Revisiting the tsunamigenic volcanic flank collapse of Fogo Island in the Cape Verdes, offshore West Africa

Abstract Volcanic archipelagos are a source of numerous on- and offshore geohazards, including explosive eruptions and potentially tsunamigenic large-scale flank collapses. Fogo Island in the southern Cape Verdes is one of the most active volcanoes in the world, making it both prone to collapse (as evidenced by the c. 73 ka Monte Amarelo volcanic flank collapse), and a source of widely distributed tephra and volcanic material. The offshore distribution of the Monte Amarelo debris avalanche deposits and the surrounding volcaniclastic apron were previously mapped using only medium-resolution bathymetric data. Here, using recently acquired, higher-resolution acoustic data, we revisit Fogo's flank collapse and find evidence suggesting that the deposition of hummocky volcanic debris originating from the failed eastern flank most likely triggered the contemporaneous, multi-phase failure of pre-existing seafloor sediments. Additionally, we identify, for the first time, multiple mass-transport deposits in the southern part of the volcaniclastic apron of Fogo and Santiago based on the presence of acoustically chaotic deposits in parametric echo sounder data and volcaniclastic turbiditic sands in recovered cores. These preliminary findings indicate a long and complex history of instability on the southern slopes of Fogo and suggest that Fogo may have experienced multiple flank collapses.

Abstract: Volcanic archipelagos are a source of numerous on-and offshore geohazards, including explosive eruptions and potentially tsunamigenic large-scale flank collapses. Fogo Island in the southern Cape Verdes is one of the most active volcanoes in the world, making it both prone to collapse (as evidenced by the c. 73 ka Monte Amarelo volcanic flank collapse), and a source of widely distributed tephra and volcanic material. The offshore distribution of the Monte Amarelo debris avalanche deposits and the surrounding volcaniclastic apron were previously mapped using only medium-resolution bathymetric data. Here, using recently acquired, higherresolution acoustic data, we revisit Fogo's flank collapse and find evidence suggesting that the deposition of hummocky volcanic debris originating from the failed eastern flank most likely triggered the contemporaneous, multi-phase failure of pre-existing seafloor sediments. Additionally, we identify, for the first time, multiple mass-transport deposits in the southern part of the volcaniclastic apron of Fogo and Santiago based on the presence of acoustically chaotic deposits in parametric echo sounder data and volcaniclastic turbiditic sands in recovered cores. These preliminary findings indicate a long and complex history of instability on the southern slopes of Fogo and suggest that Fogo may have experienced multiple flank collapses.
Volcanic archipelagos are home to numerous onand offshore hazards such as changes in sea-level, storms, volcanic eruptions, slope instabilities, large flank collapses and tsunamis (Casalbore 2018). These events can have devastating consequences for people, nature and infrastructure, both onshore and offshore. A variety of internal and external factors can precondition and trigger flank instabilities. These include: dyke and sill intrusions; volcanic eruptions and tremor; earthquakes; flank over-steepening; the weight of new volcanic material on the island flanks; weakening of the volcanic edifice by weathering and hydrothermal activity and, for smaller landslides, by the effects of wave, wind and storm activity (e.g. Siebert 1984;Begét and Kienle 1992;McGuire 1996McGuire , 2003Murray and Voight 1996;Tibaldi 2001;Cervelli et al. 2002;Casalbore et al. 2011Casalbore et al. , 2015Gross et al. 2014;Clare et al. 2018). Such factors occur over timescales ranging from seconds to thousands of years and are globally widespread (McGuire 1996;Blahut et al. 2019).
The volume of volcanic flank collapses varies significantly, but can be as large as thousands of cubic kilometres, for example, in the Hawaiian archipelago (Moore et al. 1989). Many such events have volumes in the order of tens to hundreds of cubic kilometres, for example at Nisyros Volcano in the Aegean Sea (Tibaldi et al. 2008;Livanos et al. 2013), in the Lesser Antilles Arc (Lebas et al. 2011;Le Friant et al. 2015;Brunet et al. 2016), in the Canary Islands (Krastel et al. 2001;Masson et al. 2002;León et al. 2017) and in the Cape Verde Islands (Masson et al. 2008). However, even comparatively small-volume volcanic flank collapses, such as the 0.22-0.3 km 3 Anak Krakatau flank collapse in December 2018 (Grilli et al. 2019), may result in catastrophic tsunamis. The hazard potential of such flank collapses is widely recognized, but the magnitude, and therefore hazard potential, of the tsunamis that can be triggered by flank collapses is heavily debated (e.g. Moore and Moore 1984;McMurtry et al. 2004;Watt et al. 2012a;Goff et al. 2014;Ramalho et al. 2015;Paris et al. 2018).
Other factors that can contribute to the instability of slopes offshore are the deposition of centimetreto decimetre-thick discrete layers of volcanic ash across a wide region and the presence of buried turbidites in the volcanic apron. Studies in lacustrine settings (e.g. Wiemer et al. 2015;Moernaut et al. 2019) and on active offshore margins (e.g. Harders et al. 2010;Lafuerza et al. 2014;Hornbach et al. 2015;Kuhlmann et al. 2016;Sammartini et al. 2018) have indicated a relationship between tephra layers or turbidites and slide failure planes. Although the exact nature of this relationship is disputed (Wiemer and Kopf 2016), it is thought that tephra could behave as 'weak layers'layers of inherently lower strength than adjacent layers, which are thereby prone to failure (Locat et al. 2014). Using core logging, sedimentological and geotechnical data from the IODP Expedition 340, Lafuerza et al. (2014) showed that low hydraulic conductivity of hemipelagic sediments offshore Martinique (Lesser Antilles) could cause low rates of dewatering in turbidites and tephra layers, allowing excess pore fluid pressures to persist at depth. Moreover, Hornbach et al. (2015) suggested that even small changes in the stress regime of these layers, such as that resulting from regional strain and grain reorganization during the compaction of sediments, might trigger motion. Effectively, results from IODP 340 showed that sand layers (i.e. tephra ash and turbidites), as well as boundaries between sand and mud layers, may act as multiple decollement surfaces that promote and enhance the mobility of landslide deposits (Le Friant et al. 2015).

Geological setting
The Cape Verdes archipelago, offshore West Africa (Fig. 1), is the surface expression of a mantle hotspot (Crough 1978;Holm et al. 2008;Ramalho et al. 2010). The Island of Fogo, in the southern part of the archipelago, is one of the most active oceanic intraplate volcanoes in the world, having erupted 28 times in the last 520 years (Ribeiro 1960;Torres et al. 1997;González et al. 2015). Such volcanically active areas are well known to be associated with seismicity and the southern Cape Verdes are no different, recording frequent volcano-tectonic earthquakes (Grevemeyer et al. 2010;Faria and Fonseca 2014;Vales et al. 2014). Sediment cores collected in the region provide evidence of at least 43 large, explosive eruptions in the area in the last 150 kyr (Eisele et al. 2015), attesting to Fogo's vigorous volcanic activity. Such frequent volcanic activity, along with the accumulation of volcanic deposits on the submarine flanks, means that Fogo's flanks are potentially unstable and prone to collapsing.
A semi-circular depression (Borderia), up to 1 km high and open to the east on central Fogo, was interpreted by Day et al. (1999) as the scar of a large flank collapse, referred to as 'Monte Amarelo' (Fig. 2). Other authors, however, interpret the same morphology as two partially overlapping volcanic calderas that were later cut by a flank collapse that affected the eastern portion of the edifice (Brum da Torres et al. 1997;Madeira et al. 2008 Ramalho et al. (2015) proposed that the collapse and ensuing tsunami took place between 65 and 84 ka, with a most probable age of 73 + 7 ka. This agrees, within uncertainty, with the age recently reported by Maderia et al. Objectives A detailed analysis of the distribution of the landslide deposits and failure mechanism are critical for constraining the hazard linked to the collapse of a volcanic flank. Moreover, this characterization is crucial for tsunami hazard modelling, given that the mode, volume and run-out of a collapse will have a profound effect on the resulting tsunami waves (Grilli et al. 1997;Abadie et al. 2012;Watt et al. 2012a). The offshore distribution of the Monte Amarelo flank-collapse debris (i.e. debris avalanche deposits) and the surrounding volcaniclastic apron were previously mapped using only medium-resolution (100 m grid cell size) multibeam bathymetric data (Masson et al. 2008). Consequently, the distribution of the debris avalanche deposits and of any additional slope instabilities in the area are still poorly constrained. As Fogo's volcanic flank collapse likely triggered a mega-tsunami with a recognizable impact on the adjacent islands' coastlines (Paris et al. 2011(Paris et al. , 2018Ramalho et al. 2015;Madeira et al. 2019), a full characterization of this landslide and its related volume is crucial for improving numerical models of tsunami generation, propagation and inundation, and for constraining the hazard potential associated with large, tsunamigenic volcanic flank collapses.
Here, using recently acquired multibeam bathymetric data (50 m grid cell size) in conjunction with parametric sediment echo-sounder data and sediment gravity cores, we revisit the Monte Amarelo volcanic flank collapse and consider general slope stability in the southern distal region of the volcanic apron. We aim to (1) map out the lateral extent and characterize the acoustic nature of the Monte Amarelo debris avalanche deposits; (2) identify possible additional landslides on the slopes south of the islands of Fogo and Santiago; and (3) consider the related preconditioning and triggering processes.

Data and methodology
The data presented in this paper were collected during R/V Meteor cruise M155 (May-June 2019; Krastel et al. 2019) and are supplemented by multibeam bathymetric data collected during R/V Meteor cruise M80/3 (Hansteen et al. 2014). The bathymetric data were acquired using hullmounted Kongsberg EM120 (M80/3), EM122 (M155) and EM710 (M155) multibeam echosounders. The EM120/EM122 system has a swath coverage of up to 150°and a nominal sonar frequency of 12 kHz, and is designed to perform seabed mapping to full ocean depth. During cruise M155, the swath width was reduced to 120°in order to increase the quality and resolution of the data. The EM710 system has a swath coverage of up to 140°, and a nominal sonar frequency of 70-100 kHz. As such, the EM710 was used only in water depths less than 700 m close to the islands of Fogo and Santiago. All bathymetric data were filtered for outliers and manually edited.  from all cruises and multibeam systems are combined and gridded at 50 m. Parametric sediment echo-sounder data were collected using a parametric ATLAS DS-3/P70 system (Parasound). This system has an opening angle of 4°and operates at primary high frequencies of 18.5 and 22.5 kHz, resulting in a parametric low frequency of 4 kHz. The vertical resolution of this system is in the decimetre range. All depth scales on Parasound images presented in this study were calculated using a constant velocity of 1500 m s −1 .
Sediment cores were collected during cruise M155 using a gravity corer with tube lengths of 3 to 15 m. These cores were visually described onboard and smear slides were analysed to obtain further microscopic information about the core mineralogy, texture and composition. Dating and further geochemical and sedimentological analysis of these cores is ongoing and will form the basis of future work on the subject.

The Monte Amarelo debris avalanche deposits (MTD-A)
The Monte Amarelo deposits proximal to Fogo (MTD (mass-transport deposit)-A; distribution shown in Fig. 3a) are characterized by overlapping diffraction hyperbolae (Fig. 3b). This hummocky character is typical of debris avalanche deposits from volcanic flank collapses (Siebert 1984). The sedimentary drape covering MTD-A is relatively thin (,1.5 m thick) or not imaged, and is often characterized by two strong, positive reflections. Irregularly shaped blocks that are up to 100 m higher than the surrounding seafloor are present within the landslide debris (Fig. 2). Further away from Fogo, the hummocky topography is less prevalent in the bathymetric data, despite the presence of diffraction hyperbolae in the Parasound data. This highlights a decrease in the size of the hummocks with distance from the island. We note a progressive transition in the acoustic signature of the deposits from a hyperbolic facies with metre-scale acoustic penetration (Fig. 3b), to a mounded facies draped by c. 1.5 m of sediment characterized by the aforementioned prominent double reflections (Fig. 3d). In a few places, these double reflections are replaced by a succession of finely stratified layers (inset of Fig. 3c).
Distal deposits related to the Monte Amarelo volcanic flank collapse (MTD-B) Southwards, with increasing distance from Fogo, the acoustic character of the Monte Amarelo deposits changes from being hyperbolae-dominated to being characterized by an acoustically transparent/semitransparent facies ( Fig. 3d; hereafter referred to as MTD-B; distribution shown in Fig. 3a). In the northern reaches of MTD-B, the upper surface of the deposits is undulating, with metre-scale variations in depth over wavelengths of tens to hundreds of metres (Fig. 3d). The overlying sedimentary drape mantles the surface of the deposits and a strong internal reflector is imaged in places within the deposits (Fig. 3d). This internal reflector clearly contrasts with the overlying transparent facies, separating it from the lower semi-transparent facies. In the northern and western reaches of MTD-B, this internal reflector is either clearly imaged or not present at all, and the base of the lower part of MTD-B is not resolved (Fig. 3d).
In the eastern and southern reaches of MTD-B, the aforementioned internal reflector is continuous over a relatively long distance (Fig. 3e). In these areas, a thicker sedimentary drape (up to c. 4 m thick), containing multiple strong reflections alternating with relatively thin, transparent layers, overlies MTD-B. The upper surface of MTD-B is diffuse and undulating in parts. The base of MTD-B is sometimes marked by a prominent reflection that has a similar acoustic signature, although of lower amplitude, as the double reflector that overlies the deposits (e.g. Fig. 3e). A series of well-stratified sediments is imaged below the base of MTD-B (Fig. 3e). At the eastern lateral margin of MTD-B, the internal reflector within the slide deposits is more diffuse (Fig. 3e). The lowermost part of MTD-B remains transparent in nature, but remnant stratifications are imaged in some areas of the uppermost part. The eastern boundary is gradational over 2-3 km, and is marked by a progressive thinning of the transparent facies, and an increase in stratification within the deposits (Fig. 3e).
Constraining the thickness and lateral extent of MTD-B is challenging. In the northern reaches, where the strong internal reflector is imaged, the base of the deposits is not resolved by the acoustic system (Fig. 3c). This means that only the thickness of the upper transparent facies related to MTD-B (up to c. 7 m thick, but variable) can be estimated. In its eastern and southern extents, however, the bases of the upper and lower units of MTD-B are imaged, leading to thickness estimates of 4-6 m and 5-15 m for the upper and lower units, respectively (Fig. 3e). The large spacings between the Parasound profiles in the central and southern parts of the working area, south of Fogo and Santiago, mean that we cannot constrain the thickness of the deposits in these areas with certainty. In addition, MTD-B deposits might extend beyond the limits of the surveyed area, especially to the west and south (Fig. 3a).

Additional mass wasting events on the shallow slopes south of Fogo and Santiago
Visual analysis of gravity cores taken in the southern part of the volcaniclastic apron of Fogo and Santiago reveals mud-and nanofossil-rich facies interbedded with multiple sand units that are defined by parallel and cross-laminated, seldom normally graded beds (Fig. 4). These sandy intervals generally range from fineto coarse-sand deposits, and are typically of centimetre to decimetre scale (Fig. 4b). Microscopic analyses show that these shallow sandy turbidites are predominantly of volcaniclastic (volcanic lithics that are mostly lava fragments, tachylitic to brown glass, crystal fragments), and/or mixed volcaniclasticbioclastic composition, rather than solely bioclastic.
Multibeam-bathymetry and sediment echo-sounder data on the shallow slopes south of Fogo and Santiago reveal multiple morphological steps and several MTDs (Fig. 5). These additional MTDs are characterized by acoustically transparent and/or hyperbolic facies and occur both above and below MTD-B. Several of these MTDs are exposed at the surface and have remobilized previously failed sediments, including those of MTD-B (e.g. Figs 5 & 6). However, as for MTD-B, their full lateral extent is generally poorly constrained due to large profile spacings and their possible continuation outside the surveyed area. The morphological steps vary in length, height, strike and orientation (Fig. 5a), but are typically near vertical (Fig. 5c, d). In the following section, we describe the character of, and relation between, several of these features in more detail.
Tectonic escarpments along the western and eastern margins. A series of escarpments is imaged to the west, along the western margin of MTD-B, and stretching between 40 and 110 km south of Fogo (Fig. 5a, b). The most prominent escarpment is orientated roughly north-south and extends across c. 55 km. It is c. 50 m high and near vertical (Fig. 5c). Monte Amarelo MTD-B deposits are imaged on the eastern side (footwall) of this step. On the western (hanging-wall) side, the sediments are characterized by a semi-transparent facies with a hummocky upper surface (hereafter referred to as MTD-X; Fig. 5c). In places, a pair of strong reflectors is identified within the sedimentary drape that overlies MTD-X. Proximal to the escarpment, the base of MTD-X cannot be resolved. Further west of the scarp, however, stratified sediment is imaged below the base of MTD-X (inset 1 of Fig. 5c). The lateral extent of MTD-X is constrained to both the north and east by the morphological steps (Fig. 5b, c), resulting in a minimum area of 900 km 2 for MTD-X.
Southern Scour Complex. A prominent, elongated scour-shaped feature is evident in the central part of the southern distal region (hereafter referred to as the Southern Scour Complex (SSC), Fig. 6a). This feature is c. 60 km long and covers an area of c. 340 km 2 . The western and eastern sides of the SSC are constrained by up to 40 and 55 m high escarpments, respectively (Fig. 6). In Parasound data crossing the complex, acoustically transparent facies with a strong internal reflector characterizing the Monte Amarelo MTD-B are evident on both the eastern and western sides of the SSC (Fig. 6b & c). In the central part of the SSC (Fig. 6c), acoustically transparent MTDs with some internal structure are covered by c. 3 m of stratified sediment. In the northern part, the material within the SSC is characterized by a semi-transparent facies with a hummocky upper surface, and the deposits are largely exposed at the surface (Fig. 6b). An additional instability is visible on the eastern side of the SSC (hereafter referred to as the Eastern Scour (ES) ; Fig. 6d). The ES reaches a thickness of up to 35 m thick in its centre and covers an area of c. 120 km 2 . Parasound data across the ES show that this failure primarily remobilized the Monte Amarelo MTD-B deposits in this region along the same basal glide plane as MTD-B (Fig. 6d). To the south, the SCC is fan-shaped, with fingers of unfailed stratified sediments standing up to c. 20 m above failed sediment (Fig. 6d). Semitransparent deposits characteristic of MTD-B are imaged on the tops of these stratified fingers.

Discussion
The two-fold nature of Fogo's Monte Amarelo flank collapse Based on the difference in their acoustic character, we distinguish between two main types of deposits related to the Monte Amarelo volcanic flank collapse: hummocky, debris avalanche deposits proximal to Fogo (MTD-A), and finer-grained, acoustically transparent landslide deposits (MTD-B) at greater distances from Fogo that were previously not recognized (Fig. 3). The blocky debris avalanche deposits of MTD-A cover a surface area of 3180 km 2 . If we also include the region in which the size of the hummocks decreases with distance from the source (marked by the yellow shaded region in Fig. 3a), this increases the total area of MTD-A to 6820 km 2 ;  more than four times the previous estimate (c. 1470 km 2 ; Masson et al. 2008). As the base of the volcanic debris avalanche deposits is not resolved by the Parasound data, we cannot, at this stage, revisit estimates of the volume involved. Monte Amarelo MTD-B is characterized by an acoustically transparent to semi-transparent facies that is overlain by a series of two to four strong reflectors (Figs 3, 5c & 6b-d). A prominent, internal reflection is clearly observed in places within the deposits; separating MTD-B into two main parts (Fig. 3d, e). The uppermost section is of relatively constant thickness (4-6 m). Contrastingly the thickness of the lower part of MTD-B is highly variable (up to 15 m thick). Interestingly, the upper and lower parts of MTD-B have a similar areal distribution within the volcaniclastic apron; covering a minimum of 18 400 km 2 (upper) and 19 500 km 2 (lower) (Fig. 3a). This implies minimum volumes of 92 km 3 (upper) and 195 km 3 (lower), assuming average thicknesses of 5 m (upper) and 10 m (lower). Their similar distribution implies that the upper and lower parts of MTD-B might share the same source or, at least, be genetically linked. We therefore infer that the upper and lower part of MTD-B might have been emplaced (almost) contemporaneously. The absolute timing of these two phases, however, remains unclear. Of particular interest is whether these two failures occurred as two phases of the same event, or as two distinct events separated by some time. Further analytical work on the sediment cores will help to reveal more information about the nature of the prominent internal reflection, along with the relative timing of these two depositional episodes.
A similar two-fold nature of volcanic flankcollapse depositshummocky debris avalanche deposits accompanied by acoustically transparent to seismically chaotic deposits with a comparatively smooth upper surfacehas also been reported for flank collapses at other locations, including in the Lesser Antilles (Watt et al. 2012a, b;Le Friant et al. 2015;Brunet et al. 2016), at La Réunion (Indian Ocean; Lebas et al. 2018) and at Ritter Island (Papua New Guinea; Karstens et al. 2019;Watt et al. 2019). Drilled cores retrieved during IODP Expedition 340 in the Lesser Antilles indicated that widespread, seismically chaotic deposits (interpreted as equivalent to our MTD-B) primarily consist of hemipelagic mud interbedded with a combination of tephra, volcaniclastic layers, or bioclastic turbiditic deposits, which have undergone varying degrees of deformation (Le Friant et al. 2015;Brunet et al. 2016). To explain these findings, Le Friant et al. (2015) proposed a failure model where the loading of seafloor sediment by volcanic debris avalanche deposits triggered sediment destabilization and progressive downslope-propagating failure along a decollement. In this way, the deformation can propagate great distances away from the flank collapse, affecting seafloor sediments that were otherwise stable (Le Friant et al. 2015). The acoustic character of our MTD-B bears a strong resemblance to the seismically chaotic deposits in the Lesser Antilles. We, therefore, interpret MTD-B to be the result of the failure of pre-existing seafloor sediments following the loading of the Monte Amarelo debris avalanche deposits (MTD-A).

A history of mass wasting and remobilization in the southern Cape Verdes
Repeated mass-wasting events at Fogo? Preliminary stratigraphic correlations based on visual similarities of cores in the southern distal part of the working area indicate at least seven volcaniclastic or mixed volcaniclastic-bioclastic sandy turbidite layers above and below the Monte Amarelo flank-collapse deposits (Fig. 4). These turbidite layers are dominated by mafic glass, crystals and lava fragments, which, based on their petrography, suggests that they originated from Fogo volcano. This indicates that smaller mass-wasting events may have occurred at Fogo in addition to the Monte Amarelo volcanic flank collapse, as also recently suggested by Marques et al. (2019). Correlation and origin of these volcaniclastic deposits will, however, be verified by future analytical work on the sediment cores.
Regional tectonic and volcanic influences south of Fogo? The multiple fault-related escarpments and MTDs identified in the acoustic data south of Fogo further highlight a long and complex history of instability and slope failure in the region (Fig. 5). Ramalho et al. (2010) found that the island of Santiago has uplifted at a rate of c. 100 m Ma −1 over the past 4 Ma, and attributed this uplift to a combination of magmatic intrusions under or within the island edifice, together with episodic swell-wide uplift that affected the whole archipelago. Although geologically recent uplift has not been reported at Fogo, the neighbouring island of Brava has experienced one of the most dramatic intrusion-related uplift trends of any ocean island in the world, with up to 400 m of uplift in the last 1.8 Ma (Madeira et al. 2010). The presence of such widespread uplift across the archipelagoand particularly around the southern Cape Verdesthus suggests that the series of escarpments and landslides observed on the slopes south of Fogo may be the surface expression of a combination of regional tectonics, associated with the growth of the Cape Verdes hotspot swell, and crustal intrusions in the vicinity of the islands. As several of the exposed faults have landslide deposits on their hanging wall side, we interpret that these MTDs are most likely the results of movement along the faults and, as such, the result of neotectonic activity.
The SSC, which is exposed at the seafloor, postdates the Monte Amarelo flank collapse (Fig. 6). Monte Amarelo MTD-B deposits are imaged on the footwall east of the SSC (Fig. 6b-d), and on the tops of the unfailed, stratified fingers at the southern extent of the SSC (Fig. 6d). This indicates that the SSC formed after the Monte Amarelo flank collapse, remobilizing MTD-B. In the northern and eastern parts of the SSC, acoustically transparent mass-wasting deposits are exposed at the seafloor, highlighting more recent mass wasting and remobilization of MTD-B (Fig. 6b). We interpret the more recent mass-wasting deposits in the northern and eastern parts of the SSC as evidence of retrogressive development of the SSC in these directions (Fig. 6e).
The presence of fault-related escarpments on the slopes south of Fogo, frequent volcanic episodes at Fogo (Ribeiro 1960;Torres et al. 1997), and ongoing uplift at the neighbouring islands of Santiago and Brava (Madeira et al. 2010;Ramalho et al. 2010), together imply the occurrence of frequent earthquakes in the past. The volcanic activity at Fogo has resulted in the widespread deposition of numerous tephra layers (Eisele et al. 2015). Subsequent earthquake shaking may have resulted in liquefaction of overlying layers, which, together with sedimentary overpressure following further sediment deposition, can help to precondition the shallow slopes for failure (Moernaut et al. 2019). Even small changes in the stress regime of tephra layers have been found to be able to trigger failure (Hornbach et al. 2015). Consequently, it is possible that some of the MTDs observed on the shallowly dipping slopes south of Fogo, and unrelated to escarpments, may result from such a process. Future work, particularly on the sediment cores collected during cruise M155, will help to shed light on the sources and processes responsible for the MTDs mapped on the shallow slopes south of Fogo and Santiago.

Implications and conclusions
In this study, we revisited the Monte Amarelo volcanic flank collapse of Fogo Island and found that the deposition of the debris avalanche material may have triggered subsequent failures of pre-existing seafloor sediments in (at least) two phases. This is similar to what has been observed in the Lesser Antilles (Le Friant et al. 2015;Brunet et al. 2016), at La Réunion (Lebas et al. 2018) and at Ritter Island Watt et al. 2019). It is not yet clear whether these two phases of seafloor sediment failure were synchronous as a result of a single flank collapse, or if they reflect multiple, distinct events. The question of whether the main flank collapse occurred as a single or as multiple events is of utmost importance for tsunami modelling because the volume and timing of individual failures are the main factors controlling the tsunamigenic potential (Løvholt et al. 2015). Water depth and landslide kinematics play a key role in controlling the tsunamigenic potential of a subaqueous landslide (Watts et al. 2000;Ward 2001;Watt et al. 2012a;Harbitz et al. 2014). As MTD-B occurred in depths exceeding 3000 m, we consider that its influence on the resulting tsunami magnitude was negligible, and that only the hummocky debris avalanche deposits (MTD-A) contributed to the mega-tsunami that inundated nearby islands.
Multibeam bathymetric, sediment echo-sounder and sediment gravity core data acquired during cruise M155 of R/V Meteor provide seafloor evidence that shows, for the first time, the presence of multiple additional mass-wasting events on the shallow slopes south of Fogo. Turbidite sands recovered in the cores have a volcaniclastic or mixed volcaniclastic-bioclastic composition, which suggests that prevailing mass-transport processes in the region may be dominated by volcanic eruptions or volcanic mass-wasting events. However, it is also possible that some of the additional mass-wasting events are unrelated to flank collapse events or submarine failures, and could instead have been generated by sedimentary overpressure and/or by the presence of a mechanically weak layer such as tephra or turbidites. Further analytical work is required to verify the sources of these events on the shallow slopes south of the Cape Verdean archipelago. As these layers are thin, have limited lateral extent and occurred on low-gradient slopes in deep water, we consider their associated tsunamigenic potential to be low.