Earthquake-generated tsunamis are generally associated with large submarine events on dip-slip faults, in particular at subduction zone megathrust interfaces e.g.,. The potential generation of a tsunami depends not only on the magnitude of the earthquake but also on the rupture process e.g.,, the geomorphology of the region e.g.,, and secondary effects, such as landsliding or mass slumping . The typically underrepresented tsunami hazard posed by large (partially) submarine strike-slip fault systems has received increasing attention since the unexpected and devastating local tsunami in Palu Bay, following the 2018 Mw7.5 strike-slip Sulawesi earthquake in Indonesia . Assessing the tsunamigenic potential of strike-slip fault systems has important implications worldwide, such as for the Dead Sea Transform fault system, the Enriquillo–Plantain Garden fault zone in Haiti, and for northern offshore sections of the San Andreas fault system in California.

Here, we focus on the ∼ 100 km long Húsavík–Flatey Fault Zone (HFFZ; Fig. ), the largest strike-slip fault in Iceland, which is part of the Tjörnes Fracture Zone (TFZ). The TFZ is a complex transcurrent fault system composed of three main lineaments. It links the Kolbeinsey Ridge (KR) as part of the Mid-Atlantic Ridge offshore north of Iceland (Eyjafjarðaráll Rift Zone) to its manifestation on land in the Northern Volcanic Zone (NVZ), which is characterized by volcanic systems and extensional faulting . Earthquake faulting in the TFZ is driven by eastward spreading of the Eurasian plate, with an average velocity of ∼ 18 mmyr-1 relative to the North American plate . The HFFZ strikes from offshore to onshore and is characterized by right-lateral (dextral) strike-slip faulting, a faulting mechanism which frequently appears subparallel to the adjacent active rift zones of Iceland . It poses the largest threat to coastline communities such as the town of Húsavík, which is located atop the Húsavík–Flatey Fault Zone at the eastern side of Skjálfandi bay.

North Iceland has experienced several large earthquakes in the past. Two magnitude MS∼6.5 earthquakes occurred in 1872 and a recent Mw6 earthquake struck the western end of the HFFZ in 2020 (Fig. ). The largest MS∼7 event in 1755 caused extensive damage, and historic reports indicate that a tsunami hit the coastline and overturned boats . Likewise, such reports include records that the events in 1872 caused rapid sea level changes resulting in a series of waves, i.e., a tsunami-like behavior. High-resolution seismic reflection data within Skjálfandi bay reveal up to 15 m of accumulated vertical offset during the last ∼ 12 000 years , indicating possible vertical deformation of the ocean bottom during past earthquakes. This emphasizes the relevance of studying the Húsavík–Flatey Fault Zone from earthquake rupture to its tsunami potential. estimate that 30 % to 50 % of the full transform motion is taken up by the HFFZ, corresponding to a geodetic slip rate of 6 to 9 mmyr-1. Thus, a locked HFFZ may host potential Mw6.8±0.1 earthquakes . Although the long-term Holocene slip rate is presumably slower than the present-day geodetic slip rate, it can be used to derive an average recurrence time of 500 to 600 years for a Mw7 earthquake on the HFFZ . calculate a recurrence interval of 32 ± 24 years for a magnitude 6 event. Recent velocities obtained from Global Navigation Satellite System (GNSS) measurements – using more than 100 continuous and campaign-style GNSS stations in total – are close to zero near the fault, indicating that the HFFZ may be fully locked .

A better understanding of the complex interaction between static and time-dependent earthquake displacements, off-fault deformation, and seismic, acoustic, and tsunami amplitudes is now possible, using realistic 3D scenarios. Non-linear earthquake dynamic rupture simulations combining coseismic frictional failure on prescribed faults and seismic wave propagation are powerful tools to investigate earthquake dynamics as a consequence of the model's initial conditions e.g.,. Empowered by high-performance computing , joint earthquake–tsunami modeling is now becoming applicable for the development of (probabilistic) tsunami forecasting and early warning systems .

In this study, we investigate the tsunami potential of the HFFZ using two techniques to couple earthquake and tsunami models. First, we apply a one-way linked approach that links the time-dependent seafloor deformation from 3D earthquake dynamic rupture with a subsequent tsunami simulation based on solving the shallow-water equations . Second, we show 3D fully coupled earthquake–tsunami models, which simulate seismic (i.e., elastic), ocean gravity (i.e., tsunami), and compressional ocean acoustic waves simultaneously and self-consistently . We extend six recent dynamic rupture scenarios (Fig. ) from a suite of physics-based dynamic rupture models . The chosen dynamic rupture models vary in their hypocenter location, spatiotemporal evolution of rupture dynamics, fault slip, and geometric fault system complexity. The simple fault geometry rupture models of coincide with historically and physically plausible earthquake magnitudes, stress drop, rupture speed, and slip distributions and produce ground motions that have been verified against empirical ground motion models (GMMs) calibrated for Iceland .

We detail the earthquake and tsunami model setups in Sect. . Section  summarizes the six dynamic rupture earthquake scenarios. In Sect. , we investigate physically plausible scenarios of potentially tsunamigenic HFFZ earthquakes by using the one-way linked earthquake–tsunami modeling approach for all six dynamic rupture scenarios. We show that the HFFZ may generate tsunamigenic earthquakes, potentially posing a significant hazard to coastline communities. Based on the results from the one-way linked simulations, we select the three earthquake–tsunami scenarios on the simpler fault geometry that cause larger wave heights for the fully coupled approach to better understand the initial tsunami genesis and complex superposition of seismic, acoustic, and tsunami waves. We compare the results for both earthquake–tsunami modeling techniques in Sect. .

We present one-way linked (see Sect. ) and fully coupled (see Sect. ) tsunami models that are sourced by earthquakes simulated as dynamically propagating shear rupture on seismically locked pre-existing faults.

We use six earthquake scenarios based on a suite of 3D spontaneous dynamic rupture simulations developed in that can match local GMMs and reproduce historic earthquake magnitudes. Dynamic rupture modeling includes solving for the spontaneous frictional failure non-linearly linked to the propagation of seismic waves (Fig. ) with the purpose of gaining knowledge about the underlying physical processes. Such physically self-consistent descriptions of how faults yield and slide have been developed for complex and/or poorly instrumented earthquakes in various tectonic contexts e.g.,. In contrast to kinematic earthquake source modeling, fault slip is not prescribed, but the rupture dynamics evolve based on an empirical friction law and chosen initial conditions. Here, the initial conditions of the dynamic rupture models, including fault geometries, pre-stress, and fault strength, are constrained by seismic, geodetic, and bathymetry observations, as briefly summarized in the following sections. For details and a sensitivity analysis of the HFFZ dynamic rupture simulations and their initial conditions, we refer to .

Submarine ruptures across strike-slip fault systems were long assumed to produce only minor vertical offsets and hence no significant disturbance of the water column. Linked and fully coupled earthquake dynamic rupture and tsunami modeling for the 2018 Mw7.5 Sulawesi earthquake in Indonesia suggest that coseismic-induced seafloor displacements critically contributed to the generation of an unexpected and devastating local tsunami in Palu Bay e.g.,. Widespread liquefaction-induced coastal and submarine landslides likely also played an important role e.g.,. Our simulations of six earthquake dynamic rupture scenarios show that the Húsavík–Flatey Fault Zone can host tsunamigenic earthquakes. This may have important implications for tsunami hazard assessment of submarine strike-slip fault systems in transform and transtensional tectonic settings worldwide. For example, the North Alfeo Fault in the Ionian Sea may be capable of generating a Mw≈7 strike-slip earthquake but is often not considered in tsunami modeling.

Our earthquake dynamic rupture scenarios can generate enough vertical seafloor displacements to source a localized tsunami. The scenarios on the simpler fault geometry may be considered worst-case events because the ruptures break over the entire main fault length, accumulating large fault slip (equivalent to ∼Mw7.3). The moment magnitudes of our dynamic rupture models on the complex fault geometry are lower (Mw6.7–Mw7.0) and involve more segmented slip due to rupture-jumping across the highly segmented fault network. Large-scale geometric fault complexity can act as an “earthquake gate” . An earthquake gate is a mechanical barrier to earthquake rupture , which has the potential to alter the rupture extent. In the case of the simpler fault geometry, the restraining and releasing bend east of Flatey island may be considered to be such an earthquake gate, since this smooth main fault bend does allow some ruptures to propagate across while terminating others depending on the local pre-stress and dynamic stress evolution . The segmented, more complex fault geometry is more effective at dynamically arresting earthquake rupture . However, large earthquakes are still dynamically possible with rupture capable of jumping across several fault stepovers before eventually terminating.

For all scenarios, we observe pronounced dynamic rake rotation near the surface, which we consider to be a plausible dynamic mechanism for generating increased coseismic vertical offset. The dynamic deviations from pure right-lateral strike-slip faulting are on the order of ± 20∘ and introduce significant shallow dip-slip motion. Thereby, vertical seafloor displacements in our simulations are enhanced, which are critical for tsunami generation. Shallow rake rotation has been inferred for surface-breaking earthquakes using geological slickenlines and simple dynamic rupture models . Vertical stress changes at the rupture front cause this change in rake angle, which is more pronounced near the surface due to smaller confining stresses . No rake rotation is expected directly atop the hypocenter, which is confirmed in Fig. . In Fig. , a gradual increase in the rake rotation can be observed away from the hypocenter for both unilateral and bilateral rupture scenarios on the simpler fault geometry. More complex patterns of rake rotation result in the scenarios on the complex fault geometry, and we observe a dependence of the spatial distribution of rake rotation on the fault segment length and on hypocenter location. Changes in rake for the right-lateral strike-slip earthquake dynamic rupture scenarios cause mostly uplift in the compressional quadrants. investigate the 1971 San Fernando thrust faulting earthquake using aerial stereo photographs and discuss a rotation of rake away from the pre-stress direction. use InSAR, GNSS, and optical data to study the 2019 Ridgecrest Sequence and report vertical cumulative coseismic surface displacement after the sequence, interpreted as an indication of prominent coseismic rake rotation. Other studies focusing on finite-fault models, allowing for rake variations in their inversions for slip, show rake rotation most prominently in patches which are near the surface and have a large slip magnitude .

In contrast to the suggested important contribution of off-fault deformation to strike-slip tsunami generation in Palu Bay , the effect of off-fault plasticity is likely small in our simulations (Fig. ). We find that off-fault deformation contributes only about ∼ 3 % of the total seismic moment. Accounting for the potential existence of additional shallow, weak sediments, which are more prone to off-fault plastic deformation, may increase local uplift .

Modeling tsunami scenarios for hazard assessment or rapidly after submarine earthquakes often relies on simplifications, such as the negligence of source time-dependency, only considering vertical seafloor deformation without bathymetry effects, solely planar fault geometries, or neglecting tsunami dispersion and acoustic wave effects. introduce non-dimensional parameters allowing us to quantify the validity of certain modeling assumptions. Our average water depth H can be approximated as ∼ 200 m, the source width σr as given by the length of the HFFZ (∼ 100 km); the source duration σt of 30 s constrained by the rupture duration (cf. moment rates Fig. ); the gravitational acceleration g = 10ms-2; and acoustic wave speed c0 = 1500 ms-1. Based on these approximations, we can calculate the three non-dimensional numbers posed by , as specified in Table .

We see that the shallow water limit is fulfilled (H/σr≪1), which justifies using our one-way linked earthquake–tsunami modeling approach. While we use time-dependent seafloor displacements, our source should appear effectively as instantaneous to tsunami waves (gH⋅σt/σr≪1) due to the relatively short rupture duration and shallow water depth. This fact explains the similarity in the tsunami propagation and shape of the tsunami wavefronts for the simple fault geometry scenarios (i.e., Fig.  after 2 min (second column) and 10 min (third column)), which all break the entire main fault length and lead to similar fault slip distributions. In contrast, the scenarios on the complex fault geometry differ distinctly in their final fault slip distributions and areas of seafloor displacements, depending on the chosen epicenter location. However, to compare the tsunami generation phase, it is indispensable to consider a time-dependent source model for both approaches, the one-way linked and fully coupled method, for comparability.

From the average water depth H being much smaller than c0⋅σt (Table ), we expect that it is justified to neglect acoustic wave excitation, since their amplitudes should be small. However, our fully coupled simulations include acoustic wave generation with high vertical velocity amplitudes that are larger than the tsunami signals (Figs. , , and ). Dynamic rupture reaching the Earth's surface can cause strong radiation , including the generation of high-frequency seismic waves due to the locally strong deceleration at the rupture front e.g.,. Part of this seismic wave energy is converted to ocean acoustic waves at the seafloor e.g.,, as observed during the 2011 Tōhoku-oki e.g., and the 2003 Tokachi-oki earthquakes e.g., using ocean-bottom pressure sensors. Earlier studies found that the conversion between seismic and ocean acoustic waves occurs predominantly at slopes of the seafloor e.g.,. Here, however, local bathymetry is generally flat, and the conversion is dominated by dynamic source complexity, such as surface rupture and the associated shallow rake rotation.

In our study, we compare a simple fault geometry representing the Húsavík–Flatey Fault Zone and a very complex fault network consisting of 55 individual fault segments. and proposed a linear relationship between the thickness of the seismogenic crust (brittle upper crust) and the length of fault segments for strike-slip geometries. This would imply a relatively short average fault segment length, given the locking depth of 6–10 km for the HFFZ . Therefore, during a large strike-slip earthquake along the HFFZ, the rupture may segment into several subevents , as resembled in our Complex-Middle scenario. However, Iceland, offering a unique geologic complexity, is located atop the Mid-Atlantic Ridge and influenced by the underlying mantle plume , with significantly varying crustal thickness over the last 56 Ma , potentially altering established scaling relation. Hence, it is important to include different fault structural complexity within earthquake and tsunami simulations to accurately capture plausible earthquake scenarios.

The three tsunami scenarios sourced by dynamic rupture simulations across the complex fault geometry cause significantly smaller tsunamis. This is due to lower and more segmented fault slip, leading to less vertical seafloor displacements which are spatially more restricted. The largest total wave height (i.e., crest to trough difference) of ∼ 40 cm is observed at the synthetic tide gauge stations near Grímsey island for scenario Complex-Middle and near Ólafsfjörður for Complex-West. The tsunami from scenario Complex-East does not have a significant impact on the virtual tide gauges, since much of the coseismic ground displacement occurs onshore. The town Ólafsfjörður is also highly exposed to tsunami signals in the scenarios using the simpler fault geometry, with wave heights reaching 0.9 m.

We find that our scenario Simple-East poses the largest impact for coastal communities, except for Húsavík. Here the hypocenter is near the town, which may experience strong ground shaking but not a large tsunami. However, Húsavík can be affected by scenario Simple-West, causing nearly 60 cm crest to trough difference. This unilateral rupture nucleating at the western end of the HFFZ builds up energy while propagating towards Húsavík, explaining the larger observations. We find that none of our scenarios endanger the town of Akureyri, which is shielded by the narrow Eyjafjörður. The modeled tsunami does not amplify but loses energy due to multiple reflections within the bay and due to the protection by Hrísey island.

performed a preliminary investigation of the tsunami potential for the Húsavík–Flatey Fault Zone using a uniform fault slip earthquake dislocation source with a moment magnitude of 7.0, located in the middle of the fault system. They utilized the Okada method , with instantaneous sourcing of the tsunami by the final static displacements. Their maximum synthetic crest to trough difference of ∼ 30 cm also occurs at Ólafsfjörður. While this is slightly larger than the maximum crest to trough difference of 26 cm, which we observe for the scenario Complex-Middle, it is a factor of 2.5 smaller than our scenario Simple-Middle (77 cm).

These differences may be due to our dynamic rupture models including dynamically evolving relatively large shallow fault slip (up to ∼ 8 m for Simple-East), with no SSD for scenarios on the simpler fault geometry and near-surface rake rotation (± 20∘). This results in higher-than-expected coseismic vertical displacements (± 1 m). In addition, we include local bathymetry and, to a smaller extent, off-fault plastic deformation, all contributing to the tsunami generation. We do not consider combined earthquake and landslide-induced tsunami scenarios for the HFFZ, which can additionally increase the local tsunami height .

We extend recent 3D dynamic rupture models by to tsunami modeling. In these scenarios, relatively high peak fault slip localizes near the free surface, while the average fault slip is overall ∼ 40 % smaller than the peak fault slip (Table ). show that the synthetic ground motions produced by such dynamic rupture models are in good agreement with the latest regional empirical ground motion model . While the moment magnitudes of some of these models are larger than previous slip-deficit-based estimates of the accumulated moment along the HFFZ (e.g., Mw6.8 ± 0.1; ), they have comparable moment magnitudes to historic events. Slip-deficit-based magnitude estimates are typically relying on several assumptions, including, in this case, a complete stress relaxation of the 1872 Mw6.5 earthquake and subsequent steady stress accumulation. The dynamic rupture scenarios are consistent with the average fault slip and effective rupture area scaling relations of , which have recently been validated for the Southern Iceland Seismic Zone (SISZ; ). The SISZ is similar in its tectonic and seismic context to the TFZ in northern Iceland.

Our simulated maximum fault slip occurring within the shallow offshore part of the HFFZ, i.e., the part which is relevant for the subsequent tsunami generation, is comparable to geological observations from earthquakes rupturing along faults with similar length to the HFFZ. Examples of strike-slip ruptures as summarized by comprise Neodani, Japan (length 80 km, max slip 7.9 m; ); Luzon, Philippines (length 112 km, max slip 6.2 m; ); and Landers, California (length 77 km, max slip 6.7 m; ). A recent example includes the second event of the devastating Kahramanmaraş earthquake sequence (length 150 km), resulting in up to 8 m fault slip near the surface . While we use a LSW friction law, considering a rate-and-state-dependent friction law would allow accounting for shallow velocity-strengthening behavior, which may decrease slip in the shallowest parts of the fault e.g.,.

We exclude inundation in the one-way linked approach to enable a more meaningful comparison with the fully coupled method. Our fully coupled simulations are computationally demanding and do not allow us to model inundation e.g.,. We show that the fault geometry in our six one-way linked scenarios can influence the subsequent tsunami generation. Here, future studies may explore potential variations in fault dip, which may further enhance the vertical seafloor displacement during the earthquake rupture. Changes in earthquake source parameters are known to affect the maximum tsunami height e.g., and resulting inundation e.g.,.

Our 3D fully coupled simulations include unexpectedly high-amplitude acoustic waves, which may serve as a rapid indicator of surface dynamic rupture. A better understanding of such acoustic wave signals may improve tsunami early warning, since these can be detected earlier in the far-field, e.g., at ocean-bottom pressure sensors, in comparison to the tsunami recorded at conventional DART buoys . However, in the near-field, ocean acoustic waves can superimpose onto tsunami signals, impeding early warning efforts. In addition to the seismo-acoustic wave excitation in the fully coupled simulations, we observe dispersion of tsunami propagation velocity . show that enhanced dispersion effects are expected for earthquakes with moment magnitude 8 and less as opposed to less dispersive tsunamis caused by the largest earthquakes. We do not account for dispersion in the one-way linked tsunami simulations that are based on solving the non-linear hydrostatic shallow water equations. However, Boussinesq-type tsunami models can account for dispersion e.g.,. Accounting for dispersion effects can be important if the resulting series of excited oceanic waves locally interfere constructively and amplify, which has been observed in tsunami scenarios of the South China Sea and outer-rise normal faults . Here, we do not detect significant differences in wave height or tsunami arrival times compared to our one-way linked scenarios, despite dispersion effects. Likely reasons include the relatively shallow water depth and the close proximity of the HFFZ to the coast, preventing interferences.

We present a suite of realistic earthquake–tsunami scenarios for North Iceland, comparing one-way linked and 3D fully coupled modeling techniques. Both approaches agree in the resulting tsunamis from strike-slip earthquake dynamic rupture scenarios on the Húsavík–Flatey Fault Zone. We investigate two distinct fault system geometries – a simpler fault geometry with three fault segments and a highly complex fault system composed of 55 fault segments – to represent the 100 km long Húsavík–Flatey Fault Zone striking from onshore to offshore. Our study showcases how dynamic earthquake source mechanisms, including dynamic rupture rake rotation near the surface (of ± 20∘) combined with shallow fault slip, cause coseismic vertical displacement of the order of ± 1 m and the generation of high-amplitude acoustic waves without strong bathymetric slopes. We find that our earthquake–tsunami scenarios in a less segmented fault system, in particular with a hypocenter in the east near the town of Húsavík, generate the largest wave heights of ∼ 0.9 m near the local community of Ólafsfjörður. Húsavík is the only town that is more affected by a scenario with a hypocenter in the west of the HFFZ, causing a maximum tsunami crest to trough difference of ∼ 0.4 m. None of our scenarios, regardless of the source complexity, endangers the town Akureyri, which is shielded by its narrow Eyjafjörður Fjord from the coseismically sourced tsunami. The 3D fully coupled scenarios include source dynamics, seismic, acoustic, and tsunami waves and result in complexities not present in our one-way linked simulations. We observe the excitation of tsunami normal dispersion and unexpectedly large acoustic waves, which may serve as a rapid indicator of surface-breaking dynamic rupture. Our findings highlight the importance of considering tsunamigenic strike-slip earthquakes in tsunami hazard assessment. Accounting for the dynamics of earthquake source effects and fully coupled tsunami generation may be useful to enhance tsunami hazard assessment and facilitate improvements to early warning systems.