Seismogenic Low-Angle Normal Faults: Observations of Earthquakes

Evidence for seismogenic normal faults at shallow dips in contnental rifts

Geoffrey A. Abers Department of Earth Sciences, 685 Commonwealth Ave., Boston University, Massachusetts, USA, 02215 email: [email protected] Phone: 1-617-353-2616 Fax: 1-617-363-3290

In: Wilson, R.C.L., Whitmarsh, R.B., Taylor, B, & Froitzheim, N (eds): Non-volcanic rifting of continental margins: a comparison of evidence from land and sea; for Geological Society Lond. Spec. Publ., Nonvolcanic Rifted Margins, in press 2000.

Number of words of text : 7794 (incl. references and figure captions) Number of References: 60 Number of Figures: 9 Number of Tables: 1 Abbreviated Title: Dips of seismogenic normal faults

Abstract.

Several recent observations indicate that normal faulting earthquakes occasionally occur on faults dipping < 35°, dips often considered shallow. Most of these occur in the Woodlark and Aegean rifts. The Woodlark and Aegean rifts are found to generate significantly more earthquakes than others, and are the most rapidly extending, so display the widest variety of fault behavior. Even within the Woodlark rift system extension rates vary along strike, with the shallowest-dipping faults confined to the most rapidly rifting segment. Here, several events (MW 6.0 – 6.8) feature nodal planes dipping 23-35°. These planes are subparallel to shear zones bounding nearby metamorphic core complexes, including one imaged to 8-9 km depth by seismic reflection profiling. In the western Gulf of Corinth at least one large event (Mw 6.4) occurred on a fault dipping ~33°. As the Woodlark example, this rift segment exhibits a high opening rate (10-20 mm/yr). Several other cases elsewhere, based on older historic data, microseismicity, or geologic inference suggest seismic slip at similar or shallower dips. Still, no documented large earthquake exhibits seismic slip on sub-horizontal surfaces (dip <10-15°). Stress rotation may explain the 23-35° dips, but thus far no realistic mechanism has been found. More likely, these faults represent surfaces somewhat weaker than surrounding rock, through some combination of modest cohesion of the surrounding rock and slightly lower frictional coefficients on the fault. Such weakening may be a consequence of high slip rates, which rapidly generate large offsets and mature fault systems.

Abers: Dips of seismogenic normal faults 12/15/00 p. 1

Although fault mechanics appears to preclude brittle 35°, so models of fault generation requiring brittle slip at slip on low-angle normal faults, dipping less than 30° (e.g. dips of > 45° require revision. The 23-35° dips are consis- Anderson 1951), several structures found in the geologic tent with frictional properties expected for mature faults record appear to suggest such slip is possible (e.g. Wer- with well-developed gouge zones (e.g. Byerlee & Savage nicke 1995). The seismic record, until recently, was absent 1992), and unusual mechanics may not be needed. Second, of any well-documented normal-faulting earthquakes with at present, no large earthquake has been found showing slip such dips (Jackson & White 1989), in part prompting the on a fault unequivocally dipping less than 20°, although development of theories for generating low-angle faults such slip can be found in microseismicity. However, the without requiring brittle slip at those dips (e.g. Buck 1988; record of reliable focal mechanisms is no more than 35 Wernicke & Axen 1988). In many of these theoretical years long, much less than large-earthquake repeat times, studies, dips of the active portion of normal faults are lim- even in the few rifts where 25-35° dips are found. Thus, ited to > 45° (e.g. Lavier et al. 1999). However, in recent the presence of low-angle normal faults in the geologic years several examples of normal faulting earthquakes have record remains enigmatic – we cannot yet conclude been documented for which dips less than 35° have been whether shallower-dipping faults are truly aseismic, or claimed, some of which may show dips < 30° (Abers 1991; merely occur infrequently. Abers et al. 1997; Bernard et al. 1997; Doser 1987; John- The use of the term “low-angle” varies considerably; it son & Loy 1992; Miller & John 1999; Rietbrock et al. most often refers to normal faults dipping less than 30° but 1996; Rigo et al. 1996). This evidence has been used as is sometimes applied to a wider range of dips. Although the demonstration that low-angle normal faulting can occur 30° dip threshold has mechanical significance, it lies in the seismically. middle of the uncertainty range of dips for many of the The core of this paper is a review of these observations earthquake-generating faults discussed here, so use of the and the settings in which they occur, with focus on the term “low-angle” may confuse. The events discussed here Woodlark Rift of Papua New Guinea. A global analysis of are interesting because they show faulting at dips close to seismicity data shows that background seismicity rates in or less than 30° or because they have been claimed to be rifts correlate well with extension rates, so it should come “low-angle”; in any case, the faults slip despite the rela- as no surprise that most evidence for these “unusual” earth- tively high shear stresses required for neighboring rock. quakes comes from those rifts most rapidly extending An alternative term is needed to describe these earthquakes. (Woodlark and the Aegean). Such a correlation may reflect Rather than use the term “low-angle”, I will describe these sampling bias (Wernicke 1995), or may reflect more favor- fault geometries as “shallow-dipping”, referring to specific able conditions (fault weakening, block rotation, etc.) at dip ranges wherever possible, and hopefully will offend high extension rates. The focal mechanisms for the poten- fewer readers. tial low-angle earthquakes reveal two things. First, several large earthquakes clearly rupture fault planes dipping 23-

Table 1. Active Continental Rift Systems Opening Maximum Extrapolated number of events N with Length Width, Rate, V Earthquake M > 4 per year, per Rift L, km km mm/yr Depth H, km L (×10-5 m-1 yr-1) LVH (× 10-7 m-3) Woodlark 600 100 25-40 7-9 6.8 ± 0.9 2.6 ± 0.7 Aegean 800 500 35 10-15 12.8 ± 0.2 2.9 ± 0.5 Basin-Range 1200 600 10 15 1.7 ± 0.1 1.2 ± 0.3 Red Sea/Suez 2200 250 7-12 5-10 1.2 ± 0.1 1.7 ± 0.6 Baikal 1500 100 4.5 25 1.4 ± 0.2 1.2 ± 0.7 East Africa 3000 100 <3 35 2.0 ± 0.1 3.8 ± 3.1 Rhine 400 100 0.5-1 20 0.09 0.6 ± 0.3 RioGrande 1000 100 <3 15 0.1 ± 0.1 0.5 ± 0.4 Descriptive data from Ruppel (1995) and other sources as discussed in text. Opening rates and lengths refer to continental segments only. Annual earthquake rates extrapolated from maximum-likelihood regressions (Figure 1), normalized to along-strike length of rift, or to product of L, seismogenic zone depth H, and opening rate V. Uncertainties are 1-σ, for N from regression; propagated uncertainties in V and H assumed uniformly distributed at 5 mm/yr and 2.5 km, respectively, unless range is shown. Seismicity data for Rhine Graben from Ahorner (1983), ISC catalog elsewhere.

rifts to known extension velocities shows good correlation Seismicity in Rifts (Figure 1, Table 1). Other compilations have primarily concentrated on the largest earthquakes by cataloging fault As in other tectonic settings, seismicity in rifts is con- plane solutions (e.g. Jackson & White 1989) or cumulative trolled by a combination of plate motion rates, seismic seismic moment (Jackson & McKenzie 1988), and so pro- coupling, fault geometry, and the short duration of instru- vide useful insights into extension rates, as the largest mental records of earthquakes. To illustrate this, a com- earthquakes constitute the majority of the extension ac- parison of the seismicity rates for several of the world's

Abers: Dips of seismogenic normal faults 12/15/00 p. 2

commodated by earthquakes. However, such studies are often hampered by the relatively short period of instrumen- Table 1 compares seismicity rates with rift opening tal recording compared to the recurrence time of most rates for the several rifts studied; Ruppel (1995) discusses faults, making the present comparison a useful comple- other parameters. Seismicity rates calculated here are ment. based on seismic moment estimates derived from standard catalogs (Figure 1), normalized by the along-strike length

100 L= 600 km 100 L= 800 km of the rifts. These data provide both seismicity rates as a function of moment-magnitude and estimates of detection 10 10 thresholds in different rift. Again, the seismicity rates do not provide estimates of overall extension rates in a region, 1 1 which is dominated by the occurrence of infrequent, large earthquakes, but describe the background seismicity that 0.1 0.1 Woodlark-W. 5.92 - 1.08 M Aegean 6.06 - 1.01 M makes up most of the earthquake record. Rift opening rates are compiled from a number of 3 4 5 6 7 3 4 5 6 7 sources, with preference given to geodetic rates in the Ba- 100 L= 1200 km 100 RioGrande 3.84 - 0.95 M L= 1000 km sin and Range (Dixon et al. 1995), Aegean (Le Pichon et al. 1995), the Rio Grande (Savage et al. 1985), Baikal 10 10 (Calais et al. 1998), and the Rhine Graben (Ahorner 1983).

1 1 Elsewhere, opening rates are based on magnetic lineations in the Woodlark Rift (Taylor et al. 1995) and the Red 0.1 0.1 Sea/Gulf of Suez. Finally, an upper limit of extension rates Basin+Range 4.92 - 0.91 M for East Africa can be gained by closure of global plate 3 4 5 6 7 3 4 5 6 7 velocity circuits, following DeMets et al. (1990). Uncer- 100 L= 3000 km 100 L= 2200 km tainties typically lie in the range of a few mm/yr, so differ- ences between fast and slow rifts are significant. 10 10 These compilations illustrate several trends in the rift- seismicity record. First, seismicity rates correlate well to 1 1 plate opening rates. When normalized to extension rates (V) and depth of seismogenic zone (H), seismicity rates show 0.1 0.1 E.Africa 6.04 - 1.07 M Red Sea 5.56 - 1.03 M remarkable constancy (Table 1). This suggests that large

3 4 5 6 7 3 4 5 6 7 variations in seismic coupling sometimes inferred from seismic moment release rate may be partly attributable to 100 L= 1500 km 100 Rhine 3.38 - 1.01 M L* = 250 km the short historical record for large earthquakes, as may be

10 10 true for subduction zones (McCaffrey 1997). Second, detection of earthquakes is poor in many but not all rifts. For ) per year 1 1 example, the Woodlark Rift shows evidence of incomplete Mw sampling at M < 5.2 and almost no earthquakes in the

(> W

N 0.1 0.1 Baikal 5.80 - 1.11 M ISC catalog with MW < 4.8, while earthquakes as small as 3 4 5 6 7 3 4 5 6 7 MW =3 .0 are routinely reported for the Aegean and Basin Mw Mw and Range. Fig. 1. Seismicity rates from the world’s active continen- After correcting for sampling, two rifts stand out as tal rifts (Table 1). Plots show number of events (N) larger having high seismicity, the Woodlark and Aegean (Table than a given moment magnitude (MW), from a combination of ISC and Harvard CMT (HCMT) catalog, spanning 1). These are the two most likely places to see rift-related 1964-1995 and 1977-1995 respectively. Within each re- earthquakes, and they are also regions where extensional gion normal faulting predominates. Magnitudes are taken geometries that accommodate large amounts of extension (in order of preference) from the HCMT catalog, MS and are favored. mb values from the ISC catalog. MS and mb are converted to MW following Ekström & Dziewonski (1986) and Nuttli (1985), respectively. Lines show fits to log (N) = a – bMW Large Earthquakes from a maximum-likelihood regression (Weichert 1980) that determines a and b. Minimum magnitudes for fits (inverted triangles) are those below which estimates of b The Woodlark Rift, Papua New Guinea become unstable and low. In all regions, the maximum Setting. The Woodlark Rift (Figure 2) shows a full possible magnitude is assumed to be 7.0, although varia- transition from sea-floor spreading in the east to extension tion of this parameter by one MW unit changed results in- significantly. Data and regression result for the Rhine in quasi-continental crust farther west (e.g. Mutter et al. graben are taken from Ahorner (1983), utilizing his ML - 1996; Taylor et al. 1999). The continental section features MW relation, rather than the insufficient ISC catalog; “L*” development of several metamorphic core complexes along gives rift length covered by Ahorner (1983). These regres- the rift (Davies & Warren 1988). Extension takes place sions are used to predict the number of events with MW > about a pole somewhere west of 147°E, with rates in the 4.0 on Table 1, after normalizing to an along-strike rift continental section ranging from 10 to 40 mm/yr (Taylor et length (L). Note the wide range in both seismicity levels al. 1999). The easternmost extensional feature, Moresby and detection. “Seamount”, lies immediately west of the oceanic rift tip

Abers: Dips of seismogenic normal faults 12/15/00 p. 3

and also shows a north-dipping master fault zone that ex- north and appear to lie in a single extensional domain. poses basement rocks. The westernmost identified core Subordinate deformation, reflected in seismicity and sub- complex, the Dayman Dome, lies on the Papuan Peninsula surface faults south of the D’Entrecasteaux Islands, sug- near 149°E longitude (Davies & Warren 1988) and appears gests that some deformation may occur there west of to be active (Ollier & Pain 1980). Although the fault sys- 150.5°E (e.g., Mutter et al. 1996); its kinematic relationship tems connecting these structures are poorly known except to the main deformation zone is unclear. by seismicity, all show bounding shear zones dipping to the

New Sea Floor/Rift Solomon Sea Earthquakes MCC-bounding 8ºS fault Papuan Trobriand Peninsula Platform D'En trec Is. asteaux G F D M N 10º Woodlark Basin

Coral Sea km 0 100 200

148ºS 150º 152º 154º Fig. 2. Tectonic setting and seismicity of the Woodlark Rift, relocated, from Abers et al. (1997). Large circles: events for which waveform-derived focal mechanisms are available; small circles: others. The D’Entrecasteaux Islands are Goodenough, Fergusson, and Normanby, labeled G, F, and N respectively. Other labels, M: Moresby “Seamount”; D: Dayman Dome, a core complex on the Papuan Peninsula. Thick black lines show spreading center in eastern oceanic half of the rift (Taylor et al. 1999), or faults bounding core complexes in western half with barbs on hanging walls (Davies & Warren 1988). Dashed line surrounding white region denotes new (< 5 Ma) sea floor.

The large core complexes of the D'Entrecasteaux Is- D'Entrecasteaux Islands core complexes dip consistently to lands (Figure 2) appear very young, and likely are still in the north at 20-25°, although some evidence of more com- the process of being exhumed. Despite tropical erosion plex faulting patterns exists (Hill et al. 1992). conditions, the bounding shear zones still control the Earthquake locations. Seismicity has been reproc- shapes of land surfaces and in many places appear fresh essed and relocated utilizing three-dimensional velocity near their base, which along with other geomorphic obser- models and fully nonlinear relocation procedures, as re- vations suggest continued activity (Ollier & Pain 1980). ported elsewhere (Abers et al. 1997). The arrival times Their footwall rocks show some of the most recent rapid constraining the locations largely come from teleseismic uplift known, with footwall rocks experiencing 700-900°C catalogs, limiting earthquakes to magnitudes greater than and 5-6 kbar conditions as recently as 3-4 Ma, and 4-5 kbar ~4.8 (Figure 1). The resulting epicenters (Figure 2) show at similar temperatures at 1.5 – 2 Ma (Baldwin et al. 1993; seismicity to be localized in a ~25 km wide band and con- Hill & Baldwin 1993), requiring unroofing from midcrustal fined to the north side of the rift zone near 9.5°S, at least depths within the last 2 Ma. Late cooling recorded by apa- east of 150°E. Most of these epicenters lie near the major tite fission track ages on shear-zone rocks occurred at 0.4- fault zones bounding the core complexes of the Fergusson 1.0 Ma (Baldwin et al. 1993). Although such dates reveal and Goodenough Islands and the Dayman Dome, and lie little about the last 0.5 Ma history, the continued structural just north of Moresby Seamount (e.g. Mutter et al. 1996; control of land surface morphology suggests that the rapid Taylor et al. 1999). Hence, much of the seismicity appears Quaternary exhumation continues through the present (e.g. to follow a contiguous system of faults along strike, and is Senior & Billington 1987). generally consistent with movement on faults that lie on the The very rapid unroofing of these rocks correlates well north side of the core complexes. Such a pattern would be with the rapid extension here, which ranges from 20 to 40 expected were the faults responsible for core complex ex- mm/yr across the core complex belt over the ~6 Ma history humation presently active of extension. The dominant shear zones bounding the

Abers: Dips of seismogenic normal faults 12/15/00 p. 4

-7º 13 2. 5. C6. 47 C1. 31 C5. 36 1. 3. C11. 35 -8º 23 14 8. 43 32

CMT Fig. 4 G Fig. 5 2 event 15 23 BW shallower M dip

6. 7. 16 C4. 44 C12. 51 C15 -11º 11 10 9. 4 38 40 26

147º 148º 149º 150º 151º 152º 153º

Fig. 3. Focal mechanisms of the Woodlark Rift determined from waveform inversion. Earthquake locations same as Figure 2. Focal mechanisms are lower-hemisphere equal-area projections; (black) from body-wave studies (Abers 1991; Abers et al. 1997); (gray) from Harvard CMT catalog (Dziewonski et al. 1981). Event numbers above focal mechanisms correspond to Table 1 of Abers et al. (1997), as do dips of shallower planes labeled inside dilatational quadrants for normal faults.. Locations of cross-sections (Fig. 4, 5) shown. Other labels, G: Gameta; M: Moresby Seamount.

Earthquake fault planes. First-motion focal mecha- Papuan Peninsula rift segment (147°-150.3°E) all dip be- nisms (Ripper 1982; Weissel et al. 1982) document the tween 38° and 51°. This variation correlates with extension extensional nature of the Woodlark-D'Entrecasteaux Is- rates: rates are 20-40 mm/yr between 150.5°E and 153°E, lands region. These methods can have relatively large un- and lower to the west. In at least two of the eastern cases, certainties, as the ray geometries are rarely ideal. Studies discussed below, there is good evidence that the shallower of Woodlark Rift earthquakes based on waveform inversion plane is the fault plane. Hence, the tendency toward shal- provide much clearer constraints on the geometry of fault- low-dipping fault planes appears to correlate with extension ing, and are discussed here (Abers 1991; Abers et al. 1997). rate. Overall, focal mechanisms reveal north-south extension 1985 Gameta Earthquake. The largest event (MW = (Figure 3), consistent with magnetic lineations (Taylor et 6.8; #1 on Figure 3) occurred 29 October 1985, 10-40 km al. 1995; 1999; Weissel et al. 1982). Normal faulting ex- east of the eastern core complex on Fergusson Island as tends west to 148°W, the central Papuan Peninsula, determined by relocation, directly along-strike from the throughout the region of metamorphic core complex devel- dominant shear zone that bounds the Oiatabu core complex opment; the pole of opening must lie farther to the west. of northeast Fergusson Island (Abers 1991). The focal Waveform modeling shows source depths consistently less mechanism shows one plane dipping NNW at 23°; wave- than 8-9 km below sea level (Abers et al. 1997); this cut-off forms rule out dips for that plane as steep as 30° (Abers depth is somewhat shallower than seen at other continental 1991). Unfortunately, no direct method exists to choose rifts (e.g. Jackson & White 1989) and is consistent with the fault plane from among the two nodal planes, but com- elevated temperatures during rapid rifting. parison with the nearby shear zone favors the shallow- The most unusual aspect of faulting here is the abun- dipping plane. First, a north-dipping plane would be ex- dance of normal faulting mechanisms with dips of 20-35° pected were the shear-zone bounding faults active in their near the oceanic rift tip (151.7°E; Figure 3). Between present orientation. The shear zone dips 20-25° to the 150.5°E and 153°E, the dips of the shallower nodal planes north. Second, the north-dipping plane is consistent with for normal-faulting earthquakes all lie between 23° and uplift of the core complex while the south dipping plane 36°. By contrast, shallow planes for earthquakes in the suggests that the island is submerging (Figure 4). The for-

Abers: Dips of seismogenic normal faults 12/15/00 p. 5

mer scenario is preferred, since the core complexes appear Moresby Seamount Earthquakes. A normal fault to be actively exhuming, as discussed above. The alterna- dipping ~30° seems likely along the north-dipping fault that tive, that major faulting occurs today on south-dipping bounds Moresby Seamount, the easternmost active fault faults, suggests a radical (and ad hoc) termination to several block (Abers et al. 1997; Mutter et al. 1996; Taylor et al. million years of core complex exhumation in the last 0.4 1995). Here, several earthquakes show normal-faulting Ma, while the geomorphic expression (e.g. Ollier & Pain focal mechanisms with one plane dipping near 30° (Figure 1980) continues to show core complex exhumation. 3). Multi-channel seismic lines reveal the presence of a planar north-dipping surface here that appears to be a downdip extension of the northern flank of the seamount. SNFrom an on-axis line this feature’s dip is constrained via dip-moveout techniques to lie between 25° and 35° (Abers MCC et al. 1997), and parallel off-axis seismic lines have been used to infer dips closer to 25° (Taylor et al. 1995). One A significant earthquake, the closest to the central seismic profile, shows a north dipping fault plane dipping 25-35° or nearly coincident with the imaged surface (Figure 5). This earthquake suggests that the feature imaged on the seismic line is a fault, and one that exhibits brittle failure (i.e. earth- MCC quakes) along its surface. This interpretation was confirmed by ODP Leg 180 drilling, which sampled the B exposed fault surface at Site 1117 (arrow on Figure 5).

Undeformed rock (gabbro) at ~100 m below sea floor Fig. 4. Illustration of the two possible fault planes for grades upward toward the fault surface into increasingly the 1985 Gameta earthquake, and the uplift pattern pre- sheared breccias and mylonites; the upper 4 meters below dicted by each. Only the north-dipping plane is consis- sea floor are interpreted as fault gouge of brittle origin tent with the ongoing exhumation of the Oiatabu core (Shipboard Scientific Party 1999). complex, which lies immediately south of the earthquake.

0 S N 2 Moresby 4 Seamount 6

Depth, km 8 ? 10 v.e. 1:1 151.7º E 12 Fig. 5. Interpretation of seismic line L1218 across Moresby Seamount, and focal mechanism of nearest earthquake (6 January 1970). Data are discussed and presented in Abers et al. (1997). Focal mechanism is side looking, projected into the plane of the profile, shown at the preferred depth. Note the near-coincidence of the north-dipping nodal plane with the inferred fault surface. Vertical arrow shows approximate position of fault gouge sampled by ODP leg 180 (Shipboard Scientific Party 1999).

The water depth overlying the Moresby Seamount fraction of extension in the Aegean, perhaps 15 mm/yr of earthquake (MW = 6.2; #8 on Figure 3) is confirmed by the the ~35 mm/yr total (Clarke et al. 1998). Recent GPS geo- presence of prominent water multiples in its seismograms, detic measurements in the western Gulf, east of Patras, and places tight constraints on the epicenter of this event have shown that ~12 mm/yr extension is localized across (Abers et al. 1997). The period of these reverberations is the single fault system underlying the Gulf (Clarke et al. exceedingly sensitive to the exact water depth overlying the 1997), accompanied by abundant microseismicity in the source, and here requires a water depth of 3.0 ± 0.2 km, Patras area (Rigo et al. 1996). Utilizing waveform cross- placing it under the deepest part of the basin (Figure 5). correlation methods, Rietbrock et al. (1996) were able to There, only the shallow-north dipping plane shows promi- reduce hypocentral uncertainties for the microseismicity to nence in the seismic reflection record; the conjugate south- a few tens of meters, and show that most events lay in a dipping fault system is overlain by only ~2 km of water. plane dipping north at 12-20°. Although focal mechanisms Hence the choice of fault plane is clear, and it seems likely of some microearthquakes show a nodal plane of similar that this event occurred on a normal fault dipping near 30°. dip (Rigo et al. 1996) reanalysis of an expanded data set shows that the north-dipping planes most often dip at ~ 30° Gulf of Corinth at depth in the western Gulf (Hatzfeld et al. 2000). Per- haps, the plane of seismicity reflects the brittle-ductile tran- The Gulf of Corinth (Figure 6) accommodates a large sition rather than a fault plane (Hatzfeld et al. 2000).

Abers: Dips of seismogenic normal faults 12/15/00 p. 6

2000). 40º Gulf of N. Greece S A Corinth N 0 . 33 10 10 km km

Gulf of Corinth Fig. 7. Schematic diagram illustrating seismic zone in 38º the western Gulf of Corinth. Gray cloud shows region of background seismicity, abstracted from Rigo et al. 34 mm/yr (1996). These show focal mechanisms consistent with dips of 10-25° to the N (12-20° in the relocations of Peloponnesus Rietbrock et al. 1996). Star shows focus of 1995 Aigion mainshock and heavy line shows fault plane in- Aegean Sea ferred from seismic waveforms and geodetic observations (Bernard et al. 1997). N-S section transects western Gulf near 22.0° E longitude. The “A” marks the km town of Aigion, which was heavily damaged. 0 50 100 36º 22º 24º Fig. 6. Extension in the Gulf of Corinth region. Small Peruvian Andes circles: earthquakes in ISC catalog with seismic mo- 16 In the Peruvian High Andes, normal faulting occurs at ment estimated at > 10 Nm as estimated from magni- high elevations over an otherwise convergent plate bound- tudes as in Fig. 1. Large circles: earthquakes in Har- vard CMT catalog with apparent normal-faulting ary (e.g. Dalmayrac & Molnar 1981). Much of the evi- mechanisms (P-axis plunge > 45°). Focal mechanism dence for extension comes from Quaternary fault scarps, as shows waveform-based result (Bernard et al. 1997) for earthquakes are relatively rare. The largest normal-fault the June 15, 1995 Aigion earthquake (MS 6.2); dip of earthquake in the Andes occurred near Ancash, Peru, in shallow plane labeled. Large arrow shows motion of 1946, and produced extensive surface rupture along the Peloponnesian peninsula relative to Eurasia, averaged west-dipping Quiches fault (Bellier et al. 1991). First mo- from vector velocities on peninsula of (Clarke et al. tions and waveform modeling show the fault plane dips 30° 1998); 10-20 mm/yr of this rate is taken up across the for this event, with a source depth of 15-17 km and an Mw Gulf. of 6.8 (Doser 1987); a 10° uncertainty in dip is claimed by the author. The actual uncertainties could be larger, as In June 6, 1995, a large earthquake (MS 6.2) ruptured throughout the center of the microearthquake and GPS instrumentation in 1946 was both poorly distributed and of network, causing extensive damage at the town of Aigion uncertain calibration. The seismic moment and surface (Tselentis et al. 1996). Detailed waveform inversion of this rupture give a source dimension of many tens of kilome- earthquake shows normal faulting at a depth of 7.2 km with ters; a slightly larger seismic moment (Mw = 7.0) was esti- a plane dipping north at 33°±5° (Bernard et al. 1997), de- mated from fault offset and length (Bellier et al. 1991). termined using methods similar to those for the Woodlark Trenching across the surface trace reveals 1.7-3 m of offset Rift studies. In this case, good microseismic and geodetic across a fault, which together with one previous post- data confirm that the shallow-dipping nodal plane is the glacial rupture imply a vertical displacement rate less than fault. That fault geometry (Figure 7) coincides with GPS 0.25 mm/yr. Hence, the Ancash earthquake represents a and interferometric estimates of surface displacement, third region in which moderately low-angle normal faulting which suggest a fault slip of 0.9 m at depths between 2.5 may have produced an earthquake. Unlike the previous and 10 km (Bernard et al. 1997). The fault dips steeper two examples, it occurred in a setting where the long-term than inferred from the microseismic studies discussed extension rate is fairly low; geodetic data limit extension above, although depths are similar. across the Altiplano to less than a few mm/yr (Norabuena In summary, the western Gulf of Corinth shows an ex- et al. 1998). ample of normal faulting at the shallow end of dips commonly observed (i.e. 33°), a geometry supported by geo- Pre-Historic Examples detic and seismic observations. As with the Woodlark The geologic record contains many examples of normal examples, extension here is concentrated along a single faults that appear to have moved at shallow dips (Axen fault system which accommodates very rapid slip (10-20 1992; Wernicke 1995). In the two examples below, some mm/yr here); such rates are only possible in the most rap- evidence exists for seismic slip or surface offset; low-angle idly opening rifts (Table 1). Also as with the Woodlark faulting is claimed to occur via earthquakes. example, along-strike changes in opening rates correlate Johnson and Loy (1992) showed that late Quaternary with changes in fault dips: in the eastern Gulf of Corinth fault scarps along the Santa Rita fault, southern Arizona, fault plane dips increase to ~45° while extension rates de- extend to depth as planar features dipping 20o on seismic crease to ~6 mm/yr (Clarke et al. 1997; Hatzfeld et al.

Abers: Dips of seismogenic normal faults 12/15/00 p. 7

reflection sections. While no earthquakes have been ob- static frictional coefficient F, while the surrounding me- served on this fault, geomorphic evidence suggests move- dium is cohesive. At the relatively shallow depths of fault- ment within the last 100,000 years, over an area corre- ing observed (Figures 4-7) the effective maximum com- sponding to an earthquake with Mw = 6.7-7.6. Although pressive (vertical) stress σ1’ is low and the tensile strength seismic slip is not confirmed, this fault likely moved at T limits rock strength in extension (Figure 8b). The mini- depths normally considered within the brittle regime and at mum effective stress during faulting (σ3’) may be negative o a dip near 20 . in this situation: as long as σ3’ > -T, tensile failure or hy- Near the Chemehuevi-Sacramento Mountains, deposi- draulic fracture does not occur. To estimate this limit, tional patterns led Miller and John (1999) to infer earth- parabolic Griffith – Coulomb failure criterion is assumed quakes along a low-angle fault (dip < 30°), active between for the intact rock (Sibson 1998), while the fault is cohe- 23 and 12 Ma. They argue on the basis on the geometry sionless, with stresses calculated as in Figure 8. In this and distribution of coarse-grain strata that sediments were scenario, faulting continues on the cohesionless fault at deposited directly upon a detachment fault while that fault non-optimal angles until either Coulomb (shear) fracture or lay at shallow dip (< 30°) and was active. Megabreccias tensile (hydraulic) fracture occurs in the surrounding rock. represent catastrophic depositional events, which they ar- While the true stress state is likely affected by bending and gue were triggered by earthquakes with magnitude > 6.0, topography (e.g. Buck 1993; Forsyth 1992), this calculation the only clear source for which is the detachment fault it- provides a first-order estimate of the physical limits to fault self. Fault slip rates here were fairly rapid, 7-8 mm/yr dur- dip, with few assumptions of plate rheology. ing peak extension. Below sea floor, effective vertical stress is σ1’ = ρg(z – zw)(1 - λ), Discussion where ρ is the rock density, g is gravitational acceleration, z – zw is the depth below the sea floor, and λ is the ratio of These observations suggest that earthquakes do occa- pore pressure to lithostatic pressure (~ 0.4 for hydrostatic sionally occur on normal faults that dip near 30°, the low pore pressure) (e.g. Brace & Kohlstedt 1980). If the fault end of the range found by Jackson and White (1989). slides frictionally at coefficient of friction F and dip δ, a However, their occurrence is rare and perhaps tied to spe- relation exists between σ ’, σ ’, δ, and F on the fault, so cial geologic circumstances. In most cases such faulting is 1 3 that a nondimensional horizontal stress can be defined by: found in fault systems of high slip rate (>10 mm/yr), the −σ ' æ1− F cotδ ö one exception being the Ancash, Peru event, on which dip S = 3 = (1− λ)ç ÷ ρ ()− è + δ ø constraints are relatively weak. Furthermore, except for g z zw 1 F tan microseismicity the record does not yet include an earth- (following Sibson 1985). The hydrofracture condition, -σ3’ quake on a low-angle fault dipping <20°. Whether that < T, can then be recast in terms of dimensionless parame- absence reflects a fundamental physical limit of normal ters such that hydrofracture does not occur so long as faulting or merely a longer recurrence time for such events T S < = T’. (Wernicke 1995) remains to be seen. ρg()z − z Explanations for such faulting generally fall into two w Figure 8c shows values of S vs. δ for several appropri- categories, those that rely upon basal tractions (e.g. Melosh λ 1990; Yin 1989) or magmatic intrusion (Parsons & Thomp- ate values of and F. For typical rocks with simple faults son 1993) to rotate stress into a favorable geometry, and F ~ 0.6 (e.g. Brace & Kohlstedt 1980), but in well- those that appeal to fault weakening (e.g. Axen 1992). developed fault zones with thick gouge layers, F ~ 0.5 may While the former approach has many attractions, it has be more appropriate (see review of Lockner 1995). Many been shown that most stress-rotation models do not gener- scenarios (and some direct evidence) call for super- λ ate sufficient shear stresses to produce brittle sliding or (in hydrostatic conditions with > 0.4 (e.g. Axen 1992; the magmatic intrusion case) promote low-angle faulting in Hickman et al. 1995) although that may not be necessary very small areas (Wills & Buck 1997). Some situations here. The hydrofracture (tensile failure) condition is less with extreme flow rates at the base of a brittle layer may clear; T in crystalline rock varies from 10 to 40 MPa with produce the sufficient differential stresses (Westaway values of 25 MPa typical for mafic rocks (Lockner 1995), 1999). Such a situation may indeed be a consequence of comparable to cohesion in intact rock assumed in more rapid extension, particularly if extension is driven from complex faulting models (Buck 1993; Forsyth 1992; Lavier below in some way (e.g. Abers et al. 1997). Still, without et al. 1999). In the Woodlark Rift, earthquakes do not ex- constraints on the temperature, flow and rheology of the ceed 8 km depth and water depth varies from 0 to 3 km, so ρ lower crust, these models are difficult to test and remain g(z – zw) ~ 130 – 210 MPa, giving a hydrofracture thresh- speculative. old of T’ = 0.11 - 0.19 Failure requirements at the base of the seismogenic zone as indicated on Figure Analytical estimates. Do the observed dips actually 8c. T’ could be considerably larger at shallower depths. require very weak faults? Here, we calculate the stress Additional calculations show that extensional failure should conditions observed on faults for the dips observed, for be reached before shear (Coulomb) fracture as long as T’ > simple (i.e. Anderson-like) driving forces, and find that ~0.25(1-λ), for reasonable internal friction coefficients, only slight deviations from Byerlee’s law are necessary to over the relevant depth range, so Coulomb fracture can be explain the observed fault geometries. Figure 8 illustrates ignored. Thus, the limitations set by extensional failure the calculation: a fault with dip δ is cohesionless with

Abers: Dips of seismogenic normal faults 12/15/00 p. 8

(hydrofracture) control the minimum allowed fault dip. A modest decreases in strength or increases in pore pressure. slight weakening and small cohesion can be very important Normal faults rapidly enter the hydraulic fracture regime as in this scenario, as effective stresses are small. dips decrease below 10-15°, and slip on such faults would require other physical mechanisms (e.g. Axen 1992; Wer- nicke 1995). All observed large earthquakes occur with δ σ ' = σ ' = ρgz - P 1 V dips above this limit, and mechanical problems exist only if the microseismicity evidence for dips <20° (e.g. Rietbrock et al. 1996; Rigo et al. 1996) reflects large-scale stresses. σ ' = σ ' - ∆σ F The large event with the shallowest dip, the Gameta earth- 3 1 quake, has a centroid depth of 3 km (Abers 1991). At these A. depths the hydrofracture limit is not reached until S = 0.3, a condition easily reached on faults dipping near 20° pro- vided cohesion remains high. Thus, we conclude that observed normal faults could be understood without requiring τ unusual frictional conditions. failure Topographic effects. The stress calculation just de- 2T scribed assumes simple stress geometry, in particular one frictionσ' (τ , σ ) F with no stress rotations. Topographic variations could 0 0 τ = produce large changes in stress fields at least near the surface (e.g. McTigue & Mei 1981). One example is given in F = tanφ Figure 9, based on topographic effects caused by Moresby φ 2δ Seamount (calculation described in Abers et al. 1997). σ σ σ B. T 3' 1' ' Here, tectonic stresses are applied by stretching the crust until frictional failure at optimal angles (F = 0.7) is achieved somewhere at each depth. To isolate the topog- 1.4 raphic effect, cohesion is ignored and hydrostatic pore pressure is assumed. The faulting region is in a state of 1.2 F λ 0.6 0.4 decreased shear stress owing to a stress shadowing effect, 1 but stresses are rotated slightly to favor low-angle faulting. ) 0.6 0.7 w The net result is that failure is possible on the fault dipping 0.8 0.5 0.4 25° at F near 0.3. The added effects of elevated pore pres-

g(z-z 0.6 0.5 0.7

ρ sure or cohesion should allow sliding at F closer to 0.5 – '/

3 0.4 0.6, as suggested in the previous section. Similar topog- σ 0.2 Hydrofracture limit raphic stresses are expected for the Gulf of Corinth, where = -

S topography likewise defines a narrow rift basin between tall 0 mountains. -0.2 -0.4 0.5 Moresby Seamount Fault Fault Dips 25º to N 0 102030405060708090 0.4 on fault

C. dip δ, º n 0.3 Fig. 8. Relationship between fault dip (δ), rock tensile τ/σ σ σ strength (T), and effective principal stresses ( 1’ > 3’) 0.2

σ ” required for low-angle normal faulting when –T < 3’ < 20 0. (A) Assumed normal fault geometry in cross-section; fault has frictional coefficient F and no cohesion. (B) 10 plunge,

Mohr circle diagram showing relation between shear 3 stresses (τ) and effective normal stresses (σ’). Stresses σ 0 on the fault lie on the circle; the line labeled “friction” is the frictional failure envelope (F = 0.6) and that la- 2 4 6 8 beled “failure” is the assumed Coulomb-Griffith frac- Fault Depth, km ture envelope for intact surrounding rock with tensile Fig. 9. Stresses predicted on the Moresby Seamount strength T (after Sibson 1998). Hydraulic (tensile) frac- Fault, Papua New Guinea, in the presence of both to- ture occurs if σ ’ becomes less than –T. (C) Calculated pographic and extensional forces, from Abers et al. 3 (1997). Fault is assumed to be planar and dip ~25°, as stress ratio S, defined in text, vs. fault dip δ for various Fig. 5. Extension drives region to frictional failure be- assumptions of F and λ (pore pressure ratio). Gray bar neath its weakest point (beneath the “seamount” mas- shows the range of hydrofracture limits, T’, estimated in sif), assuming a coefficient of friction of 0.7 and hydro- text for the base of the seismogenic zone in the Wood- static pore pressure beneath sea floor. Additional lark Rift. forces are not considered, such as basal tractions or transient effects, and material is assumed cohesionless. Interpretation. Figure 8c shows that faults dipping as low as 25° should slip without hydraulic fracture for all These two sets of calculations are not exhaustive, and conditions, and this limit may be decreased to 15-20° with many alternative mechanical simulations can be found in

Abers: Dips of seismogenic normal faults 12/15/00 p. 9

the literature. Nevertheless, they show that reasonable Anderson, E. M. 1951. The Dynamics of Faulting. Oliver and assumptions about material strength and stress can account Boyd, Edinburgh. for normal-faulting earthquakes with dips of 20-35°. Axen, G. J. 1992. Pore pressure, stress increase, and fault weakening in low-angle normal faulting. J. Geophys. Res. 97, 8979- 8992. Conclusions Baldwin, S. L., Lister, G. S., Hill, E. J., Foster, D. A. & McDou- gall, I. 1993. Thermochronologic constraints on the tectonic evo- Seismicity in rifts tends to correlate with overall rate of lution of active metamorphic core complexes, D'Entrecasteaux extension. This observation explains the tendency of un- Islands, Papua New Guinea. Tectonics 12, 611-628. usual behavior such as low-angle faulting to be reported Bellier, O., Dumont, J. F., Sebrier, M. & Mercier, J. L. 1991. Geo- most commonly in the Woodlark Rift and Aegean, the two logical constraints on the kinematics and fault-plane solution of most rapidly opening rifts. In these locales, normal faults the Quiches fault zone reactivated during the 10 November 1946 ° Ancash earthquake, northern Peru. Bull. Seismol. Soc. Amer. 81, dipping 23-35 are observed in the regions of fastest exten- 468-490. sion: 25-40 mm/yr near the rift tip in the Woodlark- Bernard, P., Briole, P., Meyer, B., Lyon-Caen, H., Gomez, J. M., D’Entrecasteaux region, and 10-20 mm/yr across the west- Tiberi, C., Berge, C., Cattin, R., Hatzfeld, D., Lachet, C., Le- ern Gulf of Corinth. These observations only slightly ex- brun, B., Deschamps, A., Courboulex, F., Larroque, C., Rigo, tend the range of previously-reported dips for normal A., Massonnet, D., Papadimitriou, P., Kassaras, J., Diagourtas, faults, but demonstrate that the low end of the dip range is D., Makropoulos, K., Veis, G., Papazisi, E., Mitsakaki, C., commonly achieved; models of faulting that restrict dips to Karakostas, V., Papadimitriou, E., Papanastassiou, D., Chou- liaras, G. & Stavrakakis, G. 1997. The MS = 6.2 June 15, 1995 > 45° can not explain many earthquakes. Several other Aigion earthquake (Greece): evidence for low angle normal examples of potentially seismogenic low-angle normal faulting in the Corinth rift. J. Seismol. 1, 131-150. faults have been reported in the literature but none as well Brace, W. F. & Kohlstedt, D. L. 1980. Limits on lithospheric stress constrained. These observations suggest that processes imposed by laboratory experiments. J. Geophys. Res. 85, 6248- associated with high extension rates (and high overall ex- 6252. tension) on individual fault systems may produce the condi- Buck, W. R. 1988. Flexural rotation of normal faults. Tectonics 7, tions needed for low-angle faulting. For example, well- 959-973. worn faults should achieve effective frictional coefficients Buck, W. R. 1993. Effect of lithospheric thickness on the formation of high- and low-angle normal faults. Geology 21, 933-936. near 0.5. Such coefficients do not require very unusual Byerlee, J. D. & Savage, J. C. 1992. Coulomb plasticity within the materials but may be expected in faults which achieve large fault zone. Geophys. Res. Lett. 19, 2341-2344. offsets in short time spans, so can develop mature, unce- Calais, E., Lesne, O., Deverchere, J., San'kov, V., Lukhnev, A., mented gouge zones. Miroshnitchenko, A., Buddo, V., Levi, K., Zalutzky, V. & Bash- The physical conditions for such faulting remain poorly kuev, Y. 1998. Crustal deformation in the Baikal Rift from GPS understood. Except for the Aegean case, little is known measurements. Geophys. Res. Lett. 25, 4003-4006. locally about the geometry of faulting, and neither the con- Clarke, P. J., Davies, R. R., England, P. C., Parsons, B., Billiris, ditions within the fault zones nor the behavior of the deeper H., Paradissis, D., Veis, G., Cross, P. A., Denys, P. H., Ashke- nazi, V., Bingley, R., Kahle, H. G., Muller, M. V. & Briole, P. ductile crust are well known. In particular, the role of the 1998. Crustal strain in central Greece from repeated GPS meas- lower crust in these processes may be critical, yet few di- urements in the interval 1989-1997. Geophys. J. Int. 135, 195- rect observations exist on its state. Also, direct measures of 214. the cohesion, mechanical strength and permeability struc- Clarke, P. J., Davies, R. R., England, P. C., Parsons, B. E., Billiris, ture of these active fault zones would allow several models H., Paradissis, D., Veis, G., Denys, P. H., Cross, P. A., Ashke- to be tested. Future experiments and observations aimed at nazi, V. & Bingley, R. 1997. Geodetic estimate of seismic haz- resolving these ambiguities may be necessary. ard in the Gulf of Korinthos. Geophys. Res. Lett. 24, 1303-1306. Dalmayrac, B. & Molnar, P. 1981. Parallel thrust and normal faulting in Peru and constraints on the state of stress. Earth and Acknowledgements. This work was inspired by discus- Planet. Sci. Lett. 55, 473-481. sions at the Geological Society workshop on non-volcanic Davies, H. L. & Warren, R. G. 1988. Origin of eclogite-bearing, rifted margins. It benefitted from reviews by J. Jackson domed, layered matamorphic complexes ("core complexes") in and T. Reston. This research is supported by the National the D'Entrecasteaux Islands, Papua New Guinea. Tectonics 7, 1- Science Foundation, grants OCE-9012730, OCE-9730567 21. DeMets, C., Gordon, R., Argus, D. & Stein, S. 1990. Current plate and OCE-9996317. motions. Geophys. J. Int. 101, 425-478. Dixon, T. H., Stefano, R., Lee, J. & Reheis, M. C. 1995. Con- References straints on present-day Basin and Range deformation from space geodesy. Tectonics 14, 755-772. Abers, G. A. 1991. Possible seismogenic shallow-dipping normal Doser, D. I. 1987. The Ancash, Peru, earthquake of 1946 Novem- faults in the Woodlark-D'Entrecasteaux extensional province, ber 10; evidence for low-angle normal faulting in the high Andes Papua New Guinea. Geology 19, 1205-1208. of northern Peru. Geophys. J. R. astr. Soc. 91, 57-71. Abers, G. A., Mutter, C. Z. & Fang, J. 1997. Shallow dips of nor- Dziewonski, A. M., Chou, T. A. & Woodhouse, J. H. 1981. De- mal faults during rapid extension: Earthquakes in the Woodlark- termination of earthquake source parameters from waveform D'Entrecasteaux rift system, Papua New Guinea. J. Geophys. data for studies of global and regional seismicity. J. Geophys. Res. 102, 15,301-15,317. Res. 86, 2825-2852. Ahorner, L. 1983. Seismicity and neotectonic structural activity of Ekström, G. & Dziewonski, A. M. 1986. Evidence of bias in esti- the Rhine Graben system in Central Europe. In: Seismicity and mations of earthquake size. Nature 332, 319-323. Seismic Risk in the Offshore North Sea Area (edited by Ritsema, Forsyth, D. W. 1992. Finite extension and low-angle faulting. A. R. & Gürpinar, A.). D. Reidel, Boston, 101-112. Geology 20, 27-30.

Abers: Dips of seismogenic normal faults 12/15/00 p. 10

Hatzfeld, D., Karakostas, V., Ziazia, M., Kassaras, I., Papadimit- Ruppel, C. 1995. Extensional processes in continental lithosphere. riou, E., Makropoulos, K., Voulgaris, N. & Papaioannou, C. J. Geophys. Res. 100, 24,187-24,215. 2000. Microseismicity and fault geometry in the Gulf of Corinth Savage, J. C., Lisowski, M. & Prescott, W. H. 1985. Strain accu- (Greece). Geophys. J. Int. 141, 438-456. mulation in the Rocky Mountain states. J. Geophys. Res. 90, Hickman, S., Sibson, R. & R. Bruhn. 1995. Introduction to special 10,310-10,320. section: Mechanical involvement of fluids in faulting. J. Geo- Senior, B. R. & Billington, W. G. 1987. Structural interpretation of phys. Res. 100, 12,831-12,840. northwest Fergusson Island, Papua New Guinea. In: Proc. Pa- Hill, E. J. & Baldwin, S. L. 1993. Exhumation of high pressure cific Rim Congress 87. Australasian Institute of Mining and metamorphic rocks during crustal extension in the D'Entre- Metallurgy, Gold Coast, Australia, 387-392. casteaux Islands, Papua New Guinea. J. Metamorph. Geol. 11, Shipboard Scientific Party. 1999. Leg 180 summary. In: Proc. 261-277. ODP Init. Repts., 180 (edited by Taylor, B., Huchon, P., Klaus, Hill, E. J., Baldwin, S. L. & Lister, G. S. 1992. Unroofing of active A. & et al.). Ocean Drilling Program, College Station, TX, 1-77. metamorphic core complexes in the D'Entrecasteaux Islands, Sibson, R. H. 1985. A note on fault reactivation. J. Struct. Geol. 7, Papua New Guinea. Geology 20, 907-910. 751-754. Jackson, J. & McKenzie, D. 1988. The relationship between plate Sibson, R. H. 1998. Brittle failure mode plots for compressional motions and seismic moment tensors, and the rates of active de- and extensional tectonic regimes. J. Struct. Geol. 20, 655-660. formation in the Mediterranean and Middle East. Geophys. J. Taylor, B., Goodliffe, A., Martinez, F. & Hey, R. 1995. Continen- Int. 93, 45-73. tal rifting and initial sea-floor spreading in the Woodlark basin. Jackson, J. A. & White, N. J. 1989. Normal faulting in the upper Nature 374, 534-537. continental crust: Observations from regions of active extension. Taylor, B., Goodliffe, A. M. & Martinez, F. 1999. How continents J. Struct. Geol. 11, 15-36. break up: Insights from Papua New Guinea. J. Geophys. Res. Johnson, R. A. & Loy, K. L. 1992. Seismic reflection evidence for 104, 7497-7512. seismogenic low-angle faulting in southeastern Arizona. Geol- Tselentis, G. A., Melis, N. S., Sokos, E. & Papatsimpa, K. 1996. ogy 20, 597-600. The Egion June 15, 1995 (6.2 ML) earthquake, Western Greece. Lavier, L. L., Buck, W. R. & Poliakov, A. N. B. 1999. Self- Pure and Appl. Geophys. 147, 83-98. consistent rolling-hinge model for the evolution of large-offset Weichert, D. H. 1980. Estimation of the earthquake recurrence low-angle normal faults. Geology 27, 1127-1130. parameters for unequal observations for different magnitudes. Le Pichon, X., Chamot-Rooke, N., Lallemant, S., Noomen, R. & Bull. Seismol. Soc. Amer. 70, 1337-1346. Veis, G. 1995. Geodetic determination of the kinematics of cen- Weissel, J. K., Taylor, B. & Karner, G. D. 1982. The opening of tral Greece with respect to Europe; implications for eastern the Woodlark Basin, subduction of the Woodlark spreading sys- Mediterranean tectonics. J. Geophys. Res. 100, 12,675-12,690. tem, and the evolution of northern Melanesia since mid-Pliocene Lockner, D. A. 1995. Rock failure. In: Rock Physics and Phase time. Tectonophysics 87, 253-277. Relations, AGU Ref. Shelf 3 (edited by Ahrens, T. J.). Amer. Wernicke, B. 1995. Low-angle normal faults and seismicity: a Geophys. Union, Washington, 127-147. review. J. Geophys. Res. 100, 20,159-20,174. McCaffrey, R. 1997. Statistical significance of the seismic cou- Wernicke, B. & Axen, G. A. 1988. On the role of isostasy in the pling coefficient. Bull. Seismol. Soc. Amer. 87, 1069-1073. evolution of normal fault systems. Geology 16, 848-851. McTigue, D. F. & Mei, C. C. 1981. Gravity-induced stresses near Westaway, R. 1999. The mechanical feasibility of low-angle nor- topography of small slope. J. Geophys. Res. 86, 9268-9278. mal faulting. Tectonophysics 308, 407-443. Melosh, H. J. 1990. Mechanical basis for low-angle normal fault- Wills, S. & Buck, W. R. 1997. Stress-field rotation and rooted ing in the Basin and Range province. Nature 343, 331-335. detachment faults; a Coulomb failure analysis. J. Geophys. Res. Miller, J. M. G. & John, B. E. 1999. Sedimentation patterns sup- 102, 20,503-20,514. port seismogenic low-angle normal faulting, southeastern Cali- Yin, A. 1989. Origin of regional rooted low-angle normal faults: A fornia and western Arizona. Geol. Soc. Amer. Bull. 111, 1350- mechanical model and its tectonic implications. Tectonics 8, 1370. 469-482. Mutter, J. C., Mutter, C. Z. & Fang, J. 1996. Analogies to oceanic behaviour in the continental breakup of the western Woodlark basin. Nature 380, 333-336. Norabuena, E., Leffler-Griffin, L., Mao, A., Dixon, T., Stein, S., Sacks, I. S., Ocola, L. & Ellis, M. 1998. Space geodetic observations of Nazca-South America convergence across the Central Andes. Science 279, 358-362. Nuttli, O. W. 1985. Average seismic source-parameter relations for plate-margin earthquakes. Tectonophysics 118, 161-174. Ollier, C. D. & Pain, C. F. 1980. Active rising surficial gneiss domes in Papua New Guinea. J. Geol. Soc. Austr. 27, 33-44. Parsons, T. & Thompson, G. A. 1993. Does magmatism influence low-angle normal faulting? Geology 21, 247-250. Rietbrock, A., Tiberi, C., Scherbaum, F. & Lyon-Caen, H. 1996. Seismic slip on a low-angle normal fault in the Gulf of Corinth: Evidence from high-resolution cluster analysis of microearthquakes. Geophys. Res. Lett. 23, 1817-1820. Rigo, A., Lyon-Caen, H., Armijo, R., Deschamps, A., Hatzfeld, D., Makropoulos, K. & Papadimitriou, P. 1996. A microseismic study in the western part of the Gulf of Corinth (Greece): Impli- cations for large scale normal faulting mechanisms. Geophys. J. Int. 126, 663-688. Ripper, I. D. 1982. Seismicity of the Indo-Australian/Solomon Sea plate boundary in the southeast Papua region. Tectonophysics 87, 355-369.

Abers: Dips of seismogenic normal faults 12/15/00 p. 11