Fractal Study of the 1997–2017 Italian Seismic Sequences: A Joint Analysis of Seismological Data and DInSAR Measurements

: During the last 20 years (1997 to 2017), four seismic sequences with M w ≥ 5.5 mainshocks nucleated along the Central and Northern Apennines chain (Italy), causing casualties and damage: the 1997 Colﬁorito, the 2009 L’Aquila, the 2012 Emilia, and the most recent 2016–2017 Central Italy seismic sequences. In this work, we perform a novel joint analysis of seismological and remote-sensing data to achieve new insights into the faulting process evolution during the considered seismic sequences. To this aim, we study these seismic sequences by exploiting the available seismological data and by applying fractals theory to them. In particular, we characterize the di ﬀ erent behavior of compressional and extensional seismic sequences by examining the temporal evolution of the fractal dimension values. In addition, we compare the Di ﬀ erential Synthetic Aperture Radar Interferometry (DInSAR) displacement maps relevant to the considered seismic events (already published in our past papers) and the performed spatial and temporal seismological analyses, in order to emphasize some signiﬁcant aspects of the di ﬀ erent faulting processes active during these Italian seismic sequences. The analysis of the fractal dimension values shows that over time extensional seismic sequences are spatially distributed within a volume, whereas compressional ones are aligned along a preferential surface. These spatio-temporal patterns are conﬁrmed by: (1) the spatial distribution of hypocenters for the events that occurred between the mainshock and the post-seismic synthetic aperture radar (SAR) acquisition; (2) the spatial extension of coseismic DInSAR ground-deformation patterns. The proposed seismic and ground-deformation analyses can thus typify di ﬀ erent geodynamic contexts in Italy, providing a distinct image of articulated faulting processes.


Introduction
Every day moderate-to large-magnitude earthquakes occur, generating instantaneously coseismic ground deformations of the Earth's crust, and Italy can be considered as one of the most tectonically active regions of the world [1][2][3]. Figure 1 shows the distribution of historical and recent earthquakes [4]; compressional focal mechanisms are prevalent at the fronts of the Alps and the Apennines, while extensional mechanisms dominate along the Apennines belt [5].
The Italian peninsula extends for more than 1000 km within the central Mediterranean, forming a narrow (<200 km) mountain belt (Figure 1a) between the European and African converging plates. Because of this convergence process and of the consequent subduction of the Adriatic-Ionian lithosphere, which has been active since the Cretaceous, two different orogens (i.e., the Alps and the Apennines) have developed [6,7]. These two mountain belts and, mainly, the Apennines, represent the locations of the  UTM 33) in which the main tectonic structures (derived from [31]) are reported and highlighted by different colors as a function of the geodynamic context: extensional faults, thrusts and strike-slip faults are represented by blue, red and green lines, respectively. The historical seismicity is also shown as a function of magnitude (the higher the magnitude, the bigger the squares) [4]; the mainshocks of the analysed seismic sequences are represented by stars with different colors: the light blue star for the 1997 Colfiorito, the yellow one for the 2009 L'Aquila, the green ones for the 2012 Emilia and the red ones for the 2016-2017 Central Italy mainshocks. The black rectangle indicates the area considered in the following panel.

Seismological Data
To perform our analyses, we considered the seismological data collected in the INGV catalog (National Institute of Geophysics and Volcanology, Italy) [32], in which the overall information about the occurred worldwide earthquakes are reported.
In order to achieve comparable and homogeneous seismic sequences, we adopted the following criteria during the selection of earthquake sequences: (i) presence of local and well-distributed seismometric stations to avoid the occurrence of spatial gaps within the seismic sequence; (ii) existence of complete seismological catalogs to avoid the occurrence of temporal lacks within the seismic sequence; (iii) maximum hypocentral depth of 50 km (i.e., crustal earthquakes); (iiii) mainshocks of M w ≥ 5.5 to achieve representative aftershock sequences. Moreover, the improvement of the Italian seismic network over the last 20 years has allowed to record with high accuracy a major number of small magnitude earthquakes, thanks to the increase in its spatial coverage and to the technological progress. Therefore, due to the high completeness of this catalog, we did not adopt a threshold magnitude considering all the occurred seismic events.
Accordingly, we analyzed the major seismic sequences occurred in Italy since 1997 until now and selected the following sequences: the 1997 Colfiorito (M w 6.0), the 2009 L'Aquila (M w 6.3) and the 2016-2017 Central Italy sequences within extensional tectonic settings; the 2016-2017 Central Italy sequences can be divided, in turn, into the Amatrice (M w 6.0), the Norcia (M w 6.5) and the Campotosto (M w 5.5) sequences. Within compressional tectonic settings, we took into account the 2012 Emilia (M w 6.1) seismic sequence in the Northern Apennines (Figure 1b).
In the following, we briefly describe each seismic sequence investigated in this work.

The Colfiorito Seismic Sequence
On 26 September 1997, a M w 6.0 earthquake struck the Central Apennines region, in Central Italy ( Figure 1b) [33,34]. The Colfiorito mainshock occurred at 7.5 km depth along the~12 km-long Mt. Pennino-Mt. Prefoglio normal fault, which belongs to the Apennine extensional setting [35]. The mainshock was followed by 5410 aftershocks with M w ≥ 1.6. The strongest aftershocks occurred on 6 October and 14 October with M w 5.4 and 5.6, respectively. Furthermore, six months later, on 3 April 1998, another M w 5.1 earthquake nucleated ( Figure S1a).

The L'Aquila Seismic Sequence
On 6 April 2009, a M w 6.3 earthquake struck again the Central Apennines area, in Central Italy (Figure 1b). The L'Aquila mainshock occurred at 8 km depth along the~15-18 km long Paganica normal fault [36,37]. In particular, the mainshock was followed by 19,939 aftershocks with M w ≥ 0.1 and the strongest aftershock occurred on 7 April with M w 5.4 ( Figure S1b).

The Emilia Seismic Sequence
The Emilia seismic sequence started on 20 May 2012 with a M w 6.1 earthquake and nine days later, on 29 May, a second earthquake (M w 6.0) occurred about 15 km southwest (Figure 1b). Both seismic events nucleated along two adjacent compressional structures (at 6.3 km, and 10.2 km depth, respectively) which are part of the Ferrara Arc, buried below the Po plain, in the Northern Apennines fold-and-thrust belt [38,39]. During the following months, these two main seismic events were followed by 2800 aftershocks with M w ≥ 0.7 and four M w > 5 earthquakes were also recorded ( Figure S1c). The 2016-2017 Central Italy seismic sequence began with the M w 6.0 Amatrice earthquake, nucleated on 24 August 2016, activating the northernmost part of the SW-dipping Mt. Gorzano extensional fault and the southernmost segment of the SW-dipping Mt. Vettore Fault System. Then, on 26 October, two seismic events, with M w 5.4 and 5.9 respectively, nucleated nearby Visso, activating the northernmost portion of the Mt. Vettore Fault System. Finally, on 30 October the largest event of the sequence (M w 6.5) occurred near the town of Norcia along the Mt. Vettore Fault System at about 7 km depth and hit the area included between the previous events [40]. Moreover, four events with magnitude larger than 5 (M w 5.1, 5.5, 5.4 and 5.0) struck the Campotosto area, located south of Amatrice; all four mainshocks nucleated on the deepest portion of the northwestern half of the Campotosto fault, at a depth of 9-11 km [41] ( Figure S1d). The whole sequence occurred in a seismic gap located between the 1997 Colfiorito and the 2009 L'Aquila earthquakes (Figure 1b). During the 2016-2017 Central Italy seismic sequence, more than 100,000 events (M w ≥ 0.1) were recorded by the INGV seismic network and all the nucleated mainshocks were characterized by normal fault mechanisms, in agreement with the NE-SW direction of active extension across this sector of the central Apennines.

Differential Synthetic Aperture Radar Interferometry (DInSAR) Measurements
In order to investigate the ground displacements associated with the considered seismic sequences, we exploited the Differential SAR Interferometry (DInSAR) technique [42] which allows analysis of surface displacement phenomena, by providing a measurement of the ground deformation projection along the radar line of sight (LOS). In particular, we exploited DInSAR results relevant to the selected seismic sequences that we have generated and published over the years. Specifically, we applied the DInSAR technique to synthetic aperture radar (SAR) data acquired by different sensors along ascending and descending orbits (Table 1 and Figure S2). Accordingly, we generated coseismic differential interferograms (note that within the differential interferogram generation, the SRTM (Shuttle Radar Topography Mission) DEM (digital elevation model) of the investigated areas were exploited [43]) and, subsequently, their corresponding LOS displacement maps (Figures 2-7), the latter obtained through an appropriate phase unwrapping operation [44]. We highlight that when ascending and descending SAR acquisitions were available (such as for L'Aquila and Central Italy seismic sequences), we were able both to detect the LOS ground deformation and to discriminate the Vertical and East-West components of the displacements (Table S1) [45]. To achieve this task, we properly combine the LOS deformation maps computed from the ascending and descending orbits on pixels common to both maps. In particular, the sum of the ascending and descending surface deformation patterns allows us to get a picture which mostly represents the vertical motion; by contrast, the difference between the ascending and descending displacement maps provides an estimate of the surface deformation in the East-West direction. Finally, it is worth noting that no reliable results on possible North-South deformation components can be achieved from DInSAR data because they do not contribute to the measured LOS displacement component due to the (nearly) polar direction of the sensor orbits [46]. In the following, we briefly describe each coseismic deformation map investigated in this work.

The Colfiorito Seismic Sequence
The ground displacements caused by the M w 6.0 Colfiorito earthquake (26 September 1997) have been analyzed through DInSAR measurements acquired by the European Remote-Sensing Satellite (ERS) from descending pass. Figure 2 shows the deformation map, which is relevant to the SAR acquisitions of 7 September and 12 October 1997 [46]. The map reveals a maximum displacement along the radar line of sight (LOS) of about 25 cm. The hanging wall block is affected by the maximum measured subsidence and, in terms of faulting mechanism, this is consistent with a normal slip mechanism.

The L'Aquila Seismic Sequence
The ground displacements relevant to the M w 6.3 L'Aquila earthquake (6 April 2009) have been retrieved through DInSAR measurements acquired by the Environmental Satellite (ENVISAT) from ascending and descending passes ( Figure 3). Figure 3a,b show the ascending and descending deformation maps, relevant to the acquisitions of 11 March and 15 April 2009 and of 1 February and 12 April 2009, respectively [47]. These two maps have also been properly combined to retrieve the vertical and the east-west displacement components [29], which are reported in Figure 3c,d.
As also discussed in Castaldo et al. [29], the retrieved deformation pattern (Figure 3a,b) shows a non-symmetric distribution of the LOS displacements field. The retrieved deformations in the northwestern region reach the value of 17-18 cm and 23-24 cm (corresponding to subsidence processes), for the results relevant to ascending and descending orbits, respectively; moreover, in the northeastern region a maximum displacement of 4-5 cm (corresponding to an uplift of the examined zone) is visible only in the descending map ( Figure 3b). Furthermore, the vertical deformation map reveals a maximum subsidence of about 25 cm that affected the hangingwall block and, in terms of faulting mechanism, this is consistent with a normal slip mechanism.

The Emilia Seismic Sequence
The ground displacements caused by the M w 6.1 and 6.0 Emilia (20 and 29 May 2012) earthquakes have been studied through DInSAR measurements acquired by the RADARSAT-2 satellite from descending pass ( Figure 4). Figure 4 shows the deformation map relevant to the acquisitions of 30 April and 17 June 2012; therefore, it has encompassed the two main seismic events [48]. The deformation map reveals a maximum LOS displacement of about 17 cm (corresponding to un uplift of the considered seismogenic zone). The hanging wall block is affected by the maximum measured uplift and, in terms of faulting mechanism, this is consistent with a reverse slip mechanism.

The Central Italy Seismic Sequence
We analyzed the ground displacements caused by each mainshock of this seismic sequence.
In particular, we focused on Sentinel-1 (S1) data acquired over the study area from both ascending and descending orbits. Thanks to their short revisit time and small spatial baseline separation, we selected among the possible interferometric pairs those less affected by undesired phase artifacts (atmospheric phase delays, decorrelation noise, etc.), thus preserving good spatial coverage and interferometric coherence.
In the following, we describe the deformation pattern generated by each earthquake of this seismic sequence. geodynamic context: extensional faults, thrusts and strike-slip faults are represented by blue, red and green lines, respectively. The historical seismicity is also shown as a function of magnitude (the higher the magnitude, the bigger the squares) [4]; the mainshocks of the analysed seismic sequences are represented by stars with different colors: the light blue star for the 1997 Colfiorito, the yellow one for the 2009 L'Aquila, the green ones for the 2012 Emilia and the red ones for the 2016-2017 Central Italy mainshocks. The black rectangle indicates the area considered in the following panel.

The Amatrice Seismic Sequence
The ground displacements caused by the M w 6.0 Amatrice earthquake (24 August 2016) have been analyzed through DInSAR measurements acquired by S1 constellation from ascending and descending passes ( Figure 5). Figure 5a,b show the selected ascending and descending deformation maps, relevant to the acquisitions of 21 and 27 August 2016 and of 15 and 27 August 2016, respectively [26]. The analysis of these coseismic DInSAR maps revealed a spoon-like shape geometry of the detected surface deformation pattern ( Figure 5) characterized by two NNW-SSE striking main distinctive lobes. The ascending and descending DInSAR maps were also properly combined to retrieve the vertical and the east-west displacement components, which are reported in Figure 5c,d. The vertical deformation map reveals a maximum subsidence of about 20 cm that affected the hangingwall block and, in terms of faulting mechanism, this is consistent with a normal slip mechanism.

The Norcia Seismic Sequence
The ground displacements caused by the M w 6.5 Norcia earthquake (30 October 2016) have been examined through DInSAR measurements acquired by S1 constellation from ascending and descending passes ( Figure 6). As in the case of the Amatrice earthquake, we focused on S1 SAR data acquired on 26 October and 1 November 2016 along ascending and descending orbits (Figure 6a,b) [28]. The analysis of these coseismic DInSAR measurements revealed two NNW-SSE-striking main lobes. The ascending and descending DInSAR maps were also combined to retrieve the vertical and the east-west displacement components, which are reported in Figure 6c,d. In particular, the vertical deformation map revealed a maximum subsidence of about 70 cm that affected the hangingwall block, and also a maximum uplift of about 10 cm that affected the adjacent footwall block (Figure 6c). These are the biggest ground deformations ever measured in Italy by using the DInSAR technique and caused by an earthquake. In terms of faulting mechanism, the measured ground displacements are consistent with a normal slip mechanism.
The Campotosto Seismic Sequence The ground displacements caused by the M w 5.5 Campotosto earthquake (18 January 2017) have been studied through DInSAR measurements acquired by the ALOS-2 satellite from ascending and descending passes and S1 constellation from ascending pass (Figure 7). The ALOS-2 measurements were acquired on 2 November 2016 and 25 January 2017 and on 9 November 2016 and 15 February 2017 along ascending and descending orbits, respectively (Figure 7a,b), and the S1 measurements were captured on 12 January and 24 January 2017 along the ascending orbit (Figure 7c). The ascending and descending maps were also combined to retrieve the vertical and the east-west displacement components, which are reported in Figure 7c,d. In particular, the vertical deformation map revealed a maximum subsidence of about 15 cm that affected the hangingwall block (Figure 7d). In terms of faulting mechanism, the measured ground displacements are consistent with a normal slip mechanism.    (c) DInSAR LOS displacement map computed by using S1 images acquired from ascending orbits on 12 January-24 January 2017. (d,e) Vertical and E-W displacement maps computed by exploiting the ascending and descending ALOS-2 and the ascending S1 measurements shown in panels (a-c). The white star represents the Mw 5.5 Campotosto mainshock. (c) DInSAR LOS displacement map computed by using S1 images acquired from ascending orbits on 12 January-24 January 2017. (d,e) Vertical and E-W displacement maps computed by exploiting the ascending and descending ALOS-2 and the ascending S1 measurements shown in panels (a-c). The white star represents the M w 5.5 Campotosto mainshock.

Employed Method
We examined faulting and fragmentation processes by using the Fractals Theory [19,20,23]. In this context, the fractal geometries are strictly related to the faulting processes caused by earthquake nucleation. The variation of fractal parameters can thus be indicative of the temporal and spatial evolution of the fragmentation processes along a fault system in time and space. We analyzed the seismological data presented in Section 2.1 (in particular, the magnitude-frequency distribution of earthquakes) by using a dedicated software, fitting the same data with a power law, and obtaining the fractal dimension and the related coefficient of determination (i.e., R-squared or R 2 ).
The equation relevant to the above mentioned power law is: (1) where N i is the number of objects with a characteristic linear dimension r i , C is a constant of proportionality, and D is the fractal dimension (in our case, N i is the number of earthquakes with a certain magnitude ranges r i ).
For each seismic sequence, we realized semi-logarithmic graphs, in which we reported the number of earthquakes occurred in certain magnitude ranges (Figure 8 and Figure S4). In particular, we plotted the number of earthquakes on the y-axis by using a logarithmic scale and the correspondent magnitude range on the x-axis. According to Equation (1), we searched for the exponential curve that best fits the considered data. Since we worked in a semi-logarithmic space, this exponential curve turns in a straight line (see dotted line in Figure 8 and Figure S4). All the unknown parameters of Equation (1), and in particular the fractal dimension, directly derive from the performed best-fit operation.  According to this topological analysis, the fractal dimension value represents the level of irregularity of the selected fractal set [23] and is indicative of the fragmentation process that occurred during the mainshock and the following aftershocks. If D = 0, it represents the classical Euclidean dimension of a point; if D = 1, the dimension of a line segment; if D = 2, the dimension of a surface and, finally, if D = 3, the Euclidean dimension of a volume [23,49]. To check data accuracy, we calculated the coefficient of determination (R-squared) connected with Equation (1) for each seismic sequence. R-squared represents the variability between data instability and model accuracy and can range between 0 and 1. The higher the R-squared values, the higher the model accuracy; therefore, high R-squared values point to a good fit between the model accuracy and the data instability.

Results
The results retrieved from the application of the Fractals Theory show some significant aspects about seismic sequences and, in particular, about the nature and classification of the faulting processes active during the considered Italian seismic sequences with comparable magnitude. In particular, the analyses of the seismic sequences, performed through the fractal dimension computation within a 1000 days time-lapse, reveal that: In the case of the 1997 Colfiorito seismic sequence, the fractal dimension assumes an initial value of about 2.3 and reaches a maximum value of about 2.55; as regards the R-squared, its values range between about 0.982 and 0.9935 at the beginning and at the end of the established time interval, respectively.
In the case of the 2009 L'Aquila seismic sequence, the fractal dimension assumes an initial value of about 2.48 and reaches a maximum value of about 2.75; as regards the R-squared, its values range between about 0.9475 and 0.9415 at the beginning and at the end of the established time interval, respectively.
In the case of the 2012 Emilia seismic sequence, the fractal dimension assumes an initial value of about 1.73 and reaches a maximum value of about 1.79; as regards the R-squared, its values range between about 0.965 and 0.96 at the beginning and at the end of the established time interval, respectively.
In the case of the 2016-2017 Central Italy seismic sequence, the fractal dimension assumes an initial value of about 2.9 and reaches a maximum value of about 3.05; as regards the R-squared, its values range between about 0.973 and 0.968 at the beginning and at the end of the established time interval, respectively.
To summarize, we highlight that these values vary between ca. 2-3 and ca. 1-2 for extensional (i.e., the 1997 Colfiorito, the 2009 L'Aquila and the 2016-2017 Central Italy seismic sequences) and compressional seismic sequences (i.e., the 2012 Emilia seismic sequence), respectively (Figure 9a,c,e,g and Figure 10). As the fractal dimension is indicative of the geometrical features of a faulting process [23,49], it means that over time extensional seismic sequences are thus spatially distributed within a volume, whereas compressional ones are aligned along a preferential rupture surface.
Moreover, we show that the average coefficient of determination R-squared is greater than 0.94 for all the analyzed seismic sequences (Figure 9b,d,f,h and Figure S3); these high R-squared values indicate both that the retrieved results are robust and that the exploited seismological catalog presents a good completeness.
As visible in Figure 10, our detailed fractal dimension analysis (200 days) reveals that the examined faulting processes can be considered auto-similar during their temporal evolution. This tendency can be observed both in the case of a singular faulting process with one medium-high intensity mainshock (i.e., the 1997 Colfiorito, the 2009 L'Aquila and the 2012 Emilia seismic sequences) and of articulated sequences, characterized by the spatial activation of more fault segments over time (i.e., the 2016-2017 Central Italy seismic sequence). Moreover, from the beginning of the considered observation time interval (i.e., from the mainshock) the fractal dimension presents initial values decisively greater than

Discussion
Starting from the results retrieved from the fractal dimension computation (i.e., fractal dimension values vary between ca. 2-3 and ca. 1-2 for extensional and compressional earthquakes, respectively), we show that over time extensional seismic sequences are thus spatially distributed within a volume, whereas compressional ones are aligned along a preferential rupture surface.
The validity of the retrieved fractal dimension results is confirmed by correlating the 3D hypocentral spatial distribution of the earthquakes occurred in the time interval ranging between the mainshock and the post-seismic SAR acquisition (time-lapse corresponding to the red dashed areas in Figure 10), and the coseismic DInSAR ground deformation maps, see Figure 11. Indeed, we observe that in the case of the extensional seismic sequences the hypocenters present a volumetric distribution, which occupies the involved hangingwall rock block (Figure 11a,b,d,e,f); conversely, in the case of the compressional seismic sequences, the hypocenters distribution reveals a good and preferential alignment along the involved seismogenic fault (Figure 11c).
We also emphasize that the coseismic deformation patterns present different characteristics for extensional and compressional seismic sequences (see Figures 2-7). Specifically, in the first case the ground deformation pattern is typically an image of the crustal block involvement during faulting processes [29]; accordingly, the deformation pattern retrieved on the Earth's surface is significantly spatially extended with respect to the main fault location (Figures 2, 3, 5, 6, 7 and 11a,b,d,e,f) [26,[28][29][30]. Conversely, in the second case the observed ground deformation pattern is strictly linked to the fault geometry; therefore, the measured ground deformation pattern is spatially less extended with respect to the main fault location (Figures 4 and 11c) [48]. This evidence may support the different fractal dimension values that, in turn, can be considered as a signature of different evolution of fragmentation mechanism.
The performed joint analyses provide a clear image of the faulting processes consequent to the mainshock nucleation, thus suggesting the possibility to conceive the evolution of the activated faulting process also starting from the first days of the considered seismic sequences.
Finally, we propose that the different features highlighted by our analyses can be related to the different types of energy released during the earthquakes, in turn associated with the tectonic setting [24]. In particular, we suggest that extensional seismic sequences mainly dissipate gravitational energy, stored during the interseismic period, and the consequent downward hangingwall block movement is favored by gravity. This induces an increase of potential energy and facilitates fracturing processes during the coseismic phase; the potential energy is converted into kinetic energy as indicated by the double-couple mechanism of the earthquakes generated by the shear on the fault planes. In this case, most of the involved forces are distributed within the fault hangingwall block, implying a volumetric distribution of the occurred seismic events. On the contrary, we suggest that the compressional seismic sequences are characterized by dissipation of elastic energy, which is stored both within the rock volume above the thrust fault (i.e., the hangingwall block) and along the thrust fault itself during the interseismic period. The elastic energy dissipation is buffered by the gravitational force and the downward directed gravitational force is opposite to the upward sense of motion of the hangingwall. Therefore, to overcome the inertial system, most of the stored energy is used to activate the main fault, stresses are concentrated at the interface between hangingwall and footwall blocks (i.e., along the thrust) and, therefore, earthquakes nucleate in correspondence of the seismogenic thrust, showing a preferential alignment along the involved structure.
Remote Sens. 2019, 11, x FOR PEER REVIEW 20 of 24 Figure 11. 3D comparison between DInSAR measurements and earthquakes distribution. 3D view of coseismic displacement maps shown in Figures 2, 3c, 4, 5c, 6c and 7d, and seismicity (the higher the magnitude, the bigger the spheres) recorded during the time interval defined in the red dashed areas in Figure 10

Conclusions
In this work, we have performed a detailed analysis of four seismic sequences that occurred in the Italian peninsula during the last 20 years (1997 to 2017) with M w ≥ 5.5 mainshocks, and we have jointly exploited the available seismological data and DInSAR measurements, in order to investigate the different features of faulting mechanism (i.e., compressional and extensional tectonic settings).
Our main findings can be summarized as follows: We have employed the fractal analysis to investigate the fragmentation processes caused by earthquake nucleation, which is related in turn to the temporal variation of the fractal dimension values.
The analyses of the fractal dimension values show that over time extensional seismic sequences (i.e., the 1997 Colfiorito, the 2009 L'Aquila and the 2016-2017 Central Italy seismic sequences) are spatially distributed within a volume, whereas compressional ones (i.e., the 2012 Emilia seismic sequence) are aligned along a preferential rupture surface. This is clearly highlighted by the hypocentral spatial distribution of the events, as well as from the surficial spatial extensions of DInSAR coseismic deformation maps.
Our fractal dimension analysis suggests that the studied seismic sequences reveal an auto-similar tendency during their temporal evolution, in the case of both singular faulting processes and articulated ones; moreover, the retrieved fractal dimension values furnish a clear image of the faulting processes consequent to the mainshock occurrence, within compressional and extensional settings.
This joint analysis has allowed us to typify different Italian geodynamic contexts and to provide new insights into the faulting processes caused by the considered seismic sequences.  Figure S2: Interferometric SAR data pairs. Timesheet and coseismic interferometric SAR data pairs exploited for the analysis. The M w 6.0 Colfiorito, the M w 6.3 L'Aquila, the M w 6.1 Emilia mainshocks are represented by the light blue, yellow and green stars, respectively; the Central Italy mainshocks (i.e., the M w 6.0 Amatrice, the M w 6.5 Norcia and the M w 5.5 Campotosto earthquakes) are highlighted by the red stars. Figure