Structural Complexity and Seismogenesis: The Role of the Transpressive Structures in the 1976 Friuli Earthquakes (Eastern Southern Alps, NE Italy)

: We reconstructed the seismotectonic setting of the area comprising the northeastern Friuli Plain and the Julian pre-Alpine border (NE Italy) by integrating geological and seismological data. The study area represents the junction between the SSE-verging polyphase thrust-front of the south-Alpine Chain and the NW–SE-trending strike-slip faults of the eastern Friuli–western Slovenia domain. Following a multidisciplinary approach, the 3D geometry of the Susans–Tricesimo thrust system was reconstructed through the elaboration of four geological cross sections derived from the interpretation of ENI industrial seismic lines. In a second step, the seismogenic volume of the central-eastern Friuli area was investigated through hypocentral distribution analysis: the seismic events of the latest 50 years (1976–1977 and 1978–2019 time intervals) were plotted on four NE-SW-oriented seriated sections together with the fault plane’s geometry. Through this procedure, we were able to investigate the relationship between the NW-SE-striking high-angle faults, which characterize the northern Julian pre-Alps, and the WSW-verging medium-angle reverse fronts located at the piedmont of the Friuli plain, which experienced NW-SE- to NNW-SSE-oriented compression starting at least from the Pliocene. In detail, we examined the involvement of these structures during the seismic sequences of May and September 1976, in terms of activation and/or interaction. The resulting seismotectonic model highlights the interplay between transpressive/strike-slip and reverse planes. In particular, this study suggests that Predjama and Maniaglia transpressive faults strongly control the stress release and likely played a fundamental role both during the 6 May (Mw 6.5) and 15 September (Mw 6.0) Friuli earthquakes.


Introduction
Northeastern Italy is one of the areas of central Mediterranean with a high seismic hazard, and the Friuli region is one of the most seismic areas in Italy: in historical times it was hit by several I 0 ≥ 9 earthquakes (1348,1511,1873,1928,1936) [1]. The last devastating seismic sequence occurred in 1976 (6 May, Mw 6.5 and 15 September, Mw 6.0), causing about 1000 fatalities and the destruction of many towns. In particular, the seismicity of the Friuli region is mostly concentrated along the Carnian and Julian Alpine and pre-Alpine arcuate belt (Figure 1). At present, this area undergoes an NNW-SSE-oriented sigma 1 [2], and the accumulating stress is released through the interaction between the SW-NE-to WNW-ESE-trending compressional fronts of the eastern south-Alpine belt in Friuli [3] and the Dinaric NW-SE strike-slip fault systems of western Slovenia [4][5][6][7].
The study area has been affected by a polyphase tectonic history: the Mesozoic extensional event was initially followed by the Upper Cretaceous-middle Eocene (External  [9]. Historical seismicity from CPTI15 v4.0 [1] (red stars) and from CFTI5 MED [10] (green stars). Instrumental seismicity: 1976/05/06 Mw 6.5 (yellow circle) from [11]; 1976-1977 from [12] and 1978-2019 from [13].   [9]. Historical seismicity from CPTI15 v4.0 [1] (red stars) and from CFTI5 MED [10] (green stars). Instrumental seismicity: 1976/05/06 Mw 6.5 (yellow circle) from [11]; 1976-1977 from [12] and 1978-2019 from [13]. been documented within the Julian pre-Alpine area since the year 1000: the 1511 and the 1976 earthquakes. Regarding the 1511 event, on 26 March a strong earthquake (Mw 6.3) [1] affected a wide area extending from Friuli to western Slovenia. Based on the updated macroseismic field elaborated by [41], the revised epicenter has been located near Tarcento, where Io 9 MSC was estimated ( Figure 1). On the contrary, CFTI5 MED [10] reports two strong events during March and August 1511. The macroseismic epicenter of the first 1511 earthquake (equivalent magnitude Meq 7) is located near Montefosca, in the Natisone valley, while a subsequent event (Meq 6) is documented on 8 August 1511 near Banjšice, in western Slovenia (Figure 1). Concerning the possible source responsible for the 1511 events, it is worth noting that the macroseismic epicenters in both March and August from CFTI5 MED are located on the Predjama strike-slip fault. Based on macroseismic data inversion, one study [42] proposed an earthquake modeling scenario consistent with an Mw 6.9 event which activated the 50 km-long SE portion of the Idrija strike-slip fault. Differently, as already suggested by [43], one study [41] concluded that the 1511 earthquake involved both Alpine and Dinaric structures.
Recent paleoseismological analyses on the NW-SE strike slip fault system of eastern Italy and western Slovenia revealed Holocene surface rupturing referable to the 1511 event across the Colle Villano-Borgo Faris-Cividale fault system [30], across the Idrija [38,44] and Predjama faults [38].
The more recent 1976 Friuli sequences have been accurately reviewed [11] based on a large amount of bibliography. Preceded by an Ml 4.5 foreshock, a destructive earthquake (on 6 May 1976, Ml 6.4) struck central-eastern Friuli, where Imax 9-10 MSC was recorded. After the strong Ml 5.3 aftershock on 9 May 1976, seismic activity slowly decreased until September 1976. Two quakes on 11 September 1976 (Ml 5.1 and 5.6, respectively) and two others on 15 September 1976 (Ml 5.8 and 6.1, respectively) caused further damage. One year later, on 16 September 1977, another seismic event with Ml 5.2 occurred and was followed by an aftershock series which lasted more than one month. The only data interpreted as coseismic deformation [45] come from topographic leveling carried out in 1952 and 1977 [46], and triangulation measurements [47]. Regarding the Mw 6.5 mainshock, the macroseismic epicenter is located between Buia and Gemona del Friuli (Figure 1, [1]), while the instrumental location of the event [11] is placed near Lusevera, about 12 km ENE, in a sector of the Julian pre-Alps where no major settlements are found. Regarding the focal mechanism of the mainshock of May, most of the available interpretations agree with the activation of a reverse N-dipping low-angle plane [9], with only a small oblique component. Differently, concerning the focal mechanism solution of the strongest event of 15 September (Ml 6.1), different interpretations are present in the literature, which mainly propose the activation of a reverse N-dipping plane or of an NNE-dipping reverse-oblique geometry [9].
Regarding the possible source, most of the interpretations agree with the activation of two different structures (see fault traces in Figure 2). One study [48] proposed the activation of the Buja-Tricesimo thrust system for the May event and the involvement of the Periadriatic thrust (or Barcis-Staro Stelo Auct.) during September. One study [49], on the basis of hypocenter relocation, long-period surface-wave inversion, field geology and strong motion modelling, revisited the 1976 Friuli earthquake sequence and proposed that the Friuli earthquake rupture was related to a 19 km fault-related folding evolving from blind faulting beneath the Bernadia and Buia basement-involved structures to semi-blind faulting beneath the Susans structure. On the basis of geological and seismological evidence, others [3,50] have proposed the activation of the Susans-Tricesimo thrust for the May event, and assessed the activation of a deeper blind structure during the event of September. The activation of a single source for the entire sequence, whose surficial expression is compatible with the Buia-Tricesimo thrust, was proposed in [37] based on geodetic data. Conversely, another study [51] attributed most of the events to the Periadriatic overthrust and only two aftershocks of May to the inherited Dinaric structures. Recently, one study [29] confirmed the Susans-Tricesimo thrust as the source of the May event, based on the evidence of Quaternary deformation both on the surface and in seismic ENI profiles, and the possible activation of the Buia thrust was not excluded. Following the 1976-1977 sequence, the instrumental seismicity recorded two other more recent seismic events in 1998 and 2004 ( Figure 1). The April 1998 sequence (Bovec earthquake) had an Mw 5.6 [52][53][54], while an Mw 5.2 event was registered in July 2004 (Tolmin earthquake) [55]. Both sequences are associated with the NW-SE dextral strike-slip Ravne fault, which is actively propagating through interactions of individual right-stepping fault segments and breaching of local transtensional step-over zones [33,53].
Geosciences 2022, 12, x FOR PEER REVIEW 5 of 33 event, based on the evidence of Quaternary deformation both on the surface and in seismic ENI profiles, and the possible activation of the Buia thrust was not excluded. Following the 1976-1977 sequence, the instrumental seismicity recorded two other more recent seismic events in 1998 and 2004 ( Figure 1). The April 1998 sequence (Bovec earthquake) had an Mw 5.6 [52][53][54], while an Mw 5.2 event was registered in July 2004 (Tolmin earthquake) [55]. Both sequences are associated with the NW-SE dextral strike-slip Ravne fault, which is actively propagating through interactions of individual right-stepping fault segments and breaching of local transtensional step-over zones [33,53].

Structural Model Reconstruction
The interpretation of a grid of ENI seismic lines of the northern Friuli Plain allowed us to reconstruct the buried structural setting of the study area. Four of the analyzed seismic profiles have been selected for the construction of the bi-dimensional and tri-dimensional model of the area.
Starting from the upper Carnian Travenanzes Formation, the simplified stratigraphic model of Friuli region, functional to the interpretation of seismic lines, is composed of five seismo-stratigraphic units [22,57]. - The Mesozoic platforms succession includes the Dolomia Principale and the Friuli-Dinaric carbonate platform [27], which developed from the late Triassic to middle-late Cretaceous. It is composed of an about 3000 m-thick sequence of platform facies, with locally transgressive and emersion episodes [58][59][60][61][62] The lower-middle Miocene sequence is commonly referred as the "Cavanella Group" and is composed of shallow-water marine sediments [18,23,68,69]. In the context of seismic lines interpretation, the Cavanella Group represents an important regionalscale seismo-stratigraphic group of reflectors, showing an overall tabular geometry and a southwestward thinning with thicknesses spanning from tens to hundreds of meters [22]. - The middle to upper Miocene Molasse sequence represents the foredeep deposition of the southeastward verging and migrating south Alpine chain. In the Piedmont Friuli Plain, at the outer border of the pre-Alpine relieves, the upper portion of the sequence consists of very thick fan delta and alluvial deposits dated back to latest Tortonian-early Messinian (Montello Conglomerate, [56,70]), testifying the transition from terrigenous platform to continental facies [23,27]. - The Plio-Quaternary sequence develops on top of the Messinian unconformity. It is composed of thick conglomeratic channel bodies filling the narrow Messinian canyons since the successive transgression episode [27,71]. The sea ingression extended in the Tagliamento paleovalley area and north up to Osoppo, where coarse-grained Gilbert-type deltaic bodies are preserved; the Osoppo Conglomerate dates back to upper Miocene-Pliocene [72] or to the Zanclean [25,73]. The middle-to-late Pleistocene sequence covering the erosional plain surface includes alluvial and glacial facies [27,56,70,71].
A simplified seismostratigraphic model was elaborated (Table 1) for the 2D timeto-depth conversion by integrating the stratigraphic framework of the study area with the velocity logs of some ENI exploratory wells and velocity values extracted from the bibliography, in analogy with [31].
Particularly, velocity values of the Jurassic-Cretaceous carbonate platform (late Paleogene-early Eocene turbidites) were derived from the Cargnacco 1 and Amanda 1 wells, the middle-late Miocene Molasse velocity was derived from the Gemona 1 well, and the early-middle Miocene Cavanella Group and Plio-Quaternary succession velocity values were extracted from the bibliography [22,27,56,70]. The elaborated velocity model was uploaded in the 3D Move Software (by PetEx Ltd., Edinburgh, version 2019.1), where the interpretation of seismic lines was made. The detected seismostratigraphic tops (top Cretaceous carbonatic platform, top turbiditic units and top Cavanella Group) and tectonic structures were digitalized on the selected ENI seismic profiles (sections AA' to DD', see traces in Figure 2). For mapping the Messinian unconformity, we also consulted the regional database of geognostic wells gently supplied by the Friuli Venezia Giulia Region. The horizon digitalization was calibrated using the point data of pre-Quaternary bedrock depth, extracted from projected geognostic wells. Table 1. Velocity model adopted for the 2D depth seismic lines conversion. Velocity values (p-waves) derived from Cargnacco 1, Amanda 1, Gemona1 ENI well-logs and from bibliography [22,27].

Seismostratigraphic Unit
Velocity Once digitalized, each section was converted from time to depth through the Database Method of 2D Conversion tool, which considers the velocity values settled in the stratigraphic model and computes the conversion procedure as the incremental sum of the thicknesses of equal velocity. Starting from the 2D geological sections obtained from the interpretation of seismic lines, the 3D surface of the main tectonic structure was constructed through the interpolation procedure of Create new fault within the Create Surface with Boundaries tool. The Control Points for Kriging option was ticked in order to create a surface from points that are geostatistically valid.

Instrumental Seismicity Analysis
The seismogenic thickness of the studied area was reconstructed through the hypocentral distribution analysis of instrumental seismicity. Two distinct databases were created, referring to the 1976-1977 and 1978-2019 time intervals. The 1976-77 database includes the complete relocation of the sequence performed by [12] and the latest location of the mainshock epicenter elaborated by [11]. In this context, an update of the North-Eastern Italy Seismic Network catalogue was performed by [74], which revised and corrected magnitude estimation thanks to new linear empirical regressions between Ml and Md for the eastern Southern Alps. Differently, the 1978-2019 seismic database contains the events registered by the CRS-OGS Seismometric Network and freely available on the Friuli Venezia Giulia Seismometric Network Bulletin [13]. Both databases were filtered following three criteria: gap < 180 • ; vertical and horizontal error < 4 km and Md ≥ 1. In this regard, the error distribution analysis shows that among the collected events, less than 6% of the earthquakes have an error between 3-4 km, and specifically, they include events with magnitude lower than 3.5.
Focusing on the spatial and temporal earthquake distribution analysis, the events of both databases were classified based on depth and magnitude classes, and the 2D depth distribution was analyzed in terms of number of events per depth class. The 1976-77 seismicity was further categorized into three "time classes" (May-August 1976; September-December 1976; January-December 1977) in order to analyze the evolution of the sequences with time.
Regarding the 1978-2019 database, the Md values were first converted in Ml through the [75] relation, and then the corresponding energy value was calculated through the [76] formula, valid for Ml < 4.5 earthquakes. Differently, the [77] relation for strong earthquakes was only used for only Ml 4.9 event. The total released energy per depth class distribution was then analyzed, differentiating by magnitude value. Successively, the epicentral distributions of the earthquakes of both databases were analyzed by plotting the events in map view. In detail, they were categorized in three depth classes (0-7 km; 7-13 km and >13 km). In a third step, the 3D distribution of the earthquakes was analyzed with respect to the reconstructed fault surfaces by plotting seismicity on seriated sections. Particularly, four N 40 • -trending sections were elaborated, characterized by a length of 30 km, a spacing of 5 km and a projection distance of the events of 3 km.
The focal mechanisms available from the Catalogue of earthquakes of Southern Alps and surrounding area from 1928 to 2019 [9] were consulted for both databases (1976-1977 and 1978-2019). Following the recent focal mechanisms dataset revision, performed by [9], the Catalogue contains all the available focal mechanisms obtained by different authors, and the preferred solution for each event is suggested by the catalogue's authors.

Three-Dimensional Reconstruction of the Susans-Tricesimo Thrust System
The subsurface geometry of the Piedmont Friuli Plain was reconstructed through the elaboration of four geological cross sections (AA', BB', CC' and DD' in Figure 2) obtained by the interpretation of the seismic lines gently supplied by ENI (Figures 3 and 4). The analysis focused on the area between Buia, San Daniele and Tricesimo towns, located into the morainic arc of the Tagliamento river in central Friuli. It is worth noting that the distribution of seismic lines in the study area is not homogeneous (Figure 2 Concerning the Quaternary units, the interpretation of seismic lines allowed us to decipher that STTS clearly affects the Plio-Quaternary succession, suggesting recent tectonic activity for the whole thrust system. In particular, in section AA' (Figure 3A), the projected Reana well reaches the top of Upper Molasse at a depth of 260 masl, covered by a roughly 330 m-thick conglomeratic unit (pre-LGM unit) and a 78 m-thick gravel deposition (LGM unit). Conversely, about 5 km NW, the n. 61 well testifies the presence of a 40 m-thick gravel deposition directly above the turbiditic units (encountered at about 140 masl). In addition, in section AA' ( Figure 3A), BS propagates within a deep Plio-Quaternary paleochannel, probably causing the deformation of the channel itself ( Figure 4A).
The most internal structure detected is the Buia thrust (BU), a low-angle thrust fault that only in the first kilometer depth consists of a steeper ramp. BU slightly displaces the top of the carbonatic platform, and in section AA' ( Figure 3A) it is constrained at the surface by the tectonic contact between Cavanella Group and Eocene turbidites cropping out near Tarcento [27,78].
In the most internal portion of section DD' ( Figure 3D), an additional low-angle infra-Molasse reverse structure is present: the SE-verging, WSW-ENE-trending Osoppo thrust (OSP). This area is constrained by the close Col Vergnal relief and by the homonymous geognostic well located about 250 m south. The relief is formed by Pliocene units (Osoppo conglomerates [25,27,72,79]), on top of which the presence of LGM deposits (Spilimbergo Synthem-SPB) is documented. Conversely, the Col Vergnal well stratigraphy is characterized by about 30 m Holocene deposits and 60 m LGM units, which lie on top of the upper Tortonian-lower Messinian sandstone-mudstone member of Montello conglomerate (MON2), at 83 masl [27] (see the Supplementary Material S1 for the chronostratigraphi-cal correlation scheme of the Plio-Quaternary Units of the Tagliamento Basin, modified after [27]).
In the western portion of the investigated area, at the footwall of ST, three minor reverse faults (roughly E-W-striking) were detected: the San Tomaso (STOM), Majano (MAJ) and San Daniele (SDAN) thrusts. Particularly, in section CC' (Figure 3C), STOM gives rise to a well-developed anticline that involves the south-Alpine Molasse and propagates toward the surface, also affecting the bottom of the Quaternary succession ( Figure 4C). STOM, MAJ and SDAN are characterized by a medium-angle ramp geometry ( Figure     Following the elaboration of the geological cross sections, the 2D profiles were interpolated with the aim to reconstruct the 3D geometry of the active Susans-Tricesimo thrust system and Buia Th. in the piedmont Friuli Plain. ( Figure 5). It is worth noting that seismic lines allow us to reconstruct the geometry of the tectonic structures only in their shallowest portion, up to 5 km in depth. iences 2022, 12, x FOR PEER REVIEW 11 of 33

Analysis of Seismicity Distribution
We analyzed the 1976-2019 collected events in order to investigate the involvement of the tectonic structures during the 1976-1977 earthquakes. Moreover, in order to study the evolution of the seismicity of the Julian Alps and pre-Alps during the last 50 years, seismicity was divided into two different time intervals: 1976-1977 (seismic period) and 1978-2019 (interseismic period). The seismicity distribution was analyzed with respect to the tectonic structures characterizing the study area through the elaboration of four NE-SW-seriated sections (sections 1, 2, 3 and 4 in Figure 6). Moreover, in order to integrate the relationship between the tectonic structures and 1976-1977 seismicity, we realized four additional N-S-oriented sections (Sections S05-S08 in the Supplementary Material S2). The collected events were plotted together with the active faults of the area, partly reconstructed in this study, and partly extracted from the literature [27,31,56,70]. The interpretation of the deep geometry of the tectonic structures followed the detection of planes depicted by seismicity distribution in the different time intervals analyzed. Starting from these planes, we reconstructed the structural model of the investigated crustal volume [80]. In this concern, it is important to note that the identified planes are not always all recognizable in the different time intervals, but only in occurrence of their activation. The strain pattern of the studied fault surfaces was also investigated through the analysis of the available focal mechanisms [9]. Particularly, in order to better define the transpressive, rather than pure dip-slip character of the investigated faults, special attention was dedicated to the events characterized by an oblique component (Reverse-to-Strike-Slip: R-SS; Strike-Slip-to-Reverse: SS-R; Strike-Slip: SS, as in Tables 2 and 3  In particular, STTS is made by a set of three sub-parallel 40 • −45 • NNE-dipping, WNW-ESE-average-striking reverse faults. The faults surfaces are folded according to their polyphase tectonic history. We consider the ST thrust as the masterfault, while Borgo Soima and Colle Villano-North represent two minor splays. Conversely, we consider the Buia thrust as a different structure because it has a different surface strike, and it is characterized by an average low-angle dipping geometry (10-20 • ) in the first kilometers of depth.

Analysis of Seismicity Distribution
We analyzed the 1976-2019 collected events in order to investigate the involvement of the tectonic structures during the 1976-1977 earthquakes. Moreover, in order to study the evolution of the seismicity of the Julian Alps and pre-Alps during the last 50 years, seismicity was divided into two different time intervals: 1976-1977 (seismic period) and 1978-2019 (interseismic period). The seismicity distribution was analyzed with respect to the tectonic structures characterizing the study area through the elaboration of four NE-SW-seriated sections (sections 1, 2, 3 and 4 in Figure 6). Moreover, in order to integrate the relationship between the tectonic structures and 1976-1977 seismicity, we realized four additional N-S-oriented sections (Sections S05-S08 in the Supplementary Material S2). The collected events were plotted together with the active faults of the area, partly reconstructed in this study, and partly extracted from the literature [27,31,56,70]. The interpretation of the deep geometry of the tectonic structures followed the detection of planes depicted by seismicity distribution in the different time intervals analyzed. Starting from these planes, we reconstructed the structural model of the investigated crustal volume [80]. In this concern, it is important to note that the identified planes are not always all recognizable in the different time intervals, but only in occurrence of their activation. The strain pattern of the studied fault surfaces was also investigated through the analysis of the available focal mechanisms [9]. Particularly, in order to better define the transpressive, rather than pure dip-slip character of the investigated faults, special attention was dedicated to the events characterized by an oblique component (Reverse-to-Strike-Slip: R-SS; Strike-Slip-to-Reverse: SS-R; Strike-Slip: SS, as in Tables 2 and 3). The analyzed events of both 1976-1977 and 1978-2019 seismic databases are identified with a progressive number, listed in Tables 2 and 3, and classified with different colors based on their kinematics.

1976-1977 Seismicity Distribution
First, we analyzed the 1976-1977 sequence relocalized by [12]. With the aim to investigate the spatial evolution of the sequence in time, the collected events were classified in three different "time classes": May-August 1976, September-December 1976 and January-December 1977 (Figure 7). This aspect is also useful for exploring the possibility that the sequences ruptured one single structure or whether many different sources were activated.

1976-1977 Seismicity Distribution
First, we analyzed the 1976-1977 sequence relocalized by [12]. With the aim to investigate the spatial evolution of the sequence in time, the collected events were classified in three different "time classes": May-August 1976, September-December 1976 and January-December 1977 (Figure 7). This aspect is also useful for exploring the possibility that the sequences ruptured one single structure or whether many different sources were activated.  Table 2 and focal mechanisms from [81]. The frequency histograms are in terms of number of events, differentiated per magnitude classes, per depth classes.  Table 2 and focal mechanisms from [81]. The frequency histograms are in terms of number of events, differentiated per magnitude classes, per depth classes. The maps of Figure 7 highlight the northwestward migration of the sequences, as already remarked by many authors [11,[83][84][85][86]. However, it should be noted that already during the first sequence (May-August 1976), most of the seismicity is confined to the NW portion of the study area ( Figure 7A). The depth frequency histogram for each of the three "time classes" shows that most seismicity occurred during the first year. Particularly, the comparison between the two 1976 time classes shows that the mainshock of May-Aug 1976 (Ml 6.4) is located at a 7 km depth, which is one of the most active depth classes together with 11 and 12 km, while during Sept-Dec 1976, the most active thickness lies between 6 and 9 km depth, but the strongest event of September (Ml 6.1) occurred at greater depth (12 km).
Concerning the focal mechanisms of the strongest events, the authors in [81] assessed that the May earthquakes activated a low-to-medium-angle NNE-dipping plane, while the events of September are referable to an E-W-trending gently N-dipping geometry ( Figure 7A,B). In this regard, it is worth noting that both studies [87,88] agreed on a different interpretation of the strongest event of 15 September (Ml 6.1), proposing the activation of a roughly NW-SE-oriented medium-angle plane.
In order to make some assessments regarding a possible association earthquakestructure, the distribution of the 1976-1977 sequence was analyzed with respect to the tectonic structures of the area. It must be remarked that we interpret the Colle Villano-North (CV-N) and Borgo Soima (BS) thrusts as secondary splays of the Susans-Tricesimo thrust system (STTS), and we henceforth refer to the whole structure as the Susans-Tricesimo masterfault (ST). The 1976-1977 earthquakes, classified per "time classes", and plotted on the seriated sections (Figure 8, see traces in Figure 6), reveal a diffuse distribution affecting the entire seismogenic thickness. The abrupt decrease in seismicity, which is interpreted as the bottom of the seismogenic thickness, is located at about a 15 km depth, in good agreement with previous literature [89]. that at depth, seismicity is always confined by means of a frontal medium-angle surface, probably referable to the Pozzuolo thrust system (POZ), i.e., the inherited NW-SE-striking, SW-verging frontal thrust of the External Dinarides in Friuli [90], reactivated with an oblique component under the Neolpine maximum stress tensor. In the Piedmont Friuli Plain under Udine locality, POZ shows a transpressional component on the NW-SE-striking plane, and a pure dip-slip movement on the ENE-WSW-striking lateral ramp of Terenzano thrust, deforming the pre-LGM conglomerates (Friuli Supersynthem) and the LGM surface [31,56,91].  Figure 6) together with the active faults' surfaces reconstructed in this study (red thrusts) and tectonic structures known from the literature (black faults, [27]). The symbols' size is proportional to the magnitude value (Ml), the number of the events refers to Table 2 Towards the NE, seismicity is confined by means of high-angle structures. Looking at Section 01, located in the south-easternmost edge of the study area, a sub-vertical cluster possibly corresponding at the surface to the Predjama high-angle transpressional fault  Figure 6) together with the active faults' surfaces reconstructed in this study (red thrusts) and tectonic structures known from the literature (black faults, [27]). The symbols' size is proportional to the magnitude value (Ml), the number of the events refers to Table 2 Focusing on the May-Aug 1976 time interval (Figure 8, sections 01-04) we observe that at depth, seismicity is always confined by means of a frontal medium-angle surface, probably referable to the Pozzuolo thrust system (POZ), i.e., the inherited NW-SE-striking, SW-verging frontal thrust of the External Dinarides in Friuli [90], reactivated with an oblique component under the Neolpine maximum stress tensor. In the Piedmont Friuli Plain under Udine locality, POZ shows a transpressional component on the NW-SE-striking plane, and a pure dip-slip movement on the ENE-WSW-striking lateral ramp of Terenzano thrust, deforming the pre-LGM conglomerates (Friuli Supersynthem) and the LGM surface [31,56,91].
Towards the NE, seismicity is confined by means of high-angle structures. Looking at Section 01, located in the south-easternmost edge of the study area, a sub-vertical cluster possibly corresponding at the surface to the Predjama high-angle transpressional fault (PRJ) [7,[92][93][94] seems to be linked at depth to ST, forming an SW-verging positive transpres-sional structure. A similar arrangement is also visible in Section 02, where the Predjama fault and Susans-Tricesimo thrust interact at about a 10 km depth. Note that the strongest event (6 May Ml 6.4) spatially coincides with the ST (earthquake n. 2 in Figure 8, Section 02), while the Ml 4.5 foreshock is located on the Predjama structure (earthquake n. 1 in Figure 8, Section 01). Therefore, the activation of the Predjama high-angle fault during the earliest 1976 sequence can be hypothesized (see also S2- Figure S3 of Supplementary Material S2), and the possibility that its activation could have triggered the ST source, responsible of the mainshock of May 1976, could be taken into account. We rule out the activation of the Buia source (as proposed by [37]) because BU is characterized by a low-angle dipping geometry (about 20 • ) in the first kilometers of depth ( Figures 5 and 8 Section 02). Nevertheless, the results of the inversion of the geodetic data suggest that the observed coseismic deformation could be reproduced even with seismogenic sources located further south that the Buia thrust (see Figure 6A of [37]), in agreement with the hypothesis of activation of the ST source during the 1976-1977 seismic sequence. Moving towards the NW (Figure 8, sections 03 and 04), the basal seismic level still corresponds to the POZ at about 14-15 km depth (which represents the base of the seismogenic thickness), while in the northeastern inner sector seismicity is confined by the Idrija-Ampezzo Fault System (IAFS). Differently from the southeastern area (sections 01 and 02), on sections 03 and 04, ST gives rise to a steeper ramp (about 40-45 • dip) and extends up to 12 km depth. Moreover, in sections 03 and 04, seismicity mainly concentrates on high-angle structures depicted by means of a clear high-angle alignment of earthquakes. In particular, a well-developed structure, probably corresponding at the surface to the high-angle transpressional Maniaglia fault (MAN) ("Gemona del Friuli geological Sheet", [27]) interacts with the ST steep ramp at about a 12 km depth.
On the basis of the field data, we rule out the possibility that these high-angle structures could represent Gemona-Kobarid (GK) and Musi-Verzegnis (MV) thrusts, since they both are characterized at the surface by a middle dipping angle (about 45 • dip) (Gemona del Friuli Geological Sheet and out of text Figure 1, [27]). If considering the orientation of the seriated sections with respect to the thrusts' geometry, GK and MV would appear with an even lower dip angle, much different from the roughly 70 • -dipping plane observed on the seriated sections (see also S2- Figure S3 of Supplementary Material S2).
Focusing on the September-December 1976 sequence, we observe that most of seismicity develops in the northwestern sector of the study area (see also Figure 7B) and is preferably located on a set of high-angle (60-80 • -dipping) structures (Figure 9, sections 03 and 04), often characterized by oblique focal mechanisms (events n. 15, 16, 18 in Table 2 and Figure 9, Section 04), thus revealing their transpressive kinematics. In particular, sections 03 and 04 ( Figure 9) show a roughly 70 • NE-dipping alignment, probably representing the deep expression of the high-angle Maniaglia fault (MAN). The location of the mainshock of September (Ml 6.1, event n. 13 in Table 2) suggests the activation of the NE deep portion of ST, which here extends up to 12 km-deep and likely interacts at depth with the Maniaglia fault. The remarkable seismic activity at the interaction depth between ST and MAN is already evident from the May-August 1976 earthquakes' distribution ( Figure 8, Section 04). In this sector, the Idrija-Ampezzo Fault System (IAFS) also seems to be involved, acting as the northern backstop of seismicity. orientation of the seriated sections with respect to the thrusts' geometry, GK and MV would appear with an even lower dip angle, much different from the roughly 70°-dipping plane observed on the seriated sections (see also S2 - Figure S3 of Supplementary Material S2). Focusing on the September-December 1976 sequence, we observe that most of seismicity develops in the northwestern sector of the study area (see also Figure 7B) and is preferably located on a set of high-angle (60-80°-dipping) structures (Figure 9, sections 03 and 04), often characterized by oblique focal mechanisms (events n. 15, 16, 18 in Table 2 and Figure 9, Section 04), thus revealing their transpressive kinematics. In particular, sections 03 and 04 ( Figure 9) show a roughly 70° NE-dipping alignment, probably representing the deep expression of the high-angle Maniaglia fault (MAN). The location of the mainshock of September (Ml 6.1, event n. 13 in Table 2) suggests the activation of the NE deep portion of ST, which here extends up to 12 km-deep and likely interacts at depth with the Maniaglia fault. The remarkable seismic activity at the interaction depth between ST and MAN is already evident from the May-August 1976 earthquakes' distribution (Figure 8, Section 04). In this sector, the Idrija-Ampezzo Fault System (IAFS) also seems to be involved, acting as the northern backstop of seismicity.  Figure 6) together with the active faults' surfaces reconstructed in this study (red thrusts) and tectonic structures known from the literature (black faults, [27]). The symbol's size is proportional to the magnitude value (Ml), the number of the events refers to Table 2 Figure 6) together with the active faults' surfaces reconstructed in this study (red thrusts) and tectonic structures known from the literature (black faults, [27]). The symbol's size is proportional to the magnitude value (Ml), the number of the events refers to Table 2 Concerning the southeastern sector (Figure 9, Section 02), seismicity is very scarce; in analogy with the May-August distribution, we suppose that the earthquakes are bordered towards the NE by the high-angle Predjama fault (PRJ). The reactivation of ST can be assessed from event n. 10 ( Table 2).
During January-December 1977, seismicity decreased considerably, and the strongest event (n. 19 in Table 2) is located at the similar depth to the mainshock of 15 September. However, the scarcity of recorded earthquakes prevents any attribution to possible tectonic features.
Based on these results, it is worth remarking that a complex geometry characterizes the ST structure spanning from the SE to the NW. Particularly, in sections 01 and 02 (Figures 8 and 9), the ST reaches about 10 km in depth, by means of roughly 30-35 • -dipping ramp. However, in sections 03 and 04 it shows a steeper ramp (about 40-45 • dipping) and reaches increasingly higher depths (up to 12 km) (Figure 8), suggesting that the ST surface is deformed rather than planar.

The 1978-2019 Seismicity
The collected and filtered 1978-2019 time interval seismicity of the study area (blue rectangle in Figure 10) contains 2726 earthquakes with Md values spanning from 1 to 4.9. The frequency-depth graph (Figure 10), in terms of number of events per 1 km depth class, shows that 90% of earthquakes are located in the first 15 km of depth. Moreover, an abrupt increase in seismicity is evident at a 7 km depth. If considering the magnitude Md of the events, the total released energy per depth class graph shows that most of the energy is released between 9 and 12 km depth since only a few Md > 3.5 are registered in the first 5 km.   In order to investigate the geometry of the seismogenic thickness, the collected events were classified in three macro-depth classes (0-7 km; 7-13 km and depth > 13 km) and the map distribution of the earthquakes was analyzed by plotting the three macro-depth classes as stacked layers, from the deepest to the shallowest (Figure 11).  The maps of different hypocentral depths (Figure 11) show the arcuate distribution of events and highlight the main structural elements, testifying that this area represents the junction zone between the western NE-SW-trending front in the western Carnic pre-Alps and the NW-SE-trending eastern front for the Julian pre-Alps and central Friuli area. Particularly, focusing on the study area, the events located at depth greater than 13 km ( Figure 11A) depict a clear NW-SE alignment, from Tolmezzo to Gorizia. Moving up at shallower depths ( Figure 11B), the earthquakes between 7 and 13 km-deep affect a wider area, expanding south of the Tolmezzo-Gorizia alignment. The southern border of the 7-13 km seismicity is characterized by a WNW-ESE orientation in the central portion of the study area, while SE of Tricesimo, seismicity is distributed along a NW-SE alignment matching the deeper (>13 km) seismicity distribution. Any major details can be added from the shallowest seismicity distribution analysis (0-7 km, Figure 11C), which is widely distributed. Regarding surficial seismicity, it is worth remembering that only 17% of earthquakes are located in the first 6 km depth, with one event exceeding Md > 3.5. Anyhow, the three maps of Figure 11 nicely highlight two active regions: the arcuate, mostly compressive pre-Alpine and Alpine portion of the south-Alpine chain, and the strike-slip domain of western Slovenia, related by a transpressional sector.
In good agreement with the 1976-1977 sequence previously analyzed, the 1978-2019 time interval earthquake distribution highlights the same seismogenic volume activated during the seismic 76-77 period. In detail, the seriated sections ( Figure 12) show that spanning from SE to NW, seismicity affects the entire seismogenic thickness, which extends to depth up to 15 km, progressively involving a wider crustal volume. In this context, it is worth remarking that the post-1977 seismicity refers to an interseismic period, therefore the earthquake distribution nicely highlights the active seismogenic crustal volume as a whole, but the individual structures inside of it are not clearly recognizable.
The southeastern portion of the study area is characterized by a tighter seismogenic volume: in sections 01 and 02 (Figure 12), the earthquake distribution is limited towards the NE by the high-angle Predjama fault and by a frontal medium-high-angle plane towards the SW. The latter correlates with the deep steep ramp which develops at the footwall of ST, as already identified from the 1976-1977 sequence distribution (Figures 8 and 9, sections 01 and 02).
In the northwestern portion of the study area, sections 03 and 04 ( Figure 12) show that seismicity abruptly decreases towards the NE in correspondence to the Idrija-Ampezzo strike-slip Fault System (IAFS). Regarding the first 20 km length of the sections, the plotted earthquakes are limited at depth by means of a roughly 40 • NE-dipping plane, probably correlating to the ST. Contrary to the 1976 seismic sequence, only scarce seismic activity can be related to POZ (Figure 12).  Figure 6 for section traces). The filtered 1978-2019 Md ≥ 2 seismicity (dark blue circles) was plotted together with the active faults' surfaces reconstructed in this study (red thrusts) and tectonic structures known from the literature (black faults, [27]). The symbol's size is proportional to the magnitude value (Md), the number of the events refers to Table  3 The southeastern portion of the study area is characterized by a tighter seismogenic volume: in sections 01 and 02 (Figure 12), the earthquake distribution is limited towards the NE by the high-angle Predjama fault and by a frontal medium-high-angle plane towards the SW. The latter correlates with the deep steep ramp which develops at the footwall of ST, as already identified from the 1976-1977 sequence distribution (Figures 8 and  9, sections 01 and 02).
In the northwestern portion of the study area, sections 03 and 04 ( Figure 12) show that seismicity abruptly decreases towards the NE in correspondence to the Idrija-Ampezzo strike-slip Fault System (IAFS). Regarding the first 20 km length of the sections, the plotted earthquakes are limited at depth by means of a roughly 40° NE-dipping plane, probably correlating to the ST. Contrary to the 1976 seismic sequence, only scarce seismic activity can be related to POZ (Figure 12).  Figure 6 for section traces). The filtered 1978-2019 Md ≥ 2 seismicity (dark blue circles) was plotted together with the active faults' surfaces reconstructed in this study (red thrusts) and tectonic structures known from the literature (black faults, [27]). The symbol's size is proportional to the magnitude value (Md), the number of the events refers to Table 3

Discussion
The seismicity of the central-eastern Friuli region was investigated through hypocentral distribution analysis of the earthquakes registered in the last 50 years, starting from the two seismic sequences that occurred in May and September 1976. The earthquakes' distribution analysis was also compared with both the deep geometry of the tectonic structures obtained from the interpretation of ENI industrial seismic lines in 3D Move (STTS in this paper and POZ in [31]) and with the surface geological data available from the literature [27,56,70]. Both subsurface and surficial data highlight significant structural complexities derived from the polyphase deformational evolution which affected the study area. The results obtained in this work allowed us to better characterize the seismotectonic model of the central-eastern Friuli region, and also to formulate new hypotheses regarding the seismogenic sources of the strongest recent earthquakes.
The eastern south-Alpine area reveals an articulated geological evolution, within which the inherited tectonic structures have always played a fundamental role, strongly conditioning the geometry and the kinematics of the structural elements activated during the successive deformative stages. At present, the seismotectonic framework of the study area is characterized by different deformational domains: towards the east, in western Slovenia, stress is accommodated through pure strike-slip kinematics by NW-SE (Dinaric trending) fault systems [7], while the Venetian and Carnic pre-Alpine regions are characterized by a dominantly dip-slip motion accommodated by the ENE-WSW-striking eastern south-Alpine fronts [96]. At the transition between the two structural domains, the Alpine and pre-Alpine Julian regions define a compressional domain with a strong oblique component [5,30].
Focusing on the earthquakes, the seismicity of the eastern domain is mainly referable to the dextral strike-slip fault systems, as testified by the most recent events of Bovec (1998, Mw 5.6) and Tolmin (2004, Mw 5.2) associated with the Ravne strike-slip fault [33]. Differently, the genesis of the destructive historical earthquake of 26 March 1511 is more debated, but many authors refer it to the strike-slip fault systems of western Slovenia. The CFTMed05 Catalogue [10] suggests the Predjama fault as a seismogenic source, the authors in [38,42] indicate the Idrija fault, while those in [30] propose the activation of the Borgo Faris-Cividale fault. Moving to the western domain, in central Friuli, the widespread presence of many high-angle to subvertical fault segments, often connected by structural bend and step-over zones [27,92], testifies the active northwestward propagation of the dextral strike-slip fault systems. In this sector, slip is accommodated through the partitioning between strike-slip fault segments and the inherited reverse planes from the Paleogene Dinaric orogeny, which were affected by deformation during the neo-Alpine polyphase tectonics (Pliocene-Quaternary phase in [24]; Messinian-Pliocene phase in [17,20,21,25]). At the surface, the presence of dome and basin [97] structures is documented all over the central Friuli region and is related to the superimposition of the neo-Alpine (σ1 = NW-SE to NNW-SSE) tectonics over the Dinaric (σ1 = NE-SW) compression. In the central Friuli area, polydeformed structures have already been documented and mapped by [98], who assigned a tectonic origin to the so-called Arzino, Bernadia ( Figure 2) and Natisone "Ellipsoid" located in the Carnian and Julian pre-Alps. The authors in [27,99,100] described the SE-verging macrofold of Covria Mt. on the right bank of the Tagliamento river ( Figure 2). The authors in [101,102] analyzed the polyphase structural framework of the Bernadia Mts., and those in [103], through a detailed structural analysis of the area located south of the Gemona-Kobarid thrust, confirmed the relationship between Dinaric and neo-Alpine tectonics, as already suggested by [88].
Moreover, the geological field data presented by [27,56] testify that in the northern portion of the Friuli Plain, the sub-surfacing pre-Quaternary succession is characterized by dome and basin structuring, and is generally tilted towards the north. Particularly, while in the whole eastern sector and Buia area the pre-Quaternary terms consist of Eocene turbiditic units affected by ENE-WSW-trending folding, in the northwestern portion, among the Susans Hill, Osoppo and Trasaghis relieves border, a reduced succession of south-Alpine Molasse outcrops [27].
The hypocentral distribution analysis of the 1976 seismic sequences conducted in this study allowed for the 3D reconstruction of the seismogenic volume of the central Friuli region, which corresponds to the first 15 km depth, in good agreement with [89]. In detail, starting from the 2D geometry of Susans-Tricesimo thrust, reconstructed from the interpretation of seismicity distribution on the four seriated sections at the different time intervals (red dashed lines in Figures 8, 9 and 12), the 3D surface was elaborated through interpolation procedures in 3D Move (Spline Curve Method of the Create Surface from Lines Tool). In this context, it is worth highlighting that the 3D surface of the first 5 km depth, reconstructed by integrating geological surface data and interpreted industrial seismic lines (Figure 5), shows the same features of the deep 3D surface reconstructed from the interpretation of the hypocentral distribution ( Figure 13). Indeed, the elaborated model highlights the folded geometry of the Susans-Tricesimo thrust system with an about 20-30 • N-dipping fold axis (Figure 13), likely representing the result of the superimposition of the Pliocene tectonic evolution over the Dinaric compressional phase. The reconstructed deformed surface defines two distinct planes which are probably subjected to different shortening rates: the southeastern surface, characterized by a mean dipping geometry N33 • /30 • , is connected at about a 10 km depth with the Predjama high-angle transpressional fault. Differently, the northwestern segment plane shows a mean N30 • /40 • dipping geometry and reaches higher depths, interacting at about 12 km with the Maniaglia fault. On the first surface (ST-SE) lies the mainshock of May 1976 (Ml 6.4) (Figure 8, sections 01 and 02 and Figure 13), while the strongest event of September 1976 (Ml 6.1) is located on the deepest portion of ST-NW segment, right at the interaction depth with the Maniaglia fault (Figure 9 sections 03 and 04 and Figure 13). vals (red dashed lines in Figs. 8, 9 and 12), the 3D surface was elaborated through interpolation procedures in 3D Move (Spline Curve Method of the Create Surface from Lines Tool). In this context, it is worth highlighting that the 3D surface of the first 5 km depth, reconstructed by integrating geological surface data and interpreted industrial seismic lines ( Figure 5), shows the same features of the deep 3D surface reconstructed from the interpretation of the hypocentral distribution ( Figure 13). Indeed, the elaborated model highlights the folded geometry of the Susans-Tricesimo thrust system with an about 20-30° N-dipping fold axis (Figure 13), likely representing the result of the superimposition of the Pliocene tectonic evolution over the Dinaric compressional phase. The reconstructed deformed surface defines two distinct planes which are probably subjected to different shortening rates: the southeastern surface, characterized by a mean dipping geometry N33°/30°, is connected at about a 10 km depth with the Predjama high-angle transpressional fault. Differently, the northwestern segment plane shows a mean N30°/40° dipping geometry and reaches higher depths, interacting at about 12 km with the Maniaglia fault. On the first surface (ST-SE) lies the mainshock of May 1976 (Ml 6.4) (Figure 8, sections 01 and 02 and Figure 13), while the strongest event of September 1976 (Ml 6.1) is located on the deepest portion of ST-NW segment, right at the interaction depth with the Maniaglia fault (Figure 9 sections 03 and 04 and Figure 13).  On the basis of these statements, important implications in terms of seismogenesis arise: as a consequence of the structural complexities characterizing the deep geometry of the Susans-Tricesimo Thrust, combined with the articulated strain regime of the area, it is likely that the southeastern and northwestern portions of ST define two distinct seismogenic segments. Particularly, the depicted seismotectonic framework suggests that the Ml 6.4 mainshock of May 1976 ruptured the solely southeastern portion (ST-SE) and was likely triggered by the Predjama fault, which was responsible for the Ml 4.5 foreshock. This hypothesis was further investigated by computing the static Coulomb stress change [104][105][106][107] (see Supplementary Material S3). The modelling showed that the foreshock event on the Predjama fault plane induces and increase in Coulomb stress on the ST-SE plane where the Ml 6.4 mainshock of 6 May nucleated. This event caused an increase in Coulomb stress in the area of higher structural complexity affected by the September sequence, as demonstrated by [108]. In particular, the proposed structural model showed that the Ml 6.1 mainshock of 15 September ruptured the deepest portion of the ST-NW seismogenic segment, at the depth of interaction with the Maniaglia fault, which was also activated during the second part of the sequence (Figure 14).  [81], blue from [87] and green from [88].
If considering the presence of a medium earthquake during the hours immediately before the mainshock of 15 September (event n. 11 in Table 2), located on the Maniaglia fault alignment (Figure 9, Section 03), it is likely that the Maniaglia fault triggered the  [81], blue from [87] and green from [88].
If considering the presence of a medium earthquake during the hours immediately before the mainshock of 15 September (event n. 11 in Table 2), located on the Maniaglia fault alignment (Figure 9, Section 03), it is likely that the Maniaglia fault triggered the Susans-Tricesimo northwestern segment, similarly to the interplay that likely occurred during the May sequence between the Predjama fault and southeastern Susans-Tricesimo segment. In this regard, it is worth remarking that the authors in [87] proposed a focal mechanism for the mainshock of September, which shows the activation of a 30 • -dipping NW-SE-striking plane with a significant dextral component, in good agreement with the focal mechanism solution proposed by [88]. Both interpretations, which only slightly differ from the 38 • N-dipping geometry of [81], match our hypothesis regarding the activation of the ST-NW segment as the source of the Ml 6.1 September earthquake.
On the basis of the 3D seismotectonic model reconstructed, the seismotectonic parameters of the identified seismogenic segments are presented in Table 4. Particularly, starting from the length (SRL) and the area (RA) of each source, the maximum expected values were calculated through the [106] empirical relationships considering both thrust-fault (TH) and all type (ALL) kinematics. The estimated Mmax values for both ST-SE and ST-SW well represent the seismogenic potential of the strongest events of the 1976 Friuli sequences.

Conclusions
By integrating 3D structural settings reconstructed from ENI seismic line interpretation (up to 5 km in depth), the seismogenic upper crustal volume investigated through hypocentral seismicity distribution analysis (up to 15 km depth) and the surficial structural setting of the central Friuli area, we reconstructed the present seismotectonic framework of the southern Julian pre-Alps and the northern Friuli Plain. The elaborated seismotectonic model highlights many significant aspects regarding the seismogenesis of the area, suggesting new implications in terms of seismic hazard assessment. Particularly, the proposed model shows that:

•
The seismogenesis of the area is strongly defined by the structural inheritance, especially in terms of structural complexities representing the product of the combined effect of the Paleogene Dinaric orogeny and the late Miocene-Pliocene neo-Alpine compression; • Because of its deep structural complexity, the polyphasic Susans-Tricesimo masterfault is likely segmented into two distinct seismogenic sources (ST-SE and ST-NW); • The northwestward-propagating dextral transpressive fault systems of western Slovenia are affecting the inherited thrusts, and the Dinaric trending transpressive fault segments strongly control the strain accommodation in the central Friuli area, as already remarked by [29,30]. In this context, the possibility that the 15  The activation of the Buia thrust during the strongest events of 1976 is ruled out, and the involvement of the Gemona-Kobarid and Musi-Verzegnis thrusts is reduced, as confirmed by [8].