Something Still Remains: Factors Affecting Tsunami Risk Perception on the Coasts Hit by the Reggio Calabria-Messina 1908 Event (Italy)

Abstract

Mediterranean coasts are prone to tsunamis due to high seismicity in some well-known areas near plate margins. However, tsunamis have a low frequency of occurrence despite having highly destructive potential. The low frequency of occurrence and historicity of the most destructive events lead to minimizing or neglecting this risk. Past research identified socio-demographic and spatial factors that may affect tsunami risk perception. This research is based on CATI survey (Computer Assisted Telephone Interview) to a sample of 5842 respondents designed to investigate whether and how risk perception and risk knowledge were affected by a major event such as the 1908 Reggio Calabria Messina tsunami, by making a comparison between areas hit by that event and unaffected areas, also providing some explanatory hypotheses. Despite differences between Calabria and Sicily, data show higher levels of tsunami risk perception in the area affected by the 1908 event, along with a major role of interpersonal sources, playing a relevant role in information gathering and understanding. Research also suggests the need to better integrate different sources of knowledge to improve people’s understanding so as to effectively cope with tsunami risk.

1. Introduction

With its 8300 km of coastline and about 16 million residents in coastal municipalities, Italy is significantly exposed and vulnerable to tsunami risk. Tsunami are little-known events but nonetheless they are likely to occur, as shown by a number of major events that have occurred in historical times, including the Reggio Calabria-Messina earthquake of 28 December 1908, which triggered a tsunami with waves up to 11 m high [1,2]. More recently, another tsunami event occurred in Stromboli on 30 December 2002, following a volcanic eruption. Advances in scientific knowledge about the sources capable of generating tsunamis, tsunami modeling with high-performance computational systems, and early warning systems have certainly improved the prediction and management capabilities of this type of event [3]. In particular, historical and geophysical analyses have highlighted the different risk exposure of Italian coastal slopes and the particular probability distribution function of this type of phenomena: major events such as the Crete tsunami of 365 AD, which destroyed Alexandria, Egypt, or the one that struck Naples in 1348, described by Francis Petrarch [4,5,6,7] have a low probability of occurrence, while smaller but potentially locally destructive tsunamis, such as the five events that struck the Eastern Aegean between 2017 and 2021 have a much higher probability of occurrence, sometimes leading to local damage or casualties, as for the event that affected the Greek and Turkish coasts on 30 October 2020 [8,9].

However, the advancement of scientific knowledge does not automatically translate into viable knowledge for vulnerable populations, which is instead recognized as an important predictor of preparedness and thus the effectiveness of risk mitigation measures [10,11,12,13]. In the areas most exposed to tsunami risk (particularly along the Pacific coasts and in some areas of the Indian Ocean) traditional forms of knowledge about this type of phenomena, generally based on the memory of past events, are integrated into local cultures and preserved through written documents, orally passed down stories, and accounts from direct witnesses, representing a powerful factor in communities’ adaptation to this type of event [14,15,16]. The case best known to the scientific community concerns the island of Simeulue in Indonesia. Despite its proximity to Banda Aceh, one of the areas most affected by the impact of the 2004 Sumatra tsunami, and direct exposure to tsunami waves with run-up more than 30 m, very few casualties have been reported. Subsequent research showed that traditional knowledge of tsunamis, conveyed through a folk song, had enabled the inhabitants to understand the danger and come to safety [17].

In the Euro-Mediterranean area due to their long return times, tsunamis are a virtually unknown phenomenon, of which little historical and cultural evidence remains. In this respect, authors have found only two exceptions: the case of the Lynden Fjord and the Åknes fjord area, both on the west coast of Norway and the Reggio Calabria-Messina area, selected as a case study for the here presented paper.

Some Norwegian fjords are prone to rockslide tsunamis and there have been several events in the area in the relatively recent past, in 1905, 1934 and 1936, respectively, which claimed 174 lives [18]. In these areas, the historical memory of past tsunamis together with scientific advice and operational Early Warning Systems has resulted in an enhanced tsunami risk perception [19,20].

The case study dealt with in this paper concerns the area affected by the Reggio Calabria Messina tsunami of 28 December 1908, in which systematic research on risk perception has never been carried out, which, hypothetically, may reveal evidence of great interest in the historical memory of disasters and risk communication.

The present work is a part of a larger research program on tsunami risk perception on Italian coasts, partially funded by the Department of Civil Protection [21,22]. This study is precisely focused on socio-demographic and geographical factors that may potentially affect tsunami risk perception along with forms and sources of knowledge about the nature, magnitude and danger posed by such phenomena. The research program relies on a questionnaire survey (CATI) that considered a sample of 5842 respondents, residing in 450 coastal municipalities located in 8 Italian regions. A sub-sample of 1999 interviewees in 193 municipalities of Sicily and Calabria were extracted from the national sample for the present study, providing an effective snapshot of the areas affected by the physical effects of the after mentioned Reggio Calabria and Messina earthquake and tsunami as well as of their surroundings, where the tsunami has not left documented traces of its impacts.

These data revealed interesting differences between the areas affected and not affected by the 1908 tsunami and a novel understanding of possible role of local knowledge for both Disaster Risk Management (DRM) and Risk Communication strategies. The paper not only shows different representations of the phenomenon and its effects, but also different ways to approach available sources of knowledge and their impact on risk perception. In the non-affected 1908 area, the way tsunami risk is understood is affected by a strong tendency to identify the tsunami as a catastrophic and unpredictable phenomenon, also highlighting a strong influence of mediatized images of the great tsunamis of Sumatra in 2004 and northern Japan in 2011. However, for the 1908 tsunami-affected area of Reggio Calabria and Messina, a more realistic reading of the phenomenon and its effects prevails, appearing linked to a concrete and socially shared historical memory of the event. Some of the research data will be briefly presented in the report, discussing the implications for science and risk communication strategies, forecasting and prevention activities, and possible future developments of the research.

1.1. Tsunami Hazard in the Mediterranean Area and Italian Coast

Scientific studies on tsunamis, in recent decades, experienced rapid progress. The evolution of seismic and sea-level detection tools connected through broadband networks to high-performance computers makes it possible to process large volumes of data in a very short time and to rapidly create models for estimating the impacts of a tsunami on coastlines [23,24,25]. In addition, in recent years, the use of supercomputers for High Performance Computing (HPC), allowed the development of detailed models of tsunami hazard estimation at the local, basin, and global levels by simultaneously examining a very large number of variables [26,27,28]. These computations allow researchers to go beyond hazard estimation by assessing in detail the impact that a tsunami may produce on the coast to define inundation zones and ways to mitigate the effects on the most vulnerable structures and population [3,16,29].

The Mediterranean Sea has a high tsunami hazard [25] and the Italian peninsula is located in the centre of the basin. Both the position of a nation and the estimated hazard are affected by the variability of factors such as the exponential increase, in recent years, of human settlements along the Mediterranean coasts (exposure) and the strong uncertainty related to tsunami hazard assessment. These characteristics complicate tsunami risk management, since these variables are combined in non-linear ways, thus increasing the complexity of estimating hazards and the level of exposure of coastal communities thereby also the risks.

After World War II, the Mediterranean coasts have experienced extensive urbanization, and many industrial facilities, for organizational convenience, grew up on or very close to the coasts [30]. Cities such as Alexandria, Athens, Izmir, Tripoli, Tunis, Algiers, Barcelona, Marseille, Naples, or Zadar, on the Mediterranean coasts, are exposed to tsunami risk. Throughout history, the Mediterranean coasts have been affected by many tsunamis—mostly generated by high-magnitude earthquakes at sea or near the coast—sometimes of large scale and with effects observed from wide distances. Among the first documented tsunamis in historical sources is the event that, in 365 A.D., affected the eastern Mediterranean Sea and which recent studies [5,31] estimate to have originated near Crete Island. The tsunami’s strength caused widespread damage from Greece to Sicily to Dalmatia. Egyptian writers reported very accurate details of the observed effects along the coast of Alexandria, Egypt. Most of the tsunamis in the Mediterranean Sea have been generated by strong earthquakes along the Greek arc, including the 1956 Amorgos tsunami, the Aegean Sea tsunami that affected Greece and Turkey in 2017, and the 2018 Ionian Sea tsunami that mostly affected the Zakynthos Island. The most recent tsunami that affected the eastern Mediterranean Sea dates back to 30 October 2020. This event hit the coasts of Greece, especially the island of Samos and Turkey, particularly the province of Izmir where one person died due to the tsunami [9,32].

Several factors make Italian coast highly exposed and vulnerable to tsunamis. These include:

(a)

strong seismicity of the coastal and surrounding areas, related to the convergence margins of the plates (African and Arabian with the Eurasian one) as well as crustal blocks and micro-plates (e.g., Adriatic Plate);

(b)

active volcanoes capable of triggering tsunamis (aerial and submarine factors, e.g., Stromboli and Marsili Mount);

(c)

over 8000 km of coastline surrounding the whole Italian Peninsula;

(d)

presence of medium and large-sized coastal cities such as Genoa, Venice, Naples, Bari, Cagliari, Palermo, Catania and of course Reggio Calabria and Messina;

(e)

high density of tourist venues and facilities across the whole coastline;

(f)

high density of industrial hub facilities and ports across the whole coastline;

(g)

estuaries of many rivers that may facilitate inundation in the event of tsunamis.

Italy, throughout history, has been hit by several major tsunamis. Among these are the tsunamis that affected the coasts of Puglia in 1627, the eastern coasts of Sicily in 1693, the coasts of Calabria and Sicily in 1783, again the Calabrian coasts in 1905 and then the Calabrian and Sicilian coasts in 1908 [33,34,35,36].

1.2. The 1908 Reggio Calabria-Messina Tsunami

On 28 December 1908 at 05:21 a.m., an earthquake of estimated magnitude greater than 7, struck southern Calabria and north-eastern Sicily [1]. The earthquake caused a tsunami that, in a few minutes, reached the coast. The highest values were recorded in Pellaro, near Reggio Calabria, where the waves reached 13 m in height. In Stromboli waves of about 10 m were measured and in Messina the waves reached 3 m in height causing the collapse of the port quay and widespread damage in the city. The tsunami was also recorded at the tide gauges of Naples and Civitavecchia, respectively, 300 and 500 km away. It is estimated that the tsunami caused about 2000 victims or more, which added to the approximately 80,000 victims caused by the collapses and effects of the earthquake [1]. The considered area has a primary relevance for the study for the following reasons: (1) the estimated global intensity (S-A; P-I) [37,38,39] of the phenomena was up to 6 (disastrous) at a local level and the scale of the event is actually comparable to other major tsunamis such as those which occurred in Tyrrhenian Calabria in 1783 [40] and Amorgos in 1956 [41]; (2) the event took place in a very densely inhabited area—the estimated population of the cities of Messina and Reggio at the time, was 140,000 at Messina and 45,000 at Reggio Calabria [42]; (3) the event is deemed to be one of the major events occurring in the Mediterranean since the beginning of the era of instrumental data collection in seismology and disaster science; (4) given the catastrophic intensity of the phenomena, it is theoretically plausible to suppose that social memories of tsunami related events have been partly preserved and passed down through the generations, becoming part of the local culture (see Section 2.4).

2. Addressing Tsunami Risk Perception

2.1. The Elusiveness of Tsunami Risk

Among the various natural hazards, tsunami risk perception appears particularly difficult to study due to the inherent elusiveness of these phenomena. Tsunamis can be generated by strong earthquakes, volcanic eruptions, landslides, or rapid and intense changes in atmospheric pressure that cause vertical displacement of the entire water column from the bottom to the surface [11,18,43,44,45,46,47,48,49]. Tsunami waves, being pushed by gravity, gain kinetic energy, allowing them to travel across enormous distances without losing energy. There are regions and areas where strong earthquakes, volcanic eruptions and seismic or volcanic induced landslides can trigger tsunamis more frequently. In these contexts, local communities are quite likely to have learnt and assimilated in their culture behaviors that may limit loss of life and damage to coastal infrastructures.

Compared to other natural hazards, tsunamis appears instead to be particularly elusive and unfamiliar phenomena, due to the relative long return times of these events. By contrast smaller events are relatively more frequent but far from being recognized as real hazards [3,50,51]. Furthermore, tsunamis may be triggered by physical events that may be neither felt nor perceived since they occur at a very far distance, making it difficult recognize signals and to establish cause-effect relationships between signals and impacts (far-field tsunamis) [10,16,52]. By contrast, tsunamis can be caused by sources very close to the coast (near-field tsunamis), which without adequate knowledge and preparation can make early warning and evacuation operations particularly difficult [17,53,54].

Moreover, tsunamis impacts are extremely variable, due to the non-linear combination of physical phenomena related to the characteristics of the source, bathymetry and the coastal profile that can affect both the propagation of waves in the open sea and impacts near the coast [55,56]. As an example, an earthquake or a volcanic eruption can compromise the stability of a landslide by triggering a tsunami wave, as occurred in Scilla, 1783 [33]. It is worth also mentioning that tsunamis are generally preceded by unusual and ambiguous premonitory signals (ground shaking, loud noise, sea level anomalies) thus making it particularly difficult for them to be identified and recognized by people unfamiliar with this type of event, and hampering proper response (e.g., self- evacuation and other mitigation measures). Finally, despite dramatic advancements in scientific tsunami hazard and risk assessment, this growing body of knowledge is not deemed to automatically turn into effective risk communication practices [16]. In particular, early warning procedures are affected by a large uncertainty mainly resulting from unknown details of sources and can be amplified by the urge to act rapidly to start evacuation procedures [28].

In this respect, the elusive character of the tsunami is a consequence of all these factors, representing an objective obstacle to understanding. This perhaps represents the key point in tsunami risk communication, requiring a more articulate and in-depth study of the elements that can facilitate or hinder its effectiveness.

2.2. Definitions and Factors Affecting Risk Perception

The concept of risk perception is about the way people judge and evaluate risks, and it is basically referred to the process of collecting, selecting and interpreting signals about uncertain impacts of events, activities or technologies. Such signals may come from different types of sources, such as direct observation of an event (firsthand experience) or more frequently, from information received or collected from other people in the community, from media systems as well from scientific and institutional sources (secondhand experience). According to one of the most cited definitions, ‘studies of risk perception examine the judgments people make when they are asked to characterize and evaluate hazardous activities and technologies’ [57]. Risk perception is a subjective construct that may involve feelings, beliefs, attitudes and judgments about impacts associated with an event [58]. This concept is strictly related to the assessment of certainty, severity, and immediacy of disaster impacts on individuals and communities, such as death, injuries and property destruction as well as disruption of work and normal routines [59]. Risk perception is based on a combination of both individuals’ psychological and socio-cultural factors shaping people understanding and responses to risk issues and it is socially constructed through many sources of knowledge such as media, collective memories of past events, education and experiences, and the ways these sources are framed and interpreted through culture and society [16,21,60]. Although there are no conclusive observations, risk perception is also influenced by socio-demographic variables such as gender, age, educational qualification, income, religion and political orientation, which may be investigated both as possible causes and as control variables [61,62,63,64].

Knowledge about risks, as well as messages about hazards and how they are assessed and managed, circulate within broader social groups and contexts. These groups construct shared meanings about the sources of risk, their dangerousness, and the actions deemed necessary to protect oneself, being able to affect individual courses of action as well as communities, culture, and political process [65,66,67,68]. Tsunamis enhance the role of knowledge and its limits, since the main sources of knowledge about this phenomenon are certainly not related to first-hand experience and involve different sources of information such as broadcast media (e.g., TV news, radio bulletins, movies, newspapers and, to some extent, the internet), narrowcast media (books, TV documentaries, civil protection campaigns, scientific communication), and personal networks, which may convey and re-produce information from direct witness as well as to revive tales of past events. Moreover, this information is often intertwined and implies the coexistence and overlapping of both vertical and horizontal communication flows, of differentiated sources (both official and unofficial), strategies, channels and communication models, being mediated through formal education, willingness to be involved and type of communication habits [69,70].

2.3. Socio-Demographic and Spatial Variables

Our research aims to survey tsunami knowledge and study tsunami risk perception by analyzing the opinions provided by respondents residing in coastal areas affected by the 1908 event. Although tsunami risk perception is influenced by several factors individually, socially and culturally context-related, different studies show that the most influential socio-demographic characteristics are age, educational degree, gender, and residence proximity to the coast (distance from the sea). Other variables affecting tsunami risk perception are household size and property ownership (the higher the value exposed in terms of lives and property, the higher the tsunami risk perception) participation in tsunami awareness programs and specific drills [59,71,72]. Data analysis provided the socio-demographic variables that effectively convey information about the focus of the analysis (see Section 4). In addition, identifying the sample’s socio-demographic characteristics allows more effective and target-oriented tsunami risk mitigation and communication strategies and decision-making policies to be implemented [60,73,74,75].

Variables such as gender, level of education, age, family composition and income are often analyzed in this type of research, although the results appear to be very much linked to the specific contexts in which they are carried out. Regarding gender, several studies show that it may significantly affect risk perception [76,77,78,79]. In some studies women are more likely to have a higher tsunami risk perception than men [73,74,75,76] while other research does not confirm such a gender influence [80,81,82,83].

Education was considered by many authors who studied risk perception related to natural hazards and, more specifically, tsunami risk perception. In most studies, a strong relationship emerges between tsunami risk perception and educational degree. People with a higher educational qualification are more likely to have better knowledge about natural hazards in their area and basic knowledge about tsunamis. Such a knowledge, in some situations, may reduce the response time in case of emergencies and fosters appropriate and effective behaviors [11,21,22,69,74,84,85,86,87,88].

Age is also a variable that recurs frequently in research but its variability in studies on tsunami risk perception depends on contextual and cultural factors [69]. In general, the relationship between age and tsunami risk perception is associated with individuals’ historical memory and experience gained through several catastrophic events experienced over a lifetime. In fact, several studies show that elders have a higher tsunami risk perception and are better informed than younger people. Conversely, young people have greater confidence in their ability to respond in the event of a tsunami due to their physical readiness [70,74,75]. However, the relationship between age and tsunami risk perception is not generally confirmed. A study conducted in Japan [89] shows that although older people have a higher tsunami risk perception, this would not foster willingness to evacuate in the event of a tsunami.

Household composition is considered in several studies on tsunami risk perception [70,80,83,90]. Generally, this variable has a significant impact when there are young children, elderly people, or people with disabilities in the household. In these cases, tsunami risk perception is higher than in low-member households where there are no fragility factors. Households composed of many members who resided for several generations in the area and have strong territorial roots tend to have a lower tsunami risk perception since they have greater confidence in the mutual help offered by other family members [90]. Household composition is a variable most frequently addressed in studies analyzing evacuation intentions in case of a tsunami. This specific aspect has been addressed in some studies [91,92] showing how individual solidarity may lead to delays in evacuation because family members expect to evacuate together.

The influence of spatial variables on risk perception [82], however, seems to emerge more clearly: risk perception is often higher among people living near the coast [72,73,76]. In fact, those who live near the coast and are aware of hazards impending on their area of residence (hurricanes, floods, tsunamis) have a higher tsunami risk perception, and in general, proximity positively affects individual risk perception, without affecting the intention to remain [74,76]. Proximity to the coast also positively affects some key factors closely related to tsunami risk perception including: knowledge of the phenomenon, behaviors one would adopt in case of emergency, and evacuation in case of tsunami warning [72]. Furthermore, research shows that tsunami risk perception is closely related to the temporal factors of residence in each area and to the memory of events that occurred in the past [83]. Others noticed a lower tsunami risk perception among those living in cities or large urban centers than among those living in the countryside or small towns [93]. In addition to distance from risk sources, research has also considered residence in potential flood-prone areas caused by both tsunami and weather events, concluding that the presence of risk sources close to residence positively affects risk perception, and as proximity from the risk source decreases, risk perception decreases [82].

2.4. Risk Perception and Knowledge Sources

In a social science perspective, the role of knowledge is often called to account as a crucial factor to describe and explain risk perception, and theoretical and empirical research has indeed carefully considered the ways in which different people rely on different types of knowledge to ground their judgments on risk. One of the most influential approaches highlights the rise of risk in late modern society as an unintended effect of a deep crisis of scientific and technological knowledge, deemed to be less and less capable of predicting and controlling both natural and man-made hazards [94,95,96,97,98]. Both laypeople’s and experts’ risks perception are seen as the outcome of mental and social processing of information and signals which is affected by individual characteristics and takes place within different social contexts, to give rise to culturally situated knowledge systems [64,99,100].

Knowledge is recognized as a key issue in risk studies. Risk and knowledge are closely related: moving from Luhmann’s classic distinction [89] between risk and danger, where risk is intended as the outcome of human decisions and danger to events due to environmental factors that cannot be directly controlled, risk and decision are intimately and inextricably linked to knowledge as a prerequisite of decisions; this raises the question the nature and legitimacy of knowledge and the way is is shared, accepted, or contested. As a consequence, knowledge, and scientific knowledge in particular, have progressively converted hazards (being unpredictable by definition) into risks that can be explained, measured, evaluated and—to some extent—predicted. Influential scholars including Beck [94], Giddens [97], Lupton [67] and Jasanoff [101] identified science as a linchpin of modern societies, thus paying attention to the ways in which forms of knowledge are epistemologically, morally, and politically legitimized. Knowledge made it possible to turn unpleasant “acts of gods” into calculable, predictable, and manageable events [102]. However, the claim of science to be the holder of the only authentic form of knowledge, naturally opposed to public’s perception of risk as a form of false understanding of reality, has often nurtured forms of arrogance of expertise [103]. As resumed by Schulz and Zinn “the assumption of the superiority of expert knowledge over lay-people’s emotional perceptions has been challenged by science and technology studies highlighting the reasonable rationales of lay people” [104].

The theoretical and research literature of environmental social science first, and the risk society approach just later, have strongly emphasized the idea of risk as a social problem related to contested/disputed knowledge and power asymmetries related to knowledge and its uses [105,106,107]. Risk issues are at the core of rising conflicts, in which not only scientific knowledge but also ultimate meanings of technical and social progress become objects of confrontation in the arenas of economics, politics, media, interest groups, and among intellectuals. The conflicting relationship between societies, individuals, knowledge, experts, and institutions thus becomes a crucial factor in today’s modern societies [108].

However, although this position (risk realism) is conventionally accepted in theoretical debate and widely applied in the context of environmental conflicts, it should also be remembered that it is based on the unproven assumption that different forms of knowledge are necessarily in conflict with each other. The core idea of scientific understanding of risk is that only science can provide objective knowledge about risks, while laypeople’s risk perception is rather based on subjective judgment. Hence, scientific evidence is a reflexive practice rooted in subject-object-separation and abstract thinking by means of rational objectification. Scientific, rational knowledge is held to be the best knowledge to inform decisions. However, along with objective facts, managed in the realm of science, there are also subjective facts which can play a relevant role in addressing risk knowledge through a different form of rationality, as a form of knowledge rooted in direct experience of sensible facts and certitudes [104]. Other scholars, however, considering risk as a social construct rather than an ontological reality (constructionism), have emphasized the obsolescence of the idea of knowledge monopolized by science and experts that can be unidirectionally transferred to an audience of lay people; they suggest the need to overcome this model, carefully putting scientific advice in contexts where it has to be used, and considering how risk perceptions, acceptance, or tolerance are always embedded in social contexts, being understood in the light of local knowledge and experience [109].

2.4.1. Scientific and Institutional Knowledge

The advancement of scientific knowledge about risks, as well as the centrality of science in the processes of risk assessment, management and communication represent a specific feature of late modern societies [94], embodying the belief that scientific rationality and technological development could provide a true understanding of nature and ensure control upon the natural world [98]. This type of knowledge is at the core of political decisions on risk, being often considered as the only legitimate voice to inform policy making, as it is based on evidence from scientific research [110,111,112,113]. Knowledge is made available to decision makers, stakeholders and the general public through a number of specific and clearly identifiable channels: journals, books, scientific communication, schools, institutions (e.g., Civil Protection, Local Government). Nevertheless, although scientific and technical risk analyses can provide the best knowledge about physical harm, being logically or empirically connected with each possible course of action, an integration of scientific and technical expertise and social science is necessary. There are other possible consequences of risk that people might also be considered as undesirable by individuals as members of a given society. Both causes and consequences of risks are mediated through social processes [114] and tend to suffer from the limitations of the language and channels being used as well as from the difficulty of “translating” complex concepts into concrete and viable ideas for final users. This is noteworthy within tsunami science: despite the undeniable progress in the field, science communication struggles to reach vulnerable communities, leaving several relevant issues to be addressed such as vulnerability, societal effects of a tsunami, the tsunami role in multi-hazard contexts and cascading risks [26].

2.4.2. Media Knowledge

In contemporary societies, mass media hold an unprecedented power “to encode, preserve, manipulate, reproduce and circulate symbolic representations of knowledge” [115]. Such a deep mediatization is recognized as a key issue for media and communications research, because of the broad and pervasive consequences of media on the everyday life of individuals, institutions and organizations operating within the social, political, cultural and economic contexts, including science. Such a process emphasizes the role of media platforms in spreading contents “through all types of context and practice” [116] as well their overall role in the construction of the social world, far beyond the possibilities disclosed by communicating “face to face” and “here and now” [117]. The process of mediatization has affected the very nature of science communication: scientists’ competitive efforts to gain media attention are triggering some relevant changes in the way the science system operates [118], blurring boundaries, specific norms and features of science communication into media logics, that might influence the way scientists communicate their findings to the public [119]. Undoubtedly media play an important role in mediating risks that are becoming increasingly invisible, intangible and unpredictable [94]. Mediatization also applies to risk knowledge, enhancing the media role in building representation of disasters in the global media-scape, and their role in addressing the way global risks are perceived and managed: “the linkage between risk perception and mediatized disaster representations is not incidental but intrinsic to each” [120]. The increasing dependence on risk knowledge emphasizes role and responsibilities of media in producing, disseminating and utilizing information and definitions about risks. Media also provide an important stage where information from scientists, experts, policymakers, businesses, and interest groups is mediated, emphasizing the important their role as “meaning entrepreneurs” in framing and anchoring processes [121,122]. Media, both broadcast and digital, represent a relevant source of information on both hazards and disasters, while also fulfilling other important symbolic mediating functions, e.g., fostering social exchange, providing emotional support during events, evoking past experiences of similar situations, and constructing causal explanations of current events [123]. The media, in particular, play a key role in defining what, how, and why a certain object poses a risk [124], but are also able to shape understanding of sources of risk [125,126] and influence how a certain risk is perceived [127], thus becoming a major issue in public debate and policy decision-making on risks [98]. However, it is noteworthy that media have often been accused of distorting, spectacularizing, and trivializing scientific knowledge, thus contributing to public misinformation [128,129,130]. These allegations find a partial justification as media lean on well-established routines and criteria to select and adapt some aspects of reality to their organizational purposes and constraints, the so-called media logic [131,132].

Within the tsunami risk management field, the role of the media is recognized both as a relevant source to get information in case of an event [53], and as a resource to improve risk awareness [16,22,133,134]. Media play an important yet contrasting role: on the one hand, they are recognized as indispensable actors in hazard mitigation programs due their capacity to spread information on tsunamis and their outcomes [135]; on the other hand, redundant media coverage may trigger feelings of anxiety and psychological unrest [136] as well as a false sense of subjective invulnerability, possibly resulting into evacuation mishaps [137]. In the Samos/Izmir event, occurred on 30 October 2020, despite most of the people running away from the sea to evacuate, some were spotted staring at the sea to watch such an unfamiliar event [9]. However, the visual framing of tsunamis as large waves and inevitable destruction can lead to a poor understanding of the phenomenon, its anticipatory signs, and a fatalistic attitude toward the risk: the media can reinforce the idea that because the tsunami is uncontrollable, its consequences would also be uncontrollable [11]. Finally, broadcast media have played a relevant role in the 2004 Sumatra and the 2011 Japan tsunamis and are expected to support relevant advancement in dissemination of information and educational activities on mitigation measures [134].

2.4.3. Traditional/Emic Knowledge

Traditional/emic knowledge is based on the cultural codification of the experience of past events that people have observed or participated in, being used in the construction of socially shared memories, conveyed and reproduced through myths, ceremonies and rituals.

These forms of knowledge and their importance in disaster mitigation processes are well known in the disaster literature, referred to by different definitions such as “local knowledge”, “traditional environmental knowledge”, “traditional ecological knowledge”, “indigenous technical knowledge”, “endogenous knowledge” [12,15,93,138,139,140,141] and more rarely, “local wisdom” [142]. This type of knowledge can be succinctly defined as “a body of knowledge existing within local populations or acquired by them over a period of time through the accumulation of experience, society-nature relations, community practices, and institutions, and through the transmission of these between generations [143]. Moreover, scholars have stressed the strategic importance of deepening, combining and integrating these forms of traditional knowledge with scientific knowledge [144,145,146]. The imperative need of integrating different forms of knowledge and reevaluating traditional knowledge has found a perfect case study in the specific field of tsunami risk reduction. After the tsunami of 26 December 2004, which originated from a large fault in the Indian Ocean near the coast of Sumatra, scholars’ attention focused on Simeulue, an island in Aceh district only a few tens of kilometers away from the event’s origin area, which despite being exposed to waves more than 30 m high, had a unexpectedly low number of deaths. The surprising explanation for the responsiveness on the part of local people lies in the cultural memory of a tsunami that occurred in 1907 and the measures needed to escape the wave, reproduced through the generations by means of a nursery rhyme (the song of the Smong, i.e., tsunami) taught first in schools, later incorporated into Simeulue’s musical folklore [13,17,147,148,149].

Over the decades, cultural anthropology has often dwelt on these forms of knowledge about disasters, especially through the collection of oral histories concerning causal explanations of events and descriptions of their effects. These oral histories are forms of native knowledge “precipitated” after an event and, regarding tsunamis, they are most common in areas where this type of event is most likely to occur, (e.g., near subduction zones), particularly along the Pacific Ring of Fire. Indeed, there is a great deal of research on orally passed down stories, their symbolic meanings, and the possibility of integrating this kind of knowledge into Disaster Risk Management or DRM strategies.

Although the boundaries between these three different sources are partially blurred and overlapping, the ways in which knowledge is made available must adapt to the communication patterns, channels and languages considered valid in the different domains (science, media and local communities). Scientists may adapt their language by trying to be as concise as possible to exploit the occasional media attention when an event occurs in order to reach a wide audience (e.g., the Hunga Tonga-Hunga Ha’apai eruption and subsequent tsunami). Similarly, scientific/institutional knowledge can be used for educational purposes and then be incorporated into local culture (as happened in Simeulue with the Smong song) and vice versa. The aspect that we are interested in emphasising is that, regardless of the domain source of origin, knowledge must be transformed, readapted and incorporated according to the inherent logic and communicative practices of the domain where it has to be embedded in. To briefly summarize analogies and differences, as discussed in the above paragraphs, see

Table 1. Sources of knowledge and their characteristics.

	Local/Emic Knowledge	Mediated Knowledge	Scientific Knowledge
Underlying model of communication	Network/community	Broadcast	Narrowcast
Source	Members of the community	Media organizations	Scientists, experts, science communicators, institutions
Channels	Word of mouth, myths. community rituals, ceremonies	TV news, movie, newspapers, radio, Internet	Books, TV documentaries, science and institutional channels
Receivers	Members of the community: unintentional exposure	General public: unintentional exposure	Targeted publics: voluntary exposure
Content type	Stories of past events, rescue and salvation	News and dramatization of ongoing/recent events	Accurate account of past/future events and its causes
Pros	Rooted on firsthand, sensible experience	Availability to large public, timeliness, redundancy	Accuracy, focused on causes, impacts and mitigation measures
Cons	Fading memories, vagueness	Spectacularization, biased and stereotyped representations	Excessive abstraction, specialized language, complex concepts

.

3. Materials and Methods

The overall goal of risk perception research is to improve risk analysis and decision-making processes, as well as to inform sound and consistent risk mitigation strategies. Especially, it is aimed at: (1) improving methods for eliciting, collecting and classifying opinions on risks; (2) providing a basis for understanding and anticipating possible public reactions to hazards; (3) improving the communication of risk information among lay people, technical experts and decision-makers [150].

This research has been promoted to improve seismologist, geophysics and risk managers’ ability to address lay people’s understanding of tsunami physics and its related risk, as well as possibly focusing on emerging critical issues. The questionnaire survey is intended to provide a robust base of quantitative data on the resident population in the different coastal areas, with a view to having reliable data on all Italian coastal municipalities and to make comparisons between different regions and areas characterized by a higher level of hazard and/or a history of tsunamis. The Computer Assisted Telephone Interview (hereinafter CATI) is deemed to be a reliable methodology to collect a large, standardized, retraceable and cost-effective amount of data [151]. Statistical analyses have been performed with SPSS Statistics, Version 26.0 (IBM Corp, Armonk, NY, USA).

The aims of this paper are expressed by the three following research questions:

• RQ1: Can the experience of a past tsunami in a given area still affect people’s risk perception, even if the event dates back over a century?
• RQ2: What kind of socio-demographic and geographical variables may affect risk perception level?
• RQ3: How can the different sources of knowledge used by people affect their perception of risk? And in which ways?

To obtain representative and reliable data, a stratified sample was built by using three stratification variables: respondents’ age, gender and coastal areas, as to ensure the best possible correspondence between subpopulations in the sample and within the reference universe. Such sampling strategy made it possible to collect 5842 interviews performed in 450 Italian coastal municipalities across 8 coastal regions. The results of Student’s T significance test on sample mean, confirmed the hypothesis that all the three considered sub-samples belong to the same population thus allowing the research to be based on a larger and therefore more robust sample in terms of statistical reliability. The whole set of municipalities (and in turn, single cases) has been split in two areas, including on the one hand the areas where the 1908 tsunami physical effects have been registered and documented (including municipalities for which there were no direct observations, but which fell within areas between municipalities where documented accounts of the effects of the phenomenon were available). On the other hand, we considered the municipalities in areas where no effects have been reported and documented, which we roughly assume to be not affected by the event and its effects. This subdivision was made on data from the Italian Tsunami Effects Database or ITED catalogue [152,153,154], which reports observations of event-related phenomena based on direct measurements of geophysical data or data from historical sources.

The sub-samples surveyed in coastal areas were respectively named as (a) 1908 area, including the coastal areas of Calabria and Sicily where tsunami effects were documented and reported in historical catalogs by coeval authors (N = 1624); and (b) non 1908 area, which includes the remaining coastal areas of Calabria, Sicily and also of Apulia, Basilicata, Campania, Lazio, Molise, Sardinia, (N = 4218), where no effects of the 1908 tsunami were reported or documented. Municipalities affected by the impacts of the 1908 area are plotted in red, while others included in the non 1908 area are plotted in blue (see Figure 1).

4. Results

4.1. Socio-Demographichic Variables

Once considered the composition of the two subsamples (

Table 2. Coastal areas subsample.

Coastal Areas	N	%
Not exposed, no observed impacts (non 1908 area)	4218	72.2
Exposed. with observed impacts (1908 area)	1624	27.8
Total	5842	100.0

), as a first step, we analyzed the distribution of socio-demographic variables to understand whether substantial differences exist and how they might affect the interpretation of the data.

The differences in the distribution of socio-demographic variables between the two sub-samples shows a general balance, with some minor differences: in the 1908 area there is a slight prevalence of women, people living within 1 km of the coastline and people who have lived in the same area since four or five generations.

In the non-1908 area, there are slightly more people living between 1 and 3 km from the coastline and those living in the same coastal area for only one generation (

Table 3. Socio-demographic differences in subsamples from 1908 area and non 1908 area.

		1908 Area		Non 1908 Area		Total
		N	%	N	%	N	%
Gender	Man	704	43.3	1954	46.3	2658	45.5
	Woman	920	56.7	2264	53.7	3184	54.5
Education level	Low/no qualif.	602	37.1	1627	38.6	2229	38.2
	Mid	714	44.0	2106	49.9	2820	48.3
	High	308	19.0	485	11.5	793	13.6
Age	18–34 years	399	24.6	1113	26.4	1512	25.9
	35–49 years	492	30.3	1430	33.9	1922	32.9
	50–64 years	513	31.6	1377	32.6	1890	32.4
	65 and over	220	13.5	298	7.1	518	8.9
Number of generations living in that area	≥5	194	11.9	411	9.7	605	10.4
	4	254	15.6	438	10.4	692	11.8
	3	395	24.3	1110	26.3	1505	25.8
	2	435	26.8	1479	35.1	1914	32.8
	1	346	21.3	780	18.5	1126	19.3
Total		1624	100.0	4218	100.0	5842	100.0

). In the 1908 area, the most pronounced differences (>=5%) concern the higher percentage of elders and people with high educational level: the percentage of people with a university degree or higher qualification is almost twice as high as in the other areas. Furthermore, there are more people residing in that coastal area since four generations. By contrast, in the non-1908 area we found higher percentages of people with mid-education level and people staying from two generations in that coastal area.

Table 3, for brevity’s sake, shows the distributions of the variables whose modes differ by at least 5%. None of the variables referring to household composition show substantial differences in the distribution between the two areas and they have therefore been omitted. To provide a concise and easily understandable figure that allows for an effective comparison between different categories and areas, we developed a synthetic indicator. The tsunami risk perception index (hereinafter RPI) was constructed by focusing 11 specific questions from the questionnaire, concerning three major conceptual dimension of risk perception, namely phenomena features, probability of occurrence and expected impacts of tsunamis (See questions 13, 16, 18, 19, 22.1, 22.2, 22.3, 22.4, 23.2, 23.4 and 23.5 of the questionnaire (see Supplementary Materials)). Principal Component Analysis (PCA) [155,156] is a common multivariate approach based on the analysis on correlated variables, which has been employed in this study to reduce the underlying dimensions associated with tsunami risk perception across a number of indicators, to maximize the amount of synthesized information, represented by the variance, as well as to identify a single synthetic variable (latent factor) to summarize and provide a robust and consistent measurement of relevant dimensions of risk perception, in order to make comparison between the considered areas and between different interviewees.

PCA analysis is one of the most used and robust methodologies for mathematically transforming quantitative data into synthetic indices: due its capacity to summarize data dispersed across several variables, it has been used in several researches on risk perception, e.g., [157,158,159]. Before creating the RPI index (

Table 4. Risk Perception Index (RPI).

Mean	0.00
Median	−0.11
Std. Dev	0.949
Variance	0.901
Min	−2
Max	2

), Cronbach’s Alpha coefficient was used to check index reliability, and it confirmed the internal consistency and thus cohesion among the items contributing to the summary index. The Alpha value is a quite satisfying 0.66, since alpha values between 0.61 and 0.83 are generally considered optimal for validating the reliability of the variables [160].

The creation of the RPI has been performed via SPSS synthesis commands (Analyze → Factor Analysis → Principal Components) and the statistical analysis of the output was returned by the software. More specifically, principal components of the 11 variables under analysis were extracted. The first factor—which synthesises as much information as possible and is associated to an eigenvalue >1 [161] was considered as a synthesis of the 11 original variables. The factorial scores derived from the first factor, which composed the one-dimensional RPI tsunami risk perception index. The PCA, as conceived in this way, has enabled the achievement of a twofold goal: (1) to identify the best linear transformation of the variables and interpret the result using the contributions of the starting variables to create the new synthetic variable; (2) to verify the actual possibility of creating a synthetic indicator from elementary data [162].

In statistical terms, the RPI indicator can be described by the following formula:

RPI = a × v13 + b × v16 + … + z × vn

where a, b … z represent the optimal weights estimated by the algorithm to extract factors from the original variables v1, v2, …, vn.

The 11 selected questions were arranged into a single risk perception index (hereinafter referred to as RPI), through PCA so as to provide a general indication of hazard perception by the interviewees and their ability to properly identify causes and physical effects. The most significant values are associated with the variables evoking damage, destruction and danger to people and their lives, while zero represents risk perception neutrality, which is the general mean for the whole group of 8 regions considered by the survey (

Table 5. RPI in the surveyed region.

	Mean	N	Std. Dev
Calabria	0.63	404	0.827
Apulia	0.49	617	0.874
Molise	0.41	100	0.985
Basilicata	0.21	140	0.933
Sicily	−0.04	1595	0.897
Campania	−0.13	1170	0.864
Latium	−0.23	1034	0.986
Sardinia	−0.23	782	0.936
Total	0.00	5842	0.949

).

The most important finding concerns the distribution of the RPI between the 1908 and non-1908 areas: in the areas affected by the Reggio Calabria-Messina tsunami, the average values are always higher than elsewhere. Even considering only the two regions hit by the 1908 tsunami, Sicily and Calabria (

Table 6. RPI in 1908 and non 1908 areas.

	Mean	N	Std. Dev
Non 1908 Area (for all regions covered by the survey) 1	−0.06	4218	0.953
1908 Area (Sicily + Calabria)	0.15	1624	0.923
Total (all regions covered by the survey, both 1908 and non 1908 areas)	0	5842	0.949
Non 1908 Area (Sicily only)	−0.18	340	0.881
1908 Area (Sicily only)	0.00	1255	0.898
Total (Whole Sicily, both 1908 and non 1908 areas)	−0.04	1595	0.897
Non 1908 Area (Calabria only)	0.41	45	0.988
1908 Area (Calabria only)	0.66	359	0.801
Total Calabria (Whole Calabria, both 1908 and non 1908 areas)	0.63	404	0.827

¹ Latium, Campania, Basilicata, Calabria, Apulia, Molise, Sicily, Sardinia.

), the difference between affected and unaffected areas is clear and sharp. Despite these regions having different RPI values (namely −0.04 for Sicily and 0.63 for Calabria); these values are considerably higher in the areas hit by the 1908 event. These data may be interpreted as self-evident testimony of the way event has been embedded into local knowledge. More than a century later the heritage of the 1908 tsunami is still alive and could be detected through research methodologies such as the CATI questionnaire survey.

The correlation analysis (

Table 7. Correlation between age, qualification, household distance and RPI in 1908 area.

		Educational Qualification	Age (By Classes)	Household Distance from the Coast	RPI
Educational qualification	Pearson correlation (R)	1	−0.076 **	−0.182 **	0.309 **
	Sig. (2-code)		0.002	0.000	0.000
	N	1604	1604	1604	1604
Age (by classes)	Pearson correlation (R)	−0.076 **	1	−0.106 **	0.116 **
	Sig. (2-code)	0.002		0.000	0.000
	N	1604	1604	1604	1604
Household distance from the coast 1	Pearson correlation (R)	−0.182 **	−0.106 **	1	−0.167 **
	Sig. (2-code)	0.000	0.000		0.000
	N	1604	1604	1604	1604
RPI	Pearson correlation (R)	0.309 **	0.116**	−0.167 **	1
	Sig. (2-code)	0.000	0.000	0.000
	N	1604	1604	1604	1604

¹ Interviewees who were not able to recall the distance from the coast have been not considered. ** Correlation is significant at the 0.01 level (2-tailed).

and

Table 8. Correlation between age, qualification, household distance and RPI in non 1908 area.

		Educational Qualification	Age (By Classes)	Household Distance from the Coast	RPI
Educational qualification	Pearson correlation (R)	1	−0.114 **	−0.125 **	0.221 **
	Sig. (2-code)		0.000	0.000	0.000
	N	4168	4168	4168	4168
Age (by classes)	Pearson correlation (R)	−0.114 **	1	−0.094 **	0.112 **
	Sig. (2-code)	0.000		0.000	0.000
	N	4168	4168	4168	4168
Household distance from the coast 1	Pearson correlation (R)	−0.125 **	−0.094 **	1	−0.230 **
	Sig. (2-code)	0.000	0.000		0.000
	N	4168	4168	4168	4168
RPI	Pearson correlation (R)	0.221 **	0.112 **	−0.230 **	1
	Sig. (2-code)	0.000	0.000	0.000
	N	4168	4168	4168	4168

** Correlation is significant at the 0.01 level (2-tailed). ¹ Interviewees who were not able to recall the distance from the coast have been not considered.

) shows that RPI has a significant positive correlation with education level and age class in the 1908 area. Indeed, RPI is negatively associated with distance from the coast: the closer the distance toa the sea, the higher are the observed RPI mean values (excluding people who were not able to recall the distance of their household from the sea). With respect to the non 1908 area, correlation coefficients between RPI and household distance from the sea are higher, while its value is slightly lower for RPI and educational qualification. In brief, socio-demographic variables are tied to risk perception, although proximity plays a relatively strong role, especially for those who live outside the 1908 area.

RPI is also affected by other socio-demographic variables such as education level: for those who have higher qualifications, its mean value is 0.57, while those with low education have −0.24. Having verified that the two socio-demographic variables that exert the greatest influence on risk perception are level of education and proximity, it is worthwhile to go into detail by analyzing the means in the two areas considered.

Data once again underscore—ceteris paribus—a difference in risk perception between the area exposed to the 1908 event and the area not exposed. It seems clear that proximity is a very important variable in the perception of tsunami risk: the closer the household is to the sea, the higher the average value of the indicator, indicating that risk is most perceived where conditions of objective vulnerability do exist (

Table 9. RPI and proximity in 1908 and non 1908 area.

	1908 Area			Non 1908 Area
How Many km Away from the Coast Do You Live?	Mean	N	Std. Dev	Mean	N	Std. Dev
Within 1 km	0.41	431	0.881	0.34	876	0.885
From 1 to 3 km	0.11	497	0.849	−0.04	1172	0.896
Over 3 km	0.02	676	0.962	−0.24	2120	0.963
I don’t know	−0.43	20	0.934	0.05	50	0.801
Total	0.15	1624	0.923	−0.06	4218	0.953

). However, even for this indicator there is a clear and revealing difference between those who live in areas exposed to the 1908 event and those who do not. This not only affects those who live close to the sea (<1 km) but also those who, at least in theory, should feel safer, living at greater distance.

As far as the educational level is concerned (

Table 10. RPI and educational qualification in 1908 and non 1908 area.

	1908 Area			Non 1908 Area
Educational Qualification	Mean	N	Std. Dev	Mean	N	Std. Dev
High	0.57	308	0.826	0.43	485	0.865
Mid	0.26	714	0.871	−0.01	2106	0.958
Low/no qualification	−0.20	602	0.908	−0.27	1627	0.911
Total	0.15	1624	0.923	−0.06	4218	0.953

), the effect on risk perception appears stronger in the areas affected by the 1908 tsunami, being associated with higher levels of risk perception in both areas. The low/no qualification has lower values with relevant differences for both those who have higher and mid qualifications and live in the 1908 area. This can only be attributed to people’s awareness of being exposed to the risk of a tsunami, which is visibly stronger in the areas affected by the 1908 event.

4.2. Source of Knowledge, Combinations and Effects on Risk Perception

Alongside socio-demographic variables, it is important to consider in detail the role played by different sources of knowledge, and how they are currently used (and combined) by people (

Table 11. Sources used to gather information on tsunamis. (Multiple response question: % calculated on cases).

	1908 Area		Non 1908 Area
	N	V%	N	V%
TV News	1328	84.3%	3279	81.2%
Newspapers	532	33.8%	1385	34.3%
Books	333	21.1%	742	18.4%
TV documentaries	322	20.4%	938	23.2%
Internet	285	18.1%	762	18.9%
Movies	171	10.9%	498	12.3%
Friends, neighbors, relatives	159	10.1%	162	4.0%
Radio	147	9.3%	295	7.3%
Civil protection	47	3.0%	59	1.5%
Universities and research institutions	36	2.3%	52	1.3%
Local Government	31	2.0%	25	0.6%
Total	3391	215.2%	8197	202.9%

). This kind of analysis is crucial not only to understand whether and how people collect information about this particular risk, but also to understand how these sources are brought together and, above all, the implications for risk perception. TV news is by far the most relevant source of information for both residents in 1908 and non 1908 areas. This result is consistent with other sources, including the Italian National Statistical Institute: television remains the most widely used media in Italy. In 2021, 90.1% of the population aged 6 and over said they watched TV and 72.5% did so on a daily basis [163]. The second most used source are newspapers, followed by books, documentaries, Internet and movies. It is noteworthy that preferences are roughly distributed in the same way, with little differences within the two considered areas. Some noticeable differences are found for interpersonal and institutional sources. Despite friends, neighbors and relatives seeming to play a minor role, they are listed by 10.1% of the 1908 area, against only 4.0% of the non 1908 area. In addition, people who specified sources in an open-ended question were likely to mention as a source the tales from grandparents or grand grandparents, who were very likely to refer to firsthand experience of the 1908 event, thus confirming our initial hypothesis.

On the other hand, scientific institutions, civil protection and universities, along with local authorities, play a very limited role, since just one of these sources barely reaches 3% in the 1908 area (namely, Civil Protection), while research institutions and universities were mentioned by 2.3% and local authorities by 2%. Results are even more puzzling in non-1908 area, where these sources were mentioned by 1.5%, 1.3 and a 0.6% of respondents, respectively. The limited effectiveness of scientific and institutional sources is a very critical point to be carefully considered in future research as well as in approaching future risk communication strategies, as confirmed by two previous research papers referring to the Italian coasts [21,22].

The way interviewees match together different sources of information provides further relevant insights (

Table 12. RPI and sources combinations ¹ to gather information on tsunamis.

	1908 Area			Non 1908 Area
	Mean	N	Std. Dev	Mean	N	Std. Dev
Media, scientific, interpersonal sources	0.62	50	0.69	0.55	52	0.795
Scientific/institutional only	0.43	88	0.798	0.1	250	0.746
Other combinations (N < 50)	0.39	48	0.853	0.21	44	0.976
Media and scientific sources	0.31	444	0.887	0.03	1166	0.99
Media and interpersonal sources	0.28	63	0.825	0.2	66	0.867
Media only	0.02	872	0.943	−0.13	2446	0.959
Unspecified	−0.34	59	0.844	−0.26	194	0.779
Total	0.15	1624	0.923	−0.06	4218	0.953

¹Media only: combinations containing only television, newspapers, internet and movies; scientific sources: combinations consisting in TV documentaries, scientific and institutional sources; interpersonal sources: relatives, friends or neighbors.

). We considered combinations of different sources by grouping them according to homogeneous criteria with respect to the underlying communication model. The aforementioned sources were grouped as follows: media sources = broadcast media: TV News, radio, newspapers, movies; scientific sources = specialized sources: narrowcast media, directly targeting messages at interested, segmented audiences: books, TV documentaries, civil protection, scientific institutions, local government, and finally interpersonal sources = word of mouth through personal networks: neighbors, friends and relatives. If we consider that people rely on different combinations of sources, some crystal-clear differences emerge in risk perception. In the 1908 area, the higher levels of risk perception have been found between those people who mix media, scientific and institutional sources, while those who were not able to specify their sources of information show the lowest level of risk perception.

Two important issues may be pointed out. On the one hand, despite this combination of sources being found in a relatively low number of cases, it is associated with the higher levels of tsunami risk perception in both 1908 and non 1908 areas, respectively 0.62 and 0.55. Secondly, the values of this index are higher than those which have been found in association with any other socio-demographic and spatial variable.

Furthermore, in the 1908 area, other combinations of sources including scientific sources result into higher values of risk perception, while in the non-1908 area this effect appears to be much weaker. In the non-1908 area some combinations of sources are used by a very small number of interviewees. Nevertheless, people who only rely on media sources as well as on unspecified sources have lower levels of risk perception. The data appear to be consistent with the theoretical premise [164] by which is necessary to work on the integration of local and scientific culture to improve the awareness and preparedness of populations exposed to a particular risk.

As to providing further insights, it is worth considering socio-demographic characteristics of those who exclusively rely on TV News and of those who combined media, scientific and interpersonal sources. On the one hand, the socio-demographic composition of those who use the media alone as a source of information appears substantially different from that of the general sample. In particular, the differences concern a slightly higher percentage of women (57.1% vs. 55.1%) but especially a strong over-representation of people with low educational qualifications (46.9% vs. 38.1%), which is accompanied by a slightly lower percentage of people with medium educational qualifications (41.7% vs. 44.5%) and a more pronounced difference in the percentage of people with high qualifications (17.1% vs. 11%), while the breakdown by age group appears very similar, with differences of less than 1%.

Conversely, expecting a small number of cases eligible in this category, among those who combine scientific, interpersonal and media sources, women are somewhat more numerous (58.1% vs. 55.1%). Percentages of all age groups excluding the over-65s are slightly lower than in the sample, while the percentage of the over-65’s is 25.5% vs. 11.6%, and, most importantly, the percentage of people with high qualifications is very overrepresented: 55.1% vs. 11.0%.

5. Conclusions

One of the most relevant results is the clear difference in risk perception between the area affected by the 1908 Reggio Calabria-Messina event and the surrounding area, which was not affected by the tsunami. Such a difference was also found in Calabria and Sicily, even though the average level of tsunami risk perception changes significantly between the two regions: for both, the RPI values are higher in the areas affected by the 28 December 1908 event than in the locations where—according to historical sources and catalogues—tsunami local effects due to the tsunami have not been documented. This evidence appears as a self-evident testimony that after more than a century something still remains in people’s memories about that event and these memories still prove capable of affecting the way tsunami risk is perceived and understood.

However, tsunami risk perception is also affected by some socio-demographic, spatial and socio-cultural variables, proven to be relevant to address tsunami risk perception.

• Education plays a relevant role, along with proximity and the long stay of the household in the same coastal area, being interpreted as measures of both awareness and familiarity with tsunami risk. Nonetheless, these factors alone could not provide a comprehensive explanation difference in tsunami risk perceptions in the 1908 area. Despite a time span of more than a century, local knowledge about the phenomena still remains, playing a certain role in addressing risk perception, emerging across all data and possible interpretations.
• Analysis highlights a troublesome centrality of the media as a source of information. TV news represents the most widely used channel: in fact, more than 4/5 of the sample uses it as a source about tsunamis, although this process implies substantial passivity on the part of the recipients, who do not actively engage in information seeking, but merely pay (limited) attention to news coverage of the events. Other channels require a more active participation on the part of the user, who actively and voluntarily seeks out the information required, for example through books and documentaries. People who use interpersonal sources, often explicitly linked to oral forms of storytelling about past tsunamis, are significantly more numerous in the area affected by the 1908 event, also testifying about relevant oral memories, as open-ended questions often bring recall of grandparents’ stories.
• Sources of knowledge are differently handled: this does not depend on sample characteristics but on the social sedimentation of the memory of past events: apparently, the memory of the 1908 event also fostered further examination through media (TV News, newspapers, radio) and scientific and institutional sources (universities and research Institutions, Civil Protection); this is, however, associated with a very limited penetration. In fact, the percentages of those who can integrate and combine different sources are higher in the area affected by the 1908 event, suggesting some social pressure to become more informed and knowledgeable about tsunami risk, especially among those with higher educational qualifications. It is worth also mentioning that at the moment of the survey the areas were served by the same Early Warning System operating in the whole Italian coastline, thus excluding a possible influence on RPI. However, the area hosted a number of initiatives to commemorate the event or to raise awareness of seismic and tsunami risk issues, which were also considered in the questionnaire under the title of “scientific and institutional communication”.

Research data have important implications for the development of more effective risk communication strategies. First and foremost is the need to overcome linear, hierarchical, and unidirectional views of risk communication by which messages are conveyed top down, by experts or authorities to lay people (e.g., decide-announce-defend or deficit models). As unequivocally shown by our data, institutional and scientific communication are underperforming, confirming criticalities already noticed in previous research [21,22]. That is, there appears to be a clear need to simplify and make accessible scientific knowledge, along with the need of scientists and authorities to learn how to strategically improve dialogue and interaction with the media. Second, there emerges the need for a novel approach to risk communication, that can integrate different forms of knowledge (enhancing experiential knowledge), recovering where possible the orally transmitted legacy, and collecting documents and direct testimonies from people who had first-hand experience of the events.

The research presented here raises several issues that need to be adequately explored in the future. First, studies on risk perception in conjunction with PTHA maps (Probabilistic Tsunami Hazard Assessment) [50,165,166] would allow us to identify critical issues and priority areas to build risk communication strategies. These would be implemented within a stricter cooperation of geo-scientists and social scientists, so as to identify areas where low risk is accompanied by high hazard, to prioritise drills and educational programs to improve community awareness and to implement more effective risk mitigation measures.

Furthermore, data on risk perception and knowledge should more effectively address tsunami risk communication and early warning systems’ operational design, duly considering expectations, needs and actual levels of understanding and perception of risk by vulnerable populations. The need to overcome deterministic, hierarchical, and unidirectional views of risk communication suggests the adoption of a broader and more realistic definition of the communication process and more open and inclusive practices in approaching problems [16]. It seems to us necessary to integrate and to reconcile the “vertical” dimension of risk communication (stakeholders, mayors, experts, local associations, etc.) and “horizontal” ones (between peers, neighbours, local associations, members of local communities) within risk management and communication processes, so as to better support knowledge exchange and understanding of risks.

This paper provides some descriptive data, the limitations of which stem first and foremost from the structure of the questionnaire and from the way it is administered to respondents (CATI Interview). The need for robust and complete data on the one hand, and the competing need to collect valid and reliable data while limiting refusals or partial responses as much as possible, made it necessary to limit the duration of the telephone interview, and consequently the number of possible questions. Moreover, although the responses to some open-ended questions provided interesting insights, the CATI interview is not the most suitable tool to delve into them in more depth. In this regard, the authors strongly recommend the adoption of qualitative and field-based approaches aimed at reconstructing narratives in a broader, more precise, and in-depth manner through the actors’ own thoughts, collected through in-depth interviews, life stories and ethnographic analysis of documents (diaries, written documents, etc.).

Finally, the presented research constitutes a first, concise approach to the use of multivariate analysis techniques in the Mediterranean context, namely, to build an RPI index for the specific scope of this paper. The authors and the whole research team agree on the need to further explore the use of this type of techniques, with the twofold purpose of exploring in detail the analytically relevant dimensions of tsunami risk perception and to allow for more effective syntheses, aggregations and comparisons of data within specific groups or spatial areas.