1 Introduction

On 20 May 2012, at 4:03 local time (2:03 UTC), a damaging earthquake (Mw 5.9) struck a large part of the Po Valley between the cities of Ferrara, Modena, Bologna, and Mantova (Northern Italy). The epicenter was located by the Istituto Nazionale di Geofisica e Vulcanologia (INGV) seismic network ISIDe (2012) at 44.889° N and 11.228° E, approximately 30 km west of Ferrara. The event was preceded by a foreshock that occurred at 01:13 local time (Mw 4.8) and followed by a sequence that lasted for weeks, with six strong shocks with ML >5 (Fig. 1). The strongest occurred on 29 May 2012, at 9:00 local time (Mw 5.7) and 12:55 local time (Mw 5.3), causing additional heavy damage in the western part of the area already hit on May 20.

Fig. 1
figure 1

Map showing the Emilia earthquake sequence (epicenters of events that occurred up to 5 July 2012). White star main shock of 20 May (Mw 5.9); black star main shock of 29 May (Mw 5.7); gray stars Mw ≥ 5.0 aftershocks; squares 4.0 ≤ Mw < 5.0 aftershocks; large dots 3.0 ≤ Mw < 4.0 aftershocks; small dots Mw < 3.0 aftershocks (modified after INGV terremoti (2014))

Full size image

The area affected by the earthquake sequence of 2012 is characterized by a low-to-moderate level of seismicity. In an area within 30–40 km from the epicenters of the main shocks of 20 and 29 May 2012, historical information provided by the current Parametric Catalogue of Italian Earthquakes (hereinafter CPTI11, Rovida et al. 2011) does not report any significant event, with the only exception of the event that struck Ferrara on 17 November 1570 (Mw 5.5) (see also Castelli et al. 2012). In the last decades, sporadic small-to-medium magnitude earthquakes have occurred in this sector of the Po Valley, such as the events of 6 December 1986 (Mw 4.6) and 2 and 8 May 1987 (Mw 4.7, 4.6, respectively), which affected the northern sector of the Modena province (Locati et al. 2011; Rovida et al. 2011).

The macroseismic effects of the 2012 Emilia earthquakes have been carefully investigated either according to European Macroseismic Scale (EMS98, Grünthal 1998; Tertulliani et al. 2012a) or according to the Mercalli-Cancani-Sieberg (MCS, Sieberg 1930) guidelines (Galli et al. 2012); in both cases, the results are very similar.

The EMS98 survey was performed on a total of 87 localities with the specific goal of describing, as accurately as possible, the damage scenario for each locality. It was possible to define the intensity distribution for the 20 May event in about 50 localities, surveyed before the occurrence of the 29 May shocks. The assessment of EMS98 intensities is based on the observations of damage suffered by the residential building stock, previously classified in typology and vulnerability classes. The intensity assessment is based on the percentage of the different damage grades suffered by each vulnerability class. Damage on monumental and special buildings was reported too, although it is not considered for the estimation of intensity, as suggested by the guidelines of the EMS98 scale. Such kind of buildings have the characteristic of being unique or very few in one place, and therefore, they cannot be computed in a statistical way as other building typologies (Grünthal 1998).

The MCS survey was performed in 190 localities, 52 before the 29 May shocks, with the main intent of defining, as quickly as possible, the area of major damage (Galli et al. 2012) for civil protection purposes. The main difference between the use of EMS98 and MCS scales lies in the fact that the latter assesses the intensity value on the base of a comprehensive scenario, without detailing building types, damage grades, and quantities, considered individually by EMS98.

In broad terms, we can confirm that most of the damage was concentrated in old town centers, where the historic monumental buildings were heavily struck. Most of the total or near collapses involved both manufacturing plants (industrial warehouses, farmhouses, barns) and monumental buildings (fortresses, churches, and towers or belfries). Dedicated investigations confirmed that the historical/monumental building stock suffered the major damage as demonstrated in Modena et al. (2013). On the other side, the residential building stock did not suffer much damage: generally from light to moderate (Penna et al. 2013). As a result, in many sites, severe destruction was clustered in a general scenario of light damage. The final estimate leads to moderate intensity values: the maximum intensity value assigned is Imax 7–8 for the EMS98 as well as for the MCS. As a consequence, the values of macroseismic magnitude calculated both from EMS98 and MCS intensity data (Tertulliani et al. 2012a; Galli et al. 2012) by means of the code Boxer 4.0 (Gasperini et al. 2010) (Table 1) are rather lower than the instrumental magnitude Mw 5.9 of the main shock. In particular, the macroseismic magnitude computed from the cumulative intensities at the end of the sequence is Mw 5.3. Such results instill doubt that the assessment is not appropriate.

Table 1 Intensity values (EMS and MCS) and macroseismic magnitudes related to the 20 May shock and the cumulative effects for the whole sequence (last M > 5 aftershock that occurred on 3 June 2012)
Full size table

This discrepancy can also be found on the analysis of the magnitude-intensity recent regression lines based upon the CPTI catalog (Fig. 2) (Gruppo di Lavoro 2004; Pasolini et al. 2008) which suggests intensity values ranging from 8 to 9 MCS following a Mw 5.9 earthquake, significantly higher than the intensity values assessed for the Emilia seismic sequence.

Fig. 2
figure 2

Magnitude-intensity regression line after CPTI04 (redrawn from Gruppo di Lavoro 2004). Magnitude of the 20 May 2012 Emilia earthquake and expected intensity are also indicated

Full size image

What are the possible causes of such discrepancy? Considering that the regression is mainly based on historical earthquakes, some possible hypotheses arise:

  1. 1.

    The 2012 Emilia sequence shows lower intensity values because it affected urban contexts more heterogeneous and much less vulnerable than those in the past;

  2. 2.

    Some historical earthquakes, especially those occurred ages ago and provided with little information, could tend to be overestimated due to the scarcity of data available.

Those topics were recently analyzed by Rong et al. (2011), comparing CPTI catalog with the Switzerland seismic catalog, revealing that the Italian side magnitudes were significantly higher by 0.5–0.6 unit, especially for small-to-moderate earthquakes. They conclude that a possible cause was the use of different catalog sources and an overestimation of pre-1975 earthquake intensity. Analogously, Hough (2013) demonstrates that intensity values of historical North American earthquakes, based on fragmentary accounts, can lead to overestimation, considering the most severe rather than representative effects. In the following paragraphs, we will try to clear up such questions and give a response to them.

2 Is it true that the 2012 Emilia earthquakes show lower intensity values because they affected a context more heterogeneous and much less vulnerable than that in the past?

In the course of time, urban settlements have undergone several changes, and intensity assessments have been tuned to comply with these changes. It is a fact that most of present Italian urban settlements consist of a delimited old town center, with mostly stone or masonry buildings, and a large modern expansion area, mostly composed of reinforced concrete buildings. Therefore, building typologies characterized by different vulnerability levels coexist within the same settlement, with the result of a high variability of damage in the same locality (Meroni et al. 2000; ISTAT 2001). Modern intensity assessments take into account the whole building stock, oldest and newest, considering their own statistical weight. As a consequence, towns with recent developments show an overall reduced vulnerability in respect to those that have not experienced any changes in times. Moreover, the estimation of the intensity in a recent town is assessed on an urban sample that is considerably more developed, as compared to the past; therefore, the whole damage is spread out over a larger building sample. These changes could contribute to lower intensity values in comparison with those assessed for earthquakes that occurred ages ago, when settlements were smaller and building stock was theoretically more vulnerable.

To demonstrate the above statement to be true, we have tried to handle the 2012 Emilia earthquake as if it had occurred in the past before the recent building development. As the macroseismic survey performed in EMS98 (Tertulliani et al. 2012a) provided us with a detailed dataset, where buildings are classified according to vulnerability typologies, we assessed the damage on the historical building stock only, adopting the same methodology used to study historical earthquakes. Basically, to obtain a building stock similar to one of the past, we filtered our data discarding recent buildings (reinforced concrete buildings), and we analyzed the filtered data in the frame of the MCS scale, traditionally still used in Italy to examine historical earthquakes. This approach allows us to assess the effects as a whole, without considering building types, damage grades, and quantities individually.

The two steps, (a) discarding recent buildings and (b) analyzing the data with a MCS approach, can be viewed as a sort of historical reading of the earthquake. We believe that the historical reading represents a reliable method to read a present-day earthquake in the perspective of a comparison with the historical seismicity of the area.

2.1 The historical reading: a macroseismic analysis of earthquake effects in old town centers

In order to exclude from our analysis the recent urban development, we restricted our investigation to stone and the masonry buildings only. More precisely, we selected those localities for which a delimited old town center was clearly distinguishable (Fig. 3), with a homogeneous building stock and negligible presence of concrete buildings.

Fig. 3
figure 3

Example of the analysis applied to the town of Mirandola. The central part of the town (inside the polygon) is well distinguishable also from the aerial view and represents the old town of Mirandola. In this part of the city, both residential and monumental buildings are masonry structures (clay bricks). The residential stock is almost entirely composed by old traditional two-story brick houses, with negligible presence of concrete buildings

Full size image

Out of 87 localities, 33 display the useful characteristics for our analysis, having the old town center well identifiable. Data collected during our surveys were then examined in order to evaluate the intensity in terms of MCS scale. The results are shown in Appendix. For each locality, the WBS row displays EMS98 results (WBS stays for whole building stock); we believe that this scale is the most robust tool to depict the effects of a present earthquake. The OTC row is the new set of data that represents the effects of the 2012 Emilia earthquake on old town centers only. In our opinion, the latter depicts the effects of an earthquake that struck the very same area of the Emilia seismic sequence, but as if it had occurred before the introduction of concrete in building. From this new dataset, we assessed new intensity values according to the MCS scale (hereinafter named IOTC, where OTC means old town center) and calculated new macroseismic magnitude.

Within this approach, the maximum intensity value 8 MCS was assigned to seven localities and coincides with I0.

EMS98 macroseismic intensities from Tertulliani et al. (2012a) (IWBS) and MCS intensities (IOTC) are plotted in Figs. 4 and 5, respectively. It has to be pointed out that we refer to the cumulative intensity due to the whole sequence effects.

Fig. 4
figure 4

Cumulative intensity distribution for the whole building stock (IWBS EMS98). The black star represents the macroseismic epicenter computed by Boxer 4.0 code (Gasperini et al. 2010)

Full size image
Fig. 5
figure 5

Cumulative intensity distribution for the old town centers (IOTC MCS). The black star represents the macroseismic epicenter computed by Boxer 4.0 code (Gasperini et al. 2010)

Full size image

The IOTC are on average higher than the IWBS. This trend can be highlighted in Fig. 6 where frequency of IWBS and IOTC intensities is reported for different intensity classes.

Fig. 6
figure 6

Frequency of IWBS and IOTC intensities for different intensity classes

Full size image

The IWBS distribution is centered on values lower than the IOTC ones. The mean value for the IWBS intensity distribution is 6.1, while the mean value for the IOTC distribution is 6.7.

The epicentral locations are very similar (Figs. 4 and 5). We computed the macroseismic magnitude either from the WBS and OTC datasets, obtaining, respectively, Mw 5.3 and Mw 5.7. The macroseismic magnitude from WBS dataset coincides with that reported in Table 1 computed from the whole original dataset. The result shows that MOTC (5.7) is considerably higher than the MWBS (5.3).

In the light of these results, we can affirm that, if the 2012 Emilia earthquake had occurred in the past (for example in the XIX century, that is to say a period antecedent to the use of concrete in building), we would have assessed intensities on average higher than those assessed for a present-day event. This kind of historical reading of the 2012 Emilia earthquakes shows results that are consistent with the current magnitude-intensity regression line (Pasolini et al. 2008; Gruppo di Lavoro 2004) based upon the 2004 version of the Italian parametric earthquake catalog (CPTI Working Group 2004). In other words, applying the MCS scale to old town centers only, the gap between assessed and expected intensities decreases.

3 Scarcity of data and overestimation of historical earthquakes: the 1570 Ferrara earthquake as an interpretation key

Historical and monumental buildings (i.e., fortresses, churches, bell towers, and historical palaces) are not considered truly representative of the macroseismic intensity affecting a given area or single settlement. It is not unusual that churches turn out to be the only affected buildings in a given locality struck by a moderate seismic event. In fact, this kind of buildings usually has its own particular vulnerability to ground shaking and shows a higher tendency to be damaged by earthquakes (Cattari et al. 2013). That is why modern macroseismic scales (i.e., EMS98, but also MCS) suggest caution against considering only such kind of buildings for the intensity assessment.

However, it is also true that historical/monumental buildings often have a particular importance (i.e., artistic, religious, cultural, or even political/administrative). So, they usually are under a sort of magnifying glass; consequently, they tend to receive particular attention from reporters, chroniclers, or observers. It is quite frequent that historical sources tend to mention effects on monumental buildings rather than effects on simple dwellings, on the base of the occasional extreme effects only. Indeed, to assess intensity through scarce data and descriptions referred to particular kind of buildings only is one of the main challenges of historical seismology (Musson 1998; Ferrari and Guidoboni 2000; Bakun et al. 2011; Hough 2013). Could this condition affect the earthquake scenario and lead to overemphasize the image of the earthquake impact?

To test this hypothesis, we first of all needed a historical earthquake to be used as key of interpretation, with intensity points estimated almost exclusively from damage to monumental buildings.

The 1570 Ferrara earthquake has the suitable characteristics, as its site and location resemble those of the 2012 event, and above all, its accounts are mainly related to effects on monumental heritage.

This event was carefully studied by Guidoboni et al. (2007). The 17 November 1570 main shock (Mw 5.4) started a complex seismic sequence that lasted until 1574. In the city of Ferrara, most of the buildings suffered severe structural damage and a few totally collapsed; the major damage was sustained by churches, bell towers, and monumental palaces.

In this study, Guidoboni et al. (2007) assessed the intensity of most localities starting from a poor documentation. In fact, excluding Ferrara which is well documented, 32 localities out of 35 have intensity equal or greater than 6 MCS and are documented through only one source very often referring to a single monumental building. This source is a transcription of a coeval account of the apostolic visit of Cardinal Maremonti (Maremonti 1574) and reports the damage produced by the earthquake on religious buildings.

On the base of these reports, we deduced the interpretative scheme used by Guidoboni et al. (2007) for the assessment of the intensity (Table 2). In Table 2, two examples of intensity assignment from Guidoboni et al. (2007) are given: In the first row, the assessed intensity is 6 MCS for a damage that can be mended; in the second row, the assessed intensity is 7–8 MCS for heavy damage to the roof and the belfry, in danger of collapsing.

Table 2 Ferrara earthquake of 17 November 1570: examples of the interpretation key used by Guidoboni et al. (2007) for the localities of Francolino and S. Maria Codifiume, and relative intensity assessments
Full size table

The following step is the attempt to apply the same scheme to the 2012 Emilia earthquake, as a further and more extreme historical reading, in order to see how the scenario of that earthquake changes. Therefore, we filtered the data from the EMS98 macroseismic survey (Tertulliani et al. 2012a) selecting information related to monumental buildings only (see the examples in Table 3).

Table 3 2012 Emilia earthquake sequence: two examples of the 1570 interpretation key application for the localities of San Giovanni in Persiceto and Bondanello and relative intensity assessments
Full size table

In synthesis, we suppose that if the 2012 Emilia earthquake had happened in XVI century, very likely the only information preserved would have been those about important buildings such as monumental and public buildings.

The results of this procedure show a general increase of the assessed intensities (Appendix), considering exclusively the damage on monumental buildings (hereinafter IMON).

For instance, in San Felice sul Panaro, where the IWBS assessed is 7 EMS98, the intensity based on monumental heritage is IMON = 9 MCS.

As regards the macroseismic magnitude, the solution inferred from the data (Gasperini et al. 2010) is MMON = 6.0.

In Fig. 7, the frequency of intensities of the three methods (WBS, OTC, MON) is reported for different intensity classes.

Fig. 7
figure 7

Frequency of IWBS, IOTC, and IMON intensities for different intensity classes; (white for WBS [whole building stock], pale gray for OTC [old town centres], dark gray for MON [monumental buildings])

Full size image

The white and pale gray bars are the same as in Fig. 6 while dark gray bars represent the frequency of intensities assessed on monumental damage for different classes. The mean value has considerably increased, and starting from 6.1 MCS for the WBS distribution, it has reached 7.3 MCS value for the IMON. Even in this case, the results of the historical reading, selecting only the damage on monumental buildings, are consistent with the (magnitude-intensity) regression relationships (Fig. 2).

The macroseismic epicenter, calculated by Boxer 4.0 code (Gasperini et al. 2012), differs in respect with those calculated for IWBS and IOTC datasets and moves eastward due to the intensity increase in the easternmost localities (Fig. 8).

Fig. 8
figure 8

Cumulative intensity (IMON) distribution. The black star represents the macroseismic epicenter computed by Boxer 4.0 code (Gasperini et al. 2010)

Full size image

4 Discussion and conclusions

The aim of the present work is to discuss the fact that the intensity values observed during the 2012 Emilia earthquakes were lower than we could expect for an Italian Mw 5.9 earthquake (see Pasolini et al. 2008; Gruppo di Lavoro 2004; Gasperini et al. 2010). As magnitude-intensity relationships are derived from seismic catalog records, mainly based on historical earthquakes, the suspect that some biases could have affected their intensity assignments leading to inflated assessments (Rong et al. 2011) arose. In fact, several recent studies carried out to revise historical events came to drop original intensity values (see for instance Hough et al. 2000; Tertulliani et al. 2012b; Hough 2013; Huysken and Fujita 2013). To tackle this issue, the effects of the 2012 Emilia earthquake were historically reread, as if the event had occurred in historical times. Such exercise can represent an advice on how interpretations of documentary sources can condition the assessment of a historical earthquake.

Starting from the damage survey (Tertulliani et al. 2012a) on the whole present-day building stock, developed during centuries, we have filtered the macroseismic information in order to obtain datasets representing different historical times. At first, we considered a time period before the use of reinforced concrete, when dwellings were characterized by a more homogeneous seismic vulnerability than nowadays, thus restricting the intensity assessment to old town centers. The results of this first step revealed that very likely the intensity of the 2012 Emilia earthquake would have been larger, with the maximum intensity from 7 to 8 MCS for many localities (Fig. 6). The second step is another leap in the past, when the records on damage on special monumental buildings were often the only information that could be retrieved in historical accounts about earthquake effects. In this case, the 2012 Emilia earthquake would have been classified as a strong earthquake with damage effects up to 9 MCS degree. Our journey back in time demonstrates that starting from intensity 7.5 EMS assessed on the present building stock (WBS), we have evaluated growing intensity values both for masonry building stock (XIX century-like; OTC) and monumental heritage (XVI century-like; MON). In Fig. 9, the three different datasets (WBS in white; OTC in pale gray; MON in dark gray) are simultaneously compared through a radial representation. Each radius corresponds to a different locality. It is evident that MON values are most of the times higher than the OTC ones, which in turn are larger than WBS values (the complete scheme of evaluation is in Appendix). This is particularly evident for localities such as San Felice sul Panaro, Mirandola, Finale Emilia, Camposanto, Mirabello, Stuffione, and some more. On the other hand, Cavezzo represents a very peculiar case, where intensity values from the three readings tend to coincide (8 MCS and EMS). This is mainly due to the fact that heavy damage in Cavezzo affected several modern reinforced concrete buildings as well as monumental and older ones, resulting in similar scenarios. Unlike Cavezzo, a few localities (i.e., Canaletto, and to a minor extent also S. Agata Bolognese) show MON and WBS intensity values both identical and slightly lower than the OTC one, because of the heavier damage affecting several masonry buildings in the old town center in comparison with damage observed in churches. In several cases, like Reggiolo, Concordia sulla Secchia, Novi, Moglia, Pegognaga, Gonzaga, MON and OTC values coincide, while WBS intensities are comparatively lower. Such cases occur in localities where damage both to monumental and masonry buildings (relatively old ones mainly in ancient town centres) is comparable, whereas outer modern expansion areas, mostly composed of reinforced concrete buildings, show much slighter damage levels.

Fig. 9
figure 9

A radial representation showing the comparison between the three different intensity datasets (WBS white, OTC pale gray, and MON dark gray)

Full size image

Apart from the peculiar cases of Cavezzo, Canaletto, and Sant’Agata Bolognese, we can highlight a sort of general trend: All the surveyed localities show the increase of the intensity values as we narrow down the reference damage scenario from the whole building stock to the old town center, and from the latter to the special monumental buildings.

It is evident that such variations of the intensity assessment for different building stocks result in a different calculated magnitude. As for the 2012 Emilia case, to evaluate the difference in magnitude, we computed macroseismic magnitudes by means of the code Boxer 4.0 (Gasperini et al. 2010) (Table 4). The difference of one degree in intensity entails a change in the macroseismic magnitude of 0.7.

Table 4 Macroseismic magnitudes calculated from WBS, OTC, and MON intensity data
Full size table

The exercise here presented, on only one case, opens a substantial issue and would deserve a larger-scale effort to systematically validate such preliminary results. Anyway, it is evident that the size of the discrepancies here shown could have a significant impact when considered for seismic hazard assessments (Ambraseys and Douglas 2004; Rong et al. 2011). This issue is therefore well framed in the growing and renewed interest in the interpretation of key historical earthquakes that might have been overestimated and play a crucial role for the calibration of intensity prediction relations.

In conclusion, we emphasize the need to define more homogeneous criteria in the interpretation of original accounts to reduce possible biases in the assessment of historical intensity.