Deposition of climbing-ripple beds: a flume simulation

Transcription

1 Sedimentology (1982) 29, Deposition of climbing-ripple beds: a flume simulation GAIL M. ASHLEY*, JOHN B. SOUTHARDt and JON C. BOOTHROYDS * Department of Geological Sciences, Rutgers-The State University of New Jersey, New Brunswick, New Jersey 08903, U.S.A., 7 Department of Earth and Planetary Sciences, Massachusetts Institute of Technology, Cambridge, Massachusetts , U.S. A., and $ Department of Geology, University of Rhode Island, Kingston, Rhode Island 02881, U.S.A. ABSTRACT Thirteen runs were made in a small recirculating flume to simulate the deposition of the climbingripple sequences commonly present in fine-grained facies of fluvial and deltaic deposits. These sequences consist of intergradational climbing-ripple cross laminae and draped laminae. The experiments were based on the assumption that stratification type depends mainly on near-bottom flow structure and uniform sediment fallout from an overloaded flow. Various combinations of curves of velocity versus time and of sediment feed versus time in runs lasting from 45 to 840 min were used in an exploratory program; conditions for each run were selected on the basis of experience in previous runs. The runs verified that Type A (erosional-stoss) climbing ripples are produced by aggradation rates that are small relative to ripple migration rate, and Type B (depositional-stoss) climbing ripples are produced by aggradation rates that are large relative to ripple migration rate. Draped lamination results from continued fallout of sediment from suspension after ripple migration ceases or almost ceases. Comparison of geometric details of the ripple stratification produced in the flume runs with that in natural sequences, supplemented by considerations on maximum and minimum migration rates of ripples, suggests times of no more than a few tens of hours for the deposition of the climbing-ripple portions of sequences cm thick. Runs in which deposition of a 20 cm sequence took more than 10 h produced such atypical features of ripple geometry as sharp crests, planar lee-side laminae, and angular toeset-foreset contacts. INTRODUCTION Multiple climbing-ripple or ripple-drift beds are common in fine-grained fluvial and deltaic deposits (Fig. 1). These beds consist of combinations of three kinds of laminae: (1) climbing-ripple cross laminae; (2) sinuous, parallel or congruent laminae, hereafter called draped lamination (Gustavson, Ashley & Boothroyd, 1975): and (3) in some cases, planar horizontal laminae. Angle of climb typically varies from fairly small, or even zero, to very steep within a single continuous unit, depending upon the nature of the upward transitional contact: climbing ripples to draped lamination, climbing ripples to planar lamination, draped lamination to climbing ripples, or planar lamination to climbing ripples. Field /82/ $ International Association of Sedimentologists descriptions from rivers (McKee, 1938, 1965 ; Davies, 1966; Boothroyd &Ashley, 1975) and deltas (McKee, 1939; Jopling &Walker, 1968;Aario, 1972; Gustavson et al., 1975) have led to a multiplicity of terms for such structures (Table 1) and some lack of agreement on their hydraulic and sedimentologic interpretation. Attempts to simulate even portions of such climbing-ripple units in laboratory channels (Jopling, 1960; McKee, 1965) were rudimentary prior to the work of Allen (1971a). In Allen s experiments, uniform thin sequences of realistic climbing-ripple stratification were deposited under steady conditions of flow and sediment fallout to produce a wide range of climb angles. Our experiments were designed to take such work a step further by generating thick climbing-ripple sequences while flow and sediment 5-2

2 68 G. M. Ashley, J. B. Soictliard and J. C. Boothroyd Fig. 1. Integradational packages of ciimbing-ripple lamination and draped lamination exposed in a glaciolacustrine delta. fallout varied with time, in order to simulate the deposition of natural sequences showing characteristic vertical variations in angle of climb and type of lamination. Of special interest in the experiments was the transition from climbing-ripple cross-lamination to draped lamination, because there have been widely divergent views on the origin of draped lamination. Jopling & Walker (1968) and Banerjee (1977) believed that such lamination is formed at low flow velocities (10 cm sec-l) as a kind of ripple structure, and the former consider sediment cohesion to be a major factor. Allen (1963), McKee (1965), Gustavson et al. (1975), and Hunter (1977) suggested that it results from deposition of a sediment blanket from suspension over inactive ripples. The goals of this study were to: (1) reproduce some of the characteristic vertical successions of structures found in natural climbing-ripple sequences, and (2) put constraints on parameters (mainly current velocity, rate of aggradation, and time) important in determining the nature of the structures. EXPERIMENTS Ripples climb at some angle 6 whose tangent is V,/ V,, where V, is the net rate of aggradation of the bed and V, is the downstream migration rate of the ripples (Fig. 2, Allen, 1970). Any number of combinations of V, and V, correspond to a given B so that 6 reflects relative rates rather than absolute rates. Allen (1973) drew a useful distinction between type of climbing-ripple cross-lamination, based largely on an erosional or depositional relationship between sets of laminae in a limited vertical interval of the cross-laminated bed, and pattern, based on vertical change in climb angle and grain size within the bed. Following Jopling & Walker (1968), Allen distinguished between Type A, or erosional-stoss, and Type B, or depositional-stoss, climbing-ripple cross-lamination. Also, he formally recognized three categories of pattern-climb angle constant, increasing upward, and decreasing upward-and left a fourth category to include all the more complicated

3 = Increasing Suspended Load Increasing Bed Load L Deposition of climbing rippie beds 69 Draped Lamina tion Fig. 2. The type of sedimentary structure produced is related to the relative importance of rate of ripple migration and the rate of bed aggradation. The ripples climb at some angle 0 whose tangent is the mean aggradation rate V, divided by the downstream migration rate V,. cases. Our runs produced a variety of patterns of vertical variation in climb angle. The key quantities that must be varied to produce the spectrum of climbing-ripple types and patterns (assuming lower-flow-regime conditions and appropriate grain size) are aggradation rate and ripple migration rate. Ripple migration rate can be varied simply by varying the current strength, but as discussed below, simulation of aggradation rate is less straightforward. Climbing-ripple structures are developed when bed-load transport rate over migrating ripples decreases downstream (sediment storage then resulting from spatial change in sediment transport), or when the ripples receive sediment from suspension (storage then resulting from either temporal or spatial change in suspended-load transport, spatial change certainly being the more important). Allen (1970, 1971a) presents one of the few explicit discussions of the consequences of mass balance in sediment transport over ripples. The steep angles of climb commonly found in climbing-ripple sequences can originate only by fallout from suspension. In order to produce steep climb by downstream decrease in bed-load transport, the current strength would have to decrease so rapidly downstream that the ripple train could include only a small number of ripples, whereas climbing-ripple sequences commonly show great streamwise uniformity. In contrast, where suspended-sediment concentrations are high, deposition involving even a small percentage decrease in concentration downstream can lead to large 0 if V, is not great. We therefore used a reach of uniform flow, many ripple spacings long, into which sediment was showered to produce net deposition from suspension. This is a good model for deposition by slow downstream decrease in suspended-sediment concentration,except that the inevitable gradual downstream changes in structures, bed thickness, and grain size cannot be studied. In any unidirectional-flow system, flow strength and local excess in suspended sediment, which together largely govern V, and V, and therefore the nature of the climbing-ripple structures, vary with I time in a complicated way that is somehow dependent upon the overall hydraulics of the system. To the extent that this time variation is the result of some well-defined event like a river flood or a turbidity current, the climbing-ripple structures should tend to show a characteristic vertical sequence, which will, however, vary greatly in detail depending on the particular locale and event. Our experimental arrangement was designed to afford complete uncoupling of migration rate, thereby allowing complete freedom to generate arbitrary patterns of variation of these two variables and thus to explore the range of combinations needed to simulate climbingripple sequences. The danger in this approach is that some combinations that would reproduce a given sequence just as well as others might be overlooked. However, the alternative-a survey of stratification types developed from all possible combinations of time variations in V, and V,-would be an impossible task. We therefore limited ourselves to searching for combinations that seem to account well for some distinctive natural climbing-ripple sequences. We made thirteen runs in a small recirculating flume to produce climbing-ripple deposits each about 20 cm thick (Southard, Ashley & Boothroyd, 1972). For each experiment the following conditions were selected: (1) duration of run, (2) curve of velocity versus time, and (3) curve of sediment-feed rate versus time. Each run was planned based on the results of the previous ones. In the experiments, the following quantities were independent, in that they were imposed upon the system and characterized all aspects of behaviour of flow and sediment :flow velocity, flow depth, sediment size, and aggradation rate. Of these, only velocity and aggradation rate were varied. All variables describing sediment transport and bed forms, as well as the resulting sedimentary structures, were dependent variables. The thirteen runs ranged from 30 to 840 min long; most were min long. Velocity variations con-

4 70 G. M. Ashley, J. B. Southard and J. C. Boothroyd sisted of linear deceleration curves, and symmetrical and asymmetrical acceleration-deceleration curves, in addition to various constant velocities. Sedimentfeed schedules included various constant rates ( cms min-1 of bulk sediment), linearly decreasing rates, and symmetrical and asymmetrical curves of increasing and then decreasing rates. Fig. 3 summarizes the variables for each run. The runs were made in a tilting rectangular open channel 17 cm wide, 30 cm deep, and about 6 m long (Southard & Boguchwal, 1973, fig. 2). The flow passed without a free overfall from the channel into a tailbox, and was pumped continuously through a return pipe to the upstream end of the channel. The slope could be changed by a support jack. A grid of plastic tubes at the head of channel straightened the flow. Discharge was measured with an orifice meter in the return pipe. Ideally, sediment feed should have been continuous and uniform over the entire bed surface. Because this is difficult to arrange, we used a feed hopper that traversed the entire length of the channel periodically at constant speed while dropping into the flow a curtain of sediment extending from one wall to the other. Rate of deposition was thus steady and uniform on the average, but any point on the bed experienced unsteady deposition. In each run, sediment-feed passes were made always with the same volume of sediment, from 350 to 1000cm3; average rate of deposition was varied by changing the time interval between passes. The passes were made according to a predetermined schedule at intervals from 12 sec to a few minutes, to produce the desired time history of net aggradation rate. The sediment was a well-sorted quartz sand (median size 0.15 mm) containing a very small percentage of flaked graphite to make laminations distinct. When the quartz sediment settled out after a pulse of sediment addition, the graphite, with its lower settling velocity, became concentrated in the ripple troughs, and on stoss surfaces when depositional. To vary current velocity a schedule of manometer settings for the corresponding discharges (assuming 9cm water depth) was drawn up. Discharge was adjusted at intervals of min. A separate schedule was used for addition of water to the system, to make up for pore space of the deposited sediment. Every 5-10min flow depth was checked approximately through the channel sidewall by measuring downward from the water surface to an estimated average bed surface, and the pre-established volumes VELOCITY (CmIseC) YS. TIME (mid z:lw 401 Fig. 3. A summary of the variables (velocity versus time and aggradation rate versus time) in the 13 flume runs. of water to be added were corrected accordingly. Flow velocity, however, was not monitored directly. Since water depth probably varied by ko.5 cm during the runs, the patterns of velocity variation shown in Fig. 3 are accurate only to within about 10 %. As the deposit developed during each run, the reach of uniform flow over fully developed ripples became shorter, because at the upstream end the bed tapered upward from the headbox and at the downstream end there was a progressively longer avalanche face leading into the tailbox. But even at the end of the runs there was a reach at least 4 m long of uniform flow and deposit. Only the deposits in that reach were studied. Time-lapse motion pictures were taken through a sidewall during a few of the runs, and a complete photographic record in the form of an overlapped composite picture of the entire deposit, through one or both of the sidewalls, was taken at the end of each run.

5 Deposition of climbing-ripple beds 71 Fig. 4. (A) Flume run 8. The run was designed to last 200 min with an asymmetrical velocity curve (maximum velocity = 25 cm sec-') and a symmetrical aggradation-rate curve; see Fig. 3. Total accumulation was 18 cni. Flow was from left to right. Starting with a train of ripples that had reached equilibrium with an earlier, stronger flow, deposition began (at arrow) with draped lamination (DL) followed by Type A (erosional-stoss) cross-lamination (A), Type B (depositional stoss) cross-lamination (B), and a final blanke: of draped lamination. (B) A climbing ripple sequence exposed in a glaciolacustrine delta (Bennett's Brook Delta, glacial Lake Hitchcock, Massachusetts, USA) exhibits a sequence of sedimentary structures similar to that produced in run 8. Flow was from left to right. The sequence begins by deposition over draped lamination (at arrow). Type A climbing-ripple crosslamination grades into Type B and finally into draped lamination at the top.

6 72 G. M. Ashley, J. B. Southard and J. C. Boothroyd RESULTS All runs produced draped lamination and at least one type of climbing-ripple lamination. The vertical succession of lamination types varied depending on the patterns of variation in current velocity and aggradation rate. At our chosen flow depth of 9 crn, ripples were active at current velocities ranging from 15 cm sec-l. At an aggradation rate of 9 cm h-i, about the average for the thirteen runs, Type A (erosional-stoss) climbing-ripple cross-lamination generally formed at velocities greater than 25 cm sec-', B (depositional-stoss) climbing-ripple cross- lamination developed at velocities from 15 to 25 cm sec-l, and draped lamination accreted at velocities below 15 cm sec-l. In themselves, these numbers are of no special significance ; obviously, increasing the aggradation rate raises the velocity envelopes for each of the three structure types, and reducing the aggradation rate lowers them. Only runs 8, 11, and 4, selected as being representative, are discussed in detail. Run 8 (Figs 3 and 4A), lasting 200min, was designed to simulate a waxing and waning flow. The velocity curve is asymmetrical, starting at 5 cm sec-l, increasing to 25 cm sec-' by 60 min, and then decreasing slowly ~ Fig. 5. (A). Flume run 11. This run was designed to last 45 min with linearly decreasing velocity (60 cm sec-l to zero) and linearly decreasing aggradation rate (50 cm h-' to zero). Flow was from left to right. Deposition began at the base with planar lamination, followed by Type A and then Type B climbing ripple cross-lamination, and draped lamination at the top. (B). A climbing-ripple sequence exposed in a glaciolacustrine delta (glacial Lake Hitchcock, Massachusetts, U.S.A.), showing a sequence similar to that produced under waning flow in run 11. Flow was from left to right.

7 Deposition of climbing-ripple beds 73 Fig. 6. Flume run 4. This run was designed to last 100 min, with a maximum velocity of 38 cm sec-' and a maximum aggradation rate of 25 cm 11-l, both occurring midway through the run (Fig. 3). Flow was from left to right. The sequence began with draped lamination, followed by Type B and Type A climbing-ripple cross-lamination. At that time insufficient water was added, causing an unplanned decrease in flow depth and deposition of parallel laminae under upper-flow-regime conditions at the point shown by the arrow. Subsequent addition of water increased the flow depth to the desired value, and Type A structures were again produced under lower-flow-regime conditions. A steady decrease in velocity at the end of the run produced Type B structures and a final blanket of draped lamination. to 15 cm sec-'. The sediment feed curve is symmetrical, starting at zero and increasing to a maximum halfway through the run, with an average aggradation rate of 6 cm h-l. Prior to the run the flume was operated for 2 h, until the train of ripples had reached equilibrium. Sedimentation began with a thin unit of draped lamination over these ripples, followed abruptly by Type A climbing ripples at the peak velocity. While velocity waned, Type B ripple drift developed, followed by a final blanket of draped lamination. The climbing-ripple sequence is remarkably similar to many natural deposits (Fig. 4B). Run 11 (Figs 3 and 5A) was planned to start under high-velocity flat-bed conditions and last 45 min. Velocity was decreased linearly from 60 to locm sec-l ; aggradation rate also was decreased linearly from 50 cm h-' to zero. After a few centimetres of flat-bed deposition, ripples formed, small and fastmoving at first, but becoming larger and slower. The climbing-ripple cross-laminae passed upward from low-angle Type A to high-angle Type B and then into draped laminae. The sequence deposited in run 11 under a waning flow, although qualitatively different from that in run 8, is also similar to many natural deposits (Fig. 5B). Run 4 (Figs 3 and 6) was planned to last 100 min and have a peak velocity of 38 cm sec-l and a maximum aggradation rate of 25 cm h-l midway through the run. However, in the middle of the run the flow was inadvertently allowed to pass briefly into upperflat-bed conditions, at flow velocities estimated to have been cni sec-l, because flow depth decreased to 6.5 cm at that time, when flow depth could not be monitored adequately. As the flow velocity increased, the angle of climb became very small as the ripples migrated faster and faster. At the highest flow velocities, for a brief time, the ripples were effaced and the flow was in the flat-bed regime. Ripples reappeared as flow velocity decreased, and more Type A climbing-ripple structure was formed, at first with a small angle of climb. Since flat-bed conditions persisted for only a few minutes, there is only a thin layer with horizontal lamination. Although difficult to trace, the transition between low-

8 74 G. M. Ashley, J. B. Southard and J. C. Boothroyd Table 1. Terms in the literature Parallel or congruent laminae Climbing ripple (cross) laminae. Ripple laminae-in-phase (McKee, 1939, 1965) Ripple drift (Sorby, 1859) Sinusoidal lamination (Jopling & Walker, 1968) Ripple-drift cross-lamination (Walker, 1963, 1969) Draped lamination (Gustavson et al. 1975) Climbing ripple structure (McKee, 1965) Rippleform lamination (Hunter, 1977) Climbing ripple cross-lamination (Allen, 1971a, b, 1973) Climbing translatent structure (Hunter, 1977)

9 Deposition of climbing-ripple beds 75 Fig. 7. A comparison of ripple morphology from the field with two examples of similar thickness (18 cm) produced at different time-scales in the flume. Flow is from left to right in each example. (A) Climbing-ripple cross-lamination from run 6. Deposition of 18 cm was stretched out over 500 min. Ripple crests are sharp, lee-side laminae are planar, and toeset-foreset contacts are angular. (B) Climbing-ripple lamination from run 8. Deposition of 18 cm took 200 min. Ripples have rounded crests and toeset-foreset contacts similar to those from the field. (C) Climbing-ripple cross-lamination from the field. Sediments were deposited on the prodelta slope of a glaciolacustrine delta. angle climbing ripples and horizontal lamination seems abrupt, with a non-zero angle of climb at the transition. In order to examine the factor of time, two long runs were made: run 6,500 min, and run 13,840 min. The rationale was to use a lower aggradation rate and a lower current velocity for a longer time to produce a deposit with the same thickness and qualitatively similar succession of stratification types, and then compare the details of the structures with those in the shorter runs. In run 6 the velocity curve was symmetrical, varying from 5 to 25 cm sec-l and back to 5 cm sec-l. Sediment-feed rates were slow (1-5 cm h-l) and symmetrical (Fig. 3). Sedimentation started as draped lamination, flattened out through Type B into Type A midway through the run (at the time of maximum velocity), and then passed symmetrically back to vertical. In detail, ripple geometry and stratification were different (Fig. 7A) from those in the shorter runs (Fig. 78): ripple crests were sharp, lee-side stratification was planar, and foreset-toeset contacts were angular. These details of ripple geometry are substantially different from those in naturally deposited sequences showing the same overall succession of climbing-ripple types and climb angles. The details of ripple geometry in the shorter runs (Figs 4A and 5A) resembled the field occurrences (Fig. 7C) more closely. Although we made no special measurements, ripple trains seemed to show about as much variability in geometry and migration rate during the runs, while conditions of flow and aggradation were changing, as during equilibrium migration of ripples in the absence of sediment feed. Because of this variability, and the attendant births and deaths of ripples, seldom could a single ripple be traced throughout the entire vertical sequence of climbing-ripple lamination produced in a run. In this respect the flume deposits are like many natural climbing-ripple sequences that involve fairly large and irregular ripples. If the runs had started with regular trains of small and newly developed ripples and ended before ripples had developed to the point of being irregular, then the deposits would probably have resembled the many natural examples of extremely regular climbing-ripple sequences.

10 76 G. M. Ashley, J. B. Southard and J. C. Boothroyd DISCUSSION The extremely shallow flow depth and fairly small channel width do not make the experiments a very close replica of what we envisage to be the real depositional situations. The validity of the simulation was based on the assumption that ripple processes depend mainly on near-bottom flow and uniform sediment fallout rate from an overloaded flow. This view is substantiated by: (1) the weak dependence of ripple characteristics on flow depth in flume experiments generally, provided only that depth is greater than about half the average ripple spacing; and (2) the general similarity between current ripples in laboratory channels and those in natural unidirectional flows, except for some differences in regularity of plan pattern by effects of flow depth and/or widthto-depth ratio (Allen, 1969,1977; Banks & Collinson, 1975). These differences in plan configuration would have a substantial effect on structures in the experimental deposits as viewed in spanwise section but only a minor effect on streamwise section. An obviously unrealistic element of the experiments was the unchanging grain size of the sand introduced into the flow: in natural climbing-ripple sequences at least some correlation between current velocity and grain size of the settling sediment should be expected. The monumental extra effort involved in varying sediment size would have drastically reduced the number of runs. But since size, shape, and migration rate of ripples are not strongly dependent on sediment size, this deficiency introduced at worst only some quantitative distortion in the details of the structures. All sedimentation was under lower-flow-regime conditions. Flow must be fast enough (greater than about 15 cm sec-l at these flow depths) for ripple migration, but not so high (greater than about 50 cm sec-i) that dunes or flat-bed conditions develop. Higher velocities and lower aggradation rates produce Type A (erosional-stoss) climbing-ripple lamination, whereas lower velocities and higher aggradation rates produce Type B (depositionalstoss) climbing-ripple lamination. Draped lamination develops only by deposition from suspension on an inactive rippled bed. It is seen by comparison of the flume runs (Figs 4A and 5A) with field examples (Figs 4B and 5B) that our experiments simulate many natural climbingripple sequences closely. However, there obviously are a great many unused combinations of curves of aggradation rate versus time and of flow velocity versus time that would have produced similar sequences in general if not in specific. Allen (1970, 1971a) developed a model for vertical variation in angle of climb in climbing-ripple sequences deposited by unsteady and non-uniform flows in which spatial and temporal changes in flow are sufficiently small that Bediment transport is in equilibrium with the flow. Srodon (1974) broadened these results by considering other effects on vertical pattern: independent change in grain size; development of ripples to equilibrium ; and limited sediment availability. Although our experiments were not designed to test the patterns predicted by Allen s theory, the results are qualitatively in agreement, in that reasonable spatial and temporal variations in flow velocity can be used in the theory to produce patterns qualitatively similar to those in our runs. However, this similarity seems not to be especially significant. It seems likely that in many cases of climbing-ripple deposition (for example, in front of glacial-lake deltas, or behind levees or in certain other overbank areas in rivers) the flow, having just experienced sharp convective deceleration, must be strangely overloaded with suspended fine sand or coarse silt. Grain size and net fallout rate would then not be rigorously correlated with local flow velocity, and fallout rate might indeed depend only weakly on velocity, so that vertical pattern of climb angle might depend in great part only on temporal changes, random or systematic, in flow velocity. This effect would act concurrently with those analysed by Allen, and might even be predominant. Such a model is equally consistent with our experiments, in which time variation in sediment fallout rate was purposely uncoupled from that of flow velocity. In fact, patterns of climb angle closely similar to those in the runs, which in turn correspond closely to patterns common in glacial-lake delta foreset deposits, could have been obtained equally well by varying sediment feed rates even less and magnifying the changes in flow velocity. Therefore, in accounting for at least some fluvial and glaciolacustrine climbing-ripple patterns, it seems more realistic to appeal to disequilibrium effects related to suspended-sediment overloading than to an eqyilibrium-flow model. In any case, we concur with Srodon s (1974) view that attempts to deduce conditions of flow and sediment transport from observed vertical patterns in climbingripple beds are largely unrewarding, because of the multiplicity of effects that can combine to produce closely similar vertical patterns of variation in climb angle and grain size.

11 Deposition of climbing-ripple beds 71 The time involved in deposition of natural climbingripple sequences is an intriguing question. Are sequences like those in Figs 4B and 5B deposited in hours, tens of hours, hundreds of hours, or even thousands of hours? The experiments provide two tentative lines of evidence. The first involves ripple profiles and the general configuration of the climbingripple structures. Whereas angle of climb depends on both current velocity and aggradation rate, geometry of the ripples themselves depends mainly on current velocity, leaving aside the effects of flow width and flow depth. A comparison of structural details of the ripples in the natural sequences (Fig. 7C) with those deposited in the flume at a variety of time-scales (Fig. 7A, B) but having the same overall structural pattern reveals a relationship between time-scale and stratification. Stratification in one of the slowest runs (run 6), in which 20 cm of sediment was deposited in 500 min, definitely looks less natural than stratification in the shorter runs; foreset-toeset contacts are angular; crests, where preserved, are sharp; and leeside and stoss-side laminae are largely planar. The presence of this atypical stratification in the longer runs suggests time-scales of less than ten hours for deposition of the climbing-ripple part of natural sequences like those shown in Figs 4A and 5A. It could be contended that the 'unnatural' details of stratification in run 6 are an artifact of the sedimentfeed system. The episodic sedimentation must indeed have affected the details of the stratification to some extent. However, the effects on ripple geometry and behaviour seem to have been minor. Without a means of continuous sediment feed we could not sort out possible effects of episodic feed from inevitable changes in ripple geometry and migration as a function of aggradation rate alone. But some insight was gained from a preliminary run in which sediment was fed slowly but episodically on to ripples previously at equilibrium with the flow. Since aggradation rate was very low, if the method of feed was of no effect, the ripples should have been closely similar to those at zero aggradation rate, and this was in fact the case. Also, structures developed at the sidewalls, where current strength is lower, might show lower velocity features than near the channel centreline. If this effect were important, the truly representative sedimentary structures produced in the runs might not have an atypical appearance even for deposition times considerably longer than those of our longest runs. But sectioning of the deposits along the centrelineafter these longer runs revealed that the structures in the interior of the deposits are in fact qualitatively similar to those displayed at the sidewalls. Another indication of time can be gained by considering migration rates of ripples. The maximum migration rate of ripples in a given sand size under conditions just short of the transition to dunes (or to flat bed at shallow flow depths, as in our runs, or in very fine sediment sizes), together with assumptions about angle of climb, places lower limits on time for deposition of a climbing-ripple sequence. Assuming a maximum ripple migration rate of the order of 1 cm sec-l in this sand (a crude value based only on experiments, but certainly a good order of magnitude estimate consistent with systematic observations by Dillo, 1960) and a very steep angle of climb (e.g. 45") a 20 cm deposit could be formed in less than a minute, provided that at such an astronomical feed rate the ripples would probably look very unusual. At more reasonable angles of climb, say less than lo", a deposit with the same thickness could be built in something like a few tens of minutes by ripples migrating at the maximum rate. Such conditions are closely similar to those that existed for a brief time near the middle of run 4, and the stratification would presumably resemble the low-angle Type A climbing ripples seen in that deposit. At the other extreme, ripple migration rate presumably approaches zero smoothly with decreasing current velocity, although we know of no systematic supporting data. In any case, ripples in fine sand can migrate extremely slowly at equilibrium with the current, at rates certainly as low as centimetres per day. Under such conditions, deposition of climbing-ripple sequences even with steep angles of climb would involve very long times: again assuming 45" climb, ripples migrating at lop5 cm sec-l would build a 20 cm deposit in about 106 sec, or about ten days-and this is certainly not the maximum possible figure. But for ripples to remain active during deposition at such slow rates, current velocity would have to remain within an extremely narrow range. Considering the variability usually observed within natural climbing-ripple sequences and their common occurrence in a wide variety of settings, such a restricted range of current velocities seem very unlikely. It therefore seems likely that single climbingripple sequences less than a few tens of centimetres thick are commonly deposited in less than a few tens of hours. These qualitative results on time-scales are broadly consistent with earlier estimates by Kuenen (1967) and Allen (1971b) based on somewhat different considerations.

12 78 G. M. Ashley, J. B. Southard and J. C. Boothroyd Of course, such time considerations can apply only to the part of the deposit generated by migrating ripples; the draped part can be deposited much more slowly, over tens of hours or even much longer (many days or even many weeks). Moreover, if there is degradation or zero net aggradation for an appreciable time during ripple migration, an extensive discontinuity would be evident in the climbing- ripple structures, as is commonly observed in natural climbing-ripple sequences. Obviously, the time represented by such a hiatus cannot be ascertained. ACKNOWLEDGMENTS The flume runs described in this paper were made in the Laboratory for Experimental Sedimentology (Massachusetts Institute of Technology) from May 1972 to January 1973 with financial support from the Petroleum Research Fund of the American Chemical Society (PRF no AC2). William M. Galen provided able and valuable assistance in making the flume runs. The photographic work by Paulette Hervi- Hughes is much appreciated. CONCLUSIONS Deposition of climbing-ripple sequences is easily simulated in a flume. Several characteristic patterns of vertical variation in structures found in nature were duplicated by varying current velocity and aggradation rate, and also the duration of the runs. The main conclusions of this study are the following. (1) The experiments verified directly the obvious conclusion that Type A (erosional-stoss) climbingripple cross-lamination is produced under combinations of relatively low aggradation rate and high ripple migration rate, and that Type B (depositionalstoss) climbing-ripple cross-lamination is produced under conditions of relatively high aggradation rate and low ripple migration rate. In addition, the runs showed that draped lamination results from continued fallout of sediment from suspension after cessation of ripple migration, as proposed by Allen (1963), McKee (1965), Gustavson et al. (1975), and Hunter (1977). No effects of cohesive sediments need be invoked, as believed by Jopling & Walker (1968). (2) Climbing-ripple sequences about 20 cm thick that most closely resemble characteristic natural sequences are deposited in h, at flow velocities of 154Ocm sec-l and aggradation rates of 5-1 5cm h-l. (3) Constraints placed on rates of aggradation of realistic-looking structures by what is known or can be assumed about migration rates of ripples suggests that deposition of natural climbing-ripple sequences like those in Figs 4A and SA, a few tens of centimetres thick, is unlikely to take longer than a few tens of hours. Comparison of geometric details of the ripple stratification in the experiments at various time-scales (Fig. 7A, B) with those in naturally deposited sequences (Fig. 7C) suggests even shorter times for deposition, less than 10 h. REFERENCES AARIO, R. (1972) Associations of bed forms and paleocurrent patterns in an esker delta, Haapajarri, Finland. Ann. Acad. Scientiarum Fennicae, Ser. A, Part 111, No. 111,55 pp. ALLEN, J.R.L. (1963) Asymmetrical ripple marks and the origin of water-laid cosets of cross-strata. Lpool. Manchr. geol. J. 3, ALLEN, J.R.L. (1969) On the geometry of current ripples in relation to the stability of fluid flow. Geogr. Annlr 51 A, ALLEN, J.R.L. (1970) A quantitative model of climbing ripples and their cross-laminated deposits. Sediment- OlOgy, 14, ALLEN, J.R.L. (1971a) A theoretical and experimental study of climbing ripple cross lamination, with a field application to the Uppsala Esker. Geogr. Annlr 53 A, ALLEN, J.R.L. (1971b) Instantaneous sediment deposition rates deduced from climbing-ripple cross-lamination. J. geol. SOC. London, 127, ALLEN, J.R.L. (1973) A classification of climbing ripple cross-lamination. J. geol. Soc. London, 129, ALLEN, J.R.L. (1977) The plan shape of current ripples in relation to flow conditions. Sedimentology, 24, BANERJEE, I. (1977) Experimental study on the effect of deceleration on the vertical sequence of sedimentary structures in silty sediments. J. sedim. Petrol. 47, BANKS, N.L. & COLLINSON, J.D (1975) The size and shape of small scale current ripples: an experimental study using medium sand. Sedimentology, 22, BOOTHROYD, J.C. & ASHLEY, G.M. (1975) Processes, bar morpholcgy, and sedimentary structures in braided outwash fans, northeast Gulf of Alaska. In: Glacioffuviaf and Glaciolacustrine Sedimentat ion (Ed. by B.C. MacDonald and A.V. Jopling). SOC. econ. Paleont. Miner., Tulsa, 23, DAVIES, D.K. (1966) Sedimentary structures and subfacies of a Mississippi River point bar. J. Geol. 74, DILLO, H.G. (1960) Sandwanderung in Tideflussen. Technische Hochschule Hannover, Franzius-Institut fur Grund- und Wasserbau, Mitteilungen, 17,

13 Deposition of climbing-ripple beds 79 GUSTAVSON, T.C., ASHLEY, G.M. & BOOTHROYD, J.C. (1975) Depositional sequences in glaciolacustrine deltas. In : Glaciofluvial and Glaciolacustrine Sedimentation (Ed. by B.C. MacDonald and A.V. Jopling). Soc. econ. Paleont. Miner., Tulsa, 23, HUNTER, R.E. (1977) Terminology of cross stratified sedimentary layers and climbing ripple structures. J. sedim. Petrof. 47, JOPLING, A.V. (1960) An experimental study on the mechanics of bedding. Unpublished Ph.D. thesis. Harvard. 358 pp. JOPLING, A.V. & WALKER, R.G. (1968) Morphology and origin of ripple drift cross-lamination, with examples from the Pleistocene of Massachusetts. J. sedim. Petrol. 38, KUENEN, PH. H. (1967) Emplacement of flysch-type sand beds. Sedimentology, 9, MCKEE, E.D. (1938) Original structures in Colorado River flood deposits of Grand Canyon. J. sedinz. Petrol MCKEE, E.D. (1939) Some types of bedding in the Colorado River delta. J. Geol. 47, MCKEE, E.D. (1965) Experiments in ripple lamination. In: Primary Sedimentary Structures and Their Hydrodynamic Interpretation (Ed. by G.V. Middleton). Soc. econ. Paleont. Miner., Tulsa, 12, SORBY, H.C. (1859) On the structures produced by currents present during the deposition of stratified rocks. Geologist, 2, SOUTHARD, J.B., ASHLEY, C.M. & BOOTHKOYD, J.C. (1 972) Flume simulation of ripple-drift sequences, Abst. Prog. geol. SOC. Am. 4, 672. SOUTHARD, J.B. & BOXJCHWAL, L.A. (1973) Flume experiments on the transition from ripples to lower flat bed with increasing sand size. J. sedim. Petrol. 43,, SRODON, J. (1974) An interpretation of climbing ripple cross-lamination. Roczn. pol. Tow. geol. 44, (Manuscript received 4 February 1981 ; revision received 24 April 1981)