EVOLUTION OF FLOW CELLS WITHIN A MASS-TRANSPORT COMPLEX: INSIGHTS FROM 1 THE GORGON SLIDE, OFFSHORE NW AUSTRALIA 2

11 Mass flows evolve longitudinally during emplacement, but can also vary laterally by forming discrete 12 flow cells with different rheological states separated by shear zones. Despite being documented in 13 many field and subsurface studies, the initiation, translation, and cessation of the flow cells remain 14 unclear. We use five, high-quality post-stack time-migrated (PSTM) 3D seismic reflection datasets to 15 investigate the evolution of flow cells in a seabed mass transport complex (MTC), the Gorgon Slide, on 16 the Exmouth Plateau, offshore NW Australia. The slide originated from a 30 km-wide, NE-SW trending 17 headwall scarp that dips steeply (c. 30) seaward, and was translated to the NW over a basal-shear 18 surface that deepens downslope (up to 500 m below seafloor). The slide is dominated by chaotic 19 seismic facies with discrete packages of coherent reflectors, which is interpreted as a debrite that 20 carried megaclasts (c. 0.05-1 km-long) derived from the headwall domain. The morphology and 21 orientation of the basal-shear surface focused the pathway of the slide, resulting in clustering of 22 megaclasts in proximal parts of the translation domain. The megaclasts cluster became an obstacle to 23 flow, which resulted in the formation of two flow cells (Cells A and B) separated by a longitudinal shear 24 zone. The interaction between the two cells is recorded by sinuous shear bands within, and ridges on 25 the top surface of, the slide. Along the longitudinal shear zone, the shear bands and ridges of Cell A 26 were dragged downslope, due to Cell A impeding the movement of Cell B. This interaction suggests 27 that Cell B travelled faster, and/or further, than Cell A, due to the absence of any flow obstacles. The 28 abrupt cessation of Cell A is recorded by positive seabed relief, whose amplitude decreases updip. The 29 transport processes of the Gorgon Slide show how entrainment and abrasion of megaclasts induced 30 velocity perturbations during emplacement causing: (i) changes to the flow rheology, and (ii) initiation 31 and cessation of flow cells. A better understanding of how flow cells evolve during MTCs transport 32 may help to refine modelling of the potential impact of MTCs on submarine structures (e.g. pipelines, 33 cables, etc.). 34


INTRODUCTION 35
The degradation of submarine slopes drives emplacement of large mass-transport complexes (MTCs), 36 which are deposits of gravity-driven depositional processes that include slides, slumps and debris 37 flows (e.g. Dott 1963; Nardin et al. 1979;Nemec 1991 To reconstruct emplacement processes, the Gorgon Slide is synthesized in terms of its kinematic 167 indicators and internal seismic facies from four domains: (i) headwall; (ii) upper translation (UTD); (iii) 168 lower translation (LTD); and (iv) toe (see Fig. 3B).
Besides the main headwall scarp, features observed in the headwall domain (see Fig. 6) include a small 171 scarp, circular depressions, and linear depressions updip from the main headwall scarp (see zoomed-172 in image in Fig. 6A). The small scarp (c. 10 m-high) is likely to be older than the main headwall scarp 173 due to their cross-cutting relationship (Fig. 6B) The initial failed volume of the Gorgon Slide that was removed from the headwall domain ranges from 187 31 to 43 km 3 , which is 12-16 times smaller than the deposited volume (c. 500 km 3 ) (Nugraha et al. 188 2019a). This volume discrepancy is interpreted as a result of significant erosion and substrate 189 entrainment of the carbonate ooze substrate during transport (see Nugraha et al. 2019a).

Upper translation domain 191
Basal-shear surface.-Grooves observed in the updip part of the upper translation domain (Fig. 7A), 192 are a continuation of those within the evacuation zone (see Fig. 6), displaying similar dimensions and 193 geometries (see Headwall domain section). However, the grooves in this domain converge downslope 194 towards the NE lateral margin (Fig. 7A), which contrasts to the more commonly described downslope-195 diverging grooves (e.g. Posamentier  The ramps record basal erosion by the overlying slide that are commonly expressed by truncated 207 reflections of underlying substrate by a basal-shear surface (e.g. Bull et al. 2009). However, as the 208 ramps in this domain represent relatively small steps (i.e. 10 m), the basal-shear surface does not 209 truncate more than one reflector. 210 Adjacent to the NE lateral margin, there is an area comprising highly discontinuous reflections on the 211 variance map (Fig. 7A). Some of these discontinuous reflections form lineations oriented oblique to 212 the NE lateral margin. This area is characterised by low-to-medium amplitude, discontinuous 213 reflections at, and immediately beneath, the basal-shear surface (Fig. 7D). We interpret the substrate in this area to have been compressionally deformed due to stress exerted by the slide, forming a 215 'basal-shear zone' (Butler and McCaffrey  Immediately downdip from the central part of the Area A megaclast cluster, partially-disaggregated 239 materials contained within SF-2 are aligned to form a series convex-upslope bands (Fig. 7B). These 240 bands are sub-parallel to the geometry of the cluster (Figs. 7B and 8A). In contrast, downdip from the 241 eastern margin of the cluster, the bands show a convex-downslope geometry terminating at the NE 242 lateral margin (Figs. 7B and 8A). A NW-SE trending, narrow area (c. 500 m-wide and c. 10 km-long) of 243 SF-1 defines the boundary between these two sets of bands (Fig. 7B). There is a similar occurrence of 244 convex-downslope bands downdip from the cluster in Area B (Fig. 7B). In seismic section, these bands 245 are expressed as low-frequency, medium-amplitude folded reflections (Fig. 7D). 246 The scattered megaclasts in the proximal part of this domain are clustered, possibly due to downslope-247 convergence of the pathway of the Gorgon Slide based on the orientation of the grooves on the basal-248 shear surface (see Fig. 7A). The clusters of megaclasts in Areas A and B are interpreted to have been 249 initially emplaced as a single cluster. Subsequently, they were cross-cut by the longitudinal shear zone, 250 which was also initiated at this area (Fig. 7B). 251 The clusters of megaclasts were likely induced internal velocity perturbations within the slide during 252 transport. This internal velocity variation is evidenced by the convex-downslope shear bands, in both 253 Areas A and B, which are located downflow from the convex-upslope shear bands adjacent to the 254 cluster of megaclasts in Area A (Fig. 7B). Another indicator of internal velocity variations is the narrow 255 area within Area A that separates the convex-downslope and -upslope shear bands ( The ramps, deformed substrate, and shear fractures indicate that erosion and deformation also 276 occurred in this lower translation domain (Fig. 9A). There is a close spatial relationship between the 277 deformed substrate and the concentration of shear fractures (Fig. 9A), which also coincides with the 278 thickest slide occurrence in Area A (Fig. 3A). This could imply that the basal and lateral substrate 279 deformation was more severe due to increased stress exerted by the thickest part of the Gorgon Slide. Area B, suggests a cross-cutting relationship (Fig. 11B). Thus, we interpret that the thrusted megaclasts 359 had been emplaced at their present location prior to the emplacement of the Gorgon Slide. Some 360 thrusted megaclasts (i.e. indicated by high RMS amplitude) are observed within the frontal part of 361 Area A (Fig. 11B). However, these thrusted megaclasts are distinctly different from those of the thrust 362 system within Area B (Fig. 11D). This suggests that the thrusted megaclasts were only entrained by 363 the Gorgon Slide in the frontal part of the Area A (Fig. 11B) . 10 m, Fig. 9). 378 The ridges at the Area A frontal margin could indicate a buttressing effect of the slide against the 379 thrusted megaclasts (Fig. 11C), which then formed ridges that decrease in height upflow. In contrast, 380 the ridges in Area B are not as high as those in Area A. Thus, the slide was not buttressed against the 381 thrusted megaclasts and it translated further downdip (Fig. 11C). These differential processes in the 382 toe domain of Areas A and B reflect the merging between Area B lateral margin and the longitudinal 383 shear zone (Fig. 11C). In this study, the Gorgon Slide appears to comprise two intra-MTC (second-order) flow cells. These are 397 represented physically by Areas A and B, and for the purpose of this process-based interpretation are 398 re-named as Cells A and B, respectively. The emplacement processes of the Gorgon Slide are captured 399 in a schematic model that recognises three stages of development (Fig. 12). Downflow from the shadow zone, ridges of Cell A show convex-upslope geometries adjacent to the 437 longitudinal shear zone. In contrast, ridges of Cell B consistently exhibit convex-downslope geometries 438 (Fig. 12C). These geometries indicate that Cell A resisted the downslope translation of Cell B. As a 439 result, the ridges of Cell A were dragged downslope, and the ridges of Cell B were dragged upslope 440 (Fig. 12C). Therefore, we suggest that Cell A was travelling more slowly than Cell B. Furthermore, the 441 high seabed relief of Cell A at the frontal margin suggest that it was forced to stop its translation by 442 the pre-existing thrusted megaclasts (Fig. 12C), and, thus, can be considered as "stopping structures" 443 that were formed during cessation (Masson et al. 1993;Gee et al. 2006). In contrast, the position of 444 the thrusted megaclasts allowed Cell B to translate further downdip than Cell A. Thus, Cell B was not 445 only travelled faster, but also further, than Cell A. 446

Impact of flow cells formation on MTCs flow behaviour 447
Submarine debris flow can travel for tens to hundreds of km across low gradient (c. <1 o ) continental 448 slope, despite its cohesive nature (Gee et al. 1999;Lastras et al. 2005). This mobility can be explained 449 by, for instance, sustained pore-fluid pressure within the flow during transport (Major and Iverson frontal part of the flow (i.e. hydroplaning, Mohrig et al. 1998). Ultimately, a debrite is formed by en 452 masse freezing of the debris flow (e.g. Talling et al. 2012), where materials at flow margins (i.e. frontal 453 and lateral) cease moving first, followed by materials in the main body of the flow (Iverson 1997). 454 The Gorgon Slide provides evidence of how a mass flow split into two smaller flow cells (Cell A and B, 455 Relatively continuous, sub-parallel reflections at the base of the thrusted megaclasts could indicate 493 that they had not been translated (Fig. 11D). This might support the interpretation that they are in-494 situ megaclasts. However, the NE-dipping thrusts originating from the base, and folded reflections 495 toward the top of the megaclasts, record contractional strain as a result of broadly NE-SW trending σ 1 stress (Fig. 11D). This stress was unlikely exerted by the Gorgon Slide onto the in-situ megaclasts, as the Gorgon Slide was transported towards the NW. Likewise, it is unlikely that the NE-dipping thrusts 498 were formed by an older MTC that translated to the SW. This is because there is no possible source of 499 MTCs towards the NE (see location of the NW Australian shelf , Fig. 1A). 500 If the thrusted megaclasts were to be deformed or translated by an MTC, the MTC should be sourced 501 either from the Exmouth Plateau Arch (i.e. to the SW from the megaclasts) or from the NW Shelf of 502 Australia (see Fig. 1A). As the thrusts of the megaclasts are NE-dipping (Fig. 11D) Here, the megaclasts were re-oriented from perpendicular to become sub-parallel 515 to transport direction with increasing distance from headwall scarp. 516

CONCLUSIONS 517
We use 3D seismic reflection data covering a recent mass-transport complex (MTC), the Gorgon Slide, 518 from the Exmouth Plateau, offshore NW Australia, to investigate how flow cells within an MTC was 519 formed, translated, and finally ceased. This study concludes that:  . 1.---A) Location of the study area. Regional seismic line (orange) across several wells (see Fig. 2). 774 B) Seabed map of the Gorgon Slide, and industry well data (red dots) available for this study. The 775 Gorgon Slide is expressed as rugose relief on the seabed. Both evacuation and most of deposition 776 zones are imaged within the 3D seismic reflection data. C) Outline of the deposits of the Gorgon Slide 777 (dark grey), where a minor area (c. 7%) of the total slide area in the NW (dashed line) is not imaged 778 within the 3D seismic reflection data. This minor part is delineated using 2D seismic lines (green). Five 779 3D seismic reflection datasets (Gorgon, Acme, Draeck, Duyfken, and Io-Jansz) were used in this study. 780 Bathymetry and topography data are from Geoscience Australia. 781