The Andean subduction zone between 22 and 25°S (northern Chile): precise geometry and state of stress

~"i

':

~-~-,

":

'

"j~ill

TECTONOPHYSICS

ELSEVIER

Tectonophysics259 (1996) 81-100

The Andean subduction zone between 22 and 25°S (northern Chile)" precise geometry and state of stress Bertrand Delouis

a, *

~

Armando Cisternas a Louis Dorbath a,c Luis Rivera Edgar Kausel b 9

~

a9

a Institut de Physique du Globe, 5 rue Ren£ Descartes, 67084 Strasbourg Cedex, France b Dpto de Geologia y Geofisica, Universidad de Chile, Casilla 2777 Santiago, Chile c ORSTOM, 213 rue Lafayette, 75480 Paris Cedex 10, France

Received 17 November 1994; accepted 14 June 1995

Abstract

One year of seismicity recorded by a local network is used to obtain more precision about the geometry and the stress regime of the Andean subduction between 22 and 25°S in the northern Chile seismic gap. A sharp image of the Wadati-Benioff Zone (WBZ) is obtained down to 270 km in depth. A seismically quasi-quiescent zone is observed in the WBZ below the volcanic arc, between 150 and 210 km in depth. Hypocentres of distant intermediate depth earthquakes located with the local network are compared with worldwide seismic network hypocentres in order to evaluate the accuracy of the WBZ image at depth greater than 100-150 km in depth. No shallow microearthquakes have been observed in the continental crust but some seismic activity is likely to occur locally at the deep root of the Atacama Fault. The stress field and the characteristics of faulting along the subducted slab are investigated. Underthrusting and localized reverse faulting earthquakes define the seismically coupled plate interface from 20 to 50 km in depth (Locked zone). Downdip, intra-slab normal faulting prevails (Tensile zone), but some strike-slip faulting is observed. A transition between normal faulting with variable fault azimuth and normal faulting with nearly homogeneous NNW- to NW-oriented fault plane is found at about 80 km in depth. It is found that the stress axes ~rI and tr 3 in the Locked zone are oriented in the convergence direction (75-80°E). Downdip, in the tensile zone, cr3 has a mean azimuth 60-65°E. There, the slab is hence submitted to a tensional force (slab pull) oblique relatively to the convergence. The transition between seismic underthrusting and intraplate normal faulting downdip occurs at the depth where the continental Moho encounters the Wadati-Benioff Zone, suggesting that a relationship exists between seismic coupling and the presence of continental crust at the plate interface. The pre-seismic state of this segment of the Andean subduction zone is confirmed by the occurrence of strong earthquakes located by the global network around the presumed rupture area and by the stress regime found along the Wadati-Benioff Zone.

1. I n t r o d u c t i o n

* Corresponding author.

The Andean margin between 16 and 22°S (southern Peru and northern Chile) has been recognizec originally by Lomnitz (1970) and Kelleher (1972) as

0040-1951/96/$15.00 © 1996 Elsevier Science B.V. All rights reserved SSDI 0040- 195 1(95)00065 -8

82

B. Delouis et a l . / Tectonophysics 259 (1996) 81-100

a seismic gap with high potential of occurrence of a great earthquake. The recurrence time for great interplate earthquakes of magnitude higher than 8.5 in the area can be inferred from the analysis of the last 500 years of historical seismicity. Such a recurrence time has been estimated to be of the order of 100 years (Dorbath et al., 1990; Comte and Pardo, 1991). The last major events in the region occurred in 1868

I

18S

",

I

I

~.o

~

O0

i

M=6.4

q:)i_

i0

I

I

0

~26180 iI ~=6.~ (~

¸¸~

~

0-

0

0

~

o

~ . . . . . . . . . ~ ") I 12/11/80 ~ ' 0 ! M=6.1

i'

I0

oOO°c

~

O0

I I

)

)

20 S

(southern Peru) and 1877 (northern Chile), with magnitude estimated between 8.5 and 9 (Lomnitz, 1970; Abe, 1979; Kausel, 1986; Dorbath et al., 1990). Consequently, this seismic gap can be considered to be in its pre-seismic phase, close to rupture. The epicentral map of earthquakes recorded by the global network for the last 15 years shows that strong events surround the presumed rupture area

0

~

I I ,

0(~

05124/93 t M~.6 I

~

5~(

22 S

~_ . . . .

oo~

~

0

M=6.1

1 0 _ .E)E)_ 0

24 S

I C

I

1

o/-4 ~ I

; 72 W

03/05/87 0 0 M=7,3 ~ I 07111/93 ~ I

70 W

M:&I

~

i

68 W

66 W © I

M<6.0 6.0< M < 7.0 7.0 < M < 8.0

Fig. l. Epicentres of the earthquakes of magnitude above 5.0 located by the global network (NEIC) in northern Chile and neighbouring regions within the time interval 1980-1993.

B. Delouis et al./ Tectonophysics 259 (1996) 81-100

(Fig. 1). An epicentre has been determined by Milne (1880) for the 1877 earthquake at 21.33°S and 71.25°W from the arrival times of the seismic and tsunami waves at different localities. The intensity VIII area of the 1877 earthquake, which can be taken as an estimate of the rupture area (Dorbath et al., 1990) or epicentral area (Kausel, 1986), would extend from 19°S to the latitude of the MejillonesAntofagasta region. It is not clear whether the Antofagasta city has to be included or not in the epicentral area (Nishenko, 1985; Kausel, 1986). In any case the rupture length of the 1877 earthquake would reach at least 400 kin. South of Antofagasta is a seismic gap of 150-200 km (23-25°S) where no large earthquake has been reported historically at least since the beginning of the nineteenth century. However, an intra-slab tensional earthquake of magnitude M s = 8 occurred at intermediate depth behind this gap in 1950 (Kausel and Campos, 1992). The use of teleseismic data from the worldwide seismic network permits to define the general trend of the Wadati-Benioff Zone (Kausel and Lomnitz, 1968; Barazangi and Isacks, 1976; Chowdhury and Whiteman, 1987; Cahili and Isacks, 1992) but a higher precision on the location of hypocentres can be obtained with a local network, at least in the vicinity of the network. A first study was carried out by Olea (1967) by using a local network implanted by the Carnegie Institution of Washington. A twomonth field experiment (September-October 1988) was carried out, later on, in the same area with 29 portable stations in order to obtain a precise insight into the microseismicity of the area and to prepare the installation of a permanent network. The definition of the geometry of the subducted plate could be significantly improved. Nineteen individual and composite focal mechanisms were determined and a transition from compressional to tensional stress regime was observed at a depth of 70 km (Comte et al., 1994). The research project in this area was settled in 1988 by a joint French-Chilean team (Institut de Physique du Globe de Strasbourg and Departamento de Geoffsica, University of Chile) and the permanent telemetric seismological network was installed in June 1990 in the surrounding of the Antofagasta city and the Mejillones peninsula (northern Chile), in the geographical place of what can be considered as the

83

southern extremity of the 1877 earthquake rupture zone. This network and a similar one installed in the region of Arica in November 1994, will provide preand post-seismic coverage of the seismicity and permit the improvement of the knowledge of the present state of the subduction zone and the study of the temporal evolution of the seismic activity in the rupture area. The western part of the network spans over the Coastal Cordillera and part of the N-S-trending Atacama Fault System which extends for over 1000 km between 20 and 30°S. The eastern part of the network reaches the foot of the Precordillera. Recent works in the area have demonstrated that the fault system has undergone left-lateral displacement contemporaneous with the development of the Jurassic-Early Cretaceous magmatic arc in the present Coastal Cordillera (Naranjo et al., 1984; Herv& 1987a; Thiele and Pincheira, 1987; Scheuber and Andriessen, 1990). Furthermore it as been observed that part of the Atacama Fault System has been reactivated since the late Tertiary in a normal sense (Arabasz, 1971; Okada, 1971; Hervd, 1987b; Naranjo, 1987). On the other hand, Armijo and Thiele (1990) proposed a predominant left-lateral displacement for the recent activity of the Atacama Fault System near Antofagasta. Those two points of view are contradictory and further work on this problem is needed. The present work describes results obtained from the analysis of about one year (June 1990-August 1991) of locally recorded seismicity with the Antofagasta permanent network. As a first objective, a more precise geometry of the Wadati-Benioff Zone is looked for. Moreover, the characteristics of the stress field and its variations along the subducting slab are investigated using an algorithm of simultaneous determination of the orientation and shape of the local stress tensor, and of individual fault plane solutions (Rivera and Cisternas, 1990). Special attention is given to the characterization of the transition between seismic underthrusting at the plate interface and deeper intra-slab faulting. It was difficult to have well-constrained stress tensors for the 1988 experiment due to the limited number of focal mechanisms (nineteen). Moreover, the algorithm was applied to two groups of seven and twelve earthquakes but without a strict separation by region (see Comte et

84

B. Delouis et al. / Tectonophysics 259 (1996) 81-100

al., 1994). Experience shows that a larger number of events is generally necessary to constrain the stress orientations. On the other hand, the use of the data from a permanent local network enabled us to gather a sufficiently large number of earthquakes for that purpose.

2. Data method

acquisition

and

hypocentre

mometer and signals from the nine stations are digitally recorded on DAT tapes. Recording occurs only when a seismic event is detected at four different stations. Thus, filling of the tapes with non-tectonic local events can be avoided. Most of the recorded events occur within the subducting plate or on the plate interface. Crustal earthquakes are usually too small to be detected at four different stations. Accurate time is provided by the global Omega system. The network covers the coastal range and the longitudinal valley up to the foot of the Precordilleran range, between 22030 ' and 24°30'S (Fig. 2). The hypocentres are located with the HYPOINVERSE program (Klein, 1978). The crustal velocity model, represented by a model of flat homogeneous layers (Table 1), is chosen to match the average crustal velocity (6.6 k m / s ) and the average Moho

location

The permanent telemetric network consists of eight one-vertical-component L4C (1 Hz) seismometers distributed within 150 km distance of a central reception point situated a few kilometers northeast of the city of Antofagasta. The central station is equipped with a three-component L22 (2 Hz) seis-

\

il.//

! Atacama Fault

22

1

F$ •

~

\,

\~

'

~

"eep ee o ~Ok ° • - - ~ - ~ °e

~ • ••o .....

23

(

,~ ~ , ~

•*I

• ~..

24

~

•

~o ~

....

"~

~p <

~,~

." . ~ " : " •

e - 4 ~oeo

e~

-~.:...

~

~

/

-

_

•

~

~

•

e~ ~

.

•

~ ~

..

~ ~

. -

e ~ e~e~e~~ ~ ~

25

.

....

~ ,

~ ~

•

~, Telemetric stations -~ ®

Active volcanic front epicenters 0

71

70

100 km

69 68 67 LONGITUDE (W) Fig. 2. Epicentralmap of the 412 selectedearthquakeslocallyrecordedduring the periodJune 1990-August 1991.

B. Delouis et a l . / Tectonophysics 259 (1996) 81-100 Table 1 Velocity structure used in the location process Depth

P-velocity

(kin)

(~/s)

0 4 10 48

6.0 6.4 6.7 7.9

depth (48 km) determined by the seismic refraction profiles in the area (Wigger et al., 1991). The mantle velocity has been estimated by using a RMS minimization process and a value of 7.9 k m / s was found. Travel times are corrected for the elevation of the stations. The Vp/V s ratio was determinated from Wadati plots. A constant value of 1.76 is found for earthquakes with focal depths between 20 and 150 km but the Vp/V s ratio drops to 1.70 for greater depths. Statistical errors of time residuals at the stations proved to be highly variable as a function of focal depth but moderate in size ( < 0.2 s). Accordingly, no stations delays were used. Each earthquake was located with different trial depths so as to minimize the effect of dependence of the final hypocentral determination on the initial trial solution. Trial depths are varied from 1 to 280 km with a 10-km increment. We retained as the " b e s t " solution the one combining low RMS and the higher possible number of P and S arrivals taken into account. In order to avoid bad quality and poorly constrained hypocentres we submitted the hypocentral determinations to a sorting base on the following criteria: RMS less than or equal to 0.25, total number of phases (P + S) taken into account greater or equal to 7, at least two S-phases, computed horizontal and vertical errors less than or equal to 5 km.

3. Results of locations We obtained 412 hypocentres over the period June 1990-August 1991 by using the selection criteria indicated above, and restricting the latitude to the 22°20'-24°40'S interval. We present the distribution of the epicentres in Fig. 2 and an E - W cross section

85

in Fig. 3. The trench, the Atacama Fault System, the pre-Andean depression of the "Salar de Atacama" and the active volcanic front are shown for geographic reference. Almost all of the earthquakes are located along a narrow band of concentrated seismicity related to the subducted Nazca plate, i.e., defining the Wadati-Benioff Zone (WBZ), as can be seen on the cross section. The upper limit of the WBZ is particularly sharply defined. The WBZ dips with an angle of 17-18 ° up to about 100 km in depth. Along the deeper part of the slab, the dip increases, but this feature is discussed later on in this paper. A zone of nearly quiescent seismic activity can be observed between 150 and 210 km in depth on the cross section and is located beneath the active volcanic front. No reliable hypocentres were located below 270 km in depth. The surface projection of the deepest part of the slab corresponds essentially to a limited cluster of seismicity centered at latitude 24°10'S and longitude 67°10'W (Fig. 2). West and north of the cluster, the seismically quasi-quiescent zone can be observed to extend up to 60-80 km westward of the curved active volcanic front. Some tests were performed so as to probe if the quiescent zone could be an artificial gap produced by the location process and if the sharp definition of the deep cluster could result from an artificial concentration of hypocentres. Synthetic hypocentres with exact calculated P and S arrival times were produced in the quiescent zone of the slab and at the cluster position. Then, the exact arrival times were perturbed by adding +0.1 s (P arrivals) and _+0.2 s (S arrivals) to them and the so modified synthetic events were located using the location process described above. Events with at least one S phase were located within 5 km distance of the initial synthetic hypocentres, weather in the quiescent zone or in the cluster region, thus discarding a bias in the location process for the gap and the cluster. The only superficial event (depth < 10 km) which triggered the recording system of the local network and which could be located is an explosive blast at the Mantos Blancos mine. Some deep crustal activity is situated at a 120-140 km distance from the trench (Fig. 3). The main feature of this deep crustal activity consists of five vertically aligned hypocentres

86

B. De louis et al. / Tectonovhvsics 259 (1996) 81-100

LONGITUDE (W)

71

70

I

[

69

68

67

I

I

I

E

W Network , Atacama Fault

Altiplano-Puna I 5000 =..

Atacama" ~

Refraction

50

Volcanic Front

"Salar de

Moho

100 E ""r"" 150

o

%

I'EL ILl

c~ 200

250

LATITUDE : 22S20' - 24S40'

300

o 0 ~

100 k m ~

Fig. 3. East-west cross section showing the hypocena'al location of the 412 selected earthquakes locally recorded. The topographic profile in dark gray is ~aken at latitude 23°15'S. The refraction Moho is from Wigger et al. (1991) (interpreted seismic profile at latitude 24°15'S). The shaded area above the Moho (light gray) and its prolongation towards the trench and the east represent the continental crust.

between 20 and 50 km in depth whose epicentres are located on the main branch of the Atacama Fault.

4. Stress tensor and focal mechanisms inversion

In order to investigate the stress field and the characteristics of faulting along the subducted slab, we used an algorithm which provides simultaneous inversion of the orientation and shape of the stress tensor and of individual focal mechanisms for a population of earthquakes (Rivera and Cistemas, 1990). The absolute values of the regional stresses

cannot be determined but a shape factor or aspect ratio gives the relative value of the principal stress 0.2 with respect to 0.1 and 0.3. The method assumes that the stress tensor is locally homogeneous over the area considered. Hitherto, many analyses of the stress field from seismic data were often limited to the description of the pressure ( P ) and tension (T) axes obtained from the focal mechanisms. But simple consideration of P- and T-axes is insufficient, because they may not correspond to the maximum and minimum regional stress axes, 0.1 and 0.3- Indeed, 0.~ and 0"3 may lie at any point within the dilation and compression quadrants, respectively (McKenzie,

B. Delouis et al./ Tectonophysics 259 (1996) 81-100

1969). The advantage of the method is that we obtain the stress tensor not from previously and individually determined focal mechanisms which contain a certain degree of arbitrary choice, but rather from the original data of first motion polarities. The orientation in space of the three eigenvalues of the stress tensor ~rx, ~y, ~r., and the shape factor R = (~r. - ~.,.)/(~ry - ~r.~.) are obtained, with ~r~ > % and ~r. being the eigenvector closest to the vertical. The shape factor R which varies between - ~ and + ~ is a measure of the relative stress magnitudes. The inverse problem is non linear and it is solved iteratively starting from a trial stress tensor and a set of initial fault planes. As it is often the case in this type of inverse problem, the final solution may be dependent of the initial or trial solution. We explored the possibility that several final solutions of similar quality may coexist. In that case, the range of variation was determined and only those focal mechanisms which do not vary significantly with the different stress tensor solutions were retained. The interpretation of the focal mechanisms at depth requires the most trustworthy locations. Therefore, we submitted the previously selected hypocentres to an additional selection criterion. We tested the stability of the hypocentral locations with different velocity models. Earthquakes were located with five different models in which the mean crustal velocity varied between 6.42 and 6.77 k m / s (P wave), the Moho depth between 42 and 54 km, and the mantle velocity between 7.8 and 8.0 k m / s . In addition, the V p / V S ratio was varied between 1.74 and 1.78. We retained as a final selection those earthquakes whose location varied less than 5 km in latitude and longitude and less than 6 km in depth. One way to detect variations in the stress regime in the Wadati-Benioff Zone is to define a sliding window along the slab. We limited our investigation to the upper part of the slab (west of 69°W, depth < 100 km). Deeper earthquakes are too distant from the local network to permit a good enough constraint of their focal mechanisms. Only earthquakes with a minimum of six polarities were included in the inversion. Performing the first inversions of the stress tensor and the focal mechanisms for different depth intervals, we soon identified a sharp transition between an upper zone where focal mechanisms exhibit un-

87

derthrusting and reverse faulting and a lower zone where normal faulting prevails. The transition is located at about 50 km in depth. Accordingly, we isolated the first portion of the slab, from 20 km in depth down to the point where the upper limit of the Wadati-Benioff Zone (WBZ) reaches the depth of 50 km, and we will refer to it as the Locked zone. The deeper portion of the slab was divided into two parts of about the same length, the Tensile zones parts 1 and 2. We decided to divide the normal faulting-tensile zone into two parts in order to investigate the possibility to have a variation of the stress regime between 50 km and 100 km in depth. Focal mechanisms resulting from the inversion were submitted to a last selection criterion. We examined the possibility to have some variations of the focal mechanisms related to changes in azimuth and incidence of the P arrivals when the velocity model was varied. Strong variations of take-off angles can occur when the hypocentre is located within the possible depth range of the Moho discontinuity or close to it. When the simultaneous inversion gave several solutions of similar quality (Fig. 4) we identified those earthquakes whose focal solution varied strongly with the stress tensor. Only focal mechanisms that proved to be stable are presented (Fig. 5; Appendix A). Some focal mechanisms may appear poorly constrained individually due to the small number of first motion polarities (6-9) or because they correspond to earthquakes located outside or near the border of the network coverage (Appendix A). However, an external constrain is added when the focal mechanisms have to be compatible with a local stress tensor. Unfortunately, the few crustal earthquakes recorded do not have enough polarities, or have focal mechanisms poorly constrained, in order to be considered. 4.1. The Locked zone

The inversion concerns 43 earthquakes concengated along the most superficial part of the WBZ or located immediately below it. Most of the focal mechanisms are of thrust or reverse type with a nodal plane oriented close to north-south (Fig. 5; Appendix A). The fact that we have mainly mechanisms of similar kind makes the inversion of the

B. Delouis et al./ Tectonophysics 259 (1996) 81-100

88

(a)

LOCKED ZONE

(b)

T E N S I L E ZONE PART 1

G2

Score = 0.954

Score = 0.954

Score = 0.976

Shape Factor : R = -1.4

Shape Factor : R - 1.8

Shape Factor : R = 2.6

0"I : a z i m u t h 258"

dip 25*

0"I : a z i m u t h 247"

dip 52"

0"2 : a z i m u t h 167"

d,ip 02"

0"2 : a z i m u t h 342"

dip 04"

0"3 : a z i m u t h 072"

dip 65"

0"3 : a z i m u t h 075"

dip 38"

(~I : azimuth 223"

dip 68"

0"2 : a z i m u t h 330"

dip 0 7 '

0"3: azimuth 063 ° dip 21"

TEINSILE ZONE PART 2

(c) 0"2

•

Score = 0.994

Score = 0.989

Shape Factor : R - 2.0

Shape Factor : R = - I . I

O l : a z i m u t h 238"

dip 78"

O l : a z i m u t h 205"

dip39"

O2: azimuth332"

dip01"

0"2: azimuth 307"

dip I I "

O3:

dip 12"

0 3 : azimuth 050 ° dip 49"

azimuth062"

Fig. 4. Stress tensor delermination for the three studied segments of the Wadati-Benioff Zone. (a) Locked zone. (b) Tensile zone part 1. (c) Tensile zone part 2. As several solutions of similar quality are found for the Locked zone and the Tensile zone part 2, two stress tensors solutions are presented in (a) and (c) so as to show the range of variation of the stress axes orientations in those cases. Stress tensors are represented on the lower hemisphere equal area projection.

stress directions harder. Nevertheless, a coherent pattern emerges. The maximum and minimum principal stress cr~ and ~r3 lie within a vertical plane striking 75°E ( + 50, the intermediate stress cr2 being horizontal and striking 345°E (___5°). In all cases, ~rI dips towards the west and ~r3 towards the east, but the inclination of the two axes is not strongly constrained. The maximum stress ~r1 may move from a

nearly horizontal direction (dip 2 0 - 2 5 °) to a more vertical position (dip 50-55°). Conversely, the range of the dip of cr3 is 3 5 - 7 0 ° (Fig. 4a). The shape factor R may take negative or positive values indicating either a compressional or an extensional stress regime, depending on which of the ~3 and ~rI axes is the closest to the vertical. This result indicates nevertheless that the relative value of the intermediate

B. Delouis et al,/ Tectonophysics 259 (1996) 81-100

stress ~r2 is clearly differentiated from the values of ~r1 and ~r3. The score, defined as the overall percentage of polarities consistent with the stress tensor, is 95%. Two equally likely stress tensor solutions are presented in Fig. 4a so as to show the range of variation of the stress axes orientation. Focal mechanisms of the locked zone presented in Fig. 5 and Appendix A are calculated with the stress tensor with shape factor equal to - 1.4 (Fig. 4a, left). Underthrusting with a nodal plane dipping slightly towards the east is characteristic of the upper part of the Locked zone between 20 and 35 km in depth. Focal mechanisms in the range 35-50 km in depth

89

are rather of reverse type, with nodal planes dipping in a steeper way towards the east (events LZ-4, -5, -8, -11, -17, -19, -21 and -26, Fig. 5). For three of those events, LZ-8, -19 and -21, the inversion favored such a focal mechanism solution, which moves away the nodal planes from the first motion data (see Appendix A). However, a slightly different focal mechanism, with a N-S-oriented fault plane dipping 17° towards the east, could also be found in compatibility with the stress tensor (Appendix B). On the other hand, no underthrusting focal mechanism solutions could be found for the reverse faulting events LZ-4, -5, -11, -17 and -26.

LONGITUDE (W) 71

70

69

68

67

I"11 m

z

50

E 100 ~. -r" Ira tu 150

200

250 Fig. 5. Focal mechanisms and the stress field along the Wadati-Benioff Zone (WBZ) up to 100 km in depth. The WBZ has been divided into three consecutive segments corresponding to different depth ranges [dark gray = Locked zone (LZ); medium gray = Tensile zone part 1 (721); light gray = Tensile zone part 2 (TZ2)]. In each segment the fecal mechanisms and the stress tensor have been determined from the first motion polarities using the Rivera-Cisternas algorithm (1990). Focal mechanisms and stress tensors are represented on the lower hemisphere equal area projection. The shaded areas in the focal mechanisms are the compressional quadrants. The shaded areas in the stress tensor projections indicate the range of variation of the stress axis solutions of similar quality obtained by distinct inversions.

90

B. Delouis et al./ Tectonophysics 259 (1996) 81-100

It is not clear whether it would be justified or not to treat the 20-35 km and 35-50 km depth ranges separately since events LZ-19 and -21, located at about 50 km in depth, could be alternatively interpreted as underthrusting earthquakes. In any case, the reduced number of events between 35 and 50 km in depth would not allow a separate inversion of the stress field and of focal mechanisms. Although some reverse faulting events exist, we interpret the earthquakes of the upper part of the Wadati-Benioff zone, between 20 and 50 km in depth, as defining the seismically coupled part of the plate interface. The relative displacement of the two plates is stopped in this locked area during the interval of time separating two great decoupling earthquakes. We use the definition given by Tichelaar and Ruff (1991, 1993)) for the seismically coupled zone, namely the depth range of the plate interface capable of producing underthrusting earthquakes. The normal faulting event, located 100 km east of the trench, lies 15-20 km beneath the interface (event LZ-24, Fig. 5).

4.2. The Tensile zone (part 1) The inversion in this zone, based on 23 earthquakes, produced a very good result, with a better constraint on the stress directions as the focal mechanisms are of different types. The minimum stress ~r3 strikes 65°E ( + 10°) and dips 20-25 °. The maximum stress 0-1 dips 60-70 ° in the 225°E ( + 10°) direction. The intermediate ~r2 is close to the horizontal and strikes 330°E ( + 10°). The stress regime is clearly extensional (Fig. 4b). The relative value of 0-2 is well differentiated from those of 0-1 and 0-3. The overall score is 98%. Focal mechanisms show predominantly normal faulting, but some strike-slip faulting is observed. In particular, four strike-slip events are located close to the lower boundary of the Locked zone, some 120-140 km east of the trench. Normal faulting events are characterized by a strong variability in the azimuth of the nodal planes (Fig. 5).

4.3. The Tensile zone (part 2) The inversion in this zone concems 26 earthquakes located in the deepest part of the Wadati-Be-

nioff Zone we have analyzed. Many focal mechanisms are of the same kind, presently normal faulting with a nodal plane oriented in a NNW to NW direction, and the stress tensor is rather poorly constrained, though its characteristics prove to be coherent. The minimum stress 0"3 strikes 60°E ( _+20 °) and dips 10-50 °, whereas 0-~ strikes 225°E ( + 2 0 °) and dips 40-80 ° . The best constrained axis is 0-2 which strikes 325°E (-I-20 °) and dips 0 - 1 0 ° (Fig. 4c). There is no control on the shape factor value R and the relative stress magnitudes cannot be determined. The overall score is 99%. Two equally likely stress tensor solutions are presented in Fig. 4c. Focal mechanisms of the Tensile zone part 2 shown in Fig. 5 and Appendix A are calculated with the stress tensor with shape factor equal to 2.0 (Fig. 4c, left). Although the stress tensor, and principally the 0-~ and 0-3 inclinations, may present large variations, the focal mechanisms do not exhibit significantive changes, except for only six events. The deepest eastern earthquakes exhibit a homogeneous pattern with a nodal NNW- to NW-oriented plane whereas earthquakes close to the Tensile zone part 1 show a more complex pattern with normal faults of variable azimuth.

5. Discussion

5.1. Geometry of the Wadati-Benioff Zone The most striking features of the sharp image obtained, as observed in cross section (Fig. 3), are the well-defined upper limit of the Wadati-Benioff Zone (WBZ), the existence of a seismically quasiquiescent zone beneath the active volcanic front and the rather linear cluster formed by the locations of the deepest earthquakes which give the image of an overall continuity down to 270 km in depth. The sharpness of the locations can be explained as follows. First, and most obvious, is the selection criterion. This criterion acts like a filter, eliminating hypocentres with large errors or those poorly controlled, but it does not influence the locations themselves. A second factor is the sweeping of the trial depth over the possible range of depth, with a small enough increment. For the deepest earthquakes, such a process proved to be absolutely necessary in order

B. Delouis et a l . / Tectonophysics 259 (1996) 81-100

to reach the optimum location. Otherwise, the algorithm will, quite often, reach only a secondary extreme. To test the importance of the sweeping procedure, we relocated all the 419 earthquakes with a trial depth fixed to 33 km. Events shallower than 65 km were generally well relocated. Part of the events initially located between 65 and 120 km were relocated at a smaller depth, generally close to 33 km. The slab continuity did not reach farther than 120 km in depth and all the deepest events were relocated around 33 km in depth (Fig. 6). It should be noticed that the HYPOINVERSE program (Klein, 1978), though very performant, is particularly sensitive to the trial depth. Third, the farthest events have the advantage to occur at focal depths and epicentral

91

distances of similar value. If the focal depth would be much smaller than the horizontal distance the depth uncertainty would be higher. The synthetic tests discussed earlier proved that the quiescent zone beneath the active volcanic front is not directly produced by the location process. However, another possible origin for an artificial gap would be the existence of large lateral heterogeneities not taken into account in the simple flatlayered velocity model. However, quite the same quiescent zone can be observed on the cross section based on teleseismic data (for example, Cahill and Isacks, 1992). Leaving the source area, seismic rays recorded at the local stations travel upwards while rays recorded at the teleseismic stations travel down-

LONGITUDE (W) 71

70

I

I

69 I

68

67

I

I

E

W Network

50

100 E ~ --r In ILl

150

~

200

250 0

100 km

300 Fig. 6. Same as Fig. 3 except that the trial depth is fixed to 33 km in the location process instead of being varied between 1 and 280 km with a 10 km increment. Compare with Fig. 3 and see text for further explanation.

92

B. Delouis et al. / Tectonophysics 259 (1996) 81-100

wards. Thus, lateral heterogeneities cannot have the same influence in both cases and we conclude that the quiescent zone we defined is not an artifact. From southern Peru to northern Chile (19-24°S), a strong concentration of hypocentres can be observed between 80 and 140 km in depth under the Precordillera up to the volcanic front (Wigger, 1988; Cahill and Isacks, 1992). The quiescent zone starts immediately downdip of it and extends up to 180200 km. Seismicity between 180 and 280 km in depth occurs in separate clusters. Faulting in the 80-140 km depth range may be facilitated by an effective dehydration of the oceanic crust and by the basalt-eclogite phase transformation (Liu, 1983; Haak and Giese, 1986; Kirby and Hacker, 1993). The onset of the quiescent zone may correspond to the limit beyond which those mechanisms end a n d / o r to the proximity of the asthenospheric wedge beneath the volcanic front. Lateral heterogeneities, for instance a velocity contrast between the slab and the mantle above, may have an influence on the dip of the deepest part of the slab. McLaren and Frohlich (1985) showed that if a slab/mantle velocity contrast of a few percent exists (higher velocity in the slab), locations with a classic flat-layered velocity model can produce an increase in the apparent dip of the WBZ. Moreover, we verified in the case of our simple velocity model that an increase of 2 - 3 % of the mantle velocity produced a global displacement of the deepest hypocentres of about 15-20 kin, in the upward-eastward direction. Consequently, the dip of the lowest part of the slab calculated with the local network should be considered with some care, as it is strongly related to the velocity structure. Teleseismic locations are not so much affected by the slab/mantle velocity contrast. The WBZ obtained either with the teleseismic data (Cahill and Isacks, 1992) or with the local network (present study) may differ significantly below 100 km in depth. At the depth of 250 km, the "local image" of the WBZ can be shifted 40 km downwards and westwards relative to the "teleseismic image". Local seismic stations between 67 and 69°W would permit to constrain the actual dip of the subducted slab at that depth. On the other hand, the 17-18 ° dip of the upper part of the slab down to 100 km in depth beneath the seismic network proved to be very insensitive to variation of the velocity

model. We verified that the use a simple model such as a half space with a unique P-velocity of 6.8 k m / s does not affect this result.

5.2. Interplate, intraplate faulting and the related stress regime The Locked zone, extending up to 50 km in depth, exhibits a clear pattern of thrust to reverse focal mechanisms with a nodal plane roughly parallel to the plate margin. The slip vectors on the nodal planes of underthmsting earthquakes with a shallow east dip are mostly oriented in ENE direction, and therefore are compatible with the relative motion of the Nazca and South American plates. Although the rupture and the seismic moment release of great earthquakes may extend further down, the Locked zone represents the present state of the seismically coupled part of the plate interface. Strike-slip faulting occurs at about 50 km in depth and below we observe tensional focal mechanisms with nodal planes of variable azimuth. Finally, normal faulting becomes more homogeneous as the WBZ reaches 8090 km in depth with a NNW- to NW-oriented nodal plane close to the vertical or dipping steeply towards the east (Figs. 5 and 7). This characteristic nodal plane is geometrically incompatible with faulting along the 17-18 ° E-dipping plate interface and it would indicate intra-slab faulting. In the Rivera and Cisternas method for joint determination of the stress tensor and focal mechanisms it is possible to discern, in most cases, which one of the nodal planes is the actual fault. The stress tensor determines the slip vector on a given fault plane, and hence the second nodal plane of the focal mechanism. The action of the same stress tensor on this second nodal plane will not generate, in general, the same focal mechanism. Hence the symmetry between both nodal planes is broken when the stress tensor is given. However, discrimination is not possible if the stress tensor has cylindrical symmetry (R = 0, R = 1, R = o~) or if the fault plane passes through one of the principal axes of stress (Rivera and Cisternas, 1990). In the Locked zone the interface contact dipping 17-18 ° is expected to be a major fault plane. However, we verified that the fault plane identified by the inversion does not always coincide with the shallow

93

B. Delouis et al./ Tectonophysics 259 (1996) 81-100

homogeneity of the deepest focal mechanisms. As normal faulting becomes homogeneous with a NWto NNW-oriented nodal plane, the minimum stress 0.3 may be oriented up to 50°E. Unfortunately the inversion algorithm cannot effectively distinguish the fault plane from the auxiliary plane for those homogeneous focal mechanisms because they have at least one of their nodal planes very close to the 0.2 axis. Nevertheless, we interpret the nodal plane dipping steeply towards the east-northeast to be the fault plane because the conjugate shallow dipping plane is unfavorably oriented with respect to the stress tensor. The dip of 0"3 is not strictly constrained but if we take it to be equal to the slab dip at that depth, i.e., 20 °, this conjugate plane typically lies at an angle of 60-70 ° to the greatest compressive stress 0"l, and therefore, the normal stress should be very high over this plane. Furthermore, the steeply dipping NW- to NNW-oriented nodal plane is characteristic of nearly all the large magnitude tensile earthquakes occurring at intermediate depth in the area (Stauder, 1973; Astiz et al., 1988), including the 1950 earthquake

nodal plane dipping towards the east. The algorithm cannot discriminate the actual fault plane because underthrusting focal mechanisms have nearly N-Soriented nodal planes very close to the 0"2 axis. We verified that both nodal planes could be chosen as the fault plane in compatibility with the first motion data and the stress tensor. In such a case, the choice of the fault plane by the algorithm depends mostly on the trial fault plane solution taken at the beginning of the iterative process. The stress tensors found for the Tensile zones (parts 1 and 2) are mutually compatible and reflect the downdip tensional regime prevailing in the upper part of the oceanic plate below the Locked zone. There is no clear geophysical difference between the two tensional parts apart from the fact that faulting becomes more homogeneous in the deepest part of the Tensile zone part 2. Both are characterized by a minimum stress 0"3 of mean azimuth 060-065 ° . The relatively large uncertainty on the determination of the stress directions for the Tensile zone part 2, may be attributed to the lack of constrain due to the

I Normal Faulting with ' variable _:__.azimut. x~

~

x~o

I I

._.,.,oot~i,,,~b~" s~ts~ ~~

~ ~

~

,,~,~

" " ~ G ~ I"~" ~

~

Normal Faulting with a nodal plane of nearly, constant azimut

:

/~

~

OE

~

~ ~ 20 km

~

50kin

•

~ -

~

~:l~MIc ~N7.~.,~,~4c ~

~ / ~'- ~ _~2

~

,~ 80

km

. . . . .

'~

- --/

F ,~ E V '~, ,y

~ Z

°es4,v~e ~oe,%~.e s ~

---~-

-

~

,~.~'~ ~ 0 ~

Strike-slip Faulting [ Localized Reverse l Faulting Fig. 7. Schematicrepresentationof the upper part of the Wadati-BenioffZone based on the presentlocal seismicityanalysis.

94

B. Delouis et al. / Tectonophysics 259 (1996) 81-100

south of Antofagasta (M s = 8, Kausel and Campos, 1992). Finally, if the conjugate shallow dipping nodal plane was the fault plane, the lower part of the slab would be uplifted with respect to the deeper part of the slab. Such a relative displacement could hardly be developed on a large scale in the downgoing plate. No clear bending of the subducted plate can be seen on the cross section below 50 km in depth (Fig. 3). Normal faulting events below 50 km in depth do not constitute the lower seismic plane of a double seismic zone, nor do we observe an upper seismic plane characterized by down-dip compression such as the one found beneath northeastern Honshu in Japan (Yoshii, 1979). This fact indicates that unbending of the oceanic plate at such depths does not control the state of stress. Hence, tension in the oceanic plate downdip of the Locked zone is more likely to be connected to a downdip tensional force, or slab pull force. It is generally proposed that this force is related to the density contrast and the viscous coupling between the slab and the surrounding mantle (Forsyth and Uyeda, 1975; Spence, 1987). If we consider that this force is oriented in the (r 3 direction, our results suggest that the slab pull force is slightly oblique with respect to the convergence direction (75-80°E), especially below 80 km in depth in the Tensile zone part 2. The general shape of the subducted Nazca plate beneath southern Peru and northern Chile has its steepest dip around latitude 18-20°S. To the south, the slab flattens slowly up to latitude 32°S (for example, see Cahill and Isacks, 1992). The slab tensional force may be oriented in a more NE direction than the convergence in northern Chile because of the slab tendency to dip more steeply towards the north. In the Locked zone, a stronger mechanical coupling between the slab and the continental plate of 05°-10°E main structural trend may prevent such an important NE-oriented deviation of the ~r~ and cr3 directions. It is very likely that nearly all the studied earthquakes in the shallow part of the slab (Locked zone) are located in the shear zone at the plate interface. Although the shear zone may be some kilometers wide, the fault planes do not vary much from the expected low-angle (15-20 °) E-dipping plane down to 35 km in depth. Between 35 and 50 km in depth reverse faulting events may indicate localized varia-

tions in the dip of the contact zone possibly produced by strong irregularities at the plate interface o r / a n d intra-slab faulting. The available set of focal mechanisms concentrated at the plate interface in the Locked zone give a good insight into the plate kinematics but does not permit to remove the uncertainty on the stress tensor determination. This is a very distinct situation from that of an intraplate environment where the faulting is not restricted to a kinematic boundary but can occur in a rather homogeneous volume in response to the regional stress field created by distant plate kinematics. In that case, earthquakes occurring on preexistent faults have generally more varied focal mechanisms and the stress field is easier to characterize (for example, see Delouis et al., 1993). In any case, the direction of (r~ and (r3 which lie in the vertical plane of azimuth 75 ° as well as the slip vectors in the Locked zone are in good agreement with the convergence (75-80°E). The transition between interplate underthrusting and intraplate normal faulting can be related to the change in the mechanical behavior of the interface which would undergo a unstable (seismic)-stable (aseismic) frictional slip transition. The conditions that permit unstable slip (stick-slip) along the subduction interfaces at depth of 20-50 km and the nature of the further transition between unstable and stable slip are not well understood. A complex association of factors should interfere (Shimamoto, 1985; Tichelaar and Ruff, 1991): the temperature and pressure gradients; the presence or the absence of sediments along the interface; the frictional properties of the material in the contact zone; the elastic and inertial properties of the loading system; the presence or abundance of water; and the depth of the basalt-eclogite phase transformation. In the present case, the lower limit of the underthrusting earthquakes occurs approximately at the depth where the Wadati-Benioff Zone and the Moho seem to diverge, suggesting that the crust to mantle transition in the continental plate may control the transition in slip stability at the plate interface. The 20-50 km depth range for seismic coupling inferred from the local data is in agreement with the depth of well-studied large interplate earthquakes in the region. The hypocentre of the shallower large underthrusting earthquake (December 8, 1971,22.9°S and 70.1°W, M,~ = 6.1) is located at 24 km in depth

B. Delouis et al./ Tectonophysics 259 (1996) 81-100

(Tichelaar and Ruff, 1991) and the deepest one (December 21, 1967, 21.9°S and 70.1°W, M w = 7.1-7.4) at 47-48 km in depth (Malgrange and Madariaga, 1983; Tichelaar and Ruff, 1991). Malgrange and Madariaga (1983) noticed that this latest event seemed to be too deep for the interface between the oceanic and continental plates. However, the image of the slab we obtain from the local data (Fig. 3) indicates that this earthquake can be considered to occur in the interplate area. Some teleseismic tensional events occurring below the underthrusting earthquakes have been reported in northern Chile, suggesting that the slab pull force acts within the shallower part of the subducted Nazca plate (Stauder, 1973; Chinn and Isacks, 1983; Lay et al., 1989; Malgrange and Madariaga, 1983). In the present study, we located one tensional event 15-20 km beneath the interface (event LZ-24, Fig. 5). Four strike-slip faulting events of very similar mechanisms have been located at about 50 km in depth just downdip of the Locked zone in the oceanic plate (Fig. 5). There, the subducted oceanic plate may not be so strongly mechanically coupled to the overriding plate. In that case, high stresses accumulate by the slab pull at the base of the Locked zone may be responsible of some differential slip within the slab. Strike-slip related to slab pull in the subducted oceanic plate has been observed for instance in the Sunds subduction zone by Spence (1987). Underthrusting earthquakes were observed only down to 47 km in depth in Comte et al. (1994), but a change in the Wadati-Benioff zone stress regime from compressional to tensional was observed at 70 km in depth. It was proposed that 70 km could be an alternative value for the maximum depth of coupling. This scheme relied essentially on two reverse faulting earthquakes located at 56-63 km and 69 km in depth in the continuity of the plate interface. One of this events corresponded to a composite focal mechanism and hence is less certain. The other one, which had only six polarities, was the least constrained focal mechanism of the study. Thus the compressional character of the plate interface region between 50 and 70 km was not so strongly established, and it is not confirmed by the data from the permanent network. We do not observe thrust or reverse faulting focal mechanisms, indicative of a compressional

95

stress regime below 50-55 km in depth, with the data from the permanent network. Furthermore, normal faulting earthquakes are observed to occur from 55 km in depth downwards. Hence, we postulate that the transition from thrust or reverse to normal faulting focal mechanisms and the seismic-aseismic transition at the plate interface coincide at about 50 km in depth. Seismic coupling would end at that depth, at least for small to moderately large earthquakes.

5.3. Deep crustal activity At about the same distance from the trench (120140 km), some deep crustal seismicity occurs between 20 and 50 km in depth. In a "normal" continental environment, P - T conditions would prevent seismic rupture at such depth. However, the proximity of the cold subducted slab may affect the thermal structure of the continental crust in such a way as to lower the temperature below 300°C. Thermo-mechanical modeling by Van den Beukel and Wortel (1988) support this assumption. The temperature of 300°C corresponds to the unstablestable slip transition for crustal rocks (Scholz, 1989). The deep crustal seismic activity is restricted to a 50 km long cluster of N - S trend located beneath the main branch of the Atacama Fault on the northern limit of the network. Hence, the observed deep crustal events may represent seismic activity at the deep root of this major fault.

6. Conclusions

The seismic monitoring of a segment of the northem Chile subduction zone with a local network incorporating nine seismic stations provided a very sharp image of the Wadati-Benioff Zone (WBZ). The absolute location of the deeper part of the slab has to be considered with care as it is very sensitive to the velocity model. The subduction zone is in a pre-seismic state in the segment between Antofagasta and Arica. This conclusion is confirmed by two observations: the occurrence of strong earthquakes around the presumed rapture area, and the variation of the stress regime along the subduction interface. The difficulties encountered in constraining indi-

96

B. Delouis et al. / Tectonophysics 259 (1996) 81-100

vidual focal mechanisms with a limited number of first motion polarities (6-9) can be overcome by simultaneously inverting for the stress tensor and the focal solutions. Stresses and focal mechanisms are found to vary along the upper part of the WadatiBenioff Zone (up to 100 km in depth) in a very coherent fashion. From 20 to 50 km in depth, most reliable hypocentres appear to be located in the shear zone between the two plates, defining the seismically coupled part of the interface. Nearly all focal mechanisms indicate underthrusting or localized reverse faulting (Locked zone). We did not detail the longitudinal (N-S-oriented) variation of the seismic coupling because the number of underthrusting events was not sufficient. However, as the network will be operating permanently, the two-dimensional mapping of the seismically coupled interface will become one of our main objectives in the near future. We aim at the identification of asperities on what should be the rupture surface of the next great subduction earthquake in the area. The interface appears aseismic below 50 km in depth and normal faulting occurs within the upper portion of the oceanic subducted plate. From 50 to 100 km in depth, normal faulting indicates that the slab is submitted to a tensional force (slab pull) oriented more or less downdip (Tensile zone). It is only at about 80 km in depth that normal faulting becomes homogeneous along NNW- to NW-oriented planes with a steep east dip. We evidenced the existence of a transition zone between 50 and 80 km in depth where normal faulting seems to occur in all directions. Stress orientations have enough resolution to evi-

dence a rotation of the stresses along the slab. The maximum and minimum stress axes ~r~ and tr 3 lie in a quasi-vertical plane all along the slab. In the Locked zone the plane is oriented along the convergence direction (75-80°E) but it undergoes a 10-20 ° counterclockwise rotation in the Tensile zone below. We propose that this more NE orientation in the Tensile zone is connected to the three-dimensional effect of the general trend of the subducted slab which dips more steeply towards the north. Also, we infer that this stress rotation and the seismic-aseismic transition along the interface, occurring at 50 km in depth, may be related to the crust/mantle transition in the overlying plate, the continental crust promoting strong mechanical coupling between the two plates in the Locked zone.

Acknowledgements This work has been supported by a grant from ORSTOM and by the ECOS program of cooperation between the Ministry of Foreign Affairs (France) and CONICYT (Chile). Partial financing has been given by the DBT program of the National Institute of Science of the Univers (INSU). We thank B.W. Tichelaar and an anonymous reviewer for improving the manuscript with constructive criticisms and helpful discussions. We would like to thank M. Schmitz for helpful discussions about the velocity structure of the crust as seen through the seismic refraction data in the area.

97

B. Delouis et al./ Tectonophysics 259 (1996) 81-100

Appendix A

sional polarities. As discussed in the text, only those focal mechanisms which do not vary significantly with the velocity model and the stress tensor are presented.

Focal mechanisms resulting from the inversion. All mechanisms are represented on the lower hemisphere equal area projection. Dots are the compres-

LOCKEDZONE

LZ-1

LZ-2

@@

LZ-3

LZ-

LZ-5

LZ-9

LZ- 10

LZ-6

LZ-7

LZ-8

LZ- 16

LZ- 17

LZ-18

o@ @ @@ LZ- 19

LZ- 20

o

LZ- 21

LZ- 22

LZ- 23

LZ- 25

•

LZ- 26

•

B. Delouis et al. / Tectonophysics 259 (1996) El-100

nodal plane dipping towards the east. The algorithm cannot discriminate the actual fault plane because underthrusting focal mechanisms have nearly N-Soriented nodal planes very close to the u2 axis. We verified that both nodal planes could be chosen as the fault plane in compatibility with the first motion data and the stress tensor. In such a case, the choice of the fault plane by the algorithm depends mostly on the trial fault plane solution taken at the beginning of the iterative process. The stress tensors found for the Tensile zones (parts 1 and 2) are mutually compatible and reflect the downdip tensional regime prevailing in the upper part of the oceanic plate below the Locked zone. There is no clear geophysical difference between the two tensional parts apart from the fact that faulting becomes more homogeneous in the deepest part of the Tensile zone part 2. Both are characterized by a minimum stress o3 of mean azimuth 060-065”. The relatively large uncertainty on the determination of the stress directions for the Tensile zone part 2, may be attributed to the lack of constrain due to the

93

homogeneity of the deepest focal mechanisms. As normal faulting becomes homogeneous with a NWto NNW-oriented nodal plane, the minimum stress u3 may be oriented up to 50”E. Unfortunately the inversion algorithm cannot effectively distinguish the fault plane from the auxiliary plane for those homogeneous focal mechanisms because they have at least one of their nodal planes very close to the u2 axis. Nevertheless, we interpret the nodal plane dipping steeply towards the east-northeast to be the fault plane because the conjugate shallow dipping plane is unfavorably oriented with respect to the stress tensor. The dip of u3 is not strictly constrained but if we take it to be equal to the slab dip at that depth, i.e., 20”, this conjugate plane typically lies at an angle of 60-70” to the greatest compressive stress u,, and therefore, the normal stress should be very high over this plane. Furthermore, the steeply dipping NW- to NNW-oriented nodal plane is characteristic of nearly all the large magnitude tensile earthquakes occurring at intermediate depth in the area (Stauder, 1973; Astiz et al., 1988), including the 1950 earthquake

Normal

Faultingwith

/ 6%alized

Fig. 7. Schematic

,

Reverse

1 Faulting representation

j of the upper part of the Wadati-Benioff

Zone based on the present local seismicity

analysis.

B. Delouis et al, / Tectonophysics 259 (1996) 81-100

99

Appendix A (continued) TZ2- 13

TZ2- 12

TZ2- 14 ~,

/

\ •

~ ~-. \-~

_

~<~

~

TZ2- 17 < / - t ~ - ~ - _ _ _- . \

\

•

"

~

•

~ q

\\~

d

N~ ,

o

~

"

..

•

\\

~

~

.....

,

....

~

~

~

/

,

'~ ', (.... •

\,~

.,

'

,

/' /

~ o o. _ _- -

~~ p "

TZ2- 19 -' - , /'/ '~

/~/

~.

t~

~

~-.~

TZ2- 18 / ..... ~ - , ~

TZ2- 15..__ ~ _

_~_

o

\

•

/'/

.,,

\\\ •

,,~,

,;.)

,~

'\~/ ~

\'

~\

o~ ~

~ 2 - 20

.:~

. . .

, "" ~ . _ ~ _ _ ~

_ _~<~

-,_ 2-

~

~

.....----'"

Appendix B Alternative focal mechanisms with a N-S-oriented fault plane dipping 17° towards the east for three earthquakes of the Locked zone. The steeply dipping auxiliary plane is determined by the stress tensor (Fig. 4a, left). The focal mechanisms of those three earthquakes resulting from the inversion are of reverse type (Appendix A) but it is demonstrated here that the first motion data of those three events are consistent with underthrusting focal mechanisms. LZ- 8

LZ- 19

~,,,,,/ '?,

. . . .

-

LZ- 21

Appendix C A destructive M w - - 8 earthquake took place at the southern end of the gap, along the subduction zone under Antofagasta, while this paper was in print. The epicenter was located within our permanent local network (23°43'S, 70°48'W, depth 36 km), and the focal mechanism was a shallow angle thrust. This large event may be regarded as a precursor of the even larger one which is expected to break the segment between Antofagasta and Arica. A team

/" /

,~

" ,,,

..,., •-

~

o \ • • ~\

. ~-~.

"', ~

,, ,

/'~ ....... . •

• \

~ ~ ~__.~-~>~

"

-

•

TZ2-16

_~

-

[

.s~. ,/' ~ : : ~ . < .

from Strasbourg has travelled to the epicentral area in order to install complementary seismic stations and collect other information in collaboration with the Department of Geophysics of the University of Chile.

References Abe, K., 1979. Size of the great earthquakes of 1837-1974 inferred from tsunami data. J. Geophys. Res., 84: 1561-1568. Arabasz, W.J., 1971. Geological and geophysical studies of the Atacama Fault Zone in Northern Chile. Ph.D. Thesis, California Inst. Technol., Pasadena, CA, 264 pp. Armijo, R. and Thiele, R., 1990. Active faulting in northern Chile: ramp stacking and lateral decoupling along a subduction plate boundary? Earth Planet. Sci. Lett., 98: 40-61. Astiz, L., Lay, T. and Kanamori, H., 1988. Large intermediate-depth earthquakes and subduction process. Phys. Earth Planet. Inter., 53: 80-166. Barazangi, M. and lsacks, B., 1976. Spatial distribution of earthquakes and subduction of the Nazca plate beneath South America. Geology, 4: 686-692. Cahill, T. and Isacks, B.L., 1992. Seismicity and shape of the subducted Nazca plate. J. Geophys. Res., 97: 17,503-17,529. Chinn, D.S. and isacks, B.L., 1983. Accurate source depth and focal mechanisms of shallow earthquakes in Western South America and in the New Hebrides island arc. Tectonics, 2(6): 529-563. Chowdhury, D.K. and Whiteman, S.K., 1987. Structure of the Benioff zone under southern Peru to central Chile. Tectonophysics, 134: 215-226. Comte, D. and Pardo, M., 1991. Reappraisal of great historical earthquakes in the northern Chile and southern Peru seismic gaps. Nat. Hazards, 4: 23-44. Comte, D., Pardo, M., Dorbath, L., Dorbath, C., Haessler, H.,

100

B. Delouis et al. / Tectonophysics 259 (1996) 81-100

Rivera, L., Cisternas, A. and Ponce, L., 1994. Determination of seismogenic interplate contact zone and crustal seismicity around Antofagasta, northern Chile, using local data. Geophys. J. Int., 116: 553-561. Delouis, B., Haessler, H., Cisternas, A. and Rivera, L., 1993. Stress tensor determination in France and neighbouring regions. Tectonophysics, 221: 413-437. Dorbath, L., Cistemas, A., Dorbath, C., 1990. quantitative assessment of great earthquakes in Peru. Bull. Seismol. Soc. Am., 80: 551-576. Forsyth, D. and Uyeda S., 1975. On the relative importance of the driving forces of plate motion. Geophys. J. R. Astron. Soc., 43: 163-200. Haak, V. and Giese, P., 1986. Subduction induced petrological processes as inferred from magnetotelluric, seismological and seismic observations in N-Chile and S-Bolivia. Berl. Geowiss. Abh. (A), 66: 231-246. Hervr, M., 1987a. Movimiento sinistral en el Cretacico inferior de la zona de falla Atacama al Norte de Paposo (24°S), Chile. Rev. Geol. Chile, 31: 37-42. Herv~, M., 1987b. Movimiento normal de la falla Paposo, zona de falla Atacama, en el Mioceno, Chile. Rev, Geol. Chile, 31: 31-36. Kausel, E., 1986. Los Terremotos de Agosto de 1868 y Mayo de 1877 que Afectaron el Sur del Peru y Norte de Chile. Bol. Acad. Chil. Cienc., 3: 8-12. Kausel, E. and Campos, J., 1992. The M s = 8 tensional earthquake of 9 December 1950 of Northern Chile and its relation to the seismic potential of the region. Phys. Earth Planet. Inter., 72: 220-235. Kausel, E. and Lomnitz, C., 1968. Tectonics of Chile. PanAmerican Symp. on the upper-Mantle, Mexico, 1969. Kelleher, J., 1972. Rupture zones of large South American earthquakes and some predictions. J. Geophys. Res., 77: 20872103. Kirby, S. and Hacker, B., 1993. Earthquakes at the deep roots of are volcanoes. EOS, 76: 70. Klein, F.W., 1978. Hypocenter location program HYPOINVERSE. U.S. Geol. Surv., Open File Rep. 78-694. Lay, T., Astiz, L., Kanamori, H. and Christensen, D.H., 1989. Temporal variation of large intraplate earthquakes in coupled subduction zones. Phys. Earth Planet. Inter., 54: 258-312. Liu, L-g., 1983. Phase transformations, earthquakes and the descending lithosphere. Phys. Earth Planet. Inter., 32: 226-240. Lomnitz, C., 1970. Major earlhquakes and tsunamis in Chile during the period 1535 to 1955. Geol. Rundsch., 59: 938-960. Malgrange, M. and Madariaga, R., 1983. Complex distribution of large thrust and normal fault earthquakes in the Chilean subduction zone. Geophys. J. R. Astron. Soc., 73: 489-505. McKenzie, D.P., 1969. The relation between fault plane solutions for earthquakes and the directions of the principal stresses. Bull. Seismol. Soc. Am., 59(2): 591-601. McLaren, J.P. and Frohlich, C., 1985. Model calculations of regional network locations for earthquakes in subduction zones. Bull. Seismol. Soc. Am., 75(2): 397-413. Milne, A., 1880. The Peruvian earthquake of 9th May, 1877. Trans. Seismol. Soc. Jpn., 1I: 50-97. Naranjo, J.A., 1987. Interpretation de la actividad Cenozoica

superior a 1o largo de la zona de falla de Atacama, Notre de Chile. Rev. Geol. Chile, 31: 43-55. Naranjo, J.A., Hervr, F., Prieto, X. and Munizaga, F., 1984. Actividad Cretacica de la falla Atacama al Este de Chanaral: milonitizacion y plutonismo. Comun. Univ. Chile, Santiago. Nishenko, S.P., 1985. Seismic potential for large and great interplate earthquakes along the Chilean and Southern Peruvian margins of South America: a quantitative reappraisal. J. Geophys. Res., 90: 3589-3615. Okada, A., 1971. On the neotectonics of the Atacarna fault zone region--Preliminary notes on Late Cenozoic faulting and geomorphic development of the Coast Range of Northem Chile. Bull. Dep. Geogr., Univ. Tokyo, 3: 47-65. Olea, R., 1967. Determinacion de hipocentros y su aplication al estudio del notre de Chile. Mere. Ing. Civil Minas, Univ. Chile, Santiago. Rivera, L. and Cisternas, A., 1990. Stress tensor and fault plane solutions for a population of earthquakes. Bull. Seismol. Soc. Am., 80(3): 600-614. Scheuber, E. and Andriessen, P., 1990. The kinematic and geodynamic significance of the Atacama fault zone, northern Chile. J. Struct. Geol., 12: 243-257. Scholz, C.H., 1989. Mechanics of faulting. Annu. Rev. Earth Planet. Sci., 17: 309-334. Shimamoto, T., 1985. The origin of large or great thrust-type earthquakes along subducting plate boundaries. Tectonophysics, 119: 37-65. Spence, W., 1987. The 1977 Sumba earthquake series: evidence for slab pull force acting at a subduction zone. J. Geophys. Res., 91: 7225-7239. Stauder, W., 1973. Mechanism and spatial distribution of chilean earthquakes with relation to subduction of the oceanic plate. J. Geophys. Res., 78: 5033-5061. Tichelaar, B. and Ruff, L., 1991. Seismic coupling along the Chilean subduction zone. J. Geophys. Res., 96:11,997-12,022. Tichelaar, B. and Ruff, L., 1993. Depth of seismic coupling along subduction zones. J. Geophys. Res., 98: 2017-2037. Thiele, R. and Pincheira, M., 1987. Tectonica transpresiva y movimiento de desgarre en el segmento Sur de la zona de falla Atacama, Chile. Rev. Geol. Chile, 31: 77-94. Van den Beukel, J. and Wortel, R., 1988. Thermo-mechanical modelling of arc-trench regions. Tectonophysics, 154: 177193. Wigger, P., 1988. Seismicity and crustal structure of the central Andes. In: K.J. Reutter, E. Scheuber and P.J. Wigger (Editors), The Southern Central Andes. Contributions to Structure and Evolution of an Active Continental Margin. (Lecture Notes in Earth Sciences.) Springer, Berlin, pp. 209-227. Wigger, P., Araneda, M., Giese, P., Heinsohn, W.-D., RiSwer, P., Schmitz, M. and Viramonte, J., 1991. The crustal structure along the Central Andean transect derived from seismic refraction investigations. In: R. Omarini and H.-J. G/Stze (Compilers), Global Geoscience Transect 6, Central Andean Transect, Nazca Plate to Chaco Plains. ICL and AGU, Washington DC, pp. 13-19. Yoshii, T., 1979. A detailed cross-section of the deep seismic zone beneath Northeastern Honshu, Japan. Tectonophysics, 55: 349-360.