Direct full tensor measurements of the Earth gravity field from GOCE-SGG (Satellite Gravity Gradiometer) that are further combined with GOCE-SST (Satellite-to-Satellite Tracking) lead to an excellent performance from medium to long wavelengths (Bruinsma et al., 2010; Pail et al., 2011) in newly Global Gravitational Models (GGM). Models are also improved on their spectral content and accuracy by including data from other missions such as GRACE and LAGEOS (Laser GEOdynamics Satellite). In this work, we derived the vertical gravity gradient (Tzz) from the spherical harmonic coefficients of the satellite GOCE static model GO_CONS_GCF_2_DIR_R6 (Bruinsma et al., 2014 developed up to N = 300), a full combination of the above-mentioned data, up to degree/order N = 200 on a regular grid of 0.05° grid cell size (Janak and Sprlak, 2006). The vertical gravity gradient component is the most suitable/sensitive to mapping earthquake deformations when compared to horizontal components.

The Tzz, expressed in Eötvös 1   E ö t v ö s = 10 − 4 m G a l m , represents a better theoretical resolution for detecting shallow crustal structures with high-density contrast variations (e.g., determination of the edges of anomalous masses) than the gravity vector itself (Li, 2001) as the last presents a spread signal and highlights deeper sources (Braitenberg et al., 2011; Alvarez et al., 2012). By limiting the degree/order of the harmonic expansion up to N = 200, decomposition of the gravimetric signal as causative mass depth increases is allowed, being the approximate auscultation depth of Z = 30 km for this degree/order (see Featherstone 1997; Alvarez et al., 2017a). In previous works, we qualitatively found that the best correlation between slip distribution and Tzz is obtained by limiting the harmonic expansion to degree and order N = 200 (Alvarez et al., 2017a; Alvarez et al. 2019; Alvarez et al. 2021), which agrees with the approximate depth of the seismogenic zone (megathrust). Half wavelength resolution for degree/order of N = 200 approximate to 100 km (λ/2=πR/N, Barthelmes, 2013). In addition, different authors point out that the models derived from GOCE reach their best performance at degree/order N = 200 of the spherical harmonic development (e.g., Bruinsma et al., 2010; Pail et al., 2011).

After calculating the vertical gradient from the disturbing potential derived from the aforementioned GOCE model, the effect of the topographic masses it reduced taking into account the WGS-84 ellipsoid as a reference surface, which is necessary to highlight anomalous geological structures and density contrasts within the upper crust. For calculation of the topographic contribution over the Tzz, we discretized the digital elevation model ETOPO1 (Amante and Eakins, 2009) utilizing spherical prisms with constant standard densities of 2,670 kg/m3 for masses above sea level and 1,030 kg/cm3 for seawater. The selected calculation height (Hc=3,200 m), when calculating the effect generated by the topographic masses on the vertical gravity gradient, is enough to ensure that all values are above the topography. The calculation was performed using the Tesseroids Phyton package from Uieda et al. (2010), Uieda et al. (2016) which uses spherical coordinates to take into account the Earth’s curvature to avoid considerable errors over the large study region (https://www.fatiando.org/ https://doi.org/10.5281/zenodo.4685960).

In previous works (e.g., Alvarez et al., 2014, Alvarez et al., 2019, and Alvarez et al., 2021 and references therein) we highlighted the inverse relationship between Tzz and co-seismic slip (i.e., high slip [m] over negative Tzz [mGal/m]) both in qualitative and quantitative ways. By sampling over a regular grid the Tzz field versus different co-seismic slip distribution models along the South American margin, we found that this anti-correlation improved as event magnitude and consequently rupture area increased (e.g., the correlation coefficient for the Maule 2010 earthquake was of approx. -0.70). In this work, we now computed the spectral coherence between the topography corrected vertical gravity gradient (Tzz), derived from GOCE satellite up to N = 200, and three co-seismic slip distributions (Shao et al., 2011; Wei et al., 2012; Lee et al., 2011) for the Thohoku-Oki earthquake.

The cross-spectral estimation was computed in the frequency domain assuming that the Tzz is the input and the slip is the output data (see Generic Mapping Tools software GMT, Wessel et al., 2019). First, we selected a rectangular area (dashed lines in Supplementary Figures A1–A3 in Supplementary Appendix SA) comprising the region of maximum slip. Geographical grids (in degrees) were scaled to meters via a “Flat Earth” approximation using the WGS-84 ellipsoid parameters. Then, both fields were de-trended (by removing the best-fitting linear trend) and tapered (data were extended and tapered to zero) before the computation of the 2-D forward Fast Fourier Transform (Supplementary Figures A1–A3 in Supplementary Appendix SA). Finally, we performed FFT and then calculated the power spectrum in the radial direction of both signals and the spectral coherence (plus standard deviation error estimate) between both fields (Figure 2). Discussion of coherency at low wavelengths, i.e., for a spatial resolution up to 100 km, is irrelevant as GOCE resolution is about 80–90 km at equator. For wavelengths around λ = 200 the coherency between the two signals improves substantially up to 90%.

The three highly positive values from Tzz (>+10 Eötvös), along the coastal line (East of Hachinohe, Sendai, and Ibaraki Prefecture, see Figure 2), are consistent with the residual gravity anomaly obtained by Bassett et al. (2016) albeit at very different wavelengths. In that model, these authors also found high gravity values over the forearc close to the Japan trench axis, up to approximately 15 km depth, with an interspersed gravity low (along the slab/Moho intersection) between these two across-strike gravity highs (Figure 1 of Bassett et al., 2016). The Tohoku-Oki Mw = 9.0 2011 earthquake nucleated close to a positive Tzz maxima lobe (located to the East of Sendai) with the main rupture pattern migrating (trenchwards) to the closer negative minima in Tzz (with less than -20 Eötvös), where it reached its maximum amplitude (e.g., Shao et al., 2011 in Figure 4, Wei el al. 2012 in Figure 5, and Lee et al., 2011 in Figure 6). Earthquake nucleation close to a local positive high gradient zone towards the coastline had been previously observed for other great earthquakes (e.g., Valdivia 1960; Arequipa 2007; Maule 2010; Illapel 2015; Musine 2015; Sumatra-Andaman 2004; see Alvarez et al., 2019 and Alvarez et al., 2021). This had been initially noted by Tassara (2010), who showed that ruptures along the Andean margin generally nucleated at the edge of geological heterogeneities in the forearc, and Hicks et al. (2014) who reported that the Maule earthquake nucleated in a region of high Vp and positive gravity anomaly.

The vertical gravity gradient off-shore northern Honshu is divided by a trench-parallel strong gradient with positive values towards the coastline (more than +10 Eötvös) and negative values (less than −20 Eötvös) to the trench (Figure 2). This across strike abrupt change in the gradient signal is almost centered over the slab/Moho intersection (obtained from Bassett et al., 2016). Wei et al. (2012) explained that there was little expectation that future interplate earthquakes along the eastern coast of Honshu would be different from the max. Mw = 8.5 earthquakes documented historically (Avouac 2011). Future earthquakes were expected to occur along the plate interface patch that had remained locked during the interseismic period (but determined from the modeling of geodetic strain, measured on-shore by Hashimoto et al., 2009; Loveless and Meade, 2010; Suwa et al., 2006). No major rupture areas were documented in the region between 1896 and 2011 (e.g., Bassett et al., 2016), beyond several earthquakes with over Mw/J > 7.0. The largest previous earthquake reported in this region was the Sanriku earthquake of 1896 (Kanamori, 1972) with a moment magnitude about 8 times lower than the Tohoku-Oki earthquake. When looking at historical earthquakes along this portion of the eastern Honsu Island margin, it is observed that only earthquakes that ruptured mostly or entirely between this line (slab/Moho intersection) and the trench line, reached magnitudes greater than M8 (specifically the 1896 M8.5 and the Mw = 9.1 Tohoku-Oki earthquake). We can make an analogous observation for the Mentawi region, where many intermediate magnitude earthquakes occurred along this trench parallel higher Tzz closer to the coast (e.g., Chlieh et al., 2008) while great megathrust earthquakes and historical great ruptures are restricted to the region with lower Tzz (see Alvarez et al., 2021).

Along the South American margin, a trend of low-density values was observed off-shore along the entire marine forearc (i.e., from the trench to the coast), being the −5 Eötvös contour indicated as the down-dip limit of the seismogenic zone, roughly coincident with the 30 km iso-depth plate contour (e.g., for the Maule 2010 earthquake and the Ecuador-Musine 2017 earthquake; see Alvarez et al., 2019 and Alvarez et al., 2017a). Along the central to southern Chilean margin, the Tzz has a positive signal just along the coastal line, which has been related to dense, ultramafic material due to under-plating (e.g., Alvarez et al., 2014; Bassett and Watts 2015; Alvarez et al., 2021). These parallel trench differences in the gradient signal could be explained as a result of a higher dipping angle in the Mentawi-Sumatra and Eastern Honsu margin regions (10°) when compared to the South American margin (5°). In other words, the main seismogenic zone, where type B asperities are located (Lay et al., 2012) can be mapped using a low gravity gradient signal, regardless of what type of margin we are analyzing.

When superimposing the different slip models to the Tzz (Figures 4–6), a slight lateral shift is observed between both quantities. This shift was also observed for other earthquakes (e.g., Sumatra-Andaman) and is interpreted as the positive effect of the subducting plate over the gradient signal, masking the low Tzz trend observed along the marine forearc. The forebulge presents a highly positive signal, which rapidly decreases (contours from > +10 Eötvos to <−10 Eötvos) as we move away from the trench in the landward direction, masking any lower densities signals near the trench, at least at these wavelengths.

There are three well-defined Tzz minima lobes along the central minima gradient: a < −25 Eötvös anomaly located to the south of Ibaraki Peninsula (named 4 in Figure 2), a central minima lobe to the east of Sendai, where the maximum slip was achieved (N° 2), and another with < −20 Eötvös to the north (N°1). The section between the southern minima Tzz lobe (4) and the central minima lobe (2) presents a smooth pattern, as a transitional zone, with about −17 Eötvös (named 3 in Figures 2–7). In this region, Bassett et al. (2016) reported higher densities and a high Residual gravity anomaly (mGal). These authors reported this prominent structural boundary along strike off-shore south-central Honsu, and proposed it as a possible extension of the Median Tectonic Line-MTL that separates two contrasting geological terrains along the forearc (red dashed line in Figure 2): a ∼20 km thick granitic upper-crust (north side on-shore) (Isozaki et al., 2010) and a crust entirely composed of variably metamorphosed Late Mesozoic to Cenozoic accretionary complexes (south side off-shore). The location of the MTL is coincident with a latitudinal segmentation of the Tzz signal, probably indicating the existence of an along-strike seismic barrier or attenuator (see Alvarez et al., 2014). The former authors proposed that stress heterogeneities, on either side of the MTL, may explain both the small slip area and the large slip amplitude of the Tohoku-Oki earthquake rupture, being the extent of co-seismic slip limited by the inability to rupture into the low-stress or velocity-strengthening fault segment south of the MTL. On the other hand, the highly stressed or strongly velocity-weakening area with high rates of tectonic loading north of the MTL resulted in a large slip amplitude (Basset et al., 2016).

Opposite to this, Tormann et al. (2015) explained that there is no significant variation in the low b-values to suggest that the plate interface is segmented in a way that might limit potential ruptures. These authors also found no indication for a long assumed lateral segmentation of the megathrust plate interface into smaller areas that would only rupture in isolation, limiting the maximum magnitude. Our results, based on densities distribution derived from the gravity field, disagree with this hypothesis about the absence of identified barriers. At approximately ∼36.5°N 142°E, a seismic barrier can be identified (Figure 2) indicated by the slight narrowing of the −17 Eötvös contour (orange dashed contour) which coincides with a subducted seamount chain (orange dot and dashed contour) reported by Mochizuki et al. (2008). In this sense, Nakatani et al. (2015) who analyzed the Tohoku-Oki earthquake aftershocks, by using dense ocean bottom seismic array data, suggested that the large coseismic slip zone ended in the vicinity of the frontal region of the Daiichi-Kashima subducting seamount and may not have extended off Ibaraki. This seismic barrier, not only marks the southern ending of the rupture propagation for the Tohoku-Oki earthquake but also separates the 2011 M7.7 rupture to the south from the 1938 M7.5 rupture area to the north. The Tzz signal between 36.75°N and 37.5°N encloses relatively minima values (identified as C in Figure 2), indicating the possible location of a seismic asperity (Alvarez et al., 2019; Alvarez et al., 2021 and references therein); this region that encloses the 1938 earthquake rupture area was also activated by the Tohoku-Oki earthquake (see the 5 m contour in Figure 2). The relatively higher Tzz value at 37.5°N (and approx. 142.5°E) indicates the location of another seismic barrier (orange dashed contour in Figure 2) that separated the 1938 M7.5 earthquake from the 2003 M6.8 rupture area. To the north, at about 39.5°N 143.25°E, the rupture ended in a region with a pronounced Tzz signal segmentation, indicating the location of the third barrier to seismic energy.

Shao et al. (2011) modeled the kinematic rupture history (consistent with Ide et al., 2011) with a time interval of 15 s, revealing a complex rupture process with a weak rupture during the first 5 s, initiating at an approximate depth of 23 km (Figure 4). During the following 25 s, the rupture propagated in a speed of apparently less than 1 km/s, illustrated by the lag of rupture front at 30 s (outlined by the 4-m contour of fault slip in red gamma) relative to the 1.5 km/s pseudo-rupture front (grey dashed circles, see Figure 4 from Shao et al., 2011). Then, as explained by Shao et al. (2011) the apparent rupture propagation significantly speeded up from 30 s to 45 s, accompanied by a larger slip rate and producing the first peak in the moment rate function; being the rupture limited to faults with centroid depths similar or deeper than the hypocenter. At this time, the rupture that was propagating mostly in the downdip direction reached the 0 Eötvös contour (with more positive Tzz values to the coast) where the slip was attenuated. Then at about 45 s, as explained by Shao et al. (2011), the “explosive” failure of the near trench asperity took place (in the opposite direction or up-dip) reaching up to 60 m of slip.

Similar results were shown in previous works (e.g., Alvarez et al., 2014; Alvarez et al. 2019; Alvarez et al. 2021), where we have characterized these abrupt changes in the Tzz signal, either across or along strike, as seismic barriers or attenuators to the propagation of seismic energy; being the high Tzz signal across the continental forearc (parallel to the coast, and beneath the coastline along south-central Chilean margin) indicative of the down-dip limit for rupture propagation.

From a more precise source of information, Lee et al. (2011) performed a joint source inversion (teleseismic body waves, seafloor, and terrestrial GPS coseismic deformation data and dense local strong motion records) finding a better-constrained model (Figure 5). These authors mapped the slip and accumulated slip through time, finding a large-scale repetition of slips during the temporal rupture process. Their inverted source model shows several time periods of energy release with three main peaks, suggesting repetition of large-scale slip on the largest of the three found asperities. They denoted the largest asperity, developed around the hypocentre, with a maximum slip over 50 m and a broad area of about 200 × 200 km², as Asperity I. It concentrated slip in two areas (Figure 5): Asperity IA around the hypocenter (depth between 10 and 50 km), and other Asperity IB located slightly north of the hypocenter (at about 10–30 km depth). In this model, extremely large Asperity IA activated for 40–90 s and extended from shallow to the down-dip along the fault plane. Then, when the rupture front moved slowly to the south gradually forming Asperity II (between 100 and 140 s approximately), the area just above the hypocentre started to slip again expanding northward and merging with Asperity IA that had slipped before, producing the second slip concentrated area on Asperity I, i.e., Asperity IB (Lee et al., 2011 for more details).

This main asperity of approximately 80 km wide inferred by different source models (e.g., Shao et al., 2011 at T = 45 s–90 s or Lee et al., 2011 at 30–22 and 74–76 s), located right above the hypocentre, is over the minima Tzz lobe with less than −20 Eötvös (named N° 2 and highlighted with a black solid ellipse in Figure 7). These Tzz minima were proposed as seismic asperities (Alvarez et al., 2019; Alvarez et al., 2021) in the sense proposed by Lay and Kanamori (1981) and by Lay et al. (1982), Lay et al. (2012). Results from Bletery et al. (2014) reinforce this hypothesis, where the stress drop along the fault was dominated by two very large patches (localized halfway between the epicenter and the trench) with amplitudes exceeding 8 MPa. These patches, occurring after large stress-drop after earthquake nucleation, are thought to be the result of the rupture of locked asperities loaded in stress during the interseismic period (Chlieh et al., 2007; Konca et al., 2008) as explained by Bletery et al. (2014).

Using the vertical seafloor deformation, Wei et al. (2012) separated the contributions to the slip model into shallower (S0–S1) and deeper (S2–S4) portions by performing a forward prediction for the 4 nearest Deep-ocean Assessment (ocean bottom geodetic) and Reporting of Tsunamis (DART) records. They found that the S0–S1 portion (at shallow depths and with high slip amplitude) produced the most marked seafloor deformation and the higher tsunami recorded and found a nearly complete stress drop on the upper fault portion (S0). The Tzz signal shows two well-differentiated areas along the marine forearc (along the megathrust) in consonance with observations made by Wei et al. (2012), who proposed that the shallower (S0–S1) and deeper (S2–S4) portions of the rupture have fundamentally different characteristics. These authors proposed that these differences are related to a strong and very efficient dynamic weakening process, possibly thermal pressurization (the existence of abundant free water and low permeability, due to the large clay fraction in the sediments accumulated in the trench and dragged along the megathrust), which produced a large displacement rupture at shallow depth with little high-frequency seismic radiation (as region B in the sense of Lay et al., 2012); while the deeper rupture was characterized by relatively small fault displacement with relatively strong high-frequency radiation (as region C in the sense of Lay et al., 2012). Practically all the vertical deformation predicted by this model (Wei et al., 2012) is located to the east of the 0 Eötvös contour, where the Tzz signal is negative (Figure 6).

Tormann et al. (2015) resolved a low-b-value structure in the subsequent high-slip area of the Tohoku-Oki main shock (between 38°N and 39°N), indicating locally a specifically strong stress accumulation. This asperity on the plate interface (low-b-value structures can be usefully interpreted as mapping asperities; Schorlemmer and Wiemer 2005), which latitudinally matches the Tzz minima lobe (<−20 Eötvös), extends about 200 km north-south and 100 km east-west, reaching b-values <0.5, and seems to have formed over several years before this mega-earthquake occurrence (white dashed line in Figure 7). This segment of the plate interface is a hard-to-break segment that remained largely locked or with strong coupling (Hashimoto et al., 2009), can sustain high levels of stress, and tends to release the accumulated slip deficit in large ruptures as occurred (Tormann et al., 2015). The hypothesis of the existence of asperity in this region is reinforced by the fact that the areas of large co-seismic slip during the Tohoku-Oki event exhibited a significant increase in b-values after the mainshock (returning to values similar to those seen between 1998 and 2003 in the same period that the aftershock rate to reach the average long-term level), representing a strong stress release in these areas 21, 28 (Figs 2T2 and T3 and 3; Tormann et al., 2015). Different interseismic coupling models exhibited a high degree of coupling (close to 100%; Suwa et al., 2006; Perfettini and Avouac 2014) and a high back slip rate (close to 10 cm/yr, Loveless and Meade 2011) centered to the east of the Oshika Peninsula in agreement to the region where the Tzz minima low is observed and occurred maximum slip in the Tohoku-Oki earthquake. In a former work, Alvarez et al. (2015) reported a positive correlation between the Tzz and b-value (previously observed by Sobiesiak et al., 2007 but using Isostatic Residual Anomalies) proposing to use gravity as an indirect tool for mapping low b-value zones and more indirectly potential zones of asperities along the seismogenic zone that could trigger large events.

The relationship between the Tzz signal and seismic behavior in the region of the Tohoku-Oki earthquake agrees with previous observations along the South-American margin (Alvarez et al., 2014; Alvarez et al., 2019) and along the Sunda subduction zone (Alvarez et al., 2021). In all these regions, rupture propagated mainly along negative Tzz, which in turn showed a high degree of interseismic coupling (e.g., Chlieh et al., 2008; Chlieh et al., 2011; Metois et al., 2016), slip enhanced at Tzz minima lobes, and was attenuated at intermediate values and arrested at Tzz highs both along and across strike.

By using offshore 2D/3D seismic survey data, Baba and Yoshida, (2020) analyzed in detail the geological structure over the NE Japan forearc, demonstrating that the structure in the hanging-wall plate of the subduction system consists of a series of structural blocks (segments) separated by NW–SE trending strike-slip faults. They also proposed that the inherited geological structural configuration of the overriding plate showed a close relationship with the seismic activity related to the 2011 M9.0 Tohoku-Oki megathrust earthquake (i.e., distribution of foreshocks, mainshock, and aftershocks, coseismic slip models of the coseismic slip area of M-7 class earthquakes, quasi-static slip rates, back slip rate, and seismic tomography images). By comparing two different coseismic slip models, Baba and Yoshida, (2020) highlighted that the trenchward forearc of the structural blocks between the offshore Hidaka tectonic line and the Honjo-Sendai tectonic line (yellow dashed lines in Figure 7) fitted well with the coseismic slip area of the 2011 Tohoku-Oki earthquake. These strike-slip faults coincided with the observed segmentation of the Tzz identified as barriers (white dashed contours in Figure 2) that confine the minima Tzz lobe N° 2 (Figure 7) related to the main asperity of the Tohoku-Oki earthquake.

On the other hand, Bletery et al. (2014) highlighted the crucial role of normal faulting in the overriding plate and reported that the largest slip occurred just in the up-dip direction of an important normal fault revealed in seismic lines (Tsuji et al., 2011, 2013) co-seismically reactivated. Bletery et al. (2014) also explained, based on Cubas et al. (2013), that this normal fault (yellow line in Figure 7) could have activated coseismically (under very low dynamic friction conditions) allowing a large amount of slip to propagate up to the free surface. These authors proposed that low friction authorized large slip with a low-stress drop in the shallowest part of the megathrust defined as block A (consistent with the large Tsunami), which is a coherent unit separated by the normal fault from another block B (Figure 7) located to the west (see Bletery et al., 2014 for more details). Finally, they proposed that the rupture of a long-term locked asperity on the megathrust (located below the normal fault) caused block A to move as a coherent unit with respect to block B. Our results show that the eastern edge (i.e., trenchwards) of the low Tzz lobe N° 2 coincides with the location of this normal fault (from Tsuji et al., 2011) separating very low Tzz values (<−20 Eötvös) on the ceiling block (B) with respect to higher Tzz values over the floor block (A) (Figures 7, 8).

In this way, both Bletery et al. (2014) and Baba and Yoshida, (2020) proposed a block motion behavior segmenting the forearc but analyzing different fault systems. The first, proposes that the motion of block A, which is decoupled from block B by a trench parallel normal fault, allows it to slip freely with respect to both the incoming plate and the rest of the overriding plate. This proposal is consistent with the absence of aftershocks in the frontal wedge (i.e., block A) observed by Obana et al. (2013) and because the amount of slip does not significantly decrease as the rupture reaches the free surface. The last (i.e., Baba and Yoshida, 2020), suggests that the structural blocks (in that work referred to as Blocks B, C, and D, named in Figure 7 as BB, BC and BD respectively), bounded by the offshore Hidaka tectonic line and the Honjo-Sendai tectonic line, slipped rapidly trenchward when the 2011 Tohoku-Oki earthquake occurred (Figure 7). Thus, the low Tzz lobe (identified as N° 2) could be mapping a main crustal block along the marine forearc, limited along-strike by strike-slip faults (Baba and Yoshida, 2020) and up-dip by a trench parallel normal fault (Tsuji et al., 2011; Tsuji et al., 2013). This main block is probably limited towards the down-dip direction by another fault just over the slab/Moho intersection (mapped by Bassett et al., 2016). The above-mentioned Tzz lobe is located along the large and relatively uniform region with unstable sliding frictional properties that generate modest levels of diffuse short-period radiation, high coseismic slip, high seismic moment release, and high-stress drop from the central domain B (i.e., asperities) at source depths spanning from 15 to 30 km where most megathrust events occur (see Lay et al., 2012).

Similarly, blocks segmenting the forearc were mapped from Tzz and geological data along the southern Chilean margin, where the Maule 2010 Mw = 8.8 earthquake occurred (Alvarez et al., 2019). For this earthquake, some authors (e.g., Moreno et al., 2012) suggested that splay faults in the upper plate limited rupture propagation in the up-dip and along-strike directions. In this region, four structures that segment the upper plate and control rupture behavior could be mapped from Tzz. (1) A north-south trending Thrust Ridge (TR in Figure 9) coincident with the up-dip limit to rupture propagation is depicted by the −5 Eötvös contour. This structure is associated with splay faults (Geersen et al., 2011) that coincide with the mechanical discontinuity between the frontal accretionary prism and the forearc paleoaccretionary structures (Contreras-Reyes et al., 2010; Moscoso et al., 2011). (2) The Santa María Fault (SMF), which consists of a series of back-thrusts extending between ∼36°S and 37°S rooted in the plate interface (Melnick et al., 2009), separating the northern main rupture lobe from the southern minor lobe. (3) The Lanalhue Fault Zone (LF), where transpressional deformation has been associated with the displacement of a forearc sliver, matches the southern ending of the rupture and coincides with Tzz segmentation. (4) To the North, the Pichilemu Fault (PF) which coincides with Tzz along-strike segmentation, marks the northern ending of the Maule earthquake rupture (Figure 9).

For the Tohoku-Oki earthquake, the negative Tzz signal along the marine forearc (located approximately over 10 km and 30 km of slab contours), which coincided with the shallow fault portion S0 (Figure 6) proposed by Wei et al. (2012), could also be indicative of regions where thermal pressurization is favored, enhancing the probability of occurrence of a great megathrust earthquake. These authors explained that the theoretical framework from Noda and Lapusta (2010) provided one possible explanation of many aspects of the Tohoku-Oki earthquake. The last authors considered two patches, X and Y, (side-by-side with distinct hydraulic diffusivity (being lower within patch Y allowing for thermal pressurization within that patch only) and found that most events nucleated in X and sometimes propagated into Y where slip enhanced due to thermal pressurization. Then, as rupture enters patch Y, the pulse grows broader and the slip velocity increases reaching a much larger final slip in patch Y than in X. For the NE Japan forearc, Wei et al. (2012) proposed that the shallow fault portion S0 (Figure 6), where thermal pressurization has been favored, would rupture only occasionally in very large ruptures as occurred in the Tohoku-Oki earthquake (like patch Y in the simulations of Noda and Lapusta (2010)); while deeper regions should rupture more frequently during more moderate events (like patch X of Noda and Lapusta (2010)), agreeing with the more frequent M7 seismicity. These deeper regions, which present a positive Tzz signal, could be related to the A side from Noda and Lapusta (2010) with less thermal pressurization.

Forearc blocks mapped from the gravity-derived signal (e.g., negative Tzz lobes as asperity N° 2, or those in Figure 9 for Maule) could present a sustained subsidence behavior during the interseismic period (Figure 10) as they are under a high coupling degree (in the cases that the elastic deformation is not recovered during earthquakes, in agreement to that proposed by Sugiyama 1994 and Wells et al., 2003). The existence of this behavior can be found in a previous work, where Alvarez et al. (2015) found that the area where intense foreshocks and a slow slip event preceded the 2014 Iquique Mw = 8.1 earthquake (e.g., Ruiz et al., 2014), and later, maximum slip occurred (e.g., Hayes et al., 2014), presented subsidence in the late interseismic period. These results are also in agreement with the foreslip model of Socquet et al. (2017) (see Alvarez et al., 2015 for more details). Contreras Reyes (2018) inferred these subsided coastal blocks, characterized by a gravitational collapse of the outermost fore-arc, from 2-D velocity-depth models at different latitudes along the South American margin. The location of these blocks (Figures 4–6 from Contreras Reyes, 2018) agrees with low Tzz lobes (Alvarez et al., 2015; Alvarez et al., 2017b), which in turn present a great correlation with the maximum slip for different megathrust earthquakes (e.g., Pisagua-Iquique Mw = 8.2 2014, Tocopilla Mw = 7.7 2007, Illapel Mw=8.3 2015 earthquakes).