On April 6, 2009, a Mw 6.2 earthquake hit the city of L’Aquila, one of the largest urban centers in the Abruzzo region (Central Italy) with about 70,000 inhabitants, causing 308 deaths and vast destruction in the town itself and surrounding areas. The earthquake represents one of the largest shocks in Italy since the beginning of the instrumental age, after the 1980 Mw 6.9 Irpinia, the 2016 Mw 6.5 Norcia, the 1976 Mw 6.4 Friuli earthquake. The earthquake was generated by the seismic rupture of the Paganica segment of Upper Aterno Valley - Paganica fault system consisting of four NW-SE trending normal fault segments (Galadini et al. 2018). The causative fault borders the eastern side of the Aterno River Valley, a geological depression of tectonic origin related to the extensional tectonic stresses acting in Central Italy, as depicted in Fig. 1. Note that the L’Aquila historical center is located on the surface projection of the causative fault inside the Aterno valley, at only about 2 km from the epicenter of the earthquake, suffering therefore from the largest ground shaking owing to the close proximity to the seismic source and to the site amplification effects of the valley sediments.
The post-earthquake macroseismic survey revealed a maximum intensity degree of IX-X in the Mercalli-Cancani-Sieberg (MCS) scale in the towns of Onna and Paganica, while other 14 towns and villages, including L’Aquila, reached an intensity degree between VIII and IX (Galli et al. 2009). The isoseismal lines of maximum intensity showed a pattern strongly elongated in the NW-SE direction, from the epicenter toward the southern part of the Aterno Valley. This pattern reflects the geometry and the dynamics of the fault rupture, with prevailing directivity effects in the direction SE of the epicenter (see discussion in Section 3).
After the earthquake, a comprehensive survey campaign, coordinated by the Italian Department of Civil Protection (DPC), was carried out with the aim of collecting damage and usability information on the affected building stock. This information was then conveyed in a damage database (Rosti et al. 2018; 2020a; 2021a, b; 2022), now available in the Observed Damage Database – Da.D.O. (Dolce et al. 2019). This is a web-GIS tool, operated by the Italian DPC, collecting and archiving data on the construction and structural characteristics, as well as on seismic damage, of ordinary buildings inspected after seismic crises of national importance, from 1976 onwards.
The final L’Aquila database, including only data regarding residential buildings in municipalities which were completely surveyed (with completeness ratio larger than 90%, see discussion in Rosti et al. 2021a, b), consists of about 37000 buildings, which are shown in Fig. 1. For each inspected building, the database collects information on the geographical coordinates, the construction material (masonry or RC), the typological building classification according to Table 1, the number of storeys (see height classification in Table 2), the construction age, the Damage State (DS) according to the European Macroseismic Scale EMS (Grünthal et al. 1998). Furthermore, as clarified in the following section, the database has been enriched by associating to each damaged building a value of ground motion IM coming from both the ShakeMap and the PBS computations. Referring to Rosti et al. (2018) for further details on the criteria for the attribution of the damage grade, it is important to underline herein that a global damage level is assigned to each building based on the maximum damage observed on the most damaged building component.
For the objective of this study, which is to compare the effect of different approaches for the characterization of ground shaking in the empirical fragility analysis, a subset of the damage dataset, corresponding to the detailed study area in Fig. 1 encompassing the L’Aquila municipality, is considered. An area of limited extent is studied mainly because of the computational limits associated with the generation of simulated broadband time histories over a dense grid of sites covering the entire area of the SPEED model. Furthermore, the detailed study area covers the region of larger engineering interest, because of the high density of severely damaged buildings. The considered dataset, hereafter referred to as detailed dataset, counts 7987 residential buildings, 4564 (57%) of which refer to masonry, whereas 3423 (43%) are RC buildings (Fig. 2a). Note that in Table 1 the number of buildings belonging to each typological class is specified. Figure 2 shows the spatial distribution and percentage of the different damage levels, from 0 (no damage) to 5 (collapse), for both masonry (b) and RC buildings (c).
Figure 3 shows the frequency of buildings in each damage state according to the different building typological and height classes: buildings are grouped into four main classes - a) Irregular layout or poor-quality masonry, with flexible or rigid diaphragms, w/wo connecting devices (e.g. tie-rods and/or tie-beams); b) Regular layout and good-quality masonry, with flexible or rigid diaphragms, w/wo connecting devices; c) RC seismically designed, pre 1981; d) RC seismically designed, post 1981 – and the corresponding height classes are considered to compute the statistics. The histograms indicate, as expected, that passing from more vulnerable classes, such as irregular masonry buildings, to less vulnerable classes (RC post 1981), there is a decrease of the percentage of higher damage states (DS4-DS5). Furthermore, there is a positive correlation between the percentage of DS4-DS5 with the height of the building, for any building class.
In the territories more distant from the earthquake source, post-earthquake surveys tend to be limited to damaged buildings, as they are carried out only upon the owners’ request. As a consequence, resulting damage distributions could be biased by an underestimated number of undamaged buildings. To counteract this issue and to suitably account for the negative evidence of damage in sites less affected by the ground shaking, the post-earthquake dataset was integrated by 197’528 non-inspected buildings, located in the non-surveyed and partially-surveyed (with completeness ratio lower than 10%) municipalities, which were reasonably assumed undamaged (e.g. Rosti et al. 2022). The number of undamaged buildings was retrieved from national building census (ISTAT 2001). Allocation of undamaged RC buildings to predefined building types was immediate, as national building census classifies buildings based on construction material, construction age and number of stories. Differently, mapping of undamaged masonry buildings, identified by census parameters, to predefined building typologies was performed via the empirical correlation between census and typological building attributes reported in Rosti et al. (2022).
Table 1
Typological building classification according to Rosti et al. (2021a;b). In the first column, the percentage of buildings belonging to each class within the detailed study area are reported into brackets.
Building
typology
|
Construction
Material
|
Masonry
type
|
Intermediate
diaphragm
|
Connecting
Devices?
|
Design
Level
|
IRR/F/NCD (18.5%)
|
Masonry
|
Irregular layout or poor-quality (IRR)
|
Flexible (F)
|
No (NCD)
|
-
|
IRR/F/CD (6.8%)
|
Yes (CD)
|
-
|
IRR/R/NCD (2.5%)
|
Rigid (R)
|
No
|
-
|
IRR/R/CD (3.8%)
|
Yes
|
-
|
REG/F/NCD (4.6%)
|
Regular layout and good-quality (REG)
|
Flexible (F)
|
No
|
-
|
REG/F/CD (3.1%)
|
Yes
|
-
|
REG/R/NCD (3.2%)
|
Rigid (R)
|
No
|
-
|
REG/R/CD (14.7%)
|
Yes
|
-
|
RC/Seismic-Pre81 (17.7%)
|
RC
|
-
|
-
|
-
|
Seismic-Pre 1981
|
RC/Seismic-Post81 (25.1%)
|
-
|
-
|
Seismic-Post 1981
|
Table 2
Building height classes for masonry and RC buildings.
Construction material
|
Height class
|
No. storeys
|
Masonry
|
L
|
1–2
|
MH
|
>2
|
RC
|
L
|
1–2
|
M
|
3–4
|
H
|
>4
|