Slip-rate variations over multiple seismic cycles play a fundamental role in controlling the behaviour of active fault systems, as they are linked to spatio-temporal earthquake clustering and can influence the recurrence patterns of adjacent faults. However, processes that produce slip-rate fluctuations are yet to be fully defined. Despite their importance, the physical mechanisms responsible for such slip-rate fluctuations remain only partially understood. In this study, we investigate whether interactions between neighbouring along-strike brittle faults and their underlying viscous shear zones can generate slip-rate variability associated with synchronous earthquake clustering and fault system synchronization. We focus on nine normal faults and related shear zones within the Central Apennines fault system (Italy), arranged in six along-strike fault pairs characterized by different fault spacings and strike geometries. We integrate cosmogenic 36Cl dating of tectonically exhumed fault scarps with numerical modelling of differential stress transfer between interacting fault–shear-zone pairs. The results identify a mechanism capable of producing simultaneous earthquake clusters, driven by the synchronization of high driving stresses within the viscous shear zones beneath the brittle faults. This behaviour is strongly modulated by along-strike fault spacing and strike variations. In settings with closely spaced fault pairs and limited strike variations, earthquake clusters induce positive differential stress variations on neighbouring shear-zones of sufficient magnitude to induce positive slip-rate variations on their overlying brittle faults. This produces positive feedback mechanism that sustains the occurrence of earthquake clusters that will continue to positively load the neighbouring shear zones. These findings provide new insights into fault system dynamics across multiple timescales and have important implications for seismic hazard evaluation.

Recurrence intervals and magnitude distributions of earthquakes are key parameters in probabilistic and time-dependent seismic hazard assessments, yet they are difficult to constrain because the time window of instrumental and paleoseismic records often capture only a smaller fraction of the earthquake cycle of large earthquakes. Physics-based seismic cycle simulators can help to overcome these limitations by generating synthetic catalogues that span thousands of years, offering valuable insights into the statistical behaviour of fault networks. Despite the increasing use of these simulators, the physical mechanisms governing earthquake timing and size distributions remain incompletely understood, in particular the role of fault interactions and spatial variations in long-term slip rates. Here we use the boundary-element code QDYN to simulate earthquake cycles on normal fault networks of increasing geological complexity, ranging from simplified two-fault configurations to realistic fault networks derived from field data in the Central and Southern Apennines (Italy). Our results show that both fault geometry and slip-rate variability critically influence earthquake recurrence and magnitude distributions. Networks with multiple across-strike interactions produce more complex seismic sequences, irregular recurrence intervals, and broader ranges of rupture sizes and moment magnitudes (Mw) compared to simpler configurations. Similarly, spatially variable slip-rate profiles promote diverse rupture behaviours, including partial ruptures and slow-slip events, that increase variability in stress redistribution, magnitude-frequency relationships and recurrence times. In contrast, models using uniform slip-rate profiles tend to produce regular recurrence patterns and characteristic earthquake magnitudes. These findings highlight the importance of incorporating realistic fault geometries and spatially variable slip rates in physics-based earthquake simulators used to inform seismic hazard assessments.

To undertake fault-based seismic hazard assessment, we need to accurately identify source faults and assess their slip-rate and kinematics through pertinent data collection. For example, converging slip vectors may be used to deduce whether isolated fault strands are connected at depth. Kinematic (slip vector) data can be collected through offset piercing points or from striations preserved on fault scarps. However, striations are surficial features and may therefore be readily eroded and not preserved or visible on degraded scarps. Tensional fracture networks are ubiquitous on bedrock fault scarps and extend deeper into the scarp, and therefore have a greater preservation potential when compared to striations. In this work we characterise fracture-scarp (F-S) lineation patterns across eight faults in Italy (Central Apennines) and Greece (Perachora Peninsula) to explore how these patterns relate to fault plane geometry and slip-vector.Various fracture-scarp (F-S) lineation patterns (including sinistral/dextral en-echelon arrays, slip-parallel/-perpendicular fractures, and conjugate sets) are recognised. These patterns show evidence of progressive growth during exhumation. This suggests F-S lineations formed near the surface as the footwall uplifts, with larger features becoming more connected and smaller ones remaining ‘isolated’. The orientations of F-S lineations align within a pure or Riedel shear geometry where the shear sense is related to the rake of the slip vector. We propose that the observed patterns are controlled by fault plane orientation relative to a 3D strain ellipsoid and the progressive reduction of effective normal stress during footwall exhumation. As fractures form under the same stress regime as striations, they can serve as a kinematic indicator even on highly degraded active fault scarps.

The development of geodetic tools, such as Interferometric Synthetic Aperture Radar (InSAR), has revolutionized our exploration of earthquake physics and the assessment of seismic hazard. Over the past 20 years, InSAR has been increasingly used to determine the interseismic strain rate across major seismogenic faults. Strain derived from geodetically mapped crustal deformation rates serves as an indicator of a fault’s earthquake potential, in alignment with classical elastic rebound theory. However, InSAR observation periods are often relatively short compared to much longer large earthquake recurrence intervals. This raises questions about how well geodetic strain rates represent the long-term strain accumulation on faults. It is therefore critical to understand how strain rate evolves during the interseismic period. We observe the interseismic period prior to the 2021 Mw 7.4 Maduo Earthquake: a left-lateral strike-slip earthquake that ruptured a slow-moving fault approximately 70 km south of the major block-bounding East Kunlun fault in the Eastern Tibetan Plateau. Using six years of Sentinel-1 data, we explore the temporal evolution of strain rate over time. We derive eastward velocity and maximum shear strain rate for the six-year period prior to the Maduo earthquake, before segmenting the time-series and analysing strain rate with a two-year moving time window. Our results indicate that the geodetically derived strain rate may not be constant over the interseismic period, implying that strain may not accumulate at a fixed rate in the seismogenic crust. Additionally, strain rate on the seismogenic fault does not appear to accelerate prior to the Maduo earthquake, at least on the timescales resolvable by InSAR used in this study.

Active faults with low extension rates can generate large magnitude earthquakes with severe damages, as exemplified in the southern Apennines (Italy) by the Irpinia earthquake (Mw 6.8) in 1980 and the Val D’Agri earthquake (Mw 7.1) in 1857. These earthquakes occur within a network of faults, and geological evidence (e.g. paleoseismic trenching) suggest that earthquake activity varies from decennial to millennial time scales on such fault systems. Therefore, improving our understanding and forecasting capabilities of seismic sequences in these areas is crucial. However, studying fault behaviour in slowly deforming regions can often prove challenging due to the long recurrence intervals and low slip rates of these faults, which results in limited instrumental, historical and paleoseismological records. To address this issue, we use physics-based numerical models, since they allow for controlled experiments that can span thousands of years with relatively low computational costs, thus they are valuable tools to investigate the causal dynamics between seismic events. Here, we model a system of NW-SE oriented normal faults in the southern Apennines, accounting for the variable slip rates and geometry of the faults. The study region is characterized by areas with variable number of across-strike faults, thus it is suitable to study the effects of fault network geometry (across-and along-strike interaction) on the seismic cycle and earthquake statistics (e.g. recurrence time, coefficient of variation) of a geologically realistic fault network. We use the boundary-element code QDYN which incorporates rate-and-state friction and elastic interactions to examine relevant inputs for seismic hazard assessment, including inter-event time within and between faults, magnitude-frequency distribution, and nucleation location. We are able to simulate spontaneous ruptures following power-law relationships of frequency-magnitude distribution. Differences in the recurrence time (periodic vs. aperiodic cycles) and rupture extent (characteristic vs. non-characteristic seismicity) in the fault planes seem to correlate with the number of faults that exist across strike. Our simulations demonstrate how quasi-dynamic earthquake simulators can provide insights into how fault network geometry impacts earthquake occurrence and seismic hazard assessment.

Areas of crustal extension often contain pre-existing structures that can reactivate or influence the geometry and growth of new, overlying faults. As strain accumulates, it is well know that new faults may link down-dip with pre-existing faults. Such linkage invariably leads to an increase in fault surface area, which is empirically linked to increasing seismic hazard. However, the timescales over which this linkage may occur, and its effects on throw-rate evolution and related seismic hazard, are poorly constrained. We use high-resolution 3D seismic reflection and borehole data from offshore NW Australia to investigate the growth and throw-rate evolution of two normal faults. By mapping age-constrained seismic horizons and constructing throw-length and -distance plots we show evidence of growth via dip-linkage, along all or part of their mapped length. We find that linkage did not occur simultaneously along the fault, but over a protracted period especially near fault tips. The absolute timing of fault linkage is influenced by the total throw on the pre-existing fault, throw accumulation rates during subsequent rifting, position along the fault, and the intervening stratigraphic thickness. Progressive dip-linkage increases fault surface area and is associated with an increase in throw-rate on linked segments. Given these effects on fault geometry and displacement patterns, we argue that dip-linkage should be integrated into fault growth models and considered in seismic hazard assessments in rifted regions with inherited structures.

We present slip versus time histories derived from in situ 36 Cl cosmogenic dating for three active normal faults in the southern Apennines, Italy. In this region the total extensional strain is accommodated by either a small number of faults located across strike from each other or, in places, a single fault where no other active faults exist across strike. We investigate how strain‐rates on individual faults vary through time in the context of the overall geometry of the fault system. The 36 Cl results confirm that the San Gregorio Magno, Auletta, and Vallo di Diano faults were active in the Holocene, with each fault exhibiting alternating periods of relatively rapid and slow, or even absence of, slip. During periods of rapid slip, lasting a few millennia, the faults accumulate up to ∼5 m of slip, which we interpret as earthquake clusters. At other times, the faults exhibit no slip for time periods lasting multiple millennia. The fluctuations in slip‐rates reveal the migration of activity between faults and out‐of‐phase behavior. Such fluctuations have important consequences for tectonic evolution and crustal rheology, and in particular for hazard estimation because they introduce considerable variability and hence uncertainty in earthquake probability calculations.

We report the first example where the timing of earthquake slip from in situ 36Cl cosmogenic exposure dating of an active normal fault scarp can be verified using independently 14C dated Holocene coastal notches which are deformed along the strike of the fault. We have remodelled 36Cl data from the active Pisia-Skinos normal fault, Greece, published by Mechernich et al. (2018), which indicates that the fault slip rate fluctuated through time. We model the expected coastal uplift and subsidence induced by slip on the fault using elastic half-space models and surface ruptures observed following the 1981 Pisia-Skinos earthquakes. Coastal uplift is constrained by elevation measurements of Holocene coastal notches that have previously been dated using 14C by Pirazzoli et al. (1994) and agree with time periods consistent with Holocene climate stability. We mapped the elevations and numbers of notches along the strike of the Pisia-Skinos fault, including measurements made underwater for locations where fault slip has submerged the notches below the present-day shoreline. We show that the spatial patterns and timing of uplift and subsidence from the notches agrees with the timing of periods of high slip associated with earthquake clusters and quiescence associated with anti-clusters from the slip histories derived from 36Cl data, and with the uplift and subsidence derived from elastic half-space modelling. In particular, where modelled subsidence is highest, Holocene notches that formed between 6-2 ka can be preserved but are submerged. Notches could form at this time because the 36Cl data show that the Pisia fault had entered a period of relative quiescence with a slip-rate of

Earthquakes on normal faults cause negative Coulomb stress transfer (CST) onto receiver faults located across-strike, which, in theory, delays or prevents rupture. Nevertheless, earthquakes on such across-strike faults are frequently observed. This study explores why, and how, large earthquakes can be triggered on faults repeatedly receiving negative coseismic CST. Two triggering mechanisms are hypothesised: (1) Positively stressed patches or segments of faults, resulting from heterogeneous stress transfer, act as triggers to rupture earthquakes on faults that are on average negatively stressed. (2) Negative coseismic CST is compensated by interseismic loading and other processes building up positive stress on the source faults. To test both hypotheses, a 400-year earthquake sequence is modelled, located in the fault network of the Western Anatolian Extensional Province (SW Türkiye). The fault network features multiple active faults located along-strike and across-strike of another, as well as faults in a variety of orientations, suitable to explore stress-triggering mechanisms in a structurally complex setting. Detailed information on fault location, geometry, and mechanism is compiled from field investigations, literature review, and geodetic data. Based on instrumental and historical earthquake catalogues, and the suitability of the source fault network, a sequence of earthquakes is determined to investigate the two hypotheses by comparing the effects of coseismic CST and interseismic loading. Results show that, out of 28 modelled large (MW ≥6) earthquakes, 6 were triggered on faults receiving significant negative coseismic CST. For five of these, negative coseismic CST is compensated by processes increasing CST. Only for one studied example, highly stressed positive fault segments on an otherwise negatively stressed fault could have been the driving mechanism leading to rupture of a large earthquake. Given all model uncertainties, stress-heterogeneities cannot be validated as a probable triggering mechanism for faults in the stress shadow of neighbouring faults. In contrast, earthquakes on normal faults located across-strike of another can only delay, but in most cases not prevent failure, as interseismic loading usually exceeds negative coseismic CST. To reinforce these results, the impact of the modelled fault geometry and slip rates, used to calculate interseismic loading, is evaluated. Models of strike- and dip-variable faults, following the actual surface fault traces, are compared with simplified, planar fault models. Simplified models feature exaggerated areas of positive and negative CST on source faults prior to earthquakes, essentially distorting the stress field and causing stress-heterogeneities that are less pronounced in more realistic models. This observation highlights the necessity of modelling fault geometry as realistic as possible, especially when models are used in fault-based SHA. The impact of slip rates on model results is less drastic, so long as slip rates are used that are determined on similar time scales as the model duration. For the studied fault network short-term (geodetic) and long-term (‘geologic’) slip rates vary from the ‘Holocene’ slip rates by an order of magnitude. If used for modelling interseismic loading, the stress state and recurrence intervals of faults would be drastically under- or overestimated.

Tectonics and active faults are studied using a broad range of techniques and observations, and each of these datasets have both strengths and limitations. Ideally, a multidisciplinary approach should be used when studying active faults, to mitigate against gaps in data and knowledge, and to span the spatial scales of deformation. Furthermore different approaches can provide insights into how faults and fault networks behave over a wide range of timescales, from annual behaviour (e.g. geodesy, seismology) to millennia (e.g. paleoseismology) and millions of years (e.g. seismic reflection). By using a range of techniques to study fault behaviour over a range of timescales, we gain insights into how faults behave and interact, which ultimately can improve our understanding of the resultant seismic hazard.For seismic hazard studies, it is important to quantify fault geometry, dimensions and connectivity as these factors influence the magnitude and propagation of earthquakes. However these are typically difficult to constrain from onshore continental faults where sub-surface information is often limited. Another important aspect to consider for seismic hazard studies is the slip rate of faults, but an aspect that is rarely considered is how slip rates vary spatially and temporally. Using seismic reflection datasets of inactive normal faults, we can study how slip rates vary over far longer timescales than can be considered from field studies alone. While it is challenging to study onshore faults using the same approach, what our findings indicate is that slip rates can vary by more than an order of magnitude over the lifetime of a single fault. Additionally, faults are almost never a single isolated structure, and instead form fault networks, with variable spacing, orientation and lengths. Understanding how a fault network behaves and interacts over time is also important to gain insights into seismic hazard.Ultimately to gain a comprehensive understanding of fault behaviour in time and space, a range of complementary studies, including observations and modelling, are needed to span the broad range spatial and temporal scales that need to be considered when assessing active faults.

This study investigates slip behaviour on overlapping, en echelon normal faults by analysing the slip histories of the Skinos and Pisia active normal faults over the past similar to 20 kyrs using in situ Cl-36 cosmogenic dating. New Cl-36 data from the Skinos Fault and published Pisia Fault Cl-36 data were modelled, with both sample sites located within an overlap zone and separated by an across-strike distance of 1-2 km. Our analysis reveals fluctuating slip rates, with the two faults alternating between out-of-phase and simultaneous slip. The Pisia Fault exhibited a slip rate of similar to 0.5-0.75 mm/yr from similar to 20 ka to similar to 9.6 ka, increasing to similar to 1.25 mm/yr until similar to 5.2 ka. It then slowed to similar to 0.25 mm/yr or less until similar to 2.0 ka, before accelerating again to similar to 1.25-1.5 mm/yr to the present day. The Skinos Fault maintained a low slip rate of similar to 0.25 mm/yr or less from similar to 20 ka to similar to 6.4 ka, before accelerating to similar to 2.0-3.0 mm/yr, persisting to similar to 1.0 ka or possibly the present-day. Comparing their slip histories, the faults show periods of simultaneous slip between similar to 6.4 ka to similar to 5.2 ka and similar to 2.0 ka to similar to 1.0-0.0 ka, and out-of-phase slip occurred between similar to 9.6 ka and similar to 6.4 ka, and from similar to 5.2 ka to similar to 2.0 ka. Out-of-phase behaviour on faults across strike has now been observed on faults spaced across-strike at distances of 1-2 km, 10-20 km, and similar to 100 km, raising the question of why it occurs. Possible mechanism(s), including rheological fluctuations within fault/shear-zone structures linked between the brittle upper crust and viscous lower crust and stress interactions, are discussed to explain the out-of-phase and simultaneous slip behaviour.

Earthquakes can trigger thousands of shallow landslides across mountainous terrain, reshaping landscapes and posing severe hazards. Predicting their spatial distribution remains challenging because most existing models are empirical, event-specific, and lack physical interpretability. We introduce ShallowLandslider, a physics-based component within the open-source Landlab framework for regional coseismic landslide prediction. The model extends the classical Newmark sliding block approach to three dimensions, incorporating transient seismic accelerations, slope geometry, and variable soil properties on structured grids. Instability is assessed using critical acceleration thresholds, and a probabilistic selection scheme represents natural variability in failure occurrence. We validate ShallowLandslider against landslide inventories from two subregions affected by the 2015 Mw 7.8 Gorkha earthquake in Nepal. Model performance is evaluated using non-parametric distributional metrics (Kolmogorov–Smirnov, Kuiper, and Wasserstein distances) across landslide area, elevation, slope, and aspect. Results show that realistic soil-depth parameterisations and moderate cohesion values (10-15 kPa) are essential for reproducing observed topographic clustering and size distributions. While pixel-level prediction remains impractical, ShallowLandslider captures first-order spatial and statistical patterns of coseismic landsliding, offering a reproducible, physically grounded tool for regional hazard assessment. Its modular design enables coupling with other Earth-surface process models, providing a foundation for integrated simulations of landscape response to seismic forcing.

Stress interactions between neighbouring faults plays a key role in controlling earthquake recurrence and size, and therefore in the seismic hazard posed by individual faults within a fault network. In this study, we investigate how differences in the predominant arrangement of faults, specifically, whether it is along-strike or across-strike, affect earthquake recurrence rates and magnitude of earthquakes. To address this topic, we use the boundary-element code QDYN to simulate earthquake cycles of two fault systems within the actively extending region of the Italian Apennines: one to the south where faults are predominantly arranged along-strike, and another in the central Apennines where faults are predominantly arranged across-strike. The different styles of fault network between the Central and Southern Apennines, and high seismic hazard of the region, make this the ideal area to investigate the role of fault geometry on earthquake behaviour across multiple seismic cycles in this region.The models account for variable fault slip rates between faults and network geometry to determine their impact on seismic cycles and earthquake statistics. These simulations produce spontaneous ruptures, with slip modes encompassing full and partial ruptures as well as slow-slip events. We found a good fit between the modelled magnitudes and the ones derived from historical ruptures and empirical relationships. Fault networks with multiple across-strike faults produce more complex seismic sequences, including greater variability in recurrence times and higher proportion of partial ruptures, compared to fault networks with faults arranged predominantly along-strike. Lastly, we assessed the seismic hazard in the studied regions based on the modelled earthquake rates and magnitudes. Our findings show that the spatial distribution of peak ground acceleration corresponding to a 50-year exceedance probability has a greater heterogeneity compared to classical seismic hazard assessment approaches. Hazard levels are elevated in areas where multiple faults overlap, highlighting the influence of fault interactions on regional hazard patterns. These findings show the influence of fault system geometry on how stresses redistribute across multiple earthquake cycles and associated seismic hazard.

Understanding the mechanisms behind the characteristics of earthquake cycles on normal faults is challenging due to their long recurrence times. Despite their moderate magnitude, normal faulting earthquakes can produce considerable damage. We investigate the effects of fault network geometry and spacing on the seismic cycle of two normal faults modeled with rate‐and‐state friction and elastic interactions. Our analysis examines how variable along‐strike and across‐strike distances between faults influence cycle periodicity, synchronicity, nucleation location, magnitude‐frequency distribution, and rupture characteristics. To isolate network‐geometry effects from dimensional and frictional effects, we model faults with a seismogenic width (W) over characteristic nucleation length ( L ∞ ) ratio such that isolated faults produce periodic cycles with a characteristic magnitude (Mw) of 5.1. The cycle periodicity and Mw of earthquakes change depending on the spacing and geometry of the fault network. Faults become less periodic at short across‐strike distances (≤0.2 km). Decreasing the across‐strike spacing leads to variable hypocenter locations and the emergence of partial ruptures, producing magnitudes down to Mw 4.4 at spacings ≤0.2 km. Cycle periodicity and Mw remain unaffected by along‐strike spacing. The long‐term synchronization state of the faults' seismic cycle is influenced differently by across‐strike and along‐strike distances. Closely spaced faults (≤10 km) across‐strike display variable degrees of synchronization, whereas faults arranged along‐strike tend to evolve toward more synchronized states as along‐strike separation decreases. Fault network geometry plays a prominent role, with across‐strike distance having a larger effect on interevent time and rupture style variability than along‐strike distance.

Spatial variations in long-term (e.g. Holocene) slip rates along faults are a source of uncertainty in fault-based seismic hazard assessment (SHA), but their effect on seismicity rates and magnitude-frequency distributions remains underexplored. We conduct numerical simulations of earthquake cycles on the Parasano-Pescina normal fault in Central Italy, using multiple along-strike slip rate measurements, to investigate how spatial variations in slip rate influence slip modes, rupture extent, seismicity rates and magnitudes. We compared synthetic catalogs generated using different profiles with variable slip rates with those based on a single measurement and estimated associated ground-shaking intensities. Profiles with variable slip rate affect earthquake magnitude and recurrence by modulating stress accumulation. Highly variable slip rate profiles lead to more complex rupture patterns, including partial ruptures and slow slip events, contributing to variability in stress distribution in earthquake recurrence, magnitude-frequency distributions and ground-shaking intensities. Simplified slip rate profiles based on a single measurement produce less realistic seismic catalogues with earthquakes of characteristic magnitude and regular recurrence. This highlights the need for detailed profiles to improve SHA, particularly for faults with relatively limited rate-and-state weakening behavior and sharp along-strike gradients in long-term slip rate.

We investigate the Coulomb stress changes due to 30 strong earthquakes occurring on normal faults since 1509 A.D. in Calabria, Italy, including the influence of both coseismic and interseismic loading in our modelling. We compare the results to existing studies of stress interaction from the Central and Southern Apennines, Italy. The three normal fault systems have different geometries and long-term slip-rates. The Central Apennines hosts a complex fault system, with many faults across strike, so that when an earthquake occurs, many of the surrounding faults experience a stress decrease. The Southern Apennines and Calabria have a simpler geometry, with fewer faults, and faults are located predominantly along strike, therefore when an earthquake occurs the dominant process on the neighbouring faults is stress increase. We investigate how stress transfer may influence the occurrence of future earthquakes and what factors may govern the variability in earthquake recurrence in different fault systems. Within the analysed time period, the Calabrian, Central Apennines, and Southern Apennines fault systems have 91%, 73% and 70% of faults with a mean positive cumulative Coulomb stress change, respectively; this is due to fewer faults across strike, more across strike stress reductions, and greater along-strike spacing in the three regions respectively. In regions with close along strike spacing or few faults across strike, such as Calabria and Southern Apennines, the stress loading history is mostly dominated by interseismic loading and most faults are positively stressed before an earthquake occur on them (96% of all faults that ruptured in Calabria; 94% of faults in the Southern Apennines), and some of the strongest earthquakes occur on faults with the highest mean cumulative stress of all faults prior to the earthquake. In the Central Apennines, where across strike interactions are the predominant process, 79% of the earthquakes occur on faults that are positively stressed. The results highlight that fault system geometry plays a central role in characterizing the stress evolution associated with earthquake recurrence, and can possibly influence the occurrence of propagating triggered earthquake sequences.

We present an in situ 36Cl dataset recording the exhumation of 27 active normal fault planes by earthquake slip for the central Apennines, Italy. We do this to constrain the characteristics of earthquake clustering and anticlustering across the entire extending orogen, and in an attempt to constrain the reasons why clustering and anticlustering occurs. We show that duration and magnitude of clustering and anticlustering, and their characteristics, can be explained by a model where the transfer of differential stress between faults and their underlying shear-zones, and between neighbouring fault/shear-zone structures, produces changes in strain-rates on underlying viscous shear zones which drive periods of rapid or reduced slip-rate on their overlying faults. We suggest that stress increase on an underlying shear zone produced by coseismic slip on its overlying fault could be the mechanism that initiates an earthquake cluster. We suggest that stress reductions on shear-zones from coseismic slip located across strike could be the mechanism that initiates an earthquake anticluster. The durations of anticlusters are controlled by the summed stress decreases through time on shear zones, because although these shear zones are slipping relatively slowly, eventually they will load their overlying fault to failure initiating a new cluster, with anticlusters induced across strike. Thus, there is dynamic feedback both up and down dip between faults and their underlying shear zones and crucially across strike between neighbouring fault/shear-zone structures. If the dynamics producing clustering and anticlustering can be constrained, it may be that observations of these phenomena should be included in probabilistic seismic hazard assessments (PSHA) and also interpretations of regional deformation rates and crustal rheologies based on geodetic data. Multi-millennial clustering and anticlustering should become a subject for discussion in these scientific communities.

We report agreement between the timing of slip on an active normal fault recovered from in situ Cl-36 cosmogenic fault scarp dating with independently C-14 dated Holocene coastal notches deformed along the strike of the fault, reinforcing the validity of slip-rate timing and magnitude fluctuations implied by Cl-36 fault scarp dating. The Cl-36-dated Pisia fault, central Greece, shows slip-rate fluctuations but the timing of slip derived from this cosmogenic isotope have not been confirmed with an independent dating approach. However, Holocene coastal notches dated with C-14 on fossils occupying the notches exist around the Pisia fault, these can only form when the interplay between eustatic sea-level and tectonics result in stable relative sea-level. The Cl-36 site close to the center of the Pisia fault records ongoing slip from similar to 9.6 to 5.2 (+/- 0.5) ka and 2.0 +/- 0.5 ka to the present day which was interrupted by a low slip-rate period. Holocene sea-level stabilized close to its current elevation after 7.0-6.5 ka, so the combination of low slip-rate and stable sea-level allowed notch formation. During this time, notches were uplifted by slip on the offshore Strava fault, indicated by elastic half-space modeling. Toward the center of the Pisia-Skinos fault, these notches were then submerged during the high slip period from 2.0 +/- 0.5 ka. Our findings reveal that spatial patterns of deformed radiocarbon-dated Holocene notches agree with the timing of high slip earthquake clusters/quiescent anti-clusters from Cl-36 slip histories and support use of Cl-36 to investigate normal faults, crustal rheologies and seismic hazard.

In this study, we present scaling relationships between fault lengths, fault slip-rates and historical seismicity for an active normal fault system, seismically accommodating crustal extension within the upper plate of the Ionian subduction zone (southern Italy). This crustal extension is confirmed by historical seismicity and instrumental geodesy, with GNSS-derived values of horizonal deformation within a range of 2-3 mm/yr throughout Calabria and the Messina Strait region. We collated data for fault slip-rates, fault lengths and historical earthquakes for a given fault to explore whether fault slip-rates are correlated with fault size and their geometric moment. We present new results showing a robust correlation between fault lengths and fault slip-rates, which supports the idea of a relationship for a given fault between fault slip-rates and the geometric moment. We discuss our results in terms of how these correlations should be used if regional deformation is accommodated by localised strain on faults mostly arranged along strike rather than distributed strain on multiple faults across-strike. For instance, we compare our empirical correlation between fault lengths and fault throw-rates over the Middle-Late Pleistocene in Calabria and the Messina Strait with those from Central and Southern Apennines over the Holocene, characterized by strain distributed on multiple faults across-strike and strain localised on faults mostly arranged along-strike, respectively. Tectonic and seismic hazard implications are discussed for future investigations based on fault slip-rates, fault size and historical seismicity.

Loading and deformation during the interseismic period of the earthquake cycle is often considered to be constant for continental faults, therefore assuming that the short-term (annual-decadal) deformation is representative of longer-term deformation. Based on this assumption, geodetically-derived deformation rates are sometimes used to infer the slip-rates and thus seismic hazard of faults. However geological observations indicate that deformation and slip rates are variable over a range of timescales, and we present an observation of variable deformation across an active normal fault occurring on an annual timescale. The Pisia-Skinos normal fault in the Gulf of Corinth, Greece, is a well-known fault which slipped most recently during a sequence of damaging earthquakes in 1981. Using vertical deformation data, available from the European Ground Motion Service (EGMS), we observe uplift/subsidence of the footwall/hangingwall of the Pisia fault between 2016-2021. Of particular interest is our observation that the deformation is not uniform over the 6 year time period, instead there is an up to 7-fold increase in the vertical deformation rate in mid-2019. We hypothesise that this deformation is aseismic as there is no temporally correlated increase in the earthquake activity (M>1). We explore four possible causative mechanisms for observed deformation; shallow slip, post-seismic after-slip, deep slip on an underlying shear zone, and post-seismic visco-elastic rebound. Our preferred hypothesis is that the transient deformation is caused by centimetre-scale slip in the upper 5km of the Pisia fault zone, based on the magnitude and spatial extent of the deformation. Our results suggest that continental normal faults can exhibit variable deformation over shorter timescales than previously observed, implying that the interseismic period of the earthquake cycle on continental faults may be more variable than previously hypothesised. This also highlights potential pitfalls of using slip rates measured over short-timescales to infer seismic hazard.