Geophysical Research Letters

Potential-field modeling of collapse-prone submarine volcanoes in the southern Tyrrhenian Sea (Italy)



[1] Hydrothermal alteration may weaken volcanic rocks, causing the gravitational instability of portions of active volcanoes with potentially hazardous collapses. Here we show high-resolution multibeam, magnetic and gravity surveys of the Marsili seamount, the largest active volcano of Europe located in the southern Tyrrhenian back-arc basin. These surveys reveal zones with exceptionally low densities and with vanishing magnetizations, due probably to the comminution of basalts during hyaloclastic submarine eruptions and to their post-eruptive hydrothermal alteration. The location of these regions correlates with morphological data showing the occurrence of past collapses. Similar evidence has been obtained from pre-existing data at Vavilov Seamount, another older volcanic system in the Tyrrhenian back-arc basin. Here a large volume of at least 50 km3 may have collapsed in a single event from its 40 km long western flank. Given the similarities between these volcanoes, a large collapse event may also be expected at Marsili.

1. Introduction

[2] Collapses of large portions of submarine volcanic rocks represent a major risk for tsunami generation, producing some of the most destructive events [Smith and Shepherd, 1996; McGuire, 2006]. Hydrothermal processes, in conjunction with pre-existing weak and faulted rocks, plays a dominant role in causing these catastrophic collapses [Lopez and Williams, 1993; Finn et al., 2001; Reid, 2004]. Particularly, the Tyrrhenian Sea shows evidence of past tsunamis [Tinti et al., 2004].

[3] Marsili (Figure 1), a 70 km long, 30 km wide volcanic edifice elongated NNE–SSW, rises 3000 m from the relatively flat 3500 m deep southern Tyrrhenian Sea back-arc basin, where extensional tectonics related to the subduction of the Ionian slab [Rosenbaum and Lister, 2004] has generated spreading and injection of oceanic-type basalts in the last 2.0 Myr [Kastens et al., 1998; Marani and Trua, 2002]. Cocchi et al. [2009] also showed that Marsili started its vertical accretion approximately 1 My ago; it can be interpreted as a superinflated ridge due to a strong pulse of magma upwelling through a tear in the subducted Ionian slab [Marani and Trua, 2002]. Fractional crystallization studies of Marsili basalts by Trua et al. [2002] suggest the presence of an active magma chamber roughly 2.5 km below the summit of the volcano.

Figure 1.

Geological setting of the Marsili Basin (MB) and survey line-path. The line with triangular barbs represents the subduction of the African plate below the European plate, and the Aeolian Islands are the corresponding volcanic arc.

[4] We performed gravity, magnetic and multibeam bathymetric surveys at Marsili with the R/V Universitatis, completing a 1400 km long set of parallel lines orthogonal to the ridge trend. Magnetic data were acquired by using a Caesium magnetometer towed 200 m from the ship's stern and the measured field was corrected for heading, lag and diurnal variation. We detected a widespread region with low magnetic anomalies, less than 500 nT, running along the volcano's ridge, approximately enclosed within the black rectangle in Figure 2a. A negative circular anomaly with values down to 100–200 nT is located above the Marsili's top. The standard magnetization of basaltic rocks would generate a positive anomaly of more than 1000 nT at the Marsili's top (Figure 3) and along the ridge inside the black rectangle in Figure 2a. These results indicate vanishing magnetizations. Reversed magnetization is an unlikely explanation of this pattern, since the summit of the volcano is made of rocks that acquired their magnetization during the present normal geomagnetic field [Cocchi et al., 2009], i.e., magnetic chron c1n [Cande and Kent, 1992]. The E and W flanks of the volcano are made instead by rocks older than 0.78 Ma [Cocchi et al., 2009] showing reversed magnetization, i.e., magnetic chron c1r.1r [Cande and Kent, 1992].

Figure 2.

Magnetic and gravity anomaly maps. (a) Magnetic anomaly map (total-intensity). The red line shows the location of the profile modeled in Figure 3. The entire region of the volcano's ridge enclosed in the black rectangle is characterized by low values of the magnetic anomaly acquired during the current normal polarity chron C1n (≤0.78 Ma). The lateral symmetrical negative anomalies are originated from reversed magnetization (≥0.78 Ma). (b) Complete Bouguer gravity anomaly compiled with a reduction density of 2.6 g/cm3, where the low-density region along the volcano's ridge is clearly visible.

Figure 3.

(a) Magnetic and (b) gravity model along a profile intersecting the Marsili's summit. (d, e) The corresponding models of a seamount made of uniform non-altered oceanic crust are shown for comparison, as are (c, f) the corresponding cross sections.

[5] Gravity data have been acquired by using a Lacoste&Romberg Air/Sea system II, and have been processed for tide, Eotvos and drift corrections. The complete Bouguer anomaly, computed with a density of 2.6 g/cm3 for the oceanic crust (Figure 2b), shows a large elongated minimum beneath the ridge axis, with corresponding rocks well below the density of basalts, i.e., 2.4–2.8 g/cm3 [Carlson and Herrick, 1990; Gilbert et al., 2007]. Correlation between gravity and bathymetry suggests an average density of 2.0 g/cm3. Even introducing Airy or flexural compensation [Watts, 2001], the estimated densities do not exceed 2.2 g/cm3. Moreover, the young age of Marsili and the local Moho depth derived from seismic data [Sartori et al., 2004] exclude the possibility of deep compensating roots beneath the seamount, but rather show a gentle eastward deepening of the crust-mantle interface towards the continental slope. These data suggest that compensation, if present, is either mostly of Pratt type or is due to a compensation level deeper than the Moho. However, the characteristic wavelengths of compensation are larger than the scale of our survey; thus, they should not contribute significantly to modelling the observed gravity data in the volcano's core region. The comparison between Figure 2a and Figure 2b show that the black rectangle encloses the low-density and low-magnetization regions. In particular, the low-density region appears wider. It is likely that the Marsili's ridge is characterized by low-density rocks, which provide a good basis for the hydrothermal fluid circulation, but the alteration of the magnetic properties of the rocks occurred just in some preferred regions.

2. Modeling and Discussion

[6] For simplicity of representation and due to the elongation of the volcano's ridge in the perpendicular direction we show a joint 2.5D forward model of both Bouguer and magnetic anomalies, performed along a survey line intersecting the volcano's summit approximately 500 m b.s.l. (Figure 3), where the correlation between gravity and magnetic anomalies is particularly evident. The base of the model is constrained from available Moho information [Sartori et al., 2004] and the top and strike lengths by bathymetry. This model, though simple, gives a clear picture of the Marsili's structure, which indeed is confirmed by supplementary inversions we performed on the measured anomalies [Cocchi et al., 2009; Caratori Tontini et al., 2008]. The model in Figure 3 consists of a two-layers oceanic crust, representing seismic layers 2 and 3, with densities of 2.6 g/cm3 and 2.8 g/cm3, respectively. The lateral layers with reversed magnetization (chron c1r.1r) belong to the oldest portion of the Marsili volcano. The model for the volcano's summit consists of a large altered region, with a vanishing magnetization of 0 A/m and a density of 2.0 g/cm3, which is needed to fit the negative magnetic and gravity anomalies simultaneously. The existence of an active, hot, non-magnetic magma chamber 3 km below the summit [Trua et al., 2002], with density 2.3 g/cm3 based on volatiles concentration [Wallace, 2005], plays an important role in reproducing the measured negative magnetic and gravity anomalies. The same accounts for the heat-flux data [Della Vedova et al., 2001].

[7] Two factors may have contributed to creating low-density/low magnetization masses on Marsili: (a) fragmentation of the basaltic lava upon eruption on the seafloor; and (b) post-eruption hydrothermal alteration of the basalts. Mechanisms of submarine basaltic eruptions depend on viscosity of the melt during eruption and on eruption rate; they can give rise to different products, such as sheet flow, pillow lavas and hyaloclastites [Bonatti and Harrison, 1988; Schiffmann et al., 2006]. Hyaloclastites can reach very low densities (∼2.0–2.2 g/cm3) and magnetizations. Fragmentation and comminution of the melt can occur due to thermal shattering during submarine eruptions, and/or to gas exsolution and expansion during decompression. The latter process can be important if the volatile content of the melt is high and if pressure is low, i.e., if the eruption takes place at relatively shallow depths. The summit of Marsili is presently 500 m below sea level; its depth could have been less than 100 m during Pleistocene low sea level stands [Wang et al., 1991]. Moreover, the volatile content of Marsili melts is likely to have been relatively high, given their origin in a subduction-related backarc setting [Wallace, 2005]. Pillow lavas and hyaloclastites could have been the prevalent eruptive forms at Marsili, and extensional faults may dissect the volcano.

[8] The high permeability and the heat source present within and below Marsili [Della Vedova et al., 2001] favour intense hydrothermal circulation and subsequent basalt alteration. Hydrothermal alteration of basalt with formation of secondary hydrated phases can further lower its density and drastically decrease its magnetization [Woodward and Mumme, 1993]. Evidence of hydrothermal circulation on Marsili derives from sea water 3He/4He anomalies found just above the summit [Lupton et al., 2008] and from metalliferous deposits found on its flanks, which also confirm that a shallow magma chamber is still active [Dekov et al., 2006]. We obtained direct evidence of the presence of hydrothermal chimneys from the Marsili's summit by towed cam images, and altered basalt samples were also dredged at the volcano's summit at 650 m depth.

[9] Pruis and Johnson [1998] and Cochran et al. [1999] estimated porosities as high as 38% from gravity measurements in young oceanic crust. However, a density of 2.0 g/cm3 can be realized only if, in addition to a high porosity, the volcano contains large volumes of low-density altered basalts and hyaloclastites. Calculations of the loads introduced by these low density regions produce a nearly compensated model, with a deviation of 3% from a pure Pratt-like compensation. If the Marsili altered region and magma chamber were made of fresh basaltic rocks, no negative magnetic and gravity anomalies would be present at the volcano's summit (Figures 3d and 3e), with differences of approximately 1000 nT (magnetics) and 40 mGal (gravity) from the observed values, supporting the proposed model.

3. Conclusion

[10] The volume of the altered region, estimated as nearly 100 km3, represents a large mass of weak rocks. Given the volcano's steep flank slopes and the large porosity of the altered region, eruptive or seismic events could disrupt the unstable equilibrium of these rocks, generating collapse of large masses. Multibeam morphological data give evidence of past collapses of portions of Marsili (Figures 4a, 4b, and 4c) deducted from asymmetric slopes of the volcano's flanks and differentiations in the contour irregularity [Grosse et al., 2009]. Similar collapse features have been also identified at Vavilov seamount (Figure 4d). The Vavilov system, with dimensions and orientation similar to those of Marsili, is located 200 km NW of Marsili in a backarc setting that was active before the Marsili system (Figure 1). Vavilov is older (3.0 Ma) than Marsili, but with a similar genesis and rock composition, thus representing a possible scenario of Marsili's fate in the forthcoming 2 Myr. Moreover, negative Bouguer anomalies suggest that Vavilov, just as Marsili, contains low-density masses [Morelli, 1970]. The entire 40 km long W flank of Vavilov (a volume of at least 50 km3) appears to be missing, in a collapse that may have taken place in a single event (Figure 4d). Similar large collapses may well take place in the future at Marsili. Both Marsili and Vavilov are located close to circum-Tyrrhenian highly populated areas, i.e., southern Italy, Sicily and Sardinia, and collapses of large masses from these volcanoes have the potential for generating major tsunamis.

Figure 4.

Evidence of past collapses. (a) Morphological details of past collapses at Marsili seamount in the yellow box in Figure 4c. (b) Asymmetric bathymetric profile (vertical exaggeration 10×) at Marsili along the yellow dashed line in Figure 4c. The steep gradient of approximately 18° on the E side is compatible with a major collapse. (c) Bathymetric map of Marsili, with magnetic polarities and the locations of the box (Figure 4a) and the yellow profile (Figure 4b). (d) Bathymetric map of Vavilov seamount, showing the major collapse of the W flank with an estimated volume of at least 50 km3.


[11] The data have been acquired as part of the Prometheus project cruise. Financial and logistic support was provided by Prama S.r.l and particularly thanks to Diego Paltinieri and to Patrizio Signanini. We thank the captain, officers and crew of R/V Universitatis of Conisma for their assistance during the Prometheus project cruise, and Eni E&P Div. for the use of the gravity meter. D. Scheirer, D. P. Hill, and an unknown reviewer provided helpful suggestions. Maps were produced with the Wessel and Smith's GMT package [Wessel and Smith, 1991].