Upper plate stressing and seismicity in the subduction earthquake cycle

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 103, NO. B10, PAGES 24,523-24,542, OCTOBER 10, 1998 Upper plate stressing and seismiity in the subdution earthquake yle M. A. J. Taylor, R. Dmowska, and J. R. Rie Division of Engineering and Applied Sienes and Department of Earth and Planetary Sienes, Harvard University Cambridge, Massahusetts Abstrat. We investigate upper plate stressing during the earthquake yle in a subdution segment, using three-dimensional (3-D) elasti models to address the effets of strongly heterogeneous oupling along strike of the interplate interfae. We show how heterogeneity ontrols the loations and mehanisms of seismiity in the upper plate. Oblique subdution segments, two from the Aleutians (Andreanof Islands 1986 and Rat Islands 1965) and one from Indonesia (Biak 1996) are studied. All examples of upper plate seismiity from the Aleutians represent events ourring toward the beginning of a new yle, while in B iak, Indonesia, the examined events our both toward the end of one yle and the beginning of the next. In the majority of ases studied, the loation and mode of the upper plate seismiity are onsistent with spae- and time-dependent stressing as predited by modeling. This onfirms earlier observations that seismiity in the viinity of large/great subdution earthquakes (toward the outer rise, at intermediate depth, and now in the upper plate) depends, in an interpretable manner, on the stage in the earthquake yle as well as on distribution of oupling along the interplate interfae. 1. Introdution Studies of seismiity near subdution margins, in the outer rise as well as at intermediate depth [Christensen and Ruff, 1983, 1988; Astiz et al., 1988; Drnowska et al., 1988; Lay et al., 1989], provide evidene that there is time dependene of the style of seismiity in large regions adjaent to a strongly oupled thrust interfae. In the present study we doument and interpret suh effets in the upper plate. Assoiation of regional seismiity with a major underthrusting event is most evident for the seismiity of the outer rise (inluding outer slope trenh). Global surveys [Astiz et al., 1988; Lay et al., 1989] show that outer rise earthquakes with tensional normal faulting our abundantly over an extended period after the thrust event (most happening within 10 to 20 years), whereas suh events are rare or absent as the yle matures, at whih time the less frequent ompressional outer rise events may our. Suh response is readily explained [Drnowska and Lovison, 1988; Drnowska et al., 1996a; Taylor et al., 1996] when we onsider that the outer rise and outer slope trenh is loaded by a bending stress distribution, whose propensity to ause seismiity is modulated by small extensional or ompressional stress hanges assoiated with the thrust earthquake yle. Further, Drnowska and Lovison [1992] studied the seismi response of outer rise zones in the regions of the Valparaiso (Chile) 1985, Alaska 1964, and Rat Islands (Aleutians) 1965 earthquakes. They found that the seismiity tended to luster along strike of the subduting margin in assoiation with known asperities (i.e., regions of highest moment release) inferred by others through modeling seismi radiation from the underthrusting events. Modeling of outer rise stress hanges Copyfight 1998 by the Amerian Geophysial Union. Paper number 98JB /98/98 JB ,523 when large, strong heterogeneities were distributed along strike in the seismogeni zone [Dmowska et al., 1996a] gave an explanation of this lustering. This inluded the way that ative zones in the outer rise, in the time following the thrust event, are systematially offset relative to asperity loations in regions of strongly oblique subdution as for the Rat Islands 1965 zone. Earthquake yle effets are also present, if less. marked, in the seismiity of the slab at intermediate depth [Astiz et al., 1988; Lay et al., 1989], in a way that is approximately onsistent with what is expeted from superposition of stress flutuations in the earthquake yle on a time-averaged stress state dominated by slab pull [Taylor et al., 1996]. Dmowska and Lovison [1992] showed that suh intermediate depth seismiity also lustered along strike in assoiation with asperities, but for strongly oblique onvergene like for Rat Islands 1965, the lustering is offset along strike to the opposite side of the asperity from the offset in the outer rise. In the work reported here, we examine suh earthquake yle, asperity, and obliquity effets in the upper plate. This is done through studies of seismiity in the regions of the Andreanof Islands 1986 and Rat Islands 1965 subdution events in the Aleutians and of the Biak, Irian Jaya, Indonesia 1996 subdution event. We estimate the Coulomb measure of stress hange in the upper plate assoiated with strongly heterogeneous (asperity ontaining) modes of slip on the subdution interfae for whih stress hanges are alulated from three-dimensional (3-D) models. We use suh stress hanges to rationalize the upper plate seismiity. We find that regions in the upper plate whih show earthquakes in the period of days to a few years after the thrust event are those whih underwent an inrease of Coulomb stress. That seismiity is loated well outside what would normally be alled the aftershok zone. In ases of tetonially ative areas for whih the Coulomb stress dereases, suh postevent seismiity is not present, illustrating the "stress shadow" onept. This manner of

2 24,524 TAYLOR ET AL.: UPPER PLATE STRESSING AND SEISMICITY interpreting seismiity has been very suessful in ontinental regions, partiularly in assoiation with events along the San Andreas fault system [e.g., Harris and Simpson, 1992; Jaurne and Sykes, 1992; Stein et al., 1992; Reasenberg and Simpson, 1992; Simpson and Reasenberg, 1994; King et al., 1994; Harris et al., 1995; Harris and Simpson, 1996]. Moreover, stress shadows from the great 1906 [Simpson and Reasenberg, 1994; Harris and Simpson, this issue] and 1857 [Harris and Simpson, 1996; Deng and Sykes, 1997] events have onviningly been argued to have masked the atual seismi hazard of tetonially ative areas for many years. model [Minster and Jordan, 1978].) It was followed in the first 1.5 months by a series of shallow upper plate earthquakes lose to Atka Island, of whih the largest are shown in the right orner of Figure 1, all with strike-slip mehanisms [Ekstr6m and Engdahl, 1989]. The five largest events range from Mw = 5.3 to Mw = 6.5 and are onsistent with fight-lateral motion on ar-parallel fault planes. EkstrOm and Engdahl [1989] interpret these events as ourring along a weak arparallel strike-slip shear zone in the upper plate, near the volani line, and aommodating the strike-slip part of the oblique subdution (plate onvergene diretion shown in Figure 1). 2. Examples of Upper Plate Seismiity Convergent Margins We present three speifi ases of upper plate seismiity assoiated with the subdution earthquake yle Andreanof Islands, Aleutians The May 7, 1986 (M w = 8.0), Andreanof' Islands earthquake ruptured an -280 km streth of an oblique segment of the entral Aleutians (Figure 1 [after EkstrOm and Engdahl, 1989] approximate boundaries of the aftershok zone from Engdahl et al. [1989]), where the estimated plate motion is 82 m/yr. (This and subsequent plate motions quoted are from the RM2 at 2.2. Rat Islands, Aleutians: The February 4, 1965 (Mw = 8.7), Rat Islands earthquake ruptured a 600-km segment along the western end of the Aleutian Islands, where subdution is more oblique than in the Andreanof Islands segment with an estimated plate onvergene rate of 86 m/yr. The map view of the first-order asperity distribution from Bek and Christensen [1991] is shown in Figure 2, together with the aftershok zone (dashed line). On July 4, 1966, a shallow (depth 13 km) strike-slip mb = 6.2 earthquake [Stauder, 1968b] ourred at the eastern end of the aftershok zone, the mehanism being shown in Figure 2. Stauder [1968b] interpreted that earthquake as left-lateral on a N-S trending fault, transverse to the struture of the island ar, in a plae perhapseparating the Andreanof Islands and 53N June 1986, 5.3 < Mw < N 51N 50N 178W 176W 174W 172W Figure 1. Bak ar strike-slip ativity along the Andreanof Islands setion of the Alaskan/Aleutian trenh following the May 7, 1986, Mw = 8.0 earthquake [from EkstrOm and Engdahl, 1989]. Approximate boundaries of the aftershok zone from Engdahl et al. [1989].

3 . TAYLOR ET AL.: UPPER PLATE STRESSING AND SEISMICITY 24,525 m b = 6.0 m b = 5.9, Feb. 2, 1'975 ß : ß! i i I i ß i i : slip 0.4 -' 0.2 ß, i o i '. ' ' ".,-, ',....._ L.. B slip DISTANCE ALONG FAUbT STRIKE (KM) _ l ' '-... " '"'it'... '... 'O"'[" -'- I... 54øN July 4, 1966 m b = 6.2, ø N 170øE ø 176 ø ø Figure 2. Strike-slip seismiity in the Rat Islands, Aleutians: Map view of asperity distribution of the Rat Islands Mw = 8.7 earthquake of February 4, 1965, from Bek and Christensen [1991]. Aftershok zone shown with dashed line. Mehanism of July 4, 1996, aftershok from Stauder [1968b] and of February 2, 1975 doublet from Newberry et al. [ 1986]. Rat Islands tetoni bloks. Figure 2 also shows two other strike-slip events (mehanisms from Newberry et al. [1986]), lose to Near Island, north of the westernmost asperity of the 1965 event, forming a doublet of rn b 5.9 and 6.0 of February 2, This doublet was interpreted as right-lateral strike-slip motion along a northwesterly fault plane IComtier, 1975b], as an extension of strike-slip faulting desribed by Corntier [1975a] in the Komandorsky Islands. Newberry et al. [1986] prefer to hoose the northeastrending nodal plane for these earthquakes, noting that Agattu Canyon to the south of Attu Island and major faults on Attu Island itself are aligned with the trend of this nodal plane. Agattu Canyon has been interpreted as the surfae expression of the boundary between two tetoni bloks of the Aleutian Ar, at least as far north as the 100-m bathymetri ontour. Choie of the northeast striking nodal plane as fault plane for these events would suggesthat this boundary may extend north of Attu Island Irian Jaya, Indonesia The February 17, 1996 (Mw = 8.2), Biak earthquake ruptured at least 270 km along the New Guinea trenh, the event being a thrust in a zone of very oblique subdution with estimated relative plate motion of 13 m/yr. A tetoni map of Irian Jaya with the New Guinea Trenh in the upper right hand omer is shown in Figure 3 [from Puntodewo et al., 1994]. The map inludes Biak Island, most devastated by the earthquake. As for the other regions disussed, in the Biak area there is a disrepany between the plate motion, very oblique to the New Guinea Trenh, and earthquake slip vetors (short arrow), loser to a diretion perpendiular to the trenh (but still oblique). This is illustrated in Figure 3 by a dashed box next to Biak inluding mehanisms of earthquakes from that area, with a sum mehanism shown outside the box. The long arrows show expeted motion of the Paifi plate relative to Australia, while the short arrows show the slip vetor azimuth derived from summing moment tensors of earthquakes within the box [Puntodewo et al., 1994]. The main event was followed by a number of events in the upper plate, the two largest of whih were the Mw 6.4 (February 18) and 6.5 (February 17) events that ourred SW of the mainshok within 2 days. The mainshok was preeded by a series of events from 1979 (probably triggered by a Mw = 7.5, September 12, 1979, Yapen earthquake, lose by to the southwest), in similar (but not oinident) positions to the postmainshok seismiity. All of these events are left-lateral strike-slip (or with mehanisms lose to that), and together they form an approximately linear feature parallel to the main rupture, loated to the SW. All mehanisms are shown in Figure 4, inluding the main event and two early aftershoks. For larity, earthquakes before the main event have mehanisms shown in the top part of the figure; events after the main event are shown in the middle and bottom parts of the figure. Also, Figure 4 has been reoriented, with north pointing downward, to make later omparisons between the three ases simpler.

4 24,52.6 TAYLOR ET AL.: UPPER PLATE S7RF3SING AND SEISMICI'rY I'N Mapia o IøS Bird' s BIAK x 8 Bird' s 2øS 3øS TAFZ 5øS 132øE 133øE 134øE 135øE 136øE 137øE 138øE 139øE 140øE 141øE Figure 3. Tetoni map of Irian Jaya showing fault plane solutions, faults (solid lines), and inferred faults (dashed lines). Long arrows near enlosed regions show expeted motion of Paifi relative to Australia, and the short arrows show the slip vetor azimuths from summing moment tensors of all earthquakes in boxes. From Puntodewo et al. [1994]. 3. Calulation of Coulomb Shear Stress In order to interpret the bak ar seismiity in eah of the above examples, we adopt the onept of Coulomb shear stress. This onept has already been developed in onnetion with seismiity along faults of the San Andreas system [e.g., Harris and Simpson, 1992; Jaume and Sykes, 1992; Stein et al., 1992; Reasenberg and Simpson, 1992; Simpson and Reasenberg, 1994; King et al., 1994; Harris et al., 1995; Harris and Simpson, 1996] and has proven valuable in understanding how stress hanges from an earthquake indue (or retard) seismiity in adjaent regions. The idea of relating Coulomb stress hanges to aftershok loation seems to have emerged in the work of Stein and Lisowski [1983] and Rybiki et al. [1985], although the onept that suh ombination of shear and effetive normal stress was appropriate to understanding faulting was in use earlier [e.g., Hubbert and planar fault with unit normal n, and unit vetor in the slip diretion s, the hange in Coulomb shear stress, Ao' is defined by A O'ns = A O'ns + f A O'nn (1) where f is the frition oeffiient and Ao' is the hange in the stress tensor. The inreases or dereases in Coulomb shear stress are onsidered to be superimposed on the preexisting stress field in the viinity of the subdution zone and simply modulate that overall stress field, thus promoting or inhibiting seismiity aordingly. The method pursued here is to impose oseismi slip on part of the thrust interfae of our three-dimensional model of a subdution zone, in a way whih allows free slip elsewhere on the interfae (to represent negligible stress drop there), and to alulate the oseismi stati stress hanges that our in the upper plate in response to slip on the interfae. Rubey, 1959; Rayleigh et al., 1972]. Further, assoiations between shear stress hange due to earthquakes and aftershok loations were suggested by Chinnery [1966a, b], and Rybiki 4. Three Dimensional Modeling: Heterogeneous [1970, 1973], and effets of stress hanges of major Slip Along Strike earthquakes on future large events were disussed by Smith and Van de Lindt [ 1969]. Here we extend the appliation of Coulomb stress to upper For large (M > 8) subdution zone earthquakes, the plate seismiity in the viinity of subdution zones. For a distribution of slip along-strike is, typially, highly

5 TAYLOR ET AL.: UPPER PLATE STRESSING AND SEISMICITY 24, E 137E! A C A 135E A i01679c A Mw = D Mw = 6.5 N Figure 4. Upper plate seismiity in Irian Jaya, Indonesia preeding and immediately following the February 17, 1996, Mw = 8.2 earthquake. nonuniform with isolated asperity regions of highest slip. This is exemplified in the inversion of moment distribution for both the 1986 Andreanof Islands [Hwang and Kanamori, 1986, Boyd and Ntib lek, 1988; Houston and Engdahl, 1989; Das and Kostrov, 1990; Yoshida, 1992] and 1965 Rat Islands [Bek and Christensen, 1991] (Figure 2) main events. In order to study the resulting Coulomb shear stress hanges from heterogeneou slip along-strike we examine a simple 3-D model of a generi subdution zone (Figure 5) as used in Dmowska et al. [1996a] for alulating individually the extensional and shear stress hanges due to oblique slip on asperities of various shapes and distributions. Our approah is similar here. We inorporate an isolated region of high slip on the thrust interfae (an "asperity"), indiated by the dark shading in Figure 5. The dip angle of the subduting slab, 0, the diretion of obliquity from the trenh-normal,, and the downdip width of the seismogeni thrust interfae, W, are also shown. All of the seismogeni width slips during the simulated event, though by a spatially variable amount. We attempt to model the effet of greater slip at asperities, where there may be presumed to be a higher degree of oupling between the subduting and overriding plates, while the remainder of the thrust interfae, whih is less oupled, experienes less slip in the main event. (It is assumed that the less oupled part slips aseismially or by low magnitude seismiity during the period between large thrust events.) This is simulated by applying slip of magnitude D on the asperity on the thrust interfae and alulating the slip elsewhere on the seismogeni interfae so as to assure that there is no net oseismi hange in shear stress there during the rupture. Hene the thrust interfae is modeled as "freely slipping" outside the asperity zone. The amount of slip is found to diminish rapidly from D as we move outside the asperity zone, and the slip distribution tapers to zero at the downdip edge of the seismogeni zone. Details of suh modeling are given by Drnowska et al. [1996a]. (We have, however, notied that ontour intervals reported there for dimensionless plots of oseismi stress hange were listed inorretly; all ontour intervals reported in Dmowska et al. [1996a] for normal stress hanges should be inreased by a fator of 6, and for shear stress hanges by a fator of 3.) This approah may be 'ontrasted with standard 3-D elasti disloation modeling, in whih slip is imposed, in a possibly nonuniform way, everywhere along the fault zone, without regard for the patterns of stress drop (or inrease) that are thereby indued.

6 24,528 TAY[DR ET AL.: UPPER PLATE STRF. SSING AND SEISMICrI Finite Element Mesh: Thrust interfae with an asperity: (downdip) no slip trenh/ I oblique slip freely slipping imposed on asperity lip as required for no hange vetor (along strike) (slips in shear trations) Figure 5. Three-dimensional finite element model with oblique slip, heterogeneous along-strike: finite element mesh and distribution of slip on thrust interfae with an asperity [after Dmowska et al., 1996a] Seletion of Model Parameters In order to generate results from our models that we an ompare with the loations and mehanisms of upper plate seismiity in the three regions studied here, we require approximate values for the angle of dip, 0, the angle of oblique slip from the trenh-normal in the main event,, the downdip width of the thrust interfae, W, and an upper plate frition oeffiient, f, for eah region. 1. For the Andreanof Islands, in the Aleutians, fairly good estimates for all the required parameters an be obtained from Ekstr6m and Engdahl [1989] in their examination of the soure parameters and seismiity surrounding the May 7, 1986, main event. In partiular, Figure 5 from their paper, showing the ross setion aross the Aleutian Ar with the position and nodal-planes of events on the thrust interfae, allows a good estimate of 0 and W. Although the thrust interfae undergoes a marked inrease in dip at a depth of about 30 km, it an be fairly well approximated by a plane dipping at onstant angle of ø from base of the rupture zone right up to the trenh, a distane of about 138 km (W). Ekstriim and Engdahl [1989] find that although the angle of onvergene of the Paifi and North Amerian plates is about N30øW from the trenh normal, the average slip vetors for earthquakes on the thrust fault are typially ø less than that (as is typial for subdution zones of oblique onvergene [Yu et al., 1993]), and so we examine a model where the angle of obliquity, q = 18 ø. 2. Parameters for the 1965 Rat Islands event are slightly harder to obtain as it ourred before the advent of the Harvard

7 TAYLOR ET AL.: UPPER PLATE STRESSING AND SEISMICITY 24,529 entroid moment tensor (CMTs) and no detailed study of the geometry and veloity struture of the region has been arried out (like that of EkstrOrn and Engdahl [1989] and Engdahl et al. [1989] for the Andreanof Islands) whih would give us both estimates of the variation in angle of dip and the length of the seismogeni zone. However, an inversion of the moment distribution and an approximation of the dimensions of the rupture zone from the subsequent distribution of aftershoks has been arried out by Bek a.-d Christensen [1991] (both shown in Figure 2). The soure mehanism for the main event has also been inverted in a study by Wu and Kanarnori [ 1973]. Assuming a onstant angle of dip between the hypoenter of the 1965 event and the trenh gives 0 = 20 ø (from both the International Seismologial Centre (ISC) data and Wu and Kanarnori [1973]). This is also onsistent with the results of Stauder [1968a], who finds that most of the thrust events in the aftershok series following the main event have a dip angle between 15 ø and 20 ø. Both the downdip rupture zone length and the angle of oblique slip an be estimated from the map from Bek and Christensen [1991] (Figure 2) whih gives W = 90 km and ø (also onsistent with the results of Wu and Kanamori). 3. The main Biak event ourred in February 1996, and at the time of writing, we have no information about any inversions for the slip distribution along the rupture plane, although we do have CMTs for the period sine the event. From the CMT from the main event the dip of the rupture plane an be asertained to be 0 = 11 ø, and the diretion of slip on the thrust interfae is around ø from the trenh normal ( measured as negative to indiate left-lateral oblique slip on the interfae, opposite to that in the Aleutians). In the absene of a knowledge of the downdip extent of the rupture zone we annot reasonably estimate W. However, the island of Biak underwent subsidene during the main February event. Its NE side subsided by 2 m, while the south and west sides dropped by < 1 m (R. MCaffrey, private ommuniation, 1997). By examination of our 2-D modeling of uplift versus distane along the overlying plate of a generi subdution zone, [from Taylor et al., 1996, Figure 13a], it is seen that vertial displaement hanges are only negative (subsidene) beyond the downdip end of the rupture-zone (in a line measured from the trenh and inreasing out over the upper plate). Hene it seems reasonable to assume that the northern edge of the island of Biak provides an upper bound on the extent of rupture from the trenh (and thus W). Here, we do not a priori assume a value of the frition oeffiient but instead employ a range of values with the stress fields omputed from our model. Ideally, one ould try to onstrain f by omparing the results from the different values with the observed seismiity in the regions examined, but the data seldom suffie for this Finite Element Modeling For subdution in the Aleutians the Paifi plate is subduting roughly SE to NW underneath the North Amerian plate, so we impose slip on the asperity on our thrust interfae in the same sense with angles of oblique slip p = 18 ø and p = 42 ø from the trenh normal (see Figure 5), typial values based on the events in the Andreanof Islands and Rat Islands, respetively. The finite element model, whih, for this simplified geometry, has a onstant angle of dip of 0 = 20 ø, was run using the ode ABAQUS. Extensional and right-lateral stress hanges (the relative plate motion indues right-lateral shear along ar-parallel strike-slip features in the bak ar region) were obtained from the set of elements omprising the surfae of the upper plate, and these were ombined with a frition oeffiient to alulate the right-lateral Coulomb shear stress hange for the whole of that region. For the five bak ar events in the Andreanof Islands the distribution of epienters in an ar-parallel line seems to indiate that they represent right-lateral slip on ar-parallel faults, and so it is suffiient to examine right-lateral Coulomb shear stress hanges on ar-parallel faults. However, for the July 4, 1966, and February 2, 1975, bak ar events in the Rat Islands, it is not lear whih of the two nodal planes is the atual fault plane. Hene we resolve the alulated stresses onto planes appropriate for eah of the possible fault plane solutions to examine both if the position and timing of the seismiity is onsistent with the oseismi stress hanges form the main event, and if we an resolve whih, if either, of the fault plane solutions is more onsistent with those hanges. Subdution in Indonesia is in the opposite sense to that in the Aleutians, with the Paifi plate subduting left-laterally roughly ENE to WSW underneath the Australian plate. Hene in order to alulate the Coulomb shear stress distribution in the upper plate, slip is imposed on the asperity region in the finite element model in the opposite sense to that used to produe the profiles for the Aleutians, i.e., the angle of obliquity p of Figure 5 is negative and hosen as p = -13 ø. The finite element analysis proeeds in the same way, but with a model with onstant angle of dip of only 11 ø and hanges in left-lateral Coulomb stress, Cr_xz =-Crxz + f Crxx, are alulated in the upper plate using a range of frition oeffiients, beause the relative plate motion now indues left-lateral shear along ar-parallel strike-slip features in the upper plate region. (We write the Coulomb shear stress hange Arns as ACr_xz here to signify that if we hoose the negative x diretion for strike-slip fault normal n, then s oinides with the positive z diretion.) Again, the left-lateral ar-parallel strike-slip ase an be identified with diretion n = -x and s = z in equation (1) or, equivalently, with n = x and s =-z, with the result denoted either as Cr_xz or Crx(_z). We use the former notation here. The February 18 aftershok shown in the upperplate to the east of the entroid of the main event in Figure 4 exhibits a left-lateral mehanism for an ar-parallel fault, but the preeding February 17 aftershok to the west (also shown in Figure 4) reveals a prinipally extensional mehanism. The diretion of extension in the mehanism is -30 ø antilokwise from the diretion of the trenh normal. Hene, in order to investigate whih regions of the upper-plate undergo extensional stress hanges in a diretion onsistent with the February 17 aftershok being due to slip in the main event, we alulate the extensional stress hanges resolved onto a plane perpendiular to a vetor S30øE from the trenh normal. We have also examined effets of small rotations of the n and s diretions from +x and-z, as observed for the February 18 event Field: Changes in Individual Aaxz, Aaxx, and A zz Components of Stress Before examining how the Coulomb shear stress hanges relate to eah of the three ases, it is instrutive to examine the

8 24,530 TAYLOR ET AL.: UPPER PLATE STRF3SING AND SEISMICITY response of the individual omponents of stress hange to plots extend from just beyond the trenh at x = 0 to the bak of understand how they govern the overall Coulomb stress the model (x ~ 3.25W) and for a distane z - 7W along-strike, distribution. Figure 6 shows the oseismi stress hanges with the asperity in the enter of the plot at z = 0. The alulated near the surfae of the upper plate for the three labeling on the X and Z axes should be interpreted as X = nonzero omponents of the stress field; the shear stress 100(x/W) km and Z = 100(z/W) km, respetively. The asperity, hange, Aaxz (Figure 6a), and the extensional stress hanges, where slip D is imposed (Figure 5), extends from 0.3W to 0.9W Aaxx (Figure 6b), and Aazz (Figure 6), from a 3-D finite downdip and from z =-0.5W to 0.5W along strike, and its element model with parameters harateristi of the Rat Islands projetion onto the surfae of the upper plate is indiated by a event, i.e., 0 = 20 ø, = 42'. This ase was hosen here as the retangle with dashed outline. Shaded regions indiate a most oblique of the three ases overed, and so learly shows derease in eah stress omponent, and unshaded regions the effet of obliquity on the resulting stress distributions indiate an inrease. The stress hanges are plotted in terms of ompared to those expeted form a purely dip-slip event. The nondimensional parameters of the model, and the differene in (a) x'x Derease in Oxz ', :. ", 20 ' 100 Trenh (b) -a5o -$oo -ese -eo eo -1 1 X (km)ao ø Position of Asperity / aozw Stress ontours: -- '-D-- = Diretion of Slip, D 42 ø , 0 (Ix z fi0 Trenh O0 - o o o oo sa 200 ' 00 1 Position of Asperity A (Ixx W Stress ontours: = 0.082!-tD Diretion of Slip, D 42 ø Oxx Figure 6. Coseismi hanges in individual omponents of the stress tensor for Rat Islands-like parameters, 0 = 20', = 42'. Along-trenh axis Z = 100(z/W) km, and trenh-perpendiular axis X = 100(x/W) km. Dashededge box indiates projetion of position of asperity onto surfae. (a) Change in right-lateral shear stress on ar-parallel faults, AO'xz. (b) Change in trenh-perpendiular normal stress, A xx. (b) Change in trenhparallel normal stress, A zz.

9 TAYLOR ET AL.: UPPER PLATE STRESSING AND SEISMICITY 24,531 () õ0,,,, ' Position of Asperity / A ( zz W Stress ontours: = IU, D / - 42 ø Diretion of Slip, D Ozz Figure 6. (ontinuexi) stress between neighboring ontours is expressed in terms of The distribution of oseismi Ao'zz (Figure 6) is broadly At W/!sD in eah figure individually. Note thathe zero stress split into inreased AO'zz for z < 0 and dereased Ao'zz for z > ontour (that separating shaded and unshaded regions) should be treated as an additional ontour, inluded to delineate the regions of positive and negative stress hange, and near areas of zero stress hange, the hange in stress Ao'W/!sD is atually the differene between the first ontour in the region of positive stress hange and the region of negative stress hange. The pattern of oseismi At xz shown in Figure 6a is fairly omplex but has peaks of greatest stress inrease at the trenh, above the right-hand end of the asperity (Z ~ 60 km), and greatest stress derease above the downdip end and left-hand edge of the asperity (Z ~ 0 and -60 km, respetively). For the 0, with the greatest inreases and dereases entered over the edges of the asperity at z = _+W/2. The effet of imposing oblique slip on the asperity has the same effet as seen for A xx and auses a rotation of the distribution abouthe y axis by an intermediate angle, about midway from 0 to q. While Figure 6 is drawn for q = 42 ø, it atually lets one estimate stress hanges for any value of q, at least for this partiular ase of dip, 0 = 20 ø. That is first beause the diagrams an be reinterpreted at one for q reversed in sign, to q = -42 ø in this ase. In suh transformation, q -.-q, the ontour lines of Figures 6b and 6 have a mirror refletion about Z = 0, whereas those of Figure 6a have a mirror refletion followed by a reversal of sign (regions of shear stress inrease beome derease and vie-versa). Then any general slip at an angle other than the q = 42 ø illustrated an be resolved, by vetor addition, into one slip along q = 42 ø and another along region x > W it separates into a broad regio.n of inreased rightlateral shear stress to the right of the asperity (z > 0) and inreased left-lateral shear stress to the left (z < 0). It is important to appreiate that shear stresses are indued in the upper plate even for pure dip slip on the asperity in the 3-D model beause of the edge effets from slip on a finite region of the thrust interfae. The pure dip-slip ase produes regions q = -42 ø. The stress hanges in response to eah separate slip an be obtained from Figure 6 and its transformation when q - -q and summed to get the total stress hange. of inreased (for z > 0) and dereased (for z < 0) shear stress that are symmetri in distribution, but opposite in sign, about the line z = 0. The introdution of oblique slip has rotated (about the y axis) the pattern in the diretion of ½ and also inreased. Comparison of 3-D Model Results to Upper the magnitude of derease in right-lateral shear stress, seen in Plate Seismiity Figure 6a. The oseismi At xx shown in Figure 6b exhibits the harateristi pattern of two peaks of inreased extensional Now we examine the alulated stress hanges from the 3-D stress, one over the top of the thrust interfae, and another, finite element modeling for the speifi parameters pertaining smaller one, in the bak ar (for x > W). The region of to the Andreanof Islands, Rat Islands, and Irian Jaya. inreased extensional stress is onentrated in a band of width I&l-- 2w running bak over the upper plate in a line oriented 5.1. roughly between the diretion of oblique slip, q, and the trenh normal with the first peak in stress entered over the asperity Andreanof Islands and the other, smaller peak, entered around x = 3W/2. To either side of this band the hange is minimally negative. Figure 7 shows the oseismi right-lateral Coulomb shear stress hanges, for ar-parallel faults in the upper plate, for

10 24,532 TAYLOR ET AL.' UPPER PLATE STRESSING AND SEISMICITY x (km) Derease in!jxz Inrease in (xz 2OO June '86 bak ar -- 15o seismiity loo 50 Trenh Position of Asperity Frition oeffiient: f- 0.4 A Oxz W Stress ontours' ø Diretion of Slip, D xz!q4 o - Oxz+ f (Ixx Figure 7. Coseismi hange in right-lateral Coulomb shear stress Ao'xCz on ar-parallel faults, for Andreanof Islands-like parameters, 0 = 20 ø, = 18 ø, and for frition oeffiient f = 0.4. parameters appropriate to subdution in the Andreanof Islands. The region of the upper plate for whih stresses are plotted, the axes and the shading should all be interpreted in the same way as for Figures 6a-6. Results were obtained for a wide range of for the same values as above for/.t and D. The five rosses ('x') in Figure 7 represent the orresponding position of the epienters of the June 5, 1986, bak ar events obtained by measuring the distane of the strike-slip seismiity alongfrition oeffiients, f, but Figure 7 shows the stress hanges strike in Figure 1 from the position of the asperity from the for f = 0.4 whih we show here as an midrange value for the purpose of omparison with subsequent figures for different inversion by Das and Kostrov [1990]. All five of the events lie within the lobe of inreased right-lateral Coulomb shear model parameters. The differene in stress between stress in a line spanning a range of stress inreases from 1 to 2 neighboring ontours is AO'xCz = 0.065/ D/W. The highest bars (again, for the same parameters for /.t and D). In other onentration of stress inrease is near the trenh, just to the studies, alulated stati stress hanges of the order of 0.1 bars right of the asperity, where the oseismi inrease in stress, or greater have been deemed suffiient to be onsistent with A O'xz _ max = 0.6 t. td/w ' whih for typial model parameters, triggered seismiity following a main event, e.g., in the outer say, W = 100 km, /.t = 30 GPa and D = 3m, is equivalent to rise following a main thrust event [Taylor et al., 1996] and A O'xz _ max -- 5 bars ' The general trend shows an inrease subsequent seismiity following a large strike-slip event on right-lateral Coulomb shear stress, onsistent with seismiity the San Andreas fault [Stein et al., 1992]. Here the alulated with right-lateral strike-slip mehanisms on ar-parallel faults, bak and to the right of the asperity and dereased stress hanges, suppressing suh seismiity, elsewhere. However the region of interest here is that whih orresponds to distane from the trenh of the June 1986 bak ar strike-slip stati Coulomb shear stress hanges are of the order of a bar and so seem to favorably support the interpretation of the position and timing of the June 1986 bak ar seismiity in the Andreanof Islands being attributed to Coulomb shear stress hanges indued by nonuniform slip on the thrust interfae of seismiity (Figure 1). For modeling the oseismi stress hanges due to slip on a disrete asperity we use the inferred position of the area of highest moment release of the 1986 main event from Das and Kostrov [1990] whih lies in the western half of the skethed aftershok zone (Figure 1) and is oriented roughly NW (between the trenh normal and angle of overall plate onvergene). With W = 138 km, measuring the orresponding distane from the trenh of the June 1986 bak ar seismiity from Figure 1 gives X = 130 km in Figure 7, and this is indiated by a dashed line. The maximum hange in Coulomb shear stress at that distane from the trenh is AO'xCz = 2.2x10-3/ D km -1 whih orresponds to about 2 bars the May 7, 1986, main event, with highest slip (an asperity) loated south and west of the subsequent bak ar seismiity. In addition to the inversion of the 1986 Andreanor Islands earthquake by Das and Kostrov [ 1990] that we use here, hosen as being the most detailed, there are several other inversions, namely by Hwang and Kanamori [1986], Boyd and NdbMek [1988], Houston and Engdahl [1989], and Yoshida [1992]. The inversions of Hwang and Kanamori, Boyd and NfiteVlek, Yoshida, and Das and Kostrov all show a region of highest moment release to the west of the epienter but differ on exatly where it lies. The Das and Kostrov inversion plaes the main asperity the farthest west, with that of Boyd and

11 TAYLOR ET AL.: UPPER PLATE STRESSING AND SEISMICITY 24,533 N fib lek about 50 km farther east (toward the epienter, but still about 30 km west of the down dip juntion in the skethed aftershok zone in Figure 1), and the main asperities from Hwang and Kanamori and Yoshida lie about midway between these. The inversion of Houston and Engdahl, however, plaes the main asperity just northeast of the epienter, and diretly trenhward of the June 1986 bak ar events. If instead of the asperity position from Das and Kostrov [1990], we were to take the most easterly position of the asperity, from Houston and Engdahl [1989], in order to find the relative loation of the five bak ar strike-slip events, this would move them to the left, parallel to the Z axis in Figure 7, and enter them around Z = 0. It is apparent that they would still all lie well within the zone of oseismially inreased right-lateral Coulomb shear stress, and so all of the inversions are onsistent with the above onlusion assoiating the position and timing of the bak ar seismiity with the stress hanges from the mainshok. EkstrOrn and Engdahl [1989] postulate that if the region of highest moment release was diretly trenhward of the bak ar strike-slip events, as suggested by Houston and Engdahl [1989], then a derease in ompressive stress on the ar-parallel fault, indued by slip in the main event, ould have been responsible for triggering suh seismiity. We agree with this possibility but here extend it by examining patterns of Coulomb shear stress hanges, thereby inorporating the effet of the shear stress hanges in addition to simply the extensional stress hanges. We are thereby able to show that the region of the thrust interfae experiening the highest slip in the main event that is onsistent with the bak ar seismiity is not restrited to lying just trenhward of the events but extends to the southwest also. Consequently, the position of highest moment release from all of the inversions are onsistent with the June 1986 strike-slip events Rat Islands, July 4, 1966, Event The main differene, for the generalized parameters examined here, between the Rat Islands and Andreanof Islands events is the angle of oblique slip p in the main event, around 42 ø for the February 1965 event ompared to around 18 ø for the May 1986 event (Figures 1 and 2). Figure 8a shows the oseismi hange in right-lateral Coulomb shear stress over the surfae of the upper plate for Rat Island-like parameters. The frition oeffiient used here f = 0.4 is to allow diret omparison with Figure 7 (for a smaller p). The general features of the plot, dereased AO'xz everywhere apart from a large peak of inreased AO'xz over and to the right of the position of the asperity and a lobe of inreased A O'xz strething bak and to the right from the trenh, are extremely similar in both plots. There is also very little differene in the inrement in stress hange between neighboring ontours, with AO'xz = 0.061pD/W for the Rat Islands parameters. The main differene arises in the absolute value of the stress hanges, the maximum stress inrease (above the asperity) being about 20% less for a given slip D, and the maximum stress derease being around!5% greater (more negative) in Figure 8a ( p = 42 ø) than in Figure 7 ( p = 18ø). The stress hange in the main lobe of inreased Coulomb shear stress is also muh greater in Figure 7, as indiated by the higher onentration of ontours of the same ontour spaing, then in x (km) 300 Derease in!jxcz Inrease in!j xz July 14, '66 bak ar _2_0_0 seismiity Trenh -: Position of Asperity Frition oeffiient: f = 0.4 Diretion of Slip, D 42 ø Stress ontours: A OxCz W gd {Jxz = {Jxz + f {Jxx Figure 8a. Coseismi hange in right-lateral Coulomb shear stress Islands-like parameters, 0 = 20 ø, p = 42 ø, and for f = 0.4. AO'xz on ar-parallel faults, for Rat

12 24,534 TAYLOR ET AL.: UPPER PLATE STRESSING AND SEISMICITY x (km) O Derease in!jxcz Inrease in {Jxz July 14, '66 bak ar seismiity 150 loo 50 Trenh Position of Asperity Frition oeffiient: f = 0.8 A axz w Stress ontours: = 0.072!.tD Diretion of Slip, D 42 ø {jxcz- {Jzx+ f Oxx Figure 8b. exept for f Coseismi hange in right-lateral Coulomb shear stress =0.8. At xz for parameters as in Figure 12a Figure 8a. These features are onsistent with the patterns of oseismi extensional and shear stress hanges indued for subdution in the sense found in the Aleutians and are plotted for parameters speifi to the Rat Islands in Figure 6. The linear ombination of At xz (Figure 6a) with AO'xx (Figure 6b) weighted by f produes the distribution of right-lateral Coulomb shear stress hange in Figure 8a. Inreasing the angle of oblique slip for the same slip D on the asperity inreases the omponent of slip along-strike and redues it,downdip in the main event. This auses an inrease in the nondimensional left-lateral shear stress hange -At xz, and a derease in the extensional stress hange, AO'xx over the surfae of the upper plate ompared to that for a smaller. Consequently, the hange in right-lateral Coulomb shear stress, At xz = At xz + fat xx, is more negative in those regions whih experiened inreases in left-lateral shear stress or less positive in those experiening inreases in rightlateral shear stress. This is born out by the hanges in Coulomb shear stress distribution observed between Figures 7 Coulomb shear stress) is only about 4% less in Figure 8b. This is attributable to the fat that, as outlined above, the extensional stress hanges are onentrated in two main peaks whih oinide with the largest hanges in right- and leftlateral shear stress hanges, and so they enhane the former and partially anel the latter when ombined in the rightlateral Coulomb shear stress distribution. The largest derease in right-lateral Coulomb shear stress then lies at around x = W, z = 0, between the peaks in inreased extensional stress, and so remains relatively unaltered by hanges in amplitude of extensional stress and hene frition oeffiient. As with the Andreanof Islands ase, we are most interested in the positions on the Coulomb shear stress plots whih orrespond to the subsequent bak ar seismiity in the region. Figure 2 shows the known bak ar strike-slip ativity, one event on July 4, 1996 (with mehanism from Stauder [ 1968b]), and a doublet on February 2, 1975 (with mehanisms from Newberry et al. [ 1986]). In the Andreanof Islands ase the five bak ar strike-slip events of the same mehanism strung and 8a. out in a roughly E-W diretion strongly support the Figure 8b shows the right-lateral Coulomb shear stress hypothesis that they are due to fight-lateral slip on ar-parallel hange for the same Rat Islands-like model parameters as for transforms and aommodate part of the slip defiit between Figure 8a but with a frition oeffiient off= 0.8. The general the diretion of plate onvergene and slip in the main event. stress distribution is extremely similar to that for f = 0.4, and However, for the isolated event of July 1996 in the Rat Islands the main differene is simply the inreased magnitude of the there exists ambiguity over whether it also represents rightright-lateral Coulomb shear stress hanges due to the extra "weighting" of the extensional stress inrease by the larger lateral slip on an approximately trenh-parallel fault, frition oeffiient. As a measure of this, the maximum right- onsistent with slip partitioning in a zone of oblique lateral Coulomb shear stress inrease is about 40% greater in onvergene, or instead represents left-lateral slip on an Figure 8b than in Figure 8a, where as the maximum derease approximately trenh-perpendiular fault (an interpretation (or, equivalently, the maximum inrease in left-lateral favored by Stauder [ 1968b]) as part of the rotation of the most

13 eastern of the tetoni bloks in the Rat Islands sequene, as subsequently haraterized by Geist et al. [ 1988]. Figures 8a and 8b show the Coulomb shear stress hanges on ar-parallel faults, Arrxz. A dashed line in Figures 8a and 8b at about x = 1.95W is inluded to indiate the relative distane from the trenh of the July 14, 1996, bak ar seismiity in terms of the nondimensionalized model parameters. A ross indiates the orresponding along-strike position of the epienter measured from Figure 2. From Figures 8a, 8b, and 2 it appears that right-lateral bak ar seismiity would be favored on ar-parallel faults at that distane from the trenh and along-strike of the asperity from the main event. This is indeed onsistent with the piture from Figure 2 where the July 14 event ours bak and to the east from the trenh and of the largest of the three asperities in the moment distribution from Bek and Christensen [ 1991 ]. Figure 9 shows the stress distribution required to examine the alternative interpretation of the fault-plane solution, i.e. the hanges in left-lateral Coulomb shear stress for ar- TAYLOR ET AL.: UPPER PLATE STRESSING AND SEISMICITY 24,535 perpendiular faults, Arr x, for the same model parameters and frition oeffiient (f = 0.4) as Figure 8a. Apart from a narrow region near the trenh for z < 0, the distribution of positive and negative regions of stress hange is very similar in Figures 8a and 9, with the main lobe of Coulomb shear stress inrease being bak and to the east of the asperity (albeit for different omponents of the stress tensor in the two ases). A diret omparison of the stress inrease at the orresponding position of the July 14 event (indiated by the ross) in the two ases reveals that in Figure 8a the inrease in right-lateral Coulomb shear stress, AO'xz_bak_ar - O.041.tD/W, and in Figure 9 the inrease in left-lateral Coulomb shear stress, AO' x_bak_ar = O.051.tD/W. Hene the Coulomb shear stress hanges from our model are equally onsistent with the July 14 bak ar event ourring on an ar-perpendiular transform fault, as proposed by Stauder [1968b]. Unfortunately, then, in this instane, examination of the relevant omponents of Coulomb shear stress hange does not allow us to distinguish between the two possible fault planes for the bak ar seismiity, although it does onfirm that slip on an isolated asperity in a main event would produe stati Coulomb shear stress hanges onsistent with the position, timing and mehanism of suh an bak ar event Rat Islands, February 2, 1975, Doublet These two events (Figure 2), a doublet, represent slip on faults that are neither ar-parallel nor ar-perpendiular. The more westerly of the two events ourred just over an hour after the other and was probably a diret result of the stress hanges indued by the first event. Consequently, we examine the mehanism of only the first, more easterly of the two events for onsisteny with the stress hanges indued by the 1965 mainshok. Measuring the angles of the two possible fault planes relative to the trenh from Figure 5 of Newberry et al. [1986] gives the two possible interpretations of the mehanisms as either aommodating left-lateral strike-slip motion on a vertial fault oriented roughly 70 ø antilokwise from the trenh, or right-lateral strike-slip motion on a vertial fault oriented roughly 20 ø lokwise (i.e., faults striking approximately NE and NW, respetively). The stress field alulated from the finite element model used to produe x (km) Derease in C Inrease in {jzcx July 14, '66 bak ar _ 0o seismiity 150 loo 50 Trenh Position of Asperity Diretion of Slip, D Frition oeffiient: f = 0.4 o AOzCx w Stress ontours: !-tD!Jzx = IJzx + f IJzz Figure 9. Coseismi hange in left-lateral Coulomb shear stress AO' x Islands-like parameters, 0 = 20 ø, p = 42 ø, and for f = 0.4. on ar-perpendiular faults, for Rat

14 24,536 TAYLOR ET AL.: UPPER PLATE STRESSING AND SEISMICtrY the Rat Islands-like parameter ontour plots in Figures 8 and 9 for ar-parallel and ar-perpendiular faults an be resolved onto a fault of any orientation for slip in any diretion on that fault by speifying a normal n and slip diretion s in equation (1). Figure 10 shows the oseismi hange in left-lateral Coulomb shear stress for vertial faults with n oriented 70 ø (measured positive in the same sense as ) from the trenh normal and s in the plane of the fault, as indiated in the bottom right of Figure 10, and Figure 1 l a shows the oseismi hange in right-lateral Coulomb shear stress for vertial faults with n oriented at -20 ø from the trenh normal, and s in the fault plane (in the opposite sense to Figure 10), again as indiated by the shemati drawing in the bottom right. Figures 11a and 1 lb are plotted for f= 0.4 as a representative midrange value of the frition oeffiient. The distane bak from the trenh of the epienters of the February 2 bak ar seismiity from Figure 2 is indiated on the plots by a dashed line at X km (using a value of W = 90 km). We interpret the Coulomb shear stress hanges at the position along-strike of the February 2 seismiity to be due to slip on the third (and smallest) of the three asperities indiated in Figure 2 (labeled C in the moment inversion), and the orresponding along-strike distane from the enter of the asperity is indiated by a ross (using the same value of W as the saling parameter). Figure 10 shows that the oseismi hange in right-lateral Coulomb shear stress on faults oriented at 70 ø from the trenh dereases markedly at this position. The stress derease at the ross is C ; ACrns / D km-1 whih for representative values of parameters, / = 30GPa and D = 3 m, orresponds to about 2 bars. Figure 1 l a shows that in the same region the oseismi hange in left-lateral Coulomb shear stress on faults oriented -20 ø from the trenh dereases to ArnCs =-0.001/ D km -1, whih represents a derease of about 1 bar for the same model parameters. From the oseismi stress hanges it appears that Coulomb shear stress hanges, with an f of 0.4, resulting from slip on the third (most westerly) asperity from the inversion of the February 4, 1965, main event in the Rat Islands would tend to suppresseismiity on either of the fault planes suggested by the February 2, 1975, bak ar events. However, the February 2 events ourred 10 years after the main event, and so the distribution of stress hanges in Figures 10 and 1 l a would have altered somewhat in that time due to any visoelasti relaxation of the asthenosphere and ontinued aseismi slip both downdip from and to either side of the asperity, whih presumably remains loked between large events, and from ongoing tetoni loading. Although it is a feasible extension of our modeling to alulate the flutuating stress field throughout an entire earthquake yle, as done for a 2-D model (homogeneous along-strike) by Taylor et al. [1996] and Zheng et al. [1996], here we have simply alulated the oseismi stress hanges. Suh modeling shows that the pattern of oseismi stress hange gradually reverses (positive to negative and vie versa) during the earthquake yle, at least in the idealized ase when eah yle periodially dupliates previous ones. This reversal ours linearly with time if a purely elasti response from the model is assumed and in a nonlinear fashion if a visoelasti, or other time-dependent response is introdued. It is likely that even 10 years represents only a relatively small fration of the yle time in the Rat Islands, and so the distribution of Coulomb stress hanges would still be positive and negative where shown as suh in Figures 10 and 1 l a, but the magnitude x (km) 3OO 250 Derease in ns C Inrease in 200 Feb. 2, '75 bak ar seismiity 50 loo Trenh Position of Asperity Diretion of Slip, D Frition oeffiient: f = ø ( nn Stressontours: Ao, s W 70 ø /.t D = ns = ns + f O'nn " ' Figure 10. Coseismihange in left-lateral Coulomb shear stress AO'ns on faults oriented at 70 ø antilokwise from the trenh, for Rat Islands-like parameters, 0 = 20 ø, q = 42 ø, and for f = 0.4.

15 TAYLOR ET AL.: UPPER PLATE STRESSING AND SEISMICrI 24,537 X (km) Derease in ljncs Inrease in ljns 200 Feb. 2, '75 bak ar seismiity 15ø loo Trenh Position of Asperity Frition oeffiient: f = 0.4 Diretion of Slip, D 42 ø Strike, Z (krn) A w Stress ontours: gd! ns =! ns + f! nn / ns Figure 11a. Coseismihange in right-lateral Coulomb shear stress AO'ns on faults oriented at 20 ø lokwise from the trenh, for Rat Islands-like parameters, 0 = 20 ø, q = 42 ø, and for f = 0.4. x (kin) Derease in IJnCs Inrease in lj ns 200 Feb. 2, '75 bak ar -- - seismiity 15o loo 50 Trenh Position of Asperity Diretion of Slip, D Frition oeffiient: f = ø nn A Ons W Stress ontours: = 0.083!-tD ( ns = ( ns + f ( nn 20 ns Figure 11b. Coseismi hange in right-lateral Coulomb shear stress A ncs on faults oriented at 20 ø lokwise from the trenh, for parameters as in Figure 14a exept for the frition oeffiient f = 0.8.

16 24,538 TAYLOR ET AL.: UPPER PLATE STRESSING AND SEISMICYI of the hanges would be diminished from their oseismi values. Although the magnitude of the Coulomb shear stress derease at the position of the February 2 bak ar events is less for the SE-NW striking plane (in Figure 1 l a), even 10 years after the main event, the derease in stress with both ases would still be a signifiant fration of a bar (for typial model parameters and an f = 0.4). Hene it is unlear whether the bakground stress field ould have inreased suffiiently over that time to ause a net inrease in Coulomb shear stress, and so we annot rule out the possibility that the bak ar seismiity was indued by stress hanges from ativity on other nearby faults, not from the main event. Another onsideration regarding the temporal variation in Coulomb shear stress hanges is the dependene of the distribution on f. Figures 10 and 1 la are both for a typial, midrange value f = 0.4. One possibility regarding the dependene of f on pore fluids in the fault zone whih has been raised [Simpson and Reasenberg, 1994] is that oseismially, fluids trapped in the rok result in a low effetive oeffiient of frition (e.g., f = 0.2 or lower) beause extensional normal stress hange indues a hange of opposite sign in pore pressure but that subsequently, as fluids drain away, the effetive oeffiient of frition inreases (e.g., up to f = 0.8). Figure llb shows the oseismi right-lateral Coulomb shear stress hanges for the same (SE-NW striking) fault plane as in Figure 1 l a, but with f = 0.8. The prinipal differenes in Figure l lb (ompared to Figure 11a) are the inrease in maximum stress hange, both above the asperity and in the main lobe bak and right from the trenh (indiated by more stress ontours of greater A(YnsW/#D ) and that the main region of inreased stress hange has rotated slightly about the y axis in the diretion of inreasing q. Both of these are due to the inreased weighting of the extensional stress hanges in the distribution. The orresponding position of the Feb. 2 bak-ar seismiity (marked by a ross) for f = 0.8 lies just inside the region of inreased stress, ompared to inside a region of dereased stress for f = 0.4 (Figure 1 l a). If indeed there is a time dependene in the effetive value of the oeffiient of frition, the propagation of the region of inreased Coulo nb shear stress as f inreases during the earthquake yle ould provide another mehanism for triggering of subsequent seismiity in the bak ar other than simply the oseismi response to the main event. It is important to note that the position of the last, third asperity is the most poorly onstrained in Bek and Christensen's [1991] inversion, whih is refleted in the unertainty in the relative position of the February 1975 events (ross) in Figures 11a and 1 lb. The ritial point is therefore not that the ross falls right on the border of inreased/dereased right-lateral Coulomb shear stress for an f = 0.8 but rather that the region of inreased right-lateral Coulomb shear stress expands westward with inreasing f (e.g., 0.2 to 0.8, Figures 11a and l lb, respetively). The results shown here are all from models whih use an asperity of the same downdip and along-strike dimensions, whih are haraterized in terms of a single parameter W. However, the inversion used by Bek and Christensen [1991] does not allow the determination the downdip extent of their asperities for the 1965 Rat Islands event (Figure 2). The along-strike length of the asperity in our model used to generate plots 6-13 is W. Changing this would affet the distribution and magnitude of the stress hanges and so raises the question of the validity of using the same model (and width of asperity) to interpret both the stress distribution for the July 4, 1966, bak ar seismiity (Figures 8 and 9) in assoiation with asperity A and the February 2, 1975, bak ar seismiity (Figures 10, 11a, l lb) in assoiation with asperity C. However, preliminary runs with an asperity of length 2/5 W show that the main differenes in stress distribution our only very lose to the trenh and are small and that bak from the trenh beyond z- W the distribution is almost idential to that for the larger asperity. In this region the magnitude of the stress hanges is dereased almost uniformly by a fator of -2/3 as a refletion of the response to the same magnitude of slip D being applied over a smaller area in the main event. However, even taking into aount the unertainty in position and size of the asperity, we annot definitively rationalize this doublet, 10 years after the Rat Islands event, as a onsequene of stress hanges in the event. 5.4 Irian Jaya For Irian Jaya, Figure 12 shows the oseismi hange in left-lateral Coulomb shear stress in the upper plate on arparallel faults for the same range of x and z as Figures 6-10, and for f = 0.4. The diretion of slip on the assumed asperity (z = 0) is indiated as q = -13 ø, and the ontours between lines of onstant stress hange are shown with separation A(r_xzW/#D = 0.136, with dereases in stress indiated by shaded regions and inreases by unshaded regions. The general features of Figure 12 losely resemble the plots of Coulomb shear stress hange on ar-parallel faults due to subdution in the opposite sense (Figures 6, 7a, and 7b for the Andreanof and Rat Islands, respetively) but with the pattern of stress hanges refleted about the line z = 0, as one would expet. The inrement in stress ontours is over twie that found in the plots for the Aleutians. Part of the reason is the shallower angle of obliquity, whih tends to deemphasize the hange in shear stress and to emphasize the hange in extensional stress, thus reating higher overall hanges in Coulomb shear stress, but the more dominant effet is from the shallow angle of dip in Irian Jaya (0 = 11 ø in the finite element model), whih means the stresses resulting from slip on the thrust interfae have a shorter distane in whih to attenuate before reahing the surfae. The lak of a seismi inversion for the Biak event at time of writing again leaves us without reliable values to fit to the parameters of the model, and so no dashed line appears in Figure 11 to indiate the position of upper plate seismiity as no definitive value is known for W, only the upper bound noted earlier. However, we would expet the upper plate events of Figure 4 to lie at least beyond the downdip end of the thrust zone, whih, for the 11 ø dip angle used in the finite element model, orresponds to X > 98 km. Figure 13 shows the extensional stresses indued by slip at q =-13 ø on the asperity resolved on a plane 30 ø antilokwise from the trenh (orientation shown in bottom right of Figure 13). Dereases in extensional stress are shown by shaded regions and inreases are shown by unshaded regions. The pattern of extensional stress hange for this orientation is almost exatly opposite of that of the left-lateral Coulomb shear stress hange in Figure 12, with a region of stress inrease bak and to the right of the asperity and dereases in stress everywhere else. Again the highest stress hange is loated right above the asperity, but the region we are interested in lies beyond X ~ 98 km.

17 TAYLOR Ef AL.: UPPER PLAID S1RESSING AND SEISMICITY 24,539 X(km) Inrease in r~xz Position of Asperity Frition oeffiient: f = 0.4 Diretion of Slip, D Stress ontours: ~ r~xz W = J.1D Figure 12. Coseismi hange in left-lateral Coulomb shear stress flu -Xl: on ar-parallel faults for Irian Jaya-Iike parameters, (J = 11, q> = -13, and for /= 0.4. X(km) Inrease in r nn Position of Asperity Diretion of Slip, D Stress ontours: = Figure 13. Coseismi hange in extensional normal stress flunn on faults oriented at 30 antilokwise from trenh for Irian Jaya-like parameters, (J = I I, q> = -13.

18 24,540 TAYLOR ET AL.: UPPER PLATE STRESSING AND SEISMICITY From omparison of the positions of the two upper plate events (Figure 4) and the orresponding alulated stress hange profiles from Figures 12 and 13, we an infer the approximate position of an asperity on the thrust interfae whih is onsistent with the position, mehanisms, and timing of these events. Figure 14 shows all the seismiity in the region from the various Harvard CMT atalogs, M > 5, from February 17 until July 1997, whih inludes some of the smaller aftershoks not reported immediately (and not shown in Figure 4) and also onsiderable upper plate seismiity. The earthquakes have been split into three groups: (1) those with prinipally thrust mehanisms onsistent with the orientation and shallow dip angle of the thrust interfae, shown along the base of Figure 14, (2) those with primarily extensional mehanisms, shown along the right of Figure 14, and (3) those with right-lateral strike slip or right-lateral strike slip with a groups, and so a tentative sketh of the position of suh an asperity is shown by the hathed ellipse in Figure 14 above where the thrust interfae appears to be, based on the position of the entroid and aftershoks. The fat that there are 13 events whih all seem to follow this pattern of stress distribution makes this explanation of an asperity loation fairly ompelling, and it will be of interest to examine any subsequent seismi inversions of the February 17 event, to ompare the distribution of moment release from suh an inversion with what we have inferred from the upper plate seismiity. The high tetoni omplexity of the upper-plate in the Biak area (Figure 3) may be the reason for the abundant seismiity following and preeding the 1996 main event. thrust Component whih are shown in the upper left of Figure 6. Disussion and Conlusions 14. Most of the upper plate seismiity exhibits distintly nondouble-ouple solutions and so these divisions are fairly general, although surprisingly distint none the less. It is apparent that the upper-plate extensional and strike-slip mehanisms 2 and 3 fall into distint groups relative to the position of the entroid of the main event, expliitly, the extensional events to the "right" (west) and the strike-slip events to the "left" (east). This is exatly onsistent with Our study of 3-D models with strongly heterogeneouslip reveals omplex distributions of both shear and extensional stress hange whih an be related to the edge effets of the finite region of largest slip. The distribution show a strong dependene on the various model parameters, 0,, W (Figure 5), and f. Inreasing the angle of dip 0 redues the magnitude of the stress hanges everywhere in the upper plate and ompresses the distribution toward the trenh. Inreasing the Figures 11 and 12 if slip ourred on an asperity between these slip obliquity angle inreases the weighting of the shear 138E 136E 135E D N A A B Figure 14. Bak ar seismiity in Irian Jaya, Indonesia, following the February 17, 1996, M w = 8.2 earthquake (February-July 1997).

19 TAYLOR ET AL.: UPPER PLATI- STRESSING AND SEISMICITY 24,541 stress hanges over the extensional stress hanges in the Coulomb shear stress hange distribution, as well as of left-lateral shear stress are reated to one side of an asperity, whereas stresses favoring more extensional faulting are inreasing the angle at whih the peaks in extensional stress indued on the other side. This pattern is exemplified by the inrease are oriented over the bak ar. Reduing the alongstrike length of the asperity from the total interfae width W has an almost negligible effet on the Coulomb shear stress distribution beyond z > W and simply redues the magnitude of the hanges uniformly in that region. Inreasing frition f has the opposite effet to that of inreasing ½ and inreases the weighting of the hanges in extensional stress over the shear stress hanges. postmainshok February 18 and 17 events (Figure 4) and subsequent seismiity up to the present (Figure 14) provided that the asperity is loated in the rupture zone between the positions of those two earthquakes. Suh a position of an asperity also seems onsistent with the loations of the 1979 premainshok earthquakes, whih are observed to be diretly adjaent to it (Figure 4). Those were apparently indued by the September 12, 1979 (M w = 7.5), The Coulomb shear stress distributions due to heterogeneous earthquake on the Yapen fault. The 1979 sequene has slip show qualitative onsisteny with the observations for ompressional omponents, and the late yle trenhthe Aleutians and Irian Jaya. Suh models predit that in perpendiular ompressive stress should indeed be greatest in presene of oblique slip direted like in the Aleutians, beyond the upper plate, diretly bak of a loked asperity [Dmowska et the downdip end of the thrust interfae, loalized zones of al., 1996b]. Also, the 1979 strike-slip earthquakes of Otober 16 and 17 have loations (Figure 4) somewhat to the west of right-lateral A xz on ar-parallel faults are reated postseismially just to one side of an asperity in the bak ar the abundant strike-slip ativity (Figure 14) ourring after the subdution event. We may reall that the gradual stress region and that A xcz is left-lateral (disouraging ativity) hanges during the earthquake yle must, approximately, be the same as the oseismi stress hanges, but with signs reversed. Thus the region in the upper plate under highest left- elsewhere (Figures 7, 8a, and 8b). We show results for f = 0.4 and 0.8 and for slip obliquity of 18 ø and 42 ø here, and ases that we studied with other obliquity angles and with f = 0.0, 0.2, or 0.6 gave similar features of loalized zones of right lateral A xz to one side of an asperity and left-lateral A xcz elsewhere (whih is good for robustness of preditions but unfortunately shows that strongly heterogeneous oupling will not provide any onstraints on f). Work addressing Coulomb stress interations in southern California [Stein et al., 1992; King et al., 1994; Harris et al., 1995; Harris and Simpson, 1996] also failed to definitively pik a best f to explain the data, and so typially 0.4 has been hosen as a representative value. The pattern in Figure 7 is onsistent with the position and timing of the postmainshok series of right-lateral strike-slip events in the Andreanof Islands (Figure 1) and the pattern in Figures 8a and 8b shows onsisteny with the July 1966 seismiity in the Rat Islands (Figure 2) if it ourred on an arparallel fault. Examination of the left-lateral Coulomb shear stress hanges on ar-perpendiular faults for the Rat Islands (Figure 9) shows that these are also onsistent with the position and timing of the July 1966 event. The Coulomb shear stress hanges resolved onto the two possible fault planes for the February 1975 seismiity in the Rat Islands show a derease in both ases for f = 0.4 (Figures 10 and 1 l a). However, the derease for the SE-NW striking plane (as proposed by Cormier [1975a, b]) is less in the same region than for the alternative, and a plot using an f = 0.8 (perhaps For the ase of upper plate seismiity following the 1996 Biak, Irian Jaya, event, we interpret all the earthquakes, premainshok and postmainshok, as aommodating the strike-slip part of the oblique subdution, with their positions in spae and time (before or after the mainshok) refleting the heterogeneous oupling along the main rupture zone, as modeled in 3-D for oblique subdution segments with asperities. Based on the spatial positions of all these earthquakes, we tentatively attempt to estimate the region of highest seismi slip (and the highest oupling) in the Biak 1996 earthquake, observing that models of strongly heterogeneous slip show that postseismially, loalized zones lateral Coulomb stress in 1979 should oinide with the region whih underwent oseismi derease of left-lateral stress (Figure 12) and was thereby put into a "stress shadow." Suh is onsistent with the relative loations of the premainshok and postmainshok strike-slip events in the upper plate. From the above limited overage of the upper plate seismiity and omparisons with our 3-D models of subdution earthquake yles with heterogeneous oupling along the interplate interfae we onlude that in the majority of ases the loation and mode of suh seismiity is onsistent with spae- and time-dependent stressing as predited by modeling. This onfirms earlier observations that seismiity in the viinity of large/great subdution earthquakes (in the outer rise, at intermediate depth and now in the upper plate) depends on time in the earthquake yle as well as on oupling patterns along the interplate interfae. If suh oupling patterns are roughly onstant at least from one yle to another, then the observations of this kind of seismiity ould be important in understanding earthquake yles and, perhaps, in antiipating slip distributions in future subdution events. Aknowledgments. Support for this work has been provided by USGS grant 1434-HQ-96-GR through the National Earthquake Hazards Redution Program and for studies of subdution zones outside U.S. territories, like for the 1996 Biak earthquake, initially by the George Merk Fund in Community Funds, In. through the New York Community Trust and subsequently by NSF Geophysis Program grant justified by pore pressure relaxation given the 10 year period EAR The ABAQUS finite element ode was made available sine the main event) atually reveals a very modest inrease under aademi liense from Hibbitt, Karlsson and Sorensen In., in Coulomb shear stress there. Pawtuket, Rhode Island. We thank G6ran Ekstr6m for many helpful disussions onerning upper plate seismiity along the Aleutians and providing seismiity plotting software, Gutuan Zheng for his stress ontour-plotting ode, and Eri Geist for ideas and disussions of the work on the Aleutians. We thank Emile Okal for bringing the February 1996 Biak earthquake to our attention and for subsequent disussions. We thank Rob MCaffrey, Colleen Stevens, and Fauzi for omments and data regarding that event. Figure 3 is reprinted from Puntodewo et al. [1994] with permission from Elsevier Siene. Referenes Astiz, L., T. Lay, and H. Kanamori, Large intermediate depth earthquakes and the subdution proess, Phys. Earth Planet. Inter 53, , 1988.

20 24,542 TAYLOR ET AL.: UPPER PLATE STRESSING AND SEISMICITY Bek, S.L., and D.H. Christensen, Rupture proess of the February 4, Jaum6, S. C., and L. R. Sykes, Change in the state of stress on the 1965, Rat Islands earthquake, J. Geophys. Res., 96, , southern San Andreas fault resulting from the California earthquake Boyd, T. M., and J. L. Nfib lek, Rupture proess of the Andreanof sequene of April to June 1992, Siene, 258, , Islands earthquake of May, 1986, Bull. Seismol. $o. Am., 78, King, G. C. P., R. S. Stein, and J. Lin, Stati stress hanges and the 1673, triggering of earthquakes, Bull. Seismol. So. Am., 84, , Chinnery, M. A., Seondary faulting, 1, Theoretial aspets, Can. J. Lay, T., L. Astiz, H. Kanamori, and D. H. Christensen, Temporal Earth Si., 3, , 1966a. variation of large interplate earthquakes in oupled subdution zones, Chinnery, M. A., Seondary faulting, 2, Geophysial aspets, Can. J. Phys. Earth Planet. Inter., 54, , Earth Si., 3, , 1966b. Minster, J. B., and T. H. Jordan, Present-day plate motions, J. Geophys. Christensen, D. H., and L. J. Ruff, Outer-rise earthquakes and seismi Res., 83, , oupling, Geophys. Res. Lett., 10, , Newberry, J. T., D. L. Lalair, and K. Fujita, Seismiity and tetonis of Christensen, D. H., and L. J. Ruff, Seismi oupling and outer-rise the far western Aleutian Islands, J. Geodyn., 6, 13-32, earthquakes, J. Geophys. Res., 93, , Puntodewo, S.S. O., R. MCaffrey, E. Calais, Y. Bok, J. Rais, C. Cormier, V. F., Tetonis near the juntion of the Aleutian and Kurile- Subarya, R. Poewariardi, C. Stevens, J. Genrih, Fauzi, P. Zwik, and Kamhatka ars and a mehanism for Middle-Tertiary magmatism in S. Wdowinski, GPS measurements of rustal deformation within the the Kamhatka Basin, Geol. So. Am. Bull., 86, , 1975a. Paifi-Australia plate boundary zone in Irian Jaya, Indonesia, Cormier, V. F., Composite foal mehanism solution of the Near Islands Tetonophysis, 237, , (Aleutian Islands) earthquakes of February 2, 1975 (abstrat), Eos, Rayleigh, C. B., J. H. Healy, and J. D. Bredehoefi, Faulting and rustal Trans. AGU, 56, 393, 1975b. stress at Rangely, Colorado, in Flow and Frature of Roks, Geophys. Das, S., and B. V. Kostrov, Inversion for seismi slip rate history and Monogr. Ser., vol 16, edited by H. C. Heard et al., pp , AGU, distribution with stabilizing onstraints: Appliation to the 1986 Washington, D.C., Andreanof Islands earthquake, J. Geophys. Res., 95, , Reasenberg, P. A., and R. W. Simpson, Response of regional seismiity to the stati stress hange produed by the Loma Prieta earthquake, Deng, J., and L. R. Sykes, Evolution of the stress field in the southern Siene, 255, , California and triggering of moderate-siz earthquakes: A 200 year Rybiki, K., Dependene of the spatial distribution of aftershoks on the perspetive, J. Geophys. Res., 102, , site of the disloation area at the time of the main earthquake, Publ. Dmowska, R., and L. C. Lovison, Intermediate-term seismi preursors Inst. Geophys. Pol. Aad. $i., 36, , for some oupled subdution zones, Pure Appl. Geophys., 126, 643- Rybiki, K., Analysis of aftershoks on the basis of disloation theory, 664, Phys. Earth Planet. Inter., 7, , Dmowska, R., and L. C. Lovison, Influene of asperities along Rybiki, K., T. Kato, and K. Kasahara, Mehanial interation between subdution interfaes on the stressing and seismiity of adjaent neighboring ative faults - stati and dynami stress field indued by areas, Tetonophysis, 211, 23-43, faulting, Bull. Earthquake Res. Inst. Univ. Tokyo, 60, 1-21, Dmowska, R., J. R. Rie, L. C. Lovison, and D. Josell, Stress transfer and Simpson, R. W., and P. A. Reasenberg, Earthquake-indued stati-stress seismi phenomena in oupled subdution zones during the hanges on entral California faults, U. S. Geologial Survey Prof. earthquake yle, J. Geophys. Res., 93, , Pap F, F55-F89, Dmowska, R., G. Zheng, and J. R. Rie, Seismiity and deformation at Smith, S. W., and W. Van de Lindt, Strain adjustments assoiated with onvergent margins due to heterogeneous oupling, J. Geophys. Res., earthquakes in southern California, Bull. $eismol. So. Am., 59, , , 1996a. 1589, Dmowska, R., M. A. J. Taylor, J. R. Rie, and E. A. Okal, Constraining Stauder, W., Tensional harater of earthquake foi beneath the slip heterogeneity of the February 17, 1996 Biak (Mw 8.2), Indonesia, Aleutian trenh with relation to sea floor spreading, J. Geophys. Res., earthquake from pre- and post-event seismiity, Eos, Trans. AGU, 77 73, , 1968a. (17) Spring Meet. Suppl., S 184, 1996b. Stauder, W., Mehanism of the Rat Island earthquake sequene of Ekstrtm, G., and E. R. Engdahl, Earthquake soure parameters and February 4, 1965, with relation to island ars and sea-floor spreading, stress distribution in the Adak Island region of the entral Aleutian J. Geophys. Res., 73, , 1968b. Islands, Alaska, J. Geophys. Res., 94, , Stein, R. S. and M. Lisowski, The 1979 Homestead Valley earthquake Engdahl, E. R., S. Billington, and C. Kisslinger, Teleseismially reorded sequene, California: Control of aftershoks and postseismi seismiity before and after the May 7, 1986, Andreanof Islands, deformation, J. Geophys. Res., 88, , Alaska, earthquake, J. Geophys. Res., 94, , Stein, R. S., G. C. P. King, and J. Lin, Change in failure stress on the Geist, E. L., J. R. Childs, and D. W. Sholl, The origin of summit basins southern San Andreas fault system aused by the 1992 magnitude = of the Aleutian Ridge: Impliations for blok rotation of an ar 7.4 Landers earthquake, Siene, 258, , massif, Tetonis, 7, , Taylor, M. A. J., G. Zheng, J. R. Rie, W. D. Stuart, and R. Dmowska, Harris, R. A., and R. W. Simpson, Changes in stati stress on southern Cyli stressing and seismiity at strongly oupled subdution zones, California faults after the 1992 Landers earthquake, Nature, 360, J. Geophys. Res., 101, , , Wu, F. T., and H. Kanamori, Soure mehanism of February 4, 1965, Harris, R. A., and r. W. Simpson, suppression of large earthquakes by Rat Island earthquake, J. Geophys. Res., 78, , stress shadows: A omparison of Coulomb and rate-and-state failure, Yoshida, S., Waveform inversion for rupture proess using a non-flat J. Geophys. Res., this issue. seafloor model: Appliation to 1986 Andreanof Islands and 1985 Harris, R. A., and R. W. Simpson, In the shadow of 1857-Effet of the Chile earthquakes, Tetonophysis, 211, 45-59, M8 Ft. Tejon, California, earthquake on subsequent earthquakes in Yu, G., S. G. Wesnousky, and G. Ekstrtm, Slip partitioning along major southern California, Geophys. Res. Lett., 23, 3, , onvergent plate boundaries, Pure Appl. Geophys., 140, , Harris, R. A., R. W. Simpson, and P. A. Reasenberg, Influene of stati stress hanges on earthquake loations in southern California, Nature, Zheng, G., R. Dmowska, and J. R. Rie, Modeling earthquake yles in 375, , the Shumaginsubdution segment, Alaska, with seismi and geodeti Houston, H., and E. R. Engdahl, A omparison of the spatio-temporal onstraints, J. Geophys. Res., 101, , distribution of moment release for the Andreanof Islands earthquake with reloated seismiity, Geophys. Res. Lett., 16, , R. Dmowska, J. R. Rie, and M. A. J. Taylor, Division of Engineering Hubbert, M. K. and W. W. Rubey, Role of fluid pressure in the and Applied Sienes, Harvard University, 327 Piere Hall, 29 Oxford mehanis of overthrust faulting, Geol. So. Am. Bull., 70, , Street, Cambridge, MA ( taylor@esag.harvard.edu) Hwang, L. J., and H. Kanamori, Soure parameters of the May 7, 1986 Andreanof Islands earthquake, Geophys. Res. Lett., 13, , (Reeived July 30, 1997; revised January 23, 1998; aepted February 26, 1998.)