Finite-fault Inversion Research Articles

Efficient Bayesian uncertainty estimation in linear finite fault inversion with positivity constraints by employing a log-normal prior

Efficient Bayesian uncertainty estimation in linear finite fault inversion with positivity constraints by employing a log-normal prior

Multi-Data-Type Source Estimation for the 1992 Flores Earthquake and Tsunami

We revisit the source of the 1992 Flores earthquake and tsunami using finite-fault inversion. We simultaneously invert teleseismic body and surface waves together with coseismic uplift/subsidence datasets. We then verify the inverted source against tsunami run-up heights along the northern coast of Flores Island and the only tide gauge recording of the tsunami. Our preferred source model provides a good fit to all the datasets, whereas previous models only explained a subset of the available data. We show that the fault geometry implies segmentation of the back-arc thrust system in the eastern Sunda Arc.

Teleseismic finite-fault inversion of two Mw = 6.4 earthquakes along the East Anatolian Fault Zone in Turkey: the 1998 Adana and 2003 Bingöl earthquakes

The East Anatolian Fault Zone is a continental transform fault accommodating westward motion of the Anatolian fault. This study aims to investigate the source properties of two moderately large and damaging earthquakes which occurred along the transform fault in the last two decades using the teleseismic broadband P and SH body waveforms. The first earthquake, the 27 June 1998 Adana earthquake, occurred beneath the Adana basin, located close to the eastern extreme of Turkey’s Mediterranean coast. The faulting associated with the 1998 Adana earthquake is unilateral to the NE and confined to depths below 15 km with a length of 30 km along the strike (53°) and a dipping of 81° SE. The fixed-rake models fit the data less well than the variable-rake model. The main slip area centered at depth of about 27 km and to the NE of the hypocenter, covering a circular area of 10 km in diameter with a peak slip of about 60 cm. The slip model yields a seismic moment of 3.5 × 1018 N-m (Mw ≅ 6.4). The second earthquake, the 1 May 2003 Bingol earthquake, occurred along a dextral conjugate fault of the East Anatolian Fault Zone. The preferred slip model with a seismic moment of 4.1 × 1018 N-m (Mw ≅ 6.4) suggests that the rupture was unilateral toward SE and was controlled by a failure of large asperity roughly circular in shape and centered at a depth of 5 km with peak displacement of about 55 cm. Our results suggest that the 1998 Adana earthquake did not occur on the mapped Goksun Yakapinar Fault Zone but rather on a SE dipping unmapped fault that may be a split fault of it and buried under the thick (about 6 km) deposits of the Adana basin. For the 2003 Bingol earthquake, the final slip model requires a rupture plane having 15° different strike than the most possible mapped fault.

Using Tsunami Waves Reflected at the Coast to Improve Offshore Earthquake Source Parameters: Application to the 2016 Mw 7.1 Te Araroa Earthquake, New Zealand

AbstractThe Te Araroa earthquake occurred on 2 September 2016 (local time) offshore of the northeastern coast of New Zealand's North Island (Mw 7.1). When this event occurred, ocean bottom pressure gauges (OBPs), installed ~170 km south of the source area, clearly recorded direct tsunami from the source to OBPs (~ −1.5 cm), and tsunami from coastal reflections (~2 cm). We estimate the centroid location that best reproduces the OBP waveforms. When using the direct wave alone, the centroid location is poorly constrained, with a horizontal uncertainty of ~100 km. By combining both direct and reflected tsunami waveforms, we obtain a centroid location near the Global Centroid Moment Tensor centroid (~80 km northeast from the coast) with smaller uncertainty (~40 km). We also estimate the earthquake source dimension (length and width) and found that the models using coastal reflections require a source dimension larger than ~30 km long. Based on the slip distribution obtained by the finite fault inversion, we obtain an energy‐based stress drop ΔσE of 1.0 MPa, consistent with typical earthquake stress drop values. This study shows that the information added by coastal reflected tsunami provides much tighter constraints on the centroid location, source dimension, and stress drop of offshore earthquakes, which is difficult to obtain from the onshore seismic data alone. Future studies should utilize the information provided by coastal reflected waves to improve earthquake source modeling using ocean bottom pressure data.

Seismic source characteristics of the intraslab 2017 Chiapas-Mexico earthquake (Mw8.2)

Inversion of the parameters characterising the seismic source of the instraslab 2017 Chiapas Mexico earthquake (Mw 8.2) shows a simple rupture process with a unidirectional propagation and directivity towards the North-West and a duration of the rupture process around 75 s. The initial point source values of strike, dip and rake are 316°, 80° and −91° respectively. The focal mechanism indicates a normal fault type within the oceanic Cocos plate, with an almost vertical fault plane for a focal depth of 59 km. The seismic data was obtained from 51 seismic stations of the global seismic network IRIS for the epicentral distances between 30° and 90°. In the finite-fault inversion, 75 seismic signals between P and SH waves were used. The epicenter is on the southeast margin of the large slip zone which extends 75 km to the northwest, this large slip zone is located to the south of the city of Arriaga. The scalar seismic moment was estimated at 2.55×1021Nm, equivalent to a moment magnitude of Mw 8.2. The maximum dislocation or slip is 14.5 m. As a coseismic effect, a local tsunami was generated, recorded by several tidal gauges and offshore buoys. The deformation pattern shows a coastal uplift and subsidence.

Resolving Horizontal Rupture Directivity of Moderate Crustal Earthquake in Sparse Network With Ambient Noise Location

AbstractEarthquake rupture directivity can remove fault plane ambiguity in earthquake moment tensor solutions; thus, it is helpful to improve reliability of shake map and understanding the seismogenic processes. In this paper, we propose an algorithm to resolve the horizontal rupture directivity of moderate crustal earthquakes in two steps. The first step is to determine the two candidate fault planes from focal mechanism inversion. Then, rupture directivity is estimated via the difference between the hypocenter and the centroid location which is constrained on the candidate fault planes. Accurate centroid location is determined with ambient noise location method, in which the noise cross‐correlation functions (NCFs) provide a stable calibration for the path effects. This technique is applied to the 2011 Oklahoma earthquake, and the estimated centroid location is about 2–4 km southwest to the hypocenter, indicating that the mainshock ruptured along the 234° plane for about 4–8 km. The result is consistent with directivity estimation using a foreshock as reference event and also agrees with early aftershock distribution as well as finite fault inversion. More applications to the 2004 Parkfield earthquake and the 2014 Napa earthquake are performed to explore the effectiveness and limitation of this method. This method is helpful for determining rupture directivity of earthquakes in sparse network but requiring one or two nearby stations. Alternatively, temporary stations can be installed in the area of high seismic hazard, then a library of NCFs and path calibrations would be available for rapidly determining rupture directivity of future earthquakes.

The 12 June 2017 Mw 6.3 Lesvos Island (Aegean Sea) earthquake: Slip model and directivity estimated with finite-fault inversion

On 12 June 2017 (UTC 12:28:38.26) a magnitude Mw 6.3 earthquake occurred offshore Lesvos Island in SE Aegean Sea, which was widely felt, caused 1 fatality, and partially ruined the village of Vrisa on the south-eastern coast of the island. I invert broad band and strong motion waveforms from regional stations to obtain the source model and the distribution of slip onto the fault plane. The hypocentre is located at a depth of 7km in the upper crust. The mainshock ruptured a WNW-ESE striking, SW dipping, normal fault, projecting offshore and bounding the Lesvos Basin. The strongest and most aftershocks clustered away from the hypocentre, at the eastern edge of the activated area. This cluster indicates the activation of a different fault segment, exhibiting sinistral strike-slip motions, along a plane striking WNW-ESE. The slip of the mainshock is confined in a single large asperity, WNW from the hypocentre, with dimensions 20km×10km along fault strike and dip, respectively. The average slip of the asperity is ~50cm and the peak slip is ~1m. The rupture propagated unilaterally towards WNW to the coastline of Lesvos island at a relatively high speed (~3.1km/s). The imaged slip model and forward modelling was used to calculate peak ground velocities (PGVs) in the near-field. The damage pattern produced by this earthquake, especially in the village of Vrisa is compatible with the combined effect of rupture directivity, proximity to the slip patch and the fault edge, spectral content of motions, and local site conditions.

The effects of core-reflected waves on finite fault inversions with teleseismic body wave data

The effects of core-reflected waves on finite fault inversions with teleseismic body wave data

THE 2014 MW 6.9 NORTH AEGEAN TROUGH (NAT) EARTHQUAKE: SEISMOLOGICAL AND GEODETIC EVIDENCE

A strong earthquake (Mw 6.9) on 24 May 2014 ruptured the North Aegean Trough (NAT) in Greece, west of the North Anatolian Fault Zone (NAFZ). In order to provide unbiased constrains of the rupture process and fault geometry of the earthquake, seismological and geodetic data were analyzed independently. First, based on teleseismic long-period P- and SH- waveforms a point-source solution yielded dominantly right-lateral strike-slip faulting mechanism. Furthermore, finite fault inversion of broad-band data revealed the slip history of the earthquake. Second, GPS slip vectors derived from 11 permanent GPS stations uniformly distributed around the meizoseismal area of the earthquake indicated significant horizontal coseismic slip. Inversion of GPS-derived displacements on the basis of Okada model and using the new TOPological INVersion (TOPINV) algorithm permitted to model a vertical strike slip fault, consistent with that derived from seismological data. Obtained results are consistent with the NAT structure and constrain well the fault geometry and the dynamics of the 2014 earthquake. The latter seems to fill a gap in seismicity along the NAT in the last 50 years, but seems not to have a direct relationship with the sequence of recent faulting farther east, along the NAFZ.

Rupture features of the 2010 Mw 8.8 Chile earthquake extracted from surface waves

This study used the rupture directivity theory to derive the fault parameters of the 2010 Mw 8.8 Chile earthquake on the basis of the azimuth-dependent source duration obtained from the Rayleigh-wave phase velocity. Results revealed that the 2010 Chile earthquake featured asymmetric bilateral faulting. The two rupture directions were N171°E (northward) and N17°E (southward), with rupture lengths of approximately 313 and 118 km, respectively, and were related to the locking degree in the source region. The entire source duration was approximately 187 s. After excluding the rise time from the source duration, the northward rupture velocity was approximately 2.02 km/s, faster than the southward rupture velocity (1.74 km/s). On average, the rupture velocity derived from this study was slower than that estimated from finite-fault inversion; however, several historical earthquakes in the Chile region also showed slow rupture velocity when using low-frequency signals, as surface waves do. Two earlier studies through global-positioning-system data analysis showed that the static stress drop of 50–70 bars for the 2010 Chile earthquake was higher than that for subduction-zone earthquakes. Hence, a remarkable feature was that the 2010 Chile earthquake had a slow rupture velocity and a high static stress drop, which suggested an inverse relationship between rupture velocity and static stress drop.Graphical abstract.

The 10 June 2012 MW 6.0 earthquake sequence in the easternmost end of the Hellenic arc.

The 10 June 2012 (UTC 12:44:17.3; lat. 36.441°N, long. 28.904°E, Mw6.0) earthquake sequence, 60 km to the west of Rodos Island, is studied, in an attempt to shed light to the obscure deformation pattern at the easternmost end of the Hellenic Arc. Moment tensor solutions for the mainshock and the strongest aftershocks revealed the operation of WNW-ESE dextral strike-slip faulting, with slip vector at~N295°E, approximately orthogonal to the GPS velocity vectors. The strike of the activated structure generally aligns with bathymetric linear escarpments observed in the region, bordering the eastern section of the Rodos basin. The best constrained focal depths are in the range 10 to 25 km, with the mainshock at the depth of 24 km.The slip model for the mainshock, obtained through a finite-fault inversion scheme, showed that slip was mainly concentrated in a single patch, with the locus of peak slip (~125 cm) located ~ 4km to the NW of the hypocenter. The sequence which lies in the western continuation of the Fethiye – Burdur sinistral strike-slip zone into theAegean Sea and Rodos basin, is not connected with activation of this zone. Its characteristics comply with the activation of a dextral strike-slip structure, oblique to this zone, which accommodates along – arc NE-SW extension.

Exploring the uncertainty range of coseismic stress drop estimations of large earthquakes using finite fault inversions

Exploring the uncertainty range of coseismic stress drop estimations of large earthquakes using finite fault inversions

Rapid automated W‐phase slip inversion for the Illapel great earthquake (2015, Mw = 8.3)

AbstractWe perform rapid W‐phase finite fault inversion for the 2015 Illapel great earthquake (Mw = 8.3). To evaluate the performance of the inversion in a near real time context, we divide seismic stations into four groups. The groups consider stations up to epicentral distances of 30°, 50°, 75°, and 90°, respectively. The results for the first group could have been available within 25 min after the origin time and the results for the last group within 1 h. The four results consistently show a peak slip of ∼10 m near the trench with trench perpendicular rake which is consistent with the tsunami genesis of the event. The slip location is similar to that in the preliminary U.S. Geological Survey solution. The inversion is automated and provides meaningful results within 25 min after the event. This makes the method particularly suited to emergency management and early warning at regional and teletsunami distances.

Fault slip source models for the 2014Mw6.9 Samothraki‐Gökçeada earthquake (North Aegean Trough) combining geodetic and seismological observations

AbstractThe 24 May 2014,Mw6.9, Samothraki‐Gökçeada shallow (depth: 11 km) earthquake along the North Aegean Trough (NAT), at the westward extension of the North Anatolian Fault Zone (NAFZ), is investigated using constraints from seismological and geodetic data. A point source solution based on teleseismic long‐periodPandSHwaveforms suggests an essentially strike‐slip faulting mechanism consisting of two subevents, while from a finite fault inversion of broadband data the rupture area and slip history were estimated. Analysis of data from 11 permanent GPS stations indicated significant coseismic horizontal displacement but no significant vertical or postseismic slip. Okada‐type inversion of horizontal slip vectors, using the new TOPological INVersion algorithm, allowed precise modeling of the fault rupture both as single and preferably as double strike‐slip faulting reaching the surface. Variable slip models were also computed. The independent seismological and geodetic fault rupture models are broadly consistent with each other and with structural and seismological data and indicate reactivation of two adjacent fault segments separated by a bend of the NAT. The 2014 earthquake was associated with remote clusters of low‐magnitude aftershocks, produced low accelerations, and filled a gap in seismicity along the NAT in the last 50 years; faulting in the NAT seems not directly related to the sequence of recent faulting farther east, along the NAFZ and the seismic gap in the Marmara Sea near Istanbul.

Rapidly Estimated Seismic Source Parameters for the 16 September 2015 Illapel, Chile M w 8.3 Earthquake

On 16 September 2015, a great (M w 8.3) interplate thrust earthquake ruptured offshore Illapel, Chile, producing a 4.7-m local tsunami. The last major rupture in the region was a 1943M S 7.9 event. Seismic methods for rapidly characterizing the source process, of value for tsunami warning, were applied. The source moment tensor could be obtained robustly by W-phase inversion both within minutes (Chilean researchers had a good solution using regional data within 5 min) and within an hour using broadband seismic data. Short-period teleseismic P wave back-projections indicate northward rupture expansion from the hypocenter at a modest rupture expansion velocity of 1.5–2.0 km/s. Finite-fault inversions of teleseismic P and SH waves using that range of rupture velocities and a range of dips from 16°, consistent with the local slab geometry and some moment tensor solutions, to 22°, consistent with long-period moment tensor inversions, indicate a 180- to 240-km bilateral along-strike rupture zone with larger slip northwest to north of the epicenter (with peak slip of 7–10 m). Using a shallower fault model dip shifts slip seaward toward the trench, while a steeper dip moves it closer to the coastline. Slip separates into two patches as assumed rupture velocity increases. In all cases, localized ~5 m slip extends down-dip below the coast north of the epicenter. The seismic moment estimates for the range of faulting parameters considered vary from 3.7 × 1021Nm(dip 16°) to 2.7 × 1021 Nm (dip 22°), the static stress drop estimates range from 2.6 to 3.5 MPa, and the radiated seismic energy, up to 1 Hz, is about 2.2–3.15 × 1016 J.

The isolated ∼680 km deep 30 May 2015 MW 7.9 Ogasawara (Bonin) Islands earthquake

Deep-focus earthquakes, located in very high-pressure conditions 300 to 700 km below the Earth's surface within sinking slabs of relatively cold oceanic lithosphere, are mysterious phenomena. The largest recorded deep-focus earthquake (MW 7.9) in the Izu–Bonin slab struck on 30 May 2015 beneath the Ogasawara (Bonin) Islands, isolated from prior seismicity by over 100 km in depth, and followed by only a few small aftershocks. Globally, this is the deepest (680 km centroid depth) event with MW≥7.8 in the seismological record. Seismicity indicates along-strike contortion of the Izu–Bonin slab, with horizontal flattening near a depth of 550 km in the Izu region and rapid steepening to near-vertical toward the south above the location of the 2015 event. This event was exceptionally well-recorded by seismic stations around the world, allowing detailed constraints to be placed on the source process. Analyses of a large global data set of P, SH and pP seismic phases using short-period back-projection, subevent directivity, and broadband finite-fault inversion indicate that the mainshock ruptured a shallowly-dipping fault plane with patchy slip that spread over a distance of ∼40 km with a multi-stage expansion rate (∼5+ km/s down-dip initially, ∼3 km/s up-dip later). During the 17 s total rupture duration the radiated energy was ∼3.3×1016 J and the stress drop was ∼38 MPa. The radiation efficiency is moderate (0.34), intermediate to that of the 1994 Bolivia and 2013 Sea of Okhotsk MW 8.3 deep earthquakes, indicating that source processes of very large deep earthquakes sample a wide range of behavior from dissipative, more viscous failure to very brittle failure. The isolated occurrence of the event, much deeper than the apparently thermally-bounded distribution of Bonin-slab seismicity above 600 km depth, suggests that localized stress concentration associated with the pronounced deformation of the Izu–Bonin slab and proximity to the 660-km phase transition likely played a dominant role in generating this major earthquake.

Tsunami inundation from heterogeneous earthquake slip distributions: Evaluation of synthetic source models

AbstractThis study investigates whether eight different synthetic finite fault models (SFFM) can simulate stochastic earthquake‐tsunami with similar statistical properties to “real” earthquake‐tsunami events, where the latter are represented using heterogeneous slip distributions from 66 Finite Fault Inversions (FFI) for oceanic subduction interface earthquakes. A new method is derived to estimate SFFM parameters from FFI, and predictive relations between the earthquake moment magnitude and the SFFM corner wave numbers are developed to support model applications. SFFM with more capacity to spatially localize slip are better able to simulate higher slip regions on the FFI, and this strongly influences their associated tsunami inundation, which is computed in two dimensions over idealized topography. The best performing SFFM generates tsunami inundation which envelopes the FFI inundation in 81% of cases using 10 synthetic events (close to the ideal value of 82%), while the other SFFM show greater tendencies to underestimate inundation. These differences are related to the capacity of each SFFM to produce spatially localized slip distributions. None of the SFFM showed a tendency to overpredict inundation. The results highlight that SFFM cannot be assumed to reliably quantify uncertainties in the tsunami inundation of real earthquakes, and the use of untested SFFM could create nonconservative bias in tsunami hazard assessments. However, the most successful model used here performs quite well, although it may still underestimate inundation more often than an optimal model.

Three‐dimensional dynamic rupture simulations across interacting faults: TheMw7.0, 2010, Haiti earthquake

AbstractThe mechanisms controlling rupture propagation between fault segments during a large earthquake are key to the hazard posed by fault systems. Rupture initiation on a smaller fault sometimes transfers to a larger fault, resulting in a significant event (e.g., 2002M7.9 Denali USA and 2010M7.1 Darfield New Zealand earthquakes). In other cases rupture is constrained to the initial fault and does not transfer to nearby faults, resulting in events of more moderate magnitude. This was the case of the 1989M6.9 Loma Prieta and 2010M7.0 Haiti earthquakes which initiated on reverse faults abutting against a major strike‐slip plate boundary fault but did not propagate onto it. Here we investigate the rupture dynamics of the Haiti earthquake, seeking to understand why rupture propagated across two segments of the Léogâne fault but did not propagate to the adjacent Enriquillo Plantain Garden Fault, the major 200 km long plate boundary fault cutting through southern Haiti. We use a finite element model to simulate propagation of rupture on the Léogâne fault, varying friction and background stress to determine the parameter set that best explains the observed earthquake sequence, in particular, the ground displacement. The two slip patches inferred from finite fault inversions are explained by the successive rupture of two fault segments oriented favorably with respect to the rupture propagation, while the geometry of the Enriquillo fault did not allow shear stress to reach failure.

The 2014 Mw 6.0 Napa earthquake, California: Observations from real‐time GPS‐enhanced earthquake early warning

AbstractRecently, progress has been made to demonstrate feasibility and benefits of including real‐time GPS (rtGPS) in earthquake early warning and rapid response systems. Most concepts, however, have yet to be integrated into operational environments. The Berkeley Seismological Laboratory runs an rtGPS‐based finite fault inversion scheme in real time. This system (G‐larmS) detected the 2014 Mw 6.0 South Napa earthquake in California. We review G‐larmS' performance during this event and 13 aftershocks and present rtGPS observations and real‐time modeling results for the main shock. The first distributed slip model and magnitude estimates were available 24s after the event origin time, which, after optimizations, was reduced to 14s (≈8s S wave travel time, ≈6s data latency). G‐larmS' solutions for the aftershocks (that had no measurable surface displacements) demonstrate that, in combination with the seismic early warning magnitude, Mw 6.0 is our current resolution limit.

Complementary slip distributions of the August 4, 2003 Mw 7.6 and November 17, 2013 Mw 7.8 South Scotia Ridge earthquakes

The South Scotia Ridge Transform (SSRT) plate boundary between the Scotia and Antarctic plates experienced large strike-slip earthquakes on August 4, 2003 (Mw 7.6) and November 17, 2013 (Mw 7.8). These events have overlapping aftershock zones, which is unusual. A 36°–45° southward dipping fault zone ruptured with left-lateral displacements in each event along the northern margin of the South Orkney micro-continent near 60°S. Slip distributions for the two events are determined using teleseismic body and surface wave recordings along with constraints from GPS ground motion recordings at station BORC on Laurie Island (South Orkney Islands), just south of the SSRT. The aftershock distributions, high-frequency back-projections, and unconstrained body wave finite-fault inversions permit significant overlap of the 2003 and 2013 slip zones; however, the GPS static displacements resolve differences in the large-slip regions of the two ruptures. The 2013 earthquake sequence along the SSRT initiated with Mw 6.1 (November 13) and Mw 6.8 (November 16) foreshocks located ∼50 km west of the mainshock hypocenter, and had aftershocks extending ∼250 km eastward. The rupture spread primarily eastward at ∼2.5 km/s with a total rupture duration of about 120 s, with two distinct patches of large-slip located northwest and northeast of the South Orkney Islands. The rupture swept past BORC, with high-rate GPS (HRGPS) ground motion recordings capturing the time-varying slip history of the faulting. Traditional GPS data require that the largest-slip region of the shorter rupture in 2003 is located in the gap NNE of BORC between the two patches that ruptured in 2013. There appears to be some overlap of lower slip regions. The complementary slip distributions comprise a relatively uniform offset along this portion of the SSRT, which is one of the most seismically active regions of the entire Antarctic plate boundary.