Phys. Chem. Earth, Vol. 21, No. 4, pp.

289-293, 1996 0 1997 Elsevier Science Ltd All rights reserved. Printed in Great Britain 0079- 1946/96 $15.00 + 0.00 PII: SOO79- 1946(97)00050-5

3D-Density Modelling with Seismic Constraints in the Central Andes

A. Kirchner, H.-J. G6tze and M. Schmitz
Institut fi_irGeologie, Geophysik und Geoinforrnatik, _ _ 12249 Berlin, Germany
Received 12 June 1996; accepted 1 October 1996

FU Berlin, Malteserstrasse

74-100,

Abstract. A new 3D-density model of the Central Andean lithosphere and the downgoing Nazca plate has been designed based on more recent constraints given by refraction seismics and seismology. Offset-free modelling is achieved by fitting both gravity and geoid with one density structure and by modelling relative to a reference density structure. Contributions to the Central Andean geoid by some principal parts of the lithosphere are calculated and displayed along a WE trending profile at 20 The distribution s. of the model masses down to mantle regions is used to investigate the isostatic state of the lithosphere model in comparison to Airy and VeningMeinesz isostatic models. 0 1997 Elsevier Science Ltd

64 -71W and 20-290S. Including reprocessed data of Freie Universit& Berlin and older data of South American Universities and oil and mining industry, there is now a data base of about 15.000 gravity values available, which can be used together with other geophysical and geological information for an interdisciplinary interpretation of the structure and evolution of the Central Andes (GGtze et al., 1994). The data base which includes point data and data grids of free-air-, complete Bouguer anomalies (i.e. including topographic reduction), and isostatic residual anomalies, has frequently been described by Giitze et al. (1995) and GGtze and the MIGRA Group (1996).

Introduction

The

Modelling

Method

We present the results of intensive wide angle seismic refraction and gravity research in the southern Central Andes (20 - 29 S) during the 80 and early 90 The s s. gross crustal structures could be determined covering the principal morphostructural units of the Andes between 21 and 25 S (e.g. Wigger et al., 1994). In 1994 (PISCO 94 - Proyecto de Investigacibn Sismol6gica de la Cordillera Occidental) and 1995 (CINCA 95 - Crustal Investigations Off- and On-shore Nazca Central Andes) seismic investigations focussed on the arc and forearc regions in order to provide more detailed insight into the crustal structure of this area. The structural and velocity information obtained by seismic refraction experiments is available along N-S and W-E trending profiles covering the central region of the 3D-gravity model. From 1993 to 1996 the MIGRA group (Medicibnes Internacionales de Gravedad en 10s Andes) with participants from Chile, Argentina and Germany has surveyed some 3.500 new gravity observations in an Andean Geotraverse covering N-Chile and NW-Argentina between
Correspondence to: A. Kirchner

In density foreward modelling we calculate the gravity of plausibly assumed density distributions in order to compare it with a given gravity anomaly. It is our aim to adapt the calculated field to the measured one. Unless we model the entire Earth, there is usually an offset between the measured and the calculated signals and the two fields are often shifted into each other. Various reasons could be held responsible for such an offset, e.g. edge effects of the model, calculation in the plain where model dimensions suggest a spherical calculation, a constant error in used densities or long wavelength gravity anomalies. In order to achieve shift-free modelling, model masses were constructed in an area much larger than the area of interest with a plausible density distribution in the entire model area, not only in the area of interest. The calculation was done on the sphere and the assumed densities are relative to a layered density reference model (2670 kg/m3 from 0 - 15 km, 2900 kg/m3 from 15 - 35 km and 3370 kg/m3 from 35 - 220 km), which is a simplyfied version of the Preliminary Reference Earth Model (Dziewonski and Anderson, 1981).

289

290
W PacificOcean
trench

A. Kirchner et al.
Coastal Cordilh
Anti-

PrecOrd. Preandean Depr.

PelfI

Western C.

Puna

400

500 600 UTM (km) (500 km = 69 W)

Fig. 1. Cross section with seismic discontinuities in northern Chile at lat. 23 305. Crustal structures can be traced from the Coastal Cordillera with increasing thickness towards the Western Cordillera. The shape of the subducted Nazca Plate can be inferred from Moho refractions in the Coastal Cordillera and from the distribution of the earthquake hypocenters derived from the PISCO 94-seismological catalogue (small dots, p. Asch, A. Rudloff, F. Graeber, pers. comm.). Crosses indicate the position of N-S trending refraction profiles; black triangles give the position of the magmatic arc.

Additionally, we calculated the geoid undulations of the density model used to fit the gravity field and compared it with the observed geoid (Rapp et al., 1992, OSUSlA, ) in order to interprete long wavelength gravity anomalies in the Central Andes.

in the Pacific to a 50 m high between 16 and 19 pasS sing a relative minimum above the trench and declining to 20 m in the Chaco plain. There is an isolated maximum of 45 m in the volcanic arc at 27 (Rapp et al., S 1992).

Gravity

Anomalies

and

the Geoid

4
4.1

The

3D Density

Model

and its Constraints

Onshore the Bouguer anomaly drops down to a regional minimum of about -450 mGa1 in the area of the recent volcanic arc, related to crustal thickening. The gravity effect of an isostatic compensation root was eliminated from the Bouguer gravity and the resulting anomaly serves as a residual field. The most interesting features of this field are: (1) Positive values in the area of the Jurassic forearc with isolated complexes parallel to the coastline. They are regionally caused by the presence of the dense subducting plate and locally by uplifted jurassic batholiths intruded into the volcanic La Negra formation. (2) The NNW-SSE striking positive anomaly from Calama by the Salar de Atacama to southern Puna which can be explained by a highly metamorphic and high-density Paleozoic/Precambrian structure, which is oblique to the generally N-S trending recent volcanic belt. (3) Local minima along the recent volcanic arc point to partly molten material at depths of 15-20 km. (4) Minima following a line from Ollague to Calama along 69W are caused by the Eocene volcanic arc with low-density volcanic material in the upper crust. Maps are provided by Giitze and the MIGRA Group (1996). The maximum-typegeoid in the Central Andes mainly reflects the topographic edifice. It rises from about 20 m

Forearc Crustal Data

Structures

from Seismic Refaction

In the Coastal Cordillera between 21 and 24 upper S, crustal P-wave velocities vary between 6.0 and 6.7 km/s. The top of the lower crust is located between 7 and 12 km depth, reaching down to 20-22 km, proven by prograde phases with velocities up to 7.2 km/s. The prominent discontinuity at 40 km depth (V, = 8.2 km/s) is interpreted as a Moho surface of the subducted Nazca Plate with an average velocity of 6.6-6.7 km/s (Wigger et al., 1994). Upper and lower crustal structures can be followed into the Precordillera with the upper crust increasing to 20 km thickness and the lower crust extending down to 40-45 km depth. Three distinct discontinuities with velocities from 7.3-8.0 km/s, seperated by low velocity layers, are located between 50 and 70 km depth. The average velocity down to 65-70 km decreases to 6.3 km/s. Towards the Western Cordillera the average crustal velocity once again decreases (5.9-6.0 km/s), whereas the upper crust and deeper crustal parts continue. The most striking feature in the record sections of the Western Cordillera, a discontinuity at 20 km depth (V, = 6.4 km/s) beneath Ollague, is interpreted from a structural point of view as the top of the lower crust, although

3D-Density

Modelling

with Seismic Constraints .____

Solid line Dashed Measured Bauguer Anomaly line, Calculated Gravity

291

Antotagasta

Peine

2450

560
UTM

660

(km) (500 km = 69OW)

Fig. 2. Cross section of the 3D density model at lat. 23 305 including subducting Nazca Plate with earthquake hypocenters, Andean crust, continental lithosphere and an asthenospheric wedge. Black triangles indicate the magmatic arc, black dashes a Vening-Meinesz isostatic Moho (D=35 km, 6p = 350 kg/m 3, R=10z3 Nm). According to 2D raytracing models there is material interpreted as serpentinized old mantle with densities lower as the high density matrial in the Coastal Cordillera located above. Average crustal density decreases towards the Western Cordillera including two low density zones beneath the recent arc. The transition of oceanic crust basalts to eclogite with a density of 3500kg/ m3 is completed at a depth of 60-70 km.

the crustal units below are characterised by partly reduced velocities. There are locally limited high velocity discontinuities observed in this crustal unit down to 4550 km depth. Assuming this depth range as base of the lower crust, a 25-30 km thick lower crust results with low average velocities of only 6.0 to 6.2 km/s. Fan observations between the Precordillera and the Western Cordillera give indications for a deep reflector (V, = 8.2 km/s) interpreted as the eclogitized top of the subducted Nazca Plate (see Fig. 1). 4.2 3D Modelling

A large scaled 3D density mode1 was constructed to investigate the regional structure and density distribution of the Andean lithosphere. The modelled area reaches from 12 to 35 and from 57 to 79 5 W comprising large parts of both the Nazca Plate in the West and the Brasilian Shield east of the Andean orogen as a reference lithosphere. The principal parts of the mode1 are the downgoing Nazca Plate, the Andean crust, the continental lithosphere with Brasilian shield crust and an astenospheric wedge. The mode1 resembles the new results of refraction seismics based on two WE and two NS trending ray tracing models at 24 and 23.5 and 68 and 69 S W, respectively, as described above ( 4.1 and Fig. 1). Ad-

ditionally, a balanced cross section at 21 according to S earlier refraction seismics (Schmitz, 1994) was also included. As a preliminary study the velocities of seismic models were directly converted into densities by using e.g. the Nafe & Drake relationship or similar velocitydensity relations. Initial mode1 geometry was slightly modified by the use of interactive computer graphics (S. Schmidt, pers. comm.) to verify regional trends implied by the Bouguer gravity field and the topographyreduced geoid. Fig. 2 shows the major crustal parts of the density model at 23 s. 30 In the forearc a 10 km thick upper crust with a density of 2850 kg/m3 tops a 20 km thick high-density lower crust material (2950-2930 kg/m3), followed by parts with reduced density (2930-2900 kg/m3) which are interpreted as serpentinized Pre-Andean mantle material. In the Precordillera the upper crust thickens to about 20 km with a decrease of density to 2770 kg/m3. It can be followed to the recent arc with the same thickness and further decreasing density (2730 kg/m3). Towards the back arc the upper crust base rises to 15 km and its density increases to Precordilleran values. The Coastal Cordilleran high-density material can also be pursued in WE direction with decreasing densities down to 2860 kg/m3 and its base reaching 40 km as it passes beneath the recent ,arc. There is a low-density zone of 2700 kg/m3 underneath the volcanic chain between 20

292

A. Kirchner et al. W
300.0

Trench

A- A

observed

Top/eat (a)

D -aAiry Moho (b) * * L.owcrcrust (c) o. * Slab(d) *r * Wedge(e)

-250.0

L -5Cil.0

0.0

500.0 looo.o uTM(km)(sc0kJn=69w)

1500.0

400.0

l(

11cQ.o 600.0 UTM (km) (500 km = 69W)

Fig. 3. Estimations of possible contributions to the geoid and the gravity field visualized along a WE trending profile at 20 of (1) the 5 topography (2670kg/m3) and bathymetry (-1640 kg/m3), (2 a) an Airy isostatic root (D=40 km, 6p = 300 kg/m3), (2b) the model s lower crust (6~ varying from -470 to -370 kolm31. (3) the subducting Nszca Plate with a mean density surplus of 20 kg/m3 and (4) the astheno~phericwe~ge with a lack of -10 cg/m j..

and 35 km depth resembling the low-velocity zone detected by refraction seismics. Lower crustal bodies are modelled with a density of 2980-3000 kg/m3 below 40 km placing a gravimetric moho below the arc and Altiplano/Puna to a depth ranging from 56 km at 24 to 65 km at 21 according S S to 2D ray tracing models. A seismically implied second low-density zone is modelled between 44 and 50 km with a density of 2920 kg/m3. The Brasilian shield crust consists of a 4-6 km thick sedimentary cover (2450 kg/m followed by a mid cru) stal layer with 2880 kg/m3 and a lower crust down to 40 km with a thickness of 10 km and a density of 3000 kg/m3. The shield crust lies on top of the continental lithosphere modelled with a density of 3370 kg/m3. The shape of the subducting Nazca Plate is inferred from results of Cahill et al. (1992) and the earthquake hypocenters of the PISCO 94-seismic catalogue (G. Asch, A. Rudloff, F. Graeber, pers. comm.)-. For definition of the descending Moho angle in the forearc we have observations by refraction seismics between 21 and 24 The transition of oceanic crust to a 3500 kg/m3 s. eclogite layer on top of the downgoing slab, which starts at about 50 km and is completed at 70 km depth, is simulated by a model body with a density of 3150 kg/m3 of oceanic crustal thickness within this depth range. In the model we present the subducting Nazca Plate to be modelled down to 670 km with a mean density contrast to the surrounding asthenosphere of 20 kg/m3. The model also contains an asthenospheric wedge with a density of 3360 kg/m3 filling the space between the slab and the shield lithosphere. According to studies on Andean magmatism (Kay et al., 1994) and P-wave attenuation (Whitman et al., 1992) it reaches the crustal root beneath the recent arc with an alongstrike change of geometry.

Possible Contributions to the Geoid at 20 S

The SD-density model with fitted geoid and Bouguer gravity was used to estimate possible contributions to the geoid and the gravity field of the topography and bathymetry, an Airy isostatic root, the model lower s crust, the slab and the asthenospheric wedge. To visualize these contributions (see Fig. 3) we chose an exemplary WE trending profile at 20 where the geoid and S the Bouguer Anomaly reach 46 m and -365 mGa1 respectively. In order to obtain a geoid signal comparable to the Bouguer anomaly we calculated within the entire model area the contribution of the continent topographic s masses above sea level with a density of 2670 kg/m3 and of the ocean missing masses between sea level and s ocean floor with a density of -1640 kg/m3. The calculated contribution which accounts for a maximum of 148 geoid meters at 20 was subtracted from OSUSlA. S An Airy isostatic root was determined with a normal crustal thickness of 40 km and a crust mantle density contrast of 300 kg/m3 within the entire model area. The calculated geoid contribution has a minimumof -124 m and mirrors the signal due to topography and bathymetry. The gravity effect drops to -365 mGa1. Alternatively we regarded the effects on geoid and gravity field of the model lower crustal parts. The dens sities again are taken relative to the reference value of 3370 kg/m3 and vary from -370 kg/m3 to -470 kg/m3. Fig. 3 shows a geoid curve (minimum of -212 m) being considerably lower than the Airy Moho effect, whereas s the gravity contribution of a -295 mGa1 low is a little higher than that of the Airy Moho. It should be noted that surplus masses due to the negative root in ocean areas are not taken into account here. The subducting Nazca Plate is modelled between the trench and a depth of 670 km with a mean positive den-

3D-Density Modelling sity contrast of 20 kg/m3 apart from a 5-6 km thick eclogite layer with a density difference of +130 kg/m3 between 70 km and 220 km and a metastable olivine wedge reaching into the spine1 zone with -100 kg/m3 from 400 km to 600 km. This slab structure contributes 47 m to the geoid and 60 mGa1 to the Bouguer anomaly. The asthenospheric wedge described in section 4.2, modelled with a density contrast of -10 kg/m3, accounts for -7 geoid meters and -13 mGa1. Discussion of the geoid is carried out at 20 because S the contribution from the Andean topography and the Nazca plate are largest at this latitude.

with Seismic Constraints

293

lower than the observed signals. This could be explained by the fact that the top of the modelled slab is situated too low thus leaving too much room for light weighted crustal material. A slab subduction at a lower angle is suggested in this zone.
Acknowledgements.
discussion and to G. the manuscript. Thanks Jentzsch are due U. to K. Lessel and for providing for reviewing

and

Casten

References
Cahill, A., T., Isacks, B. L., Whitman, and tectonics 1 f(5), D., Chat&in, in Jujuy 944-959, D. Province, 1992. L., Preliminary reference J.-L., and Perez, Seismicity Northwestern

Argentina,

Tectonics, A. M. and

Conclusions

Dziewonski, Earth G&e,

Anderson,

A three dimensional density model in the Central Andes explains the observed geoid and gravity without offset between the calculated and measured signals. Except the downgoing slab the model does not take into account other density inhomogeneities below 220 km. The density distribution in forearc and arc regions was calibrated to the generalized geometry of 2D ray tracing models. Crustal densities were preliminarily determined according to common velocity-density relations and slightly changed under the need to fit curves with fixed geometry. The model crust-mantle s boundary, deepest below the recent arc and the Altiplano-Puna zone, shows an alongstrike change in depth ranging from 56 km at 24 S to 65 km at 20-21%. Density contrasts at the gravimetric Moho vary between 360-370 kg/m3 in arc and backarc areas and 450 kg/m3 in the Coastal Cordillera due to low density bodies directly above the subducting slab. A Vening-Meinesz isostatic Moho derived from topographic loads with a normal crustal thickness of 35 km, a crust-mantle density contrast of 350 kg/m3 and a lithospheric rigidity of 1O23 Nm is situated well below the model crust-mantle boundary in the south at about 23 s (compare Fig. 2), but coincides well with 30 it at 21 This possibly reflects concepts of lithospheric s. delamination south of 23 (e.g. Kay et al., 1994). S The isostatic state of the model, derived by locally comparing the density distribution relative to the reference model with the topographic load, shows a significant overcompensation (mass deficit below sea level) in the forearc between W and 23s. This coincides with S a region where modelled gravity and geoid trend a little

model, H.-J.

Phys.

Earth.

Plan.
Group,

Int.,

25, 297-356,1981.
updates 1996. S., The ligravity da-

and the MIGRA Andes, B.,

Group

tabase G&e,

for Central

EOS

Transactions,
S., Andes gravity,

H.-J.,

Labmeyer, structure interpretation

Schmidt,

and Strunk, in

thospheric red from

of the Central of regional edited

(20 - 26 S) as infer-

Tectonics

of the

Southern
and 1994. G&e, H.-J.,

Central

Andes,
pp. 7-21,

by K.-J.

Reutter, Berlin

E. Scheuber, Heidelberg,

P. Wigger,

Springer-Verlag,

Schmitz, G., Araneda,

M.,

Giese, M.,

P.,

Schmidt, D, G., and

S.,

Wigger, australes

P., J., ba-

Schwarz,

Chong

Viramonte,

Las estructuras

litosfhricas

de 10s andes

centrales

sados eninterpretacionesgeofisicas,

Rev&a

Geoldgica

de Chile,
back Andes, arc J.

B(2),
Kay, volcanic nation

171-192,199s. Coira, rocks beneath B., and Viramonte, Puna J., Young Plateau, m&c

S. M.,

as indicators the Argentine 99(B12),

of continental

lithospheric Central

delami-

Geophys.
Rapp, vity

Res.,

24323-24339,1994. Y. and Gravity 0. M., High resolution data, in gra-

R. H., Pavlis, models

N. K., and Wang, terrestrial Charting edited by

combining

satellite

Mars borne
1992. Schmitz,

to Greenland: Instruments,
M.,

With

Space

From and Air-

Colombo,

Springer-Verlag,

A balanced 13(2),

model

of the

southern

Central

Andes,

Tectonics, Whitman, rez, A., Central

484-492,

1994. J.-L., Chiu, J.-M., and Pethe 19929-

D., Isacks, Attenuationof Andean

B. L., Chatelain, high-frequency J.

seismic

.saves seneath 97(B13),

Plateau,

Geophys.

Res.,

19947,1992. Wigger, Giese, P. J., Schmitz, P., Heinsohn, Central in M., Araneda, M., Asch, E., G.,

Baldzuhn, S.,
E., RGwer, of the inve-

W.-D.,

Martinez, from and

Rica%,

P., and Viramonte, southern stigations, ted by K.-J.

J., Variation Andes deduced of

in the crustal seismic

structure refraction

Tectonics Reutter, Berlin

the Southern
1994.

Central
P. Wigger,

Andes,
pp.

edi23-48,

E. Scheuber, Heidelberg,

Springer-Verlag,