1 Introduction

Earthquake environmental effects (EEE) are all the effects, from geological to hydrological, physical and meteorological, that a seismic event can induce on the natural environment (Michetti et al. 2007). Among them, the coseismic geological effects are the most hazardous. They range from surface faulting, which can reach displacements of many meters and extend for hundreds of kilometers, to landslides, rock falls, liquefaction, ground collapse and many other consequences, including tsunamis.

Earthquake environmental effects are common features produced by moderate to large crustal earthquakes, in both their near and far fields. Always recorded and surveyed in recent events, very often they are remembered in historical accounts and conserved in the stratigraphic record as paleo-earthquake markers, the latter being the basis of paleoseismology (e.g., McCalpin 2009). Both surface deformation and faulting and shaking-related geological effects (e.g., liquefaction, landslides) not only leave permanent imprints in the environment, but can also severely impact man-made structures (e.g., Hancox et al. 2002; Honegger et al. 2004; EERI 2008, 2011). Moreover, underwater fault ruptures and seismically triggered landslides can generate devastating tsunami waves (cf. Ward 2001; Harbitz et al. 2006; ten Brink et al. 2009; Ozawa et al. 2011; Satake et al. 2013; and bibliography therein).

These phenomena represent significant sources of hazard, especially (but not exclusively) during large earthquakes, substantially contributing to the scenarios of destruction. Severe damage to buildings and infrastructure from surface faulting, landslides and liquefaction is commonly experienced during moderate to strong seismic events (e.g., Dowrick et al. 2008; Michetti et al. 2009; Eberhard et al. 2010; Hayes et al. 2010; Lekkas 2010; EERI 2011; Fritz et al. 2011; Mori et al. 2011; Vittori et al. 2011; Di Manna et al. 2013; Harp et al. 2013; Mavroulis et al. 2013; Pavlides et al. 2013; Silva et al. 2013; Valkaniotis et al. 2014). Our aim was to prove that seismic hazard assessment (SHA) would benefit from a comprehensive consideration of all earthquake-related effects, including environmental ones. Macroseismic intensity and its attenuation with distance are still considered key parameters for SHA and are used to generate shake maps for early warning and rapid response planning (e.g., Sørensen et al. 2009). According to the latter authors, ground motion attenuation in terms of macroseismic intensity allows to overcome some drawbacks of the commonly applied ground motion prediction equations (GMPEs), which are the limited number of recordings and the complex and not straightforward association of ground motion with damage.

A tool devised in the recent past (Michetti et al. 2007) to improve the intensity characterization is the Environmental Seismic Intensity (ESI) scale. It is a 12 degrees intensity scale (Table 1) solely based on EEEs, whose documentation has seen a considerable growth in the past decades. Several authors have made use of geological effects in their application of the Modified Mercalli (MM) intensity scale in the past, also influencing the advent of the ESI scale (e.g., Dengler and McPherson 1993; Hancox et al. 2002). Many scientists all around the world have already applied the ESI scale in their account of recent and historical earthquake scenarios (see “Appendix”). The key contribution to SHA of the ESI scale is the improved intensity assessment, seen as the comprehensive parameter necessary to maintain the consistency between source parameters assessed for historical earthquakes and for the modern ones. The case of the 2012 Modena earthquake sequence clearly illustrated this point (e.g., Galli et al. 2012; Di Manna et al. 2013; Graziani et al. 2015). Such consistency is pivotal for reliable magnitude/intensity relationships and the value of the seismic catalogue itself.

Table 1 The ESI scale
Full size table

Therefore, based on several years of worldwide application in the field, in this paper (a) we introduce the ESI scale to the community of earth scientists (geologists, geophysicists and seismologists) and civil engineers, as a survey instrument to better characterize a seismic event, also in terms of local effects and attenuation with distance, and (b) we provide insurers, civil protection agencies and administrators with an integrated tool to assess the potential damage deriving from geological effects during a future earthquake in an area, to be added to that directly associated to seismic shaking.

In the following, first a short summary of the evolution of earthquake intensity scales is given, also considering the impact of the advent of magnitude, followed by the background and rationale of the ESI scale. Then, three representative case studies target what are deemed to be the most crucial issues related to the application of the ESI scale, including: (a) the comparability of epicentral intensity assessed based on traditional macroseismic scales and on environmental effects, (b) the capability to measure paleo-earthquakes by means of the empirical relationships of rupture parameters (rupture length and displacement) with intensity and magnitude and (c) the assessment of epicentral intensity from the total area of secondary environmental effects and from the total rupture length (Michetti et al. 2007).

2 The Evolution of Intensity Scales and the Role of Environmental Effects

Over the past decades, a number of publications have dealt with the history of the intensity scales and the analysis of their relationships, from Richter (1958) and Shebalin (1975) to Musson et al. (2010). In the following, the focus is placed on how intensity was originally conceived and for what purposes. Before any standardized intensity scale had been conceived, a set of symbols were occasionally utilized for depicting different levels of damage of an earthquake, as in the map drawn by Matteo Greuter (Fig. 1) for the July 30, 1627, Gargano earthquake (in Zecchi 2004). Something similar might already have appeared in Italy in the late 16th century, drawn by Gastaldi for the July 20, 1564, Nizza event, as cited in Gioffredo (1692), but this map, if it ever existed, is now lost. While this made it possible to compare damage caused by a given earthquake at different sites, it did not allow comparison among different earthquakes.

Fig. 1
figure 1

Matteo Greuter, Roma, 1627: distribution of damage caused by the July 30, 1627, Gargano earthquake in southern Italy; the stippled box encloses the map legend: symbols classify damage as completely destroyed, in large part destroyed, about half ruined, only partially ruined or damaged

Full size image

A number of strong events since the seventeenth century in many regions of Europe gradually increased the awareness that measuring and comparing earthquake effects was necessary, requiring a detailed documentation through a proper standardized procedure. A logical approach appeared to be the use of all the observable effects and consequences of an earthquake to represent its size. By the end of the nineteenth and in the early twentieth century, several intensity scales were conceived that divided the whole range of seismic effects into characteristic sets commonly defined as degrees, now generally 10 to 12 in number. Their basic purpose was the same: to assess the size (strength) of earthquakes based on their effects on three different classes of targets: (1) domestic objects and human perception, (2) man-made engineered structures and infrastructures and (3) geological/natural environment.

Intensity scales had a threefold purpose: (1) to document the whole set of phenomena due to the earthquake; (2) to define a level of structural/environmental damage (I i ) in all the localities where it was felt, based on direct observation or deducted from historical accounts; (3) to compare different earthquakes, based on the size of the observed effects and extent areas of similar levels of structural/environmental damage. The latter implies the possibility to classify earthquakes according to their epicentral intensity (I 0), i.e., the intensity within the epicentral area where the strongest effects usually take place. For example, when Mushketov (1890), after a reconnaissance mission in the epicentral area of the May 28, 1887, Verny earthquake, wrote, “the earthquake strength was at least X degrees on the Rossi-Forel scale,” he meant the strength of the Verny earthquake as a natural phenomenon (see Fig. 2, where the area of maximum environmental effects is outlined). In fact, the environmental effects observed in the mountains were described in an extensive chapter, supported by drawings and photographs and used in earthquake strength assessment, on a par with damage incurred in towns and villages. Thus, in the earlier applications of the intensity scales the environmental effects were taken into account not differently from the effects on humans and manmade structures. See for example their wide use in the Mercalli–Cancani–Sieberg scale (MCS) or in the several versions and variants of the Modified Mercalli scale (e.g., Dowrick et al. 2008; Comerci 2013).

Fig. 2
figure 2

Map of the effects of the Verny, 1887, earthquake in the Tien Shan range of central Asia (from Mushketov 1890). The small ellipse encloses the area of maximum environmental effects, while the large ellipse indicates the area of considerable damage. Both contours appear on Mushketov’s original map. Note that most of the environmental effects occurred in a mountainous zone located 15–20 km south of the nearest populated area (Verny, now Almaty, Kazakhstan)

Full size image

The advent of the earthquake magnitude scale (Richter 1935) greatly influenced the subsequent development of the intensity concept. Magnitude was deemed a more objective parameter, because it relates the instrumentally recorded ground displacement to its distance from the epicenter. Since then, a number of studies have appeared relating intensity to magnitude, with the goal to estimate magnitudes for historical, pre-instrumental, earthquakes. Information on their effects was collected going back as far as possible in history, sometimes over many centuries, as in Italy, Greece, Japan and China. The set of empirical relationships among intensity, source depth and magnitude, known as macroseismic field equations, takes a variety of forms. Some equations link I 0 directly to magnitude; others use generalized information on macroseismic effects in the form of isoseismals and solve the equations to obtain magnitude and source depth from ratios of isoseismal radii (I i ), and yet others are based on felt-areas (e.g., Shebalin 1972; Ambraseys 1985).

What is stressed here is that more weight was placed on intensity distribution rather than on I 0. The intensity distribution, however, depends on the availability of observations and is, therefore, biased towards populated areas. More stable magnitude assessments can be derived from isoseismals than from a single I 0. When observations are obtained in a large number of localities, a poor intensity assessment in one of them has little or no impact on the average isoseismal radius, while an erroneous I 0 assessment significantly affects the magnitude estimate. Another reason is that for many earthquakes the epicentral intensity could not be accurately defined, mainly because the epicenter was in a remote place.

Another achievement of instrumental seismology that has dramatically affected the intensity concept is the measurement of strong ground motion by accelerometers. Ground motion is mainly responsible for the observed intensity at a site (away from the surface faulting zone) and significantly contributes to its understanding (e.g., Panza et al. 1997; Wald et al. 1999). However, such recordings cannot compete with intensity observations in terms of density of information and only rarely are available at the epicenter. In Japan, intensity (JMA scale) is now estimated based on the very dense array of strong-motion instruments and seismic intensity-meters (accelerometers) deployed there (e.g., Nishimae 2004). In the US, the ShakeMap program of the US Geological Survey produces near real-time shaking intensity maps from ground motion recordings (http://earthquake.usgs.gov/earthquakes/shakemap/).

With time, the focus on the behavior of engineering structures under seismic load, based on the estimation of local damage levels, has grown steadily in importance, radically restricting the original meaning of the intensity as an earthquake sizing parameter. The European Macroseismic scale (EMS98: Grünthal 1998) openly discourages the use of EEEs with the comment: “while variations in the vulnerability of manmade structures can be presented in a reasonably coherent yet robust manner, in the case of the effects on nature, most of these depend on complex, geomorphological and hydrological features which cannot easily be assessed by the observer, or at all”. In our view, the significant improvements made in recent decades on the knowledge of the environmental effects instead allow them to be fully exploited to categorize earthquake size, i.e., assess epicentral intensity, which is the rationale of the Environmental Seismic Intensity scale described hereinafter.

3 Earthquake Environmental Effects and ESI Scale

The ESI scale has been conceived not only to supplement the existing macroseismic intensity scales, but also to work alone when the other scales cannot be applied (e.g., vast uninhabited areas). The reference macroseismic scales for calibration have been the MM and Medvedev–Sponheur–Karnik (MSK) scales, but also the MCS scale, after adjusted to align with the other (see for example the comparison table of intensities in Reiter 1990). The choice of these scales was motivated by the wealth of historical information (also on environmental effects) available from their application for over a century to earthquakes in many regions worldwide and to historical events going back in time for at least a millennium. This, and the database of surface rupture parameters (e.g., Wells and Coppersmith 1994; and analysis in Stirling et al. 2013) of events for which intensity data could be retrieved, allowed the tuning of the EEEs in the different degrees, in terms of manifestation, dimensions and density.

The extensive list of earthquakes in Table 2 helps clarify the application range of the ESI scale and its added value to macroseismic scales. In fact, many earthquakes of moderate to strong magnitude are assigned in the literature a small, unrealistic I 0, despite extensive surface faulting and large offsets, e.g., the I 0 of IX for the 2004, Sumatra, M 9.1 event. Table 2 emphasizes also how the epicentral intensity and location of historical earthquakes can be poorly constrained based only on damage, when epicenters are located in sparsely populated areas with few buildings, or when the structural damage saturates, as it is usually seen, at intensity X of the historically used XII degrees scales cited before (MCS, MM, MSK). In some cases, the rupture parameters are not measured directly, but inferred from geophysical or instrumental seismological data, as for several events in the database of Wells and Coppersmith (1994). Some occurrences in the data base are likely erroneous, e.g., the 1976 Gazli earthquake: its epicentral area was in the desert, and the expedition of the Institute of Physics of the Earth of Moscow observed only secondary effects (e.g., Hartzell 1980).

Table 2 Some earthquakes with epicentral intensities differing by 2 or more degrees between MM and ESI scales (SRL surface rupture length, MD maximum displacement)
Full size table

The intensity values in Table 2 are undeniably correct, because they refer to the strongest macroseismic effects observed. However, they are far from representing that comprehensive measure of the event necessary for comparing seismicity in space and time, especially prior to the instrumental era, which is the prime scope of epicentral intensity. This current tendency in intensity estimation affects the development of reliable intensity/magnitude relationships, which serves to make the seismic events comparable in those catalogues that extend back in time well before the instrumental period even including paleoseismic events. This is the case, for example, for the Mediterranean region, China, Japan, and the USA. In this perspective, seismic intensity remains a fundamental gauging parameter, despite the availability of more objective parameters, such as magnitude. The intensity assessment of many historical earthquakes is being updated as new archive documents are discovered or sources reinterpreted. Sometimes, the newly inferred magnitudes have a strong impact on SHA. However, to maintain or even improve its essential role, intensity should take advantage of all information pertaining to earthquake effects, on both built and natural environments, also considering that some geological effects (especially faulting, liquefaction, landslides) conserve a much longer “memory” than those on the man-made structures. This provides a unique tool for comparing and mutually calibrating prehistoric (paleoseismic), historical and modern earthquake datasets. Like macroseismic estimates, the application of the ESI scale must also take into account local effects of soil amplification and preexisting highly unstable slopes that may lead to locally overestimate the intensity. EEEs can be found only where the geological and morphological conditions exist for their occurrence, with an uneven coverage of territory. These drawbacks, however, do not lessen the value of the overall picture, similarly to what happens in all macroseismic field reconstructions.

EEEs fall into two main categories: (1) primary effects, which are the surface expression of the seismogenic source (e.g., surface faulting), normally observed for crustal earthquakes above a given magnitude threshold, generally close to 6.0 according to available sets of data (e.g., Wells and Coppersmith 1994; Yeats et al. 1997); (2) secondary effects, mostly dependent on the characteristics of seismic wave and morphologic, geologic, climatic and soil conditions (e.g., landslides, liquefaction effects, ground cracks, etc.).

A good example of the contribution of EEEs to hazard assessment and earthquake characterization is provided by the empirical relationships of (1) magnitude and faulting parameters, in particular maximum/average displacement, surface rupture length and rupture area (e.g., Wells and Coppersmith 1994; Pavlides and Caputo 2004; Stirling et al. 2013), (2) magnitude and landslide distance from epicenter (e.g., Keefer 1984; Rodríguez et al. 1999; Hancox et al. 2002; Porfido et al. 2002; Delgado et al. 2013; Comerci et al. 2013; Esposito et al. 2013) and (3) magnitude and liquefaction distance from epicenter (Galli 2000; Castilla and Audemard 2007; and references therein). The aforementioned relationships, although largely empirical and characterized by considerable scattering, are widely applied in both deterministic and probabilistic SHA studies, especially (1). The same empirical analysis can be carried out based on intensity (see for example the graphs in Michetti et al. 2004), which is the basic approach of the ESI scale, where several other environmental effects are also taken into account statistically.

The importance of EEEs as a tool to measure earthquake intensity was already outlined in the early 1990s, for example by Dengler and McPherson (1993) and Serva (1994). An MMI-derived scale has been developed in New Zealand (Hancox et al. 2002) that incorporates landsliding, liquefaction and other ground damage criteria, but excluding primary effects, regarded as a seismogenic effect unrelated to shaking and, therefore, neither to intensity.

The concept of an intensity scale exclusively based on environmental effects took form in the 1999 15th INQUA Congress, Durban. A first version of the scale was presented in 2003 (16th INQUA Congress, Reno, Michetti et al. 2004) and tested over a 4-year trial period using actual case studies worldwide. The ESI 2007 intensity scale, presented in the framework of the 2007 17th INQUA Congress in Cairns, Australia, has taken into account the results of these applications.

In the following, the basic structure of the ESI 2007 is discussed. The full text is given in Table 1 (Michetti et al. 2007) and is available on-line in several languages (http://www.isprambiente.gov.it/it/pubblicazioni/periodici-tecnici/memorie-descrittive-della-carta-geologica-ditalia/memdes_97.pdf).

In order to make the ESI 2007 scale more compatible with previous macroseismic scales, it maintains two different levels of EEE spatial generalization: site and locality.

Site corresponds to the place where a single EEE of a certain type was observed, such as a landslide or a ground crack. This is the level at which EEE descriptions have to be compiled. As these effects are strongly dependent not only on the severity of shaking, but also on many other physical factors, it is only possible to assign an interval of probable intensity values to the effect observed at the site.

Locality is the place including one or more sites where the EEE occurred and presents a level of generalization to which intensity can be assigned. It is assumed that within the spatial frame of a locality, when EEE occurred in several sites, peculiar effects associated with the specific characteristics of each site are removed. Locality can refer to any place, whether populated or not. Since a natural “locality” can be anything from a river valley to a mountain slope or a large hill, it is difficult to assign a typical dimension to it, although it must be small enough to avoid comprising separate areas with significantly different site intensities, but large enough to include more than one site and consequently to be representative for an intensity assessment. Therefore, the definition of locality boundaries is a matter of expert judgment.

An analogue of locality in traditional macroseismic studies is a village or a medium-sized town, to which an intensity value can be assigned. Sometimes, intensities differing by one or even two degrees are found for the same town, commonly due to site effect conditions. In such a case, conservatively, the highest intensity is generally taken, if the scale of representation does not allow all values to be displayed. The site is an analogue of a single macroseismic object. For a single building, the degree of damage can be defined, but not a well-defined intensity value.

In the classical macroseismic intensity assessments, a statistical analysis of damage levels for different categories of structures (differentiation is based on their vulnerability) within the city or town provides the intensity of shaking at this locality. In contrast, expert judgement of EEEs at a given locality is used for the ESI scale. This evaluation concerns both the probable range of intensities observed as well as the accuracy of the estimates. By adopting the “locality-site” concept, a merging of environmental and macroseismic effects into a unified intensity scale might be almost straightforwardly achieved.

The EEE intensity field resulting from this process provides an additional portrait of the earthquake. A comparison with the “traditional” macroseismic fields of the same earthquake has shown so far a fairly good agreement, generally within one degree (see references on its application listed in the “Introduction”). However, significant discrepancies in intensity assessments (> one degree) must be accounted for with a dedicated analysis. They can result from: (a) a site amplification selectively affecting the human or natural environment to a greater extent (for example, topographic or soft sediment amplification effects), (b) an anomalously strong or light damage because of building type and/or quality. Moreover, although it has worked satisfactorily so far, the ESI scale is still a novel tool. Some fine-tuning might emerge as useful in the future, based on experience gained by its application to more cases.

Based on the ESI scale (Tables 1, 3), the size of the epicentral area, i.e., the area where the strongest effects are observed, as well as the total area affected by EEEs, allow I 0 to be estimated. The latter also scales with rupture parameters: surface rupture length (SRL) and maximum displacement (MD) that are also rather well correlated with magnitude. This is particularly useful for assessing the size of paleo-earthquakes, where proxies of intensity/magnitude data are extremely rare. Relationships between the total area of specific secondary effects (e.g., landslides or liquefactions) and I 0, despite more complex, also make it possible to estimate the earthquake size (Table 1). In addition, recent studies confirm that factors such as the total number, size and density of landslides may provide a sound means of assessing the strength of shaking across an earthquake-affected area (Hancox et al. 2002, 2014, and references therein).

Table 3 Ranges of surface faulting parameters and typical extents of total area for each intensity degree (SRL: surface rupture length, MD: maximum surface displacement or deformation, Total Area: area of occurrence of EEEs)
Full size table

However, there are instances where the instrumental and macroseismic epicenters are far apart; the typical example is the Mexico City earthquake in 1985, where an M8.1 subduction event caused its greatest damage more than 350 km away, in the soft lacustrine setting of Mexico City (e.g., Campillo et al. 1989; Chavez-Garcia and Bard 1994). This is a particular case where the local amplifications (mostly liquefaction and ground compaction in Mexico City and sparse landslides over a wide region) have to be taken into proper consideration, together with other information, most likely leading to excluding the Mexico City data for assessing I 0. The characteristics of the EEEs of many earthquakes worldwide, including the assessed ESI intensity, are reported in the EEE Catalogue, hosted by ISPRA in the framework of the INQUA TERPRO Focus Area on Paleoseismicity project “A global catalogue and mapping of Earthquake Environmental Effects” (http://www.eeecatalog.sinanet.apat.it/terremoti). The same data set is currently also accessible through the United Nations—International Atomic Energy Agency—International Seismic Safety Center (IAEA-ISSC) webpage (issc.iaea.org).

4 Case Studies of Application of the ESI Scale

Many papers published so far illustrate the application of the ESI scale to recent, historical and paleo-earthquakes (cf. list of references in the “Introduction”). In the following, some recent and historical case studies, representative of different seismotectonic environments (location in Fig. 3), focus on some key questions. The first case study is an example of how regional and local factors affect earthquake size assessment. Then, the 1805 Italy earthquake is used to illustrate how the distribution of maximum secondary effects makes it possible to assess I 0 and to identify the seismogenic source. In a third example, the Denali 2002 earthquake serves to highlight the problem of epicentral intensity assessment, when the epicenter is far from populated areas.

Fig. 3
figure 3

Location of earthquake case-studies discussed in this paper

Full size image

4.1 Focusing on How Regional and Local Factors Affect Earthquake Size Assessment: the June 27, 1957, Muya (Baikal, Russia) Earthquake

Systematic studies of paleo-earthquakes started in the Baikal region in the early 1960s; the first regional catalogue of paleo-earthquakes was published in Kondorskaya and Shebalin (1982).

Baikal is a unique zone of active continental rifting. The almost pure normal faulting of published CMT-solutions (inset of Fig. 4) is in full agreement with this general geodynamic framework. However, the largest known earthquake in the region, the 1957 Muya earthquake (M lh = 7.6, M S = 7.5), demonstrates a much more complex faulting style, with a relevant strike-slip component found both in source modeling and in surface ruptures (Fig. 5). Surface ruptures associated with the earthquake were studied during several field investigations and summarized in reports and papers by Kurushin ( 1963 ), Solonenko ( 1965 ), Solonenko et al. (1966, 1985) and Kurushin and Mel’nikova (2008) (Fig. 5). Figure 4 summarizes the macroseismic information, based on questionnaires collected in 1958 by the Institute of the Physics of the Earth, RAS and data from Solonenko et al. (1958). Noteworthy is the far-reaching extension of the felt area, over 700 km away from the surface faulting zone.

Fig. 4
figure 4

Approximate macroseismic (white star) and instrumental (blue star, according to Vvedenskaya and Balakina 1960) epicenters of the Muya, 1957, earthquake, and MSK intensities (degrees in Roman numerals). Inset in lower left: source mechanisms in the Baikal region for events with M o ≥ 1024 dyn cm (M w ≥ 5.3) (http://www.globalcmt.org/CMTsearch.html), showing a dominant extension in the Baikal rift grading into ca. east–west left-lateral slip eastward of the Muya region

Full size image
Fig. 5
figure 5

Surface faulting of the 1957 Muya earthquake; stippled in white is a doubtful portion of the coseismic rupture (L rupture length, Va average vertical component of slip, Ha average horizontal component of slip). Orange arrows mark sections with substantial horizontal slip. Three main segments can be discerned, separated by transition zones. The focal mechanism shown, from Vvedenskaya and Balakina (1960), is very similar to the three subevents reconstructed by Doser ( 1991 ). White dot is approximate location of top photograph in Fig. 11

Full size image

According to the ESI scale, epicentral intensity can also be assessed based on the total length (SRL) and maximum displacement (MD) of the observed surface faulting. The surface ruptures of the Muya event occurred along three main WNW trending en echelon segments with partly lateral and normal slip components (Fig. 5). The reported 20–25 km of SRL corresponds to an epicentral ESI intensity X. The maximum offset (vertical) was 3.3 m, which, for normal faulting, is also in agreement with an intensity of at least X. Instead, since the villages nearest the surface rupture zone were 50 km away, the maximum observed macroseismic intensity was rather low compared to the magnitude of the event, not reaching VIII. Thus, if only macroseismic effects are considered for epicentral intensity assessment, I 0 would be underestimated by no less than two degrees.

The epicentral locations determined instrumentally differ from each other by more than 100 km (ISC 2012). The instrumental hypocenter depth varies from 10 (Doser 1991) to 22 km (Balakina et al. 1972). Tatevossian et al. (2010) have assessed a depth around 20 km, applying the macroseismic field equation for Baikal region found in Kondorskaya and Shebalin (1982) for M lh = 7.6 and a radius of felt shaking of 700 km (Fig. 4). The rather short surface faulting zone (20–25 km) is also supportive of the relatively deep source, common in the Baikal region (e.g., Déverchère et al. 2001). For the sake of comparison, the 2003 Altai earthquake (M s = 7.4) was accompanied by 70 km of surface ruptures (SRL) (Tatevossian et al. 2009). According to Wells and Coppersmith (1994), a 20–25 km SRL would correspond to M s 6.6, much lower than that measured (see, for example, Tatevossian 2011). The short SRL and large MD (maximum displacement) could also be representative of a source with high stress drop, a possible trait of the regional seismicity. Thus, the influence of regional tectonics and structural characteristics on the coseismic effects should also be considered in the paleoseismic evaluations.

4.2 Focusing on the Distribution of Maximum Secondary Effects to Assess I 0 and to Identify the Seismogenic Source: the July 26, 1805, Molise (Southern Italy) Earthquake

The 1805 Molise earthquake (Southern Italy) was characterized by a main shock (macroseismically derived magnitude M a = 6.6) followed a few hours later by two important aftershocks. The epicentral zone was centered in the Bojano plain (I 0 = X MCS; Locati et al. 2011), and an area of about 2000 km2 was affected by MCS intensities ≥VIII. Esposito et al. (1987) locate the macroseismic epicenter of the main shock (Figs. 6, 7) at Frosolone (I max = XI MCS). The epicenter of the second main shock was located at Morcone, some tens of kilometers to the southeast.

Fig. 6
figure 6

1805 Molise earthquake: comparison of MCS and ESI macroseismic intensity fields. White lines MCS isoseismals (MCS degrees in Roman numerals, after Esposito et al. 1987). The highest intensity was seen at Frosolone, slightly west of the 1456 epicenter. ESI local intensities (triangles) are after Serva et al. (2007). Only the epicenters of the three major historical earthquakes in the near area are shown. Likely, the 1456 and 1805 events occurred along the same fault

Full size image
Fig. 7
figure 7

Distribution of the main geological effects reported for the 1805 Molise earthquake (data after Porfido et al. 2002). The Bojano fault system, running northwest–southeast, has controlled the Quaternary evolution of the Bojano and adjacent intermountain basins, marked in light green, the largest of them being Morcone and Sepino (? where rupture is doubtful)

Full size image

The historical accounts of the 1805 earthquake mention about one hundred seismically induced environmental effects, mostly in the near-field area, although some were reported as far away as 70 km from the epicenter (Esposito et al. 1987; Porfido et al. 2002). The most relevant effects documented in contemporary sources are described in Table 4 and mapped in Fig. 7. Among these, vertical ground displacements of about 1.5 m at Guardiaregia and Morcone are interpreted as primary effects. Along an antithetic fault, a long fracture reported between Pesche and Miranda and northeast of Castelpetroso may also be interpreted as evidence of surface faulting. This fact would indicate that the total rupture length was about 40 km, with maximum displacements of about 1.5 m at Guardiaregia. Consequently, the ESI epicentral intensity is X.

Table 4 Descriptions of relevant environmental effects induced by the 1805 Molise earthquake
Full size table

The earthquake also triggered a number of secondary effects, notably, slope movements, hydrological anomalies and liquefaction. At least 26 mass movements (rock falls, topples, slumps, earth flows and slump earth flows) were recorded (Esposito et al. 1987, 1998; Porfido et al. 2007; Serva et al. 2007). In 29 localities, principally around Bojano (Biferno springs) and the Matese Massif, 48 hydrological anomalies were reported, mainly changes in water discharge from springs. Flow increases were observed in at least 16 springs, mainly SSW of the Matese Mountains. New springs also appeared, one of them at Bojano was active for about 2 months after the earthquake (Esposito et al. 1987, 2001; Porfido et al. 2002). Only one clear case of liquefaction was reported at Cantalupo. Flames were also seen escaping from the supposed fault rupture at Morcone, likely burning methane from palustrine sediments. Furthermore, anomalous sea waves were observed in the gulfs of Naples and Gaeta (Caputo and Faita 1984; Esposito et al. 1987; Tinti and Maramai 1996; Maramai et al. 2014), the cause of which is unidentified, possibly submarine landslides or resonance effects.

The application of the ESI scale to secondary effects has allowed local intensity values between V and X to be assigned to about 50 municipalities (Serva et al. 2007; Fig. 7). The total area of relevant ground effects (ESI local intensities ≥VII) amounts to about 5300 km2. According to Table 3, the ESI epicentral intensity is X.

In conclusion, this case study shows that both primary and secondary effects, when properly taken into account, can serve for:

  1. 1.

    Intensity assessment: in fact, two independent applications of the ESI, from the total area of secondary effects and from SRL, have yielded the same I 0 = X. This result is also consistent with the independently assessed macroseismic intensity (MCS = X).

  2. 2.

    Identification of the seismogenic source: the scenario of secondary effects allows the seismogenic source to be located on the southwestern border of the Bojano basin, in good agreement with local evidence of surface faulting reported in historical documents at Morcone, Guardiaregia and Pesche (Table 4), confirmed by recent studies (Guerrieri et al. 1999; Blumetti et al. 2000; Galli and Galadini 2003).

4.3 Focusing on Epicentral Intensity Assessment, When the Epicenter is Far from Populated Areas: the November 3, 2002, Denali (Alaska) Earthquake

The 2002 Denali earthquake (M w = 7.9) was the largest strike-slip earthquake in North America in more than 150 years (Hansen and Ratchkovski 2004; Martirosyan 2004). The epicenter was located 135 km south of Fairbanks and 280 km north of Anchorage (Fig. 8). The total surface fault rupture length was about 330 km, running ca. east–west at its northwestern tip and then WNW-ESE; the maximum displacement was 8.8 m right-lateral, measured west of the Denali and Totschunda fault junction, and over 5 m reverse on the Susitna Glacier fault (Eberhart-Phillips et al. 2003; Aagaard et al. 2004; Crone et al. 2004; Frankel 2004; Hansen and Ratchkovski 2004; Haeussler 2009). Three sub-events were identified (Eberhart-Phillips et al. 2003; Frankel 2004), and a total seismic moment equivalent to M w = 7.9 was inferred from GPS data, also consistent with that derived from InSAR data (Wright et al. 2004). The first sub-event (M w = 7.2), located near the instrumental epicenter, was associated with the rupture along the Susitna Glacier Fault. The second sub-event (M w = 7.3) was 50–100 km east of the epicenter, where the surface offset of the Denali Fault reached over 6 m. It is noteworthy that, despite this, the Trans-Alaska Pipeline did not produce any oil spill, thanks to a technical solution devised after the experience of the 1964 Alaska earthquake (Honegger et al. 2004). The third sub-event had the largest seismic moment, equivalent to M w = 7.6, and was located about 130–220 km east of the epicenter, where the maximum surface offset of 8.8 m was measured (Eberhart-Phillips et al. 2003; Frankel 2004).

Fig. 8
figure 8

2002 Denali earthquake: Community Internet Intensity (CII) and GI (Geophysical Institute, University of Alaska Fairbanks) intensity distribution based on the report materials (Martirosyan 2004). Star indicates the instrumental epicenter. The fault ruptures are plotted, simplified, according to Eberhart-Phillips et al. (2003). The box encloses the epicentral area (Fig. 9)

Full size image

The effects of rupture directivity are particularly remarkable for the Denali fault event. Ground shaking effects were reported as far away as almost 6000 km from the epicenter: for example, in Louisiana, seiches rocked boats and broke moorings (Eberhart-Phillips et al. 2003; Cassidy and Rogers 2004). Local bursts of seismic activity were also observed far away in volcanic and geothermal areas, especially if lying in the direction of the Denali rupture propagation, e.g., in the Yellowstone area, at a distance of 3100 km (Husen et al. 2004; Moran et al. 2004).

Several papers describe the surface faulting of the 2002 earthquake (Eberhart-Phillips et al. 2003; Crone et al. 2004; Haeussler et al. 2004; Haeussler 2009). The slip on the Susitna Glacier thrust fault generated structures ranging from simple folds on a single trace to complex thrust-fault ruptures and pressure ridges on multiple, sinuous strands, in a deformation zone locally wider than 1 km. A maximum vertical displacement of 5.4 m on the south-directed main thrust was measured. The principal surface break occurred along 226 km of the Denali fault, with average right-lateral offsets of 4.5–5.1 m and a maximum offset of 8.8 m near its eastern end. Finally, dextral slip averaging 1.6–1.8 m transferred southeastward onto the Totschunda fault for another 66 km.

The secondary geological effects of the 2002 Denali earthquake were mostly landslides, liquefaction and ground cracks (Fig. 9; Eberhart-Phillips et al. 2003; Harp et al. 2003; Hansen and Ratchkovski 2004; Jibson et al. 2004, 2006). Despite the thousands of landslides that were triggered, primarily rock falls, rock slides and rock avalanches, ranging in volume from a few cubic meters to tens of millions of cubic meters (i.e., the rock avalanches that covered much of the McGinnis Glacier), Jibson et al. (2006) have estimated them to be far less than expected for an earthquake of this magnitude and unusually concentrated in only a narrow zone about 30 km wide, straddling the fault-rupture zone over its entire length. The overall affected area of 10,000 km2 was significantly smaller than that triggered by other earthquakes of comparable magnitude (Harp et al. 2003, and references therein). To explain this, Jibson et al. (2006) suggest a deficit in high-frequency shaking, with the highest accelerations being confined to the vicinity of the fault zone.

Fig. 9
figure 9

2002 Denali earthquake: distribution of geological effects based on survey data from authors cited in text. The fault ruptures are plotted, simplified, according to Eberhart-Phillips et al. (2003)

Full size image

Liquefaction features were observed over a much greater distance, up to 120 km away from the rupture zone. The liquefaction was more extensive and severe to the east, near the third sub-event, on the Holocene alluvial deposits of the Robertson, Slana, Tok, Chisana, Nabesnaand Tanana Rivers (Eberhart-Phillips et al. 2003; Kayen et al. 2004). Apparently, according to the latter authors, the minimum shaking levels and duration requirements for liquefaction were reached more extensively than those needed to trigger rock falls and rock slides. Actually, the third sub-event had a longer duration and period of shaking than the previous two (Harp et al. 2003; Kayen et al. 2004; Jibson et al. 2006). However, also the stabilizing effect of permafrost may have contributed to reducing the number of mass movements. Moreover, the region surrounding the high peaks of the Alaska Range displays a rather smoother morphology with wide alluvial valleys hosting peri-glacial liquefaction-prone deposits.

Many hydrological anomalies were reported, such as water waves, water spill from swimming pools, seiches in lakes and rivers, muddy well waters, at distances up to 3500 km across western Canada and in the Seattle basin (Cassidy and Rogers 2004; Barberopoulou et al. 2006; Sil 2006).

The US Geological Survey carried out an indirect macroseismic survey (Community Internet Intensity, CII), later expanded by the University of Alaska Fairbanks (Martirosyan 2004). The combined dataset contains intensities for more than 155 inhabited locations, 29 of which reported a maximum intensity MM IX. Of these, 28 are located in the eastern part of the ruptured fault, with an average distance from the fault of 27 km (Martirosyan 2004). Most of the reported intensity data come from localities very far (10–100 km) from the fault, and more than 70 % of them are V MMI or less. Their spatial distribution is strongly inhomogeneous, reflecting the sparse population. The USGS has also produced a Shakemap of the event, merging ground motion and macroseismic data (http://earthquake.usgs.gov/earthquakes/eqinthenews/2002/uslbbl/images/AK_mmi_new.jpg) that shows instrumental intensities of at least IX, especially in the central and eastern sections of the rupture zone.

The application of the ESI scale to the diagnostic EEEs reported in the papers above provides intensities between VII and XII for 131 sites (Fig. 10). These intensities are based on evidence of surface faulting (primary effects), slope movements, liquefaction and ground rupture features (secondary effects). Based on the maximum horizontal slip of 8.8 m and the total surface rupture length of 330 km, the maximum ESI scale intensity would be XII. The spatial distribution of secondary effects (landslides and liquefactions), with a total affected area of at least 30,000 km2, suggests epicentral intensity XI. Considering only mass movements, the ESI would range between X and XI (Table 3). Such contrasting values might be justified by the multiple rupture and widespread sediments particularly susceptible to liquefaction. As a whole, intensity XI appears most reasonable. The distribution and characteristics of EEEs locates the macroseismic epicenter west of Mentasta, broadly ESE of the instrumental epicenter and near the third sub-event (Fig. 10).

Fig. 10
figure 10

2002 Denali earthquake: intensity field based on CII and ESI intensities. For symbols, see legend in Fig. 9. The assessed epicentral (I 0) ESI intensity is XI, based on the amount of slip

Full size image

5 Discussion

The three case studies presented in the previous chapter, together with Table 2, help underline the efficacy of environmental effects for improving the evaluation of earthquake size. The maximum intensity of the 1957 Muya earthquake, defined on the basis of its effects on man-made structures is two degrees less than the most reasonable epicentral intensity, simply because the town nearest to the epicenter is 50 km away. The same holds true for the much more recent 2002 Denali earthquake.

Sprawling infrastructure is not limited to urbanized areas: extensive transportation, water channels and pipeline networks run across the countryside, as well as high-hazard installations such as dams, chemical, nuclear power and Liquefied Natural Gas (LNG) plants. These facilities need be evaluated for geological hazard, and the ESI scale provides a basis for achieving this task. It allows, in a simple and effective manner, the building of scenarios for given intensity values of potential geohazards that an area is liable to face, considering solely its geomorphological and soil characteristics. Practical applicants might be structural designers, regulators, civil protection agencies and administrators, also for a more effective public communication. Even insurance companies may benefit from a comprehensive representation of the seismic hazard at a site, inclusive of geohazards. This remains valid even when dealing with critical facilities, where the most up-to-date and sophisticated methods of SHA are generally applied.

The 1805 Molise earthquake demonstrates that even when surface ruptures are not positively recognized in historical documents, the total area affected by secondary environmental effects can yield a quite accurate assessment of epicentral intensity. This is inferred from the good match of independent epicentral intensity assessments based on primary and secondary environmental effects. At the same time, this case proves the compatibility of the ESI scale with traditional macroseismic scales, where detailed and complete information on earthquake impact on buildings in the epicentral area is available.

Although it has been satisfactorily applied so far, it should be borne in mind that the ESI scale is still a novel tool. Some fine-tuning might prove useful in the future, based on the experience being gained by its application to more cases. In particular, intensity estimates based on primary effects, i.e., SRL or displacement, are a crucial aspect, not considered, for example, in the MMI-derived scale used in New Zealand (Hancox et al. 2002; Dowrick et al. 2008), which instead takes advantage of the shaking-related effects such as mass movements, liquefaction and other ground damage effects. Large fluctuations in the faulting parameters are commonly observed when compared to magnitude (Stirling et al. 2013, and references therein). The causes are manifold: depth, kinematics and multiple ruptures. The same happens with intensity; therefore, the boundary values given in Table 3 provide a reference frame that must be accompanied by a careful consideration of regional features, such as other rupturing events in the same region, thickness of the seismogenic layer and regional stress–strain state. Possibly, also a characteristic stress-drop may play a role, but there is no general consensus about what it can actually be. For instance, the anomalously short SRL observed for the M s = 7.5 Muya, 1957, earthquake (25 km) is in agreement with all reports of coseismic surface rupture lengths in the Baikal region, which never exceed 45 km (e.g., Tatevossian et al. 2010). This suggests that such a feature is “characteristic” for the region, driven by a high stress-drop or by an unusually deep seismogenic layer, limiting the portion of the source rupture emerging at surface. Also, the complexity of the rupture process, with nearly pure strike-slip at depth, revealed by focal mechanisms and substantial vertical components of deformation at the surface, points out the need for a careful interpretation of paleo-earthquake data obtained in situ, in light of the regional geodynamics.

In developing the ESI scale, it has always been considered that realistic seismic hazard assessment must be based on a sufficiently wide time-window of earthquake history. Historical accounts of geological effects commonly inform on secondary effects, such as ground cracks, liquefaction and landslide phenomena, but may also include indications on fault displacement, sometimes even rupture lengths. In addition, erosion or trenching can expose paleoseismic evidence, in the form of time-constrained fault displacements and other geological features, especially filled cracks and liquefaction. Not all EEEs have the same weight in the intensity assessment. Each type of effect depends on some specific parameters, sometimes regional (primary effects), sometimes very local (secondary effects), like morphology, stratigraphy, water table depth and saturation. Some EEEs appear to start from a degree threshold, but subsequently do not allow precise degrees to be defined; others perform better to assess intensities, but always in a statistical sense. There are no clear steps in the number and size of most EEEs between one degree and the next: Natura non facit saltus. In general, the more the available EEEs in terms of number and types, the better the intensity estimate. Despite the fact that a single piece of evidence (or several at the same site) may not provide a precise indication of the magnitude of the causative event, it may still represent a key information to confirm the indications deriving from instrumental data or active tectonics studies. Moreover, the size or simply the presence of an EEE does allow a minimum intensity to be assessed, similarly to what is commonly applied in paleoseismology to derive a minimum magnitude threshold. This minimum threshold and the time interval between independent faulting/shaking events can have a strong impact on seismic hazard assessment. In the ESI scale, the uncertainties regarding the value, in terms of earthquake characterization, to be attributed to a local, often single, evidence of faulting, liquefaction or earthquake-triggered mass movement are somehow incorporated in the discrete nature of intensity degrees, which are defined by large boundary values.

With more and more strong-motion recordings made available, attenuation relationships have proliferated based on moment magnitude as the most representative parameter of the source. At the same time, seismologists endeavor to understand the phenomena involved in source rupturing and propagation, but rarely are their insights taken into account in the design of facilities. Their studies of major recent earthquakes show that the source is complex, and parameters such as rupture length and stress drop, directly linked to slip on the fault plane, are extremely important for source characterization (e.g., Mohammadioun and Serva 2001). Stress drop, a parameter linked to maximum acceleration, is strongly variable from one seismotectonic region to another, or even from one fault or seismogenic source to another (e.g., Fry et al. 2010). Moreover, the commonly applied ground motion attenuation relationships quite often reveal considerable discrepancies between predicted values and actual data (e.g., Sørensen et al. 2009). All this calls for a reasonably simple tool to calibrate earthquake size in a straightforward manner from observations of its impact, to be as independent as possible of constraints imposed by models and assumptions.

Macroseismic intensity scales were conceived to determine the size of an earthquake by noting the effects it had caused. However, intensity assessment can be biased by the varying levels of vulnerability displayed by man-made environment. The advantage in founding an intensity scale on geological effects is the opportunity it provides to counterbalance the unavoidable inconsistencies caused by such variable vulnerability. Also, the natural environment is characterized by a varying level of vulnerability. Ground vibratory characteristics affect both secondary geological phenomena and artificial structures. However, the natural environment is more stable on the long-term scale: although ephemeral, it is capable of conserving many EEEs, compared to the man-made environment, being, therefore, able to retain the record of earthquake traces much longer than man-made structures. In fact, damaging effects on the latter strongly depend not only on foundation soil, but also on materials, design and foundation type, which undergo rapid changes, especially in the modern construction industry. This presents an obstacle to a sound comparison of present and past damage. Figure 11 shows examples of this fact, based on the epicentral areas of three large recent earthquakes.

Fig. 11
figure 11

Top Photograph shot from a helicopter in the epicentral zone of the 1957 Muya earthquake in 2005, i.e., 48 years after it occurred. The fault trace is still clearly visible, and anyone today can verify its parameters. Middle Google Earth image captured in 2010 of the epicentral area of the 1988 Spitak, Armenia, earthquake (22 years after). The two villages in the image were totally destroyed, and intensity 10 was assessed there. Today, it is no longer possible to check any damage-related feature, because the villages have been rebuilt and repaired. Bottom The same is true, for opposite reasons, in the epicentral area of the 1995 Neftegorsk earthquake in northern Sakhalin Island, as evident in the Google Earth image of 2010 (15 years after the event). Only a portion of the portrayed building compound collapsed, and intensity 8–9 was assessed for the town at that time. But Neftegorsk was completely abandoned, and this was the reason for further heavy destruction. If someone were to address the 1988 Spitak and the 1995 Neftegorsk earthquakes today, the wrong impression would be that Neftegorsk suffered much heavier damage than the villages in the Spitak epicentral area

Full size image

6 Conclusions

The main goal of this paper was to introduce the ESI intensity scale, based on environmental effects of earthquakes (EEEs), to the widest possible audience. The key message is that, despite the advent of magnitude, earthquake intensity remains a significant seismic parameter for reliable SHA, especially when EEEs are properly taken into account. In fact, this study shows that epicentral intensity may be underestimated by two or even more degrees if the contribution of environmental effects is ignored. Table 2 demonstrates this is a generalized problem of worldwide proportions. The case studies presented earlier have served to substantiate this conclusion.

When the epicentral area is located in a densely populated region and detailed information on building damage is available, the ESI scale is to be used in conjunction with one of the “traditional” scales (MSK, MM, EMS). In fact, it must always be borne in mind that intensity in a specific locality results from an informed assessment based on three categories of effects: on human, infrastructures (built environment) and natural environment. Only in such a case does the ESI, which has been developed from the MMI and MSK scales, provide an intensity assessment that is consistent with, and complements the results of, the other scales.

The use of EEEs offers the possibility of comparing earthquake intensities worldwide. In fact, EEEs are uninfluenced by cultural and technological aspects, which may differ significantly from region to region. Moreover, earthquake-prone areas can be located in sparsely or even completely unpopulated regions, where only the effects on the natural environment might be observable. In such a case, ESI becomes the only tool able to calibrate earthquake intensity. The same holds true for events with macroseismic intensity of X and above, where, with most structures being ruined, all the damage-based intensities tend to saturate. Thus, any assessment based on such damage is biased. Conversely, the size of some EEEs continues to be proportional to the intensity of the earthquake.

It is important to bear in mind that the natural environment may have a much longer memory than the built one. As the impact of an earthquake on a man-made environment depends on the distribution of urbanized areas, it is difficult to compare two seismic events that occurred in the same area, but at very different times. Conversely, this can be achieved on the basis of documented EEEs. This approach extends the time coverage of earthquake catalogues to prehistoric times. Local evidence of surface faulting and the size of secondary effects (i.e., liquefaction effects, landslide-dammed lakes, etc.) pertaining to prehistoric events can be evaluated via detailed paleoseismological investigations in natural or artificial exposures. Thus, the ESI scale allows intensity to be estimated also for paleo-earthquakes.

The ESI scale provides a very convenient guideline for the survey of EEEs in earthquake-stricken areas, ensuring their complete and homogeneous analysis. Its application, continuously improving the EEE catalogue, would consequently also improve the correlations cited above relating magnitude to surface faulting or other geological parameters, which are significantly affected by the inhomogeneity of available data, rarely documented so far with the same criteria and quality.

In conclusion, the factoring in of the geological and geomorphological effects of earthquakes allows the most comprehensive earthquake risk scenario to be built, a crucial need for all stakeholders, especially designers, geotechnical engineers, hazard analysts, regulators, civil protection agencies and insurers, as well as the general population.