1 Introduction

The occurrence of tsunamis in different areas around the world and the intense urbanization of the coastal zones (SOER 2015) drew the attention of the scientific community in the last decades. Scientists have been motivated to develop methodologies to trace the extension of areas possibly exposed to tsunami risk, as main step to the hazard assessment along the coasts.

Grezio et al. (2012) define the risk as Risk = Hazard × Exposed elements × Vulnerability in accordance with the general definition of UNESCO (1972) and Fournier D’Albe (1979); this definition was adopted for evaluating several geophysical risks (Glade 2003; Pesaresi et al. 2007; Lirer et al. 2001, 2010). Hazard is the probability of occurrence of a dangerous event in a fixed future time. Exposure measures people, property, systems or other elements present in hazardous zones that are thereby subject to potential losses. Vulnerability is the proportion of lives or goods likely to be lost and accounts for the characteristics of a system or asset that make it susceptible to the damaging effects of a hazard (Alberico and Petrosino 2015).

The contouring of coastal zones potentially exposed to tsunami waves represents an important step towards the development of a system able to react and/or to cope with such an event (i.e. adaptive capacity and/or coping capacity). In the last decade, probabilistic and deterministic approaches were developed to assess the tsunami hazard taking advantage from computer systems.

The probabilistic method requires data on a significant number of events; therefore, it is a method used mainly for tsunami triggered by earthquakes. Relationships linking earthquakes occurrence and tsunamis incidence were found to identify the probability rate of tsunamis (Tinti 1991a, b; Tinti and Gavagni 1995; Orfanogiannaki and Papadopoulos 2004; González et al. 2009). For tsunamis induced by slides or by volcanic activity, average recurrence times of the events are much more difficult to identify owing to the overall lack of data (Ward 2001). In this case, a deterministic approach (Tinti et al. 1999a, 2006a, 2011), possibly joined with the scenario concept, has to be invoked. The worst-case scenario and the multi-scenario (Papathoma and Dominey-Howes 2003; Tinti and Armigliato 2003; Tinti et al. 2008; Eckert et al. 2012; Omira et al. 2010; Nguyen et al. 2014; Dall’Osso et al. 2014; Alberico and Petrosino 2015) are the principal methods used to delineate the hazard zones. The former considers only the largest event occurred in the past, while the latter takes into consideration more events with different magnitudes.

In the present work, we have carried out a comprehensive analysis of potential and historical tsunamigenic events that might have struck the city of Napoli from near (from Venotene Gulf of Napoli to the Gulf of Salerno) and far (other Italian subregions and Western Mediterranean Sea) fields. The aim of the paper is to present a methodological approach useful to identify potential tsunamigenic sources as first step towards the tsunami hazard assessment in areas characterized by scarce data and a non-conservative environment as the Gulf of Napoli. Specifically, this approach consists in the integration of results coming from: (a) a critical review of tsunamigenic sources from the available tsunami database and scientific literatures; (b) the analysis of scientific literature concerning tsunami simulations in the Western Mediterranean Sea aimed at identifying the possible exposure of the study area to events generated in the far field; (c) the computation of the maximum tsunami amplitude that could be generated by submarine landslides through the coupling of scar morphometry with predictive tsunami amplitude equations (e.g. Watts 2000; McAdoo and Watts 2004; Rahiman and Pettinga 2006).

The main outcome of this integrated analysis is to point out the potential tsunamigenic sources and associated features that may affect the Gulf of Napoli, providing insights for a proper tsunami hazard assessment, a fundamental goal for scientific community and land-planners, mainly in consideration of the intense anthropization and tourist exploitation of this coastal sector.

2 Materials and method

The identification of tsunami events potentially hazardous for the city of Napoli outcomes from a multifaceted workflow, made up of four phases: (a) recovery and recording of data derived from available tsunami database and literature into tables, (b) data analysis, (c) rough assessment of tsunamigenic potential for submarine landslides, (d) summary of data. To better link the concept of potential hazard to the territory, the tsunami sources were grouped and analysed in far- and near-field events. The latter includes the events occurred in the Gulf of Napoli and in the adjacent areas (from Ventotene Island to the northern sector of the Gulf of Salerno), whereas the former the events occurred mostly in the Western and Central Mediterranean Sea.

In the first phase, the tsunami data available in scientific literature and EMTC database, for a time interval ranging between 2000 years B.C. to present, were collected (Fig. 1a). In addition, several extreme events out of this time interval were considered. The data were arranged into tables and managed in a Geographic Information System (GIS) framework, allowing to link all type of information to the geographical position of the tsunami source and to use the spatial analysis to draw maps.

Fig. 1
figure 1

Input data and associated sources used for the tsunami events valuation. The Italian Tsunami Catalogues (grey ellipses) and scientific works (yellow boxes) are the main source of triggering causes, exposed zones (subregion), time of occurrence, earthquake magnitude and tsunami intensity data. The data encircled in the pale yellow rectangle were collected from the scientific works reported in the coloured boxes on the right side of the image. Abbreviations meaning: CFB (Caputo and Faita Tsunami Catalogue, 1984 updated by Bedosti and Caputo in 1986), TMC (Tinti Maramai Catalogue;  Tinti and Maramai 1996), ITC (Italia Tsunami Catalogue; Tinti et al. 2004), ITC2 (Italian Tsunami Catalogue Two; Tinti et al. 2007), EMTC (Euro-Mediterranean Tsunami Catalogue; Maramai et al. 2014)

Full size image

In the second phase, five indices (Table 1), useful for the tsunami events classification, were evaluated for: (a) triggering mechanisms (earthquakes, volcanic eruptions, landslide) and exposed zones (subregion), (b) characteristics of the trigger mechanisms (earthquake magnitude, Volcanic Explosivity Index, volume range), (c) tsunami parameters (tsunami intensity). On this latter point, the scarcity of data prevented the identification of indices for other tsunami parameters (i.e. water height, run-up, tsunami amplitude and extension of inland inundation), so that their hazard potential was expressed with a value range, identified from the analysis of 17 events, occurred mainly in the southern Italian subregions.

Table 1 Indicators and indices for triggering causes, subregion, earthquakes, magnitude, Volcanic Explosivity Index (VEI) and tsunami intensity are shown
Full size table

The index for the triggering mechanisms was calculated as the ratio of number of tsunamis triggered by a specific event (e.g. earthquakes) and the total number of tsunamis recorded into the GIS framework. For the feature of triggering mechanisms (e.g. earthquake magnitude), the index was calculated as the ratio between the number of tsunamis pertaining to the single class of triggering mechanisms (e.g. earthquake magnitude values: unknown, 3.2–4.2, 4.2–5.2 etc.) and the total events with a specific triggering mechanisms. In the same way, the tsunami parameters (e.g. tsunami intensity) index was calculated.

Moreover, a brief description of tsunami simulations performed in the Western Mediterranean Sea and that could propagate in the Tyrrhenian Sea was made. For the single simulation, the main feature of triggering causes, exposed zones and tsunami amplitude were summarized.

In the third phase, a simplified method coupling data on scar morphometry with predictive tsunami amplitude equations was used off the Latium and Campanian coastline, similarly to marine geo-hazard studies realized in other marine settings (Locat et al. 2004; Goldfinger et al. 2000; Bohannon and Gardner 2004; McAdoo and Watts 2004; Rahiman and Pettinga 2006; Casalbore et al. 2011, 2017; Casas et al. 2016). In detail, the analysis of high-resolution Digital Terrain Models obtained by swath bathymetric data (D’Argenio et al. 2004; Budillon et al. 2011a, b; Casalbore et al. 2014) was carried out to identify the morphometric parameters of the submarine mass failures identified in the near field. Those parameters are needed to assess the maximum tsunami amplitude (A) according to the equation of Rahiman and Pettinga (2006):

$$\begin{aligned} \lambda & = 3.87\left( {bd/\sin \sigma } \right)^{0.5} \\ A & = 0.224 T\left[ {w/\left( {w + \lambda } \right)} \right] \times \left[ {\left( {\sin \sigma } \right)^{1.29} - 0.746\left( {\sin \sigma } \right)^{2.29} + 0.170\left( {\sin \sigma } \right)^{3.29} } \right]\left( {b/d} \right)^{1.25} \\ \end{aligned}$$
(1)

where λ is the tsunami wavelength, b is the length of the scar along the slope, w is the width of the scar parallel to the slope, d is the water depth at the landslide barycentre, T is the head scarp height (assumed as a proxy of the maximum initial failure mass thickness normal to slope), and σ is the mean slope of failure scar. Moreover, to take into account the key role of the volumes of mass failure for the hazard assessment, only the landslide scars resulting from a unique mass-wasting event were considered in the analysis.

Despite the potential inaccuracy in using equations with parametric coefficients derived for a different part of the world and with different conditions, the equation was roughly calibrated on the basis of the computed maximum amplitude and observed run-up of tsunami waves generated by the tsunamigenic landslide occurred in 2002 in the relatively nearby Stromboli Island (Casalbore et al. 2011).

In the fourth phase, a summary of all data (e.g. indices, synthetic descriptions) was made to identify the potential tsunamigenic sources for the city of Napoli, associated with events occurred both in far- and near-fields.

3 Tsunami data analysis and hazard index

At the end of eighteenth century, tsunamis were included as secondary events in earthquakes (see Mallet 1854; Baratta 1901) or volcanic eruptions catalogues (Mercalli 1883; Baratta 1897). Since 1947, when a tsunami catalogue regarding all continents was published by Heck, several tsunami databases recording new detailed information were implemented (Fig. 1). The most recent database concerning the European-Mediterranean and Eastern Atlantic Regions (EMTC—Maramai et al. 2014) records 221 tsunamis in the Mediterranean Sea, whose 72 occurred in the M2 zone (Central Mediterranean Basin). The analysis of the events allowed us to define five indices expressing the dangerousness of tsunami sourcing in both far (Western Mediterranean Sea) and near (Gulf of Napoli and in the adjacent areas—from Ventotene Island to the northern sector of the Gulf of Salerno) fields with respect to the city of Napoli.

3.1 Triggering causes and subregions

According to the formula (1), the number of events (N ic ) and the percentage of triggering causes for the M2 zone were assessed. As reported in Fig. 2a, b, the tsunami events were triggered for the 75% by earthquakes, 14% by volcanic eruptions and 3% by both landslides and combined sources classes (Fig. 2a, b). The same indices, identified for each Italian subregion, highlight that the Campania, Liguria-Côte d’Azur, Aeolian Islands and Tyrrhenian Calabria coasts record the highest number of events, whereas the Latium subregion, the lowest one (Fig. 2c, d).

Fig. 2
figure 2

Number and percentage of tsunami events classified for triggering causes (a) and graphic representation of tsunami percentage (b). The same type of data was reported for each Italian subregion in figures c and d

Full size image

3.2 Characteristics of triggering events

The data analysis showed that earthquakes with magnitude > 5.3 Mw triggered a large number of tsunamis and even a different threshold of magnitude for the genesis of tsunamis occurs between earthquakes generated offshore and on land. The number of events (N ic ) concerning both the single class of earthquake magnitude and of tsunami source (in land and offshore) is reported in Fig. 3a, b.

Fig. 3
figure 3

The number of tsunamis and hazard index are reported for the single earthquake magnitude class and for source location (on land and offshore). These value was also plotted in the diagram “b”. Similarly, the number and hazard index of tsunamis triggered by volcanic eruption was shown in figures “c” and “d”, respectively

Full size image

The Volcanic Explosivity Index is known for the tsunamis triggered by volcanic eruptions; 7 out of 10 have a VEI ranging from 2 to 3. Similar to the earthquake magnitude index, the VEI data were used in the formula (1) to identify the index reported in Fig. 3c, d.

3.3 Tsunami parameters

The analysis of tsunami intensity pointed out that the greatest part of events (event percentage: 58%) has an intensity of 2, whereas only the 16% has an intensity higher than 3 (Fig. 4a). This parameter ranges from 2 to 4 for the events induced by volcanic eruptions, from 2 to 3 for those triggered by landslides, while a wider range varying from 2 to 6 characterizes the events triggered by earthquakes (Fig. 4b–f). For the latter, the highest percentage of tsunamis has an intensity of 2 and their percentage decreases of about three times for tsunami intensity of 3 and only 10 tsunamis have an intensity higher than 3 (Fig. 4b–f).

Fig. 4
figure 4

The number of tsunamis events and hazard index are reported for each tsunami intensity classes and triggering causes (a). The figure “a” illustrates the percentage of tsunamis grouped only for intensity, while the figures from “b” to “f” point out, for the single tsunami intensity classes, the percentage of tsunami events classified for triggering causes

Full size image

The analysis of field survey data retrieved from scientific papers, coupled with the information recorded into EMTC, pointed out that few information on tsunami parameters (water height, run-up, run down, inland inundation, wave amplitude, wave length, period, frequency) are available.

The tsunami amplitude, run-up and inland inundation parameters are often available only for few events. Moreover, these parameters are deeply conditioned by the local morphology of both sea floor and emerged area that prevent the assessment of experimental laws at regional scale (Szczucinski et al. 2006; Billi et al. 2010; Nakamura et al. 2012). Despite the paucity of information, we tried to summarize the available data in order to identify, for each parameter, the range of values to better constrain the hazard of tsunami events (Table 2a, b; for detail see online Appendix 1). The tsunami amplitude varies from 0.75 to 15 m, with the highest value characterizing the Eastern Sicily followed by Messina Straits and Aeolian Islands. The run-up ranges between 1 and 11.7 m, with the maximum value recorded in Messina Straits for the 1908 event. The inland inundation varies from 5 to 2400 m, with the maximum value still recorded in Messina Straits subregion.

Table 2 Summary of data for tsunami events affecting the Italian subregions retrieved from EMTC database and scientific literature. They were grouped in three sections: the triggering causes, features of triggering causes, tsunami features. In addition, the events providing data on tsunami features are listed in (b) (for detail on single event see the online Appendix 1)
Full size table

4 Analysis of tsunami simulations performed in the Mediterranean Sea

The analysis of scientific literature concerning tsunami simulations aimed at identifying the possible exposure of the study area to events generated in the far field. The simulations of tsunami occurred in western Mediterranean Sea are widely described in Table 3, where the information on source area, main feature of triggering causes, exposed zones, tsunami amplitude and references is recorded. In contrast, we overlooked the tsunamis generated in the eastern Mediterranean Basin, because the Messina Strait acts as a barrier to their propagation in the Tyrrhenian Sea (Lorito et al. 2008).

Table 3 Summary of data about tsunami simulated in the Western Mediterranean Region and affecting the Italian subregions
Full size table

4.1 Tsunami generated by earthquakes

The geodynamic setting of Western Mediterranean basin is characterized by seismogenic faulting zones, as the Tell-Atlas thrust belt, the Southern Tyrrhenian thrust belt and the Eastern Sicily, able to generate earthquake-generated tsunamis that can pose significant hazard for the Mediterranean coasts (Fig. 5) (Tinti et al. 2005, 2011; Lorito et al. 2008; Sahal et al. 2009). Seven simulations of tsunamis triggered by earthquakes are reviewed and summarized in Table 3; they concerned the events occurred: (a) on 21 May 2003 (performed by Tinti et al. 2005), and on 10 October 1980 (simulated by Lorito et al. 2008) for the Algeria-Tunisia zone; (b) on 5 March 1823 along the offshore seismic belt (simulated by Lorito et al. 2008) in the Southern Tyrrhenian zone; (c) on 11 January 1693 (Tinti et al. 2005; Tonini et al. 2011) for the Eastern Sicily (Fig. 5). This latter event was possibly generated offshore, according to Tinti et al. (2001) and was located along the Hyblaean-Malta escarpment by Tinti and Armigliato (2003). However, Tonini et al. (2011) speculate that this event was related to a submarine landslide (volume = 4.1 × 109 m3; Argnani et al. 2012) along the Malta escarpment, which took place in waters deeper than 1800 m, triggered by a non-tsunamigenic earthquake occurred on the mainland. Tonini et al. (2011) also simulated the tsunami triggered by the strong earthquake occurred in the Messina Straits on 28 December 1908.

Fig. 5
figure 5

a Topographic and bathymetric map of Western Mediterranean domain; the line symbols represent simplified geological structures (modified from Faccenna et al. 2003; Cavazza et al. 2004; Billi et al. 2011). The orange stars are the location of tsunami simulated in the Western Mediterranean Sea, while the numbers in the brackets indicate: 1-French Italian Riviera, 2-Alboran ridge, 3-Tell-Atlas thrust belt, 4-Marsili Seamount, 5-Palinuro Seamount, 6-Stromboli, 7-Southern Tyrrhenian thrust belt, 8-Messina Strait, 9-Hyblaean Malta escarpment. The yellow box indicates the location of study area illustrated in detail in the lower part of the image; b bathymetry off the Campania-Lazio regions, in the central Tyrrhenian Sea, by a composite swath bathymetric dataset of IAMC. Deep data from Emodnet, http://portal.emodnet-bathymetry.eu/

Full size image

In the westernmost Mediterranean area, the Alboran Ridge is the only zone able to generate earthquake induced tsunamis. For this area, the results of several simulated events (Álvarez-Gómez et al. 2011) generated by earthquakes with magnitude ranging between 6 and 7 are reported in Table 3. In addition, Ioualalen et al. (2014) pointed out that the earthquakes sourcing on the French–Italian Riviera are recurrent but only strong earthquakes, such as the event occurred on 23 February 1887, could induce tsunamis (Table 2b).

4.2 Tsunami generated by landslides

A large number of tsunamigenic landslides were generated in subaerial coastal settings or shallow marine areas, typically at the shelf edge or upper continental slope (Bondevik et al. 2005; Chaytor et al. 2007; Billi et al. 2008; Lo Iacono et al. 2011; Mazzanti and Bozzano 2011). For most of these events, it is still unclear if the tsunami was caused by co-seismic slip, by earthquake or by a combination of both mechanisms. In the Western Mediterranean, most failures have limited volume, short run-out and source in relatively deep water. Therefore, only the largest although infrequent events are likely to trigger large tsunamis (e.g. Camerlenghi et al. 2010; Urgeles and Camerlenghi 2013). However, catastrophic sediment failures, such as the BIG’95 (11.500 cal year BP, the main features are summarized in Table 3) can also happen along continental margin not affected by significant earthquakes or active faulting (Lastras et al. 2002), such as those characterizing the northwestern Mediterranean Sea.

The propagation of a tsunami that may be triggered by a submarine landslide occurred on the southern flanks of the Alboran Ridge was simulated by Macias et al. (2015) (Table 3). They pointed out that shape and propagation patterns, wave amplitude and, finally, the area that may be exposed to this event (South Iberia and North Africa) are mainly influenced by the basin morphology.

Also in the Ligurian Sea, at the France–Italy border, numerous submarine landslides have occurred in the past along the narrow continental slope. The effects of the event occurred on 16 October 1979, simulated by Ioualalen et al. (2010), are summarized in Table 3. This event was divided into two main phases: the initial rupture at the shelf break involving a volume of 2.2 × 106 m3 in shallow water; the ensuing phase of erosion and increasing volume occurred at about 800 m depth along the slope. The volume comparison has revealed that the moving slurry flow increased its volume 25 times in less than 3–4 km and in a short time period. A similar event had occurred 2 years before at the head of Gioia Canyon (Southern Tyrrhenian Sea), with a mobilized volume of 5.5 × 106 m3, generating 5-m tsunami amplitude (Colantoni et al. 1992). More interestingly, in the surrounding area Casas et al. (2016), using high-resolution DEM, recognized several potentially tsunamigenic landslides occurring in shallow water.

Moreover, we also considered the inferred tsunamis which might have been triggered by landslides occurred in active volcanic area. In the last 13 ka BP, four coaxial flank collapses affected the NW flank of Stromboli volcano, the last one forming the Sciara del Fuoco (SdF) depression (Tibaldi 2001).The first of these collapses involved the Upper Vancori edifice about 13 ka BP (Tibaldi 2001) with an estimated volume exceeding 2 × 109 m3, the second occurred in the early Holocene before 5.6 ka BP with a sliding mass in the order of 1 × 109 m3 (Tibaldi 2001), the third more recent than 5 ka BP, involved a total volume of material of about 5.2 × 108–9.5 × 108 m3 (Tibaldi 2001). The tsunami triggered by the last major slope failure was simulated by Tinti et al. (2003b), and the resulting characteristics are summarized in Table 3.

Marine studies recognized large debris avalanche deposits at the base both of the northwestern and eastern flank of Stromboli (Romagnoli et al. 2009a, b). The latter deposits on the E flank of Stromboli were attributed to (at least) two large sector collapses, the older one being comparable or larger in size to the collapses affecting the NW flank and the younger one (last few thousands of years) involving about 100 × 106 m3 (Romagnoli et al. 2009b). Based on the mobilized volume and similar morphological setting with the NW flank, they have been considered responsible for the generation of large tsunamis comparable to those computed for the NW flank (Casalbore et al. 2011 and reference therein).

In more recent times, the tsunamis generated by a submarine and a 7 min later subaerial landslide (volume of approximately 10 × 106 m3, and 4–9 × 106 m3, respectively; Chiocci et al. 2008), affecting the NE part of the Sciara del Fuoco (December 2002), were investigated by Tinti et al. (2006a, b). It is noteworthy that similar events likely occur with high recurrence times (i.e. some tens of years) at the Sciara del Fuoco depression, as testified by the re-interpretation of five similar tsunami events which occurred in Stromboli in the past century (Maramai et al. 2005a) and by results of multi-lapse bathymetric surveys to monitor the morphological evolution of the 2002 landslide scar (Chiocci et al. 2008).

Amongst the active volcanic seamounts located in the Tyrrhenian Sea, the steep slopes (25°–40°) of the Palinuro Seamount, coupled with the rock alteration and high fluid pressure, contribute to a slope instability often marked by mounded facies in the accumulation zone (Passaro et al. 2010 and reference therein). Similarly, the Marsili seamount showed signs of instability and eruptive or seismic events which might disrupt the unstable equilibrium of these rocks, intensely altered, generating mass failures (Ventura et al. 2013). Furthermore, Caratori Tonini et al. (2011) recognized a collapse in the western flank of the central-northern sector of Marsili seamount. The authors claim the threat associated with possible future mass failures in the area and the need of further investigations to assess the potential hazard.

4.3 Tsunami triggered by volcanic eruption

The Aeolian archipelago is an active volcanic area that should not pose significant hazard for the city of Napoli. Here, the tsunamis directly linked to volcanic activity have had only a local impact (Tinti et al. 2003a); as a matter of fact, only two tsunamis related to hot avalanche entering the sea occurred in the last 100 years; moreover, they only posed serious hazard only along the Aeolian Islands’ coasts (Maramai et al. 2005b).

5 Tsunamis potentially generated at Napoli Gulf

5.1 Volcano-tectonic activity and tsunami events

In a volcanic area, earthquakes, rapid inflation or deflation of volcanic flanks, pyroclastic flows and lahars can occur during an eruption; the tight relationship amongst co-volcanic (or volcanic-induced) phenomena makes very difficult the identification of specific tsunami triggering factors. For the Napoli Gulf, a combined reading of historical documents, field data and simulation results is needed to solve the problem of scarcity of data relatively to the feedback between these events.

At present, the Somma-Vesuvio is characterized by a low level of seismicity (a few hundred micro earthquakes per year), with most of the earthquakes having a local magnitude (ML) < 3.0. The hypocenters are mainly located at very shallow depths (less than 5 km), and the strongest earthquake occurred in 9 October 1999 (ML 3.6) (Convertito and Zollo 2011).

At Campi Flegrei, the ground deformations are accompanied by low-energy shallow seismicity (Dvorak and Gasparini 1991). During the last bradyseismic episodes in 1969–1972 and 1982–1984, a maximum uplift of about 350 cm was observed and more than 16.000 local earthquakes were recorded. Seismicity was characterized by swarm-type activity; the largest event (MD = 4.2) occurred on 8 December 1984 (Del Pezzo et al. 1987). A ground uplift up to 4 cm occurred in March 2000 and was accompanied by two swarms of about 50 low-energy (MD ≤ 2.2) earthquakes occurred on 2–7 July, and 22 August 2000.

At Ischia Island seismic activity is concentrated in the northern part of the island around the town of Casamicciola and is characterized by a very shallow hypocentral distribution with typical depths between 1 and 3 km. Several events with maximum intensity I ≥ VII MCS have occurred in the last few centuries (i.e. 1762, 1796, 1828, 1841, 1863, 1881 and 1883; Cubellis and Luongo 1998), causing landslides (Rapolla et al. 2010). Apart from the earthquake with intensity 6.2 MCS occurred on 22 August 2017, the island was characterized by low intensity seismicity over the last 130 years.

5.2 Analysis of tsunamis data and simulation results

The occurrence of anomalous waves has taken place not only during the largest explosive eruptions in the area (79 A.D., 1631) but also during low-size eruptive events (14 May 1698; 4 April 1906—Tinti and Maramai 1996) or prior to the beginning of volcanic activity (25 December 1813)—Lirer et al. 2009) (online Appendix 2). Moreover, a sea retreat was described for the eruption of 12 August 1804 (Scandone et al. 1993) and for the 1723 eruption (Lirer et al. 2009).

During the 1631 eruption, the contemporary accounts describe the occurrence of a tsunami reaching amplitude between 2 and 5 m in the Gulf of Napoli (online Appendix 2).

In his Geographica (source in Buchner 1986), the Greek writer Strabo (64 B.C.–21 A.D.) gives the description of a sudden collapse of a sector of the Ischia Island with an associated tsunami wave, dating prior to the Greek period, and recently dated between ~ 3 and 2.4 ka BP (Ischia Debris Avalanche, IDA; de Alteriis et al. 2010). At Ischia Island, mass failures are, in fact, quite common (Della Seta et al. 2012); in many cases, they were caused by large-scale collapses of the over-steepened and faulted flanks of Mt. Epomeo resurgent block (Buchner 1986). The occurrence of these events is supported by large-scale slide scar recognizable onshore and facing coastal sector as well as by the local hummocky topography of the island offshore (Budillon et al. 2003; D’Argenio et al. 2004; Chiocci and de Alteriis 2006; de Alteriis and Violante 2009).

In a recent scientific work, the presence of tsunamis hazard triggered by the occurrence of debris avalanches at the Somma-Vesuvio volcano was inferred by Milia et al. (2012 and reference therein). They recognize two debris avalanches of about 1 × 109 m3 (DA2), deposited in a time span very close to the Avellino eruption (4.2 ka BP—Santacroce et al. 2008), and of 3 × 109 m3 (DA1), occurred during the Basal Pumice Plinian eruption (22 ka BP—Santacroce et al. 2008), respectively. They also recognize a submarine fan (volume: 3 × 108 m3) interpreted as the result of the entry into the sea of AD 79 pyroclastic density currents.

The earthquake occurred in the Apennine Chain on 26 July 1805 (online Appendix 2) caused a rise of sea water in the Napoli Gulf (Poli 1806; D’Onofrio 1805; Baratta 1901).

Two tsunami simulations were performed for in the Napoli Gulf considering as triggering mechanisms the lightest component of pyroclastic flows as those occurred during large explosive eruptions of Somma-Vesuvio (Tinti et al. 2003a) and the Ischia Debris Avalanche (Tinti et al. 2011). Moreover, Zaniboni et al. (2013) have simulated a possible mass-failure event that might involve Monte Nuovo in response to a renewal of magmatic activity (Table 3).

5.3 Morphological-derived submarine mass failure parameters

The morphometric parameters of about 470 submarine landslide scars were measured between Ventotene Island (75 km to the NW of Napoli) and the northern sector of the Salerno Gulf (60 km to the SE) (Fig. 5b) in order to estimate the landslide-generated tsunami hazard from the marine sectors surrounding the city of Napoli. The landslide scars below 700 m depth were excluded from the analysis for their intrinsic low hazard in relation to their size.

In the case of the Ventotene edifice, a total number of 126 landslide scars have been identified between − 130 and − 1150 m (Casalbore et al. 2014), with areas ranging from 0.1 × 106 to 10 × 106 m2 (Fig. 6). The scars can be grouped into two types, one affecting the flanks of the edifice and the other occurring at the edges of the insular shelf, similarly to what was observed offshore other insular and oceanic islands (e.g. Romagnoli et al. 2013; Casalbore et al. 2015). However, none of these landslides was able to generate a maximum computed tsunami amplitude higher than 1 m. Taking into account the distance from the city of Napoli (ca. 75 km) and the inactivity of Ventotene edifice (last activity ended 0.15–0.3 Ma), the tsunami hazard related to landsliding processes affecting the submarine flanks of Ventotene edifice can be considered absent or very low for the Napoli city. Indeed, typical features of landslide-generated tsunamis are high amplitudes in the near-field and a rapid attenuation of the waves in the far field (Harbitz et al. 2006).

Fig. 6
figure 6

Shaded relief map and isobaths (equidistance 100 m) of Ventotene edifice allow the recognition and morphometric characterization of 126 submarine scars (modified by Casalbore et al. 2014), based on the coupling between scar morphometry and predictive equations available in literature (see text for details); none of them resulted as being potentially tsunamigenic

Full size image

Around 350 landslides were identified along the continental margin between Volturno and the Sele river mouth. The greatest number of failures (up to 20/10 km2) is observed north, west and south of the island of Ischia, a sector of the Campania continental margin largely controlled by volcanic features, such as cones, domes, dykes (Fig. 7). Landslide scars develop in water depth ranging between − 7 and − 900 m and occur in several morphological contexts: edge of submarine terraces, flanks of volcanic cones, canyon heads and walls, open slope, sedimentary aprons. Landslide scars are clustered around two modes: 0.20–0.4 × 106 m2 and 0.6–2.0 × 106 m2 wide, with volumes not exceeding 8 × 106 m3 and 13–25 × 106 m3, respectively. At least two failures, characterized by large volume and shallow coastal settings (− 35/− 240 m), resulted as potentially tsunamigenic with a computed maximum tsunami amplitude > 1 m. Differently, the large failure recognizable offshore the village of Casamicciola, at a water depth of about 7 m, partially remobilizing the northern debris avalanches, was not considered tsunamigenic, because it has been interpreted as a slow-moving slide by de Alteriis and Violante (2009).

Fig. 7
figure 7

Shaded relief map and isobaths (equidistance 100 m) of Ischia Island-Campi Flegrei offshore. The symbols indicate: (1) debris avalanches/debris flow deposit; (2) Banco della Montagna dome-like structure; (3) underwater creeping/slump deposit; (4) volcanic reliefs and banks (modified from Aiello et al. 2001; Milia and Torrente 2007; de Alteriis and Violante 2009; Della Seta et al. 2012); (5) non-tsunamigenic landslide (maximum tsunami wave amplitude < 1 m); 6) potentially tsunamigenic landslide (maximum tsunami wave amplitude > 1 m)

Full size image

A large number of landslides has been also observed at the headwalls and along the lateral walls of Magnaghi and Dohrn Canyons (D’Argenio et al. 2004; Milia 2000), in the depth range of 100–600 m. The instability processes have progressively enlarged the headwall of the Canyon Magnaghi through very small failures (< 0.08 × 106 m2). Differently, landslides in the canyon thalweg are at least one order of magnitude larger than the previous ones (0.2–1 × 106 m2). Several landslides (0.5–1 × 106 m2) have also been observed on the open slope between the canyon valleys (Fig. 7); they have a low potential hazard because located deeper than 200 m and characterized by volumes not exceeding 7 × 106 m3. In this morphological context, few landslide scars larger than 1 × 106 m2 were found at water depth shallower than − 300 m, implying a low potential tsunami hazard (Fig. 7).

The third group of submarine landslides runs along a Pleistocene structural lineament which bounds the southern side of the Sorrento Peninsula-Capri horst (Fig. 8) and controls the extension of the shelf edge (Milia and Torrente 1999; Casciello et al. 2006). The resulting 50-km-long escarpment is very steep (locally up to 20°) and an uneven seascape made of gullies, canyons, erosive scarps and aprons has developed through time. Locally, the emerged rocky cliffs are in continuity with the submerged slope and very small littoral wedges are preserved (Sacchi et al. 2009; Violante 2009). The greatest number (70%) of landslides is in shallow waters (< − 150 m) since the failures mostly occur along the shelf edge and account for a general retrogression of the continental slope. Despite the limited extension of the landslides, very high head scarps and very steep scar slopes have been observed. However, only two failures could have had a potential tsunami hazard, with computed maximum wave amplitude > 1 m high.

Fig. 8
figure 8

(modified from Milia and Torrente 1999; Casciello et al. 2006; Violante 2009)

Shaded relief map and isobaths (equidistance 100 m) of Capri Island-Sorrento Peninsula offshore. The symbols indicate: (1) deep sedimentary fan; (2) base of slope apron; (3) mass-wasting deposit (modified from Violante 2009; Budillon et al. 2014); (4) non-tsunamigenic landslide (maximum tsunami wave amplitude < 1 m); (5) potentially tsunamigenic landslide (maximum tsunami wave amplitude > 1 m); main fault.

Full size image

The southern edge of the Salerno Valley (Fig. 8) includes about 100 single failures, mostly distributed in 120–750 m depth range and a 14 km-long landslide complex, developed at 320–850 m water depth (Budillon et al. 2014). On the whole, they are translational landslides failed along weak layers with landslide scars characterized by several metres high head scarps and flat bottoms, corresponding to the slip surfaces. The largest single failure event identified in the area, which mobilized about 245 × 106 m3 of material, is associated with a low tsunami hazard, due to the deep-water setting (> − 400 m).

The morphometric analysis of mass failures in the near field pointed out that only four submarine landslides, with volume ranging from ca. 2.0 to 10.5 Mm3 (Table 4) and occurring in relatively shallow water (15–360 m), could be considered potentially tsunamigenic, as their computed maximum wave amplitude (A) is encompassed between 1.18 and 1.71 m. This finding agrees with the recent observation that small and medium scale landslides (volumes up to tens of millions of cubic metres) in shallow waters can create local but sometimes destructive tsunamis, as for instance observed at Gioia Tauro in 1977 (Colantoni et al. 1992), Nice in 1979 (Sultan et al. 2010) and Stromboli in 2002 (Chiocci et al. 2008).

Table 4 Morphometric parameters measured for the four potential tsunamigenic landslide recognized in the study areas (red lines in Figs. 7, 8)
Full size table

A proper modelling of landslide-generated tsunami propagation is beyond the aim of this paper, however, a first estimation of tsunami amplitude at the coast can be roughly made by applying the empirical relation between depth and amplitude of tsunami waves offshore (above the landslide scar) and in coastal sector (at − 1 m) provided by Federici et al. (2006). In the study area, the shoaling factor for the computed tsunami amplitude above the slide should be roughly comprised between 2 and 4.

6 Tsunami events

The identification of tsunami events possibly affecting the city of Napoli required the analysis of tsunamis occurred not only in the homonymous Gulf (near field) but also in all the Italian subregions. This paramount vision aimed to improve the knowledge on the events possibly affecting the city overcoming the paucity of quantitative data typical of a non-conservative environment as the volcanic areas. Moreover, the results of tsunami simulations realized in the Western Mediterranean Sea have been also considered.

The analysis of tsunami data concerning the Italian subregions and the simulations realized in the Western Mediterranean Sea highlighted:

  • the main tsunami triggering causes for the Italian Peninsula are the earthquakes and volcanic eruptions with 54 (index value = 75%) and 10 (index value = 9%) events out of 72, respectively. The Campania (index value = 13.89%), Aeolian Islands and Tyrrhenian Calabria (index value: 11.11%) are the subregions more exposed to these events.

  • The earthquakes magnitude index pointed out that 38 tsunami events out of 54 (index value = 70%) were generated by seismic events ≥ 5 Mw.

  • Only 14 tsunamis out of 72 showed an intensity > 3 Mw (rather strong, according to Ambraseys 1962); in detail, these events were mostly triggered by earthquakes (sum of intensity index = 13.89), followed by volcanic eruptions (intensity index = 2.78), earthquakes and landslide (intensity index = 1.39) and eruption and landslide classes (intensity index = 1.39).

  • The summary of tsunamis features evidenced a tsunami amplitude ≤ 4 m in 14 sites; higher values characterized tsunami events sourcing at Aeolian Island, Eastern Sicily, Messina Straits and Tyrrhenian Calabria and recorded in only seven sites (for detail see online Appendix 1). For the run-up and inland inundation tsunami features, a general rule could not be found since the local topography influenced the tsunami propagation. Moreover, the greatest part of sites reached by tsunami waves were characterized by an inland inundation lower than 200 m, only in five localities this value was overcome (Table 2a). These places were exposed to tsunami events occurred on the 22 May 1919 and 20 August 1944 at Aeolian Island, on the 6 February 1783 at Scilla (Messinia Straits) and on the 28 December 1908 in the Messinia Straits (for detail see Appendix 1).

  • The analysis of tsunamis simulation triggered by earthquakes (magnitude: from 6.2 to 7.4 Mw) sourcing in Southern Tyrrhenian and along the coasts of Algeria and Tunisia might generate tsunami amplitude ranging from 0.1 to 0.5 m in the Gulf of Napoli.

  • Volcanic eruptions occurred in the far field, as those characterizing the Aeolian Islands and Etna Volcano, have had only local effects (Table 3).

  • Landslides, with a volume ranging from few to several millions cubic metres, occurring in several sites of Mediterranean Sea did not propagate in the Gulf of Napoli (Table 3).

For the near field, the analysis of data pointed out a tsunami hazard mainly related to the Neapolitan volcanoes activity. The tsunami data analysis concerning the event occurred or simulated in the near field highlighted:

  • tsunamis with low intensity (value = 2) could be triggered by volcanic eruptions. The tsunami amplitude ranges from centimetres to a few metres. In detail, during the Somma-Vesuvio eruptions, a tsunami amplitude of 30–40 cm was recorded for the 1906 eruption and of 2–5 m for the 1631 eruptions; a value of 1.8–2.7 m was reported for the earthquake of 1805 (online Appendix 2).

  • The analysis of numerical simulations of tsunami events (Table 4) allowed to estimate a tsunami amplitude of about 20 m close to the coast of Napoli city linked to the debris avalanche occurred at Ischia Island (3 and 2.4 ka BP, de Alteriis et al. 2010) and of 0.4 m due to the propagation in the sea of the lightest component of pyroclastic flows sourcing at Somma-Vesuvio.

  • The morphometric analysis of submarine landslide scars along the continental margin nearby the city of Napoli has evidenced that only 4 out of over 450 were potential tsunamigenic landslides. They could have generated tsunami amplitude of a few metres in the city of Napoli. Because of the steep gradients and concomitant presence of multiple predisposing and triggering factors for the development of instability processes along the Campania continental margin off Napoli, the possibility of future submarine landslides with size similar to the previous ones cannot be neglected.

7 Conclusive remarks

The integration of tsunami data for the events occurred in both near (from Ventotene Island to the northern sector of Salerno Gulf) and far (from other Italian subregions and Western Mediterranean Sea) fields made possible to identify the main source areas of tsunami events which might affect the city of Napoli in the future. Concerning the tsunamis generated by earthquakes in the far field (Western Mediterranean), only those triggered by high magnitude events (value ≥ 6–7 Mw) might affect the city of Napoli in about 30′; however, the expected maximum tsunami amplitude would not exceed 0.5 m, according to the analysed data.

Concerning the landslide-generated tsunamis, those sourced in the far field are far to represent a serious hazardousness for the city of Napoli, if the recurrent instability of Stromboli Island is excluded. As its instability is concerned, mass failures with volumes higher than those occurred in December 2002 (e.g. the sector collapse at 5 ka BP) could represent a potential hazard for the study area.

Concerning the landslide-generated tsunamis in the near field, the most critical source area could be represented by the southern submarine slope of Ischia and Sorrento peninsula where, in a recent past (20 ka BP), four landslides have occurred with a size and depth that could potentially trigger tsunamis with a maximum wave amplitude between 1.18 and 1.71 m. In addition, a possible sector collapse is thought to be occurred at Ischia in historical time, promoting the generation of tsunami waves up to 20 m high.

The hypothetical occurrence of a failure involving the Monte Nuovo at Ischia Island as much as the propagation of lightest components of pyroclastic flows emitted from the Somma-Vesuvio volcano could reach the city within 10–20 min with a tsunami amplitude ranging from 0.4 to 1 m.

This research aims to support the drawing up of a territorial plan sensitive to all natural hazards impending on the investigated area. The tsunami hazard has to be considered in the process of multi-hazard assessment for a correct allocation of resources dedicated to the risk reduction. In this frame, this work represents a first step towards the assessment of Neapolitan areas possibly exposed to tsunami inundation.