Submarine canyons and slides in the central-west Otway Basin: their morphology, genesis, links to groundwater discharge and tsunamigenic potential

Abstract The morphology and development of several submarine canyons offshore southeast South Australia and western Victoria are described. The existence of three of those canyons had been foreshadowed in 1963 by N. Boutakoff, who thought them to be linked to ancient courses of the Glenelg River. These canyons occur on the outer continental shelf where their heads are situated in depths shallower than 1000 m. Sinuous channels are visible within two of the canyon heads, indicating that water and sediment may still travel downslope and cause erosion, and thus are geologically recent. Several other canyons are also documented and named. Two canyons are confined to depths below 3000 m; they may be much older and seem not to be linked to ancient river courses. They are also oblique to the upper canyons. The other characteristic feature of the area is the presence of numerous undersea slides. These occur at three specific depths (∼1200, ∼1500 and ∼1800 m) and are coincident with stratigraphic horizons in which continental groundwater flows have been identified in adjacent exploratory oil and gas wells drilled into the continental shelf. Sapping of groundwater may likely have occurred during very wet periods inland. We suggest that these undersea slides could be the first step in the formation of deep-sea canyons that are not necessarily linked to ancient river courses. We also postulate that the ‘sliding’ of large piles of sediment down the continental slope has tsunamigenic potential and may have occurred during significant wet climate on land. KEY POINTS Several deep-sea canyons and numerous submarine slides are documented along the continental shelf offshore the coastline of western Victoria and southeast South Australia in the Otway Basin. Some of these canyons are linked to previous courses of the Glenelg River, which changed over time in response to regional tectonic changes such as faulting and significant upwarping. The orientation of most of these canyons tends to align with ancient terrestrial lineaments that likely extend offshore. Sinuous channels within the upper canyon reaches are clearly visible and not infilled with sediment; they are likely active and linked to ongoing groundwater seepage. It is proposed that the underwater slides are generated by the submarine discharge of continental groundwater on the continental slope, at three specific water depths with discharge associated with particular lithologies. These slides may still be active. Superimposed slides may engender the formation of new canyons. The volume of some of these undersea slides is of tsunamigenic scale.


Introduction
The Australian continental margin is characterised by numerous submarine canyons, many of which are ancient features, and may relate to the original breakup of Gondwana . Huang et al. (2014) assembled a compendium of 713 of these submarine canyons and demonstrated that they are not distributed evenly along the Australian margin, but rather are clustered in the east, southeast, west and southwest where the margin is steepest. These authors described two subtypes of slope-confined canyons: (1) 'headless canyons' (n ¼ 95) that incise the shelf edge but do not extend across the shelf, and which do not connect to river systems (Greene et al., 1991), and (2) 'blind (slope-confined) canyons' (n ¼ 618), with heads that terminate below the shelf break (cf. slopesourced canyons; Brothers et al., 2013). The distinction between these two subtypes is important because they relate to discretely different formative processes.
The Australian Navy first identified and mapped shelf-incising canyons in the southeast Asian region along southern New Guinea during WWII, and R. C. Sprigg subsequently redirected their focus to Australia's southern margin. He used that bathymetry to describe the morphology and origin of these features and inferred a link between the ancient courses of the River Murray and the marine Murray Canyon. Sprigg's (1947Sprigg's ( , 1948 original maps of the Lacepede Shelf (offshore South Australia), the ancient River Murray courses and the submarine canyons were remarkably consistent with maps subsequently produced by Hill et al. (2009, figure 3), which utilised over 1000 km of trackline and swath data on the Lacepede Shelf, including new multibeam data collected during the 2006 RV Southern Surveyor SS02/06 led by P. De Deckker and P. J. Hill. Sprigg's (1947, 1948 preliminary investigations of Australia's southern canyons were followed by those of Conolly and von der Borch (1967) and von der Borch (1968), who also documented the presence of several canyons along the southern coastline of Australia, as well as offshore Perth. These canyons, and a number of additional ones, have since been described along the southern margin of Australia, and these include the Murray Canyons Group , with the adjacent du Cou€ edic Canyon to the west and the Bonney Canyon to the east (Currie & Sorokin, 2014), the Albany Canyons to the west of the Great Australian Bight , the Perth Canyon in increasing detail (Huang et al., 2014;Nanson et al., 2022;Rennie et al., 2009), and the Bass Canyon on the eastern edge of Bass Strait (Mitchell et al., 2007). The mechanisms for initiation and expansion of these canyons have principally been ascribed to fluvial connectivity (Hill et al., , 2009Nanson et al., 2022;von der Borch, 1968). There remains, however, some ambiguity in the initiating mechanisms for these widespread features around Australia's southern margin, especially since Mahon and Wallace (2022) recently described Cenozoic incised channels in the Gippsland Basin, which they convincingly argue to have been submarine canyons not formed by fluvial activity. Figure 1 illustrates the distribution and morphology of a series of shelf-incising and blind canyons and slides within the geological Otway Basin (Figure 1, insert). Herein, we use new high-resolution bathymetric mapping to describe their morphology and link these observations with published sub-surface datasets to advance our understanding of their formative processes. In particular, we aim to identify their potential link to groundwater discharges at sea that may trigger canyon formation, and to also examine the potential tsunamigenic processes linked to underwater slides caused by groundwater sapping.

Previous work in the vicinity of the study area
Onshore paleochannels Boutakoff's (1963) comprehensive investigation of the geology and geomorphology of the Portland area of SW Victoria was a collaboration with the South Australian Geological Survey and R. C. Sprigg (1952), who had previously published a volume on the Geology of the South-East Province of South Australia. Boutakoff (1963) described how onshore faults and palaeochannel courses of the Glenelg River may have exerted some control over submarine canyon formation in the region (Supplemental data, Figures S1 and S2).
The clear pattern of conjugate NE-SW-trending onshore lineaments in southwestern Victoria was proposed by Boutakoff (1963, figure 22; see Supplemental data, Figure S3) to extend offshore and to be of a similar scale to those that also affect a large part of the Australian continent offshore. These lineaments originally formed as a result of an extensional regime during early Cambrian rifting (Jensen-Schmidt et al., 2001, figures 5-12) and have been reactivated since as an incipient fracture system (Holford et al., 2010(Holford et al., , 2014. In addition, there are faults, which are parallel to a series of anticlines, many of which were drilled by exploratory wells, including the four located on the shelf in the proximity of our study area (Figure 1). Many of the ancient courses of the Glenelg River also appear to be aligned with regional NE-SW lineaments/fractures (Boutakoff, 1963, figure 22; Supplemental data, Figure S3) that seemed to continue offshore and also define the orientation of the underwater Nelson and Keble canyons defined by Boutakoff (1963) on bathymetric contours (see also comments further below). In his figure 26 (cf. Supplemental data, Figure S2), Boutakoff (1963) extended Sprigg's (1952) observations to describe and also name the Rivoli Canyon offshore from Rivoli Bay (in South Australia; see further description below). In fact, Boutakoff (1963, figure 26; see also Supplemental data, Figure S2), using geomorphological features and the recognition of ancient fluvial deposits (mostly gravels), argued for four different ancestral courses of the Glenelg River. These were coined 'routes' by Boutakoff (1963) and are highlighted in the Supplemental data (Figures S1 and S2). Route 1 (recognised by Sprigg, 1952, in Boutakoff, 1963 is the oldest course and is associated with the development of the Rivoli Canyon. Boutakoff (1963) proposed that this initial course of the Glenelg River must have passed through the Dismal Swamp area (Supplemental data, Figure S1) and possibly also around the Mount Burr volcanic complex (Boutakoff, 1963, figure 26; Supplemental data, Figure S2; see also Murray-Wallace, 2018, figure 2.20). The development of the Gambier Upwarp subsequently diverted the rivers' course eastward (Boutakoff, 1963, p. 81). Route 2 (Supplemental data, Figures S1 and S2), where it exploited NE-SW lineaments through the Kentbruck region (Supplemental data, Figures S1-S3), must have persisted for a considerable period to contribute to the formation of the Keble Canyon, which was named and described by Boutakoff (1963, figure 22; NB: Keble Canyon appears to have been erroneously labelled as Glenelg Canyon on his figure 22). Route 3 follows the modern course of the Glenelg River until Kentbruck, where it veered west at Moleside before reaching the coast, and Route 4 is represented by the modern Glenelg River course. Note that Boutakoff (1963) in his block diagram (Supplemental data, Figure S2) identified another route, also called Route 4, which he suggested must be a very ancient one and that flowed towards the Surrey River along the 'Old Wannon River', upstream the Surrey River. The proposed outlet of those combined rivers to the sea must have been east of Portland ( Figure 1). Refer also to the Supplemental data ( Figure S4) for the proposed ancient courses of the Glenelg River following the suggestions of Boutakoff (1963).
The influence of the Glenelg River paleonetwork may not be limited to determining the course of the Otway canyons. Submarine groundwater discharge (SGD) via paleochannel courses, together with the mixing of these discharges with oceanic water and a combination of loose sediments that form 'subterranean estuaries' (STE; Moore, 1999), also have the potential to drive density currents and sediment transport on the continental shelf and slope (Taniguchi et al., 2002).
Fifty years subsequent to Boutakoff's (1963) work, Currie and Sorokin (2014) examined megabenthic biodiversity at six sites (100-2000 m) in the Bonney Canyon (Figures 1 and 2) offshore the Bonney Coast, which is known for its ephemeral upwelling (Schahinger, 1987) and, consequently, is an area of frequentation by blue whales (M€ oller et al., 2020). A multibeam bathymetry map of the Bonney Canyon provided in Currie and Sorokin (2014) shows its head starting around 900 m water depth (Figures 3 and 4). It is therefore a 'blind canyon' (sensu Huang et al., 2014). Above the canyon head, there appears to be a large number of undersea slides that extend down to the canyon head (Currie et al., 2012, figure 2) (Figures 3 and 4). The same authors identified that, during their RV Southern Surveyor cruise SS02/ 2008(4-16 February 2008, there was evidence of upwelling in the area of the Bonney Canyons (Currie et al., 2012, figure 8); this is driven by the westward underwater Flinders Current and its mesoscale eddies (Middleton & Cirano, 2002). These authors also postulated that those features could be strong enough within the canyons to Figure 1. Submarine canyons and slides characterising the continental slope offshore from the South Australian and Victorian border, illustrated here using broad regional bathymetry (250 m grid; Whiteway, 2009). The US82-31 seismic line (yellow) is labelled. Black diamonds indicate the location of the four exploratory wells drilled on the continental shelf, in which non-marine groundwater was identified by Knight et al. (2019). The rectangle shows the periphery of Figure 2. The small inset shows the location of the study area on the map of Australia.  move sediments and nutrients up slope (see modelling attempts by K€ ampf, 2018, 2021).

Paleo-canyons of the Central to west Otway Basin
Two studies describe buried (and filled) canyons offshore the Nelson-Portland area (Leach & Wallace, 2002;Pollock et al., 2002; and Supplemental data, Figure S6). In an area overlapping and immediately east of our study area, Leach and Wallace (2002) described well-preserved canyons in the Heytensbury Group's Oligocene-Miocene sequences as well as Pliocene to Recent ones in the Whalers Bluff Formation in the offshore Otway Basin. These authors identified that the Miocene canyon system consists of sedimentary deposits that migrated laterally to the west. These Miocene-filled canyons occur opposite Warrnambool in Victoria, between 141 E and 142 E to the west and east of Portland (Leach & Wallace, 2002, figure 3), $100 km east of the study area. The earliest canyon is thought to be early Miocene in age, and those canyons are broadly U-shaped, 2-7 km wide and at least 30 km long (Leach & Wallace, 2002). Their infill consists of sediments prograding towards the west. Leach and Wallace (2002) also stated that the other filled canyons are either symmetrical overall or oriented slightly to the east and argued that the modern canyons offshore Portland are located immediately seaward of their Pliocene counterparts. The Pliocene canyons are Ushaped, but V-shaped in the shoreward portion, are $12 km wide and range from 20 to 30 km in length (Leach & Wallace, 2002). The directional change in canyon migration, as suggested by Leach and Wallace (2002), resulted from oceanic changes that occurred at the Mio-Pliocene boundary, with the westerly-directed current in the Miocene that weakened afterwards. These authors also suggested that the canyon regimes were affected by tectonic activity, including uplift that began in the late Miocene (Dickinson et al., 2001).
To the west of Leach and Wallace (2002) area and overlapping our study area, Pollock et al. (2002) identified three palaeo-canyon systems within the Gambier Limestone of the South Australian Gambier Embayment (called the Gambier sub-basin by these authors). Pollock et al. (2002) identified a series of up to 20 cut-and-fill canyons/events that ranged in age from late Oligocene to mid-Miocene, but which started to form in the early Oligocene after sealevel fell and the shelf became exposed in this sub-basin; this enabled fluvial erosion to occur on the shelf. This phase coincided with tectonic uplift and a sea level drop as already assumed by Leach and Wallace (2002). The three paleo-canyons are named Robe Canyon, Lakes Canyons and Northumberland canyons (Pollock et al., 2002, figures 1 and 3; Supplemental data, Figure S6). Their respective dimensions are: Robe Canyon 30 km long and 4-5 km wide; Lakes Canyons are much smaller and 2-3 km apart, 10 km long and are 2.9, 3.5 and 2.3 km wide; and Northumberland Canyons 4.8 and 1.5 km wide and both 15 km long. Leach and Wallace (2002), Pollock et al. (2002) and Pollock (2003) presented an array of seismic profiles that illustrate the canyon geometries and their fill. In addition, Pollock (2003) also provided processed seismic sections in her unpublished PhD thesis. None of these buried canyons are visible in multibeam bathymetry.

Central-west Otway Basin
Oceanographic setting and surficial sediments Before discussing the modern canyon settings, a brief introduction to the oceanography of the central-west Otway Basin (Figure 1, insert) is warranted. Two principal oceanic currents affect the region: the Leeuwin Current, which originates in the tropics before circumnavigating the coast of Western Australia, continues along the southern Australian coast to sometimes reach northwestern Tasmania (Wijffels et al., 2018, figure 26). It is particularly strong in winter during La Niña phases, but importantly it is also shallow (<200 m). Where it reaches our study area, it is called the South Australian Current (James & Bone, 2011, figure 10.2). The South Australian Current overrides the western flowing Flinders Current, which at times can induce upwelling along the Bonney Coast (Middleton & Cirano, 2002) when westerlies are strong and located near the coast. The shelf along the Bonney Coast is relatively narrow ( 50 km) compared with the adjacent Lacepede Shelf to the west and the Otway Shelf south of Cape Otway. The geology of surficial sediments of the Bonney Coast and western part of the Otway Shelf is thoroughly described in James and Bone (2011), who reviewed the southeastern continental margin of Australia. Boreen et al. (1993) had earlier identified the modern carbonate sediment production on the Otway Shelf, which was later sub-divided by James and Bone (2011) into the Bonney Coast to the north and the Otway Shelf to the south and characterised both as highenergy, open, cool-temperate carbonates. In particular, the 0-70 m 'shallow shelf' is non-depositional and undergoes distinct erosion. The 70-130 m 'middle shelf' is a zone of wave abrasion, whereas the 'deep-shelf' (130-250 m) is a zone of accumulation of bioclastic sands with episodic reworking. The 'shelf edge and upper slope' are areas of accumulation of muddy carbonate sand, with banded pelagic carbonates that are produced by nutrientdriven upwelling. Nevertheless, James and Bone (2011) described differences between the two areas that are the result of the somewhat wider Otway Shelf. Overall, the modern Otway Shelf is sediment-starved. Nevertheless, a compilation of the surficial sediments of the shelf and slope on the Otway Shelf is provided in James and Bone (2011, figures 10.7 and 10.8), who clearly defined that the South Australian coast of the Otway Shelf is a 'zone of mixed sedimentary facies, characterised by slumping, mass wasting and downslope transport' (James & Bone, 2011, figures 10.4 and 10.7). Of note also is that the Bonney Coast is characterised by seasonal upwelling (Currie et al., 2012;K€ ampf et al., 2004;Richardson et al., 2020;Schahinger, 1987), and at times surface water temperatures can be 4 C cooler in summer than in winter (James & Bone, 2011).

Modern submarine canyons and slides
Two compilation bathymetry datasets (128 m: Spinoccia, 2020;250 m: Whiteway, 2009; Figures 1 and 2) provide complete coverage of the area of interest. We also utilise a higher-resolution (50 m), although less continuous, bathymetric dataset that extends across the continental shelf break and slope offshore from the mouth of the Glenelg River (Spinoccia, 2020). Together, these data reveal a series of submarine slides and two canyons offshore of Cape Nelson (Figures 2-4), and which to the northwest extend through the previously listed Nelson, Kebble and Rivoli canyons to Bonney Canyon (Figures 3-8). Their morphologies are described below using these bathymetric data and existing published insights into the paleochannel and canyon networks.

Keble Canyon
Keble Canyon was named by N. Boutakoff in 1963, although in his monograph, he also labelled it Glenelg Canyon. In fact, in 1949, R. Keble had pointed out to Boutakoff (1963, p. 79) that the Glenelg River must have exited at Kentbruck and would have formed a canyon at sea. This was later investigated by Boutakoff in 1951. We have retained the name of Keble Canyon, for reasons discussed below. This canyon is pronounced as shown in Figures 1 and 4. Gullies characterise its headwall (starting at 400 m water depth) and join to form a single channel that continues down the canyon. Keble Canyon is presently unconnected to a paleochannel, and headwall gullying is evidence of downslope current transport. This canyon was originally linked to route 2 of the Glenelg River, as proposed by Boutakoff (1963) and shown in the Supplemental data ( Figures S1 and S2).
At about $40 km along its course, the main canyon is joined by another canyon on its western bank in which some undersea slides are visible below 1000 m (Figures 2, 3b and 4b). Downstream of this confluence, the canyon's width increases, and there are clear incisions in the walls of the canyon that extend well down to 4000 m ( Figure 1).

Nelson Canyon
Nelson Canyon was originally named by N. Boutakoff in 1963. It is located just offshore Nelson township where the modern mouth of the Glenelg River is situated ( Figures  1-4). Similar to Keble Canyon, a narrow channel is visible from its commencement on the shelf (Figure 3b) but is not distinct below a depth of 1000 m. The eastern side of this canyon is marked by a set of undersea slides that may eventually link with the canyon with time and thereby broaden it.
Although this canyon is situated offshore from the modern-day Glenelg River, these features are disconnected across the shelf during the present sea-level highstand.

Bocara Canyon
Bocara Canyon is named here using the Gunditj Mirring name of Glenelg River. We are unsure if this canyon was formerly connected to the Glenelg River, but tectonic changes in the region indicate that it is possible (Figures 7  and 8). With no present high-resolution bathymetry across this part of the continental shelf, we are unsure of the most recent Last Glacial Maximum course of the Glenelg River when sea-level was approximately 120 m below present (De Deckker & Yokoyama, 2009;Yokoyama et al., 2000), or during former glacial periods.
The head of this canyon starts at a depth of $500 m (Figures 4b, 7 and 8), and consistent with both Keble and Nelson canyons, 'gullying' is clearly visible in its upper reaches. To its east another canyon is visible, and its head starts at a water depth of $800 m, but there is no evidence of gullying in the upper reaches of this canyon. Its formation appears to have commenced as a series of superimposed undersea slides (see white arrows in Figure 7b) that eventually became deep enough to initiate sediment failure and a subsequent canyon feature. In fact, a similar morphology is also visible on the eastern side of the adjacent Nelson Canyon, but it is visible at a water depth of $1000 m and again occurring at $2000 m water depth (Figures 7a and 8a).

Rivoli Canyon
Rivoli Canyon (Figures 3a and 4a) was named by N. Boutakoff in 1963 using information he received from R. C. Sprigg, but which had already been surveyed in 1940 by Lieutenant H. J. Stanley. Boutakoff (1963, p. 79) stated that this canyon is located off the coast, southwest of Millicent and directly south of Rivoli Bay in South Australia; he labelled it in his map on figure 26 (Boutakoff, 1963; Supplemental data, Figure S1). In that publication, he referred to comments by Sprigg, which the latter postulated must have been the first 'route' (sic Boutakoff, 1963) of the Glenelg River. In his monograph also, Boutakoff (1963) coined the name of Millicent Delta for a vast deltaic fan affecting the 27-46 m (15-25 fathom) contour north of the Rivoli Canyon, which is considered to be linked to the Millicent River. We are unable to verify this because resolution of the available bathymetry over this section of the continental shelf is too coarse (128 m; 250 m).
There is an additional canyon to the southeast of Rivoli Canyon, called here Rivoli East Canyon, and its head commences at a water depth of $3000 m ( Figures 5 and 6). Its upper reaches must be presently fed by turbidity currents discharging from the upper canyon, but it is incised into a stratigraphy that is aligned more in a southerly direction than the upper southwestern alignment. We are unsure if this is a much older canyon, and this will be further discussed below. Figure 5. Image to illustrate the deeper and wider canyon east and below of both Bonney and Rivoli canyons that have a different orientation from the canyons above them. These deep canyons, which are called Bonney East and Rivoli East canyons, commence at a water depth of approximately 3000 m are also bordered by undersea slides. The canyons proper commence at $4000 m. Bathymetric data are the 50 m grid overlaid on the 128 m grid (Spinoccia, 2020). The white areas indicate lack of data.

Bonney Canyon
An additional canyon located farther to the west is Bonney Canyon, which was thoroughly described by Currie et al. (2012;Figures 3a and 4a). The orientation of this canyon follows the same NE-SW trend seen for the other canyons located to the east (see Discussion), and it appears that this orientation is not repeated in other canyons farther west. There is no indication of a direct link to a modern (or palaeo) river, and large slide scars form a broad amphitheatre above the canyon's narrow headwall. The defined canyon headwall commences at $-1000 m and, in contrast with many of the other canyons described here, there is no evidence of gullying. Note that K€ ampf (2021) suggested that in this canyon, there is upslope sediment transport driven by upwelling, which we interpret therefore as possible evidence of sediment covering any former gullying.
Similar to the formation of Rivoli Canyon from $3000 m, a deeper canyon has also formed beneath the Bonney Canyon exit at $-4000 m, although it is offset to the east (called here Bonney East Canyon) (Figures 5 and 6). Several slump scars on the upper continental slope are coincident with the alignment of this canyon, and its origin is further discussed below.

Armand Canyon
The head of this canyon commences at $500 m water depth ( Figure 2). Its upper section (to $2000 m water depth) follows a N-S direction, whereas the lower part follows a NE-SW direction, like the lower part of the other canyons, and also the entire Bonney Canyon.
The proposed name is in memory of the late Professor Leanne Armand , who was the major leading force in the International Ocean Discovery Program and Director of the Australian and New Zealand Consortium based at the Australian National University. She was an expert on the use of diatoms in the Southern Ocean to determine the waxing and waning of sea ice around Antarctica through late Quaternary times.

Canyon long profiles
The profiles of all the canyons are illustrated in Figure 9 and in the Supplemental data ( Figure S7a, b); these show some distinction between several of the canyons that relate to the angle of the continental slope, but also the depth of the canyon headwalls. Armand Canyon initiates at the greatest depth and has a steeper slope angle. Bonney Canyon indicates a change in slope angle at $3000 m water depth, which, below that, is much steeper than the other canyons. This may be due to a change in lithology as postulated from the seismic profiles discussed in Pollock (2003) and extrapolated in Figure 11.

Development of the modern central-west Otway Basin canyons
Buried paleo-canyons beneath the continental shelf, which also extend a little on the continental slope, were described by Leach and Wallace (2002), who, prior to the availability of suitably high-resolution bathymetric data, mistakenly also illustrated the heads of the modern-day canyons as starting below the shelf edge. Pollock et al. (2002) described the development of these ancient canyons by the retreat of nick-points that were themselves initiated during lowstands and retreated by turbidity current action during sea-level highstands. These sea-level fluctuations may have been eustatic or driven by tectonic readjustments and increased substantially towards the end of the early Oligocene at ca 30 Ma (McGowran et al., 1997, figure 3). Pollock et al. (2002) also document what is now called Bonney Canyon (Currie & Sorokin, 2014;Currie et al., 2012), which they described as fault-controlled (Pollock et al., 2002, figures 3 and 14) with its axis striking NW-SE and paralleling the regional fault strike (sic Pollock et al., 2002), as originally postulated by Boutakoff (1963). It is also important to note that Rivoli Canyon is located directly seaward of the now buried Robe Canyon (Supplemental data, Figure S8), and if Boutakoff (1963) was correct in postulating that this canyon system is linked to an ancient course of the Glenelg River (Supplemental data, Figures S2 and  S3), this would indicate its great antiquity. Pollock et al. (2002) identified that incision of Robe Canyon commenced Figure 7. (a) Planview of the heads of Bocara and Nelson canyons that clearly show numerous gullies on both slides, which indicate that mass wasting does occur today, otherwise these channels would be filled with sediments. Both canyon heads commence around $450 and $500 m water depth. Bathymetry is based on the 50 m grid (Spinoccia, 2020). (b) Planview of Bocara and Nelson canyons as well as the undersea slides on either side. The white arrows indicate a series of undersea slides, which are superimposed over one another and are thought to be the commencement of new canyons. The direction of those superimposed slides follows the general NE-SW trend also followed by most the canyons in the area. Bathymetry is derived from a temporary grid and image created during the data acquisition survey. Note the different datasets used to illustrate parts (a) and (b).
in the late Oligocene and ceased in the mid-Miocene, having gone through approximately 20 successive series of 'cut-and-fill' processes. These are found buried below what is now the continental shelf. Channels within the headwall reaches of Rivoli Canyon indicate that it is recently active (Figures 3a and 4a). Boutakoff's (1963, figure 22) illustration of the lineaments across southwestern Victoria showed a clear parallel set of lineaments running NE-SW (see discussion below), which he extended offshore and which seem to coincide with one of the canyons. The direction of Nelson Canyon is also consistent with these onshore lineaments (Supplemental data, Figure S3). In fact, examination of Figures 1 and S3 shows that all the canyons, as well as the assemblage of many of the undersea slides, follow the same directional trend. Gerber et al. (2009) used models for sediment gravity flow dynamics to predict the longitudinal profiles of submarine channels, which are commonly upwards concave. The profiles shown in that article are markedly different from the canyons described here, except for their upper sections (above 2500 m depth). With the exception of the headwalls of Bonney (linear) and Bocara (convex) canyons (Figure 10), Gerber et al.'s (2009) profile predictions match these profiles with the concave ones they described, which these authors claim to have formed 'in the absence of background sedimentation and progradation'. In the absence of background sedimentation and progradation, the graded canyon long profiles are concave and described by a simple power law slope-distance relationship that arises from down canyon increases in discharge owing to flow evolution.  figure S1). In contrast, all the paleo-canyons described by Leach and Wallace (2002) from north of Nelson Canyon to the east as far as south of Warrnambool and by Pollock et al. (2002), in the same region as the canyons described here, were repeatedly incised and filled completely and are now covered by a veneer of more recent sediments. Thus, the ancient canyons must have formed and eventually filled under a different sedimentary regime. This was discussed by Leach and Wallace (2002), who suggested that a strong and persistent westward current caused some sediment to drift and eventually fill the canyons offshore Warrnambool. The likely current would have been the precursor of the modern-day Flinders Current described by Middleton and Bye (2007), but in order to have a substantial sediment drift, its strength must have been enhanced in the past. Although Leach and Wallace (2002) suggested that the Leeuwin Current (LC) would have been absent at that time, this contrasts with McGowran et al.'s (1997) findings. The modern LC is shallow (<200 m deep), and even when strong, it overrides the Flinders Current but does not affect it. The presence of warm water faunas during the time of the formation of the old canyons is explained by the LC above them.

Canyon infilling and sediment regimes
Since the modern-day canyons appear to be devoid of sediments, the area is sediment-starved, similar to the Murray Canyon Group (cf. De Deckker et al., 2021), in contrast to the older canyons that became completely infilled. The contrast between the repeated incision and infilling of the paleo-canyons and the persisting expression of the modern canyons must be the result of varying oceanographic and climatic regimes. Overall, the modern Otway Shelf is sediment-starved, with insufficient sediment to infill the modern canyons. This results from the fact that, especially in winter, the whole area is strongly affected by longperiod swells originating from the Southern Ocean that continuously sweep the shelf, with possible transport of sediments down the canyons. Moreover, Moros et al. (2021) argued that during the Last Glacial Maximum, the subpolar In contrast, the late Oligocene to mid-Miocene canyon infills could be explained by enhanced biogenic sediment production and deposition under calmer oceanic conditions during the warmer climatic regime of that period, but also because the uplift that created the Otway Ranges produced more sediment that was eventually delivered to the region. Towards the end of the Miocene, climatic conditions commenced deteriorating that eventually led to the modern-day regime with prevalence of a very cold Southern Ocean and associated oceanographic conditions (McGowran & Hill, 2015).
Underwater slides offshore the southeastern Coast of South Australia 1 Figures 3b, 4b and 11 illustrate a series of scars on the upper continental slope and at greater depths (>3000 m; Figure 11). These appear to have resulted from underwater slides of a pile of sediment caused by sudden slope instability. These slides, which have left 'scars' on the slope, range in size. The estimated volumes of two of the largest slides are 18.6 km 3 and 13 km 3 (Figure 6b; Supplemental data, Figure S5a, b), and their headwalls are situated at what appears to be consistent water depths (e.g. $1200 m and $1500 m; Figure 3a, b; Supplemental data, Figure S11). Several of the scars are composite and are aligned parallel to the main direction of the canyons (see white arrows in Figure 7b) that result in submarine mass failures (SMFs), the triggers for which are discussed below.

Discussion
SMFs have the potential to generate tsunamis, and four potential initiating mechanisms for submarine mass failure are considered and described: (1) seismic activity along a system of faults that suddenly trigger slope failure; (2) change in slope stability engendered by a change in porewater pressure caused by a sea-level rise; (3) change in pore-water pressure affected by methane gas hydrates stability change, likely engendered by a temperature increase and or pore pressure owing to sea-level change; and (4) seepage of continental groundwater at specific depths.
Seismic activity triggering undersea mass movement Masson et al. (2006), in their review paper of the processes and triggers of submarine landslides, stated that 'historical evidence suggests that the majority of large submarine landslides are triggered by earthquakes'. These authors also discussed several other processes that trigger undersea slides, some of which will be discussed in the subsequent sections.
Seismic records for the coast of western Victoria and the southeast of South Australia indicate a lack of substantial  small escarpments and therefore must still be active or were recently active. We must conclude, however, that examination of the bathymetric maps (Figures 4b, 6 and 8b) for the area discussed in this paper provides inconclusive information of a possible link between the top of some of the scars and fault lines. If a link did exist, large lineaments with extensive scars and slides running latitudinally would be visible on the sea floor.

Change in slope stability engendered by a sealevel rise
The volume and rate of post-last glacial rapid sea-level rise have been demonstrated to be sufficient to trigger submarine sediment failures by inducing seismic activity (Smith et al., 2013). While the role of a post last-glacial sea-level rise has been 'infrequently cited as a cause of such events/ features', Smith et al. (2013) argued that such events should be considered as a trigger for seismic activity. Rapid sea-level rise could change pore-water pressure, and hence the bending stresses within the sediments, leading to a change in slope stability and eventually slope failure. Brothers et al. (2013) argued this case by mentioning that a load-induced flexure of the sea floor can engender substantial changes in pore-water pressure. Both groups of authors discussed the particularly impactful and important implications of meltwater pulses (MWP1A and IB) that occurred at 14.08-13.61 ka and 11.4-11.1 ka, with MPW1A sea-level rising by 16 ± 2 m within 350 years (Camoin et al., 2012). Such a rapid change in water load on the sea floor, especially over the shelf/slope, would have had a significant effect on sediment porosity, in addition to possibly a change in sedimentation in shallow depths as recognised on the Lacepede Shelf by Nash et al. (2018).
Although the underwater slides described in this study are relatively small (less than 20 km 3 ), they may have been triggered by a change in pore pressure within the slope sediments by loading as a result of transgression, or by groundwater processes (see discussion below).

Change in the stability of methane gas hydrate within the sediments
A change in the nature of methane gas hydrate (also called clathrate), which normally occurs under high hydrostatic pressure (>5 bars) and low bottom-water temperature (<7 C) at the contact of the sea floor and ambient water, could potentially affect sediment cohesion. It is possible that methane gas hydrates would have been present in the area of investigation off the South Australian coast (Bonney Upwelling: Schahinger, 1987). Of interest is that this area is affected by seasonal upwelling linked to the Flinders Current and was discussed in Middleton and Bye (2007) and more specifically by Currie et al. (2012), who looked at fish productivity in the vicinity of the Bonney Canyon during a period of upwelling. Upwelling would cause primary productivity to increase significantly that, in turn, would cause an enhancement of organic carbon deposition on the sea floor. Eventually, this excess organic carbon could potentially turn into methane gas hydrates if preservation conditions were suitable. This was further investigated by K€ ampf and Kavi (2017), who used satellite-derived monthly estimates of chlorophyll a concentrations for the period 2003-2015 offshore southern Australia (Bonney Upwelling), to estimate the spatial extent of phytoplankton blooms in which concentrations reached >1.2 mg/m 3 over an area of $3000-6000 km 2 during upwelling events, but dropped to between 0.8 and 1.2 mg/m 3 over a much larger area (5000-10 000 km 2 ). During glacial periods, this region likely experienced much stronger upwellings owing to the enhanced winds, with the westerlies positioned farther northward (closer to Moros et al., 2021). It is possible that the presence of a substantial amount of organic carbon was therefore generated in the area and could have eventually been 'transformed' into methane once settled on the sea floor. It is important to note that sea-surface temperatures (SSTs) during the LGM offshore Kangaroo Island (documented in two adjacent cores by Calvo et al., 2007;Lopes dos Santos et al., 2013;Moros et al., 2021) were $9 C lower than today and that low SSTs persisted through the entire glacial period. Such low temperatures imply that methane gas hydrates could potentially have formed and thus remained stable at low depths such as $200 m where water pressure would have been >5 bars and temperatures well below 7 C. Middleton and Bye (2007, figure 26) obtained, for various periods between 1980 and 2003, a range of water temperatures between 11.4 and 13.2 C at 200 m in CTD profiles offshore Kangaroo Island, and at such depth we can confidently assume that temperatures would have been <7 C during the LGM. Furthermore, sites at 200 m water depth at the LGM are today 325 m deep. Hence, if methane gas hydrates were present in the sediments at 200 m water depth or deeper, it would have become unstable and 'bubbled out' during the rapid postglacial sea-surface temperature increase.
Lopes dos Santos et al. (2012) also examined a marine core offshore Kangaroo Island from which paleo-SSTs were inferred and examined the changes in productivity at sea over the last glacial/interglacial cycle (viz. the last 125 ka). These authors examined the d 13 C values of planktic foraminifera as well as total organic carbon and alkenone accumulation rates, and found that, at the LGM, productivity was at its highest between 0.1 and 0.2 g m À2 year À1 (confirming prior findings by Gingele and De Deckker (2005) for the same core) and suggested that this was mainly due to stronger westerly winds that would also bring more eolian dust to the area (De Deckker et al., 2019Moros et al., 2021), which, in combination with upwelled nutrientrich waters, would have enhanced productivity near the sea surface. It is not clear if such levels of productivity would have been sufficient to generate enough organic carbon that could become 'transformed' into methane.
The potential clue for this is that examination of some of the cores taken at great depths (where the temperature is <4 C) offshore Kangaroo Island during the AUSCAN cruise (Hill & De Deckker, 2004) did not reveal any sign of methane clathrate, any evidence of disturbance or any gaps in the sedimentary layering that would have been engendered by some degassing. Hence, slope failure caused by the change in stability of methane gas hydrate in the region examined here seems unlikely.

Continental groundwater discharge at sea
Continental groundwater discharge on the sea bed at specific points on the slope could affect pore pressure in sediments and therefore cause slope failure. This scenario has not been discussed before and will be examined in close detail here. Robb (1984) described such a phenomenon during periods of low sea-level offshore the New Jersey coast. He described that in this area, interstitial water could be recovered from below the continental shelf as far as 100 km from the coast.
There is already evidence for the presence of groundwater aquifers in the southeast of South Australia around Mount Gambier (Figure 1; Harrington et al., 2011;Love et al., 1993;Wood & Harrington, 2015) and western Victoria's Gambier Embayment (Bush, 2009). In both regions, substantial quantities of groundwater characteristically flow towards the sea. For example, at Piccaninnie Ponds, a site located <1 km upstream from the coast along an open channel, and proximal to the South Australian/ Victorian border, water discharge flows via an open channel to the ocean. The ponds are part of a karstic system that extends down >100 m below ground (Wood & Harrington, 2015, figure 1b, c). Bachmann (2016) claims that prior to European settlement ($1850), water that discharged from the karstic springs may have flowed eastward into the Glenelg estuary in adjacent Victoria.
At the Piccaninnie Ponds, the open channel flow rates from the unconfined Tertiary Limestone Aquifer vary from 12 ML per day to 27 ML per day (i.e. 0.14-0.31 m 3 /s). Although the flow is seasonal and greatest in winter (Wood & Harrington, 2015), it may have been much higher in the past prior to human water extraction. For the Glenelg River, the mean daily discharge at the gauging station closest to the ocean (Dartmoor À 45 km from Nelson; 44 years of record between 1974-2020: http://www.bom. gov.au/waterdata/) is 3.79 ± 2.42 Â 10 4 m 3 . This value is slightly larger than the discharge of water from Piccaninnie Ponds.
However, it is apparent that there are also extensive freshwater springs at sea on the continental shelf in karstic holes in the Gambier Limestone. Bush (2009) reported submarine springs exist along faults as reported by Sprigg (1952, figure 4.5) and Boutakoff (1963) at Cape Nelson and Port Macdonnell. Bush (2009) also mentioned that significant freshwater discharges at sea have been reported by fishermen, who have been known to replenish their water supplies from the surface of the ocean, approximately 2 miles offshore from Cape Nelson (P. Arkell pers. comm. in Bush, 2009, p. 139). The same phenomenon was reported by fishermen to J. Sherwood (pers. comm. to PDD in 2021), but none of these people reported exact locations for the springs at sea. These significant volumes of freshwater discharging from both on and offshore springs may drive seafloor instability in this area. These links are described in more detail further below.
The intrusion of marine water into submarine pore water occurs across the sea bed by SGD and submarine groundwater recharge (for a review, see Taniguchi et al., 2002). Hence, at Piccaninnie Ponds, Wood and Harrington (2015) measured water salinity at three different depths in the fully immersed cave and found that salinity varied seasonally and appeared to be affected by the height of the oceanic tide. Indeed, they showed that at 94 m water depth, salinity measured by electrical conductivity ranged over $3 weeks in January 2011 between 2400 and close to 3400 mS/cm (being equivalent to 2.4 to $3.4‰ salinity) (Wood & Harrington, 2015, figure 3). This implies that despite the ponds being only 600 m inland of the open coast, the flow of groundwater towards the ocean was so strong that there was little 'contamination' by marine waters. The same has been observed at the deep Blue Lake in Mount Gambier, which is $25 km from the coast, with freshwater present down to the bottom of the lake despite it being approximately 60 m below sea level (Somaratne et al., 2014). The general groundwater flow at Mount Gambier is to the south towards the ocean and the Gambier Embayment (Love et al., 1993).
Further investigations of SGD along the coastline southeast of South Australia were made by Lamontagne et al. (2015), who used a range of potential environmental tracers (temperature, salinity, 222 Rn, 223 Ra, 224 Ra, 226 Ra, 228 Ra and 4 He) that were measured in potential sources of SGD in seawater at and near the surface along a 45 km transect off the coast from Piccaninnie Ponds. These authors found that, along the coast, the SGD flux consisted mostly of recirculated seawater such that estimates of terrestrial freshwater discharge using Ra isotopes are difficult. This contrasts with the investigations of Knight et al. (2019), who found waters of very low salinities in the four exploration wells offshore (Table 1). Unfortunately, the timing of their collection of seawater samples along the transect at sea was not given; the area is well known for seasonal upwelling (Schahinger, 1987;Wijffels et al., 2018), and consequently, deep cold waters reaching the surface may have influenced the results. Nevertheless, Lamontagne et al. (2015) estimated that SGD at their location 45 km offshore is $1.2-4.6 m 3 /s, which is 10 times higher than the discharge from the spring-fed creeks in the area.
In 2014, Barandao carried out a preliminary study in an unpublished MSc thesis on the offshore extension of the hydrostratigraphy southeast of South Australia. He examined records from 10 offshore wells drilled for oil and gas exploration. His map was reproduced in Morgan et al. (2015, figure  2.18), and it shows the location of the estimated 'outcrop' of the confined aquifer that is beyond the continental shelf. More recently, Knight et al. (2019) pursued the original study of Barandao (2014) by examining a large seismic data set, combined with onshore and offshore bore-log geological profiles, which were used to explore the regional offshore hydrostratigraphy. In their case, these authors only examined four offshore wells (Figure 1), but also added data obtained from four onshore wells located near the South Australian/Victorian border opposite the important offshore Breaksea Reef exploratory well (Knight et al., 2019, figure 1), $20 km from the coast and $34 km from Piccaninnie Ponds, and were able to reconstruct salinity profiles within the four offshore wells that were derived from resistivity measurements. They found indication in the southern portion of their study area of pore water with total dissolved solids (TDS) of 2.2 g L À1 that can be found as far as 13.2 km offshore. Table 1 summarises the salinities encountered in the four offshore wells taken from Knight et al. (2019) and includes low salinities encountered at great depths (350-905 m) in the very porous lithologies (see ranges in different wells in Table 1) within the Lower Tertiary Confined Aquifer (Knight et al., 2019, figure S3); it is likely that these interstitial waters would eventually seep out as SGD along the continental slope. Morgan et al. (2015) developed a regional groundwater flow model for the southeast of South Australia and stated that 'the Lower Tertiary Confined Sand Aquifer, which comprises mainly the Dilwyn Sand aquifer in the Gambier Basin and the Renmark Sand in the Murray Basin, generally increases in thickness towards the south, being up to 800 m thick offshore to the south of Mount Gambier'. These sandy lithologies must be the ones listed by Knight et al. (2019) as well as in Table 1 and correspond to the flow and eventual SGD estimations of Knight et al. (2019). Since all these formations are also tilted and systemically increase in thickness towards the south, it is difficult to estimate exactly where the SGD are likely to occur along the continental slope.
The work of Pollock et al. (2002) and Pollock (2003) that determined the biostratigraphy of the four exploratory wells (Chama, Copa, Argonaut and Breaksea Reef) is particularly relevant to our examination and interpretation of the undersea slides offshore those wells. First, the formations containing the aquifers extend farther offshore where they become tilted over the continental slope. Figure 9 shows the extent of the Cretaceous/early Paleogene unconformity, and above it is the Dilwyn Formation. It is mostly within this formation that Knight et al. (2019) identified in the Breaksea Reef well the presence of continental waters, all of which are of low salinity. These authors identified the same feature (porous sandy sediments) in three other exploratory wells, and the depths as well as salinities are presented in Table 1. All those waters are found  in coarse sandstones, most of which belong to the Dilwyn Formation.
In the absence of seismic data for the mid to lower continental slope, we extrapolate the stratigraphic units mapped by Pollock (2003) to continue downslope, providing the conduit for groundwater flow, and potentially outcrop in areas that are now defined by head scarps at 1200, 1500 and 1800 m depths as observed in multibeam images (Figures 3-6).

Possible change in discharge at sea through time
It is not clear if groundwater sapping occurs today. However, during the Holocene and late Pleistocene, there were periods that were much wetter than today, and it may be that during those times of significant recharge of the groundwater tables, groundwater sapping would have predominantly occurred. Of note is the detailed and welldated record of two crater lakes in the volcanic district of western Victoria (which acted as 'gigantic rain gauges') that indicate these lakes overflowed during the period of 7.4-7.0 ka BP Wilkins et al., 2013). This was matched by more rainfall extrapolated from a speleothem in northern Tasmania (Xia et al., 2001) and lakes in northwest Victoria (Kemp et al., 2012). Rivers in southeast Australia also record enhanced water flows for the period 8-4 ka (Cohen & Nanson, 2007). Gingele et al. (2007) summed up the hydrological scenario by determining that the River Murray had experienced two periods of significantly high discharges: one coinciding with the high lake level phase discussed above (7.4-7.0 ka), and an earlier one peaking between 13.5 and 11.5 ka BP. As a consequence of this significant increase in rainfall in the region, we can speculate that the groundwater flows in the Gambier Embayment, both onshore and offshore, must also have been substantially enhanced compared with today. Hence, we postulate that the timing of the resulting larger SGD may have coincided with the initiation of the central to western-Otway Basin undersea slides.
Further to the Holocene peaks in precipitation, during the last interglacial the hydrological budget of a large part of Australia was also substantially enhanced. The best record of these conditions is in the Kati Thanda-Lake Eyre region where the lake was full of water between 128 and 110 ka and extended for 35 000 km 2 , up to three times its present surface area, and held 430 km 3 of water (DeVogel et al., 2004;Magee et al., 2004). At about the same time, in western Victoria, Lake Corangamite ($260 km 2 today when full) had also vastly increased in size ($1800 km 2 ; Currey, 1964, Dimmer, 1992, Edwards et al., 1996. Such huge hydrological changes would have significantly enhanced groundwater flows and associated SGDs, likely resulting in undersea slide activity at different depths on the continental slope.

Formation of canyons and associated undersea slides
The alignments of most of the canyons of the central-west Otway Basin follow ancient regional lineaments. In addition, examination of many of the undersea slides, some of which occur in parallel with some of the canyons, provides clues on how several of the canyons may have formed. Figures 3b, 4b and 7b for east of the Nelson and Bocara canyons, and Figure 6 for east of the Rivoli Canyon, clearly show that many of the undersea slides are superimposed over one another but still follow some faults (see white arrows in Figure 7b) that could be linked to the regional lineaments postulated by Boutakoff (1963). Eventually, with time and additional sliding/erosion, such features may form precursors of canyons. Importantly, these accumulations of the slides do not commence at depths above 1200 m (Figures 3b, 4b and 11a), and consequently this indicates that the heads of these 'potential' canyons are not near the continental shelf edge and more importantly are not connected to rivers. It is likely also that some groundwater would have preferentially seeped at sea via the lineaments of Boutakoff (1963) and faults recognised by Boult et al. (2005), therefore acting as groundwater conduits.
We note, however, that some of the canyons are aligned to ancient river courses as perceived by Boutakoff (1963). Those canyons at greater depths as discussed above (Bonney East and Rivoli East) ( Figure 5) are different. Where they commence at depths below 3000 m, their courses do not follow the regional lineament trends, but below 4000 m water depth, their directions change, and they are aligned to the NE-SW trend as for the shallower canyons further up the slope. We do not have an explanation for such a phenomenon, but it is likely that these deeper canyons may be much older features than the shallower ones described herewith.
We note also that it is possible that continental groundwater may in fact use some of the ancient, infilled canyons as conduits to eventually seep on the continental slope and help incise the Rivoli Canyon (Figures 1 and 3a, b) by groundwater sapping, but at this stage it cannot be verified.
Instead, along the US Atlantic continental slope, ten Brink et al. (2009) had identified submarine slope failures caused by undersea slides, which they argued are caused by horizontal ground shaking. This appears not to be the case in our area.

Undersea slides as potential tsunamigenic features
SMFs are known to generate tsunamis of varying scales and impacts. The Great Banks earthquake that occurred in 1929 off the south coast of Nova Scotia is well documented, and it generated three tsunami waves with amplitudes of 3-8 m (Fine et al., 2005). In 1998, a slope failure of 5 km 3 of sediments offshore northern Papua New Guinea initiated a devastating tsunami that caused a great loss of life (>2200 people) on land and was triggered by an SMF (Tappin et al., 2008). The Storegga Slide, which is the world's most spectacular reported SMF event, occurred 8200 years ago on the mid-Norwegian margin (Mienert et al., 2005). That SMF displaced >3000 km 3 of sediment and affected the coasts of Scandinavia, the British Isles and at least one site in Greenland (Talling et al., 2014;Williams, 2014). This event is considered to even have triggered a significant climatic shift in the northern hemisphere (Rohling & P€ alike, 2005) with a substantial temperature drop recorded in Greenland ice cores (Alley & Ag ustsd ottir, 2005). More recently, Sultan et al. (2010) studied the cause of a tsunamic wave that occurred adjacent to the airport of Nice in southern France and identified that it resulted from a submarine landslide [sic] in the Var canyon located just offshore. This tsunami resulted in several casualties and infrastructural damage (Sultan et al., 2010).
Recently, Chang et al. (2021) described, using highresolution bathymetric data, the occurrence of small but frequent undersea slides in the upper submarine slopes of volcanic islands in the central Azores. These authors also stated that these slides are potential hazards locally because of their high frequency.
In Australia, Puga-Bernab eu et al. (2013) used highresolution multibeam bathymetry and seismic profiles in the Noggin Passage region in the upper-slope offshore Cairns in northeastern Australia to characterise the Noggin block. They estimated this block to be 5.3 km 3 and a precursor to a larger SMF, and estimated that if this block were to collapse and slide down, it could generate a tsunami wave with a height of 7-11 m at its inception point and would reach a maximum run-up height of 16.5-24.8 m at the nearby coast, although this remains unclear because of the complexity of reef morphologies in this part of the Great Barrier Reef.
More recently, the seafloor along the southeastern continental margin of Australia was intensively mapped to search for evidence of undersea slides. Clarke et al. (2019) determined the presence of 260 submarine landslides [sic] between depths of 400 and 3500 m. The studied area covers about 1500 km and started opposite the northern tip of Fraser Island (24 39 0 S) down to Jervis Bay in the south (35 20 0 S). Of these, Clarke et al. (2019) estimated that 36 of such slides could have produced tsunamis with flood depths of !5 m at the coastline, assuming a downslope velocity of 20 m s À1 . However, these authors concluded that there is no evidence of any large tsunami-generated slides during the Holocene. Finally, Clarke et al. (2019) argued that the majority of these undersea slide movements would have been generated by earthquakes.
In another study, Clarke et al. (2016) examined in detail five undersea scars between Noosa (26 40 0 S) and Byron Bay (28 38 0 S), also along the coast of eastern Australia, and supporting evidence of sedimentary core analyses. The cores showed evidence of gaps in sediment accumulation and therefore unconformable sediments, and 14 C dates confirmed that these slides would have occurred less than 25 000 years ago. They also estimated the volume of the sediments displaced in the five slides, ranging between 0.4 and 3 km 3 , for slides approximately 2-8 km long and in water depths between 488 and 1167 m. These authors concluded that they favoured an earthquake mechanism for the trigger of these undersea slides.
In all the studies listed above, a variety of assumptions were made to estimate the amplitude of the tsunamis that could be generated from these undersea slides. Hence, we are not attempting to estimate the possible amplitudes for tsunami generation for the two slides examined in this study, but we note that the volumes that would have been displaced by the two slides (13 and 18.6 km 3 ) are much larger than the slides estimated from the Great Barrier Reef (Puga-Bernab eu et al., 2013) and those from the Tasman Sea (Clarke et al., 2019). As such, the larger volumes and depths of the slides described herein are consistent with the generation of tsunamis, assuming the slides represent a single failure event.

Conclusions
We have illustrated and described the morphology of several canyons that occur on the continental slope offshore the coast of western Victorian and the southeast of South Australia. These canyons are still active today, especially close to their heads where sinuous channels are clearly visible and are not infilled with sediment. Significant groundwater seepage is widespread in this region, although high-resolution bathymetric data and shallow sub-bottom profile data are not available across the continental shelf to verify the presence of paleochannel courses.
Nevertheless, it appears that many of the canyons, may be linked to ancient river courses of the Glenelg River as postulated close to 60 years ago by Boutakoff (1963). A detailed examination of the tectonism of the Portland to Cape Jaffa area would enable us to delineate the timing of the changes in the courses of the Glenelg River and link canyon developments to specific tectonic events.
Most of the younger canyons that occur in the upper slope are aligned parallel to the regional trend of lineaments that may be endorsed by faults. These lineaments follow an extensional regime that commenced during early Cambrian rifting and was re-invigorated during the breakup of Gondwana.
Numerous undersea slides in the canyon regions occur at specific depths ($1200, $1500 and $1800 m), and seismic data indicate that these depths likely coincide with horizons where groundwater seepage on the slope from the Dilwyn Formation could occur. Continental water seepage through these strata could engender groundwater sapping that has the potential to then drive the slipping of slope sediments. The undersea slides described herein need not have been caused by earthquakes or by a change in stability of methane gas hydrates in the sediments that elsewhere are considered to be the trigger for submarine mass failure (Haq, 2001;Maslin et al., 2010;Matsumoto, 2001). Rather, changes in SGD may have instead initiated the SMFs described herein. This process had already been introduced as far back as 1939 by Johnson as a result of groundwater sapping but had not been pursued by other authors. We re-invigorate that phenomenon here. More recently, Dugan and Flemings (2000) described the importance of fluid pressure identified along the mid-Atlantic continental slope that can engender slope instability.
We observed also that many of the undersea slides are superimposed, and postulate that these may also be the onset of canyon formation. Again, many of the slides are aligned with the NE-SW lineaments that control canyon alignment. At least two much older and deeper canyons were found, and the alignment of their upstream reaches (below 3000 m water depth) is inconsistent with regional lineaments, but by 4000 m water depths, they align with the regional lineament trend, consistent with the alignment of the younger canyons on the upper slope. These deeper canyons extend well over 100 km down to abyssal depths and therefore act as conduits of sediments and organic matter down to great depths.
We postulate also that some of the large undersea slides may have potentially triggered tsunamis because the volume of displaced sediments ($13 and 18.6 km 3 ) is substantially larger than tsunamigenic slides identified and modelled along Australia's east coast. Note 1. Before commencing a description of the 'undersea slides' and discussing their formation, it is necessary to justify our nomenclature. Because these features occur under the sea, it is thought inappropriate to label them as 'undersea land slides', although many authors refer to these features as such.