Ggs 083 Font

Geophysical Journal International
Geophys. J. Int. (2013) 193, 263–286 doi: 10.1093/gji/ggs083
Seismicity patterns along the Ecuadorian subduction zone: new

constraints from earthquake location in a 3-D a priori velocity model
Yvonne Font,1 Monica Segovia,2 Sandro Vaca2 and Thomas Theunissen3

1 Universitéde Nice Sophia-Antipolis, Institut de Recherche pour le Développement (UR 082), Observatoire de la Côte d’Azur, Géoazur, 250 av Einstein
06560 Valbonne, France. E-mail: font@geoazur.unice.fr
2 Instituto Geofisico, Escuela Politecnica Nacional, Quito, Ecuador
3 Geosciences Montpellier, UMR 5243, CNRS, University of Montpellier 2, France
Accepted 2012 November 22. Received 2012 July 6; in original form 2011 June 20
Downloaded from https://academic.oup.com/gji/article/193/1/263/734190 by guest on 31 July 2023

SUMMARY
To improve earthquake location, we create a 3-D a priori P-wave velocity model (3-DVM)
that approximates the large velocity variations of the Ecuadorian subduction system. The
3-DVM is constructed from the integration of geophysical and geological data that depend
on the structural geometry and velocity properties of the crust and the upper mantle. In ad-
dition, specific station selection is carried out to compensate for the high station density on
the Andean Chain. 3-D synthetic experiments are then designed to evaluate the network ca-
pacity to recover the event position using only P arrivals and the MAXI technique. Three
synthetic earthquake location experiments are proposed: (1) noise-free and (2) noisy arrivals
used in the 3-DVM, and (3) noise-free arrivals used in a 1-DVM. Synthetic results indicate
that, under the best conditions (exact arrival data set and 3-DVM), the spatiotemporal config-
uration of the Ecuadorian network can accurately locate 70 per cent of events in the frontal
part of the subduction zone (average azimuthal gap is 289◦ ± 44◦ ). Noisy P arrivals (up to
± 0.3 s) can accurately located 50 per cent of earthquakes. Processing earthquake location
within a 1-DVM almost never allows accurate hypocentre position for offshore earthquakes
(15 per cent), which highlights the role of using a 3-DVM in subduction zone. For the applica-
GJI Seismology
tion to real data, the seismicity distribution from the 3-D-MAXI catalogue is also compared to
the determinations obtained in a 1-D-layered VM. In addition to good-quality location uncer-
tainties, the clustering and the depth distribution confirm the 3-D-MAXI catalogue reliability.
The pattern of the seismicity distribution (a 13 yr record during the inter-seismic period of
the seismic cycle) is compared to the pattern of rupture zone and asperity of the Mw = 7.9
1942 and the Mw = 7.7 1958 events (the Mw = 8.8 1906 asperity patch is not defined). We
observe that the nucleation of 1942, 1958 and 1906 events coincides with areas of positive
Simple Bouguer anomalies and areas where marine terraces are still preserved on the coastal
morphology. From north to south: (1) the 1958 rupture zone is almost aseismic and is attributed
to a zone of high coupling; (2) south of the Galera alignment (perpendicular to the trench), the
1942 rupture zone presents moderate seismicity, deeper on the seismogenic interplate zone,
and abutting on the Jama cluster (to the south). This cluster is facing the Cabo Pasado cap
and positive Bouguer anomalies on the overriding margin. We suspect that this cluster reflects
a zone of local asperity (partial coupling). South of the Jama cluster, the spherical aseismic
zone in the Bahia area is interpreted as having a low seismic coupling (steady creep motion or
slow slip events). We suspect that the site that generated the three M > 7 events (1896, 1956
and 1998) correspond to a small patch of strong coupling. To the south, in the Manta-Puerto
Lopez zone, the seismicity is mainly organized in earthquake swarms (1998, 2002, 2005).
Although slow slip events have been observed in the area (Vallée et al. submitted), we infer
from the coastline shape, the marine terraces and the high positive Bouguer anomalies that the
seismicity here might reveal a significant amount of seismic coupling.
Key words: Seismic cycle; Seismicity and tectonics; Computational seismology; Subduction
zone processes; South America.

C The Authors 2013. Published by Oxford University Press on behalf of The Royal Astronomical Society. 263
264 Y. Font et al.
plosions and seismicity on well-known tectonic features for local

1 I N T RO D U C T I O N
studies.
Interplate seismogenic zones of subduction systems generate the In this paper, we propose a 3-D approach to improve earth-
most destructive earthquakes worldwide, as these faults are su- quake location in subduction zones. The application’s objective
perficial (generally shallower than 60 km depth) and extend later- is to better resolve the seismicity distribution on or at the vicin-
ally along coastlines (Pacheco et al. 1993; Tichelaar & Ruff 1993; ity of the Ecuadorian interplate seismogenic zone. In this area, no
Heuret et al. 2011). However, the precise seismicity distribution and precise tomography covering the whole active margin has been
its temporal modification, particularly for small magnitude events, published, mainly because of the network configuration (Prevot
are often poorly assessed. Our comprehension of the interplate et al. 1996). We thus construct and present the first a priori 3-
seismogenic zone behaviour is thus restricted in time, especially D P-wave velocity model established from the integration of geo-
during the later part of the seismic cycle when seismicity is sup- physical and geological investigations carried out on the Ecuado-
posed to contribute to stress build up and concentrates around the rian margin. The velocity model is available upon request to the
periphery of the (future) rupture zone (Mogi 1979; Scholz 2002; first author. Hypocentre determination uses events recorded lo-
Bollinger et al. 2004). The absolute position of earthquakes also cally during the interseismic period, between 1994 and 2007, on
contributes to the assessment of the seismic hazard because ac- the permanent RENSIG seismological network (Red Nacional de
curate event location is needed to determine magnitude and focal Sismógrafos del Instituto Geofisco, Escuela Politecnica Nacional,

mechanism, as well as to highlight active fault geometry or seis- Quito). One of the main difficulties hinges upon the network con-
mic gaps. Improving absolute earthquake location from permanent figuration (essentially short-period and vertical component) that
seismic network monitoring is therefore crucial to better under- was initially installed to monitor volcanic hazards and is essen-
stand the seismic behaviour along subduction systems before a main tially deployed along the linear Andean Chain at about 200–300 km
shock. from the trench. The subsequent station density disequilibrium re-
In most subduction zones, limitations in assessing absolute earth- quires a specific data selection to enhance the effect of the few
quake location are specifically because of (1) scarce seismic mon- coastal stations on earthquake location (relative to the Andean
itoring near coastal areas and the lack of permanent marine net- stations).
works and (2) the fact that the three-dimensional (3-D) aspect of We perform hypocentre determination using an improved version
the heterogeneous lithospheres is generally not considered. In most of the maximum intersection technique (MAXI; Font et al. 2004;
cases, earthquakes occurring on the interplate seismogenic zones Theunissen et al. 2012) as its search process is reduced to the di-
are located offshore while seismic stations are deployed onland. rect search of the three unknowns (longitude, latitude and depth).
Consequently, the network configuration causes azimuthal gaps of- The origin time of the earthquake is then solved from the solution
ten larger than 180◦ and large source-to-station distances (>50 km), position. Designing a P-velocity model alone is a choice motivated
which are not suitable for regional tomographic inversion or earth- by the lack of knowledge about the 3-D S-velocity properties and
quake location (Bondar et al. 2004). This restricts our knowledge on the network characteristics (vertical component sensors). Before
3-D structures and seismic velocity properties. Subsequently, 1-D the application to real data, we design a 3-D synthetic experiment
velocity models are generally used despite the strong lateral seis- (that we called a resolution test) to evaluate the network capacity
mic velocity variations found in subduction zones. When computed to recover true-event positions using only P arrivals. Applying the
in 1-D media, predicted seismic ray traveltimes have large errors application to real data, we evaluate the effect of the 3-DVM by
and generate significant mislocation in hypocentre determination comparing the earthquake location to the results obtained, using
(Lomax et al. 2001; Flanagan et al. 2007). Improving earthquake the same arrival-time data set and technique, within a 1-D velocity
location in subduction zones, where deploying permanent network model and to the initial solutions from RENSIG. Earthquake lo-
with efficient azimuthal coverage is difficult, requires improving cation results, from location uncertainties, confidence factors and
traveltime estimates, which in turn essentially relies on the abil- geodynamic adequacy, attest to the efficiency of the a priori 3-D
ity of the 3-D velocity model to represent structural intricacies velocity model presented in this study to improve hypocentre loca-
such as: surface topography variations (from the oceanic trench to tion in the Ecuadorian subduction zone. To conclude, we discuss
the volcanic arc), the compositional difference between an oceanic the new event catalogue compared to the rupture zones of the four
subducting plate and a continental overriding plates, lateral seismic historical earthquakes that occurred in the area.
velocity variations produced by local tectonic structures (such as
subducting ridges, forearc sedimentary basin, magmatic chambers,
backarc basin), crustal thickness and Moho discontinuity or mantle
2 A P R I O R I 3 - D P - WAV E V E L O C I T Y
hydration.
M O D E L I N E C UA D O R
A significant step in earthquake location consists in the es-
tablishment of a priori 3-D velocity model (3-DVM) to com- This study consists in the construction of a 3-D georealistic
pensate the lack of precise tomography model. A priori velocity P-velocity model (3-DVM) to compensate for the lack of a pre-
models are constructed in part or totally by integrating informa- cise tomography model. We create the a priori model by integrating
tion on structures and velocities that results from independent in- published information on structures and velocities that we could
vestigations (e.g. reflection or wide-angle seismics, gravity, sur- gather on Ecuador (or comparable subduction zones). The resulting
face geology; Nataf & Ricard 1996; Ricard et al. 1996; Bhat- model is totally independent of tomographic inversions and recov-
tacharyya et al. 2000; Pasyanos et al. 2001; Johnson & Vincent ers, as well as possible, the first-order lithospheric variations in terms
2002; Font et al. 2003; Husen et al. 2003; Rhodes 2004; Flana- of geometry and velocity. The model extends from northern Peru
gan et al. 2007; Font & Lallemand 2009). This type of models (6.5◦ S) to southern Colombia (3◦ S), between 277◦ E and 283◦ E and
improves traveltime estimates and earthquake location accuracy as reaches 300 km depth. The parametrization of the velocity model is
evidenced by the error reduction relative to reference events such discussed at the end of this chapter. The following section presents
as nuclear bombs for teleseismic hypocentre determinations or ex- the information we used to construct the velocity model.
Seismicity along the Ecuadorian subduction zone 265
2.1 Geodynamic, structural and velocity data the model construction and structural units, we refer under the term
“North Andean Block” to the part of the block that extends from the
Subduction occurs beneath northern South America since at least
trench to the western Andes foothills. The real limit of the North
Early Jurassic time (Aspden et al. 1987; Jaillard et al. 1990; Kerr
Andean Block is still debated but we know it is located eastward,
& Tarney 2005). Many works have been conducted onland and
along the Andean Chain, and that it extends from the Puna Island
offshore which allow reconstructing the geodynamic evolution of
to north of Venezuela.
the region and highlight the complexity and 3-D structure of the
The surface topography included in the 3-DVM is irregular and
Ecuadorian subduction system. For the purpose of a 3-DVM con-
presents huge variations from about −4000 m near the trench to
struction, we summarize those works, distinguishing four main
+6000 m at some active volcanoes (Michaud et al. 2006). The
structural units over the southern Colombia—northern Peru region.
trench-to-coast distance (from 50 to 190 km) and trench geometry
Each of them has its own geometry and seismic velocity properties
vary laterally along the subduction, increasing in obliquity from
that one needs to consider to improve wave propagation compu-
∼10◦ in Central Ecuador, to ∼30◦ in southern Colombia (Trenkamp
tations. The structural units are: The Nazca Plate and associated
et al. 2002).
downgoing slab, and the overriding plate composed of three re-
gions: (1) the continental Guyana Shield and sedimentary basin
(east of the Andes), (2) the Andean Chain and inter-Andean val-
2.1.1 The Nazca Plate

ley (both a volcanic arc and geological suture) and (3) the North
Andean Block composed of accreted oceanic plateaus and a thick The age of the Nazca Plate (subducting beneath the South America
sedimentary cover (Figs 1 and 2). Note that in the presentation of Plate) and its segmentation varies along the subduction zone, mainly
Figure 1. Geodynamical framework of the Ecuadorian subduction zone. Simplified faults traces modified from Vallejo et al. (2009): CF = Cosanga Fault;
CPPF: Calacalı́ – Pujilı́ – Pallatanga Fault. Bathymetry/topography map from Michaud et al. (2006). Limits of the main sedimentary basins from Deniaud
(2000): BB = Borbon Basin; MB = Manabi Basin; PB = Progreso Basin.
266 Y. Font et al.
Figure 2. (a) Example of the lithospherical and upper mantle model of Ecuador along an E–W cross-section at 1.5◦ S established for the construction of the
3-DVM; thin lines define the principal limits of the structural units (sedimentary layer, crustal basement and upper mantle, for both tectonic plates—see colour
code within the figure). (b) Cross-section within the velocity model and geophysical data existing at (or near) this latitude (active seismic refraction from
Graindorge et al. (2004) and Ocola et al. (1975); crustal model from Simple Bouguer gravity field of Feininger & Seguin (1983); seismicity distribution from
Engdahl et al. (1998, in white) and RENSIG (in grey). Thin lines define the same principal limits of the structure shown in (a); (c) Active seismic refraction
model from Graindorge et al. (2004). (d) Crustal profile determined from Bouguer anomalies (Feininger & Seguin 1983). (e) 1-D velocity model along the
Andean Chain resulting from the Nariño active seismic experiment (from southern Colombia to central Ecuador, from Ocola et al. (1975)).
because of the Farallon Plate splitting along a preexistent NNE (4) Between the Carnegie Ridge and the Grijalva Fracture Zone,
fracture zone during Oligocene time (∼ 30–25 Ma; Hey 1977). The the trench is narrow (3–4 km wide), at 4 to 4.2 km depth and contains
crust is about 30 Ma south of the Grijalva Fracture Zone, 22 Ma about 300 m of infill deposit (2–2.2 km s–1 ; Calahorrano et al. 2008).
north of it, and decreases northwards to about 10 Ma at the southern The accretionary prism widens to 3 km. The subduction channel is
limit of the Yaquina Graben (Fig. 1; Lonsdale 1978, 2005; Lonsdale thin (maximum of 0.6 km) and imaged to 25 km landward from the
& Klitgord 1978). The rupture of the Farallon oceanic plate evolved trench (∼7.5 km deep). The Nazca Plate is locally dipping landward
into the E–W trending Cocos-Nazca Spreading Center. Around 23– at 6◦ . Off the Gulf of Guyaquil, a 3–10 km wide and 1 km thick
20 Ma, the interaction between the Galapagos Hot Spot and the sedimentary prism fronts the margin (Calahorrano et al. 2008).
spreading centre results in the formation of the Cocos (and Malpelo) Sediment accumulation is favored south of the Grijalva Fracture
Ridge north of the spreading centre and the Carnegie Ridge south Zone because the trench is deepest. The trench is 10 to 15 km
of it (Sallares & Charvis 2003). Near the Ecuadorian Trench, the wide, about 4.8 km deep and filled by about 500 m of sediment.
aseismic Carnegie Ridge is about 280 km wide and 2 km high with The subduction channel is visible from the trench (0.2 km thick)
respect to the neighbouring abyssal plain. reaching a thickness of ∼1 km before thinning to only 0.3 km at
The oceanic crust thickness varies laterally beneath the overriding ∼30 km of the trench landward (low velocity 2.8 km s–1 relative
margin as observed on wide-angle data, reaching up to 19 km be- to the 3.5–3.8 km s–1 of the overlying upper-plate basement). The
neath the Carnegie Ridge crest (Sallares & Charvis 2003) and 14 km Nazca crust is at 4.5 km depth before trench.

on its southern flank (central Ecuador; Calahorrano 2001; Grain-
At greater depth (between 40 and 300 km), as imaged by local
dorge et al. 2004) and thinning down to 5 km off southern Colombia
seismicity, the downgoing slab dip-angle does not vary significantly,
(Meissner et al. 1977; Garcia 2009). Near the trench, oceanic base-
although it is slightly higher to the north and lower in northern Peru.
ment velocities vary from ∼5 km s–1 (at the top of oceanic layer
A flat slab had been proposed beneath Ecuador based on scarce
2) to ∼7.0 km s–1 at the base of oceanic layer 3 (Figs 2b and c).
teleseismic records (Gutscher et al. 1999) but is controversial based
The Moho discontinuity is represented by a strong velocity gradient
on other local seismicity observations. Indeed, in northern Ecuador,
from ∼7 to 7.8 km s–1 in less than 1 km of depth (averaged from
at depths greater than 30 km, the slab dips about 25–30◦ down to
(Graindorge et al. 2004; Agudelo et al. 2009).
140 km depth (Pontoise & Monfret 2004; Manchuel et al. 2009;
The Ecuadorian margin is highly segmented and mainly ero-
Manchuel et al. 2011), 25–35◦ down to 200 km in Central Ecuador
sional (e.g. Collot et al. 2002). Sedimentary variation of trench
(Prevot et al. 1996; Guillier et al. 2001) and reduces to about 25◦ in
filling, downgoing subduction channel thickness, physical proper-
northern Peru (Tavera et al. 2006).
ties and/or patchiness on the interplate, accretionary prism, frontal
In the 3-DVM, the slab geometry is extrapolated from both lo-
forearc basins and also shallow interplate dip angle are given by
cal (RENSIG) and teleseismic catalogues [EHB catalogue from
Calahorrano et al. (2008), Collot et al. (2004, 2008), Marcaillou &
Engdahl et al. (1998); Fig. 2b]. In the northern part of the country,
Collot (2008), Sage et al. (2006), Witt et al. (2006). The shallowest
the 1-D deep seismic profile, resulting from the Nariño active seis-
part of the 3-DVM is constrained by these structural data.
mic experiment (Ocola et al. 1975), is also taken into consideration
Along-strike structural variations may be described along four
(Fig. 2e). When not clearly constrained by seismicity, the slab depth
segments.
beneath active volcanoes is forced to range between 80 and 120 km
depth (Fig. 2b). Slab dip is considered constant down to 300 km
depth. Seismic velocities beneath the oceanic Moho are taken from
(1) North of the Esmeraldas Canyon (Manglares segment after
the iasp91 model (Kennett & Engdahl 1991).
Marcaillou et al. 2006), the trench is locally disrupted and filled
by ∼3 km of turbidites. The accretionary prism is less than 5 km
wide or absent (Collot et al. 2008). The narrow shelf (5–20 km) is 2.1.2 Guyana shield and sedimentary cover
affected by an 80 km wide and ∼1 km deep reentrant (Collot et al. The Nazca Plate underthrusts the South America Plate. East of the
2004; Marcaillou et al. 2006). Near the trench, the Nazca Plate dips Andes, the Amazon Basin (or Oriente) is a Mesozoic to Cenozoic,
about 7.5◦ eastward (at ∼4.5 km depth) and is only slightly affected hydrocarbon-rich sedimentary foreland basin that includes a plat-
by normal faults (contrarily to the northern Colombian segment; form carbonate sequence, overlying the older cratonic Guyana base-
Marcaillou et al. 2006). The downgoing subduction channel is about ment, both intruded by large granitoid batholiths along the bound-
2 km thick and is considerably reduced at 60 km from the trench ary with the Andes (Christophoul et al. 2002; Hughes & Pilatasig
(∼15 km depth; Agudelo et al. 2009). The Nazca crust thickness 2002; Cediel et al. 2003). The sedimentary thickness reaches 10 km
is about constant (6.5 km) from the oceanic basin to 60 km east of (Thomas et al. 1995) and is characterized by a velocity gradient
trench (Agudelo et al. 2009). between ∼2 to 5.5 km s–1 . From gravity modelling, the average con-
(2) Between the Esmeraldas canyon and the northern extent of tinental crust thickness is estimated between 30 to 35 km (Fig. 2d;
the Carnegie Ridge (1.5◦ N–0.5◦ N), the continental shelf widens to Feininger & Seguin 1983). Velocities increase from 5.8 km s–1 (at
50 km and shallows to less than 100 m (Collot et al. 2004). The top of continental basement) to 6.8 km s–1 near the Moho disconti-
well-marked trench deepens to ∼3.7 km and contains up to 500 m nuity.
sedimentary fill (Collot et al. 2002). The Nazca Plate lies at ∼5 km
depth beneath the trench fill. The dip angle is 3◦ along 35 km from
2.1.3 Andean chain and inter-Andean valley
the trench and increase to ∼15◦ eastward.
(3) Along the Carnegie Ridge segment, the trench shallows West of the Amazon Basin, the Ecuadorian Andes, in broad ap-
to ∼2.8 km and contains only a few turbidites (Lonsdale 1978; proximation, consists of two parallel ranges, oriented NNE–SSW
Collot et al. 2004). The interplate boundary is dipping 10◦ 15 km and reaching about 6 km of elevation: the Western Cordillera (to
from the trench. A thin subduction channel locally thickens to form the west) and the Real Cordillera (to the east). The Real Cordillera
low-velocity 600-m-thick lenses of underthrusted fluid-rich sedi- consists mostly of sub-linear belts of Palaeozoic to Mesozoic meta-
ment (Sage et al. 2006). morphic rocks, accreted to the palaeo-continental margin since
268 Y. Font et al.
the Mesozoic (Aspden et al. 1992). They are covered by Ceno- axes following E–W and N–S directions, respectively. The SW cor-
zoic to modern volcanosedimentary units. The Western Cordillera ner is assigned to be the origin point, at 277◦ E and 6.5◦ S. The
consists of a volcanic arc that lies astride a basement composed heterogeneous velocity structures are characterized by a set of non-
of accreted oceanic plateau (same basement as the coastal area; overlapping, equal volume and constant VP velocity blocks. Each
Hughes & Pilatasig 2002). Both cordilleras are separated by the block dimension is 12 × 12 × 3 km in the x, y and z directions,
inter-Andean graben filled by Pliocene and Quaternary volcanosed- respectively. Therefore, structures thinner than 3 km (such as the
imentary rocks (sedimentary velocity for about 5 km of thickness; subduction channel, for instance) are incorporated in the average
Fig. 2d). Feininger & Seguin (1983) suggest that the basement un- velocity of the concerned blocks. The entire velocity model consists
derlying the inter-Andean graben consist of Real Cordillera rocks. of 516 120 blocks. In accordance with the shortest ray path method
The inter-Andean graben and active Calacali-Pujili Fault zone (at (Moser et al. 1992), ray tracing nodes are placed only on facets,
the eastern limit of the Western Cordillera) is considered as marking edges and vertices of blocks and each of them has a constant slow-
the suture zone between the continental basement to the east and ness (1/VP ). Because blocks are of constant velocity, that is straight
the oceanic one to the west. Within the 3-DVM, seismic velocities ray path within each block, there is no need for nodes inside the
are subsequently slightly lower in the Real Cordillera than in the blocks. The interval between two adjacent nodes is 3 km along the
Western Cordillera. The crustal depth and reflective Moho is not horizontal directions and 1 km vertically. Each block consists of 82
well constrained in this area. Active seismic, seismology and grav- nodes and 15 745 384 nodes constitute the whole velocity model.

ity modelling agree with a crust thickness reaching between 50 and This parametrization is propitious for the dimension of our velocity
66 km depth (Ocola et al. 1975; Leeds 1977; Feininger & Seguin model but it shall be adapted accordingly to the size of the region
1983; Prevot et al. 1996; Guillier et al. 2001) but no strong Moho under study (Theunissen et al. 2012).
discontinuity has been seen. In the 3-DVM, east and west and the In summary, the 3-DVM constructed here integrates the data
Andean Chain, the Moho is represented, at near 30 km in depth, by known on the heterogeneous crust and mantle velocities, mimick-
a strong velocity gradient (from ∼7 to 8 km s–1 in less than 2 km ing the topography variations, the sea water and air columns, the
thick layer). Beneath the Andean Chain, the Moho gradient is sig- sedimentary layers and crustal thicknesses, the crustal lithologies
nificantly smoothed and the velocities of 7 and 8 km s–1 are marked and the slab dip angle. The model includes the effect on near-surface
at 55 and 65 km depth, respectively. seismic-wave velocities produced by some local tectonic structures
such as the Carnegie Ridge or the Inter-Andean Valley. The 3-DVM
aspires to represent, at least in broad lines, realistic velocity varia-
2.1.4 North Andean block and forearc basins tions to better control traveltime estimates in a region where local
tomography does not exist.
West of the Andes, the Coastal Plain basement is characterized by
allochthonous oceanic terranes accreted to the continental margin
during Cretaceous to mid-Eocene times that are locally exposed in
the Coastal Cordillera (Litherland & Aspden 1992; Kerr et al. 2002).
Three thick sedimentary basins overlay the oceanic basement and 3 S E I S M O L O G I C A L D ATA
cover a large part of the Coastal Plain (Daly 1989; Deniaud 2000).
Since 1990, the RENSIG is recording the seismic activity in Ecuador
From north to south, they are Borbon, Manabi and Progreso basins.
(Segovia & Alvarado 2009). This study uses the arrival times man-
Sedimentary fill may reach up to 10 km. The seismic velocities
ually picked by the staff of the Institute of Geophysics of the events
within the forearc basins increase with the sedimentary thickness,
recorded between 1994 and 2007. During this period, the RENSIG
reaching about 5 km s–1 for a thickness of about 8 km. Based on
documented about 44,000 seismic events. The network is teleme-
gravity modelling (Case et al. 1973; Feininger & Seguin 1983),
tered and essentially composed of vertical component, short-period
the North Andean Block crustal thickness varies between 22 km
stations. Accurate subduction earthquake location is complicated
(near the coast line) to 30 km (western flank of the Andes), locally
because of the seismological network configuration, which has been
thinning beneath the Coastal Cordillera. The crustal seismic velocity
for the last 15 yr essentially distributed over the Andean Chain to
gradient increases from 4.8 to ∼7 km s–1 .
monitor volcanic hazard. The distribution of the seismic stations
from Ecuadorian network shows a density discrepancy between the
Coastal (20 per cent) and the Andean regions (80 per cent).
2.2 Constructing the velocity model and grid
To avoid the excessive influence of seismic stations on the Andes
parametrization
relative to the Coastal region, we select all available stations on the
To construct the 3-DVM, we integrate on a series of vertical E–W Coastal region and one on each volcano subnetwork (Fig. 4). The
cross-sections (each 0.5◦ ) the multiple and independent geologi- choice of the volcanic station is based on number of records between
cal/geophysical observations described above to differentiate the 1994 and 2007, signal-to-noise ratio and geographic position (pref-
structural units (sedimentary layers, crustal basement and upper erentially located west of the volcano, to avoid as much as possible
mantle, for both tectonic plates) in terms of geometry as well as velocity structures related to the volcanic edifices). Note that during
P-velocity (VP ) gradients. Structural geometry and VP gradients are the hypocentre determination process, only one station at a time can
used to characterize the top and bottom of the envelopes that limit be selected in the groups vc1-mary-tamb and in ara2-retu (Fig. 4).
each unit in space. The final model is obtained by interpolating In total, the selected subnetwork is composed of 39 stations.
the seismic velocities to the whole region and by averaging seismic In addition to the difficulty of the uneven spatial distribution, the
velocities in blocks of constant velocities (Fig. 3). The technical performance of the seismic stations on the Coastal area has been
approach is similar to that used by Font et al. (2003). intermittent (Fig. 4). Only four coastal stations can be considered as
The gridding of the velocity model is parametrized into a blocky nearly permanent during the 14 yr: jama, hoja, lore, cupa with 13,
schema, following the approach presented by Zhou (1994). Hor- 12, 12 and 11.5 yr of recording, respectively. Stations magd and sali
izontally, the studied area covers 648 × 1092 km2 , with x and y arrive in second position with 10 and 7 yr of recording, respectively.
Figure 3. (a) Horizontal and (b) vertical cross-sections showing VP lateral and vertical variations within the 3-D geo-realistic a priori P-velocity model
(3-DVM) constructed in the study. The 3-DVM is established from the integration of geological and geophysical information collected on the Ecuadorian
subduction system. It extends down to 300 km depth. Triangles on both shallower sections represent the position of seismic stations involved in the earthquake
location procedure.
270 Y. Font et al.

Figure 3 – continued
Based on RENSIG hypocentral determinations, five geographical 4 A B S O L U T E E A RT H Q UA K E L O C AT I O N

groups share the seismicity distribution as following: 72 per cent TECHNIQUE
of earthquakes are located on the Andean Chain (group A),
To search event location within a 3-D medium, we use an updated
6 per cent offshore (group O), 6 per cent on the Coastal Plain
version of the maximum intersection method. A description of the
(i.e. on the margin between the coast and the western flank of
method and application to the southern Ryukyu seismicity (off-
the Andes, group M), 9 per cent south of 2◦ S (group S) and
shore Taiwan) within a 3-DVM of the region are shown by Font &
7 per cent are deeper than 50 km (group D). Considering only
Lallemand (2009), Font et al. (2004) and Theunissen et al. (2012).
P phase arrival times, we select all events that were recorded
Therefore, only a brief overview of the technique will be given here.
by at least four seismic stations, among which at least one sta-
tion belongs to the same geographic group than the event (all
available stations for group D). In total, 12 398 events are kept, 4.1 Maximum intersection technique
and 112 148 P-wave measurements compose the arrival-time data MAXI is a direct-search procedure based on graph theory and well
set. adapted to strongly heterogeneous environments. The single event

Figure 4. Station map. Triangle = seismic stations selected from the RENSIG network to improve hypocenter determination of the seismicity occurring in
the vicinity of the interplate seismogenic zone (subduction front). Triangle colour code: record frequency. Histograms represent the number of P arrivals along
time (1994–2007). Station positions and associated histograms illustrate the distribution heterogeneity, in space and time, of the permanent network (RENSIG).
Bathymetry and topography relief is from Michaud et al. (2006).
position is solved by searching for the maximum intersection num- volume is totally independent of the earthquake origin time as it
ber of equal differential time (EDT) volumes (Zhou 1994). An EDT only depends on arrival time difference and traveltime computa-
volume is described as the collection of grid nodes satisfying the tion. Subsequently, from EDT properties, the MAXI search process
arrival-time difference between two stations ± a tolerance value directly determines the three coordinate parameters (longitude, lat-
(in seconds) known as TERR. The node collection organizes in itude and depth). The origin-time is then evaluated a posteriori
space along a deformed 3-D hyperbolic volume. In a network of j once the hypocentre position is solved. Thus, the origin-time esti-
stations, C 2j = j×(2j−1) EDTs can be constructed and shall intersect mate does not affect the hypocentre search procedure. Accordingly,
around the hypocentral solution. Theoretically, an EDT volume in- except for specific situations, a good approximation of the depth pa-
cludes the hypocentre. It is important to acknowledge that an EDT rameter can be achieved using only P arrivals (if the 3-D P-velocity
272 Y. Font et al.
model well represents crust and mantle velocity heterogeneities), Broadly, the main advantages of MAXI is (1) to eliminate the ori-
from the intersections of all possible combination of EDT volumes. gin time unknown from the search procedure and subsequently to
The specific situations where the depth unknown can not be well re- retrieve the depth parameter independently from the origin time, (2)
solved concerns the cases where (1) all stations are at equal distance to show good performances when azimuthal coverage is poor, (3)
of the hypocentre, (2) the first arrival is a head wave at all stations to objectively filter inconsistent arrival times and (4) to be less sen-
and (3) the stations are so far that the curvature of the wave can be sitive to velocity model inaccuracies than techniques that minimize
neglected. In this study, thanks to the a priori station selection that traveltime residues (Theunissen et al. 2012).
always involves a near-by station, those specific cases should not
occur. 4.2 MAXI parametrization
For the establishment of the EDT volume, grid nodes are col-
lected by scanning a pre-computed reference file. This reference The MAXI earthquake location process will search for the so-
file stores the ray tracing traveltimes from each node of the dense lution showing the best consistency between input parameters
grid that covers the modelled region to each seismic station. The (such as arrival times, station positions and the discrete 3-DVM)
book-keeping system is strictly identical to the one described by and the parametrization of the system (e.g. grid geometry, TERR
Zhou (1994) who applied the shortest path method (Moser 1991; variations). A successful search process greatly depends on the
Moser et al. 1992) to compute ray tracings. parametrization implemented for the velocity grid and the MAXI

The algorithm works as follows. First, an iterative approach on process, that in turns depend on the network configuration. In this
the parameter that controls the EDT thickness (TERR) retrieves a study, because of the seismic stations distribution (∼80 per cent
spatial volume (V1) including high-probability solutions, that is all aligned on the Andean Chain), using the MAXI technique by com-
nodes that have been intersected by a maximum number of EDTs at bining all possible pairs of stations provide EDT data sets that (1) are
a given iteration. Intersection scores obtained at each iteration are spatially tangent (approximately normal to the trench) when focus-
stacked for each node of V1. V1 is then subdivided into two volumes ing on offshore earthquakes and (2) enhance the velocity anomalies
(V2 and V3) in function of the stacked EDT-intersection scores. not taken into account beneath active volcanoes.
V3 only includes best-scored nodes and is the smallest volume. To avoid these difficulties, we limit the number of EDT volumes
V2 contains good-scored node (best intersection-scores minus a to the set of pairs that include the station with the earliest arrival
confidence margin) and is larger than V3. time. For an earthquake recorded by j stations, (j−1) EDTs are
Second, a statistical analysis on the EDT intersection rate per thus used to determine the solution (at a given TERR value). This
seismic station is conducted among V2 nodes to detect and re- choice gives more weight to the first station and its surrounding
move inconsistent arrivals from the initial data set. Third, the it- velocity model, in the sense that all EDTs rely on the first station
erative TERR approach, reconducted within the V4 (i.e. volume arrival time and are combined only once with the remaining stations,
expended from V2 to account for grid approximation) on a hi- thus reducing the harmful effect of structural anomalies under the
erarchy of grid discretization and using cleaned data, better sur- volcanic arc. Consequently, this choice is particularly advantageous
veys the solution domain and focuses on a small volume (V5) that for the location process of earthquakes occurring offshore or near
may include several nodes accounting for the same maximal EDT the coast. However, it should be noted that if the first arrival is
score. The final absolute hypocentral solution is the barycenter erroneous, the hypocenter determination will be shifted in space
of V5. (Font et al. 2004).
Solution reliability of a seismicity data set can be assessed by the Based on the grid parametrization and on classical estimations of
analysis of a series of confidence factors and by discarding the de- picking errors for regional earthquakes (Zeiler & Velasco 2009), we
terminations presenting statistically extreme values of these factors. implement the TERR parameter (that controls the thickness of EDT)
Among those, the QEDT parameter quantifies the ratio between the to increase from 0.4 to 0.8 s with an increment of 0.1 s. The initial
number of EDTs effectively intersecting the solution and the theoret- node spacing is defined above; the secondary grid is composed of
ical maximum number of EDTs involved in the process. Therefore, nodes equally spaced and distant of 0.5 km.
the QEDT directly characterizes the consistency between the input
data and system parametrization. QEDT is unitless and varies from
5 S Y N T H E T I C R E S O LU T I O N
0 to 1. When QEDT is close to 1, one can consider that there is
I N V E S T I G AT I O N
a good consistency between data and system parametrization. In
those cases, MAXI confidence factors can also geometrically judge The configuration of seismic station in Ecuador between 1994 and
the way a solution is constrained thanks to graphical relations of 2007 is not constant through time and only includes a few coastal
the solution-node distribution. Two types of factors are considered: stations. To evaluate the ability of the selected subnetwork to prop-
(1) size, node density and node distribution of volumes V1, V2 and erly locate subduction earthquakes using P arrivals and the MAXI
V3 and (2) spatial coherency of the volumes and their organization technique, we design a synthetic investigation aiming to mimic the
in space, which is mainly assessed by the distances between the realistic azimuthal gap and 3-D configuration.
barycenter of the volumes V1, V2 and V3 (if the distance is too The first step consists in retrieving 3-D synthetic P-wave arrival
large, volumes do not well superimpose, which suggest multimodal times, with the objective to simulate the acquisition of seismic waves
solutions). Each factor taken independently leads to different inter- that propagate within a strongly heterogeneous medium. To do so,
pretations and only the combination of a series of factors leads to we extract all earthquake positions from the initial RENSIG data set
understanding the location process (see Theunissen et al. 2012 for (∼12 400 events). From a known event position, we retrieve the 3-D
more details). traveltimes applying the ray tracing shortest path method (Moser
At last, once the geometrical node distribution is evaluated 1991) within the 3-DVM, and using a fixed origin time, we establish
through the confidence factors, statistical analyses on traveltime the synthetic arrival time. Synthetic arrivals are only restored at the
residue rms and location uncertainties (based on Sambridge & station positions that genuinely recorded the earthquake to re-create
Kennett 1986) can confirm the reliability of the solution. the network conditions at the time of the recorded event.
Table 1. Synthetic investigation results or study feasibility. The event thetic arrival times and perform earthquake location following the
distribution (in per cent) is classified per intervals of misfit (the misfit being exact same procedure (Fig. 6). The disturbance follows a Gaussian
the distance between the hypocenter solution and the initial event position). distribution between +/− 0.3 s. Allowing for noise in the arrival
Each column refers to a specific geographic group; the number of events is time data set decreases the quality of hypocenter determinations.
given in the second line of each column. In the best conditions of a known
Still, 51 per cent of the events from group O and M have a misfit
velocity model and exact P-waves arrival times, the application of the
MAXI technique using the spatial-and-temporal distribution of P-arrivals
lower than 10 km (1050 events with less than 5 km error).
within the 3-DVM resolves ∼70 per cent of hypocenter determination The last test reproduces the synthetic experiment (with exact
offshore Ecuador with less than 5 km of error. 3-D arrival times) but processing earthquake location within a 1-D
velocity model. The 1-D model is taken from the RENSIG (Table 2)
Andes Deep South Margin Offshore
and is projected on the same grid parametrization than the one
5470 1778 1843 1447 1860 used in the 3-D case. The classification of the results per range
00 < misfit < 05 km 90.3 80.4 69.2 57.2 67.5 of misfit (Fig. 7) is shown for comparison with 3-D experiment
05 < misfit < 10 km 5.2 12.5 20.3 20 18.4 (Fig. 5 or 6). The results confirm that, for the Ecuadorian subduction
10 < misfit < 15 km 1.2 2.9 6.6 10.7 6.5 configuration, processing earthquake location within a 1-D-velocity
15 < misfit < 20 km 0.5 1.8 1.4 4.6 3 model almost never allows accurate hypocenter position for offshore
20 < misfit < 50 km 0.3 1.4 1.9 5.5 2.2 earthquakes. This experiment highlights the damaging impact of

Misfit > 50 km 0.2 0.2 0.3 0.3 0.2 traveltimes computed in a 1-D medium when seismic rays have
No solution 2.3 0.8 0.3 1.7 2.2 indeed travelled into a heterogeneous medium and enhances the
importance of a realistic velocity model.
The synthetic tests shown here validate the legitimacy of the 3-D
Second, we search for the absolute earthquake positions using approach proposed in this paper, that is the joint use of P arrivals and
the exact 3-D synthetic arrival data set, the 3-DVM and the MAXI the MAXI technique, especially because no detailed tomography
technique. Grid and technical parametrizations are identical to those exists. The azimuthal gap dispersion observed for events associated
described in the previous paragraph and that are also used for the to small misfits lead us to avoid an a priori selection of data as a
real-case application. function of gap. Therefore, the whole data set of 12 398 events is
The accuracy of earthquake location for the synthetic experiment used in the real application.
is recovered by the statistical analyses of the misfit, defined as the
distance (in kilometres) between the final solution and the initial
synthetic event position. The efficiency of the 3-D approach and
6 A B S O LU T E H Y P O C E N T E R
network configuration is assessed through the hypocenter determi-
D E T E R M I N AT I O N : P R O C E S S I N G ,
nation accuracy. The average misfit per geographic group (A, D,
R E S U LT S A N D C O M PA R I S O N
M, O and S) and per misfit interval is shown on Table 1. As ex-
pected, in broad approximation, mislocation increases when earth-
6.1 Processing
quakes occur far outside and off the seismic network. In group A,
96 per cent can be determined with less than 10 km of error while The application to the real case involves three steps. The two first
93 and 90 per cent characterize group D and S, respectively. Nev- steps consist in identifying incorrect stations and evaluating station
ertheless, 77 and 86 per cent of events can still be determined in correction terms. The last step consists in locating the earthquake
groups M and O, respectively, with less than 10 km of error, even data set applying static station delays to integrate velocity model
though source-to-station distance and azimuthal gap increase. errors under each station.
The statistical analysis on the misfit values and the spatial repar- First, the P arrivals selected from the 14-yr long RENSIG cat-
tition for synthetic events of group M and O (Fig. 5) reveals that, alogue are used to locate the 12 400 events, using MAXI (and
under the best conditions (i.e. when the 3-DVM and arrival times the parametrization described previously) within the 3-DVM estab-
are perfectly known), ∼2085 among 3300 events can be accurately lished for whole Ecuador. All 39 seismic stations from the selected
located (with a misfit <5 km; ∼2700 events with less than 10 km subnetwork are used. To evaluate static station terms, we analyse
of error). The average misfit for the whole event data set is 5.5 km the resulting phase outputs for preliminary hypocenter solutions
(±7.1) km, with dx, dy, dz reaching respectively −0.5 (±4.2) km, (i.e. before removing inconsistent arrivals from the initial data set).
0.1 (±1.9) km and 2.4 (±7.4) km. Note that the spatial position of The MAXI technique produces two types of information for each
the synthetic events does not seem to influence the misfit as the earth- phase involved: (1) EDT-intersection rate per phase (i.e. per station,
quake distribution covers the same area whatever the misfit range. because in this case only P arrivals are taken into account) and (2)
Considering the network aperture, the azimuthal gap for events of traveltime residual per station. Because of the choices in MAXI
group O associated to a misfit lower than 5 km is, on average, 289◦ procedure (i.e. all EDTs are only combined to the first arrival), we
(±44◦ ). The azimuthal gap average roughly increases by 40◦ from separate information associated to ‘a first station’ from the others
small (<5 km) to large (>50 km) misfit values for group O, and by (that we will call later on ‘the remaining phases’). The reason for
∼10◦ for group M (Fig. 5). However, while the average increases, this separation is that first station traveltime residuals are predictably
the dispersion of azimuthal gap values around the average is large very small and meaningless for static term recovering. Based on the
(± 40 and ± 60◦ for group O and M, respectively). Consequently, remaining phases, we distinguish, in terms of EDT-intersection rate
location error cannot be directly related to poor azimuthal gap. per station, EDT-in phases from EDT-out ones. Phases qualified as
The noise-free synthetic investigation indicates that, within an EDT-out are arrival times whose associated EDTs do not cross the
exact 3-D P-velocity model, the spatiotemporal configuration of solutions. By extension, we call EDT-in the phases associated to
the network is appropriate enough to accurately locate a reasonable EDTs that do cross the solution. This distinction allows us to objec-
percentage of earthquakes in the frontal part of the subduction zone. tively and statistically evaluate the phases that are inconsistent with
For a more realistic test, we introduce some noise into the 3-D syn- the remaining set of data.
274 Y. Font et al.
Figure 5. Noise-free synthetic experiment to examine the ability of the network (and its spatiotemporal heterogeneity) to solve earthquake locations for events
of group O and M (Offshore and Margin) within the 3-DVM. Initial event position (empty side of the stroke) is from the RENSIG. Synthetic 3-D P-wave arrival
times are retrieved with the genuine azimuthal gap. The parametrization of the MAXI technique and 3-DVM grid is identical to the application to real data (see
text for more details). The circles represent hypocenter solutions; the colour code indicates the azimuthal gap. Each map corresponds to a given misfit range
(see also Table 2). Below each map, the associated histogram evidences the absolute depth difference between the initial event and the solution (normalized
on the number of events involved in each range). The synthetic investigation suggests that the network spatiotemporal configuration is appropriate to resolve
70 per cent of earthquake locations offshore Ecuador.

Figure 6. Same as in Fig. 5 except that 3-D synthetic P arrivals are disturbed with noise (following a Gaussian distribution between ±0.3 s). The circle colour
code indicates the absolute depth difference between the initial synthetic event and the 3-D solution. The synthetic investigation suggests that, considering
noisy arrivals, the network’s spatiotemporal configuration is appropriate to resolve 50 per cent of earthquake location offshore Ecuador.
Table 2. 1-D velocity model (named ASW) used In the second step, we relocated all seismicity using 37 stations,
for the RENSIG hypocenter determinations. The including ecen and carm only to retrieve residual values (but with-
depth of the each layer top limit is given with out using the stations in the search process). We then computed
respect to the sea level. static station corrections following the same event data separation
VP (km s–1 ) VS (km s–1 ) Layer depth (km) as explained above. The residual statistic analysis is conducted sepa-
rately for each geographic group of events (Andes–A-, Margin–M-,
3.32 1.94 3.30
5.90 3.45 0.30 Offshore–O-, South–S- and Deep–D-). Note that the rms of the
6.20 3.63 −11.70 residual distribution are small and never exceed 0.2 s, even for
6.70 3.92 −26.70 the stations ecen and carm, attesting that traveltime residuals are
8.10 4.74 −46.70 distributed according to narrow Gaussians. Residual averages for
stations ecen and carm are about −7 and −1 s (Table 3), which
Analysing EDT-intersection rate per station, 37 stations out of 39 suggests that, in the first location trial, EDTs from the stations ecen
are qualified as EDT-in at 86 per cent (on average), indicating that and carm did affect the hypocenter solutions. From the sharp shape
those stations create, by combination with whatever other station, of the Gaussian distribution, we consider that some velocity anoma-
EDTs that do intersect the hypocenter determination 86 per cent lies under the stations had not been taken into account within the
of the time. Both remaining stations (ecen and carm; Fig. 4) are 3-DVM, except for station ecen whose large residual average is not
EDT-out at 98 and 90 per cent (with ∼1000 and 100 arrivals for easily explained except by a station mislocation or a timing error.
ecen and carm, respectively), which mean that EDTs associated to Static term values are reported in Table 3. All values are relatively
both stations almost never intersect the hypocenter determination. small.
Residual averages reach nearly −8 and −2 seconds for ecen and The third stage relocates all seismicity, separated per geographic
carm, respectively. group of events, applying the corresponding station corrections.
276 Y. Font et al.

Figure 7. Same as Fig. 5 except that the 3-D synthetic P arrivals are used within a 1-D velocity model (from the RENSIG). The circle colour-code indicates the
absolute depth difference between the initial synthetic event and the solution. The synthetic investigation suggests that a 1-D velocity model can only resolve
15 per cent of earthquake location offshore Ecuador.
To obtain a high quality data set, we select earthquakes based perpendicular to the trench, 15 km landward from the trench to
on MAXI confidence factors: (1) Qedt ≥ 0.6 (2) interbarycenter 20 km inland from the coast line (Figs 8 and 9a). This seismic pat-
distance (between volumes V1 and V2; see section 4.1) <15 km tern is known as the Galera alignment (Segovia 2001). Hypocen-
(3) distance to first station <200 km and (4) number of V2 solu- ters mainly occur from about 10 to 30 km depth, along a plane
tion nodes <1000 (see section on MAXI technique). After quality dipping gently eastward, most probably the interplate seismogenic
selection, the Qedt average is 0.94 (±0.07), the average residual zone (Fig. 8—section 1). Seismicity also scatters within the over-
rms is 0.21 (±0.13) s, the average number of stations is 7.7 (±2.0, riding crust and, in less proportion, within the subducting plate.
and the average location uncertainties dx, dy, dz are 1.5 (±2.1 km), Southward, near 0.25◦ S, a cluster of earthquakes occurs at about
1.2 (±1.9 km), 1.4 (±1.8 km), respectively. Finally, our catalogue 10 km landward from the trench, extending over about 35 km of
is composed of 7950 earthquakes with a quality considered as high distance. At depth, this cluster (further on referred to the Jama
(Figs 8 (zoom) and 9a). cluster) essentially affects the overriding plate down to the pre-
sumed plate interface (dipping 10◦ towards the east). The interplate
activity is bounded between 8 and 14 km depth. Locally, a small
6.2 Results amount of seismic activity takes place deeper within the bending
Because hypocenter location uncertainties are generally underes- slab (beneath the trench) between 30 and 80 km depth (Fig. 8—
timated (Billings et al. 1994; Shearer 1997), earthquake location section 2).
quality can be assessed from seismicity clustering or alignment To the south, centred on 1.25◦ S, seismicity recurrently strikes
along well-known tectonic features. We thus will succinctly describe the margin through large seismic swarms. This cluster (further on
the seismicity observed at the vicinity of the interplate seismogenic referred as the Manta cluster after Vaca 2007) extends along 90 km,
zone and compare them to 1-D hypocenter determinations. in a direction parallel to the trench and 15–20 km landward from
To the north, near 0.5◦ N (near the town of Muisne; Fig. 1), a it. At depth, seismicity concentrates in the overriding plate down to
60-km-long alignment of earthquakes narrowly spreads ∼NW–SE, the plate interface (Fig. 8—section 3).
Table 3. Station corrections, in seconds, computed at 1-D model used by Instituto Geofı́sico was constructed based on the
each station of the selected subnetwork from the 3-D Andean geology and is used indistinctly in all the country (Table 2).
approach statistical investigations (see text for more de- Based on the MAXI-1-D results, the Galera alignment is barely
tails). Station location is shown on Fig. 2. Bold font recognizable. At this latitude, the Galera seismicity seems to start
highlights stations ecen and carm, qualified as EDT-out
about 30 km landward compared to MAXI-3-D results (Figs 9 and
(98 and 90 per cent, respectively).
10). The depth distribution shows two distinct clusters. The deeper
A D S M O one is extending from 10 to 50 km depth, the shallower strikes in
anti 0.19 0.06 0.16 −0.04 −0.11 part in the air and sea water and affects essentially the overriding
ara2 0.07 0.11 0.33 0.00 −0.06 margin. The Jama cluster epicentral position scatters east and west
calv −0.11 0.02 0.00 0.00 0.00 from the trench. Depth determinations range between –5 km (west
cars −0.01 0.04 0.00 0.07 −0.02 of the trench) to 40 km depth (at 80 km landward from the trench),
caya 0.05 −0.03 −0.14 −0.06 −0.06 affecting essentially the crust and mantle of the subducting plate.
chi1 −0.16 −0.01 −0.26 0.16 0.20 The Manta cluster is also scattered and separates in two groups
cone −0.28 0.08 −0.11 −0.19 −0.07 with a distance of ∼20 km. The northern one extends from 20 to
cota −0.07 0.04 0.10 0.09 0.15
30 km west of the trench to near the coastline, occurring to a large
ecen 7.23 7.09 6.64 7.23 7.67
degree within the water column. The southern one is located at less
−0.14

ggp 0.17 0.09 0.14 0.27
igua 0.19 0.18 0.04 0.13 0.22 than 10 km from the trench and spread to the coastline. Seismicity
mary −0.09 0.06 −0.01 −0.14 −0.01 deepens with sea water depth, near and around the seismogenic
milv 0.18 −0.04 −0.16 −0.08 −0.18 plate interface, down to about 30 km depth.
mush 0.34 0.00 0.72 0.00 0.00 To assess the effect of the location technique, we evaluate MAXI-
palm −0.24 −0.17 −0.09 0.00 −0.25 1-D against RENSIG determinations. Note however, that this latter
palo 0.10 0.03 0.12 −0.04 0.00 comparison is not totally appropriate because the RENSIG uses P
pisa −0.03 −0.18 0.15 −0.22 −0.24 and S-wave arrival times on the whole network. RENSIG determi-
qil1 0.01 −0.02 0.12 0.19 0.21 nations are performed using the 1-D-layered model described above
retu 0.07 0.17 0.05 0.14 0.03
and the FASTHYPO technique (Herrmann 1979) which minimizes
soch −0.10 0.04 −0.19 −0.07 −0.03
traveltimes residuals similar to other location programs. This algo-
tamb −0.17 0.07 −0.09 0.08 −0.19
tena −0.17 0.05 −0.05 −0.04 0.03 rithm was modified locally to consider stations delays because of
vc1 −0.22 −0.10 −0.12 −0.16 −0.15 high altitude variations of some seismic stations.
carm 0.74 1.27 0.00 1.86 1.81 From the RENSIG determinations, the Galera alignment is
chis 0.00 0.02 0.00 −0.08 0.16 clearly visible. It initiates 10 km before trench and extends to the
coji 0.00 0.00 0.00 0.00 0.03 coastline. Seismicity occurs at 30 km depth beneath the trench and
cupa −0.14 −0.15 −0.19 −0.06 −0.04 continuously affects the crust and mantle of the subducting plate
hoja −0.11 0.02 −0.14 0.08 0.06 (between 20 and 45 km depth). Very little activity appears at the
jama 0.12 0.05 0.10 0.10 0.13 interplate seismogenic zone or in the overriding plate. The Jama
lore 0.01 −0.13 −0.54 −0.07 −0.08
and Manta clusters are both relatively well clustered and occur on
mag1 −0.11 −0.18 0.03 0.15 0.03
both sides of the trench. Both cluster seismicity delineates, at depth,
magd −0.15 −0.04 0.14 0.00 −0.22
mnta 0.04 −0.02 0.19 −0.01 −0.05 a ∼20 km-thick plane dipping eastwards with an angle of about 20-
qit2 −0.15 −0.09 0.00 −0.14 0.10 to-25◦ (when starting from the trench). The activity stops at 35 km
qiti 0.08 −0.06 −0.29 −0.11 0.07 depth, for the Jama cluster, and near 30 km depth for the Manta
sali −0.28 −0.11 −1.41 0.11 0.06 cluster.
gual 0.02 −0.08 0.18 0.02 −0.01
hier −0.10 0.07 −0.11 0.21 0.81
puca −0.16 0.20 0.06 −0.07 −0.08 6.4 Effect of the a priori 3-D velocity model
The velocity model errors have been found to have a larger effect on
The Jama and Manta clusters and their depth distribution near the location procedure than the effect of picking errors (e.g. Gomberg
expected interplate seismogenic zone are consistent with the seis- et al. 1990; Billings et al. 1994; Husen et al. 2003; Bai & Green-
mic activity known in other subduction zones. In addition, given halgh 2006). As the real heterogeneity of the crust and mantle is
that each earthquake is located independently, the tightness of the rarely well known, this conclusion mainly results from synthetic
Galera alignment tends to support the notion that in addition to investigations and the common knowledge that a 1-D model cannot
reliable confidence factors and uncertainty parameters hypocenter account for complicated velocity structure. Indeed, in real appli-
determinations resulting from the proposed 3-D approach are rea- cations, at local to regional scale and in well-instrumented region,
sonably accurate. 3-DVM obtained from tomographic inversion of local earthquakes
(starting from a minimum 1-D velocity model and associated 1-
D-hypocenter determination) only produces a few differences in
6.3 Comparison to 1-D-determinations
earthquake location, with no large systematic shift but an clustering
To assess the effect of the 3-DVM in earthquake location (real of seismicity and slight depth variations (< 3 km, in average, given
data application), we perform hypocenter determinations using a by Micheal 1988; Husen & Smith 2004). On land, a probabilistic
procedure identical to what is presented above (same P-wave arrival earthquake location in Switzerland and the integration of a priori
data set and MAXI technique) but implemented within the same knowledge in the 3-DVM generates large individual shifts but there
1-D velocity model than the one used by the Instituto Geofisico. were not systematic (Husen et al. 2003). In this study, the velocity
We name MAXI-1-D the resulting hypocenter determinations. The model is totally constrained from a priori knowledge. Its adequacy
MAXI-1-D catalogue includes the same events as MAXI-3-D. The to represent locally the earth structure can be appraised from formal
278 Y. Font et al.

Figure 8. Seismic hypocenter location map and cross-sections in the vicinity of the interplate seismogenic zone: results from the 3-D approach and P-wave
arrivals. Section 1 refers to the Galera alignment, 2 and 3 to Jama and Manta clusters, respectively. Section locations and width appear as brackets on the map.
Bold line with triangles is the trench.
earthquake location uncertainties, traveltime residual statistics per underestimated and because uncertainties depend on the velocity
seismic station (Table 3) and the geological reasonableness of the model in which the solution is produced (Sambridge & Kennett
seismicity distribution. 1986; Billings et al. 1994; Shearer 1997). In addition to the high
The effect of the 3-DVM on earthquake location can be assessed values of the confidence indicators and small uncertainties, the
by comparing MAXI-3-D and MAXI-1-D results (application to MAXI-3-D seismicity appears to be reliable because it is consistent
real data; Figs 9 and 10). The distances between 1-D and 3-D- with the plate interface location in the vicinity of where one ex-
resolved hypocenters are in average 36 ± 23 km (dx = 7 ± 35 km, pects seismogenic deformation and because its clustering in space
dy = 7 ± 16 km and dz = 4 ± 15). As the only difference between conveys the image of seismic activity on a portion of a fault. As a
both catalogues is the velocity model in which ray paths propagate, first approximation, we therefore consider that the 3-DVM estab-
the differences in earthquake position directly illustrate the effect lished to represent velocity variations in the area is trustworthy and
of the differences in traveltime estimates between 3-D and 1-D advantageous for earthquake location.
models (see Table N2-old3, for 1-D model). First, we observe that In this study, we have constructed the VP 3-DVM with care.
this effect is not systematic (Figs 9 and 10). Relatively to MAXI-3- Nevertheless, initial data on crust and mantle velocity properties
D, the earthquake position of the Galera alignment tends to move do not uniformly cover the studied area and some necessary in-
land ward while a good proportion of Jama and Manta earthquakes formation is uncertain or missing (for instance, the forearc crustal
are moving seaward. The depth parameter also undergoes a non- thickness lacks recent constraints; magmatic chambers are not taken
systematic shift, deeper along the Muisne alignment and shallower into consideration). Therefore, the 3-DVM most certainly only ap-
for the western part of Jama and Manta clusters. Second, even proximates to first order the large-scale velocity variations. If non-
though the 3-D approach handles single-event locations, epicenters existent local heterogeneities might be introduced by mistake in
appear to better cluster. Third, even if the real hypocenter position the 3-DVM, we assume they are insignificant compared to large
is not known, the MAXI-3-D evaluation of depth (more consistent scales anomalies that a 1-D layered model would inexorably gen-
with the plate interface) also appears more realistic than MAXI-1-D erate when trying to represent a non-flat configuration of crust and
(Fig. 10). Consequently, despite that the quality of the 3-DVM is mantle such as in subduction zones. In such areas, 1-D-layered ve-
not directly evaluated as it would be in a tomographic inversion, its locity model obviously degrades traveltime computations and thus
implementation produces an image of seismicity that seems more earthquake location accuracy. Consequently, in subduction zone
reliable. seismological studies, exploration of the concept of a priori 3-D
In seismology, assessing the reliability of hypocenter determi- velocity models and questioning the classical 1-D-approach seem
nations is challenging because location uncertainties are generally fully appropriate.
Figure 9. Seismicity map comparing earthquake location resulting from

three different procedures. 3-D-MAXI refers to the catalogue resulting from Figure 10. Seismicity cross-sections comparing earthquake location result-
earthquake location using: a selection of RENSIG stations, P arrivals, 3- ing from three different procedures (location on Fig. 9). See Fig. 9 for
DVM, and the MAXI technique (see text for more details). 1-D-MAXI: explanations about 3-D-MAXI, 1-D-MAXI and RENSIG procedures. The
identical procedure within a 1-D-layered model (from the RENSIG). REN- grey dash line marks an eventual position of the Nazca Plate top and serves
SIG refers to the whole network, P- and S-waves arrivals, 1-D-layered model as spatial reference.
(see Table 3) and FASTHYPO (Herrmann 1979). The traces of the Carnegie
Ridge (dashed line), the trench (bold line and triangles) and the Andean
Chain (in grey) are marked for references. 1 and 2 identify the cross-section
shown on Fig. 10.
280 Y. Font et al.
A second question concerns the use of an unconstrained Mw = 8.2 1979 earthquake (Kanamori & McNally 1982; Mendoza
S-wave velocity model in earthquake location because, from what & Dewey 1984; Swenson & Beck 1996). The rupture zones of the
is said above, the accuracy of the velocity model mainly controls three smaller events (1942, 1958 and 1979), approximated from the
the quality of earthquake location. Indeed, a priori S-wave veloc- distribution of aftershocks, abut closely without significant overlap.
ity models are not available if tomographic inversions cannot be In the study area, the 1942 epicenter has been located near 0◦ S
performed. Furthermore, when tomographic investigations can be and 279.8◦ E (Kelleher 1972; Swenson & Beck 1996; Engdahl &
performed in subduction zones, it is observed that the VP /VS ratio Villaseñor 2002). Its rupture zone propagated from south to north
is far from constant in those systems because of the high fluid con- (Kelleher 1972; Mendoza & Dewey 1984, Fig. 11). The source–
tent and fracturing (e.g. DeShon et al. 2006). Variations of VP /VS time function of this event is characteristics of a single asperity
ratios are also in receiver function investigation (e.g. Audet et al. rupture (although its location is uncertain), and most of the seismic
2009). In addition, relatively to P arrivals, measuring S-wave arrival moment is assumed to be released within approximately 50 km of
times is complicated by the P-wave coda and conversions. Zeiler the epicenter (Swenson & Beck 1996). The original rupture zone
& Velasco (2009) reported that, at local to regional distance range extends westwards from the trench. In Fig. 11, we draw its contour
(0◦ –30◦ ), most of the picking errors were associated with the S and so that it fits the trench. The 1958 event occurred to the north,
Sg phases. By comparing common arrival times reported by several near 1◦ N and 280.5◦ E. It presumably ruptured unilaterally to the
institutions, they demonstrate that the average time difference for northeast abutting, but not overlapping, the 1979 rupture area. The

S waves reached 5.5 s (compared to 0.4 s for P waves). Zeiler & epicenter is determined at the southwestern edge of the aftershock
Velasco (2009) also indicate that short-period instruments are not as area (Mendoza & Dewey 1984; Swenson & Beck 1996).
precise when used to pick arrivals. In the specific case of Ecuador, In 1901 and 1907, two other events of magnitude 7.2 and 7.0
most stations of the permanent network are composed of vertical occurred near 0◦ N and 2◦ S, respectively (Askew & Algermissen
component short-period sensors, which are known to distort the S 1985; Engdahl & Villasenor 2002). Both epicenters are not well
arrival-time measurements. Subsequently, the S measurement pre- constrained and are reported, about 150 km west of the trench, near
cision should be considered as suspicious (compared to P arrivals), 278◦ E.
and most probably flawed in the specific case of Ecuador (mainly In the area near Bahia city (near 0.5◦ S; Fig. 1), three M > 7 events
because of the network characteristics). have occurred in 1896, 1956 and 1998. The last one is known as
Given those difficulties, one can wonder what would be the effect the Bahia earthquake (Segovia 2001). Associated aftershocks are
of a non-precise S-velocity model and inaccurate S arrivals in the reported in yellow triangles (Fig. 11).
process of earthquake location. The answer to this question certainly
depends on the relative quality between the S and P velocity model.
7.2 Forearc density, marine terraces and large earthquakes
Although beyond the scoop of this paper, synthetic investigations
should be carried out using heterogeneous models and arrival-time Lateral variations on shear strength (i.e. the resistance that a fault
difference technique to bring some answers on the effect of a biases plane opposes to yield and slip under the load of external forces)
S-velocity model on earthquake location. In this study investigation, are, at least partially, because of spatial changes of the vertical stress
because of the inferred large heterogeneities of crust and mantle, loading the megathrust fault, as controlled by the density structure
the unknown quantity of water in the mantle wedge, the geometry of the overlying forearc (Tassara 2010). Although no forearc density
and positions of magmatic chambers, the lack of information on model exists in the Ecuadorian Margin, we tentatively superimposes
fracturing, and the uncertain quality of S arrival times, we favour the Simple Bouguer Gravity Anomaly map produced by Feininger
the use of a unique P-velocity model. It is worth noting that, whereas & Seguin (1983) to the epicenter positions of the main earthquakes
most seismologists commonly use both P and S waves, some local known in the area (Fig. 11).
to regional studies provide interesting results using P waves only On the coastal area, the Simple Bouguer Gravity Anomaly field
(e.g. Husen et al. 2003; DeShon et al. 2007; Arroyo et al. 2009). is dominated by a broad belt of positive anomalies (Piñon forma-
tion) that extends from Colombia to south of Guayaquil. On about
two thirds of the coastal area, several areas of negative anomalies
interrupt the continuity of the positive belt and coincides, broadly,
7 S E I S M I C I T Y PAT T E R N W I T H I N T H E
with sedimentary basins (Feininger & Seguin 1983). Although the
INTER-SEISMIC PERIOD ALONG THE
location accuracy of major earthquake epicenters might be debat-
E C UA D O R I A N S U B D U C T I O N Z O N E
able, we note that the 1942, 1958 and 1906 earthquake nucleation
remarkably falls into small areas of positive Bouguer anomalies that
7.1 Historical earthquakes known along the northern
intrude near the coastline.
Ecuador subduction zone
Repeatedly coastal uplifts resulting in the formation of marine
In 1906, a Mw = 8.8 megathrust earthquake, accompanied by a terraces have been related to the vertical motion produced by the
large far-field and local tsunami, ruptured the subduction fault zone subduction of aseismic ridges (e.g. Hsu 1992) but also to the uplift
in North Ecuador South Colombia along a 500-km-long rupture that occurred during large magnitude earthquakes (e.g. Berryman
area (Kelleher 1972; Kanamori & McNally 1982). The epicenter of et al. 2011). Along the Ecuadorian coast, marine terraces have been
this event was first located seaward from the trench (Gutenber & observed in Rio Verde, Punta Galera, Manta Peninsula, Santa Elena
Richter 1959), but from constant (S–P) time at teleseismic stations Peninsula and Puna Island (Fig. 1; Winckell & Zebrowski 1997;
(Baltimore and Victoria, Kanamori & McNally 1982) and from the Pedoja 2003; Pedoja et al. 2006). In Rio Verde and Punta Galera,
intensity distribution described by Rudolph & Szirtes (1911), we sequences of two and three marine terraces are observed (respec-
prefer the epicenter position from Collot et al. (2004) near 0.8◦ N tively) covering an area of 2 to 5 km wide and 15 to 20 km long (in
and 280◦ E (Fig. 11). Much of the rupture zone that broke during both cases). Mean uplift rates (given since ∼120 ky) are evaluated
the 1906 event ruptured again, with thrust motion, during the Mw = at less than 0.1 mm yr–1 in Rio Verde (not well constrained) and
7.9 earthquake of 1942, the Mw = 7.7 1958 earthquake and the about 0.3 (±0.02) mm yr–1 in Punta Galera (Pedoja 2003; Pedoja

Figure 11. Left: Map showing the historical earthquake epicenters (each symbol corresponds to the bibliographic source given below the main frame, the thin
dashed line represent different solutions), their asperity (dark green) and rupture zone (light green; Kanamori & McNally 1982; Beck & Ruff 1984; Swenson &
Beck 1996). The dashed grey line represent the rupture zone of the Mw = 8.8 1906 event (Kelleher 1972). Positive (red) and Negative (yellow) Simple Bouguer
anomalies are taken from Feininger & Seguin 1983. Marine terraces mean uplift rate (white arrows) are from Pedoja (2003) and Pedoja et al. (2006). The trace
of the Carnegie Ridge (dark to light grey) and the Trench (bold line with triangles) are also shown. Right: 3-D-MAXI catalogue (circles) selected from 0 to
30 km depth. Yellow triangles correspond to the aftershock sequence of the Bahia event (Segovia 2001). Blue lines are fault traces from Collot et al. 2004.
et al. 2006). Northeast of Rio Verde, the coastal area is described cumulates along the interplate seismogenic fault zone, the coseis-
as subsiding (Dumont et al. 2006). On the Manta Peninsula and mic phase that suddenly releases the accumulated stress through
La Plata Island, up to five marine terraces are observed and uplift large earthquake and, the post-seismic phase that corresponds to
rates have been estimated at 0.3 (± 0.02) mm and 0.4 (± 0.02) mm, the total stress relaxation along the plate interface. In this study,
respectively (Cantalamessa & Di Celma 2004; Pedoja et al. 2006). the 3-D-MAXI catalogue most certainly samples a part of the inter-
To the south, on Santa Elena Peninsula, a sequence of three ma- seismic period (Manchuel et al. 2011). The seismicity distribution
rine terraces is observed and the mean uplift rate is lower than 0.1 should correlate with surface rupture and asperity related to the
(±0.02) mm yr–1 . Pedoja et al. (2006) concluded that the uplift of stress built-up along the plate interface (Mogi 1979; Scholz 2002;
the Ecuador (and Peruvian) Talara Arc is reduced when compared Bollinger et al. 2004). We hereby intend to correlate the seismicity
to other regions where aseismic ridges are subducting. Although distribution (magnitude lower than 7) to the areas where stress might
the location accuracy of major earthquake epicenters might be de- be building on the interplate seismogenic fault zone. The rupture
batable, we note that the position of the 1958 and 1906 earthquake zone of the 1979 event is too far away from our catalogue sensitivity
nucleation remarkably falls in the areas where uplift indicators are and cannot be discussed here from our results.
well preserved in the coastal morphology (Fig. 11). These areas
coincide with the region of Simple Bouguer positive anomalies
7.3.1 1958 event rupture zone
(Feininger & Seguin 1983).
The scarcity of events within the known rupture zone of the Mw =
7.8 1958 event is genuine from the 3-D-MAXI catalogue (Fig. 11).
7.3 Seismicity and segmentation
This observation is coherent with the results of Manchuel et al.
Along subduction zones, the seismic cycle consists of three main (2011) on 3 months of temporary onshore–offshore seismic deploy-
stages (Reid 1910): the interseismic period during which stress ac- ment in the area, and with teleseismic observations that also image a
282 Y. Font et al.
lack of seismicity (EHB catalogue from 1964–2007, Engdahl et al. counter argument of the hypothesis is that no marine terraces are
1998). In this area, from marine geophysical investigation, Collot observed on the coast. One can nevertheless argue that this could be
et al. (2004) suggested that the Esmeraldas and Manglares fault because the coastal morphology is not well preserved in this region
(Fig. 11) could have delimited the lateral extension of the 1958 (or because onland observations have not been done yet).
rupture area (slightly displacing northwards the original rupture To the north, we can not apply the same reasoning. Indeed, if
area from Swenson & Beck 1996). In between those two faults, an we believe that the Galera alignment is bounding two segments of
outer basement high, which bounds the margin seaward of the 1958 the Ecuadorian subduction zone, its shape (i.e. a line rather than a
rupture zone, may have acted as a deformable buttress to seaward circle) is not coherent with a ‘classical’ asperity geometry.
propagation of coseismic slip. The Esmeraldas fault, such as men-
tioned by Collot et al. (2004) is not confirmed from the 3-D-MAXI
catalogue. However, a seismic alignment, parallel to the Esmeral- 7.3.3 1998 Bahia event zone
das fault and about 20 km south of it, has been evidenced from
recent investigation on microseismicity (Regnier et al. 2011; ADN The area from 0.25◦ S to ∼0.75◦ S (between Cabo Pasado and Manta;
program). From the 3-D-MAXI catalogue, we observe that the seis- Figs 1 and 11) differs widely from the 1942 rupture area. Three large
mic behavior of the interplate seismogenic zone changes drastically earthquakes occurred near Bahia during historical and instrumental
south of the Galera alignment, suggesting that this alignment also times: the Ml = 7.0 1896 event, the Ms = 7.3 1956 event (both from

reflects a limit of segment. Whatever is the southern limit of the SISRA catalogue and; Askew & Algermissen 1985) and the Mw =
1958 rupture area (from Esmeraldas fault to Galera alignment), it 7.2 1998 Bahia event (EHB, Engdahl et al. 1998). From the 3-D-
seems that the seismogenic interplate zone is further segmented MAXI catalogue, this zone appears almost aseismic on a spherical
in this region which could explain the two pulses of the source– domain of about 30 km of radius. Because the 3-D-MAXI catalogue
time function determined from deconvolution of long-period only samples high quality location events, we report on the Fig. 11
P-wave records (Swenson & Beck 1996). The lack of seismicity the aftershocks relocated by Segovia (2001, hypo71 earthquake
north of the Galera alignment might reveal that the contact zone location: class A events correspond to good quality location, with
is, at least, partially locked and stress is accumulating. A partially horizontal errors lower than 7 km, vertical error lower than 6 km
locked zone is consistent with the knowledge that a large historical and residuals lower than 0.7 s; class B events correspond to bigger
earthquake occurred in the area, the coincidence of positive Bouguer errors). In this area, although some earthquakes might be repeated
anomalies and coastal uplifts, but also with GPS campaign measure- in Fig. 11, earthquake density is still low.
ments (Trenkamp et al. 2002). Nocquet et al. (2009) mentioned that The concept of a seismic gap refers to areas of no (or little)
the average coupling coefficient might reach up 40 per cent in this recent rupture displacement. Seismic gaps are related to asperity
area (near Esmeraldas). regions that are expected to experience a large displacement during
the future rupture. Although the relation between microseismicity
during the inter-seismic period and seismic gaps (able to generate
7.3.2 1942 event rupture zone the future asperity) is not well known, we propose that the Jama
The seismicity behaviour changes between the Galera alignment cluster (∼0.25◦ S) is revealing the location of a possible asperity on
(Fig. 8—section 1) and the Jama cluster (Fig. 8—section 2). The the interplate seismogenic zone. The adjacent Bahia zone should
Jama cluster bounds both the southern limit of the 1942 rupture zone undergo relatively more aseismic slip. The fact that the area is
and its asperity. In this area, the coastline is deflected landward, the facing a margin with negative Bouguer anomaly, that the coastline
margin presents a positive Bouguer anomaly, and the seismicity is is deflected landward and that no marine terraces are observed
occurring deeper on the seismogenic zone down to 30 km depth. onland are in favour of this interpretation.
These observations converge towards the interpretation of the seis-
mogenic zone downdip limit near 30 km in depth.
7.3.4 The Manta–Puerto Lopez area
Lay & Kanamori (1981) suggested that a small earthquake repre-
sents the failure of one asperity and that its rupture zone is limited, at South of Manta (from 0.75◦ S to 1.5◦ S), microseismicity pattern
both ends, by adjacent asperities (in the sense of a barrier). Effective widely differs again. The seismicity in this area is mainly organized
width and amount of slip are relatively small and the source–time in earthquake swarms such as in 1998, 2002, 2005 (Segovia 2009;
function is simple. A great earthquake represents the failure of Vaca et al. 2009). Detailed analyses on the 2005 swarm (wave-
more than one asperity involving much larger width and slip and form inversion and waveform correlations) permits identification
generating more complex waveforms. Kanamori & McNally (1982) of multiple triggering sources between Manta and Puerto Lopez.
suggested that the 1942, 1958 and 1979 might have behaved as small Facing this area onland, the Jipijapa-Portoviejo crustal fault within
earthquakes and that the large 1906 earthquake would have involved the North Andean Block have been found seismically active and
much larger width and slip by failing the asperities of 1942, 1958 connecting to the plate interface (Bethoux et al. 2011, from local
and 1979. In that context, the 1942 earthquake failed an asperity onshore–offshore deployment and Segovia (personal communica-
(drawn on Fig. 11), the rupture propagated along the interplate con- tion, 2011) from OSISEQ project, local network recently deployed),
tact zone and have been limited by adjacent asperities. In the light therefore contributing in some extend to the subduction system de-
of the seismicity patterns, the Jama cluster (facing the Cabo Pasado formation.
cap) could be the site of an asperity (defined here as a portion of the The swarm activity affects a region where the coast is in part
seismogenic interplate fault sufficiently coupled to block the rupture deflected seaward (Manta Peninsula), where the Bouguer anoma-
propagation). If so, the seismicity there would reflect partial strain lies are greatly positive and where marine terraces are still well
release, the other part of the strain budget would be accumulated on preserved on the coastal morphology. Nevertheless, from historical
a (partially) locked plate interface. If the coastline somehow reflects records, no major earthquakes have affected this segment (nor from
the behaviour of the seismogenic interplate zone, it is noteworthy there to northern Peru). Even though a slow slip event in the area
that the coastline is locally deflected towards the Jama cluster. A (2010), recently described by Vallée et al. (2013), contribute to the
budget of the total energy release by aseismic displacement, we do well as all the staff of the Instituto Geofisico that have installed and
not reject the probability of a large event in this area. maintained the network and manually measured arrival times. We
sincerely thank the anonymous reviewer, the editor, M. Cocco and
P. Mothes, for their comments that greatly contribute to improve the
7.3.5 South of Puerto Lopez area manuscript.
The southern segment of the Ecuadorian subduction sampled by
the 3-D-MAXI catalogue indicates a wide aseismic segment from FUNDING
1.5◦ S to south of the Gulf of Guayaquil (∼2.5◦ S). A earthquake
of magnitude greater than 7 occurred in the area: the Mw = 7.2 This study was funded by the Institut de Recherche pour le
earthquake in 1901 was located around 2◦ S, about 100 km west of Développement, Geoazur and the ANR blanc 2007 (program ADN,
the trench (Fig. 11) and the Mw = 7.5, near 3.5◦ S and 279◦ E, in Andes du Nord). This work is a contribution from the Laboratoire
1953 (EHB and SISRA catalogues). Mixte International Seismes & Volcans dans les Andes du Nord
supported by the IRD and the IGEPN.
8 C O N C LU S I O N
REFERENCES

In this investigation, we present a 3-D a priori velocity model
constructed to improve the seismicity distribution in the subduction Agudelo, W., Ribodetti, A., Collot, J.Y. & Operto, S., 2009. Joint inver-
sion of multichannel seismic reflection and wide-angle seismic data; im-
zone of Ecuador. Synthetic experiments evaluate the feasibility of
proved imaging and refined velocity model of the crustal structure of
this earthquake location study based on the permanent network of
the north Ecuador-south Colombia convergent margin, J. geophys. Res.,
the RENSIG. 114(B2), doi:10.1029/2008JB005690.
The resulting 3-D-MAXI catalogue upgrades the image of the Arroyo, I.G., Husen, S., Flueh, E.R., Gossler, J., Kissling, E. & Alvarado,
seismicity distribution in the inter-seismic period that we intent to G.E., 2009. Three-dimensional P-wave velocity structure on the shallow
interpret in light of previous knowledge of the past earthquakes that part of the central Costa Rican Pacific margin from local earthquake
occurred in the area. As previously mentioned, the subduction zone tomography using off- and onshore networks, Geophys. J. Int., 179(2),
of Ecuador is highly segmented. We interpret two areas (the 1958 827–849.
rupture zone area and the Jama cluster) as eventual asperities that Askew, B.L. & Algermissen, S.T., 1985. Catalog of Earthquakes for South
might be accumulating stress, even though both areas show different America: Hypocenter and Intensity Data, Volumes 4, 6, 7a, 7b, and 7c,
Centro Regional de Sismologia para America del Sur (CERESIS), Lima,
seismicity patterns and that the coupling might be much lower on
Peru.
the Jama cluster.
Aspden, J.A., McCourt, W.J. & Brook, M., 1987. Geometrical control of
A third area, from Cabo Pasado to Manta appears almost aseismic subduction-related magmatism; the Mesozoic and Cenozoic plutonic his-
on a spherical domain of about 30 km of radius. We suspect that the tory of western Colombia, J. geol. Soc. Lond., 144(6), 893–905.
50 yr recurrence period of M > 7 earthquakes (1896, 1956 and 1998 Aspden, J.A., Fortey, N., Litherland, M., Viteri, F. & Harrison, S.M., 1992.
events) occurs on a small local asperity and that the rest of the area Regional S-type granites in the Ecuadorian Andes; possible remnants of
behaves mostly aseismically by steady creep or slow slip events. the breakup of western Gondwana, J. South Am. Earth Sci., 6(3), 123–
A fourth area (between Manta and Puerto Lopez) is more am- 132.
biguous because no historical earthquakes are known in the area and Audet, P., Bostock, M.G., Christensen, N.I. & Peackock, S., 2009. Seismic
because slow-slip events have been observed together with seismic evidence for overpressurde subducted oceanic crust and megathrust fault
sealing, Nature (London), 457, 76–78.
swarms (Vallée et al. submitted). The observation of well preserved
Bai, C.-Y. & Greenhalgh, S., 2006. 3D local earthquake hypocenter deter-
marine terraces on the coastal morphology, a coastline that is de-
mination with an irregular shortest-path method, Bull. seism. Soc. Am.,
flected seaward at the Manta and the Puerto Lopez Peninsula and 96(6), 2257–2268.
the positive Bouguer anomaly let us infer that the seismic coupling Beck, S.L. & Ruff, L.J., 1984. The rupture process of the great 1979 Colom-
might be only partial, with stress accumulating on the seismogenic bia earthquake: evidence from the asperity model, J. geophys. Res., 17,
interplate zone. 1969–1972.
The seismic segmentation in Ecuador is often regarded as the Berryman, K., Ota, Y., Miyauchi, T., Hull, A., Clark, K., Ishibashi, K.,
result of the Carnegie Ridge subduction. As the ridge subducts be- Iso, N. & Litchfield, N., 2011. Holocene Paleoseismic History of Upper-
neath the South America Plate between 1◦ N and 2◦ S, it is difficult Plate Faults in the Southern Hikurangi Subduction Margin, New Zealand,
to distinguish the precise causes of such detailed segmentation. Ma- Deduced from Marine Terrace Records, Bull. seism. Soc. Am., 101(5),
2064–2087, doi:10.1785/0120100282.
rine geophysical data on one hand and GPS models (from ADN and
Bethoux, N., Segovia, M., Alvarez, V., Collot, J.-Y., Charvis, P., Gailler,
IG-EPN projects) in another hand should help to further understand
A. & Monfret, T., 2011. Seismological study of the central Ecuadorian
the seismic pattern of the region. margin: evidence of upper plate deformation, J. South Am. Earth Sci.,
31(1), 139–152.
Bhattacharyya, J., Flanagan, M.P., Schultz, C.A., Pasyanos, M.E. & Walter,
AC K N OW L E D G M E N T S W.R., 2000. Building and validating a regional model of the Former Soviet
For their contributions to feed the velocity model, we thank A. Cala- Union, Seismological Research Letters, 71(2), 208.
Billings, S.D., Sambridge, M.S. & Kennett, B.L.N., 1994. Errors in hypocen-
horrano, B. Marcaillou, W. Agudelo, L. Garcia, J.Y Collot. We are
ter location; picking, model, and magnitude dependence, Bull. seism. Soc.
grateful to J. Trevisan for her technical help in the computational
Am., 84(6), 1978–1990.
aspects of the construction of the velocity model. G. Vetois and the Bollinger, L., Avouac, J.P., Cattin, R. & Pandey, M.R., 2004. Stress buildup
EGEODE virtual organization are thanks for the access to EGEE in the Himalaya, J. geophys. Res, 109, doi:10.1029/2003JB002911.
computer grid and their computing support. N. Béthoux, T. Monfret, Bondar, I., Myers, S.C., Engdahl, E.R. & Bergman, E.A., 2004. Epicentre
G. Nolet and M. Régnier for the discussion about the MAXI tech- accuracy based on seismic network criteria, Geophys. J. Int., 156(3), 483–
nique. H. Yepes, J. Aguilar are thanked for their help and support, as 496.
284 Y. Font et al.
Calahorrano, A., 2001. Carnegie Ridge subduction under the Ecuadorian Font, Y., Kao, H., Lallemand, S., Liu, C.-S. & Chiao, L.-Y., 2004. Hypocentre
margin: crustal modelisation from wide-angle data (in French), Master determination offshore of eastern Taiwan using the maximum intersection
thesis, 37 pp, Université de Pierre et Marie Curie, Villefranche-sur-Mer. method, Geophys. J. Int., 158(2), 655–675.
Calahorrano, A., Sallares, V., Collot, J.-Y., Sage, F. & Ranero, C.R., 2008. Garcia, L., 2009. 3D seismic imaging of the subduction zone at the
Nonlinear variations of the physical properties along the southern Ecuador Colombia-Ecuadorian frontier (in French), Université de Nice-Sophia
subduction channel; results from depth-migrated seismic data, Earth Antipolis, Villefranche-sur-Mer, 214 pp.
planet. Sci. Lett., 267(3–4), 453–467. Gomberg, J.S., Shedlock, K.M. & Roecker, S.W., 1990. The effect of S-wave
Cantalamessa, G. & Di Celma, C., 2004. Origin and chronology of Pleis- arrival times on the accuracy of hypocenter estimation, Bull. seism. Soc.
tocene marine terraces of Isla de la Plata and of flat, gently dipping Am., 80(6), 1605–1628.
surfaces of the southern coast of Cabo San Lorenzo (Manabi, Ecuador), Graindorge, D., Calahorrano, A., Charvis, P., Collot, J.-Y. & Bethoux, N.,
J. South Am. Earth Sci., 16, 633–648. 2004. Deep structures of the Ecuador convergent margin and the Carnegie
Case, J.E., Barnes, J., Paris, G., Gonzalez, H. & Vina, A., 1973. Trans- Ridge, possible consequence on great earthquakes recurrence interval,
Andean Geophysical Profile, Southern Colombia, Bull. geol. Soc. Am., Geophys. Res. Lett., 31(4), doi:10.1029/2003GL018803.
84(9), 2895–2903. Guillier, B., Chatelain, J.L., Jaillard, E., Yepes, H., Poupinet, G. & Fels,
Cediel, F., Shaw, R.P. & Caceres, C., 2003. Tectonic assembly of the northern J.F., 2001. Seismological evidence on the geometry of the orogenic sys-
Andean Block, AAPG Mem., 79, 815–848. tem in central-northern Ecuador (South America), Geophysical Research
Christophoul, F., Baby, P. & Davila, C., 2002. Stratigraphic responses to a Letters, 28(19), 3749–3752.

major tectonic event in a foreland basin; the Ecuadorian Oriente Basin Gutenber, B. & Richter, C.F., 1959. Seismicity of the Earth, Princeton Uni-
from Eocene to Oligocene times, Tectonophysics, 345(1–4), 281–298. versity Press, Princeton, New Jersey, 310 pp.
Collot, J.-Y. et al., 2004. Are rupture zone limits of great subduction earth- Gutscher, M.-A., Malavieille, J., Lallemand, S. & Collot, J.-Y., 1999. Tec-
quakes controlled by upper plate structures? Evidence from multichannel tonic segmentation of the North Andean margin: impact of the Carnegie
seismic reflection data acquired across the northern Ecuador; southwest Ridge collision, Earth planet. Sci. Lett., 168, 255–270.
Colombia margin, J. geophys. Res., 109(B11103), 14. Herrmann, R.B., 1979. FASTHYPO; a hypocenter location program, Earthq.
Collot, J.Y., Charvis, P., Gutscher, M.A. & Operto, S., 2002. Exploring the Notes, 50(2), 25–37.
Ecuador-Columbia active margin and interplate seismogenic zone, EOS, Heuret, A., Lallemand, S., Funiciello, F., Piromallo, C. & Faccenna,
Trans. Am. geophys. Un., 83(17), 189–190. C., 2011. Physical characteristics of subduction interface type seis-
Collot, J.Y., Agudelo, W., Ribodetti, A. & Marcaillou, B., 2008. Origin of mogenic zones revisited, Geochem. Geophys. Geosyst., 12(Q01004),
a crustal splay fault and its relation to the seismogenic zone and under- doi:10.1029/2010GC003230.
plating at the erosional north Ecuador-south Colombia oceanic margin, J. Hey, R., 1977. Tectonic evolution of the Cocos-Nazca spreading center,
geophys. Res., 113(B12), doi:10.1029/2008JB005691. Bull. geol. Soc. Am., 88(10), 1404–1420.
Daly, M.C., 1989. Correlations between Nazca/Farallon plate kinematics Hsu, J.T.-J., 1992. Quaternary uplift of the Peruvian coast related to the
and forearc basin evolution in Ecuador, Tectonics, 8(4), 769–790. subduction of the Nazca Ridge: 13.5 to 15.6 degrees south latitude, Quat.
Deniaud, Y., 2000. Sedimentary and structural records of the Ecuadorian Int., 15/16, 87–97.
Andes geodynamic evolution during Neogen Time: Forearc basin study Hughes, R.A. & Pilatasig, L.F., 2002. Cretaceous and Tertiary terrane accre-
and mass balance (in French), Ph.D. Thesis, Université Joseph Fourier, tion in the Cordillera Occidental of the Andes of Ecuador, Tectonophysics,
Grenoble, France, 157 pp. 345(1–4), 29–48.
DeShon, H., Thurber, C.H. & Rowe, C., 2007. High-precision earth- Husen, S. & Smith, R.B., 2004. Probabilistic earthquake resolution in three-
quake location and three-dimensional P wave velocity determina- dimensional velocity models for the Yellowstone National Park region,
tion at Redoubt Volcano, Alaska, J. geophys. Res., 112(B07312), Wyoming, Bull. seism. Soc. Am., 94(3), 880–896.
doi:10.1029/2006JB004751. Husen, S., Kissling, E., Deichmann, N., Wiemer, S., Giardini, D. & Baer,
DeShon, H., Rowe, C. & Thurber, C., 2006. Seismogenic zone structure M., 2003. Probabilistic earthquake location in complex three-dimensional
beneath the Nicoya Peninsula, Costa Rica, from three-dimensional local velocity models; application to Switzerland, J. geophys. Res., 108(B2),
earthquake P- and S-wave tomography, Geophys. J. Int., 164(1), 109– doi:10.1029/2002JB001778.
124. Jaillard, E., Soler, P., Carlier, G. & Mourier, T., 1990. Geodynamic evolution
Dumont, J.F., Santana, E., Valdez, F., Tihay, J.P., Usselmann, P., Iturralde, of the Northern and Central Andes during early to middle Mesozoic times;
D. & Navarette, E., 2006. Fan beheading and drainage diversion as evi- a Tethyan model, J. geol. Soc. Lond., 147(6), 1009–1022.
dence of a 3200–2800 BP earthquake event in the Esmeraldas-Tumaco Johnson, M. & Vincent, C., 2002. Development and testing of a 3D velocity
seismic zone: a case study for the effects of great subduction earthquakes, model for improved event location; a case study for the India-Pakistan
Geomorphology, 74, 100–123. région. Bull. seis. Soc. Am., 98(8), 2893–2910.
Engdahl, E.R., Van Der Hilst, R.D. & Buland, R., 1998. Global teleseismic Kanamori, H. & McNally, K.C., 1982. Variable rupture mode of the sub-
earthquake relocation with improved travel times and procedures for depth duction zone along the Ecuador-Colombia coast, Bull. seism. Soc. Am.,
relocation, Bull. seism. Soc. Am., 88, 722–743. 72(4), 1241–1253.
Engdahl, E.R. & Villaseñor, A., 2002. Global seismicity: 1900–1999, Intl. Kelleher, J., 1972. Rupture Zones of Large South American Earthquakes
Hb of Earthquake and Engineering Seismology, 81A, 665–689. and Some predictions, J. geophys. Res., 77(11), 2087–2097.
Feininger, T. & Seguin, M.K., 1983. Simple Bouguer gravity anomaly field Kennett, B.L.N. & Engdahl, E.R., 1991. “Travel times for global earthquake
and the inferred crustal structure of continental Ecuador, Geology, 11(1), location and phase association,” Geophys. J. Int., 105, 429–465.
40–44. Kerr, A.C. & Tarney, J., 2005. Tectonic evolution of the Caribbean and
Flanagan, M.P., Myers, S.C. & Koper, K.D., 2007. Regional travel-time northwestern South America: the case for accretion of two Late Creta-
uncertainty and seismic location improvement using a three-dimensional ceous oceanic plateaus, Geology, 33(4), 269–272.
a priori velocity model, Bull. seism. Soc. Am., 97(3), 804–825. Kerr, A.C., Aspden, J.A., Tarney, J. & Pilatasig, L.F., 2002. The nature and
Font, Y. & Lallemand, S., 2009. Subducting oceanic high causes compres- provenance of accreted oceanic terranes in western Ecuador; geochemical
sional faulting in southernmost Ryukyu forearc a revealed by hypocen- and tectonic constraints, J. geol. Soc. Lond., 159(5), 577–594.
tral determinations of earthquakes and reflection/refraction seismic data, Lay, T. & Kanamori, H., 1981. An asperity model of large earthquake
Tectonophysics, 466(3–4), 255–267. sequences, in Earthquake Prediction, pp. 579–592, eds Simpson, D.W. &
Font, Y., Kao, H., Liu, C.-S. & Chiao, L.-Y., 2003. A comprehensive 3D seis- Richards, P.G., American Geophysical Union, Washington, DC.
mic velocity model for the eastern Taiwan-southernmost Ryukyu regions, Leeds, A.R., 1977. Mantle velocities in the Colombia-Ecuador region, Univ.
Terr. Atmos. Ocean. Sci., 14(2), 159–182. Javeriana, Inst. Geofis., Bogota, Colombia, Colombia, 237–242 pp.
Litherland, M. & Aspden, J.A., 1992. Terrane-boundary reactivation; control Middle East, North Africa, and former Soviet Union. Seism. Res. Lett.,
on the evolution of the Northern Andes, J. South Am. Earth Sci., 5(1), 72(2), 293.
71–76. Pedoja, K., 2003. Les terrasses marines de la marge Nord Andine
Lomax, A., Zollo, A., Capuano, P. & Virieux, J., 2001. Precise, absolute (Equateur et Nord Pérou): relations avec le contexte géodynamique,
earthquake location under Somma-Vesuvius volcano using a new 3D PhD thesis, Université de Pierre et Marie Curie, Villefranche-sur-Mer,
velocity model, Geophys. J. Int., 146, 313–331. France.
Lonsdale, P., 1978. Ecuadorian subduction system, AAPG Bull., 62(12), Pedoja, K., Ortlieb, L., Dumont, F., Lamothe, M., Ghaleb, B., Auclair, M.
2454–2477. & Labrousse, B., 2006. Quaternary coastal uplift along the Talara Arc
Lonsdale, P., 2005. Creation of the Cocos and Nazca plates by fission of the (Ecuador, Northern Peru) from new marine terrace data, Mar. Geol., 228,
Farallon Plate, Tectonophysics, 404(3–4), 237–264. 73–91.
Lonsdale, P. & Klitgord, K.D., 1978. Structure and tectonic history of the Pontoise, B. & Monfret, T., 2004. Shallow seismogenic zone detected
eastern Panama Basin, Bull. geol. Soc. Am., 89(7), 981–999. from an offshore-onshore temporary seismic network in the Es-
Manchuel, K., Regnier, M., Bethoux, N., Font, Y., Sallares, V., Diaz, J. & meraldas area (northern Ecuador), Geochem., Geophys. Geosyst., 5,
Yepes, H., 2011. New insights on the interseismic active deformation doi:10.1029/2003GC000561.
along the North Ecuadorian—South Colombian (NESC) margin, Tecton- Prevot, R., Chatelain, J.L., Guillier, B. & Yepes, H., 1996. Tomography of
ics, 30(4), doi:10.1029/2010TC002757. the Ecuadorian Andes; evidence for continuity of the Central Andes. C.R.
Manchuel, K. et al., 2009. Seismicity and structural implication in the Es- Acad. Sci. Paris. Série 2. Sciences de la terre et des planètes, 323(10),

meraldas area (Northern Ecuador): based on the Sublime and Esmeraldas 833–840.
experiments (in Spanish), in Geology and marine and onland Geophysics Reid, H.F., 1910. The mechanism of the earthquake, in The California Earth-
of Ecuador: from the continental coast to the Galapagos Islands, edited quake of April 18, 1906; Report of the State Earthquake Investigation
by J.-Y. Collot, V. Sallares and N. Pazmino, pp. 167–179, CNDM; IRD; Commission, 2, 16–28, Carnegie Institute of Washington, Washington,
INOCAR, Guayaquil (ECU); Marseille (FRA). DC.
Marcaillou, B. & Collot, J.-Y., 2008. Chronostratigraphy and tectonic de- Regnier, M., Macquet, M., Jarrin, P., Nocquet, J.-M., Mothes, P.A., Vallée,
formation of the north Ecuadorian-south Colombian offshore Manglares M. & Ruiz, M., 2011. Local seismicity in the region of the Mw 8.8 1906
forearc basin, Mar. Geol., 255(1–2), 30–44. Colombia-Ecuador subduction earthquake: new evidences of the margin
Marcaillou, B., Spence, G., Collot, J.-Y. & Wang, K., 2006. Thermal regime segmentation and of the depth of the coupled zone. AGU Fall Meeting,
from bottom simulating reflectors along the north Ecuador-south Colom- 5–9 December, San Francisco (USA).
bia margin; relation to margin segmentation and great subduction earth- Rhodes, M.J.H., 2004. Constructing a Geodynamic A Priori Seismic
quakes, J. geophys. Res., 111(B12), B12407. (GAPS) velocity model of upper mantle heterogeneity. Geochem. Geo-
Meissner, R.O., Flueh, E.R., Stibane, F. & Berg, E., 1977. Project Narino phys. Geosyst., 5, doi:10.1029/2003GC000622.
I: Dynamics of the plate boundary in South West Colombia according to Rudolph, E. & Szirtes, S., 1911. Das Kolumbianskche Erdbeben am 31
recent geophysical measurement, pp. 157–168, ed., Univ. Javeriana, Inst. January 1906, G. Beitr., 11, 132–199, 207–275.
Geofis., Bogota, Colombia. Sage, F., Collot, J.Y. & Ranero, C.R., 2006. Interplate patchiness and
Mendoza, C. & Dewey, J., 1984. Seismicity associated with the great subduction-erosion mechanisms; evidence from depth-migrated seismic
Colombia-Ecuador earthquakes of 1942, 1958 and 1979: Implications images at the central Ecuador convergent margin, Geology (Boulder),
for barrier models of earthquake rupture, Bull. Seismol. Soc. Am., 74, 34(12), 997–1000.
577–593. Sallares, V. & Charvis, P., 2003. Crustal thickness constraints on the geody-
Michaud, F., Collot, J.Y., Alvarado, A. & Lopez, E., 2006. Batimetrı́a y namic evolution of the Galapagos volcanic province, Earth planet. Sci.
Relieve Continental, ed. a. R. d. E.P. c. y. t. d. INOCAR, p. 679, IOA- Lett., 214(3–4), 545–559.
CVM-02-Post, INOCAR, Guayaquil, Ecuador. Sambridge, M.S. & Kennett, B.L.N., 1986. A novel method of hypocentre
Micheal, A.J., 1988. Effects of the three-dimensional velocity structure on location, Geophys. J. R. astr. Soc., 87(2), 679–697.
the seismicity of the 1984 Morgan Hill, California, Aftershock sequence, Scholz, C.H., 2002. The Mechanics of Earthquakes and Faulting, Cambridge
Bull. seism. Soc. Am., 78(3), 1199–1221. University Press, Cambridge, United Kingdom, United Kingdom.
Mogi, K., 1979. Two kinds of seismic gaps, Pure appl. Geophys., 117(6), Segovia, 2001. El sismo de Bahia del 4 de agosto de 1998: caracterizacion
1172–1186. del mecanismo de ruptura y analisis de la sismicidad en la zona costera,
Moser, T.J., 1991. Shortest path calculation of seismic rays, Geophysics, Tesis de previa a la obtencion del titulo de Ingeniera Geologa thesis,
56(1), 59–67. Escuela Politecnica Nacional, Quito, Ecuador, 126p.
Moser, T.J., van Eck, T. & Nolet, G., 1992. Hypocenter determination in Segovia, M., 2009. “Análisis espacio-temporal del enjambre de Puerto López
strongly heterogeneous Earth models using the shortest path method, J. entre enero y febrero de 2005 con observaciones de la estación de banda
geophys. Res., 97(B5), 6563–6572. ancha de OTAVALO”, Master 2 thesis, Université de Nice.
Nataf, H.-C. & Ricard, Y., 1996. 3-SMAC: an a priori model of the upper Segovia, M. & Alvarado, A., 2009. Breve Análisis de la Sismicidad y del
mantle based on geophysical tomography, Phys. Earth Planet. Int., 95, Campo de Esfuerzos en el Ecuador, in Geologia y geofisica marina y ter-
101–122. restre del Ecuador : desde la costa continental hasta las Islas Galápagos,
Nocquet, J.M., Mothes, P. & Alvarado, A., 2009. Geodesy, geodynamics pp. 131–149, eds Collot, J.-Y., Sallares, V. & Pazmino, N., CNDM, IRD,
and earthquake cycle in Ecuador. In: Collot Jean-Yves (ed.), Sallares V. INOCAR, Guayaquil (ECU), Marseille (FRA).
(ed.), Pazmino N. (editors) Geology and marine and onland Geophysics of Shearer, P.M., 1997. Improving local earthquake locations using the L1
Ecuador: from the continental coast to the Galapagos Islands. Guayaquil norm and waveform cross correlation; application to the Whittier Nar-
(ECU); Marseille (FRA); Guayaquil: CNDM; IRD; INOCAR, 2009, rows, California, aftershock sequence, J. geophys. Res., 102(B4), 8269–
p. 83–94. ISBN 978-9978-92-737-3. 8283.
Ocola, L., Aldrich, L.T., Gettrust, J.F., Meyer, R.P. & Ramirez, J., 1975. Swenson, J.L. & Beck, S.L., 1996. Historical 1942 Ecuador and 1942 Peru
Project Narino; I, Crustal structure under southern Colombian-northern subduction earthquakes, and Earthquake Cycle along Colombia-Ecuador
Ecuador Andes from seismic refraction data, Bull. seism. Soc. Am., 65(6), and Peru subduction segments, PAGEOPH, 146(1), 67–101.
1681–1695. Tassara, A., 2010. Control of foreac density structure on megathrust shear
Pacheco, J., Sykes, L. & Scholtz, C.H., 1993. Nature of seismic coupling strength along the Chilean subduction zone, Tectonophysics, 495, 34–47,
among simple plate boundaries of the subduction type, J. geophys. Res., doi:10.1016/j.tecto.2010.06.004.
98, 14 133–14 159. Tavera, H., Vilca, R. & Marin, G., 2006. Inferences on the geometry of
Pasyanos, L., Walter, W.R., Flanagan, M.P. & Goldstein, P., 2001. Construc- the Nazca Plate in northwestern Peru based on data collected by a local
tion and validation of the Menafsu crust and upper mantle model of the seismograph network, Earth Sci. Res. J., 10(1), 15–24.
286 Y. Font et al.
Theunissen, T., Font, Y., Lallemand, S. & Gautier, S., 2012. Improvements of Vallée, M. et al., 2013. Intense interface seismicity triggered by a shallow
the Maximum Intersection Method for 3D absolute earthquake location, slow-slip event in the Central-Ecuador subduction zone. J. geophys. Res.,
Bull. seism. Soc. Am., 102(4), 1764–1785. doi:2012JB009899.
Thomas, G., Lavenu, A. & Berrones, G., 1995. Subsidence evolution Vallejo, C., Winkler, W., Spikings, R.A., Luzieux, L., Heller, F. & Bussy,
of the northern part of the Ecuadorian Oriente basin (Upper Creta- F., 2009. Mode and timing of terrane accretion in the forearc of the
ceous to the present), (in French), C.R. Acad. Sci. Paris, 320(7), 617– Andes in Ecuador, in Kay, S.M., Ramos, V.A., and Dickinson, W.R.,
624. eds., Backbone of the Americas: Shallow Subduction, Plateau Uplift, and
Tichelaar, B.W. & Ruff, L.J., 1993. Depth of seismic coupling along sub- Ridge and Terrane Collision: Geological Society of America Memoir 204,
duction zones, J. geophys. Res., 98(B2), 2017–2037. doi:10.1130/2009.1204(09).
Trenkamp, R., Kellogg, J.N., Freymueller, J.T. & Mora, H.P., 2002. Wide Winckell, A. & Zebrowski, C., 1997. Los paisajes costeros, in Los Paisajes
plate margin deformation, southern Central America and northwestern Naturales del Ecuador. Geografısica del Ecuador, pp. 208–319, ed.
South America, CASA GPS observations, J. South Am. Earth Sci., 15, Winckel, A., CEDIG, Quito.
157–171. Witt, C., Bourgois, J., Michaud, F., Ordonez, M., Jimenez, N. & Sosson, M.,
Vaca, S., 2007. Sismotectonics of Manta-Ecuador region”, Sophia Antipolis 2006. Development of the Gulf of Guayaquil (Ecuador) during the Qua-
University, Nice-France. ternary as an effect of the North Andean block tectonic escape, Tectonics,
Vaca, S., Régnier, M., Béthoux, N., Alvarez, V. & Pontoise, B., 2009. Seis- 25(3), doi:10.1029/2004TC001723.
micity in the Manta region: the 2005-Manta seismic swarm, In: Collot Zeiler, C. & Velasco, A.A., 2009. Seismogram picking error from analyst

Jean-Yves (ed.), Sallares V. (ed.), Pazmino N. (editors) Geology and ma- review (SPEAR); single-analyst and institution analysis, Bull. seism. Soc.
rine and onland Geophysics of Ecuador: from the continental coast to Am., 99(5), 2759–2770.
the Galapagos Islands. Guayaquil (ECU); Marseille (FRA); Guayaquil: Zhou, H.-W., 1994. Rapid three-dimensional hypocentral determination us-
CNDM; IRD; INOCAR, 2009, pp. 151–166, ISBN 978-9978-92-737-3. ing a master station method, J. geophys. Res., 99(B8), 15 439–415 455.

Ggs 083 Font

Uploaded by

Copyright:

Available Formats

Ggs 083 Font

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ggs 083 Font

Uploaded by

Copyright:

Available Formats

Geophysical Journal International

Geophys. J. Int. (2013) 193, 263–286 doi: 10.1093/gji/ggs083

Seismicity patterns along the Ecuadorian subduction zone: new

Yvonne Font,1 Monica Segovia,2 Sandro Vaca2 and Thomas Theunissen3

Downloaded from https://academic.oup.com/gji/article/193/1/263/734190 by guest on 31 July 2023

plosions and seismicity on well-known tectonic features for local

Downloaded from https://academic.oup.com/gji/article/193/1/263/734190 by guest on 31 July 2023

Downloaded from https://academic.oup.com/gji/article/193/1/263/734190 by guest on 31 July 2023

Downloaded from https://academic.oup.com/gji/article/193/1/263/734190 by guest on 31 July 2023

Downloaded from https://academic.oup.com/gji/article/193/1/263/734190 by guest on 31 July 2023

Downloaded from https://academic.oup.com/gji/article/193/1/263/734190 by guest on 31 July 2023

Downloaded from https://academic.oup.com/gji/article/193/1/263/734190 by guest on 31 July 2023

Downloaded from https://academic.oup.com/gji/article/193/1/263/734190 by guest on 31 July 2023

Based on RENSIG hypocentral determinations, five geographical 4 A B S O L U T E E A RT H Q UA K E L O C AT I O N

Downloaded from https://academic.oup.com/gji/article/193/1/263/734190 by guest on 31 July 2023

Downloaded from https://academic.oup.com/gji/article/193/1/263/734190 by guest on 31 July 2023

Downloaded from https://academic.oup.com/gji/article/193/1/263/734190 by guest on 31 July 2023

Downloaded from https://academic.oup.com/gji/article/193/1/263/734190 by guest on 31 July 2023

Downloaded from https://academic.oup.com/gji/article/193/1/263/734190 by guest on 31 July 2023

Downloaded from https://academic.oup.com/gji/article/193/1/263/734190 by guest on 31 July 2023

Downloaded from https://academic.oup.com/gji/article/193/1/263/734190 by guest on 31 July 2023

Downloaded from https://academic.oup.com/gji/article/193/1/263/734190 by guest on 31 July 2023

Downloaded from https://academic.oup.com/gji/article/193/1/263/734190 by guest on 31 July 2023

Figure 9. Seismicity map comparing earthquake location resulting from

Downloaded from https://academic.oup.com/gji/article/193/1/263/734190 by guest on 31 July 2023

Downloaded from https://academic.oup.com/gji/article/193/1/263/734190 by guest on 31 July 2023

Downloaded from https://academic.oup.com/gji/article/193/1/263/734190 by guest on 31 July 2023

Downloaded from https://academic.oup.com/gji/article/193/1/263/734190 by guest on 31 July 2023

Downloaded from https://academic.oup.com/gji/article/193/1/263/734190 by guest on 31 July 2023

Downloaded from https://academic.oup.com/gji/article/193/1/263/734190 by guest on 31 July 2023

Downloaded from https://academic.oup.com/gji/article/193/1/263/734190 by guest on 31 July 2023

You might also like