. The proportion of horizontal mass flux of ascending sand particles is generally about 0.3–0.42, this means the horizontal mass flux of descending sand particles makes an important contribution to the total horizontal sand mass flux. Both the upward and downward vertical sand mass fluxes generally decrease with height and they are approximately equal at the same height and friction velocity. The relation between upward (or downward) vertical sand mass flux and horizontal sand mass flux can be described by a power function. The present study is used to help understand the transport of ascending and descending sand particles. Copyright © 2016 John Wiley & Sons, Ltd.     

Measurement of particle rotation in a saltation layer

Two computational methods to measure particle rotations from shadow images of sand particles saltating in a wind tunnel are presented. One method calculates the maximum of the cross‐correlations through multiple angular rotations of an imaged particle. The second method polar transforms both images and then calculates the correlation coefficient for multiple pixel displacements in the θ axis, corresponding to particle rotations. The results from both methods were analysed as a function of height above sand bed (3.7–33.4 mm) and particle size (0.32–0.93 mm equivalent mean diameter). Our results indicate little evidence that particle rotation speeds depend on either their size or height above the sand bed. Though similar results were obtained from both methods, there existed different advantages and disadvantages between the methods. Erroneous results likely arose from particles that were inadequately described by a 2‐D rotation axis, or from poorly imaged particles. At a wind tunnel speed of about 12 m/s, most particles rotated at around 300–400 rev/s. Negative rotations were also found, and their proportion was approximately 15% within the total range of ?450 to 850 rev/s. The ratio of displacement kinetic energy to rotation energy was compared across the various groups and had values between 15 and 40. The quotient showed little dependence on height, though decreased with increasing particle size. Wider applicability of the measurement methodology to study snow particle rotation is also discussed. Copyright © 2014 John Wiley & Sons, Ltd.     

Near-bed mass flux profiles in aeolian sand transport: high-resolution measurements in a wind tunnel

Vertical profiles of the streamwise mass flux of blown sand in the near-bed (< 17 mm) region are analysed from high-resolution measurements made using an optical sensor in a wind tunnel. This analysis is complemented by detailed measurements of mass flux and mean velocity profiles throughout the boundary layer depth (0·17 m) using passive, chambered sand traps of small dimensions and armoured thermal anemometers, respectively. The data permit a preliminary analysis of the relations between the observed forms of the profiles of near-bed fluid stress and horizontal mass flux within a carefully conditioned boundary layer. Profiles of mass flux density are found to be characterized by three regions of differing gradient with transitions at about 2 mm and 19 mm above the bed. The exponential decay of mass flux with height is confirmed for elevations above 19 mm, and when plotted as a function of u_*²/g (a parameter of mean vertical trajectory height in saltation), the gradient of mass flux in this region scales with the wake-corrected friction velocity (u), where u > 0·30 m s⁻¹. A separate near-bed region of more intense transport below 19 mm is identified which carries 80 per cent of the total mass flux. This region is evident in some previous field and wind tunnel data but not in profiles simulated by numerical models. Ventilated passive sand traps underestimate mass flux in this region by 37 per cent. At slow or moderate wind speeds a third significant region below 2 mm is observed. These regions are likely to be related to grain populations in successive saltation, low-energy ejections and intermittent bed contact, respectively. Optical measurements reveal locally high grain concentrations at some elevations below 5 mm; these heights scale with transport rate, mass flux gradient and wind speed. Copyright © 1999 John Wiley & Sons, Ltd.     

Bed roughness and grain sorting-an experimental study over fine to medium sand beds

The relation between grain-size distribution of the bed and in suspension was critically examined under a uniform flow velocity of 50 cm/s over two beds: one of mainly fine sands and the other of medium sands. Two sections – one 2.85 m downstream and the other 6.35 m downstream in the experimental channel-were selected for sampling to study the grain-sorting pattern in the vertical direction along the direction of transport. The shape and type of the grain-size distribution pattern were critically studied with height above the bed. The change in the distribution pattern has been attributed to the change of local bed roughness causing scouring against the protruded relatively coarse grains on the bed. Such trends are important to predict the nature of river bed topography. The sand of Bed-1 initially exhibits a log-skew-Laplace distribution at different heights of suspension. The distribution pattern, however, changes but this changing pattern is not consistent along the upstream side. For Bed-2, which initially exhibits a log-normal distribution, the same pattern persists from the height of suspension at 5 cm up to 20 cm. Such consistency in log-normality is also observed at the downstream points of measurement. It is generally expected that the mean grain-size would reduce with increases of suspension height but the results of the experiments, in some occasions, differ significantly from the gradual fining upward trend. This result has been attributed to local changes of bed roughness arising from the protruded relatively coarse grains causing eddies, scouring, and turbulent phenomena which moves coarse particles higher in suspension adding a coarse tail to the distribution increasing the mean grain-size.     

The effect of roughness configuration on velocity profiles in an artificial channel

In order to determine the effect of bed roughness on velocity distribution, we used seven different configurations of bed roughness, with 16 test runs of varying discharge and slope for each configuration. For each run, one-dimensional velocity profiles were measured at 1 cm vertical increments over the crest of the roughness element, and at intervals of 4·25 cm downstream. Results indicate that velocity profile shape remains fairly constant for a given slope and roughness configuration as discharge increases. As slope increases, the profiles become less linear, with a much larger near-bed velocity gradient and a more pronounced velocity peak close to 0·6 flow depth at the measurement point immediately downstream from the roughness element. The zone of large near-bed velocity gradients increases in both length and depth as roughness concentration decreases, up to a length/height ratio of about 9, at which point maximum flow resistance occurs. Longitudinal roughness elements do not create nearly as much flow resistance as do transverse elements. Rates of velocity increase suggest that roughness elements spaced at a length/height ratio of about 9 are most effective at creating flow resistance over a range of discharges in channels with steeper slopes. © 1998 John Wiley & Sons, Ltd.     

Aeolian sediment flux decay: Non-linear behaviour on developing deflation lag surfaces

Wind tunnel simulations of the effect of non-erodible roughness elements on sediment transport show that the flux ratio q/q_s, shear velocity U*, and roughness density λ are co-dependent variables. Initially, the sediment flux is enhanced by kinetic energy retention in relatively elastic collisions that occur at the roughness element surfaces, but at the same time, the rising surface coverage of the immobile elements reduces the probability of grain ejection. A zone of strong shearing stress develops within 0·03 to 0·04 m of the rough bed because of a relative straightening of velocity profiles which are normally convex with saltation drag. This positive influence on fluid entrainment is opposed by declining shear stress partitioned to the sand bed. Similarly, because the free stream velocity U_f is fixed while U_* increases, velocity at height z and particle momentum gain from the airstream decline, leading eventually to lower numbers of particles ejected on average at each impact. When the ratio of the element basal area to frontal area σ is approximately equal to 3·5, secondary flow effects appear to become significant, so that the dimensionless aerodynamic roughness parameter Z₀/h and shear stress on the exposed sand bed T_s decrease. It is at this point that grain supply to the airstream and saltation drag appear to be significantly reduced, thereby intensifying the reduction in U_*. The zone of strong fluid shear near the bed dissipates.     

Grain-size Analysis of Surface Material Under Wind Erosion Using the Effective Surface Concept

The grain mobility and roughness of a surface exposed to wind are dependent on the grain size of the surface particles. This paper deals with the temporal variation in the grain size of surface material using an analytical method based on the effective surface concept. The analysis of grain size data obtained from a wind tunnel experiment indicated that, above the threshold wind friction velocity for all surface particles, the grain-size distribution of surface particles was very similar to that of the parent material over a time period of 10 to 15 minutes. However, the mean grain size of surface particles apparently decreased over the initial time period of 2 to 3 minutes. We therefore confirm earlier studies that on a non-uniform grain bed a larger particle could be more mobile than a smaller particle if the wind friction velocity was higher than the threshold for the larger particle. However, this does not imply that the largest particle is most mobile due to the non-linear dynamics of aeolian sediment transport processes. © 1997 by John Wiley & Sons, Ltd.     

Effect of non‐uniformity of flow on velocity and turbulence intensities over a cobble‐bed

Non‐uniform flows encompassing both accelerating and decelerating flows over a cobble‐bed flume have been experimentally investigated in a flume at a scale of intermediate relative submergence. Measurements of mean longitudinal flow velocity u, and determinations of turbulence intensities u′, v′, w′, and Reynolds shear stress ?u_fw_f have been made. The longitudinal velocity distribution was divided into the inner zone close to the bed and the outer zone far from the bed. In the inner zone of the boundary layer (near the bed) the velocity profile closely followed the ‘Log Law’; however, in the outer zone the velocity distribution deviated from the Log Law consistently for both accelerating and decelerating flows and the changes in bed slopes ranging from ?2% to + 2% had no considerable effect on the outer zone. For a constant bed slope (S = ±0·015), the larger the flow rate, the smaller the turbulence intensities. However, no detectable pattern has been observed for u′, v′ and w′ distributions near the bed. Likewise, for a constant flow rate (Q = 0·040 m³/s), with variation in bed slope the longitudinal turbulent intensity profile in the longitudinal direction remained concave for both accelerating and decelerating flows; whereas vertical turbulent intensity (w′) profile presented no specific form. The results reveal that the positions of maximum values of turbulence intensities and the Reynolds shear stress depend not only on the flow structure (accelerating or decelerating) but also on the intermediate relative submergence scale. Copyright © 2009 John Wiley & Sons, Ltd.     

The role of wind direction in eolian transport on a narrow sandy beach

Comparison of eolian transport during five high-velocity wind events over a 29 day period on a narrow estuarine beach in Delaware Bay, New Jersey, USA, reveals the temporal variability of transport, due to changes in direction of wind approach. Mean wind speed measured 6 m above the dune crest for the five events ranged from 8·5 to 15·9 ms^?1. Mean wind direction was oblique to the shoreline (63° from shore-normal) during one event but was within 14° of shore-normal during the other events. Eolian transport is greatest during low tide and rising tide, when the beach source area is widest and when drying of surface sediments occurs. The quantity of sediment caught in a vertical trap for the five events varied from a total of 0·07 to 113·73 kgm^?1. Differences in temperature, relative humidity and moisture and salt content of surficial sediments were slight. Mean grain sizes ranged from 0·33 to 0·58 mm, causing slight differences in threshold shear velocity, but shear velocities exceeded the threshold required for transport during all events. Beach width, measured normal to the shoreline, varied from 15·5 to 18·0 m; beach slope differed by 0·5°. The oblique wind during one event created a source width nearly double the width during other days. Beach slope, measured in the direction of the wind, was less than half as steep as the slope measured normal to the shoreline. The amount of sand trapped during the oblique wind was over 20 times greater than any other event, even those with higher shear velocities. The ability of the beach surface to supply grains to the air stream is limited on narrow beaches, but increased source width, due to oblique wind approach, can partially overcome limitations of surface conditions on the beach.     

Transient sand transport rates after wind tunnel start-up

Wind tunnel experiments were conducted with a well mixed, flat sand bed, 5·7 m in length, to study the initial sand flux response at three different shear velocities. In some experiments, the bed was allowed to deplete without replenishment; in others, sand was fed 10·8 m upstream of the monitored cross-section. The results indicated that the transport rate increases rapidly during the first minute, and then adjusts slowly towards a steady rate. The time to reach such an equilibrium was observed to be on the order of 2–4 min in non-fed experiments and on the order of 8–9 min in fed experiments. Many factors may affect such development and bring about non-stationarity in total sand transport rate. Among these factors are differences in the natural composition of the sand bed, changes in both the topographical features of the sand bed (ripples) and its surface texture, and any artificial features that influence the adjustment between the boundary layer profile and the sand load on the wind. A useful key to the influence of each factor is obtained by noting that each has a typical and distinct ‘time constant’. The nature and relative importance of each is discussed by reference to the reported wind tunnel experiments and to the behaviour of saltation cloud numerical models. © 1998 John Wiley & Sons, Ltd.     

Collisions of quartz grains with a sand bed: The influence of incident angle

The velocities with which grains were observed to emerge from a sand bed after an intersaltation collision at u_* = 40 cm s^?1 are reported for four bed attitudes, from horizontal bed to adverse bed slope 15°. The principal effect of bed angle is to alter the magnitude and direction of the ricochet velocity. However, emergent velocities of dislodged grains are consistent with reptation path lengths comparable to the length of the upwind face of ripples in the corresponding wind. Calculations of the loss of forward momentum at collision, using the data for the range of bed attitudes studied suggest that creep is most vigorous on the sloping upwind face of the ripple and diminishes at the crest. As a result, the crest would be expected to accumulate the coarse material which moves predominantly by creep. The saltations originating in ricochet from the sloping back of the ripple are more vigorous and more concentrated in plan than are those originating at the crest. However, the saltation path length is at least an order of magnitude greater than the ripple wavelength and the probability distribution of path lengths is quite dispersed. Consequently it is very unlikely that these spatial patterns of ricochet are preserved sufficiently distinctly in the saltation cloud and subsequent collision distribution to be the agent of ripple development. This study therefore supports a view of moving grain interaction with the bed in which saltation provides the power to mobilize grains but ripple growth is associated with reptation and particularly with a pattern of impact which develops with the bed relief. Creep is more active on upwind facing slopes than at the crest, which therefore is a zone of net creep grain deposition.