Basin inversion: Reactivated rift structures in the Ligurian Sea revealed by OBS

. The northern margin of the Ligurian Basin shows notable seismicity at the Alpine front, including frequent magnitude 4 events. Seismicity decreases offshore towards the Basin centre and Corsica, revealing a diffuse distribution of low magnitude earthquakes. We analyse data of the amphibious AlpArray seismic network with focus on the offshore component, the AlpArray OBS network, consisting of 24 broadband ocean bottom seismometers deployed for eight months, 15 to reveal the seismicity and depth distribution of micro-earthquakes beneath the Ligurian Sea. Two clusters occurred between ~10 km to ~16 km depth below sea surface, within the lower crust and uppermost mantle. Thrust faulting focal mechanisms indicate compression and an inversion of the Ligurian Basin, which is an abandoned Oligocene rift basin. The Basin inversion is suggested to be related to the Africa-Europe plate convergence. The locations and focal mechanisms of seismicity suggest reactivation of pre-existing rift structures. Slightly different striking directions of faults 20 in the basin centre compared to faults further east and hence away from the abandoned rift may mimic the counter-clockwise rotation of the Corsica-Sardinia block during ~20-16 Ma. The observed cluster events support the hypothesis of strengthening of crust and uppermost mantle during rifting related extension and thinning of continental crust.


Introduction
Earthquakes of magnitude 4 are frequently recorded in the Ligurian Basin, especially at the Ligurian margin at the junction 25 between the southwestern Alps and the Ligurian Basin (hereafter named the Alps-Liguria junction) with maximum magnitudes of Mw 6 to 6.5, indicating a moderate seismic activity (Béthoux, 1992;Courboulex et al., 2007;Béthoux et al., 2008;Larroque et al., 2012Larroque et al., , 2016. Seismic activity is highest along the Côte d'Azur and the Ligurian coast and decreases towards the central basin and Corsica (Fig. 1). The Ligurian Sea formed during Oligo-Miocene times as a back-arc basin (Burrus, 1984;Rehault et al., 1984;Faccenna et al., 1997;Gueguen et al., 1998;Rosenbaum et al., 2002), but extension stopped ~16 Ma. Today, GPS Here, we report on local seismicity in the centre of the Ligurian Basin (Fig. 1a). We analysed two earthquake clusters (Fig. 1b) that were recorded by an amphibious seismic network operated in the framework of the European AlpArray initiative (Hetényi et al., 2018). The AlpArray OBS network, a long-term ocean bottom broadband seismometer (OBS) array, recorded ground motion continuously for 8 months. The deployment of long-term OBS enables robust source estimates of earthquakes 45 far away from land stations. Nevertheless, observations from the land stations improved the estimate of fault plane solutions.
The observed earthquake clusters and their depth distribution provide constraints on the crust and upper mantle rheology and provide insights into today's regional stress field.

Geological and geodynamic setting
The Western Mediterranean, including the waters west of Apennines and Sicily, consists of several basins (Fig. 2) that formed 50 from Oligo-Miocene times to the present (e.g. Burrus, 1984;Rehault et al., 1984). The geodynamic setting is controlled by the convergence of the African and Eurasian plates (e.g. Dercourt et al., 1986) and the rollback of the Apennines and Calabrian subduction zones (Jolivet and Faccenna, 2000). The fast rollback of the westward migrating Gibraltar arc and the eastward migrating Apennines arc played a major role in the opening of the Mediterranean sub-basins (Lamotte et al., 2000;Mauffret et al., 2004;Handy et al., 2010). The Ligurian Basin, opening ~30-16 Ma (Burrus, 1984;Bache et al., 2010), and the Alboran 55 Basin, opening ~25-20 Ma (Faccenna et al., 2004), are the oldest basins in the Western Mediterranean. The Algerian Basin, opening ~16-8 Ma (Mauffret et al., 2004), and the Tyrrhenian Basin, opening ~8-0 Ma (Faccenna et al., 2001;Rosenbaum et al., 2002), are the youngest basins in the Western Mediterranean. In the latter two basins, geophysical and geological data clearly show that extension caused break-up and seafloor spreading (Nicolosi et al., 2006;Bouyahiaoui et al., 2015;Prada et al., 2016;Booth-Rea et al., 2018), while the Alboran Sea is a domain of extended continental crust modulated by arc 60 magmatism (Booth-Rea et al., 2018;Gómez de la Peña et al., 2020). The Balearic Sea is floored by oceanic crust (Gailler et al., 2009;Afilhado et al., 2015;Moulin et al., 2015). However, the extent and nature of the oceanic domain towards the northeast into the Liguro-Provencal Basin remains debated.
The Calabrian trench retreats further south-eastwards (Fig. 2, white arrow), however, recent seismological observations indicate that Tyrrhenian Sea opening ceased and the Africa-Eurasia convergence results in basin inversion at its southern rim (Zitellini et al., 2020). It is proposed that the major shortening caused by the convergence between Africa and Europe is accommodated in the Maghrebides ranges in North Africa (90-100%) and that the Ligurian Basin and the Corsica-Sardinia block are rigid (Nocquet and Calais, 2004;Béthoux et al., 2008;Nocquet, 2012). However, passive seismic studies observe signatures of compression in the Ligurian Basin (Béthoux, 1992;Baroux et al., 2001;Eva et al., 2001;Courboulex et al., 2007;Béthoux et al., 2008;Larroque et al., 2012Larroque et al., , 2016. Moreover, active compressional structures were imaged in seismic reflection profiles (Bigot-Cormier et al., 2004) offshore the Ligurian coast. Along the Corsica margin, seismic reflection data do not 80 image significant seismic ruptures that reached the surface since ~5 Ma, but earthquakes of ML 5.5 and ML 4.4 occurred in July 2011 offshore Corsica (Larroque et al., 2016) (Fig. 1, grey star). While the Ligurian margin is narrow and steep with a few listric normal faults, the Corsica margin is wider and several listric faults were imaged in seismic reflection data (Finetti et al., 2005). Within a short distance of 30-50 km, the crust-mantle boundary (CMB) deepens from ~15 km depth in the basin to ~25 km depth at the continental margins (Contrucci et al., 2001;Gailler et al., 2009;Dessa et al., 2011).

Data and results
The AlpArray OBS network consisted of 24 broadband OBS ( Fig. 1) that continuously recorded ground motion for ~8 months

Events, picking and location
During the long-term OBS deployment, 39 seismic events were detected within the Ligurian Basin excluding the Alps-Liguria junction zone (Fig. 1a, magenta circles). Our work focuses on two earthquake clusters in the centre of the Ligurian Basin, near OBS A423A (magenta circles in Fig. 1b; Table 1). The first cluster (C1) consists of 13 events that occurred from June to November 2017 (Fig. 3a, blue bars). The second cluster (C2) consists of three events that occurred during one day in January 105 2018, about 25 km southwest of C1 (Fig. 3a, red bars). Two events (grey bars in figure 3a) had few observations and high uncertainties, they were not used in this interpretation. Two more small events were only observed at station A423A and were not further analysed (Fig. 3a, black bars). We show that a seafloor network could detect more events than those in the ISC catalogue (Tab. 1), whose magnitude of completeness is 2.2 in the region, but the four events (grey and black bars in Fig. 3a) indicate that the AlpArray OBS station spacing is too large to render a more precise picture of the local seismicity.

110
Initially, the ISC bulletin (Storchak et al., 2017) was used to detect seismic events in the AlpArray OBS data. The phases were picked manually on all stations plotted with orange triangles, orange squares and white hexagons in Figure 1b. Only the seismometer components were used in our analysis since the signal-noise ratio on most of the hydrophones was too low to identify local seismic events.
Three onsets predominated the observed events: the P-wave, a converted Ps-wave and the S-wave (Fig. 3b). The P-wave is 115 weak in amplitude and followed by a stronger Ps-phase. Both phases indicate an apparent P-wave velocity (Vp) of ~8.1 km/s and they are separated by a nearly constant time difference of ~1 s. The S-wave phase has a high amplitude, compared to the P-onsets, and shows an apparent S-wave velocity (Vs) of ~4.7 km/s (Fig. 3b). The waveforms are highly coherent as shown for the vertical component of station A423A (Fig. 4a). Picking was done for the P-onset on the vertical component and for the S-onset on the horizontal components using SEISAN (Havskov and Ottemoller, 1999).

130
We use only stations within the basin (orange triangles in Fig. 1) to locate the events. In this way, we can use a 1-D seismic velocity model for the basin and avoid errors introduced by the extreme topography and changes in crustal thickness near the https://doi.org/10.5194/se-2021-9 Preprint. Discussion started: 5 February 2021 c Author(s) 2021. CC BY 4.0 License. margins. We applied the 1-D velocity model (Fig. 4b) that is based on P-wave velocities from the seismic refraction profile p02 (Dannowski et al., 2020) (Fig. 5e). A Vp/Vs ratio of 2.0 at the seafloor and 1.87 above the CMB was assumed to convert Vp to Vs (Fig. 4b). The observed apparent velocities of mantle refracted waves Pn and Sn (Fig. 3b) were used as Vp and Vs 135 for the uppermost mantle in the 1-D velocity models (Fig. 4b). After an initial event location, using SEISAN (Havskov and Ottemoller, 1999), events of the two clusters were relocated with hypoDD, a double-difference earthquake algorithm for relative relocations (Waldhauser and Ellsworth, 2000). Because our clusters containing only a few events and the background velocity model is well-constrained, we used hypoDD's singular value decomposition (SVD) solver. Epicentres of events observed in both the ISC bulletin and our study, were compared. Pairs are connected by black lines in the 145 seismicity map (Fig. 1). The epicentres based on the ISC bulletin are spread over a 50x20 km zone. In contrast, the same events located solely based on OBS data plot in a very narrow 3x3 km zone ( Fig. 1b and 5b). Additional events with magnitude ML <2 were detected using the OBS; these events were also recorded by the land stations ( Fig. 1)

Focal mechanisms
To estimate fault plane solutions we used the stations shown in orange in figure 1. The station distribution provided a good azimuthal coverage for the Ligurian Sea. Additionally, the permanent station VSL in southern Sardinia (located outside the 155 map of Fig. 1a) was used. The first motion direction of the P-wave was determined for on-and offshore stations where clearly visible. The amplitude ratio of P-and S-wave was determined on the vertical component at land stations only. The ocean bottom stations showed unusual small amplitudes for the P-wave compared to the S-wave, indicating systematic faulty recording. Polarity and amplitude ratios were used to derive the fault plane solution of events 2, 6, 8, and 9 ( Fig. 4; Tab. 1) by means of the program FOCMEC (Snoke, 2003). In general, the events of cluster C1 show stronger amplitudes compared to 160 C2. The four fault plane solutions of cluster C1 indicate thrust faulting ( Fig. 4c and Fig. 5a). The arrivals from events of C2 were not of sufficient quality to calculate focal mechanisms.

Basin inversion
Geodetic measurements in Corsica and Sardinia show residual velocities <0.5 mm/y with respect to stable Europe (Nocquet and Calais, 2004), which the authors use to conclude that the shortening at the Alps-Liguria junction, expressed in widely spreading centre, since no spreading axis had been mapped. Our study corroborates this hypothesis, since clusters C1 and C2 are located in the basin centre and would be even closer to or at the proposed spreading axis and so ridge-push forces would be higher. Additionally, no spreading axis was mapped in more recent seismic studies and the LOBSTER seismic refraction profile p02 (Dannowski et al., 2020) proposes that rifting failed before seafloor spreading was initiated.
To summarise previous studies, sources for the regional compressional stresses could be: (1) Africa-Europe convergence, (2) 190 CCW rotation of the Adriatic plate (Larroque et al., 2016), or (3) north-eastward motion of the Tyrrhenian Sea towards stable Europe (Nocquet, 2012). Geodetic stations in Northern Africa are still missing, excluding reliable constraints on plate motions based on geodetic measurements (Nocquet, 2012). The latest plate motion models (Nocquet, 2012;Le Breton et al., 2017, 2020van Hinsbergen et al., 2020) are based on seismicity and other geophysical and geological information and indicate that the majority (90-100%) of Europe-Africa convergence is accommodated in the Maghrebides. The Ligurian Basin and the 195 Corsica-Sardinia block are seen as rigidly attached to stable Europe (Nocquet, 2012).
The epicentres of cluster C1 were located in the uppermost mantle (one event in the lower crust) and their mechanisms are thrust faulting (Fig. 5), while the epicentres of cluster C2 were located in the lower crust, above the CMB. The 2011 events were also thrust faulting events which occurred in the crust and uppermost mantle (Larroque et al., 2016), roughly 50 km E-SE of clusters C1 and C2. If we project C1 and C2 on line A-B that follows the push direction of the thrust events, they map 200 in a slightly tilted vertical plane dipping south-eastwards (Fig. 5d). Independently of the source of regional stresses, we rifting-related fault planes was observed in other areas, for example in the Tyrrhenian Sea (Zitellini et al., 2020), in the Gulf of Cadiz  and in northern Honshu, Japan (Kato et al., 2009). Therefore, like the Tyrrhenian Sea (Zitellini et al., 2020), the Ligurian Sea may have entered a stage of basin inversion.

Implications on rifting history
210 While earthquakes are spread over the entire Ligurian Basin (Fig. 1), the C1 and C2 events cluster within small areas. C1 and C2 are separated by ~25 km in a NE-SW direction, suggesting that both clusters may originate from the same fault zone. The focal mechanisms of cluster C1 are similar to the 2011 events (Larroque et al., 2016). In detail, we observe a difference of ~15° between the striking direction of the fault planes of C1 and the striking direction of the 2011 events. As discussed before, we interpret the three clusters as caused by the reactivation of Oligocene-Miocene rifting structures. Normal faults that were 215 created during the extensional phase are turned into reverse faults and their strikes are those of the normal faults during the extensional phase. Since the 2011 events are located more to the southeast, they represent an older phase of rifting compared to C1 and C2. Thus, we speculate that this difference in strike might be connected to the 23° CCW rotation of the Corsica-Sardinia block compared to stable Europe that took place between ~20-16 Ma (Speranza et al., 2002;Le Breton et al., 2017).

220
Earthquakes in continental domains normally occur in the upper crust, while the lower crust is relatively aseismic and earthquakes are rare in continental mantle lithosphere (Maggi et al., 2000). Whereas earthquakes certainly occur in oceanic crust and mantle lithosphere (Wiens and Stein, 1983). Similar to oceanic lithosphere, Maggi et al. (2000) suggest that the strength of continental lithosphere is controlled by the temperature structure and the amount of water in the formation and resides in one seismogenic layer within the crust. They propose that continental mantle lithosphere is relatively weak, and 225 thus, aseismic. However, episodes of rifting may affect crustal and mantle rheology.
Rifting models suggest that the rheology of the continental domain changes during extension at non-volcanic rifted margins (Pérez-Gussinyé and Reston, 2001). With increasing stretching, the portion of the crust that becomes brittle increases.
Stretching of the lithosphere will cause lower temperatures at the crust-mantle boundary, which in turn, will strengthen the entire crust. When the entire crust is brittle, faults can cut through the crust into the mantle and act as fluid pathways, a pre- The cluster events discussed here are interpreted to reflect pre-existing normal faults generated during rifting. We recognise a good relationship between our data and the rifting model of Pérez-Gussinyé and Reston (2001), suggesting that the entire 235 continental crust may have evolved into a brittle domain during extension. The C1 events at mantle depths are more puzzling.
Earthquakes within continental mantle lithosphere were also observed in the Gulf of Cadiz, in the southern Iberian old Jurassic mantle lithosphere . An explanation could be that during extension, mantle material moves closer to the surface than before stretching, causing mantle temperatures to decrease (Sandiford, 1999). Thus, the mantle in the basin centre might become stronger and more brittle than the surrounding mantle (Sandiford, 1999). Such a scenario is supported by 240 cluster C1 occurring in the uppermost mantle and supporting a low Vp/Vs-ratio of 1.72 (Fig. 4b). The crustal structure in the vicinity of clusters C1 and C2 is well imaged by the LOBSTER seismic refraction profile p02 (Dannowski et al., 2020), which provides no indication of a high amount of mantle serpentinisation at its southern end. A high sedimentation rate during rifting High heat flow (>100 mW/m²) in the Ligurian Basin (Pasquale et al., 1994) may contradict a cool CMB at the basin centre.
However, Hansen and Nielsen (2002) showed that a combination of a very thick sedimentary cover (>8 km), extremely low conductive sediments (<1.5 W/mK) and very shallow and localised crustal radiogenic heat production allow for a temperature 250 maximum at the CMB beneath the basin centre. The thick sedimentary cover in the Ligurian Basin (up to 7 km, Schettino and Turco, 2006) might cause thermal blanketing, reducing the lithospheric heat loss, in line with the observed high heat flow values in the basin centre compared to the margins.

Conclusions
The entire Ligurian Basin is characterised by sparse but wide-spread micro-earthquakes of magnitude <3.

285
The authors declare that they have no conflict of interest.

Author contributions
MT carried out data analysis and contributed to the manuscript. AD wrote the manuscript and contributed to the data analysis.

Team list
The complete member list of the AlpArray Working Group can be found at: http://www.alparray.ethz.ch/en/seismic_network/backbone/data-policy-and-citation/.