A model for the headward of submarine induced by downslope-eroding flows

Lincoln F. Pratson Lamont-Doherty Earth Observatory, Columbia University, Palisades, New York 10964 Bernard J. Coakley }

ABSTRACT dence for headward erosion by mass erosion by the canyon and carv- wasting with the stratigraphic evidence for ing a sinuous . A simple, physically based computer canyon inception by downslope-eroding sed- The differences between shelf-indenting model of continental slope evolution is used iment flows. and slope-confined canyons first noted along to investigate the sequence of submarine the mideast U.S. continental slope by canyon formation. The model simulates INTRODUCTION Twichell and Roberts (1982) and Farre et al. submarine canyons as evolving under the (1983) have subsequently been observed influence of sedimentation, slope failure, In one of the first regional studies of a along other continental slopes throughout sediment flow erosion, and topography. In- continental slope using GLORIA side-scan the world (e.g., McGregor, 1985; Dingle and teractions between these factors are mod- sonar imagery, Twichell and Roberts (1982) Robson, 1985; Nelson and Maldonado, eled as being governed by local sea-floor pointed out that two general populations of 1988; Klaus and Taylor, 1991). Their idea slope, which in the model determines the submarine canyons occur along the mideast that slope-confined and shelf-indenting sub- extent of sea-floor failures, directs the U.S. margin: a relatively few, large, often marine canyons represent different stages of downslope path of sediment flows triggered sinuous canyons with heads that indent the canyon evolution has provided a framework by the failures, and scales the amount of shelf break, and a much greater number of for interpreting the relative ages of canyons sea-floor erosion caused by the sediment smaller, more linear canyons with heads on a continental slope. Furthermore, the flows. Based on these interactions, the hundreds of meters below the shelf break on Farre et al. (1983) theory has provided an model simulates a three-stage sequence for the continental slope. They suggested that explanation for why slope-confined and submarine canyon formation: (1) the ero- the slope-confined canyons had initiated on shelf-indenting canyons co-exist. sion of pre-canyon by sediment flows the continental slope, and that the shelf-in- Aspects of the Farre et al. (1983) theory initiated at sites on the upper slope over- denting canyons had evolved from slope- have recently been called into question. Us- steepened by sedimentation; (2) localized confined canyons and thus were older. ing multibeam bathymetry, seismic reflec- slope failure of the walls and/or floor of the This same idea was independently tion profiles, and borehole data, Pratson et rills at one or more mid- to lower-slope sites reached by Farre et al. (1983) after studying al. (1994) identified and mapped a number destabilized by sediment flow erosion; and the morphology of a number of submarine of buried canyons in an area of the New (3) evolution of the failure into a headward- canyons on the New Jersey and Maryland Jersey Slope investigated by both Farre et al. eroding canyon that advances upslope along continental slopes imaged in higher-resolu- (1983) and Twichell and Roberts (1982). the rills by sediment-flow-driven retrogres- tion SeaMARC I side-scan sonar imagery. Pratson et al. (1994) found that at least sev- sive failure. Through this sequence, the They went a step further than Twichell and eral of the larger, ‘‘slope-confined’’ canyons model simulates canyon and intercanyon Roberts (1982) and hypothesized a scenario in the area have buried, upslope extensions morphology that successfully reproduces for the evolution from ‘‘youthful’’ to ‘‘ma- that may once have indented the shelf break crosscutting relations observed between ture’’ canyon morphology. Farre et al. but now are infilled and cannot be discerned Lindenkohl Canyon and adjacent erosional (1983) proposed that the youthful stage of in side-scan sonar imagery. They also found slope rills on the passive-margin New Jersey submarine canyon evolution begins with that a number of existing canyons exploited continental slope, and between slope fail- slope failure. Retrogressive mass wasting of the lower slope reaches of the older, buried ures and long, narrow dendritic the continental slope along the canyons, suggesting that the existing can- that enter into the Aoga Shima Canyon on failure headwall leads to the formation and yons were initiated by downslope-eroding the convergent-margin Izu-Bonin fore arc. upslope extension of a relatively straight, sediment flows rather than upslope-eroding These results suggest that the model may be steep-walled chute. If this headward-migrat- retrogressive failures. applicable in explaining submarine canyon ing chute breaches the shelf break, the can- To explain the formation of these can- formation along a variety of continental yon taps into a new sediment source of outer yons, Pratson et al. (1994) combined ideas of margins. More significantly, in illustrating shelf sands and enters into a mature phase Farre et al. (1983) with those of Daly (1936), how sediment flows might repeatedly trigger of canyon evolution. Failure of shelf sedi- who was the first to suggest that submarine retrogressive failures, the model presents a ments in the vicinity of the canyon head in- canyons are eroded by turbidity currents. new explanation for submarine canyon for- itiates coarse-grained turbidity currents, Pratson et al. (1994) proposed that canyon mation that reconciles morphologic evi- which become an important agent in canyon initiation began with depositional over-

GSA Bulletin; February 1996; v. 108; no. 2; p. 225–234; 8 figures; 1 table.

225

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/108/2/225/3382446/i0016-7606-108-2-225.pdf by guest on 25 September 2021 PRATSON AND COAKLEY

Figure 1. Shaded image of the bathymetric grid of the New Jersey slope used as a standard of submarine canyon morphology for evaluating model results. Grid is based on almost 100% SeaBeam bathymetry coverage of the sea-floor area shown in the inset. Grid cell size is 300 m2.

steepening and localized failure of the upper their hypothesis to the formation of (1983). We demonstrate these possibilities continental slope. The failures ignited ero- smaller submarine canyons on the New through a new computer model of subma- sive sediment flows, which were routed Jersey Slope, which are confined to water rine canyon evolution that simulates the in- downslope through preexisting bathymetric depths Ն1500 m and cut into areas of shal- teraction between slope failure, sediment lows, including sea-floor troughs that overlie low-buried and exposed Eocene chalk. flow erosion, and topography as conceptu- the buried canyons on the middle to lower McHugh et al. (1993) have shown that the alized by Pratson et al. (1994). The model slope where they become only partially in- formation of these canyons was controlled results are contrasted to the morphology of filled. The erosion caused by these flows es- by diagenetically induced fracturing of the the New Jersey continental slope in the vi- tablished passages along which canyon de- chalk. But the events that triggered this cinity of Lindenkohl Canyon (box, Fig. 1), a velopment ensued. Subsequent sediment fracturing remain speculative. well-mapped subregion of the area where flows deepened the evolving canyons and We show in this study how the formation Twichell and Roberts (1982), Farre et al. oversteepened their walls, leading to a cor- of these lower slope canyons could have (1983), and Pratson et al. (1994) have all responding widening of the canyons through been triggered by the same sediment flow suggested that submarine canyons in dif- retrogressive canyon-wall failure. erosion that appears to have ultimately led ferent stages of evolution are represented. Pratson et al. (1994) used their hypothesis to the formation of larger slope-crossing The applicability of the model to canyon to explain the initiation of a number of canyons within the area. In so doing, we also formation in other continental margin set- large, slope-crossing canyons that incise show how downslope-directed sediment tings is then discussed in a comparison of the Miocene through Quaternary muds flow erosion could induce the type of head- the results to the morphology of the Aoga that form the upper New Jersey Slope ward canyon erosion conceived of by Twich- Shima Canyon on the Izu-Bonin fore arc (Robb et al., 1981). They did not apply ell and Roberts (1982) and Farre et al. off Japan.

226 Geological Society of America Bulletin, February 1996

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/108/2/225/3382446/i0016-7606-108-2-225.pdf by guest on 25 September 2021 EROSION OF SUBMARINE CANYONS

Figure 2. Schematic illustrating the downslope pattern of sedimentation simulated by the model. This pattern was uniform along the model slope.

THE MODEL Canyon. In this study, variations in the type maximum slope threshold. Thus, in the and strength of the sediments composing model, the extent of a slope failure is deter- Few studies have simulated the formation the slope are not modeled. mined by the steepness of the surrounding of submarine canyons. Tetzlaugh and Har- Sedimentation. Sedimentation is simu- continental slope area. baugh (1989) used their SEDSIM3 model to lated by raising the elevation of each grid Sediment Flow Erosion. Sediment flow simulate the formation of Simpson Canyon, cell on the continental slope a small amount erosion is initiated in the model at a grid cell an ancient submarine canyon buried be- during every model iteration. This amount that fails. The failed sediment is removed neath Alaska’s Arctic coast. More recently, varies randomly from one grid cell to the from the grid cell as a sediment flow that is Cao and Lerche (1994) have simulated the next (i.e., spatially) and with each model it- one grid cell in areal extent (0.09 km2). The formation of a generic canyon with their eration (i.e., ‘‘temporally’’) but is scaled by a unit size sediment flow, or ‘‘floxel’’ (Fig. 3B), MOSED3D model. The principal focus of background depositional pattern that deter- follows the steepest topographic descent these studies was not modeling submarine mines the maximum amount of sediment a downslope. As it moves, the floxel erodes canyon formation, but the erosion, sediment grid cell can receive during a single iteration the slope surface beneath it by reducing the transport, and caused by gravity- (Fig. 2). In this study, the depositional pat- elevation of each grid cell it passes over. driven sediment flows. The submarine can- tern reflects the accumulation of Quater- This amount is equal to the square root of yons were modeled as being eroded by tur- nary sediments along the New Jersey Slope. the local slope gradient times the volume of bidity currents initiated at predefined model These are thickest on the upper continental sediment the floxel is already transporting. locations. slope and thin seaward, suggesting they This parameterization represents a non– In this study, we present a new ‘‘seascape’’ were input from a shelf-edge line source time dependent linearization of the equa- evolution model for simulating the morpho- (Pratson et al., 1994). For simplicity, this tion by Komar (1977) for the body flow of a logic evolution of a passive, clastic continen- source is approximated using a background turbidity . Importantly, the influence tal slope. The model simulates sedimenta- depositional pattern that decreases linearly of other hydrodynamic factors in sediment tion, slope failure, and sediment flow from a maximum of 5 m per iteration at the flow erosion, such as friction, water entrain- erosion by discrete grid cell interactions. shelf edge to a minimum of 0 m per iteration ment, and flow density (Parker et al., 1986), The parameterizations for these processes at the slope base. is not considered here. are simple, but physically founded. They Slope Failure. Slope failure occurs in the In the case that the failure of a grid cell emulate the fundamental mechanics of each model when sedimentation raises the eleva- causes neighboring grid cells to retrogres- process but, in this study, not the rate at tion of a grid cell above the elevation of its sively fail, each of these generates its own which the processes occur. A factor common neighbors such that the slope between the floxel (Fig. 3C). The individual floxels inde- to all of the parameterizations is local sea- grid cell and its lowest neighbor exceeds a pendently move down the continental slope floor slope. In the model, local slope governs maximum slope threshold (Fig. 3A). This and erode the sea floor as they are initiated; the location and extent of sea-floor failure, maximum slope threshold is the single free however, the downslope path each floxel fol- directs the downslope path of the sediment parameter that can be varied from one lows is influenced by the eroded topography flow triggered by the failure, and scales the model run to the next. It represents the max- left by the floxel that immediately preceded amount of sea-floor erosion the sediment imum angle of repose for the continental it. Likewise, as it erodes the sea floor, a flow can cause. slope sediments and reflects the limit of floxel influences the path of any following Initial Model Surface. The initial model their material strength or stability. When a floxel. This dynamic feedback between ero- surface is a rectangular grid of sea-floor el- grid cell exceeds the maximum slope thresh- sion and surface evolution leads the individ- evations. For this study, a smooth, dipping old and fails, its elevation is reduced by an ual floxels to converge and form a multigrid elevation grid covering an area 30 km ϫ 30 amount that brings the grid cell 5% below cell sediment flow train (Fig. 3C). The larger km is used. The grid cells are 300 m on a the maximum slope threshold. If the failure the retrogressive failure, the larger the re- side. The grid ranges in water depth from of the grid cell oversteepens any of its im- sulting sediment flow train. 100 to 2100 m and has a gradient of ϳ4Њ, mediate neighbors, these fail too (Fig. 3C). When a floxel or sediment flow train which is similar to the New Jersey continen- This chain reaction of failures continues un- reaches the base of the continental slope, it tal slope between Lindenkohl and Carteret til the failure headwall is regraded below the exits the model. The model then iteratively

Geological Society of America Bulletin, February 1996 227

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/108/2/225/3382446/i0016-7606-108-2-225.pdf by guest on 25 September 2021 PRATSON AND COAKLEY

Figure 3. Schematic illustrating the possible sequence of slope failure and sediment flow erosion simulated by the model at sites on the model slope oversteepened by sedimentation.

adds sediment upon the new continental ures occur within these rills where the fre- corresponding widening of the canyon by slope surface until another slope failure is quency of floxel capture and wall erosion is side-wall failures farther down the slope. initiated, and the sequence is repeated. greatest. Eventually, these failures become A consequence of this pattern of erosion, large enough that they trigger a chain reac- which was unanticipated during model de- RESULTS tion of retrogressive slope failures (Fig. 4C). velopment, is that the simulated canyons are These retrogressive failures radiate from the mostly excavated by a few, large, retrogres- The style of submarine canyon evolution initial failure site, regrading, expanding, and sive slope failures (Fig. 5). When the walls of simulated by the model is illustrated with altering the form of the into a canyon- a floxel-cut rill begin to fail, the volume of two model runs made with different maxi- like excavation (Fig. 4D). Floxels of failed sediments removed by these failures in- mum slope thresholds. In one, a maximum sediments join to form sediment flow trains creases orders of magnitude within a mod- slope threshold of 12Њ is used, which approx- that erode pinnate into the walls of erate number of failure events (Fig. 5). The imates the mean gradient of the walls of sub- the simulated canyon as they are funneled magnitude of the slope failures then pla- marine canyons on the New Jersey conti- into its emerging thalweg (Fig. 4D). There teaus in size for about the same number of nental slope as measured from gridded they converge with other sediment flow failure events. It is during this phase of the multibeam bathymetry of the area (Fig. 1). trains to form a ‘‘catastrophic’’ erosional model that the most significant changes in In the other, a maximum slope threshold of event that excavates the downslope reach of simulated canyon morphology occur. Each 30Њ is used, which approaches the steepest the simulated canyon to the base of the big, retrogressive slope failure advances the gradient for the walls of these canyons. model slope. head of the evolving canyon up the conti- The sequence of canyon evolution in both Once the walls of the simulated canyon nental slope, forms new canyon tributaries, model runs is the same. Sedimentation in are established, they become focal points for and widens the canyon (Fig. 5). The big ret- the model gradually oversteepens the upper further slope failure for they already recline rogressive failures then diminish in fre- continental slope, and grid cells begin to fail. near the maximum slope threshold. Contin- quency once the head of the canyon ap- At first, the failures involve only single grid ued sedimentation along the canyon wall proaches the top of the continental slope in cells. These initiate floxels that cut narrow, rims and downcutting within the canyon by the model. shallow rills to the base of the continental sediment flow erosion steepen the canyon The headward canyon erosion simulated slope (Fig. 4A). Subsequent floxels moving walls beyond this limit, and thus they fail by the model is strikingly similar to the downslope through topographic lows nearby more frequently than the surrounding slope. youthful, mass-wasting phase of submarine are directed into the rills where their walls These failures are most common along the canyon evolution envisioned by Farre et al. become steeper than the adjacent slope sur- headwalls of the canyon toward the upper (1983). However, the model simulation de- face. The floxels then follow the rills to the slope where sedimentation rates and the in- parts from the Farre et al. (1983) theory in base of the slope, eroding and deepening troduction of floxels are greatest. These fail- several important ways. First, although the rills before they are infilled by ures lead to the headward growth of the sim- Farre et al. (1983) proposed that localized sedimentation. ulated canyon (Fig. 5). At the same time, failure on the continental slope begins head- With time, some of the rills are deepened they initiate sediment flow erosion within ward canyon erosion, they did not speculate to the point that portions of their walls are the canyon, which deepens the thalweg and as to the possible causes for such a failure. In oversteepened and fail (Fig. 4B). The fail- undercuts the walls. This in turn leads to a the model, the onset of headward canyon

228 Geological Society of America Bulletin, February 1996

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/108/2/225/3382446/i0016-7606-108-2-225.pdf by guest on 25 September 2021 EROSION OF SUBMARINE CANYONS

Figure 4. Simulations showing the sequence of submarine canyon evolution generated by the model.

erosion begins when repeated sediment flow tion simulated by the model is insensitive to Canyon in its thalweg profile (Fig. 7; Ta- downcutting in an area oversteepens the sea the value of the maximum slope threshold, ble 1). Both simulated canyons are much floor and triggers failure. the canyon morphology produced by the larger than Lindenkohl Canyon (Table 1) Second, Farre et al. (1983) proposed that model is not (Figs. 6 and 7). The canyon and have wider canyon floors that are sediment flows become an important factor simulated using a maximum slope threshold broader nearer to the base of the continen- in canyon erosion only after the head of a of 12Њ is characterized by a wide canyon tal slope (Figs. 6 and 7). canyon breaches the shelf break. Prior to floor and a broad, flowering headwall. The that point, other, continental slope–based canyon simulated using a maximum slope DISCUSSION mechanisms cause the retrogressive failures threshold of 30Њ has a deeper canyon floor that advance the canyon upslope. By con- and a narrower headwall. The results show The model results predict that the bulk of trast, the model results predict that sedi- that the maximum slope threshold influ- canyon formation is carried out by a few ment flows trigger the retrogressive failures ences the width and relief of the submarine large-scale retrogressive failures (Fig. 5). where they enter into the canyon and/or un- canyons simulated by the model. Some submarine canyons, like the Missis- dercut its walls. Thus, in the model, sedi- The two simulations are also compared to sippi, show morphologic and stratigraphic ment flows initiated upslope of a canyon are a bathymetric grid of Lindenkohl Canyon evidence of having undergone this type of an important factor driving the headward constructed from near-complete SeaBeam catastrophic formation (Goodwin and Prior, erosion of the canyon throughout its multibeam bathymetry of the New Jersey 1989). But others, such as Oceanographer evolution. continental slope (Figs. 6 and 7). The grid off Georges , exhibit evidence of cyclic A third difference is that Farre et al. has a grid cell spacing of 100 m and covers erosion and infilling dating as far back as the (1983) proposed that after the head of a can- the same area as represented in the two sim- Cretaceous (Ryan et al., 1978). The latter yon breaches the shelf break, downcutting ulations. Visually, the form and width of the observations suggest that many submarine by sediment flows leads to the development simulated canyon produced using a maxi- canyons evolve over long time periods from of a meandering thalweg. In the model, can- mum slope limit of 30Њ come closest to ap- repeated erosional events, presumably of yons develop meandering through proximating the morphology of Lindenkohl varying magnitude and due to multiple the slope failures that advance the canyon Canyon (Figs. 6 and 7). The simulated can- causes (Shepard, 1981). This more gradual headwall up the slope. yon produced using a maximum slope limit and varied form of canyon evolution is not Although this sequence of canyon forma- of 12Њ, however, is more like Lindenkohl addressed by the present model. It is biased

Geological Society of America Bulletin, February 1996 229

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/108/2/225/3382446/i0016-7606-108-2-225.pdf by guest on 25 September 2021 PRATSON AND COAKLEY

Figure 5. Time series showing sequential variations in the magnitude of erosion generated during a model run. Selected simulations demonstrate the significant growth and headward erosion that occur in the simulated canyons during a relatively few, large-volume retrogressive failures (gray area).

in its predictions toward canyon formation ied sediments should be less erodible, for yons having V-shaped cross sections, rather by catastrophic erosion. they generally have undergone compaction than the U-shaped cross section predicted This bias is attributable to at least two and have been hardened by consolidation by the model (Fig. 7). model simplifications. One is the use of a (Karig and Hou, 1992). As they are ex- A second simplification in the model is single maximum slope threshold, which im- humed, consolidated sediments should be that sediment flow erosion is simulated with- plies all slope sediments are of uniform able to sustain greater angles of repose than out factoring in the influence of flow accel- strength. A floxel in the model can downcut can surface sediments and be more resistant eration and related flow hydrodynamics canyon floor sediments exhumed from to slope failure and sediment flow erosion. (bed friction, water entrainment, flow den- depth the same amount as newly deposited This resistance would slow canyon evolu- sity, etc.). Using the parameterization in the sediments on the open slope. In reality, bur- tion. It is also the likely cause for many can- model, a simulated canyon develops a thal-

230 Geological Society of America Bulletin, February 1996

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/108/2/225/3382446/i0016-7606-108-2-225.pdf by guest on 25 September 2021 EROSION OF SUBMARINE CANYONS

Figure 6. Examples of simulations generated with the model using two maximum slope thresholds and their comparison to Lindenkohl Canyon. Letters in all three images correspond to analogous morphologies: A, upper slope sediment buildup; B, narrow erosional rills; C, lower slope failure excavations; and D, heads of erosional rills truncated by canyon sidewalls and tributaries. Note that in the simulation .using a maximum slope threshold of 30؇, sedimentation in the model raised the elevation of the upper continental slope above sea level This was because no process other than downslope erosion was simulated to offset sediment , such as subsidence or shallow water current and wave erosion. Although unrealistic, this consequence is not considered to have bearing on the results discussed in the text.

weg grade that reflects the volume and fre- area. This sequence of events is as follows: When followed downslope, the rills coa- quency of floxels that pass through it. How- (1) erosion of precanyon rills by sediment lesce and enter three to four amphitheater- ever, the magnitude of erosion a sediment flows initiated at sites on the upper slope shaped failure scars on the middle to lower flow can cause, and the amount of sediment due to depositional oversteepening (Fig. 4A); slope (Fig. 6C). Each of these scars consti- it can entrain, is dependent on the acceler- (2) slope failure along the rills at one or tutes one of the slope-confined canyons that ation of the flow (Parker et al., 1986; Nor- more mid- to lower-slope sites destabilized incise an Eocene chalk, the surface of which mark and Piper, 1991). The present model by sediment flow erosion (Fig. 4B); and (3) is generally covered by a thin veneer of requires a more physically based parameter- evolution of the failure into a headward- Pliocene-Pleistocene siliciclastic sediments ization of sediment flow behavior to assess eroding canyon that advances upslope along (Farre and Ryan, 1987). The walls and floors the role of sediment flow acceleration on the chute(s) by sediment-flow-driven retro- of the slope-confined canyons follow frac- submarine canyon formation. ture surfaces in the chalk that were initiated gressive failure (Figs. 4C–4D and 5). Variable sediment strength, sediment by pore-fluid expulsion during chalk diagen- The intercanyon area surrounding Lin- flow acceleration, and other improvements esis (McHugh et al., 1993). The fractures denkohl Canyon is incised by a series of rel- are being incorporated into a new version of apparently expanded following the erosion the model. The version discussed in this atively closely spaced slope rills (Fig. 6B) of overburden above the chalk and, in exfo- study, however, already succeeds in repro- with headwalls that cut into a steepened sec- liating, fostered the slope failures that exca- ducing several fundamental elements of tion of the upper slope (Fig. 6A). Pratson et vated the canyons. canyon and intercanyon morphology in the al. (1994) interpreted these slope rills to What process could have removed late vicinity of Lindenkohl Canyon on the New have been eroded by failure-induced sedi- Cenozoic sediments from above the Eocene Jersey continental slope. This success sug- ment flows, triggered by depositional over- chalk is unknown. McHugh et al. (1993) pre- gests that, despite its biases and simplicity, steepening of the upper slope. These slope sented several possibilities, including over- the model may be correctly simulating the rills may thus be a natural analog for the burden removal by mass wasting. Our model general sequence of events that led to the precanyon rills generated in the initial stages results suggest to us that erosion by sedi- formation of submarine canyons within this of the model results (Fig. 4A). ment flows passing over these areas is an

Geological Society of America Bulletin, February 1996 231

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/108/2/225/3382446/i0016-7606-108-2-225.pdf by guest on 25 September 2021 PRATSON AND COAKLEY

equally viable mechanism. In the model, failure-induced canyons are simulated as forming on the middle to lower slope by ret- rogressive failures. The mechanism that triggers these failures is oversteepening of the walls of the precanyon rills. The sedi- ment flows that passed through the New Jer- sey Slope rills could have caused similar fail- ures or simply downcut to a critical depth to trigger the exfoliation and subsequent fail- ures that formed the lower slope canyons. Thus, we suggest the lower slope canyons may be representative of the early stage of headward canyon erosion simulated by the model (Fig. 6C). Whether the lower slope canyons post- date the slope rills is uncertain and cannot be determined from the morphologic rela- tion between these features. But the slope rills do appear to predate the present areal extent of Lindenkohl Canyon and Carteret Canyon to the northeast. The walls of these canyons cut into and terminate the heads of adjacent slope rills (Fig. 6D). This distinc- tive crosscutting relation is successfully re- produced by the model. As the headwall and sidewalls of simulated canyons advance up and across the slope by retrogressive failure, they erode into and terminate the upslope extensions of precanyon chutes formed ear- lier nearby (Fig. 4C). In so doing, the sim- ulated canyons capture the upslope drain- age area of the precanyon rills. This capture shadows the rills from further downslope sediment flow erosion, aborting their con- tinued development. Growth of the inter- canyon slope rills cut into by the walls of Lindenkohl and Carteret Canyons appears to have been suspended by similar circum- stances (Fig. 6D). The parallels between the model results and the slope morphology in the vicinity of Lindenkohl and Carteret Canyons (Fig. 6) suggest these canyons evolved first from slope rills, and possibly later from slope- confined canyons. The model shows how slope-confined and shelf-indenting canyons Figure 7. A. Bathymetric profiles down the thalwegs of Lindenkohl Canyon and the two in this area can be related through a com- simulated canyons shown in Figure 6. B. Four profiles across Lindenkohl Canyon and the mon evolution involving retrogressive slope two simulated canyons. The mean depth has been removed from these across-slope profiles failures induced by sediment flows initiated to facilitate comparison of canyon widths and depths. upslope of the failures. The model thus pre- sents an explanation for the formation of submarine canyons on the New Jersey con- tinental slope that reconciles the morpho- logic evidence for headward canyon erosion reported by Twichell and Roberts (1982) and Farre et al. (1983) with the stratigraphic evidence for canyon inception by down- slope-eroding sediment flows presented by Pratson et al. (1994).

232 Geological Society of America Bulletin, February 1996

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/108/2/225/3382446/i0016-7606-108-2-225.pdf by guest on 25 September 2021 EROSION OF SUBMARINE CANYONS

Figure 8. The Aoga Shima Canyon as interpreted from seismic reflection data and full-coverage SeaMARC II side-scan sonar imagery. Modified from Klaus and Taylor (1991).

Because our model of canyon formation is ures. For instance, it does not seem likely canyon and at the headwalls of a slope-con- linked to sedimentation, it predicts that the that turbidity currents were involved in the fined canyon re-entrant (Fig. 8). process and pattern of canyon evolution is formation of slope-confined canyons on the The slumping along the Aoga Shima Can- dependent upon when and where deposi- sheltered seaward flanks of accretionary yon led Klaus and Taylor (1991) to infer that tional oversteepening of the continental ridges along the convergent Oregon margin the canyon evolved from headward erosion slope occurs. According to our model, can- (Orange and Breen, 1992). by mass wasting as proposed by Farre et al. yon evolution should be most active when Because many submarine canyons are not (1983). However, they noted that the den- sediment influx to the slope is greatest. Off- protected from sediment flow erosion, how- dritic tributaries that enter the head of the shore New Jersey, this would be during rel- ever, we believe that the model results pre- canyon on the inner fore arc (Fig. 8) are ative sea level lowstands. Studies of the ac- sented here may be applicable in explaining more likely to have been eroded by sediment tivity of submarine canyons within this the formation of a number of canyons in flows than mass wasting. They also noted region support this notion (Prior et al., 1984; that aspects of the canyon drainage pattern, diverse continental margin settings. One Stanley et al., 1984). Sediment flow source such as the bifurcation of the canyon around candidate is the Aoga Shima Canyon on the locations are also important, for their move- a frontal arc high (Fig. 8), are difficult to central Izu-Bonin fore arc south of Japan ment between sea-level cycles or within a reconcile with headward canyon erosion (Fig. 1B), a region that is the tectonic op- single cycle could lead to the abandonment strictly by mass wasting, for such canyon posite to the passive New Jersey margin. of one canyon head and the formation of growth should follow the steepest slope another, or to the abandonment of the can- Klaus and Taylor (1991) have mapped the gradients. yon altogether (Felix and Gorsline, 1971). Aoga Shima Canyon using SeaMARC II A number of the observations made by We recognize that all slope failures are side-scan sonar imagery and seismic reflec- Klaus and Taylor (1991) can be explained in not associated with depositional oversteep- tion data. They show that the canyon is fed the context of our model results. As alluded ening and sediment flow erosion. Earth- by a number of long, narrow, shallow den- to in the comparison to the New Jersey quakes, gas-hydrate decomposition, wave- dritic tributaries (Fig. 8), similar to the slope Slope rills above, the dendritic tributaries loading, and a variety of other processes can rills offshore New Jersey. Those dendritic that feed into the Aoga Shima Canyon ap- cause slope failure as well (Schwab et al., tributaries that enter into the canyon on the pear to have been eroded by sediment flows. 1993). We also recognize that there are re- middle to outer fore arc are cut in their mid- The arcuate slumps that cut the midsections gions where submarine canyons appear to section by arcuate slumps (Fig. 8). Arcuate of these tributaries (Fig. 8) may have de- have formed strictly from retrogressive fail- slumps also occur along the sidewalls of the rived from where erosion by the sediment

Geological Society of America Bulletin, February 1996 233

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/108/2/225/3382446/i0016-7606-108-2-225.pdf by guest on 25 September 2021 PRATSON AND COAKLEY

flows destabilized the sea floor. The most confined canyons. Our model results predict ontologists and Mineralogists Special Publication No. 33, p. 25–39. active tributaries would have fostered retro- that these deposits should consist of a few, Felix, D. W., and Gorsline, D. S., 1971, Newport submarine can- yon, California: An example of the effects of shifting loci of gressive slumping, leading to the formation large slope failure deposits, such as debris sand supply upon canyon position: Marine Geology, v. 10, and headward erosion of the canyon re-en- flows, underlain by and interbedded with p. 177–198. Goodwin, R. H., and Prior, D. B., 1989, Geometry and deposi- trant and presumably the canyon itself. Im- sediment flow deposits containing shallow- tional sequences of the Mississippi Canyon, Gulf of Mexico: Journal of Sedimentary Petrology, v. 59, p. 318–329. portantly, the model results predict that this water materials from the upper continental Karig, D. E., and Hou, G., 1992, High-stress consolidation exper- iments and their geologic implications: Journal of Geophysi- headward erosion would have occurred up slope and possibly outer continental shelf. A cal Research, v. 97, p. 289–300. the dendritic tributaries through which the cursory review of Miocene-Pleistocene age Klaus, A., and Taylor, B., 1991, Submarine canyon development in the Izu-Bonin forearc; a SeaMARC II and seismic survey of failure-triggering sediment flows were being sediments drilled on the New Jersey upper Aoga Shima Canyon: Marine Geophysical Researches, v. 13, p. 105–130. funneled. This prediction is consistent with continental rise just seaward of several Komar, P. D., 1977, Computer simulation of turbidity current flow the drainage pattern of the Aoga Shima slope-confined canyons appears to support and the study of deep-sea channels and fan sedimentation, in Goldberg, E. D., McCave, I. N., O’Brien, J. J., and Steele, Canyon (Fig. 8). The headwalls of both the this prediction. Deep Sea Drilling Project J. H., eds., The sea, Volume 6: New York, Wiley-Inter- science, p. 603–621. canyon and its re-entrant connect to den- Sites 604 (Shipboard Scientific Party, 1987a) McGregor, B., 1985, Role of submarine canyons in shaping the rise dritic tributaries. The prediction is also con- and 613 (Shipboard Scientific Party, 1987b) between Lydonia and Oceanographer canyons, Georges Bank: Marine Geology, v. 62, p. 277–293. sistent with the canyon re-entrant being cut were drilled ϳ10 km seaward of North Car- McHugh, C. M., Ryan, W. B. F., and Schreiber, B. C., 1993, The role of diagenesis in exfoliation of submarine canyons: around rather than up the frontal arc high, teret Canyon and ϳ6 km seaward of North- American Association of Petroleum Geologists Bulletin, for sediment flows follow topographic lows, east (respectively, Fig. 1), whereas v. 77, p. 145–172. Nelson, C. H., and Maldonado, A., 1988, Factors controlling dep- and this is the path along which a dendritic Ocean Drilling Program Site 905 (Ship- ositional patterns of Ebro turbidite systems, Mediterranean Sea: American Association of Petroleum Geologists Bulle- would have been cut. Finally, the board Scientific Party, 1994) was drilled ϳ35 tin, v. 72, p. 698–716. Normark, W. R., and Piper, D. J. W., 1991, Initiation processes model results would predict that the arcuate km seaward of Berkeley Canyon (Fig. 1). and flow evolution of turbidity currents; implications for the slumps bordering the sidewalls of the Aoga The Miocene-Pleistocene sediments recov- depositional record, in Normark, W. R., and Piper, D. J. W., eds., From shoreline to abyss: Society of Economic Paleon- Shima Canyon were widening the canyon in ered in these boreholes consist of multiple tologists and Mineralogists Special Publication No. 46, p. 207–230. response to thalweg deepening by sediment and possibly thick (potentially Ͼ200mat Orange, D. L., and Breen, N. A., 1992, The effects of fluid escape flow erosion, which undercut the canyon ODP Site 905) debris flow deposits com- on accretionary wedges, 2. Seepage force, slope failure, headless submarine canyons, and vents: Journal of Geo- walls. posed of slope materials, and interbedded physical Research, v. 97, p. 9277–9295. Parker, G., Fukushima, Y., and Pantin, H. M., 1986, Self-acceler- turbidites containing shallow-water sands ating turbidity currents: Journal of Fluid Mechanics, v. 171, CONCLUSIONS and fauna. Further investigation is needed p. 145–181. Pratson, L. F., Ryan, W. B. F., Mountain, G. S., and Twichell, to determine how these deposits relate to D. C., 1994, Submarine canyon initiation by downslope- eroding sediment flows: Evidence in late Cenozoic strata on On the basis of the results of our seascape the formation of the submarine canyons im- the New Jersey continental slope: Geological Society of evolution model, we propose that Linden- mediately upslope of the boreholes. America Bulletin, v. 106, p. 395–412. Prior, D. B., Coleman, J. M., and Doyle, E. H., 1984, Antiquity of kohl Canyon, Carteret Canyon, the Aoga the continental slope along the middle Atlantic margin of the United States: Science, v. 223, p. 926–928. Shima Canyon, and other submarine can- ACKNOWLEDGMENTS Robb, J. M., Hampson, J. C., Kirby, J. R., and Twichell, D. C., 1981, Geology and potential hazards of the continental yons along other continental margins were slope between Lindenkohl and South Toms canyons, off- formed from headward erosion driven by This study was made possible through shore mid-Atlantic states: U.S. Geological Survey Open- File Report 81-600, p. 1–33. sediment flow downcutting. In this scenario, funds provided by the Office of Naval Re- Ryan, W. B. F., Cia, M. B., Miller, E. L., Hanselman, D., Nesteroff, W. D., Hecker, B., and Nibbelink, M., 1978, Bedrock geol- retrogressive failure of the headwall of a search (ONR grant no. N00014-93-1-0126). ogy in New England submarine canyons: Oceanologica slope-confined submarine canyon does not The authors are grateful to David Piper, Acta, v. 1, p. 233–254. Schwab, W. C., Lee, H. J., and Twichell, D. C., 1993, Submarine fortuitously ‘‘capture’’ tributaries cut by sed- Gerard Middleton, and G. Parker, whose landslides: Selected studies in the U.S. Exclusive Economic Zone: U.S. Geological Survey Bulletin 2002, 204 p. iment flows, as suggested by Farre et al. critical reviews greatly improved the manu- Shipboard Scientific Party, 1987a, Sites 604 and 605, in van Hinte, (1983). Instead, such tributaries are the pre- script. We also thank W. Ryan and C. Pir- J. E., Wise, S. W., and others, Initial reports of the Deep Sea Drilling Project, Volume 93: Washington, D.C., U.S. Gov- cursors of a submarine canyon and establish mez for their thoughts on submarine canyon ernment Printing Office, p. 277–413. Shipboard Scientific Party, 1987b, Site 613, in Poag, C. W., Watts, the template along which submarine canyon formation and their constructive criticisms A. B., and others, Initial reports of the Deep Sea Drilling formation ensues. The sediment flows that of the original manuscript. Preparation of Project, Volume 95: Washington, D.C., U.S. Government Printing Office, p. 155–241. cut these passages promote slope failure the manuscript benefited from suggestions Shipboard Scientific Party, 1994, Site 905, in Mountain, G. S., Miller, K. G., Blum, P., and others, Proceedings of the and, if frequent enough, induce upslope ero- by J. Weissel and M. Steckler as well. Ocean Drilling Program, Initial Reports: Washington, D.C., U.S. Government Printing Office, p. 255–308. sion of the failure headwall along sediment Shepard, F. P., 1981, Submarine canyons; multiple causes and long-time persistence: American Association of Petroleum flow passages toward their source. In so do- REFERENCES CITED Geologists Bulletin, v. 65, p. 1062–1077. ing, the failure can evolve into a slope-con- Stanley, D. J., Nelsen, T. A., and Stuckenrath, R., 1984, Recent Cao, S., and Lerche, I., 1994, A quantitative model of dynamical sedimentation on the New Jersey slope and rise: Science, sediment deposition and erosion in three dimensions: Com- fined canyon and ultimately a shelf-indent- v. 226, p. 125–133. puters & Geosciences, v. 20, p. 635–663. Tetzlaff, D. M., and Harbaugh, J. W., 1989, Simulating clastic sed- ing canyon, which, depending on the process Daly, R. A., 1936, Origin of submarine ‘‘canyons’’: American Jour- imentation: New York, Van Nostrand, 202 p. nal of Science, ser. 5, v. 31, p. 401–420. Twichell, D. C., and Roberts, D. G., 1982, Morphology, distribu- sourcing the sediment flows (fluvial/deltaic Dingle, R. V., and Robson, S., 1985, Slumps, canyons and related tion, and development of submarine canyons on the United features on the continental margin off East London, SE Af- input, shelf-currents, etc.), may or may not States Atlantic continental slope between Hudson and Bal- rica (SW Indian Ocean): Marine Geology, v. 67, p. 37–54. timore Canyons: Geology, v. 10, p. 408–412. connect with the mouth of a on the Farre, J. A., and Ryan, W. B. F., 1987, Surficial geology of the continental margin offshore New Jersey in the vicinity of continental shelf. Deep Sea Drilling Project Sites 612 and 613, in Poag, C. W., The results of our seascape evolution and Watts, A. B., and others, Initial reports of the Deep Sea Drilling Project, Volume 95: Washington, D.C., U.S. Gov- model are tested in this study against the ernment Printing Office, p. 725–759. Farre, J. A., McGregor, B. A., Ryan, W. B. F., and Robb, J. M., morphology of existing submarine canyons. 1983, Breaching the shelfbreak; Passage from youthful to MANUSCRIPT RECEIVED BY THE SOCIETY AUGUST 12, 1994 The model results can also be tested by ex- mature phase in submarine canyon evolution, in Stanley, REVISED MANUSCRIPT RECEIVED JULY 17, 1995 D. J., and Moore, G. T., eds., The shelfbreak: Critical in- MANUSCRIPT ACCEPTED AUGUST 18, 1995 amining the deposits at the base of slope- terface on continental margins: Society of Economic Pale- LAMONT-DOHERTY EARTH OBSERVATORY CONTRIBUTION NO. 5426

Printed in U.S.A.

234 Geological Society of America Bulletin, February 1996

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/108/2/225/3382446/i0016-7606-108-2-225.pdf by guest on 25 September 2021