The study area, struck by Mw 5.1 Montecilfone earthquake, is about 1000 km2 and is located in the NE-side of the Molise region in a portion between Frentani Mts. and Adriatic coast, representing the external part of the Apennine chain.
This region is characterised by a NE-verging thrust and fold belt given by the stacking of the most external tectonic units of the southern sector of the chain, locally covered by thrust-top deposits and a late Miocene-Pliocene terrigenous succession29-31. This geological setting derived from the early Miocene to Pleistocene uplift of the chain, which led to the current regional morpho-structural setting30.
More in particular, the sector of the Apennines under investigation is the result of three important tectonic phases32, which involved both surficial and deeper tectonic units deriving from the deformation of the buried Apulian domain: i) compressive phase (Low. Miocene - Up. Pliocene), characterised by the development of E-vergent surficial thrusts; ii) transcurrent phase (Up. Pliocene - Low. Pleistocene), characterised by the presence of N-S and E-W faulting; iii) extensional phase (Mid. Pleistocene - Holocene), characterised by direct SW-NE and NW-SE fault systems, still active in the study area, as suggested by seismicity and landslide activity.
The study area presents a thick marine sequence of chain domain (Cenozoic), belonging to the Lagonegrese-Molisan Basin29,30; it is mainly composed of scaly clays followed by arenaceous-marly and calcareous-marly flysch deposits29, evaporites and chaotic terrains29,30 while thick silico-clastic sequences (Plio-Pleistocene) outcrops in foredeep. Since the middle Pleistocene, the study area was affected by moderate tectonic uplift phenomena33,34 resulting in the formation of regressive sequences and terraced marine deposits, outcropping onto hilly reliefs of Guglionesi, Petacciato and San Giacomo degli Schiavoni villages35. The successions of both domains (i.e., chain and foredeep) are covered by continental, marine and transitional deposits (Quaternary) that widely outcrop in alluvial areas and close to the Adriatic coast, where are thick up to tens of meters29,36,37.
The geotechnical units outcropping in the study area can be described as follows29 (Fig. 1):
- Quaternary deposits: composed by i) landslides and ii) terraced alluvial deposits of both continental and marine environment, near the river network and close to the Adriatic coast respectively;
- Foredeep domain: composed by i) Blue Clays which pass upwards to sands and conglomerates; ii) marine conglomerate-sandy deposits (Plio-Pleistocene);
- Chain domain: composed by i) Pelitic Flysch belonging to Vallone Ferrato Fm. (Messinian - Tortonian), mainly composed by marls and clayey marls interlayered with sandstones and siltstones; ii) Arenaceous Flysch belonging to Faeto Fm. (Tortonian-Serravallian); iii) Scaly Clays (Low. Miocene - Up. Cretaceous), tectonic melange consisting of clays and marly clays, with typical coloring from green to reddish, in chaotic structure and with micritic intercalations;
- Top-thrust basin transitional domain: composed by i) Calcarenites belonging to Palombaro, Casalanguida and Larino Fm. (Mid. Pliocene - Low. Pliocene), ii) Marly Clays (Mid. Pliocene - Low. Pliocene).
The outcropping lithologies justify a hilly relief with elevations up to 650 m a.s.l. and river valleys of variable extension. The central sector of the study area belongs to the lower basin of the Biferno river, while northern and southern sectors bound the basins of the Trigno and Fortore rivers.
The geotechnical features of the outcropping units make the Molise region one of the most landslide-prone areas in central and Southern Italy38. In fact, it is typified by the presence of more than 28000 landslides of different size and type of movement that cover an area of about 4461 km2 39-42. In the epicentral area of the 2018 earthquake, about 7100 landslides were recognised over time, with an average size ranging from 100 m2 to 3 km2. Regarding the state of activity, dormant landslides are about 72%, while active or stabilised ones are about 24% and 4%, respectively. These landslides show low to very low velocity and can be classified as flows, slides or complex, according to Hungr et al.43. Moreover, very slow viscous deformations of the soil cover, such as soil creep and solifluction, can affect the gently dipping slopes. Landslides are often driven by the presence of tectonic alignments or by fluvial networks and by surface erosion of run-off waters. Gravity-induced deformations mainly involve quaternary deposits and the more shallow and weathered portions of the bedrock. Deep phenomena involve the thick sequences of Plio-Pleistocene clays, especially in coastal areas44-46.
Seismic and rainfall events on August 16th, 2018
The study area was struck by a Mw 5.1 earthquake occurred on August 16th, 2018 at 18:19:04 (UTC) with epicentre in Montecilfone (CB - Molise Region - Central Italy) (LAT 41.87, LONG 14.86, hypocentral depth 20 km), represented the main shock of a longer seismic sequence, started on 25 April 2018 and ended on 4 September 2018, and was followed by almost 840 low-magnitude earthquakes47 (Mw < 2.0).
The 2018 Montecilfone mainshock resulted in a low Peak Ground Acceleration (PGA) up to 0.12 g (http://shakemap.rm.ingv.it). Macroseismic surveying was achieved by the INGV QUEST Group48. In general, despite a low-damage scenario on buildings in Montecilfone, Acquaviva Collecroce and Castelmauro municipalities (grade V-VI of EMS98 scale), a higher intensity level (up to VII) can be attributed if earthquake-induced ground effects are considered in accordance with the Environmental Seismic Intensity (ESI) scale49. The distribution of the epicentres appears to be slightly elongated around E-W, similarly to the one caused by the 2002 San Giuliano di Puglia seismic sequence, located about 20 km south far from the Montecilfone epicentral area. The focal mechanism of the 2018 mainshock indicates a dextral strike-slip rupture47, which is consistent with the one elaborated for the 2002 earthquake and attributed to the Individual Seismogenic Source coded as ITIS052 (San Giuliano di Puglia) and to the Composite Seismogenic Source coded as ITCS003 (Ripabottoni-San Severo) by the Database of Individual Seismogenic Sources v. 3.2.1 (DISS v. 3.2.1- http://diss.rm.ingv.it/diss/). Valensise et al.50 added more information for the 2002 San Giuliano di Puglia earthquake, considering it due to a composite rupture of a 20 km-long segment of an important E-W trending pre-existing sub-vertical fault. This event is located on the westward prolongation of the Mattinata fault zone, about 15 km west of the presumed western end of the Mw 6.7 rupture of 30 July 1627 (corresponding to the ITIS054: San Severo of the DISS; Terremoto della Capitanata), 30 km east from the Apennines axis. Valensise et al.50 suggest also that the entire system is active, but that it is segmented up into 10-25 km-long portions. Given the relatively low magnitude of the event, the seismogenic source will not be included in the DISS catalogue. Notwithstanding, the 2002 and 2018 Molise seismic sequences highlighted the occurrence of active faulting in this area, where relevant earthquakes have not been previously recorded. This zone is located about 50 km east of the extensional Appenninic belt and about 50 km west of the dextral strike-slip Mattinata fault51.
During the three days before the seismic event, the epicentral area also experienced intense rainfall ranging from 120 mm to 150 mm, as cumulative value, recorded by the closest rain gages to the epicentre (Ponte Liscione and Palata weather stations, respectively). To avoid the action of rain alone as the main triggering factor of the surveyed landslides, a statistical analysis of maximum daily and hourly rainfall intensity data was performed by Martino et al.28 testifying that the August 2018 rainfall event cannot be classified as exceptional with respect to other rainfall events occurred in the same area during the last decades.
The EqTLs resulted from the coupled action of simultaneous preparatory and triggering factors, characterised by different hazards. The lower one is referred to as the seismic trigger, which has a return period ranging from 150 to 200 years. The higher one can be attributed to rainfall triggers, characterised by maximum return periods of 17 and 13 years if a time window of 1 hour and 3 days have been considered, respectively. It follows that the contribution of the earthquake, although prepared by rainfalls, appears to be a necessary condition to justify the landslide scenario surveyed immediately after the August 2018 mainshock, giving it the meaning of a co-seismic scenario.
Landslide scenario induced by the August 16th, 2018 earthquake
In the days immediately after the 2018 Montecilfone earthquake, a detailed field survey of earthquake-induced ground effects was performed by Sapienza fast-response team28, inventorying 84 Earthquake-Triggered Landslides (EqTLs)52 and 4 co-seismic ground cracks.
The observed effects were surveyed and geo-localised in-field following the standardised procedure exposed in Martino et al.53, focused on the distinction of clear clues for landslides activation (or reactivation). In this way, all surveyed landslides were compared with available landslides catalogues (i.e., IFFI) to assess the number of landslide reactivations with respect to first-time occurrences54 in co-seismic time, revealing that 37.5% of EqTLs was a reactivation of pre-existing landslides. The prevalent landslide mechanisms consisted in earth slides and earth flows, which involved slopes with an inclination ranging from 10° to 15° according to lithological units outcropping in the study area. In particular, the landslides mainly affected surficial soil covers in gentle slopes made up of clays and marly clays (60 %) and, secondarily, flysch (27 %) and alluvial or debris deposits (13 %).
The surveyed effects were inventoried according to the criteria established for the Italian Catalogue of Earthquake-Induced Ground Failure database55 (CEDIT) that, at same time, did not document evidence of past earthquake-induced ground effects in the study area. Only a few landslides and ground cracks are referred to historical 1805 (Sant’Anna), 1627 (Gargano), 1894 (Lesina), and 1980 (Irpinia) earthquakes and took place in the surrounding area.
The analysis of frequency distributions of total landslides and landscape with respect to topographic metrics (i.e., elevation and slope) or lithology provides a summary of the susceptibility to landslide of the area56. The frequency distribution of the landslides available in the catalogue shows a peaked unimodal right-skewed (positive Kurtosis and Skewness) distribution for both elevation and slope, with mode values equal to 200-250 m a.s.l. and 20°.
EqTLs showed a unimodal distribution with peaks centred to lower values of 150-200 m a.s.l. and gradients of 15°. It remarks the importance of seismic input in the controlling landslide occurrence (i.e., the observed scenario) since they occur even at slighter slopes and lower elevation (consistent with ones that characterise the epicentral area).
Based on the first analysis reported in Martino et al.28, the surveyed EqTLs also show a slight NE-SW directivity in their spatial distribution in accordance with the main morpho-structural alignments, such as the fluvial and drainage networks, which represents counter prints of morpho structural setting.
This directivity seems to be more pronounced for first-time landslides with respect to the reactivations28. However, the distribution of EqTLs does not provide to date evidence supporting the link between the directivity and the strike-slip fault mechanism. Moreover, regarding the distribution with respect to the epicentral distance, a typical decreasing frequency of EqTLs was observed with increasing distance, with an almost symmetrical distribution, especially parallel to the strike of the reliable fault plane (Fig. 1).
Maximum distances of 7 km for disrupted landslides (i.e., mainly represented by rock falls) and of 18 km for coherent landslides (i.e., earth landslides) were surveyed. Based on these epicentral distances, given the magnitude of the earthquake, all the disrupted landslides occurred below the maximum expected distance provided by Keefer6 on world-based statistics and by Martino et al.57 for the Italian territory. On the contrary, 43 out of 75 coherent landslides took place beyond the maximum expected distance, placed about 3.5 km from epicentre. As inferred by Martino et al.28, the occurrence of coherent landslides beyond the established thresholds can be attributed to the coupled and quasi-simultaneous action of preparatory factors (i.e., rainfall) and triggering factors (i.e., earthquake). The unexpected distance of occurrence Keefer affected area, based on the registered magnitude of the event, allows delineation of the perturbed area, where possible evidence of increased post-seismic activities can be assumed.
Despite the low magnitude of the seismic event, the combination of preparatory and trigger factors (i.e., rainfall and seismic shaking) resulted in an increased impact on the territory, as testified by the surveyed scenario28. These destabilising factors act on mainly clayey (both Blue and Scaly Clays belonging to the former pelagic Molisian Basin58) and flysch lithologies, characterised by very low friction angles44,59 (lower than 22°) resulting in gentle slopes, affected a-priori by abundant landslides. The latter show reactivations every year involving shallow soils and eluvium-colluvium covers (Fig. 2). The illustrated landscape, devoid of vegetational or anthropic disturbance elements and featured by a large abundance of landslides, has already been revealed to be proper for investigating landslide triggering using DInSAR analysis60. For these reasons, evidence of reactivations or first-time occurrence of landslides were monitored over a medium term consisting of two years before the seismic event and the first year after the mainshock, to assess how even relatively weak seismic events can affect, in predisposed areas, the spatial and temporal distribution of landsliding.
In the following, evidence or proxies of increased activity in the have been recorded and discussed in the frame of seismotectonic, geo-lithological and morpho-structural setting of the area.