50,000 yr of recurrent volcaniclastic megabed deposition in the Marsili Basin, Tyrrhenian Sea

Megabeds, or “megaturbidites” (sensu Bouma, 1987), have been identified in a variety of deep-water basins and are interpreted to be the deposits of major catastrophic events. Known megabeds in the Mediterranean Sea include the Augias megaturbidite (Hieke, 2000; Rebesco et al., 2000; Polonia et al., 2016), the Balearic abyssal plain megabed (Rothwell et al., 1998; Cattaneo et al., 2020), and megabeds in the Vavilov Basin (Gamberi and Marani, 2009). Outside of the Mediterranean, megabeds have been identified in the Onnuri Basin offshore Korea (Cukur et al., 2021), the Horseshoe Basin offshore Portugal (Lebreiro et al., 1997), the Madeira abyssal plain (Weaver et al., 1992), and the Indus Fan (Bourget et al., 2013).

Megabeds are enigmatic because they are significantly thicker (several to tens of meters thick) than most turbidite beds (centimeters). Their exceptional thickness has stimulated questions about the processes responsible for their generation and if they represent single or multiple events. By comparison, the tsunamigenic 2002 Stromboli volcano flank collapse (Bonaccorso et al., 2003) is recorded by a turbidite less than 10 cm thick (Di Roberto et al., 2010). Similarly, the large slope failures of the Canary Islands (Hunt et al., 2013) and the 1929 Grand Banks M 7.6 earthquake deposits (Normandeau et al., 2019) are composed of individual beds of centimeter scale. A notable exception occurs in hadal basins along the Japan Trench, where acoustically transparent units up to 5 m thick have been linked to known earthquakes (Kioka et al., 2019; Schwestermann et al., 2020).

The Tyrrhenian Sea is a back-arc basin adjacent to a slow-moving convergent margin, where active volcanism and strong (but infrequent) earthquakes occur. The surrounding coastlines are vulnerable to tsunami submarine landslides (Urgeles and Camerlenghi, 2013), volcanic eruptions (Keller et al., 1978), and flank collapses (Bonaccorso et al., 2003; Chiocci et al., 2008). The catastrophic 1908 Messina earthquake and related tsunami caused at least 60,000 casualties (Billi et al., 2009). Furthermore, the southern Tyrrhenian Sea hosts the largest active volcanic edifice of Europe, the Marsili Seamount, which rises 3500 m from the seafloor to a depth of 489 m below sea level (Fig. 1). Recent volcanic and hydrothermal activity has raised concerns that the Marsili Seamount may pose risks to adjacent coastal communities (Gallotti et al., 2020). Nonetheless, there is little information on the past types, volumes, and hazard potential of activity in the recent sedimentary record of the basin floors adjacent to Marsili volcano.

With new high-resolution profiler data, we discovered a previously unknown series of at least four megabeds within the western Marsili Basin (Fig. 2). Considering the hazard potential of Marsili Seamount and surrounding seismically and volcanically active margins, our goals were to answer the following questions: (1) What are the nature and origin of the observed distinctive megabed units? (2) What relationship, if any, is there between the megabeds and the active Marsili volcano? (3) How large and how often do these events occur? (4) Are the megabeds single events or composed of amalgamated multiple events?

We combined geophysical data acquired as part of the CHIANTI project (Urgeles and CSIC-Unidad de Tecnología Marina, 2015) and core-derived data from Site 650 of Ocean Drilling Program (ODP) Leg 107 (Fig. 1; Kastens et al., 1987). In the Marsili Basin, ~920 km of high-resolution ATLAS Parasound parametric subbottom profiles were acquired with maximum vertical resolution of 0.2 ms (~15 cm). Parasound profiles imaged the megabeds as acoustically transparent units with mappable top and base reflectors (Figs. 2 and 3). We used the profile that crossed ODP Site 650 to define the four megabeds (Fig. 2), with megabed 1 (“M1”) being the shallowest and megabed 4 (“M4”) being the deepest. M1 was not penetrated by Site 650; therefore, we focused most of the analyses on M2–M4.

The study interval corresponded to the upper ~70 m at Site 650 (cores 1H–8H; Figs. 1–3; Kastens et al., 1987; Hieke et al., 1990). This interval is younger than the tephra dated to 60.3 ka in core 8H (Paterne et al., 1986; Calanchi et al., 1994). Ages of the megabeds were estimated from stratigraphic constraints and tephrachronology provided by Calanchi et al. (1994), assuming constant sedimentation rates in between age datums. Age errors were determined from intervals containing the age datums, and errors in the ages of individual megabeds were propagated to vary linearly in between datums.

To link lithology and acoustic facies, we computed a synthetic seismogram by convolving Site 650 P-wave velocity and bulk density data with a Ricker wavelet. The well tie is most robust in the uppermost section where the Parasound signal is strongest (Fig. 3). We used IHS KingdomSuite software for mapping, gridding, and computing thickness maps. For thickness maps, we applied a conservative velocity of 1500 m/s; therefore, calculated thicknesses are considered minimum estimates.

The acoustically transparent facies of the megabeds corresponds to a unit of mud, sandy mud, and muddy sand, which overlies a high-amplitude facies of coarse-grained, normally graded volcaniclastic sand (Figs. 2 and 3; Kastens et al., 1987), likely part of the same event. These acoustic and lithological characteristics make those megabeds more complex compared to megaturbidites (Rebesco et al., 2000; Cattaneo et al., 2020; Polonia et al., 2022), which present a more classical fining-upward sequence. In contrast to M2 and M4, the muddy top of M3 is highly deformed and is classified as a volcaniclastic debris flow (Kastens et al., 1987). Despite the starkly transparent seismic facies, cores revealed occasional thin interbeds of coarse material within M2 and M3 (Fig. 3, lithological log; Kastens et al., 1987). These interbeds were below the resolution of the subbottom profiler data; therefore, it is not clear how laterally continuous they are, nor what processes were responsible for these coarse layers, but they may also reflect different stages in the process(es) originating such volcaniclastic deposits.

The megabeds are widespread throughout the western Marsili Basin (Fig. 4). Maximum megabed thickness ranges from 9.8 to 24.7 m (Figs. 2–4; Table S1 in the Supplemental Material¹). Megabeds are thickest where they pond within antecedent lows and thinnest where they pinch out along the basin edges and local highs (Figs. 2 and 4). Estimated minimum volumes range from 1.3 km³ (M1) to 13.3 km³ (M4). However, addition of the basal layered facies, as suggested by ODP Site 650 core descriptions (Fig. 3, lithological log; Calanchi et al., 1994), would nearly double the volumes of M4 and M3 and may be key to the interpretation of the origin of the megabeds.

The Marsili megabeds record episodic major sediment flux events. The Marsili megabeds originated from large mass-gravity flows rather than settling out of the water column, as evidenced by the ponded geometries that pinch out toward the basin margins and thicken in the basin centers (Figs. 3 and 4). Correlations to sedimentary facies nevertheless indicate strong lateral heterogeneity, much larger than the diameter of sediment cores. Such heterogeneity could be explained by (1) multiple flow-entry points due to the complex network of canyons feeding the basin, (2) complex flow (reflection, localized conditions), and (3) sediment deformation at large horizontal scales, thus creating largely stretched bedding that could appear horizontal. Furthermore, the presence of erosive bases and the injectite and dewatering structures (Fig. 4) are suggestive of rapid and energetic deposition. There is a clear distinction between these transparent megabeds and the stratified acoustic response of the sediments in between. The latter sediments are typical of classical centimeter- to decimeter-scale deep-sea turbidites also present at Site 650 (Kastens et al., 1987) and common in deep-sea basins globally. We estimated the ages of the megabeds as 36 ± 6 ka (M4), 32 ± 7 ka (M3), 18 ± 3 ka (M2), and 8 ± 1 ka (M1) (Table S2).

A likely source region for the Marsili megabeds is the Campanian volcanic province to the north. In the past 130 k.y., volcanic sediments in the Marsili Basin have been predominantly sourced from the Campanian volcanic province (Calanchi et al., 1994) and are distinguishable from other volcanic provinces (Paterne et al., 1986). Alternative source regions include the active Stromboli Canyon to the east (Kidd et al., 1998), yet sediment flows entering from the east would have to course around the Marsili Seamount to be deposited in the western part of the basin. If derived from the Campanian margin, the Campi Flegrei would be a likely source area, given the known supervolcanic events that have occurred within the time period represented by the megabeds, most notably, the Campanian ignimbrite (CI) at 39.2 ka; (Silleni et al., 2020) and the Neapolitan Yellow Tuff (NYT) at 14.9 ka (Deino et al., 2004). The Marsili Seamount is a potential local source region of the megabeds. However, swath bathymetric data do not reveal major collapse structures on any of its flanks. The seamount itself is ~925 km³ in volume, and, therefore, given the volume of the megabeds (Table S1), individual events would be expected to have left evidence of significant scars if they originated from its flanks. Furthermore, the abundant shallow-water benthic microfossils recorded at Site 650 (Kastens et al., 1987) in the megabeds rule out the Marsili Seamount as the source location.

The 39.2 ka CI supereruption had pronounced effects on climate, human history, and ecosystems (Thunell et al., 1979; Milia et al., 2020; Silleni et al., 2020). The event involved a bulk volume >500 km³, including a pyroclastic flow on land that spread over 7000 km² (Silleni et al., 2020). Silleni et al. (2020) predicted that up to 75 km³ represent submarine deposits, yet their locations and actual volumes have been unconstrained. At Site 650, McCoy and Cornell (1990) correlated (based on glass chemistry) the volcaniclastic debris flow in bed 27, located between 60 and 82 m below seafloor, to the “Y-5” distal marine tephra of the CI (Keller et al., 1978). Interestingly, this depth interval encompasses both M3 and M4 in the Parasound data (Fig. 2). The age estimates and glass chemistry strongly support an interpretation of M4 as part of the CI marine deposits. We tentatively suggest that M3 was also related to the CI event because of the distinctive Y-5 glass chemistry, the age estimate of 32 ± 7 ka, which is within error of the CI event, and the close association between M3 and M4 (there is only a meter, or less, of layered material separating M3 and M4; Figs. 2 and 3). The depositional mechanism of M3 (debris flow) could indicate that deposition of the CI triggered associated major slope instability of the Campanian slope. Given that the CI event resulted from multiple collapses in a restricted time span, there could also have been a time lag in between megabeds. With M4 and tentatively M3 associated with the CI, the 14.9 ka NYT may correspond to M2 (18 ± 3 ka). After the NYT eruption, less-energetic events occurred at Campi Flegrei (Smith et al., 2011). It is possible that M1 (8 ± 1 ka) may be linked to one of these younger, less-energetic events.

Subaqueous mass-gravity flows originating from the Campanian margin could be routed to the deep basins of the Tyrrhenian Sea via the Magnaghi and Dohrm Canyons, which both originate in the CI source area within short proximity to the shoreline (Fig. 1). Sea level at the time of the CI would have been 70–80 m lower than today and thus even more favorable for sediment-gravity flows traversing south to have been intercepted by the canyons (Silleni et al., 2020). Milia et al. (2020) reported finding some CI-related material trapped within minor basins and/or left within the canyon as observed by sediment cores.

The Marsili megabeds may correlate with those in the neighboring Vavilov Basin to the west (Fig. 1; Gamberi and Marani, 2009). If true, the megabeds would be larger than areas represented in the maps (Fig. 4). Given that our volume estimates for M4 and M3 of 13.3 and 11.8 km³, respectively, would account for up to 40% of the estimate of submarine CI deposits (Silleni et al., 2020), the additional volume could be accounted for in the Vavilov Basin (Gamberi and Marani, 2009).

The apparent 10–15 k.y. recurrence interval of megabeds in the Marsili Basin (~17 k.y. if M3 and M4 are considered part of the same event) in the past 50 k.y. is not generally seen in such short depth and time intervals in other megabed localities. This may be related to the combination of highly active volcanic provinces around the Tyrrhenian Sea to trigger catastrophic flows and seafloor canyons that directly funnel flows from multiple entry points to the deep basins. Further, the Marsili Basin, because of its mostly enclosed configuration with high surrounding walls, may be an ideal basin to preserve the full mud components of giant turbidity flows. Previous megabed studies inferred eustatic lowstands (Rothwell et al., 1998) and earthquakes (Kioka et al., 2019) as triggering mechanisms. In the Marsili, sea-level fluctuations and earthquakes cannot be ruled out as playing a role in the origin of, at least, some of the megabeds. The preponderance of volcaniclastic material in the megabeds may simply indicate the source material and not necessarily be a direct indicator of volcanic processes. Further, tephra layers from volcanic activity may have served as a weak layer, favoring slope instability, and preconditioned the triggering of mass-gravity flows along the Marsili flanks and the surrounding slope sectors. Regardless of origin, the Marsili Basin megabeds are an important archive of catastrophic events with valuable information for understanding megabeds globally and geohazards in the circum–Tyrrhenian Sea regionally.

We thank the scientific party, officers, and crew of CHIANTI Cruise, Leg 4, on the R/V Sarmiento de Gamboa. This study was part of projects HADES and ICEFLAME (grants CTM2011-30400-C02-01-R and PID2020-114856RB-100) funded by the Ministerio de Ciencia e Innovation (MCIN)/Agencia Estatal de Investigación (AEI), https://doi.org/10.13039/501100011033. D.E. Sawyer was partially supported by National Science Foundation grant 1945543. We appreciated the constructive reviews from Antonio Cattaneo, Cecilia McHugh, and Christian Gorini, which greatly improved the manuscript.