Landslide Frequency and Failure Mechanisms at NE Gela Basin (Strait of Sicily)

Abstract

Despite intense research by both academia and industry, the parameters controlling slope stability at continental margins are often speculated upon. Lack of core recovery and age control on failed sediments prevent the assessment of failure timing/frequency and the role of prefailure architecture as shaped by paleoenvironmental changes. This study uses an integrated chronological framework from two boreholes and complementary ultrahigh-resolution acoustic profiling in order to assess (1) the frequency of submarine landsliding at the continental margin of NE Gela Basin and (2) the associated mechanisms of failure. Accurate age control was achieved through absolute radiocarbon dating and indirect dating relying on isotope stratigraphic and micropaleontological reconstructions. A total of nine major slope failure events have been recognized that occurred within the last 87 kyr (~10 kyr return frequency), though there is evidence for additional syndepositional, small-scaled transport processes of lower volume. Preferential failure involves translational movement of mudflows along subhorizontal surfaces that are induced by sedimentological changes relating to prefailure stratal architecture. Along with sequence-stratigraphic boundaries reflecting paleoenvironmental fluctuations, recovered core material suggests that intercalated volcaniclastic layers are key to the basal confinement and lateral movement of these events in the study area. Another major predisposing factor is given by rapid loading of fine-grained homogenous strata and successive generation of excess pore pressure, as expressed by several fluid escape structures. Recurrent failure, however, requires repeated generation of favorable conditions, and seismic activity, though low if compared to many other Mediterranean settings, is shown to represent a legitimate trigger mechanism.

Key Points

• Nine major failure events occurred at NE Gela Basin within the last 87 kyr, predominantly during conditions of sea level fall or lowstand
• Foraminiferal records suggest additional, syndepositional processes of lower volume, counteracting the emplacement of larger landslides
• Preferential failure along subhorizontal weak layers (stratigraphic boundaries and volcanoclastic layers) likely triggered by seismicity

1 Introduction

Submarine slope destabilization is a common phenomenon along active and passive continental margins and is the subject of intense research, mainly owing to the inherent hazardous consequences associated with landslide activities (e.g., Hampton et al., 1996; Masson et al., 2006; Twichell et al., 2009). Considerable effort has been undertaken by both academia and industry in order to elucidate the mechanisms and processes at the origin of such events (e.g., Vanneste et al., 2014, and references therein). As a consequence, our knowledge on predisposing factors and trigger mechanisms has vastly improved over the last decades. Still, key parameters controlling sediment instabilities are often speculated upon due to the lack of core recovery and are largely dependent on a multitude of environmental parameters, which have to be assessed individually for each case study. It is essential to study not only the spatial distribution of submarine landsliding but as well their temporal evolution and response to changes in environmental conditions (e.g., Brothers et al., 2013; Canals et al., 2004; Lee, 2009; McAdoo et al., 2000; Twichell et al., 2009; Urlaub et al., 2013). Evaluation of the recurrence of landslide processes is key not only to the assessment of potential hazards (e.g., Locat & Lee, 2002) but also to unravel the influence of paleoenvironmental changes on preconditioning factors and triggering mechanisms. This task requires great care in order to discern individual failure events and is facilitated through unprecedented quality of modern acoustic devices and the availability of long cores that provide age control on sediment successions.

In this study, we present a detailed investigation of the historical landslide record intercalated in the late Quaternary sedimentary sequences at the continental margin of NE Gela Basin (Strait of Sicily) and discuss potential failure mechanisms related to these events. We introduce a chronostratigraphic model of a 35.6 m long drilled core recovered from the failed material on the lower slope (site GeoB14401, ~614 m below sea level (bsl)) that integrates evidence from (1) foraminifera-based radiometric datings, (2) stable oxygen (δ¹⁸O) and carbon (δ¹³C) isotopic compositions from the benthic B. marginata and the planktonic G. ruber, and (3) biostratigraphic considerations related to changes in foraminiferal assemblages. Together with published records from a 54.6 m long core recovering undisturbed shelf-edge sediments (site GeoB14403, ~182 m bsl) and associated sequence-stratigraphic interpretations, this forms the chronostratigraphic framework for a successive analysis of landslide frequency in the study area. A dense network of ultrahigh-resolution parametric subbottom profiles was used to identify and map landslide deposits within the study area. Core-seismic correlation and tracing of dated marker horizons in the acoustic record provided full stratigraphic control on these deposits.

On the basis of this wealth of data, we present an estimate on the recurrence of failure at NE Gela Basin and evaluate the temporal distribution of landslide events. Detailed analysis of two recovered basal bedding planes allows us to draw further conclusions on the role of prefailure stratal architecture and associated emplacement of weak layers and may provide new evidence for the potentially destabilizing effect of intercalated volcaniclastic layers.

2 Regional and Local Setting

2.1 Geology and Seismicity

Filled with as much as 2,500 m of shallowing upward marine sediments, the Gela Basin represents the most recent (Pliocene-Quaternary) foredeep basin of the Maghrebian fold-and-thrust belt, which reflects the ongoing subduction of the African plate beneath the Eurasian plate (Figure 1a) (Argnani, 1990; Colantoni, 1975). In the north, it is overthrusted by the Gela Nappe—a southwest migrating contractional front representing the southernmost thrust wedge of the Maghrebian chain (Figure 1b) (Argnani, 1990; Butler et al., 1992). Plunging in the direction of the basin, the surface slope in the frontal area of Gela Nappe contributes to instability of overlaying sediments and manifests in frequent failure along the shelf edge in the study area (Figure 2) (Kuhlmann et al., 2014; Minisini et al., 2007; Minisini & Trincardi, 2009). Subsequent to the emplacement of the Gela Nappe, a phase of general uplift associated with widespread volcanic activity characterized the mid-Pliocene to late Quaternary, favoring the intercalation of eruptive material in the study area (Gardiner et al., 1993). To the east, the Gela Basin is confined by the Malta Plateau, the seaward counterpart of the Hyblean Plateau on mainland Sicily (Figure 1b). The southwestward dipping depositional ramp of the Hyblean-Malta Platform is overlain by a Pliocene-Quaternary succession of progradational sequences, resulting in a peculiar setting of SW dipping sediment bodies in the study area (Figure 3). The southern margin of the Gela Basin corresponds to the extensional area of the NW-SE trending Sicily Channel rift zone including the basins of Pantelleria, Malta, and Linosa. The corresponding rifting phase initiated in the late Miocene to early Pliocene and lasted through the Quaternary (Grasso, 1993).

Figure 1

(a) Simplified sketch of the central Mediterranean plate boundaries with arrows indicating direction of plate motion (after Dilek & Sandvol, 2009; Camerlenghi et al., 2010). (b) Bathymetric map of the study area in the Strait of Sicily illustrating main tectonic and geographic elements (after Jenny et al., 2006; Accaino et al., 2011; Civile et al., 2010; bathymetric data taken from GEBCO database). Superposed on this map is the U.S. Geological Survey (USGS) earthquake record since 1970, both size and color coded depending on the magnitude and depth of the event, respectively (see legend on the left-hand side). The Pantelleria graben (PG), the Malta graben (MG), and the Linosa graben (LG) are the principal tectonic depressions in the area. AB = Adventure Bank; MP = Malta Plateau; GB = Gela Basin.

Figure 2

Multibeam shaded relief of the NE Gela Basin illustrating the position of prominent features, drill sites, and acoustic profiles used in this study.

Figure 3

Overview of the chronologic framework for the undisturbed shelf-edge sediments at NE Gela Basin (modified after Kuhlmann et al., 2015). (a) Chronology at site GeoB14403 illustrating drill scheme, stable isotope records from benthic B. marginata (blue line) and planktonic G. ruber (red line), radiocarbon datings (black filled circles), and interpreted Marine Isotope Stages and Substages (MIS) and Dansgaard-Oeschger oscillations. See Figure 4a for a complete legend. (b) Along-shelf acoustic profile at the study site indicating the thick sedimentary successions deposited on the continental margin and the signature of the basin-wide Father Slide event. (c) Core-supported sequence-stratigraphic interpretation of the acoustic profile in Figure 3b revealing Milankovitch-type forcing of depositional architecture and major bounding surfaces. Inset: position of MIS as well as of internal sequence boundaries relative to the sea level curve of the last glacial-interglacial cycle (from Waelbroeck et al., 2002). Legend: HST: Highstand Systems Tract; TST: Transgressive Systems Tract; LST: Lowstand Systems Tract; FSST: Falling Stage Systems Tract.

As indicated by the instrumental and historical series of the USGS earthquake catalog (U.S. Geological Survey, https://earthquake.usgs.gov/earthquakes/search/, accessed on 25 May 2017), seismicity at Gela Basin is relatively low if compared to other Mediterranean areas (Figure 1b). Within nearly half a century (since 1970), seven epicenters were registered in the immediate proximity of NE Gela Basin that showed low to moderate magnitudes ranging from 2.6 to 4.2 M.

2.2 Oceanographic Setting

Oceanic currents in the Strait of Sicily are driven by the antiestuarine thermohaline circulation system of the Mediterranean Sea and involve an eastward directed fresher surface layer of Modified Atlantic Water (MAW; first 200 m of water column) and a concurrent westward flow of relatively more saline Mediterranean subsurface waters (mostly Levantine Intermediate Waters, LIW) (Robinson et al., 1999; Sammari et al., 1999). Along the continental slope of the study site, a regional branch of LIW reaches velocities of 13 cm/s (Lermusiaux & Robinson, 2001). This branch has been linked to the emplacement of muddy shelf-edge contourites suggested to be capable of generating excess pore pressures and, hence, reducing slope stability and (indirectly) influencing the position of postdrift slide events (Kuhlmann et al., 2014; Verdicchio & Trincardi, 2008).

2.3 Local Morphology

As previously reported by Trincardi and Argnani (1990), Trincardi et al. (2003), Minisini et al. (2007), and Kuhlmann et al. (2014), the continental slope in the study area documents recent seafloor-exposed landslides as well as remnants of old and partially buried slide scars (Figure 2). The most prominent feature is the Twin Slide complex, two recent coeval landslides on the upper slope that display subrounded scarps and feature reliefs exceeding 100 m. Head scarps incline by a maximum of 32° (Kuhlmann et al., 2014), and the complex is characterized by runout lengths of 10–12 km, involving material with an estimated volume of <1 km³ (Minisini et al., 2007). Both landslides have been shown to relate to a multistage failure involving a debris avalanche and a successive retrogressive slide/slump (Kuhlmann et al., 2014; Minisini & Trincardi, 2009) that partly covers the debris avalanche deposits. The shelf edge/upper slope area in between these scarps accommodates a depression that dips steeper (~8°) than the surrounding shelf deposits, interpreted as the buried head scarp of an older landslide event termed Father Slide. Remnant bathymetric evidence for the emplacement of additional mass transport deposits (MTDs) within the timespan between the occurrence of the Father Slide and the Twin Slides is discernible in the form of paleoscarps of up to 10 m in relief (Figure 2).

2.4 Local Stratigraphy

Within the study area, high sediment accumulation rates (up to 2 m/kyr (Kuhlmann et al., 2015)) and strong bottom currents favor the formation of thick progradational shelf clinoforms on the continental shelf. Kuhlmann et al. (2015) related these depositional sequences to cyclic fluctuations in sea level, which in turn are paced at Milankovitch-type (100 kyr and 23 kyr) and subordinate (Dansgaard-Oeschger events, D/O) frequencies. Upslope of the failure zone, age control is provided by sequence-stratigraphic interpretations and a chronostratigraphic model of a quasi-continuous sediment record recovered from the undisturbed shelf-edge at site GeoB14403 (54.6 m below seafloor (bsf)). The introduced age model integrates radiocarbon datings with stable isotope stratigraphic and biostratigraphic considerations (Figure 3a, see Kuhlmann et al., 2015, for details).

Figure 3b illustrates the late Quaternary shelf-depositional sequence of the last glacial-interglacial cycle at NE Gela Basin comprising, from bottom to top, (1) alternating highstand (HST: Highstand Systems Tract) and forced regression units (FSST: Falling Stage Systems Tract) correlating to the Marine Isotope Stages and Substages (MIS) of MIS 5; (2) a lowstand unit (LST: Lowstand Systems Tract) that is confined at the base by a subordinate sequence boundary, a downward shift of the shoreline imparted by a sea level fall related to a higher-frequency sea level fluctuation, and correlates to MIS 4; (3) a shelf margin wedge comprising alternating HST/FSST units that formed during the falling sea level limb associated with MIS 3-2 until glacial lowstand; (4) a faintly stratified transgressive drape (TST: Transgressive Systems Tract) that accumulated during the phases of drowning of the continental shelf driven by the postglacial sea level rise; and (5) a progradational shelf wedge forming since the onset of the modern sea level at about 5.5 kyr B.P. (HST).

3 Materials and Methods

This study is based on a dense grid of acoustic subbottom profiles, high-resolution bathymetric data, and several complementing sediment cores acquired aboard R/V Odin Finder (2000), R/V Urania (2003, 2004, and 2005), and R/V Maria S. Merian (2010). The methods of acquisition as well as laboratory analyses performed on the recovered sediment archives are introduced below. Figure 2 provides an overview of core locations as well as acoustic datasets considered in this study and highlights representative profiles presented herein.

3.1 Hydroacoustics

High-resolution bathymetric data derive from a RESON 8160 multibeam operating at 50 kHz and the multibeam echosounder systems Kongsberg Simrad EM-120 (12 kHz), EM-300 (50 kHz), and EM-1002 (95 kHz). Subsurface imagery was conducted with an Atlas Parasound system operating at 4 kHz and a Benthos (Datasonics) CAP-6600 Chirp-II sonar system using a 2–7 kHz sweep-modulated bandwidth. Vertical resolution for both systems is around 0.2 m. Core-seismic correlation, tracing of dated marker horizons and mapping of mass transport deposits (MTDs), was conducted within the Kingdom software (ver. 8.6).

3.2 Coring

This study introduces the previously unpublished drill site GeoB14401 (35.6 m bsf) that samples stacked MTDs from the accumulation zone on the lower slope (Figure 2). It comprises a single gravity core (GC), which is accompanied by two drilled sequences recovered with the Bremen seafloor drill rig MeBo (Freudenthal & Wefer, 2007, 2013). An overview of main core metadata is provided in Table 1.

Table 1. Overview of Core Metadata at Site GeoB14401

		Water depth	Position		Top depth	Bottom depth	Length
Name	Type	(m bsl)	Latitude	Longitude	(m bsf)	(m bsf)	(m)	Recovery rate
GeoB14401-2	Gravity core (GC)	634.3	36°47.17′N	14°11.91′E	0.00	4.18	4.18	100%
GeoB14401-3	MeBo	613.8	36°47.20′N	14°11.90′E	0.00	20.49	20.49	26%
GeoB14401-5	MeBo	601.3	36°47.19′N	14°11.89′E	14.45	35.6	21.15	84.8%

3.3 Core Logging

A GEOTEK Ltd. multisensor core logger (MSCL) housed at MARUM (University of Bremen, Germany) was used for nondestructive measurements of core-physical properties including gamma-ray-attenuated density, compressional wave velocity, and magnetic susceptibility. Measured density values were corrected using moisture and density parameters (MAD) determined according to the International Ocean Drilling Program (IODP) shipboard practices (Blum, 1997).

Geochemical logging for light elements (Al to Fe) was performed on an Avaatech II core scanner housed at MARUM using a generator setting of 10 kV, 0.2 mA, and a sampling time of 20 s. Elemental intensities were measured in a step interval of 2 cm with a slit size of 1 cm downcore and 1.2 cm crosscore. Following suggestions by Richter et al. (2006) and Weltje and Tjallingii (2008), results are presented in logarithmic elemental ratios (i.e., log[A/B]) due to the inherent measurement principles.

3.4 Micropaleontology

For foraminiferal analysis, subsamples of 1 cm (GC) and 2 cm (MeBo) thickness were obtained with a vertical resolution of at least 20 cm throughout the cores (a total of 93 at site GeoB14401). The subsamples were freeze-dried, soaked in distilled water, and washed over a 63 μm mesh sieve (Schröder et al., 1987). The foraminiferal content was examined with an optical microscope and a semiquantitative analysis was performed on the fraction >63 μm mesh sieve as (1) part of the development of a robust age model and (2) evidence for transport processes. Since benthic foraminifera species are specialized bottom dwellers that are characteristically linked to certain environments (e.g., outer shelf and inner shelf facies), they are excellent indicators of paleoenvironmental conditions and, indirectly, sea level (e.g., Jorissen, 1987). The occurrence of individual specimens in an untypical environment can hence be linked to transport processes and may be used to infer landsliding or erosional reworking by bottom hugging currents.

3.5 Stable Isotope Analysis

Subsamples obtained for micropaleontology were equally used to determine oxygen and carbon isotopic compositions of the benthic Bulimina marginata and the planktonic Globigerinoides ex gr. ruber. Hand-picked specimens showing no signs of diagenetic alteration were ultrasonically cleaned to remove contaminants such as overgrowths, coccoliths, and detrital infilling. Analyses were performed at MARUM using a Finnigan MAT 252 mass spectrometer coupled with a carbonate preparation device type “Bremen.” Isotopic composition is expressed as per mil (‰) deviation with respect to the Pee Dee Belemnite (PDB) standard, with a long-term laboratory analytical standard deviation of <0.07‰ for the stable oxygen and <0.05‰ for the stable carbon isotope data. Note that none of the isotopic data presented here have been corrected for the ice volume effect.

3.6 Radiocarbon Dating

Sediment samples were taken from selected core horizons in order to (1) corroborate stratigraphic interpretations inferred from micropaleontological and stable isotope data and (2) provide age control on submarine landslide deposits. Radiocarbon analyses were performed by the National Ocean Sciences Accelerator Mass Spectrometry Facility (NOSAMS, Woods Hole Oceanographic Institution, USA) on monospecific or mixed planktonic foraminifera in the size fraction >63 μm (Table 2). Obtained AMS ¹⁴C ages were converted into 2σ calibrated ages using Calib 7.10 Radiocarbon Calibration Program (Stuiver et al., 2017) and the Marine13 calibration data set (Reimer et al., 2013). The reservoir correction was selected from the Calib database for the Sicily Channel, with a calculated weighted mean ΔR value of 71 years and a standard derivation of 50 years (Siani et al., 2000). All dates reported are given in calibrated thousands of years before present (cal kyr B.P.).

Table 2. AMS ¹⁴C and Calibrated Ages Performed on Selected Foraminifera

Lab #	Core	Section #/depth (cm)	Material	^14C age (yr B.P.)	2σ cal. intercept (cal kyr B.P.)	Reference
OS-106893	GeoB14401-2 (GC)	IV 80-84	N. pachyderma, G. bulloides	16400 ± 95	19.22 ± 0.29	this study
OS-106666	GeoB14401-5 (MeBo)	1P-1 40-46	N. pachyderma, G. bulloides, N. dutertrei	15800 ± 55	18.59 ± 0.18	this study
OS-106889	GeoB14401-5 (MeBo)	7P-2 0-5	N. pachyderma	25300 ± 290	28.88 ± 0.66	this study
OS-106890	GeoB14401-5 (MeBo)	9P-2 62-67	N. pachyderma	26700 ± 350	30.31 ± 0.71	this study

• Note. Calibration is based on Calib 7.10 Radiocarbon Calibration Program (Stuiver et al., 2017). Calibration data set: Marine13 according to Reimer et al. (2013). Reservoir correction: the calculated weighted mean ΔR value is 71 with a standard derivation of 50 (Siani et al., 2000).

3.7 Grain Size and Microanalyses

Grain sizes were measured from 1 ml syringe samples with the Laser Diffraction Particle Analyzer (Beckman Coulter) LS 13320 at MARUM grain size lab and are reported in volume percentages of 92 size classes from 0.4 to 2000 μm. Thin sections of MTD bedding planes were prepared from selected horizons and were investigated by optical microscopy and by EDX on a Cameca SX100 electron microprobe at the Department of Geosciences, University of Bremen.

4 Chronostratigraphic Framework

In order to correlate upslope and downslope sedimentary units and to provide an accurate temporal control on the emplacement of MTDs in the area, this study relies on a multiproxy chronological framework integrating evidence from (1) stable isotope data of the benthic foraminifer B. marginata and the planktonic G. ruber, (2) microfaunal bioevents as reflected in the foraminiferal assemblage, and (3) foraminifera-based radiometric datings at site GeoB14401 (35.6 m bsf). This framework forms the basis of subsequent core-seismic correlations and is summarized schematically in Figure 4a (see Figure 3a and Kuhlmann et al., 2015, for details on the chronology of site GeoB14403 on the shelf-edge).

Figure 4

(a) Overview of the chronostratigraphic framework at borehole GeoB14401 illustrating drill scheme, δ13C and δ18O records from benthic B. marginata and planktonic G. ruber, position and calibrated ages from radiocarbon datings, and interpreted MIS. The microscope image shows shelf-benthic foraminifera specimens in the deepwater core, suggesting reworking. The legend refers to the symbology used in Figure 3a as well. (b) Line scans of recovered sediment sections showing evidence of reworking (1 and 2) as well as sedimentological anomalies at the base of identified mass transport deposits (3, 4, and 5).

The sequence recovered from the landslide accumulation zone at site GeoB14401 in ~614 m water depth provides a stratigraphic record of sediment successions to a depth of 35.6 m bsf, with poor core recovery in the depth window of 4.5–14.5 m bsf that corresponds to a thick acoustically transparent unit (Figure 4a, GB I). Except from an abrupt shift of isotopic composition (> 2‰ for δ¹⁸O values) in the upper section recovered at site GeoB14401-2, the isotopic curves lack significant variation that could be fine-tuned with well-known events in the NGRIP record (North Greenland Ice Core Project Members, 2004)—especially in the well stratified section between 14.5 m bsf and 27 m bsf, where G. ruber shows a quite heavy δ¹⁸O composition ranging between 2.5 and 3.0‰, comparable with the values recorded during MIS 2 by Capotondi et al. (1999) and Sbaffi et al. (2004) in the Tyrrhenian Sea and by Rossignol-Strick and Paterne (1999), Lourens (2004), and Ducassou et al. (2007) in the eastern Mediterranean. This suggests an extended interval of steady environmental conditions, which is supported by radiocarbon datings inferring a sedimentation rate of 2 m/kyr for this interval. Similar evidence comes from micropaleontological investigations, which reveal an abundance of oxidized clasts and fine terrigenous material (suggesting proximity to the coastline) as well as cold and productive waters (planktic assemblage dominated by Globigerina quinqueloba common, Neogloboquadrina pachyderma r.c., Neogloboquadrina dutertrei, Globigerina bulloides, Globigerina glutinata, Globigerina scitula rare, and G. ruber very rare (Hemleben et al., 1989; Pujol & Vergnaud Grazzini, 1995; Thunell, 1978)). Based on all these observations, the interval from about 15–27 m can confidently be ascribed to the Last Glacial Maximum chronozone (LGM), which is defined between 19 and 23 kyr B.P. (Mix et al., 2001) and is coeval with the lowest stand of sea level (e.g., Yokoyama et al., 2000). On the upper shelf borehole GeoB14403, the equivalent to this interval is condensed in the prominent erosive unconformity ES1 (see Figure 3 as well).

For the sedimentary succession of the upper 2.6 m of borehole GeoB14401, foraminifera assemblages indicate a Holocene age (Figure 4a). The lowest part of this section comprises sediment coeval with the deposition of Sapropel 1 in the eastern Mediterranean (S1 equivalent), recognized by the absence of Globorotalia truncatulinoides (reduction of the winter mixing of the water columns) and by a benthic assemblage suggesting a higher accumulation of organic matter (Uvigerina peregrina abundant) along with some degree of scarcity of bottom oxygenation within the sediment (presence, although not abundant, of deep infaunal benthic species Globobulimina spp. and Chilostomella spp. (De Rijk et al., 1999; De Stigter et al., 1998; Jorissen, 1987, 1999; Schmiedl et al., 2000)). The benthic assemblage of the overlying highstand (HST) unit (Uvigerina mediterranea, U. peregrina, Cibicidoides pachyderma, Cassidulina laevigata carinata common, Hyalinea balthica, Melonis barleeanum, and Gyroidinoides altiformis, Articulina tubulosa) indicates a relatively less stressed environment. The base of this unit can be positioned at approximately 1.5 m (above which Globigerinoides sacculifer shows common presence and reoccurrence of G. truncatulinoides r.c. is reported (Incarbona et al., 2010; Minisini et al., 2007; Rouis-Zargouni et al., 2010; Sprovieri et al., 2003)).

Moreover, lacking of the older phase of S1 equivalent is quite probable, as a reference core collected nearby (core P09 in Minisini et al., 2007; see Figure 2) reports a frequency peak of deep infaunal benthic species, while at borehole GeoB14401 these species are present but not common as expected. Further evidence supporting this hypothesis is seen in the G. ruber δ¹⁸O composition of this interval, which records values slightly heavier than expected in comparison with nearby records (GeoB14403 and GeoB14414; Kuhlmann et al., 2015). Also, the older pre-Boreal interval, expected below the S1 equivalent record, is not recorded because G. truncatulinoides, along with the warm planktic assemblage in which it peaks (Minisini et al., 2007; Sprovieri et al., 2003), is absent.

In the depth interval related to a second acoustically transparent section (i.e., 27–35 m bsf, GB II), the stable carbon isotopic record reveals very similar values for both G. ruber and B. marginata. The planktonic assemblage in this acoustically transparent body, as well as downcore, lacks G. inflata, implying an age younger than 38 kyr B.P. (see Kuhlmann et al., 2015, for further details regarding the temporary disappearance of this biomarker within MIS 3). In addition, the benthic microfauna is well preserved and displays a composition quite similar to the one present in the overlying stratified section and compatible with a mesotrophic slope environment (C. laevigata carinata + B. dilatata + B. aenariensis abundant, accompanied by M. barleeanum, H. balthica, B. marginata, G. umbonatus, G. neosoldanii, C. pachyderma, A. tubulosa, and U. peregrina).

The lowermost 40 cm of the core is characterized by very scarce terrigenous material, differently from the overlying sediments, and by a very abundant cold water planktonic foraminifera assemblage (G. bulloides, N. pachyderma r.c., N. dutertrei, G. glutinata, G. scitula, G. quinqueloba, and rare G. ruber). The benthic assemblage is quite similar to the one described in the upper sediments, although shallow water specimens are near absent. Based on these observations, this part of the core appears to indicate sediment not reworked, in situ. Radiometric dating of this section suggests an age of about 30 kyr B.P.

5 Results

5.1 Identification of MTDs

The assessment of failure frequency is closely linked to the identification of deposits generated by mass wasting events. Evidence for such deposits in the study area is provided by drilled sediment sequences recovered at site GeoB14401 (Figure 4b) and related micropaleontological analyses as well as through interpretations of acoustic subbottom profiles.

GeoB14401 sediments reveal a dominant lithology of silty nannofossil clay with quartzose silt, which is accompanied by minor amounts of foraminifera and authigenic pyrite as well as volcaniclastic material. Drilled sections that penetrate acoustically transparent facies GB II (e.g., 27–35 m bsf, Figure 4a) show clear signs of reworking indicating the presence of a MTD (Figure 4b: 1 and 2). This is supported by the foraminiferal assemblages, which reveal a scarcity of shelf species within this unit (such as Elphidium spp., Asterigerinata spp., and Neoconorbina terquemi (Jorissen, 1987; Langer, 1993)) along with the presence of unbroken specimens of quite fragile species, such as A. tubulosa, suggesting a short-distance transport of sediment from shallower areas of the slope. The base of this reworked unit GB II corresponds to a prominent acoustic reflector and a marked sediment anomaly (Figure 4b: 3), below which the foraminiferal assemblage indicates sediment not reworked. Similar observations can be made at the base of the drilled section correlating with the acoustically transparent unit GB I at 4.5–14.5 m bsf (Figures 4a and 4b: 4 and 5). Within the study area, multiple acoustically transparent units with a similar appearance to GB I and GB II can be observed, all of which share the following characteristics and were interpreted as MTDs (see section 5.2 for details):

1. They are usually confined upslope by sharply truncated stratified sections and induce a seaward thinning of sediment packages due to the downslope transport of material at this lateral interface.
2. They show lateral variations in thickness and preferentially fill out underlying irregularities.
3. There is an absence of thick acoustically transparent units that align with the regular stratification patterns over the entire length of the slope, which could hint towards fine stratified clinoforms not resolvable in the acoustic record (rather than MTDs).

Apart from the evidence of MTDs that relates to the presence of acoustically transparent units, further evidence of landsliding may derive from the micropaleontological record at site GeoB14401, where reworked shelf-benthic specimens (mainly Elphidium) can be observed. Interestingly, these appear predominantly in the zone associated to the LGM chronozone (i.e., 15–27 m bsf, Figure 4a) and show no signs of damage (microscope image in Figure 4a). Finally, no sorting events seem to have clearly affected the foraminifera assemblages in this interval, which comprise specimens with different sizes (e.g., both adult and young specimens of the same species). This observation suggests a syndepositional transport process of lower volume, that is, a short-distance transport of material by a slide movement.

5.2 Mapping of MTDs

The dense network of high-resolution acoustic profiles available in the study area (see Figure 2) allowed for a comprehensive mapping of MTDs. If not compromised by attenuating signals, acoustic reflectors bounding these MTDs could be traced upslope and into the drill sites, thus offering chronostratigraphic control on the individual events. In the following, identified MTDs are introduced using type sections along and across the slope (Figures 5 and 6).

Figure 5

(a) Along-slope acoustic profile at the Northern Twin Slide correlating main marker surfaces from the shelf area (see legend) with the accumulation zone on the lower slope, thus providing age control over the emplacement of deposits and/or glide planes. Between the buried Father Slide and the recent Northern Twin Slide, six major events can be distinguished (termed GB I–GB VI). (b) Magnification of the failure zone of the Father Slide showing two additional events that are stratigraphically older (GB VII and GB VIII). (c) Evidence of midslope failure during relative lowstand of sea level in MIS 5.2. (d) Example of clearly visible dewatering structures and evidence for high-frequency, but very low-magnitude landslides processes.

Figure 6

Interpreted acoustic profiles (a) across the lower slope and (b) along the slope at the Southern Twin Slide revealing stacked pattern of small-scaled landslide deposits as recognized in the profile presented in Figure 5.

Exhibiting thick acoustically transparent facies (locally up to 50 m), the deposits of the Father Slide event (FS) are ubiquitous in deep sedimentary units and form the stratigraphically oldest MTDs that are continuously traceable throughout the area of investigation (Figures 5 and 6). The deposits are bounded basally by subhorizontal bedding planes, along which movement occurred, and appear to cut across reflectors depending on the location (Figure 6, profile 2). Despite their considerable thickness, they appear acoustically transparent and lack any evidence of internal structuring, while their upper surface is rather irregular and shows local bulges indicating the presence of compression ridges (Figure 5a). Farther upslope, younger MTDs are restricted within a pronounced head scarp, which sharply crosscuts the surrounding slope well into Eemian deposits (MIS 5.5; see Figure 3).

The deposits associated with the Father Slide event are covered by a stack of multiple thinner MTDs showing laterally varying thicknesses that appear to preferentially infill local depressions and smooth out seafloor irregularities (thus regulating the disturbances relating to the Father Slide event, Figures 5 and 6). A total of six major deposits of this kind can be traced in the Gela Basin (GB I–GB VI), all being bounded basally by subhorizontal bedding planes but with varying lateral extents. Typically, they lack signs of scarping and induce subvertical acoustic wipeouts in overlying stratified layers that represent variable scattering and absorption of the acoustic signal (Figures 5d and 6). Analogous to the FS deposits, there is sporadic evidence for truncating reflectors (e.g., GB I in Figure 6, profile 2). Two additional events associated with this group could be traced in the stratigraphic section reworked by Father Slide (GB VII and GB VIII; Figure 5b). These are, however, excluded from the estimation of landslide frequency due to difficulties of their tracing in deeper stratigraphic units and resulting lack of age control.

Some of the stacked MTDs were reworked by the recent failures of the Northern Twin Slide (NTS) and the Southern Twin Slide (STS). While the NTS appears to be bounded basally by subhorizontal bedding planes (Figure 5), the STS reveals clear signs of scarping and cannibalization of underlying strata (Figure 6).

5.3 Bedding Planes of MTDs

Recovery of the basal parts of GB I and GB II offers full analytical control on the transition from undisturbed background sedimentation to the overlying MTDs. In order to understand the mechanisms involved in their emplacement, material properties and microfabrics of the respective core sections and bedding planes were investigated by means of geophysical (MSCL) and geochemical (XRF) logs as well as microscopic and energy dispersive X-ray (EDX) analysis of thin sections (Figures 7 and 8).

Figure 7

(a) Selected material properties around the potential bedding plane (light yellow) at the base of event GB I in borehole GeoB14401 testifying a higher terrigenous fraction than in the surrounding material (elevated magnetic susceptibility; low Ca and high Fe concentrations with regard to Ti). (b) Thin section analysis of GB I basal bedding plane showing elongated mud clasts in plane-polarized light (1) interpreted as preferential direction of shear (2). The organic-rich brown glide plane reveals evidence of significantly higher gain size (3) and basaltic volcaniclastic origin: plagioclase (Pl) under cross-polarized light (4); plagioclase (Pl), basaltic glass (sideromelane, Bg) and palagonite (Pbg) under crossed polarizers (5); detail of irregular shaped fragments of (palagonitic) glassy basalt displaying small laths of plagioclase (with swallowtails) and transparent tiny microliths, in plane-polarized light (6), and under crossed polarizers (7).

Figure 8

(a) Analogue to GB I, the potential bedding plane (light yellow) of event GB II in borehole GeoB14403 is characterized by a prominent terrigenous fraction (elevated magnetic susceptibility; low Ca and high Fe concentrations with regard to Ti). Note that no p wave velocity data are available for the displayed core section. (b) Thin section overview of GB II basal bedding plane in plane-polarized light (1 and 2): evidence of sorting processes and presence of a thin band with preferential alignment of minerals, possibly indicating a shear zone: under plane-polarized light (3) and crossed polarizers (4 and 5); plagioclase (Pl) under cross-polarized light (6); electron backscatter image showing fragments of orthopyroxene (Opx), clinopyroxene (Cpx), plagioclase (Pl), and quartz (Qtz); mineral aggregate with clinopyroxene (Cpx), plagioclase (Pl), K-feldspar (Kfs), titanomagnetite (Mt), and associated biotite (Bt): under plane-polarized light (8), crossed polarizers (9), and as electron backscatter image (10).

Core-to-seismic correlation and micropaleontological observations reveal that the drilled bedding plane of GB I appears to correlate well with a section dominated by a thin, irregular band of darker sediment that features reduced concentrations of Ca and K with respect to Ti. At the same time, Fe concentrations and magnetic susceptibility are elevated with respect to the surrounding sediment (Figure 7a). The interval below displays a homogeneous unit exhibiting numerous oxidized clasts and relatively constant physical and chemical composition related to the LGM chronozone. The section just above the band of darker sediment supposedly correlating with the base of GB I is characterized by increased mean grain sizes and reduced densities and reveals enhanced Ca concentrations with respect to Ti, which suggests an increase with foraminiferal content.

Thin sections of the band of darker sediment reveal a clear distinction from the homogenous background material, both in terms of color as well as mineral content and particle size (Figure 7b). Small mud clasts show preferential elongation along the position of counterdirected deformation bands (Figure 7b: 1 and 2), while subrounded mineral fragments with significantly increased grain sizes are frequently observed (Figure 7b: 3). Unrepresentative sampling of background sediment in between the deformation bands masks this in the mean grain size data, while the grain size spectrum reveals a bimodal distribution. The mineralogical composition of this sedimentary anomaly shows widespread occurrence of plagioclase (Pl) fragments (Figure 7b: 4 and 5). Equally, the presence of irregularly shaped fragments of brown basaltic glass (sideromelane, Bg) can be observed that display small laths of plagioclase (some showing swallowtails) and transparent tiny microliths (Figure 7b: 5, 6, and 7). Some of these glasses are locally palagonitized due to seawater alteration (Figure 7b: 6 and 7).

The proposed bedding plane of GB II appears to exhibit a similar sediment character (Figure 8) but is thicker and shows greater lithological deviation from the background sediment. Mean grain sizes significantly increase with regard to the underlying deposit of MIS 3, along with slightly enhanced bulk densities and a strong positive magnetic anomaly (Figure 8a). While Ca concentrations are reduced with respect to Ti, concentrations of Fe and K are clearly elevated compared to the surrounding sediment.

Associated thin sections support these findings and reveal a high content of subrounded to angular mineral fragments in a dark brown matrix (Figure 8b: 1 and 2). The lower boundary of the bedding plane displays small bands with highly variable content of such fragments and shows some evidence of sorting (Figure 8b: 3 and 4). In one of these bands, minerals appear to be preferentially aligned subhorizontally (Figure 8b: 5). Analyses of individual mineral fragments (Figure 8b: 7) and within a mineral conglomerate (Figure 8b: 8–10) reveal the presence of orthopyroxene (Opx), clinopyroxene (Cpx), plagioclase (Pl), biotite (Bt), K-feldspar (Kfs), and titanomagnetite (Mt) within this sedimentological feature.

6 Discussion

6.1 Timing of Landslide Deposits

In order to estimate the timing of landslide events we used a mixed strategy that depends on the availability of direct and indirect dating possibilities (see Urlaub et al. (2013)). Where possible, we directly dated hemipelagic background sediment deposited after the event (for a minimum age) as well as prior to the event (for a maximum age) by means of ¹⁴C AMS dating to provide a time bracket. Alternatively, indirect dating through stable isotope stratigraphies and/or biostratigraphic considerations was applied. In case of lacking core control on observed landslide deposits, we relied on indirect dating through the tracing of respective marker horizons into absolute or indirectly dated surfaces. This, however, introduces larger uncertainties associated with the tracing of acoustic horizons itself as well as the lack of control on dated reference horizons.

A total of nine stacked MTDs with the Father Slide event at their base could be recognized and mapped at NE Gela Basin. It is noteworthy that we referred to more proximal acoustic profiles for their recognition in order to include landslides featuring small runout lengths while avoiding overlaps with landslide deposits from other areas of the basin that extend into the distal parts of the study area. Below we summarize the age constraints of the individual landslide events as introduced above. A graphic representation of their timing and distribution outlining the evolution at the continental slope at NE Gela Basin is depicted in Figure 9:

1. NTS/STS. These units represent the youngest landslide events of the investigated succession. There is no age control on the exact timing available. However, lack of smoothing of bathymetric features and little-to-no drape on top of the MTDs attest that the complex is of very recent age. Although both events involve multiple failure stages, they are counted as one single event, given the temporal proximity of individual stages and the lack of control on other potential multi-stage failures (e.g., Father Slide).
2. GB I. This deposit was investigated in a previous study and was interpreted as an 8.5 kyr B.P. mudflow overlain by thin turbidites (Minisini et al., 2007). This timing is consistent with the foraminifera assemblage at borehole GeoB14401, which indicates the absence of the lower Sapropel S1 equivalent, and hence is adopted herein. The basal bedding plane as recovered in borehole GeoB14401 is only marginally younger than the 18.59 kyr radiometric dating from a sample just below. However, in contrast to most other observed landslides in this group, GB I appears to truncate reflectors (profile 2 in Figure 6) and is locally confined.
3. GB II. This deposit with strong local variations in thickness and continuity provides age control by two radiocarbon datings just below its base and above its top. The latter suggests a deposition at ~24 kyr B.P., while the former indicates subhorizontal movement along a basal failure plane with an age of about 30 kyr B.P.
4. LGM. This group encompasses syndepositional emplacement of very small-scaled landslide events that are not resolvable in the acoustic record and can only be fully recognized by the micropaleontological analysis. Their deposition is linked to the occurrence of the LGM chronozone (i.e., 23–19 kyr B.P.).
5. GB III. A basin-wide failure event that has reworked sedimentary units of MIS 3. This deposit is confined up-section by a 30 kyr B.P. radiocarbon dating. Basal movement occurred along a surface deposited at around 45 kyr (according to upslope tracing into borehole GeoB14401).
6. GB IV. Deposits associated with this event occur preferentially in the area of STS and extend, if present at all, only with a thin deposit in the area allocated to NTS. Since the basal surface cannot be clearly traced into a dated borehole sequence, we estimate its age to 50–52 kyr B.P. based on its position between the two surrounding marker surfaces (a radiocarbon dating of 44.56 kyr and the isotopic MIS 4/3 boundary at 59.44 kyr) and presuming a constant sedimentation rate within this interval.
7. GB V. This deposit involves the locally confined failure of the lowstand wedge associated with MIS 4 and may be dated at 59 kyr B.P. based on isotopic correlations. Movement takes place along the subordinate sequence boundary related to the MIS 5/4 transition at ~70 kyr B.P.
8. GB VI. This unit involves several smaller and localized landslides that followed the emplacement of FS and subsequent rapid infilling of the associated scars. These deposits formed within the time window comprised between FS and GB V (i.e., 87–70 kyr B.P.).
9. FS. The thick deposits of Father Slide are basally bounded by surfaces that cannot be traced upslope due to attenuation of the acoustic signal. However, the associated head scarp on the upper slope suggests a penetration at least into Eemian sediments (MIS 5.5, Figure 3b). The timing of failure is more intriguing, since associated shear planes hamper the tracing of acoustic reflectors. From the acoustic evidence presented in Figure 5, we assign the time of failure to within the stadial-interstadial boundary (MIS 5.2/5.1, ~87 kyr B.P.). Multistage failure cannot be ruled out, as attenuated acoustic signals may mask recognition of internal structuring.

Figure 9

Schematic overview of major slide events since the emplacement of Father Slide deposits at about 87 kyr B.P. Blue and red outlines correspond to lateral extents of MTDs and their head scarps; arrows indicate the direction of motion as well as local confinements of deposits.

From the above introduced deposits, age estimates for GB I, GB II, and LGM can be considered the most reliable given the direct dating on hemipelagic sediments deposited shortly after MTD emplacement. The same consideration is true for GB III, though the reported minimum age might be underestimated due to uncertainties regarding the lack of core control. Even the proposed age for GB IV should be relatively accurate given the close proximity of dated marker horizons. Higher uncertainties can be expected for GB V, GB VI, and FS, since these lack marker horizons that could confidently be dated.

6.2 Frequency Estimations

In order to assess the frequency of landsliding, the time span of observation (i.e., the emplacement of FS until modern time) is simply divided by the amount of recognized events. Disregarding any paleoenvironmental constraints on the timing of failure (e.g., with regard to the position of sea level) and acknowledging the varying scale and style of described events, these nine groups of major slide events provide an estimated recurrence of about one event every 10 kyr at the continental slope of NE Gela Basin. This is about 3 times lower than originally estimated by Minisini and Trincardi (2009), who lacked full stratigraphic control on the emplaced MTDs and proposed a recurrence of one failure every 3–4 kyr. Still, this frequency is too high to imply a correlation to cyclic sea level variations, as proposed for a variety of cases (e.g., Brothers et al., 2013; Lebreiro et al., 1997; Lee, 2009; Piper et al., 2003). When observing the dated failure times with regard to the respective relative sea level (Figure 10), the data do not show any significant patterns in landslide abundance, supporting the findings of Urlaub et al. (2013): disregarding the three oldest deposits due to expected age uncertainties (i.e., GB V, GB VI, and FS), a total of three events may be related to the falling limb of the sea level curve (GB II, GB III, and GB IV), while two events appear to correlate with rapid sea level rise (GB I) or highstand (NTS/STS). While the 8.5 kyr mudflow depositing GB I might temporally coincide with statistical landslide agglomerations presented by Maslin et al. (2004) or Lee (2009) and with proposed enhanced landslide occurrences during rapid sea level rise as suggested by Brothers et al. (2013) or Smith et al. (2013), there is no obvious cluster observable for the introduced landslide record. Given the strong regional focus of this database, slope stability in the study area may thus relate to factors other than sea level change.

Figure 10

Occurrence of dated MTDs with regard to the relative sea level curve of the last glacial-interglacial cycle (Waelbroeck et al., 2002).

One recognized event that may be of interest in this context is given by the small-scale landslides occurring during the time of the LGM chronozone (19–23 kyr (Mix et al., 2001)). The micropaleontological record reveals frequent sliding on a scale below the resolution of available acoustic records during this interval. The frequent (i.e., syndepositional) displacement and downslope transport of limited sediment volumes may have counteracted the development of major events during this time by reducing slope angles as well as hampering the increase of pore pressure on the upper slope.

6.3 Predisposing Factors and Failure Mechanisms

6.3.1 Sedimentation Rates and Excess Pore Pressure

Sedimentation rates at the study site are found to reach maximum figures of 2 m/kyr during the last glacial-interglacial cycle. This is an order of magnitude higher than at the neighboring ODP Site 963 at NW Gela Basin (Emeis et al., 1996; Sprovieri et al., 2003) and was suggested to be related to the presence of strong along-slope bottom currents and increased sediment discharge during hyperpycnal river floods (Kuhlmann et al., 2015). Both factors favor the rapid deposition of fine-grained and undercompacted sediment (e.g., Warrick et al., 2008) and thus the generation of excess pore pressure in underlying sediments, as demonstrated by Dugan and Flemings (2002) in a hydrodynamic simulation of the New Jersey continental slope margin. Steady bottom currents induce additional deposition of contourites with good sorting and thus high water contents (Laberg & Camerlenghi, 2008; Verdicchio & Trincardi, 2008), contributing to a state of general weakness of the slope. Another mechanism increasing the overpressure within the sediments is the deposition of landslide events themselves, which exert a rapid loading on the buried material, thus inducing the upward migration of pore fluids and successive generation of overpressure (Minisini & Trincardi, 2009). Evidence for this mechanism is documented in the seismic profiles by the frequent wipeout structures related to channelized fluids within well-stratified units topping landslide deposits (Figures 5d and 6). Additional evidence of such fluid escape structures in the region is documented by Taviani et al. (2013).

6.3.2 Prefailure Stratal Architecture and Weak Layers

The vast majority of recognized MTDs in the study area can be related to translational mudflow deposits involving movement along sub-horizontal bedding or glide planes. This basal confinement suggests a strong influence of geological key surfaces (i.e., weak layers) on the failure mechanisms. Such surfaces may involve a change in accumulation rate (e.g., a dramatic increase in sediment accumulation rate above a condensed interval), lithology (e.g., mineral types and grain size) or stratal geometry and possible paths of fluid migration. For NE Gela Basin, previous studies speculated about the role of sequence-stratigraphic boundaries in this context that relate to paleoenvironmental changes in the depositional setting, such as (cyclic) variability in sediment flux and sea level (Kuhlmann et al., 2014, 2015). Such surfaces typically coincide with distinctive physical attributes with regard to the surrounding material and hence favor preferential sliding in shallow burial. These attributes can be ascribed to lithological changes as well as alteration processes such as cementation, compaction, or biogenetic processes (e.g., Laberg & Camerlenghi, 2008; Lee et al., 2007).

However, the recovered deposits at the base of GB I and GB II suggest a second type of key surface that may precondition the slope to failure and that relates to an origin other than paleoclimatic variability. Enriched in both Ti and Fe and with elevated magnetic susceptibilities, these layers possess a higher fraction of terrigenous input and feature significantly different lithologic characteristics with regard to the background sedimentation of silty nannofossil clay. Microscopic and instrumental investigation of minerals and textures in thin sections from these layers testify a volcaniclastic origin of the material, likely deposited by a flow as indicated by the lamination and alignment of elongated minerals (Figures 7 and 8). Though the presence of a volcaniclastic layer at the base of both MTDs drilled at site GeoB14401 is intriguing, it remains unclear whether these layers indeed represent the basal bedding surface of the associated deposits or whether they are simply a reworked part of the slide deposited at its base—maybe even a clast of what was once the weak layer. Given the lack of core control on an undisturbed, prefailure equivalent, there is no unequivocal evidence for one or the other. This is especially true for the volcaniclastic layer at the base of GB I, as core control on the overlying MTD is rather limited and counterdirected deformation bands indicate rotational movement associated with shearing of the volcaniclastic layer itself (Figure 7a). In contrast, unambiguous signs of reworked sediment above (Figure 4b) and undisturbed sediment below (see section 4) the volcaniclastic layer at the base of GB II provide considerably more arguments for the presence of an actual weak layer. Indeed, the detachment of GB II deposit and upper volcaniclastic layer (see line scan in Figure 8) may suggest a similar mechanism to that proposed by Harders et al. (2010), who related failure processes to glass shard rearrangement during seismicity-induced ground shaking and associated particle internal water expulsion. Similarly, Wiemer & Kopf (2017) observed major weakening of soft-grained and porous volcanic material due to excess pore pressure buildup during undrained monotonic shear experiments. Although volcaniclastic content of the GB II basal bedding plane may be reduced if compared to these studies, significant differences in grain size, porosity, and the presence of microfracturing within the volcaniclastic layer may lead to a similar process when subjected to seismic shaking. This is supported by the study of Kuhlmann et al. (2016) on marine marker tephra Y-7 recovered from the undisturbed proximal part of the study area at site GeoB14403, who proposed failure initiation at the interface of the volcaniclastic layer and above lying strata due to transiently elevated pore water pressures triggered by seismic shaking.

6.3.3 Seismic Activity

Considering the high frequency of submarine landsliding within Gela Basin (section 6.2), a trigger mechanism of repetitive character is required in order to recurrently invoke conditions capable of destabilizing the sedimentary successions on the continental slope. As frequently suggested, seismic activity is the most plausible mechanism for this. Ai et al. (2014) showed in a geotechnical study of borehole GeoB14403 that at current state, a relatively small horizontal acceleration of 0.03–0.08 g may be sufficient to cause failure on the upper slope. Such an external stimulus could be provided by either moderate earthquakes (M = 4.0–4.8) in the direct vicinity of the study area (<10 km) or stronger (M = 7) events in an epicentral distance <20–80 km. As introduced earlier, the instrumental and historical series of the U.S. Geological Survey National Earthquake Information Center (USGS NEIC) database document comparably low seismicity at Gela Basin (Figure 1b), with only seven epicenters in the immediate proximity showing low to moderate magnitudes of M = 2.8–4.0 (since 1970). Though these numbers may not suggest seismicity as a strong and recurrent local trigger mechanism for submarine landsliding, they indicate significant seismic frequency if extrapolated to longer temporal intervals. Minisini et al. (2007) proposed a rough estimate of about 2000 seismic shocks over the period of the post-LGM interval. They additionally argued that full account of the maximum possible earthquake magnitude and spatial distribution of the epicenters in the historical record may have been hindered by a variety of reasons: (1) the difficulty to infer past offshore epicenters, (2) the bias of onland distribution of historical earthquakes associated with the location of ancient settlements, and (3) the incidental absence of potential extreme events in the short interval of observation. Hence, the regional seismic potential may be underestimated with regard to the limited temporal interval at the base of the USGS database and seismic activity is considered as a main agent controlling the generation of sediment landslides at NE Gela Basin.

7 Conclusions

This study presents a detailed analysis of landslide activity and failure mechanisms at the continental margin of Gela Basin. A total of nine MTDs could be identified along the lower slope section, representing predominantly disintegrative mudflows that display preferential movement along subhorizontal, bedding-parallel layers. Accurate timing of the individual events was achieved through acoustic correlation of the deposits with two drilled boreholes that provide centennial to millennial scale age control by integrating biostratigraphic reconstructions, carbon and oxygen isotope stratigraphies, and radiocarbon dating. Results reveal a basic failure return interval of ~10 kyr, but temporal distribution shows no clear predominance of failure related to sea level changes. This figure excludes several syndepositional and small-scaled failure processes of lower volume that are not discernible in the acoustic signal but are suggested by the micropaleontological record.

A major factor in predisposing slope failure appears to be the depositional margin architecture as shaped by Milankovitch-type climatic oscillations, as well as the intercalation of volcaniclastic material into the rapidly accumulating sediment. Though there is a lack of concluding evidence, the latter may provide mechanically weak surfaces that act as preferential planes of failure. However, slope destabilization may require additional seismic shaking to generate transiently elevated pore water pressures in the interface between volcaniclastic layer and overlying strata. Further predisposing factors capable of reducing slope stability in this area are the exceptionally high sedimentation rates, fluid expulsion, and generation of excess pore pressures in response to rapid loading.

Another noteworthy finding is the occurrence of very small-scaled landslide processes during the LGM chronozone (~19–23 kyr B.P.), which are not discernible in the acoustic record but provide direct evidence of downslope transport in the benthic foraminifera assemblage. These may significantly add to the high sedimentation rate during this interval (~2 m/kyr) and may have counteracted the emplacement of larger landslides.

Acknowledgments

This work has been funded through DFG-Research Center/Cluster of Excellence “The Ocean in the Earth System”. We are grateful to the captain, crew, and scientific party of the R/V Odin Finder, R/V Urania, and R/V Maria S. Merian for their assistance with data acquisition. δ¹³C/δ¹⁸O and grain size analyses as well as MSCL and XRF logging were carried out at the lab facilities of MARUM. This data is made available through the PANGEA data repository (https://doi.pangaea.de/10.1594/PANGAEA.871955). The earthquake catalog is available through the USGS NEIC database (http://earthquake.usgs.gov/earthquakes/search/). Multibeam and Parasound data can be requested from the German Oceanographic Data Centre (DOD-Ref-No.20100157).