The seismic microzonation study of Pescara del Tronto (Central Italy) during and after the Central Italy earthquake sequence.

Abstract

Site effects in consequence of the 2016–2018 Central Italy seismic sequence “marked the fate” of many villages located close to the epicenters along the Apennine chain. Pescara del Tronto, a small settlement located on a mountain slope in the municipality of Arquata del Tronto, (AP) is very representative of such territories since, early after the onset of the seismic sequence, suffered very impressive site effects, including a large seismically-induced debris flow and a high number of small volume landslides. This paper describes the main results of the scientific activities carried out by the authors in this locality in the framework of the unprecedented large-scale Seismic Microzonation project funded by Italian Government in Central Italy for 138 territories. Through an intense field activity, the geological and geomorphologic setting of the study area were revised and updated, including a landslide susceptibility assessment that helped the Italian Department of Civil Protection in the emergency management for temporary housing and, later on, was included in the 3rd Level Seismic Microzonation study implemented by authors while the seismic sequence was still ongoing. A very detailed and reliable subsoil model for this municipality was defined, despite difficulties faced in performing direct and indirect investigations due to the safety restrictions for many areas. The Local Seismic Site response was finally assessed for this locality and the results discussed. A key role in the occurrence of strong site effects into Pescara del Tronto has been played by quaternary deposits having an unexpected heterogeneity under the old village.

Appendix 1: Landslide susceptibility analysis

The study started with the collection (and the critical review) of pre-existing data from national databases, repositories and archives, supported by the multiscale and multi-temporal (1954–1955, 1983 and 1988–1989 years) analysis of aerial photographs. Copernicus satellite images (https://emergency.copernicus.eu/mapping/list-of-components/EMSR177) and aerial orthophotos acquired early after the 24th August event (courtesy of DPC) were also examined. The Digital Terrain Model (3x3 m resolution) was generated starting from LiDAR data (1x1 m resolution) integrated with the already available official topographic map of the Marche region (1:2,000 scale) for some outer areas. The final DTM was used to obtain two intermediate products used in the study, that is the ground surface’s steepness map and the cardinal exposition map of the study area.

The two wide IFFI areas (see Fig. 2 in the paper) were carefully surveyed during field activity. Their geomorphologic setting was completely revised according to the collected data (i.e. field surveys, geotechnical and geophysical investigations) and only recent debris flow deposits were identified and mapped (a_1a unit, see Fig. 2 in the paper). To numerically assess the susceptibility to a deep-seated landslides the intensity scale proposed by Fell (1994) was used, considering the phenomena belonging to intensity classes equal or greater than 5.

The slope stability for a generalized collapse that could affect not only the debris layer but the rocky substrate too was numerically evaluated along the 2D geotechnical sections A–A′, B–B, C–C′, considering static and pseudo-static conditions. The Limit Equilibrium (LE) method was implemented first by using the SLIDE software (Rocscience), adopting the Mohr-Coulomb failure criterion to model the soil behavior and considering the mechanical characterization of the geological units as reported in Table 1.

Several methods were considered, particularly those proposed by Janbu (1973), Morgenstern and Price (1965) and Sarma (1979). We obtained factors of safety (FOS) always higher than 4 along all sections and regardless to the considered method, confirming what was expected according to the actual (i.e. revised) geological and geotechnical setting of the slope, i.e. that a deep-seated landslide cannot be hypothesized due to the absence of geometric and kinematic conditions for triggering. Results obtained by considering the Morgenstern and Price (1965) method are shown in Fig. 17 and only for the A–A′ section, which is characterized by the most complex subsoil setting. Excepting for the cover deposits, the geological and geomorphological settings of the geologic bedrock in the A–A′ cross-section are representative of the whole slope and the pseudo-static results in terms of bedrock shear strain and deformation from this 2D analysis may be extrapolated to the rest of the slope, including those landslide areas identified in the IFFI Catalog (see main text). Readers can refer to Fig. 12 in the paper for the detailed description of the geological units.

Fig. 17

Results of the LE analysis obtained for the A–A′ cross section in terms of FOS

The slope stability assessment for pseudo-static conditions was carried out by using the RS2 software (Rocscience) that allowed to perform analyses via finite elements method (FEM). The shear strength reduction (SSR) approach proposed by Dawson et al. (1999) was adopted, considering the same mechanical characterization as for the LE assessment. The pseudo-static condition was considered via the horizontal seismic coefficient value K_h (see Eurocode8 2004) equal to 0.11: this value represents the seismic action for a return period of 475 years to be used for the analyses according to NTC18 (2018). Results are shown again only for the A–A′ section, in terms of maximum shear strain (Fig. 18) and in terms of total displacement (Fig. 19). The maximum shear strain values focused in correspondence of the old village within the anthropogenic deposits (h) and the weakly cemented travertine unit (f1b), interesting also the base of the lithoid travertine body that, following the October 30th Mw 6.5 event, collapsed into huge blocks on the SS4-Salaria road. Accordingly, huge deformations were estimated for almost the whole Pescara del Tronto settlement (Fig. 19), whereas it can be observed along the upper portion of the slope that displacement inside the quaternary covers are not very significant and can occur only within the first 2–3 m in depth. In terms of displacement, it is observed that the geologic bedrock is not significantly affected. As the occurrence of a deep-seated landslide is conditioned on the possibility of a deep sliding surface within the bedrock, which is not compatible with the results from pseudo-static analysis, the possibility of deep-seated landslides was no longer considered.

Fig. 18

Maximum shear strain values distribution along the A–A′ cross section, considering pseudo-static conditions

Fig. 19

Total displacement values distribution along the A–A′ cross section, considering pseudo-static conditions

The stability analyses for shallow landslides, mainly consisting of slides affecting debris deposits and/or completely fractured and weathered rocks, were carried out using the SINMAP (Stability INdex MAPping) GIS tool introduced by Pack et al. (2005), based on the “infinite slope” stability analysis. For such type of landslides, a maximum thickness of 2 meters has been hypothesized, based on the observed earthquake-induced phenomena and ancient debris flow deposits. The assessment was performed through the calculation of the ratio (SI—Stability Index) between the component parallel to the slope of the gravity force and the cohesion and friction forces that oppose the movement of ground masses along a sliding plane parallel to the slope. Input parameters are the friction angle, the bulk natural weight, the cohesion and thickness, assuming the Mohr Coulomb failure criterion. The resulting classification in terms of SI values for the study area is reported in Fig. 20a. The final raster layer was characterized by SI values ranging from 0 to >1.5, categorized in 4 classes (SI <0.5, 0.5–1.0, 1.0–1.5, > 1.5).

Fig. 20

Stability Index map for shallow landslides through SINMAP GIS tool analysis (a); source areas for debris flow and related (reclassified) hazard assessment according to the balance of volumes approach (b) and to simplified “mass point” approach (c). The channels affected by debris flows phenomena are numbered 1 to 5

The debris flows triggering susceptibility evaluation was carried out through the calculation of the proneness to generate the phenomenon and an estimate of the possible magnitude; field survey data have been used to identify the possible source areas as main input, integrated by aerial photographs analysis and morphological indexes calculation, according to 2 different simplified approaches:

• (a) balance of volumes, as proposed by Iverson et al. (1998) and implemented as GIS tool by Berti and Simoni (2014)

• (b) simplified “mass point” approach proposed by Voellmy (1955).

The prediction of the debris flow inundation areas, according to the balance of volumes approach is depicted in Fig. 20b. The final raster layer was re-classified by using boolean values (1- flooded areas; 0- non flooded area). The results of the simplified “mass point” approach are reported in Fig. 20c in which the hazard was quantified in terms of kinetic energy. The 5 classes shown were subsequently categorized into a new raster layer containing Boolean 1 and 0 (presence, absence) values.

The Flow-R model developed by Horton et al. (2013) was used to define areas susceptible to rock-fall events. Flow-R is a semi-empirical numerical model which exploit algorithms based on simplified hypotheses that, completely neglecting the volume of the event, allows a good estimate of the hazard. It has been applied with good results to different case studies applied to rock-falls and avalanches (e.g. Van Westen et al. 2014; Quan Luna et al. 2011). Possible source areas for such type of landslides were identified mainly using field observations, integrated with information about the steepness of the slope (i.e. values greater than 45° were always considered). The multiple flow direction algorithm (Quinn et al., 1991) was considered here and the friction laws were used in order to determine the runout distances. The friction loss was assessed according to the two parameters friction model proposed by Perla (1980), developed for avalanches. Friction parameter values and mass-to-drag ratio were fixed after a calibration step based on the observed (earthquake-induced) phenomena. Results were expressed in terms of energy map (by considering the maximum value of overlapping propagations) and reclassified into Boolean values. In this case too all areas interested by the modeled phenomena were categorized into 1 and 0 (presence, absence) values. These areas (not shown) represent a sub-set of those already classified with Boolean 1 through the simplified “mass point” approach.

All the information was combined and a raster analysis was carried out through the overlap of the maps related to the different types of instability to define the final landslide susceptibility map. Values contained into raster layers derived from each step were combined through a matrix, generating a final value for each Terrain Units (TU—3x3 m pixel) according to the following precautionary rules:

• The very high susceptibility class (4) contains the TU where at least 2 types of hazard exist (value 1 of the indexes) or one value is 1 and the SI (Stability Index) for shallow landslides is lower than 0.5.

• The high susceptibility class (3) contains the TU where debris flow phenomena of whatever intensity exist, or SI is lower than 0.5 or where the MOPS map (see main paper) indicates the existence of areas of attention due to instability.

• The medium susceptibility class (2) contains the TU where no previous types of failure exist (indexes = 0), but the SI for shallow landslides is between 0.5 and 1.0.

• The low susceptibility class (1) is the same as the previous one, but the SI for shallow landslides is between 1.0 and 1.5.

• The very low susceptibility class (0) is the same as the previous one, but the SI for shallow landslides is higher than 1.5.