Intermittent Slip Along the Alto Tiberina Low‐Angle Normal Fault in Central Italy

The Alto Tiberina normal fault (ATF) in central Italy is a 50‐km‐long crustal structure that dips at a low angle (15–20°). Events on the fault plane are about 10 times less frequent than those located in its shallower syn‐ and antithetic hanging‐wall splays. To enhance ATF catalog and achieve a better understanding of the degree of coupling in the fault system, we apply a template matching technique in the 2010–2014 time window. We augment by a factor 5 the detections and decrease the completeness magnitude to negative values. Contrary to what previously observed on ATF, we highlight intermittent seismic activity and long‐lasting clusters interacting with sequences on the shallower splays. One of these episodes of prolonged seismic activity, detected at the end of 2013 on a 30‐km‐long ATF segment, suggest the ATF active role during an aseismic transient unraveled by geodetic data.

The Alto Tiberina Fault (ATF), located in the Northern Apennines within the Umbria-Marche region in central Italy (Figure 1), is an active LANF (e.g.,  within a fold and thrust belt under- going NE-trending extension at a rate of about 3 mm/year . This fault was identified by interpreting deep seismic reflection profiles (Barchi et al., 1998) and by field geology, gravity, magnetic and heat flow data, measurements, and modeling (Boncio et al., 2000;Collettini & Barchi, 2002;Mirabella et al., 2011;Pauselli et al., 2006).
These studies reconstructed a ≈50to 60-km-long extensional NNW-trending main fault dipping 15-20 • from the surface to at least 12-14 km depth, toward ENE. Above the ATF, synthetic and antithetic structures dip at high-angle (40-60 • ), ending at the intersection with the ATF plane at approximately 3-6 km of depth ( Figure 1).
Several geophysical studies investigated ATF mechanical behavior. Vadacca et al. (2016) through a 2-D numerical elastic model including geological and seismic data, simulated the interseismic deformation of the whole fault system. The comparison between the modeling results and GPS data supports the active role Geophysical Research Letters 10.1029/2020GL089039 played by ATF in accommodating the ongoing tectonic extension in this sector of the chain. From geodetic data, Gualandi et al. (2017) identified a small aseismic transient signal associated with the occurrence of a shallow sequence lasting about 6 months in 2013-2014 (Gubbio M W = 3.8). The transient produced an extension of about 5 mm with a moment release equivalent to an M W = 4.7, and it was modeled to occur along two minor shallow steeply dipping faults placed within the hanging-wall volume (Figure 1b).
By applying template matching to data collected between 2010 and 2014 (Figure 1), we analyze the ATF seismicity pattern to investigate the mechanical behavior of this LANF. Exploiting a high-resolution earthquake catalog, we investigate the interactions and the possible degree of coupling between the ATF and the shallower system of normal faults. To this end, we characterize the clusters of the augmented catalog in terms of swarms, mainshock-aftershock, and foreshock-mainshock sequences. Including greater detail in the earthquakes patterns and delimiting the fault portions that actively creep or aseismically slip help to better evaluate where stresses accumulate along the ATF and could provide indications on the connection between the fault system heterogeneity and the physics of earthquake initiation.

Input Catalog
Instrumental seismicity in the studied area was mainly located in the shallower dipping faults in the ATF hanging-wall and along with the ATF, while the footwall is almost aseismic (Chiaraluce et al., 2007). The high-angle synthetic and antithetic faults in the hanging-wall showed the most energetic sequences in the last 50 years (M S <5.2 in 1984Haessler et al., 1988. Small earthquakes along the ATF (M L <2.4) are released at the average rate of 2.2 events per day (Valoroso et al., 2017) and within ≈1-km-thick dipping fault ranging from ≈3 to 16 km of depth.
The procedure applied by Valoroso et al. (2017) to obtain high-quality locations consisted of using waveform cross-correlation and double-difference methods (Waldhauser & Schaff, 2008). The retrieved catalog contains 36,819 events that occurred between 1 April 2010 and 30 September 2014. Seismic events have at least 12 phase readings, azimuthal gap lower than 180 • , residuals (RMS) smaller than 0.3 s, and a formal relative location error of few tens of meters.
The 3-D geometry of the ATF as imaged by seismic reflection profiles (Mirabella et al., 2011) together with relocated seismicity has been used to select events belonging to the ATF and those occurring in its hanging-wall ( Figure 1b). The ATF related earthquakes, hereafter named ATF-Z, are the ones nucleating within 1.5 km from the assumed fault surface; these 3,685 events, in the magnitude range −1 < M L < 2.4, represent the input templates.
Under the assumption that the templates widespread along the ATF may represent a meaningful sample in time and space of the fault behavior, we apply a waveform matching technique to detect a more significant number of small earthquakes not yet recognized by conventional techniques. These events contained in the continuous recordings can only be retrieved by the application of a more sensitive and computationally demanding method.

Detection Results
By template matching (details in Text S1 in the supporting information and in Sugan et al., 2019;Vuan et al., 2017Vuan et al., , 2018, we augment the starting catalog of 16,500 new detections (hereafter named ATF-Z+), most of them having smaller magnitudes than the 3,685 templates, allowing a substantial decrease of completeness magnitude (M c ) from M L 0.6 to less than M L 0.0 ( Figure S1). It is worth noting that M c of the ATF-Z and ATF-Z+ catalogs vary with time during the 5 years decreasing at the end of 2013. This decrease is not due to a different network setting or performance but to rate changes of seismicity. Examples of detection for a template of M L 0.0 finding an M L 0.8 and an M L -1 event are shown in Figure S2.
Looking at the occurrence of the events versus time in terms of cumulative number, the first novelty unraveled by the new ATF-Z+ catalog is that we do not observe anymore a stable and gradual increase as we did for the templates. Now, the distribution in 5 years time of the new detections in comparison with the hanging-wall seismicity shows ( Figure 2) that there is some synchrony in the frequency of occurrence of the events, especially during the moderate shallow sequences at the end of 2013 to early 2014. The time-magnitude distribution for the template and ATF-Z+ catalogs are different. The greater detail gained by operating the template matching analysis highlights intermittent bursts spanning the whole period from 2010 to 2014 (Figure 2f), previously completely undetected. Moreover, in ATF-Z+, we are able to fill time windows for which we had no available templates, as in the first months of 2010 or at the end of 2014. We run the waveform matching also for these two unsampled time windows to check both the power of the method and to confirm specific seismicity patterns as the intermittent bursts we find at the end of 2014.
We also detect few new M L < 3.0 events occurring along the ATF in particular when many events happened during the shallower 2013-2014 sequence. Rising seismicity on the hanging-wall often hampers the detection of deeper and lower magnitude ATF earthquakes.
ATF-Z+ samples the fault volume more densely than original ATF-Z (Figure 3). The number of events in ATF-Z+ is considerably increased, and new detections are distributed more homogeneously in the 70 km along-strike area, also showing many clusters. From Figure 3, it is noteworthy the increase of seismicity on ATF after December 2013, spanning an area that extends 30 km along-strike beyond the fault portions activated in the hanging-wall. We also find clusters on ATF portions not included in the templates catalog occurring respectively after or in synchronous with sequences on the shallower splays of the hanging-wall. The distribution of the events at depth indicates some preferential ranges separated by a 1-to 1.5-km-thick layer, from approximately 7-8 km to 8-9 km (see Figure 3a), where seismicity considerably decreases. The decrease of seismic events at ≈8 km of depth (Latorre et al., 2016) seems to coincide with lithological changes (from acoustic to crystalline basement) and an increase in seismic velocities at depth.

Interaction Between ATF and Hanging-Wall Splays
The high-resolution catalog we generated represents an exceptional opportunity to correlate the tectonic loading variations, in terms of seismicity rate changes, to the degree of interconnection between different fault segments with varying frictional properties. Hereafter we describe the seismicity pattern of the ATF augmented catalog relative to the seismic sequences occurred along the secondary splays faults located in the hanging-wall, all with multiple mainshocks ranging from 3.0 ≤ M W ≤ 3.8. The primary sequences, extensively described by Valoroso et al. (2017) and Marzorati et al. (2014), occurred in 2010 (Pietralunga, M W 3.6), 2013 (Cittá di Castello, M W > 3) and the long-lasting 2013-2014 (Gubbio, M W 3.8; Figures 1 and 3 for map view and along-strike position, respectively).
Looking at the interplay of the seismic activity within the fault system during the hanging-wall sequence of April-May 2010 (Pietralunga), we do not observe a microseismicity rate increase on the ATF before or immediately after it. Multiple events, starting from July until the end of August 2010, are instead found by template matching on the ATF volume, south of the Pietralunga sequence, as a possible delayed response ( Figure 2).
Conversely, during the April 2013 (Cittá di Castello) hanging-wall sequence, we find many undetected events supposed to be related to the ATF (≈2,000) ( Figure 3f). The earthquakes along the ATF are almost synchronous with those on the hanging-wall. This sequence is the shallowest one, and the ATF low-angle plane could be, based on this newly discovered seismicity, interconnected with the high-angle normal faults close to Cittá di Castello.
Such a simultaneous activity is also observed during the 2013-2014 Gubbio sequence when the rate change in the ATF affects a ≈30-km-long fault segment (box in Figure 3f). Moreover, along the ATF activated zone, there are particular areas (e.g., strands) of the fault plane where the seismic activity persists over time interspersed with more quiet zones (Figure 3f). The seismic activation of the ATF started in December 2013 and extended slightly beyond the length of the Gubbio fault system. The microseismicity persisted for months after the M W > 3 seismic events between December 2013 and February 2014, suggesting a strong coupling between the system of splay faults in the hanging-wall and the ATF (Figure 3).
If the ATF and the system of splay faults are strongly coupled, we could consider the microseismicity at depth on the ATF as a tracer of a possible afterslip involving an area of ≈250-300 km 2 . The same coupling is not observed in the period 2010-2011 and during the first 6 months of 2013 when the same portions of the hanging-wall volume close to Gubbio were active ( Figure 3). Therefore, we suggest that the interchange of accelerated microseismicity (end of 2013) could be related to a transient causing a tectonic loading variation on large ATF portions.

Clustering Analysis
To further investigate the seismicity pattern and its correlation to tectonic loading variations, we analyze the clustering over time and space (see Text S2). The time-space (T-R) rescaled distribution of the events in ATF-Z+, by applying the Zaliapin and Ben-Zion (2016) approach, helps in separating the clusters from the background seismicity that spans homogeneously all along the ATF with a smooth increase from December 2013 ( Figures S3a and S3b).
The clusters, at least for the first couple of years (2010-2012), seem almost homogeneously distributed in space and time mainly in the northern portion of the fault plane (from −10 to 20 km along-strike in Figure S4a). Then, from the end of 2013, contemporaneously to the seismic activity occurred in the splays located north of the Gubbio area, the clusters occurrence increases together with the energy released along a ≈30-km-long fault portion.
In detail, we observe that until December 2013, clusters last no more than a few days and are located all along the fault plane (between 6 and 12 km of depth in Figure S4a). At the end of 2013 until the first few months of 2014, the clusters extend overtime lasting for 2-3 months and appearing as seismic strands in a 30 km along-strike fault segment. Clusters depth is generally shallower (<6 km) in the northern area of the ATF and deepening (6-10 km) in the central part. The same depth distribution of events, more superficial at the edges and deeper in the central zone of the fault, is also shown by the background seismicity ( Figure S4b). The synchronous activation of long-lasting clusters in an area of ≈250-300 km 2 suggests an increasing slip on the ATF since December 2013 and a not negligible role of ATF in the evolution of the aseismic transient observed by Gualandi et al. (2017).
We then classified the clusters in swarms, mainshock-aftershock, and foreshock-mainshock sequences following the criterion proposed by Ogata and Katsura (2012). In Figure S5, we show examples of these types of clusters. The declustering procedure applied to the ATF-Z+ finds 133 clusters with at least 10 events. Forty-eight percent of clusters are classified as foreshock-mainshock, 47% as swarm-like, and only 5% represents typical mainshock-aftershock sequences (Figure 4). Examining the magnitude distribution within each ATF cluster with more than 10 events allow us to understand their development in time and space better. Swarms, more sensitive to changes of the tectonic loading (e.g., Passarelli et al., 2018;Shelly et al., 2016;Toda et al., 2002), persist over time compared with typical foreshock sequences and are characterized by a significantly higher number of events. Few clusters classified as swarms revealed a complex mixed behavior with bursts repetitions in a few days.

Discussion
A recent analysis of the ATF seismicity (Valoroso et al., 2017) proposed that the release of small events was constant all along the main fault and widespread, without any evident change in the rate of earthquakes occurrence or evidence for long-lasting clusters. In particular, similarly to Chiaraluce et al. (2007), clusters were considered to be composed of few events (mainly doublets) and having a short duration (hours-day).
The new augmented catalog we obtained, having a greater detail, shows different peculiarities with repeated variations in the seismic release both in space and time. These sudden variations often coincide with moderate seismic sequences (e.g., Cittá di Castello 2013 and Gubbio 2013-2014) activating the shallower splays in the fault hanging-wall; other times the rate changes are instead uncorrelated (e.g., Pietralunga 2010 and some clusters in 2011). Despite in some cases the activation of specific ATF portions is then evident, and coincident with the strongest earthquakes (M L > 3.5) occurring in the structures above, the interaction between the master and secondary faults is challenging to interpret by using only seismic data.
However, we suggest an active role of the leading low angle fault during the occurrence of the slow slip event at the end of 2013. The stress load from the aseismic transient may have induced the ATF rise in seismicity or have been taken place along with the ATF itself. In both cases, the ATF seismic energy release is not enough to fill the gap between the geodetic and the seismic moment. As observed by Gualandi et al. (2017), the seismic moment associated with aseismic slip is, in fact, more than 2 times higher than the value related to the seismic one, and the additional events now included in the ATF-Z+ catalog are still not enough.
Looking at the normalized cumulative seismic moment release with time for the hanging-wall and the ATF-Z+ catalogs, we observe that within the periods when significant shallow sequences above the ATF are absent, the percentage moment growth is 3 times higher in the ATF than in the hanging-wall ( Figure S6). As proposed by Anderlini et al. (2016) and Vadacca et al. (2016), we suggest that positive stress accumulation due to ATF creep is most likely released by more favorable oriented shallow splay faults in the hanging-wall (M W > 3 sequences), whose rupture may propagate downdip to the ATF intersection.
The majority of the clusters occurring during the 2013-2014 Gubbio sequence were swarms and foreshock-mainshock sequences, suggesting a combined effect of relaxation and tectonic load that could also be facilitated by the presence of fluids. High pore pressure can induce aseismic creep (e.g., Guglielmi et al., 2015;Wei et al., 2015), and we know about the presence of high CO 2 pressure in ATF footwall observed in kilometers deep boreholes (Collettini & Barchi, 2002;Chiodini et al., 2004). Another signature of the possible combination of tectonic effects with fluids is the presence of fault seismic strands ( Figure 3) interspersed with aseismic zones we observe with the new catalog. These structures could represent small patches of velocity weakening material embedded within primarily velocity strengthening material (e.g., Collettini, 2011;Ross et al., 2020;Schwartz & Rokosky, 2007). The spatial distribution of velocity weakening and velocity strengthening patches on the fault interface (e.g., Hetland & Simons, 2010) is linked to the interseismic coupling, and thermal pressurization may control the slip velocity (e.g., Wibberley & Shimamoto, 2005).
At a larger scale, the presence of ATF portions possessing various frictional properties and then possibly corresponding to sectors undergoing seismic or aseismic deformation has been investigated by Anderlini et al. (2016), based on the mapping of the coefficient of interseismic coupling (ratio of long-term seismic slip rate to the tectonic slip rate) and by inverting GPS data (see map in Figure 4a). Surface deformation observations show that, in the interseismic phase between strong earthquakes, some areas remain locked, and others creep aseismically (e.g., Kaneko et al., 2010). The value of coupling depends on the number, size, and spacing of velocity weakening asperities. Anderlini et al. (2016) found that about half of the ATF surface below 5 km of depth is characterized by creep (e.g., low coupling), absorbing a large part of the regional long-term slip rate (1.7 mm/year), while the remaining portions are locked and then able of generating M > 6 earthquakes. Following Valoroso et al. (2017), we superposed the 2010-2014 ATF-Z+ seismicity to the map of the coupling proposed by Anderlini et al. (2016) and confirm that the creeping regions (low coupling coefficient) are surrounded by clusters and high seismic activity, while the less productive portions are supposed to be locked (Figure 4a).
We also added to the map the typology of clusters, and we observed that the highest number of events is within swarms. The location at the edges of higher coupling patches is confirmed for almost all clusters regardless of their classification and is also reflected in the location of the repeaters highlighted by Valoroso et al. (2017) on the ATF. Figures 4b and 4c also show the cluster type in relation to their maximum magnitude, depth, and duration. Swarms duration (0-120 days) generally increases with the maximum magnitude, and depths range from 3 to 8 km.
The observations made on creep and interseismic coupling suggest some mechanical implications. First, from the microseismicity pattern analysis, we confirm that a mixed distribution of locked and creeping patches seems to characterize the ATF. Second, the ATF is a thick fault zone in which heterogeneous rocks accommodate seismic-aseismic shear slip with different frictional properties. Third, since the ATF also shows a high degree of tectonic coupling with its main antithetic faults, it is likely that microseismicity and aseismic slip on ATF may control the seismic activity on the shallower splays. This ATF specific behavior can be retrieved in other tectonic environments (e.g., Harris, 2017), in laboratory experiments (e.g., Collettini et al., 2011) and in numerical simulations (e.g., van den Ende et al., 2020) and confirm the recent evidence about the control of seismicity by complex fault systems (e.g., Ross et al., 2020). A better understanding of the interplay of the seismic activity within complex fault systems, including segments undergoing both creeping and fast/slow slip earthquakes, seems to be the main road to provide better earthquake hazard characterizations (e.g., Avouac, 2015;Bürgmann, 2018).

Conclusions
The ATF is a creeping low-angle normal fault (15-20 • ) characterized by frequent moderate earthquakes on steeply dipping, shallow secondary faults. From 2010 to 2014, by using a template matching approach, we obtain a fivefold increase in ATF earthquake detections and a decrease of ≈0.5-0.7 of the completeness magnitude. We identify previously hidden features such as the variation of the rate of occurrence of the seismicity along the ATF, including the presence of at least 133 highly productive (more than 10 events) clusters. The majority of clusters behave as fore-mainshock sequences (48%) and swarms (47%); sometimes lasting more than 40 days. Clusters occur around fault patches with a low seismicity rate and higher coupling between the shallow crust and the ATF. Looking at the seismic behavior of ATF in relation with sequences occurring in the high-angle normal faults at shallow depth, we show that the ATF is mainly active at variable intermittent rates and interplaying with moderate earthquakes in the hanging-wall. This intermittency is evident along-strike for fault strands that extend below and at the edges of the shallower faults. As an example, we observe that the ATF shows a rise in the seismic activity during the occurrence of the slow slip associated with the 2013-2014 sequence.

Data Availability Statement
INGV manages the TABOO stations ) -Seismological Data Centre (https://doi.org/10. 13127/SD/X0FXNH7QFY), and data to support this article are available via EIDA (the European Integrated Data Archive infrastructure within ORFEUS) (http://www.orfeus-eu.org/webdc3/). TABOO is a multidisciplinary geophysical observatory of INGV, designed for the study of deformation processes active along with a system of extensional faults in the Northern Apennines and for the identification of earthquake preparation processes. Computations are performed at CINECA in the framework of the HPC-TRES program agreement between OGS and CINECA. PyMPA software is freely available online (https://github.com/avuan/ PyMPA37). The code development is partially supported by the project Seismology and Earthquake Engi-