1 Introduction

The subduction zone along Chile was struck by the third great earthquake in 6 years when the plate boundary offshore of Illapel ruptured from 30.25°S to 32.5°S in an M W 8.3 event on 16 September 2015 (31.570°S, 71.670°W, 22:54:33 UTC (USGS National Earthquake Information Center—NEIC: http://earthquake.usgs.gov/earthquakes/) (Fig. 1). This region had been identified as a seismic gap based on the occurrence of prior large earthquakes in 1943, 1880, and 1730 (Kelleher 1972; Nishenko 1985; Beck et al. 1998). The region just to the south (32°S to 34.5°S) ruptured most recently in the 9 July 1971 [M S 7.9, M W 7.8, (ISC-GEM)] and 3 March 1985 [M S 7.8, M W 7.9 (CMT)] Valparaíso earthquakes as well as previously in the great 17 August 1906 (M S 8.4) earthquake (e.g., Comte et al. 1986; Christensen and Ruff 1986). The region to the north (26°S to 30.25°S) ruptured in the great 11 November 1922 M S 8.3 earthquake (e.g., Beck et al. 1998) (actual rupture is likely to be offshore, as large tsunami was produced, so the ISC-GEM location in Fig. 1a is too far inland). The 1922 rupture zone has been seismically relatively quiet since then.

Fig. 1
figure 1

Maps of seismicity in the source region of the 16 September 2015 Illapel, Chile earthquake with the inset (b) showing a slip distribution inverted from teleseismic body waves. The USGS-NEIC epicenter is indicated by the red star. The global centroid moment tensor (gCMT) mechanism is shown with a line to the epicenter in both panels and plotted at the gCMT centroid in (b). gCMT solutions for other large events are shown at their respective centroid locations, color-coded by depth and scaled proportional to their moments. Mechanisms for the great 1943 and 1922 events are from Beck et al. (1998). Preceding earthquakes larger than magnitude 7.0 from 1900 to 2015 from the ISC-GEM catalog are shown by the depth-colored circles in (a) and the blue circles in (b) (or focal mechanisms, if known) with each radius scaled by estimated magnitude (year and magnitude are indicated). The seismicity of the 1997 Coquimbo swarm is outlined in yellow. The black contours in (a) are the fault locking model of Métois et al. (2012) for south of −28°, with the dashed curve indicating 50 % locking and the solid curve 90 % locking. The small green circles in (b) are USGS-NEIC epicenters of aftershocks in the first 2 weeks following the 2015 mainshock. The rectangular area in (b) is the slip distribution estimated by teleseismic P and SH inversion for V r = 2.0 km/s and dip 22° in this study (Fig. 4d). The toothed white curves indicate the trench position

Full size image

The rapid global Centroid-Moment Tensor (gCMT) solution for this event (http://www.globalcmt.org/CMTsearch.html) indicates an almost pure double-couple faulting geometry with strike, ϕ = 5°, dip, δ = 22°, and rake, λ = 106°, at a centroid depth h c = 17.8 km with a centroid location northwest of the hypocenter (31.22°S, 72.27°W) (Fig. 1). The centroid time shift t c is 47.9 s and seismic moment M 0 is 2.86 × 1021 Nm (M w 8.2). The W-phase solution from CNRS has M 0 = 2.68 × 1021 Nm, with a centroid location at 31.02°S and 72.04°W with h c = 17.5 km, and best-double-couple ϕ = 2.7°, δ = 22.3°, and λ = 94.3° (http://wphase.unistra.fr/events/illapel_2015/index.html). The USGS-NEIC W-phase solution has slightly larger M 0 = 3.19 × 1021 Nm (M W 8.3) and h c = 25.5 km, with a best-double-couple geometry of ϕ = 353°, δ = 19° and λ = 83°. Although the precise seismic moment, dip and source depth vary among these early point-source estimates due to intrinsic trade-offs and limitations of long-period determinations, the location and size estimates confirm the potential for large tsunami generation.

The Nazca plate has been underthrusting South America near the 2015 rupture at about 74 mm/year (DeMets et al. 2010) for the 72 years since the 6 April 1943 rupture, suggesting that up to 5.3 m of slip deficit may have accumulated since then. Accumulation of slip deficit on the Chilean megathrust has been measured using the progressively eastward deflection of GPS stations in Chile for almost two decades (e.g., Norabuena et al. 1998; Kendrick et al. 1999). Several early interpretations modeled GPS observations with a uniformly totally locked shallow plate boundary and a significantly deforming back-arc region (e.g., Kendrick et al. 2001; Khazaradze and Klotz 2003; Brooks 2003). Vigny et al. (2009) conclude that only 40–45 % of the total convergence rate between the Nazca and South American plates is causing accumulation of elastic deformation in the upper plate. Recent interpretations indicate that the plate boundary slip deficit is a large fraction of the plate convergence near 31°S, but reduces along 31.5–32°S (Moreno et al. 2010). Métois et al. (2012) also argue for variable coupling along the plate boundary (Fig. 1a), but find nearly 100 % coupling from 31°S to 33°S at depths from 15 to 45 km on the megathrust, corresponding to the 2015 rupture area in Fig. 1b. Weak coupling from 30°S to 31°S is indicated for the region to the north of the rupture where the July 1997 Coquimbo swarm (Fig. 1) occurred (e.g., Gardi et al. 2006; Holtkamp et al. 2011; Brodsky and Lay 2014).

Recent studies of large subduction-zone earthquakes clearly demonstrated that although the long-term strain accumulation patterns can be mapped quantitatively with GPS, the occurrence of individual earthquakes exhibit significant complexity in both spatial and temporal distributions, especially along subduction zones where events with variable rupture length have occurred. Given this complexity, when a large earthquake has occurred, very rapid assessment of the event is critically important for early hazard warning of aftershocks and tsunamis, immediate post-earthquake emergency efforts, and assessment of regional seismic potential immediately following the event. With the currently available real-time data and analysis methods, we can assess the first-order tsunami potential in 5–20 min using long-period seismic waves and high-rate GPS, and characterize the slip distribution and the spectral characteristics of the source within a few hours. In view of this importance, we report here the primary source parameters estimated from rapid seismological methods for the 16 September 2015 event and discuss the rupture in the context of the historic earthquake activity and the plate locking determinations. We examine point-source parameters and tectonic setting to constrain the faulting geometry, back-projection of teleseismic P waves to constrain the rupture dimensions and expansion velocity, and then perform finite-fault inversions for which we estimate source parameters for the event.

2 Rapid Seismic Analysis of the 2015 M w 8.3 Earthquake

2.1 Point-Source Solutions

For tsunami warning purposes, it is desirable to estimate the earthquake mechanism, seismic moment, and depth within a few minutes. We applied the W-phase inversion method (Kanamori and Rivera 2008; Duputel et al. 2012) to regional three-component ground displacements in the passband 1–5 mHz from 10 to 11 global seismic stations (22 components) within 10° of the epicenter. For source depths of 17.5–25.5 km this analysis yielded shallow-dipping (13.7°–17.4°) thrust mechanisms with moments of 3.5–3.4 × 1021 Nm, and could be completed within five to ten minutes if operated in real time, as it requires only 200–300 s of recorded signal after the origin time. Centroid time shifts, t c, of 42–41 s were found, which are only moderately longer than the value of 36 s predicted based on the average relationship found by Duputel et al. (2013) of t c = 2.59 × 10−6 M 1/30 (for M 0 in Nm). This provides a rapid determination that the source was not a tsunami earthquake, as such events have significantly larger t c shifts relative to the average (Duputel et al. 2013), even while the seismic moment, mechanism, and location of the 2015 event do suggest significant tsunamigenesis.

An operational application of regional W-phase inversion was performed by the Chilean National Seismological Center of the University of Chile, and provided a solution within 5 min using data from 22 stations within 12° of the epicenter, with M W 8.1. This information was automatically transmitted to the Navy Hydrographic Service in charge of the tsunami alerts, and the Emergency Office of the Interior deployed an evacuation plan for the coastal population. It is estimated that about 1 million people were evacuated from various exposed regions in central Chile.

Point-source inversions using global data for the 16 September 2015 Illapel mainshock reported by routine rapid seismic wave processing have moderately varying thrust plane dip, centroid depth, and seismic moment. For shallow thrust events the product M 0 × sin(2δ) is well determined by long-period inversions. However, there is strong trade-off between dip and seismic moment, such that the estimated moment is about 30 % larger for a 16° dip than for a 22° dip. We performed W-phase inversions using varying global data distributions finding centroid depths of 25.5–30.5 km and dips of 16.1° to 18.5° with centroid locations being dependent on the data set.

To provide an independent estimate of the dip to stabilize the trade-off between dip and moment, we examined the slab geometry models of Hayes and Wald (2009) and Hayes et al. (2012) which indicate a relatively planar plate boundary interface dipping at ~16.5° near 31°S. Relative to the trench position, the USGS-NEIC hypocenter is compatible with a dip of ~16° as well. Thus, there is a range of dip estimates of from about 16° to 22° quickly available, with the steeper dip requiring some reduction of dip at shallow depth on the megathrust to connect to the trench. There is no clear resolution of shallow curvature from the seismicity distributions. For our finite-fault inversions we consider models with uniform dips from 16° to 22°, recognizing that the seismic moment varies with dip and that dip may actually vary along the megathrust.

2.2 Back-Projection Analysis

We performed a back-projection of teleseismic short-period P waves recorded in North America. This is the best station geometry (Fig. 2a) for back-projection given the location of the source region and global station distribution. The broadband P wave arrivals are very coherent and readily aligned across North America by multi-station correlation (Fig. 2b). The aligned traces were then filtered in the passband 0.5–2.0 Hz and back-projected using the procedure of Xu et al. (2009) with fourth-root stacking of beam power across a constant depth target grid in the source region. Figure 2c shows the time-integrated beam power at each grid position, demonstrating that the coherent short-period energy largely originated north of the epicenter, with two concentrations, the higher power image forming about 60 km NE of the epicenter and a second region about 50 km WNW of the epicenter. The primary bursts of coherent short-period energy are found in the first 50 s of the rupture process near the coastal region and there is little evidence for short-period bursts toward the trench (Fig. 2d). The spatial extent of the rupture indicated by these back-projections extends less than 100 km from the epicenter, which together with the 47.9 s centroid time for the gCMT inversion suggests a low overall rupture expansion velocity.

Fig. 2
figure 2

a Location of North American seismic stations used in the back-projection of 0.5–2.0 Hz P wave signals for the 16 September 2015 Illapel earthquake, color-coded by average correlation coefficient with all other traces. b The aligned, unfiltered P wave data from North America used in the back-projection. c Map of the time-integrated beam power distribution obtained by back-projection. The star indicates the USGS-NEIC epicenter, which is the reference position for aligning the first-motions of the P waves. The small circles are USGS-NEIC aftershock epicenters in the first 4 days following the 2015 mainshock, scaled proportional to magnitude. d Positions of back-projection maxima at various lapse times from the back-projection image for the North American P wave data. The relative power of peaks scales the circle radii and the points are color-coded by time into the rupture. An animation of the full back-projection is provided in Supplemental Movie 1

Full size image

Time snapshots from the back-projection are shown in Fig. 3, capturing the irregular growth of the two lobes of short-period energy, first forming to the WNW from the epicenter in the first 20 s, then to the NE around 30 s, then with irregular northward expansion over the next 30 s. Significant coherent energy is imaged to at least 95 s lag time in the azimuth to North America. Estimates of rupture expansion velocity of 1.5–2.0 km/s can be made from the space–time pattern of these back-projection images. Similar overall behavior is found in the on-line back-projections posted by the Incorporated Research Institutions for Seismology (IRIS) (http://ds.iris.edu/spud/backprojection/10089385).

Fig. 3
figure 3

Snapshots from the North American P wave back-projection for the 2015 Illapel earthquake. Each image shows the spatial distribution of beam power at the lapse time indicated by the red line on the power time history shown above each map. These power peaks can be used to infer the pattern of rupture expansion with time and the associated effective rupture velocity for the portion of the 0.5- to 2.0-Hz energy that constructively interferes in the beam image. The fourth-root beam power is shown, so the image is not linearly proportional to wave amplitudes. Significant beam power is imaged for about 90 s. The small circles are USGS-NEIC aftershock epicenters in the first 4 days following the 2015 mainshock (red star), scaled proportional to magnitude. The continuous animation is provided in Supplemental Movie 1

Full size image

Surface wave source time functions display slight narrowing at azimuths northward from the epicenter, compatible with a low rupture velocity (1–2 km/s) along a 5° strike (http://ds.iris.edu/spud/sourcetimefunction/10090670) or about 30 s range in duration at an azimuth of 345° with a mean duration of about 105 s. Assuming unilateral rupture toward the NNW the latter measurements indicate a rupture dimension of only about 60 km and a low rupture velocity of ~1 km/s. Some component of bilateral rupture cannot be precluded, but the overall long-period source directivity is modest, similar to the short-period back-projection finiteness.

2.3 Teleseismic Finite-Fault Inversion

Guided by these constraints on the faulting geometry and rupture dimensions, several finite-fault slip inversions were performed using a least-squares procedure applied to 135 s long recordings of 60 teleseismic broadband P waves and 42 SH waves (Hartzell and Heaton 1983; adapted by Kikuchi and Kanamori 1991). In this kinematic inversion we prescribe a fault strike (5°) and dip (16° or 22°) based on slab geometry or long-period moment tensors. The fault is subdivided into 9–10 segments along-dip and 18–19 along-strike with 15 × 15 km subfault dimensions. The subfault source time functions are parameterized by 6 2.5 s rise time triangles offset by 2.5 s, giving possible total subfault durations of 17.5 s. Rake is allowed to vary for each subfault subevent. Rupture expansion velocities of 1.5 and 2.0 km/s are considered here, although we considered a wider range of rupture velocities finding that an optimal choice is not constrained by waveform misfit, so we imposed bounds from the back-projections and R1 time function directivity to settle on this range of possibilities. The hypocentral depth was 25 km, and Green’s functions were computed for the local crustal structure from Crust 1.0 (Laske et al. 2013) with addition of a 3 km deep water layer.

Four slip distributions inverted for the 16 September 2015 event are shown in Fig. 4. The solution with V r = 2.0 km/s and dip 22° (Fig. 4d) is placed on the regional map in Fig. 1b, as that model is most consistent with early reports of large geodetic motions north of the hypocenter. For the shallower dip of 16°, the fault model extends further offshore, and slip occurs along the up-dip portion of the megathrust with slip centroid depths of about 16 km. For the lower rupture velocity, slip is relatively uniform and as the rupture velocity increases it splits into two patches, with more slip in the north. For the 22° dip cases the slip locates deeper on the fault closer to the coastline with slip centroid depths of 20–24 km, with two slip patches, the southern one locating further up-dip than the northern one. The centroid times of the moment rate functions for the models vary from 50.5 to 57.3 s.

Fig. 4
figure 4

Slip models for the 16 September 2015 Illapel earthquake obtained by least-squares inversion of teleseismic P and SH waves in the passband 0.005–0.9 Hz. a Model for V r = 1.5 km/s and a dip of 16°. b Model for V r = 2.0 km/s and a dip of 16°. c Model for V r = 1.5 km/s and a dip of 22°. d Model for V r = 2.0 km/s and a dip of 22°. Waveform fits are very similar for each model; those for d are shown in Fig. 5. Each model grid has 15 km spacing between subfaults. The black arrows indicate the direction and magnitude of average slip in the fault model coordinates for each subfault with the color-scale indicating the absolute slip magnitudes. The red star is the USGS-NEIC hypocentral location. The pink outline indicates the subfaults with estimated moment larger than 15 % of the peak subfault moment used for the simplified stress drop calculation, Δσ 0.15

Full size image

The slip models have modest rake variability, but all share an average rake of about 96°. About 80–82 % of the teleseismic P and SH waveform power is accounted for by each model, and it is not sensible to favor a particular fault parameterization based on waveform fit. Figure 5 shows an example of waveform fits for the model in Fig. 4d. The P waves are particularly well fit for all 4 models.

Fig. 5
figure 5

Observed (black traces) and synthetic (red traces) broadband P and SH ground displacements at global stations, at indicated azimuths and epicentral distances. The peak-to-peak amplitude of the data (in microns) is indicated for each trace on the right. The synthetics are for the finite-fault model shown in Fig. 4d

Full size image

The along-strike distribution of large slip depends on the assumed rupture velocity and is about 180 km for V r = 1.5 km/s and 240 km for V r = 2.0 km/s. All of the models have a concentration of slip to the N or NW that accounts for the modest directivity seen in the R1 source time functions. All of the models also indicate a down-dip region of ~5 m slip extending below the coastline 30–60 km north of the hypocenter.

2.4 Source Spectrum, Radiated Energy and Stress Drop Estimation

The moment rate function for the model in Fig. 4d (V r = 2 km/s, dip = 22°) is shown in Fig. 6a, and there is a 95.5-s total duration, with a seismic moment of 2.67 × 1021 Nm (M W 8.2). The centroid time is 50.5 s, several seconds larger than the gCMT centroid time, which suggests that the last 10 s of the source model may be model noise. The overall shape of the moment rate function, which has very little dependence on the specific finite-fault inversion model, is very smooth and close to a truncated Gaussian in symmetry. This contributes to scalloping of the smooth source amplitude spectrum, with a deep notch near 0.03 Hz as shown in Fig. 6b. This source spectrum is constructed from the spectrum of the moment rate function for frequencies below 0.05 Hz and from log-averaging of far-field P wave displacement spectra corrected for radiation pattern, geometric spreading, and attenuation for frequencies above 0.05 Hz. A reference ω −2 spectrum with 3 MPa stress parameter and the same seismic moment is shown by the dashed line.

Fig. 6
figure 6

a Moment rate function for the 16 September 2015 Illapel earthquake from the teleseismic body wave finite-fault inversion in Fig. 4d. The seismic moment, M W, centroid time T c, and total duration, T d estimated from this model are shown. Note the overall Gaussian shape of the source function. b The average source spectrum for the 2015 earthquake constructed using the moment rate function spectrum in (a) for frequencies below 0.05 Hz (black line) and log-average spectra of teleseismic P wave ground displacements corrected for geometric spreading, radiation pattern, and attenuation for frequencies above 0.05 Hz (red line). The total radiated energy integrated to 1 Hz, Er, and the seismic moment-scaled values are indicated. Note the deep scallop in the source spectrum near 0.03 Hz; this is a result of the smooth truncated Gaussian shape of the moment rate function. The SH and P wave radiation patterns and data sampling for the average focal mechanism from the finite-fault inversion are shown in (c)

Full size image

The far-field radiated energy for the source was estimated from teleseismic P wave ground velocities following the basic procedure of Venkataraman and Kanamori (2004), adding relative contribution from low-frequency energy based on the moment-rate spectrum following the procedure of Ye et al. (2013a). For the case shown in Fig. 6, this gives a radiated energy estimate up to 1 Hz of E r = 2.9 × 1016 J, and a moment-scaled value E r/M 0 = 1.1 × 10−5. The various models yield a range of radiated energy estimates because of the differences in depth of the slip distributions, spanning values from 2.2 × 1016 to 3.15 × 1016 J, with E r/M 0 = 0.8–1.1 × 10−5. The broadband energy estimate posted by IRIS is 3.2 × 1016 J (http://ds.iris.edu/spud/eqenergy/10095476) based on the method of Convers and Newman (2011).

For the slip models in Fig. 4 we made two calculations of the stress drop. In a simplified procedure, we remove the subfaults with inverted seismic moment less than 15 % of the peak subfault seismic moment and then use the total area and average slip of the remaining subfaults (outlined in Fig. 4) in a uniform slip circular crack model, finding estimates of Δσ 0.15 = 1.7–3.4 MPa. To better account for the spatially varying slip effects we also calculate the shear stress near the center of each subfault (Fig. 7 shows the result for the model in Fig. 4d) and integrate the slip-weighted stress change to find estimates of Δσ E = 2.6–4.0 MPa following the procedure of Noda et al. (2013). The corresponding estimates of radiation efficiency η r = 0.17 (for the 16° dip models) to 0.47 (for the 22° dip model with V r = 2.0 km/s) are somewhat low values due to the relatively low moment-scaled radiated energy (Ye et al. 2013b).

Fig. 7
figure 7

The distribution of stress change associated with the finite-fault slip model for the 16 September 2015 Illapel earthquake shown in Fig. 4d, where the stress values are calculated near the center of each subfault. The black arrows indicate the direction of the shear stress, with the color-coding indicating the value for each subfault. The slip-weighted integral of stresses is used to calculate the stress drop Δσ E, whereas a simplified procedure that trims off subfaults with moments less than 15 % of the peak subfault moment and uses the remaining subfaults to estimate the rupture area and average slip for a circular stress drop estimate gives Δσ 0.15

Full size image

3 Discussion and Conclusion

The teleseismic parameters estimated here are not unique, both due to the limitations of seismic wave coverage and dependence on assumed parameters such as the dip. Joint analysis of seismic, geodetic, and tsunami data will soon establish more definitive values for the rupture process. However, it is valuable to have rapidly obtained first-order seismic solutions for applications such as tsunami warning as well as for comparison with additional non-unique analyses of geodetic and tsunami data. The W-phase method applied to regional data provided a stable mechanism, seismic moment, and depth obtainable within 5–10 min after the event, which was valuable for tsunami warning. Other seismic source attributes noted here were obtained within a few hours of the event.

The models with large slip near the trench are representative of all models we found with shallower dip values. This attribute is shared by the USGS-NEIC finite-fault solution for a dip of 19° (http://earthquake.usgs.gov/earthquakes/eventpage/us20003k7a-scientific_finitefault), and with some other on-line posted models (e.g., http://www.geol.tsukuba.ac.jp/~yagi-y/EQ/20150917/index.html). The along-strike placement of slip varies with the assumed rupture velocity, but in general such models place the slip up-dip of most aftershocks. The models with steeper dip do not extend as far off-shore, but tend to put the slip deeper on the fault as well, closer to the coast, in closer proximity to most aftershocks (Fig. 1b) and near the seaward region of short-period radiation imaged by back-projection (Fig. 2c). This is consistent with some other on-line posted models (e.g., http://www.earthobservatory.sg/news/september-16-2015-chile-earthquake). Any one-dimensional structural model does not account for the precise depth below seafloor and intersection with the trench.

The moment rate function is relatively robust for variations of rupture velocity, although the moment varies with dip. The relatively low rupture velocities of 1.5–2.0 km/s indicated by the high-frequency back-projections limit the along-strike expansion of the rupture. The lack of short-period bursts of energy from the large-slip regions near the trench indicated by the shallow dip models is similar to what has been observed for numerous large events (Lay et al. 2012). The steeper dip models do not have as strong of a discrepancy between the large-slip region and the back-projections, although the strongest feature in the back-projections is still deeper than the main slip patch. The moment-scaled radiated energy value is higher than typical of tsunami earthquakes (e.g., Ye et al. 2013b) and this event does not have the distinctive long centroid time shift of a tsunami earthquake. The report of a localized 4.7 m tsunami near Coquimbo suggests concentrated slip offshore northward of the hypocenter, compatible with the steeper dipping models we considered. The overall tsunami observations do not indicate the exceptionally large tsunami expected for a near-trench tsunami earthquake.

The peak slip estimates of 8–12 m indicated by our teleseismic finite-fault inversions raise questions about the relationship between the 2015 and 1943 Coquimbo earthquakes. While putatively rupturing in the same along-strike region of the megathrust, the estimated shallow slip for the 2015 event is as much as twice the amount of slip deficit expected to have accumulated since 1943. The 6 April 1943 rupture appears to have been smaller than the 2015 event, although it is listed as M W 8.1 in the ISC-GEM Global Instrumental Earthquake Catalog (http://www.isc.ac.uk/iscgem/; Storchak et al. 2013). Beck et al. (1998) model a few P waves and estimate the seismic moment M 0 is 6 × 1020 Nm (M W 7.9), similar to the Gutenberg-Richter MS(G-R) = 7.9 (Abe 1981). They obtained a simple source time function with 24 s duration from limited bandwidth data. The 1943 event produced far-field tsunami amplitudes in Japan (10–30 cm) (Beck et al. 1998) only about half as large as reported for 2015 (11–80 cm). There is a report of a minor local tsunami in 1943 near Los Vilos (Lomnitz 1971). We computed an MS(G-R) for the 2015 event of 7.8. A reasonable interpretation is that the 2015 event rupture extended shallower or further along strike to the north on the megathrust than the 1943 event, releasing strain accumulated there over a longer interval, possibly dating back to the earlier 1880 rupture.

Rupture of up to 15 m extending up-dip close to the trench was found by Yue et al. (2014) for two large-slip patches in the 27 February 2010 Maule, Chile MW 8.8 earthquake. The northern slip patch for that event was along the 1 December 1928 M W 7.7 (ISC-GEM) rupture zone, and the large slip may have released strain accumulated possibly back to the earlier 1835 event. Shallow slip was not found for the 1 April 2014 Iquique, Chile M W 8.1 earthquake (e.g., Lay and Yue 2014), within the great 1877 rupture zone. Thus, it is important to perform tsunami modeling to constrain the off-shore position of slip on the fault for the September 16, 2015 Chile earthquake. The rupture zone of the great 11 November 1922 earthquake to the north of the 2015 rupture, with previous large events in 1819, 1822, 1849, 1857, 1859 (Beck et al. 1998) is of concern for future great earthquake rupture in addition to the still un-ruptured portion of the 1877 zone in northern Chile.