Geodetically constrained models of viscoelastic stress transfer and earthquake triggering along the North Anatolian fault

Abstract

Over the past 80 years, 8 M W > 6.7 strike-slip earthquakes west of 408 longitude have ruptured the North Anatolian fault (NAF) from east to west. The series began with the 1939 Erzincan earthquake in eastern Turkey, and the most recent 1999 M W 5 7.4 Izmit earthquake extended the pattern of ruptures into the Sea of Marmara in western Turkey. The mean time between seismic events in this westward progression is 8.5 6 11 years (67% confidence interval), much greater than the timescale of seismic wave propagation (seconds to minutes). The delayed triggering of these earthquakes may be explained by the propagation of earthquake-generated diffusive viscoelastic fronts within the upper mantle that slowly increase the Coulomb failure stress change (DCFS) at adjacent hypocenters. Here we develop three-dimensional stress transfer models with an elastic upper crust coupled to a viscoelastic Burgers rheology mantle. Both the Maxwell (g M 5 4 3 10 18 21 3 10 19 Pa s) and Kelvin (g K 5 1 3 10 18 21 3 10 19 Pa s) viscosities are constrained by studies of geodetic observations before and after the 1999 Izmit earthquake. We combine this geodetically constrained rheological model with the observed sequence of large earthquakes since 1939 to calculate the time evolution of DCFS changes along the North Anatolian fault due to viscoelastic stress transfer. Apparent threshold values of mean DCFS at which the earthquakes in the eight decade sequence occur are between $0.02 to $3.15 MPa and may exceed the magnitude of static DCFS values by as much as 177%. By 2023, we infer that the mean time-dependent stress change along the northern NAF strand in the Marmara Sea near Istanbul, which may have previously ruptured in 1766, may reach the mean apparent time-dependent stress thresholds of the previous NAF earthquakes.

Enabling large‐scale viscoelastic calculations via neural network acceleration

“…These new and efficient computational representations of viscoelastic physics are notable in a number of ways, listed below: Acceleration : With trained ANNs, viscoelastic calculations can be accelerated by ~55,000%. Large‐scale viscoelastic calculations [e.g., DeVries et al , , ] that previously took over 2 million CPU hours could take only ~5 h on an equivalent number of GPUs. With these trained ANNs, massive viscoelastic calculations, across very large spatial and temporal scales and ranges of model parameters, can be done in a matter of minutes or hours. Ease of use and portability : To run forward predictions using the trained ANNs, only the matrices of weights w ij , biases b j , and activation functions are required.…”

“…The distributed architecture of the code (97% efficiency on 1000 cores) facilitates the calculation of viscoelastic effects of earthquakes across realistic fault systems with highly complex geometries and slip histories. However, all told, these calculations may take thousands to millions of CPU‐hours for large‐scale applications [e.g., DeVries et al , , ].…”

“…Computational requirements have often limited the scope of viscoelastic model calculations to a few fixed rheological structures and model geometries over short (<10 years) time periods at a given depth or specific location after a large earthquake [e.g., Lorenzo-Martín et al, 2006;Pollitz and Sacks, 2002;Hearn et al, 2009]. The few studies that consider a large suite of viscosity structures [Takeuchi and Fialko, 2012;DeVries et al, 2016a] require extensive computation time and resources (in the case of DeVries et al [2016aDeVries et al [ , 2016b, 2.7 million CPU hours in total). These computational requirements make global-scale viscoelastic calculations across geometrically complex fault systems and large ranges of model parameters prohibitively expensive.…”

The Role of Viscoelastic Stress Transfer in Long‐Term Earthquake Cascades: Insights After the Central Italy 2016–2017 Seismic Sequence

“…Viscoelastic relaxation of the lower crust and upper mantle has been invoked to explain the delayed triggering of many major earthquakes in California, Japan, and Mongolia (DeVries et al, , and references therein). The coupling between the near surface elastic and brittle crust and the underlying viscous layers has also been shown to play a major role in influencing the spatial pattern of faulting in continental regions (e.g., Verdecchia & Carena, ).…”

Connecting crustal seismicity and earthquake‐driven stress evolution in Southern California

“…Crustal seismicity is thought to be linked to earthquake-driven transient stress changes through numerous mechanisms, including dynamic stresses (during or shortly after seismic wave propagation), static stress changes (instantaneous and long lived), and postseismic stress changes, which are slowly evolving and may arise from a combination of afterslip and viscoelastic relaxation of the lower crust and mantle [Freed, 2005]. The importance of both dynamic stressing and static stress transfer for earthquake triggering is well documented [e.g., Harris, 1998;Stein, 1999;Steacy et al, 2005], while viscoelastic stress transfer has been suggested as an important mechanism in a few case studies [e.g., Freed and Lin, 2001;Zeng, 2001;Pollitz and Sacks, 2002;DeVries et al, 2016]. Toda et al [2005] construct a model of time-dependent seismicity rate in Southern California that accounts for coseismic stress steps following selected M ≥ 5.5 source events, and they relate them to seismicity rate via a rate-and-state friction model.…”