Tectonics of the Juan Fernandez Microplate and Nazca-Antarctic Plate Boundary

In the first part of this study, magnetic and bathymetric data from an R/V Endeavor survey are combined with earthquake mechanism studies to produce a tectonic model for the Juan Fernandez microplate at the Pacific-Nazca-Antarctic triple junction. Using rate data from the East Ridge of the microplate, a Nazca-Juan Fernandez best fit pole was determined. As is suggested by the fanning of anomalies, this pole is just north of the ridge. The other two poles were determined by closure. Plate motion data for the four-plate system were inverted to define present-day plate motions and rotation poles. The resulting tectonic model predicts that the northern boundary of the Juan Fernandez microplate is a zone of compression and that the West Ridge and southwestern boundary are spreading obliquely. The southeastern boundary is predieted to be obliquely divergent, but present data fail to constrain its geometry. A schematic evolutionary model was also constructed to illustrate the migration and evolution of the triple junction, and relates the formation of the Juan Fernandez microplate to differential spreading rates at the triple junction. Possibilities for the future evolution of the four-plate system are included as part of the model. In the second part of this study, bathymetric data collected by the Endeavor were used to construct a new contour map of the Chile transform 0 system, a major part of the Nazca-Antarctic boundary, from 100 w to its termination at the East Ridge of the Juan Fernandez microplate at 34°30'S, l09°15'W. Geophysical data from along this boundary had been extremely limited until the Endeavor survey. A generally continuous lineated bathymetric trend can be followed through the entire region,

In the first part of this study, magnetic and bathymetric data from an R/V Endeavor survey are combined with earthquake mechanism studies to produce a tectonic model for the Juan Fernandez microplate at the Pacific-Nazca-Antarctic triple junction. Using rate data from the East Ridge of the microplate, a Nazca-Juan Fernandez best fit pole was determined. As is suggested by the fanning of anomalies, this pole is just north of the ridge. The other two poles were determined by closure.
Plate motion data for the four-plate system were inverted to define present-day plate motions and rotation poles. The resulting tectonic model predicts that the northern boundary of the Juan Fernandez microplate is a zone of compression and that the West Ridge and southwestern boundary are spreading obliquely. The southeastern boundary is predieted to be obliquely divergent, but present data fail to constrain its geometry. A schematic evolutionary model was also constructed to illustrate the migration and evolution of the triple junction, and relates the formation of the Juan Fernandez microplate to differential spreading rates at the triple junction. Possibilities for the future evolution of the four-plate system are included as part of the model.
In the second part of this study, bathymetric data collected by the Endeavor were used to construct a new contour map of the Chile transform 0 system, a major part of the Nazca-Antarctic boundary, from 100 w to its termination at the East Ridge of the Juan Fernandez microplate at 34°30'S, l09°15'W. Geophysical data from along this boundary had been extremely limited until the Endeavor survey. A generally continuous lineated bathymetric trend can be followed through the entire region, i i with the transform valley being relatively narrow and well-defined from 109°w to approximately 104°30'W. The fracture zone-parallel topography then widens eastward, with at least two probable en echelon offsets to 0 0 the south at 104 and 102 W. This bathymetric data, along with new earthquake mechanism data from the transform system and additional data from the Nazca-Antarctic boundary to the east, have been compiled into a new, larger data set for this boundary. Inversion of this data set has produced a new best fit pole for the Nazca-Antarctic plate pair, providing better constraints on the relative motion along this boundary. This new best fit pole matches the data better than the best fit pole calculated from old data, though some discrepancies still remain. Additional data, particularly from the western portion of the boundary, are needed to further refine the pole location.  predicted to be obliquely divergent, but present data fail to constrain its geometry. We also present a schematic evolutionary model which relates the formation of the Juan Fernandez microplate to differential spreading rates at the triple junction, and discuss related complexities. The future evolution of the four-plate system depends on whether one or both ridges remain active.

INTRODUCTION
The existence of a small plate at the intersection of the Nazca, Antarctic, and Pacific plates, just north of the Chile Fracture Zone ( Figure 1), was proposed independently by Herron [1972a] and Forsyth [l972] on the basis of anomalous seismicity. Earthquake epicenters outline the region, called the "lower plate" by Herron [1972a], and fault plane solutions  are inconsistent with the expected Pacific-Nazca relative motion. Stover [1973] noted that the epicenter 0 0 distribution from 31 to 35 s could be interpreted to include one or two small plates, although plate boundaries could not be clearly identified.
He concluded that this anomalous seismic activity is the result of a reorientation of the ridge axis. Additional fault plane solution data by Anderson et al. [1974] et al. [1974], based mainly on focal mechanism studies because of the extremely limited marine geophysical data existing at the time, has since been modified significantly as more geophysical data became available [Engeln and Stein, 1984;R. N. Hey et al., unpublished manuscript, 1985;D. F. Naar and R. N. Hey, unpublished manuscript, 1985]. In this paper, we present magnetic and bathymetric data from the R/V Endeavor survey and focal mechanism studies for earthquakes on two of the Juan Fernandez microplate boundaries. We then combine these data and invert them using the technique of Minster et al. [1974] to define present-day plate motions and rotation poles. Finally, we propose a present-day tectonic model and a possible evolutionary model for the region consistent with our data, and discuss some of the complexities of this area.

EARTHQUAKE STUDIES
The seismicity of the Chile Rise and Chile Fracture Zone defines the Nazca-Antarctic plate boundary, with the Juan Fernandez microplate indicated by a ring in the seismicity at the western end of the Chile Fracture Zone (Figure 1). We studied four earthquakes (Figures 1, 2, and 3) on two boundaries of the Juan Fernandez microplate to provide data on present motions • . In addition, we studied six events on the Chile Fracture Zone to provide additional constraints on Nazca-Antarctic motion, a key to Juan Fernandez tectonics. Details of these six events can be found elsewhere [Engeln, 1985; s. Anderson-Fontana et al., unpublished manuscript, 1985]. All events were studied using P wave first motions and Rayleigh wave spectral amplitudes. For the Juan Fernandez events, P and SH waveform modeling [Fukao, 1970;Langston and Helmberger, 1975;Kanamori and Stewart, 1976] using the algorithm of G. c. Kroeger and R. J. Geller (unpublished manuscript, 1985) was used.
The parameters of the four Juan Fernandez events are listed in Table 1.
Three of these events are nearly pure strike-slip. Only the April 24, 1972, event has a large dip-slip component. 5 Two events large enough for detailed analysis occurred on the western boundary of the Juan Fernandez microplate. One (December 29, 1966) was studied by Anderson et al. [1974], who proposed a mechanism combining normal and strike-slip components. Our results ( Figure 2) indicate nearly pure strike-slip motion. The first motion polarities analyzed in this study allow only determination of the general trend of the nodal planes, but SH wave analysis provides stricter constraints on the mechanism. Some of the results from the P and SH wave modeling are

MAGNETIC ANOMALY DATA
Total intensity magnetic measurements were made aboard the R/V Endeavor in the region around the Pacif ic-Nazca-Antarctic triple junction. These data were reduced to anomaly form by subtracting the 1980 International Geomagnetic Reference Field [Peddie, 1982] from each data point. The magnetic anomaly data, along with the bathymetric data, were then merged with navigational data and plotted along the ship's tracks. In order to improve the regional field manipulations, residual trends were removed from selected anomaly profiles which were then interpolated to a 0.5 km data spacing for subsequent projection. 6 The locations of the magnetic anomaly profiles we examined are shown in Figure 4. We present the data in Figure 5, along with a synthetic profile for comparison based on approximate spreading rates from profile A. These profiles have been projected to an azimuth of 097°, normal to the trend of the ridge axis. The location of the central anomaly can be clearly seen on these profiles, but older anomalies become increasingly difficult to identify towards the north. This is apparently due to the rapid northward convergence of the magnetic lineations, which is more extreme on the west flank than on the east. To further support this interpretation of northward convergence, we illustrate in Figure 6 the three southernmost magnetic anomaly profiles (A, B, and C) with their respective synthetic profiles. Note the good correlation between the synthetic and actual profiles in each case. If there were a constant spreading rate along the ridge, the model for profile A should also match profiles B and C. However, examination of the data and model profiles show that this is not the case, and that models with decreasing spreading rates northward are necessary to match the data. The converging lineations produced imply rotation about a nearby pole.
We measured approximate half-opening rates averaged from anomaly 2 to the present for the east and west flanks. These are unusually large, 0 0 10.4 /m.y. for the west flank and 6.9 /m.y. for the east flank. These rates are more than twice the rates calculated for the north and south flanks of the Magellan spreading system, where a nearby Euler pole also produced asymmetric, fanning anomalies [Tamaki et al., 1979], and more than 10 times the rotational half-opening rates for the Pacific of 0.5° 0 to 0.69 /m.y. [Lancelot and Larson, 1975]. such asymmetric and extremely rapid rotation rates may suggest instability of the spreading system [Tamaki et al., 1979].  Rea, 1977Rea, , 1981Hey et al., 1985]. These two profiles show an increase in rates on the west flank during the Jaramillo to Brunhes-Matuyama time interval, followed by a decrease to the present time. A similar fluctuation was noted by Rea [1977] on the 0 0 EPR at 31 S, but confined to the east flank of the rise, and at 9.5 to 12°s on the west flank only [Rea, 1976a]. Present whole spreading rates on the East Ridge decrease northward from 5.5 cm/year. North of profile B, however, our inability to identify anomalies younger than anomaly 2 prevents calculation of present spreading rates.  [Rea, 1977]. The discontinuity of anomaly J at 31°s ( Figure 4) may suggest complications in the spreading process at that time [Rea, 1977[Rea, , 1981.
on the Pacific-Antarctic Rise directly south of the region, Handschu-9 macher [1976] estimated whole rates at 10 to 10.5 cm/year, and identified anomalies out to 5A on the Nazca plate at about 33°s, 103°w ( Figure   4). These anomalies correlate and are continuous with the newly identified anomalies produced by the East Ridge, implying that this ridge has been spreading for at least 11.5 m.y.

BATHYMETRIC DATA
The bathymetric profiles across the East Ridge have also been However, since analysis of the magnetic anomalies does not show shifting or jumping of the ridge axis at those locations ( Figures 5 and 6), we feel that the location of the spreading center is accurately shown by the dashed line in Figure 8. The significance of the depressions greater than 4000 m in depth on the west flanks of profiles c and D cannot be determined based on our present data set. Similar depressions of uncertain origin have also been mapped along the EPR at 6°s [Rea, 1976b]. In addition, a large depression can be seen on the west flank near the ridge tip (profile F). The maximum width of this depression is approximately 35 km, with a local relief of 3200 m and a maximum water depth of approximately 5100 m. It is a localized feature, is teleseismically quiet, and has no apparent magnetic signature. The tectonic significance of this depression and its location relative to the spreading axis are unclear. Topographic deeps have been observed both at propagating rift tips and at ridge-transform intersections, but are generally not of this magnitude [Macdonald et al., 1979;Fox and Gallo, 1984;D. F. Naar and R. N. Hey, unpublished manuscript, 1985]. A depression of similar dimensions has, however, been mapped near the northern tip of the East Ridge on the Easter microplate Hey et al., 1985]. Another curious zone of lineated topography was identified on these same profiles north of the lineated feature described above ( Figure 9).
This zone appears to have an arcuate shape and consists of an apparently linear depression 3500 to 4500 m in depth with an associated high directly south 1300 to 2000 m in depth. The significance of this feature is unknown, although its lack of seismicity implies that it may be a relict feature. Finally, crossings of the supposed southeast boundary, also seen in Figure 9, indicate relatively smooth topography in this region.
PRESENT-DAY TECTONICS

12
The seismicity ( Figure 1) near the Juan Fernandez microplate is clustered in a ring which defines the supposed plate boundaries, suggesting that the nearly aseismic area in the center of the plate is behaving rigidly. To test whether rigid plate tectonics adequately describes the region, we inverted the relative motion data for the fourplate (Pacific-Nazca-Antarctic-Juan Fernandez) system using the Minster et al. (1974] algorithm. The data, listed in Table 3, included the data used in model RM2  for the Pacific-Nazca and Pacific-Antarctic boundaries (minus those slip vectors now known to be on the Pacific-Easter and Pacific-Juan Fernandez boundaries). rate data along the Juan Fernandez East Ridge from this study, and slip vectors from the four Juan Fernandez events. We also used a new data set for the Nazca-Antarctic boundary which includes the RM2 data, slip vectors from Engeln (1985], transform azimuth and rate data from Herron et al. (1981], and EN-112 transform azimuth data. Engeln (1985] showed that these data produced a best fit pole much closer to the RM2 Nazca-Antarctic pole which matched the data much better than a best fit pole produced using only RM2 data. using only the rate data from anomaly 2 to the present along the East Ridge and assuming orthogonal spreading, we determined a Nazca-Juan Fernandez best fit pole. As is suggested by the fanning of anomalies, this pole is just north of the ridge (Table 4; Figure 10, Four inversions were carried out using different subsets of the Juan Fernandez data (Table 4). we assigned the rate data uncertainties of 50 0 mm/year and the slip vectors on the Juan Fernandez boundaries 20 uncertainties. The poles vary only slightly based on which data were used. Poles determined using the spreading rates from anomaly 2 (the most recent that can be identified on all three of the profiles used for the rate data) to the present are shown in Figure 10. The dots indicate the poles determined by an inversion in which the only data from the Juan Fernandez microplate were the rate data from the East Ridge. The stars indicate the poles determined by inversion of the entire data set, With their 95% confidence limits indicated by the ellipses; these poles were used to generate the relative motion vectors shown. The good fit of the data (summarized in Table 3) offers some confidence in the locations of the rotation poles for the four plate system. If the Nazca-Juan Fernandez pole were at a great distance, it would be inconsistent with the decrease in spreading rate along this boundary or the slip vectors on the other two Juan Fernandez boundaries. The small 95% confidence ellipses shown in Figure 10 indicate the internal consistency of the data, although they underestimate the true uncertainty in these poles' determinations. One possible problem is presented by the two earthquakes along the northern plate boundary; if this is a broad zone of deformation, the compressional axes may be more important to the tectonic interpretation than the slip vectors of these events.
The spreading rate along the East Ridge increases almost linearly from north to south because of the proximity of the pole. This pole's location requires that most of the northern boundary of the Juan Fernandez plate be in compression unless the boundary geometry is very different from that shown. The rough topography of this region may be a result of this compression, with the young age of the lithosphere probably preventing subduction (or obduction) from occurring.
The West Ridge is predicted to be a fast spreading ridge (greater than 100 mm/year), but must be spreading obliquely if the slip vectors from the earthquakes along this boundary and the interpretation of ridge strike J. Francheteau, personal communication, 1983] are correct. This oblique spreading, along with the rough topography to the west of the ridge observed on the two EN-112 ship cross-i S would most likely contribute to the lack of identifiable magnetic ng , anomaly patterns in this area. The transforms shown are schematic and are not bathymetrically defined; they have the predicted trend and the correct sense of offset, but their lengths are not well constrained.
The southwestern boundary is also an obliquely spreading ridge if the 15 trend mapped from EN-112 data indicates the boundary, or is a series of ridge segments offset by transforms. The remaining boundary, Antarctic-Juan Fernandez, is predicted to have slow, divergent motion across it.
However, the details of its ~geometry are totally unconstrained.
rt is important to recognize that this tectonic model is limited by the sparse data. Nonetheless, it is internally consistent and testable, facilitating future investigations.

DISCUSSION
Given the limited data along the Juan Fernandez microplate boundaries, only a general outline of its evolution is possible at present.
The age continuity of the magnetics to the east of the East Ridge ( propagation, or if the ridge was formed far from the East Ridge, this is an upper limit to its age. The question arises as to whether these topographic boundaries represent pseudofaults created by ridge propagation [e.g., Hey, 1977;Hey and Wilson, 1982]. We feel this is not the case for the East Ridge for unless the spreading is extremely asymmetric on the East Ridge, this ridge will move eastward relative to the EPR. Because the asymmetry on the East Ridge is not large, the inception of spreading on a West Ridge presumably caused the eastward migration of the East Ridge relative to the EPR (as the simplified model in Figure 11 shows).
At the Pacific-Nazca-Antarctic triple junction, relative motion is essentially collinear, with Pacific-Nazca much faster than Pacific-Antarctic motion ( Figure 1). As spreading is nearly symmetric, the Pacific-Nazca Ridge moves eastward with respect to the Pacific-Antarctic

19
It is interesting to consider three of the many paths the system might follow in its future evolution ( Figure 12). In one case ( Figure   12, top), the microplate continues to grow as a separate entity. In another case ( Figure  Another possibility, if neither ridge dies, is that the Juan Fernandez microplate will grow into a major plate. Hilde et al. [1976] suggested that the Pacific plate was formed in a somewhat similar manner in the late Jurassic.
Understanding the Juan Fernandez plate should provide crucial insights into past and present major plate boundary reorganizations.
Evidence for microplates similar to the Juan Fernandez and Easter plates appears to be preserved in marine magnetic anomalies. For example, between 8.2 and 6.5 million years ago, a small plate existed between the EPR and the then-active Galapagos Rise [Rea, 1978;Mammerickx et al., 19 80]. After the extinction of the Galapagos Rise, this microplate became part of the Nazca plate. Cande et al. [1982] proposed that a plate which existed at the      weighted residuals and importances are those for the inversion using the new poles derived in this study and in the work by Engeln [1985]. Rates are in centimeters per year. Transform fault and slip vector azimuths are in degrees measured counterclockwise from east. Sources: RM2, compiled by Minster and Jordan [1978]; H, Herron et al. [1981]; E, EN-112 (1984 survey); T, this study; F, fictitious transform (assuming orthogonal spreading).   at the Pacific-Antarctic-Nazca triple junction.   Herron [1972b], Handschumacher [1976], and Rea [1977].  to the south at 104° and l02°w. We also present six new strike-slip mechanisms along the Chile Transform and one normal fault mechanism near the northern end of the Chile Rise. We have compiled a new, larger relative motion data set for the Nazca-Antarctic boundary consisting of these new earthquake and bathymetric data in addition to other data from the eastern portion of the boundary. We have inverted these data to produce a new best fit pole for the Nazca-Antarctic plate pair, providing tighter constraints on the relative plate motions. This new best fit pole is located farther south and matches the data better than the best fit pole calculated from old data, though some discrepancies still remain. Additional data, particularly from the western portion of the boundary, are needed to further refine the pole location and improve the agreement between measured and predicted relative motions along the entire Nazca-Antarctic boundary.

INTRODUCTION The Chile Fracture Zone in the southeastern Pacific Ocean has been
Of the most poorly surveyed sections of the mid-ocean ridge system. one It is a major part of the Nazca-Antarctic plate boundary, extending from 0 0 the Chile Ridge at about 36 S, 97 30'W to the East Ridge of the recently surveyed Juan Fernandez microplate [l, 2) at 34°30'8, l09°15'W, a distance of approximately 1300 km (Fig. 1). The geometry of this boundary and the relative motion there have been poorly constrained due to the lack of sufficient geophysical data. Klitgord et al. [3] estimated the The limited bathymetric data do not define a lineation direction for the Chile Fracture Zone, but the locations of earthquake epicenters show an overall ESE trend (Fig. 1). However, a limited number of slip directions determined from focal plane mechanisms by Forsyth [5) and Anderson et al. [6] indicate a different slip direction of approximately ENE.
These slip directions conflict with those predicted by the world-wide Plate motion model of Minster and Jordan [7], while the trend of earth-quake epicenters from the western portion of the transform (Fig. 1) is nearlY parallel to the predicted motion.  Fox, personal communication, 1985;D. Gallo, personal communication, 1985).
No topographic deep or nodal basin is observed at the ridge-transform intersection, though some deepening of the East Ridge's rift valley to 0 0 its intersection with the fracture zone at 34 30'S, 109 30'W is probable (<200m).
In Figure 3, Of 102°w further implies the presence of the offset in the transeast form there that was ambiguous in the bathymetry. 0 East of 100 W, the available bathymetric data consist only of a few very widely spaced track lines, making it unrealistic to contour the fracture zone to its 0 intersection with the Chile Rise at -97 30'W. However, the earthquake epicenters continue to follow the projected trend of the fracture zone through this region (Fig. 1).

EARTHQUAKE STUDIES
The seismicity of the Chile Rise and Chile transform system also 66 defines the Nazca-Antarctic plate boundary. New mechanisms were determined for six strike-slip earthquakes along the Chile Transform and one normal fault mechanism near the northern end of the Chile Rise (Fig. 1, Table 1). All events were studied using P wave first motions and Rayleigh wave spectral amplitudes. Figure 5 shows the first motions and surface wave amplitude patterns for three of these events. Though the few stations located to the south limit the accurate determination of the fault planes, first motion polarities from all four quadrants were observed, indicating strike-slip motion. For all three events, the north-south nodal plane could be determined using the first motions alone, and both planes of the April 14, 1979, earthquake are constrained. We interpret Rayleigh wave radiation patterns as four-lobed, suggesting strike-slip motion on either north-south or east-west trending fault planes. Data for the other three Chile Transform earthquakes are shown in Figure 6. The September 13, 1965, andJuly 12, 1979, events are similar to the other three. However, the strike of the June 26 , 19 74, earthquake appears to be somewhat different from that of the other events. This apparent difference may be due to poor station coverage.
we also obtained a normal fault mechanism for one event which occurred near the northern end of the Chile Rise (Fig. 7). First motions partially constrain the steeply dipping nodal plane, and the two-lobed Rayleigh wave radiation pattern indicates primarily dip-slip 0 motion on a plane trending 335 • The strike-slip mechanisms show a discrepancy with the apparent trend of the Chile transform system. The east-west nodal plane strike is noticeably more nortberly than expected from the morphology. This situation suggests that either a "leaky" component of extension may be present, though no such events were observed, or the region consists of a series of en echelon transform faults [5,8], though the gross bathymetry suggests only two offsets, both east of l05°w. Unfortunately, no 0 events west of 102 30'W were large enough to be studied. However, it may be significant to note that the slip vector for the event at l02°3o'w, calculated to be 91° (Table 1), is close to the measured transform azimuth at that location, which is between 93° and 91°. This may suggest that the discrepancy between slip vector direction an d transform (and epicenter) trend is not as great as previously believed [5,6]. There is clearly a need for additional slip vector data west of 0 102 W to test this hypothesis.

TECTONICS OF THE NAZCA-ANTARCTIC BOUNDARY
The relative motion at the Nazca-Antarctic boundary has been poorly constrained, largely due to a lack of data along this boundary. We have compiled a new, larger data set which includes, in addition to RM2 data [ 7 ], magnetic and transform azimuth data from Herron et al. [4] and slip vector and transform azimuth data from this study. These data are listed in Table 2  approximately 60 mm/year, which is slower than the 76 mm/year rate for this ridge calculated by Klitgord et al. [3], whose data were averaged over the past 5 m.y. This discrepancy is in accord with Herron et al.'s [4] suggestion that spreading has slowed with time, implying that 76 mm/year may be an overestimation of present spreading rates. We thus elected to use the lower rates determined from Herron et al.'s [4] data to include in our data set. using the Minster et al. [10] algorithm, we inverted this new relative motion data set for the Nazca-Antarctic boundary ( Table 2) to determine a new best fit pole for this plate pair. The results of our inversions and a comparison with previous results are summarized in Table 3 and are shown in Figures 8 and 9. Since Minster and Jordan [7] did not publish a best fit pole for this plate pair, we used their data set and algorithm to produce one (Fig. 8, solid circle). The difference between the pole for this plate pair from global model RM2 [7] and the best fit pole produced using their data and algorithm indicates how poorly motion along this boundary was constrained. Our new pole (Fig. 8,star) is closer to the RM2 pole for this plate pair and matches the data better than the previous data's best fit pole (Fig. 9). This suggests that the larger data set is improving the determination of Nazca-Antarctic relative motion, and also that the world data set in RM2 was adequate to model reasonably accurately the Nazca-Antarctic pole location, even though relative motion data from that plate boundary was clearly inadequate when used by itself. Figure 9 summarizes the data and the predictions of the new best fit pole and the other poles. The differences between the predictions of RM2 and the best fit pole produced using Minster and Jordan's [7] data are very pronounced. In particular, the rates predicted by the old best fit pole fall well outside the standard deviations for the observed rates. Also, there is a clear divergence from east to west between the azimuths predicted by the old best fit pole and those predicted by the other poles and measured from the data. This divergence is also evident in Figure 10.
The predictions of the new best fit pole provide a much better fit overall to the data, indicating that our pole is an improvement over that produced by the old data. However, the quality of the fit to the west is apparently poorer than that to the east for our new pole. Also shown in Figure 9 are the Nazca-Antarctic predicted relative motions based on the Pacific-Nazca-Antarctic-Juan Fernandez four-plate inversion from . These predictions are close to those for our new best fit pole, particularly for the azimuths, but don't fit the rate data as well as the new pole.
In Figure 10   East of 104 W, the trends predicted by our new best fit pole are apparently closer to the bathymetric trend than those predicted by the other poles. However, west of 0 104 W we see a divergence from the fit between the data and transform trends predicted by the new best fit pole, suggesting that the modeled pole location is still in need of improvement.

DISCUSSION
The inversion of our new data set for the Nazca-Antarctic plate boundary has produced a new best fit pole for that plate pair. This new Pole predicts relative motions that fit the data better and agree more closely with motions predicted by the RM2 pole than a best fit pole produced by the previous data set used in the derivation of RM2. However, there is a misfit between the data and the relative motion predieted by the new best fit pole in the western portion of the transform (Fig. 10). Inspection of Figure 9 suggests that this results from a moderate amount of incompatibility between the bathymetric trends from 10 1° to l09°W and the slip vector azimuths from earthquakes between 95° and loo 0 w. one reason for this discrepancy may be that the topography is far more complex than it appears, with numerous en echelon offsets that trend more easterly (in agreement with the predicted trends) al-71 though the overall bathymetry and epicenters trend ESE. Another reason may be that the paucity of geophysical data in the western portion of the Nazca-Antarctic boundary limits the accurate determination of a best fit pole. 0 0 The general lack of focal mechanisms from 101 to 109 W and, 0 correspondingly, the lack of precise bathymetric surveys from 95 to 0 100 W are the main impediments resolving this question.
Our new best fit pole is located much closer to the 'Nazca-Antarctic boundary than the old pole, and also south of the RM2 pole (Fig. 8).
This is consistent with our data which indicates slower spreading rates and more curvature in the boundary than previously believed. It may be argued that finer bathymetric trends in the western portion of the  [13] 0 reported an event located in the transform at 105 W, which was recorded by the Global Digital Seismograph Network (GDSN), with a fault plane 0 strike of 100 , also nearly parallel to the transform azimuth. These earthquake data suggest that our reported transform azimuths are accurate and the Euler pole is indeed located further south than previous data suggested, or than we calculate with this present data set.
The fact that the curvature in the boundary appears greater than even the new best fit pole predicts implies that this pole is located even further south. While our data set has significantly improved the calculated location of the best fit pole, thus providing tighter constraints on Nazca-Antarctic relative motions, a weakness exists due to the fact that most of the data are from locations east of loo 0 w. 7 2 Additional geophysical data from the western portion of the boundary are still needed to further refine the pole location and improve the agreement between measured and predicted relative motions along the entire Nazca-Antarctic boundary.     V.E. = 13.3 Fig. 4. Bathymetry of the Chile transform system as in Figure 2 with the locations of earthquake epicenters indicated by the solid circles.
Note the increased scatter in the epicenter locations east of l04°3o'W, 0 0 particularly between 104 and 102 W, and the gaps in seismicity also at 0 0 104 and 102 w.   The magnetic and bathymetric data were collected on R/V Endeavor cruise 112 using standard geophysical instrumentation. Endeavor's echo sounding system was used to collect the bathymetric data at frequencies of 3.5 kHz and 12 kHz. The 3.5 kHz energy will penetrate through thin sediments to bedrock. Since the area surveyed is relatively young with low biological productivity, the sediment cover is thin enough to be penetrated completely using the 3.5 kHz system. The 12 kHz frequency provides less penetration, but improved resolution, facilitating onboard analyses and planning. All the bathymetric data were recorded on analog tape and real-time charts. The proton precision magnetometer was used to collect total intensity magnetic field data. These data were also recorded on analog tape and a real-time chart.
All the software used in the data reduction are available on the Graduate School of Oceanography's PRIME computer under BOBD>GEOPHYSLIB.
The first step in the post-cruise data analysis was to combine the satellite position data with dead reckoning data to produce a best fit navigation data set, using a program called "CNAV". The remoteness of the region prevented the use of Loran c navigation. The program "CRUISETRACK" was used to plot the track lines on a Mercator projection.
Obvious bad navigation points could then be removed. The next step was to merge the magnetic and bathymetric data with the navigational data using the program called "MERGE80". This included the reduction of the total intensity magnetic data to anomaly form by subtracting the International Geomagnetic Reference Field at each point [Peddie, 1982]. The magnetic anomaly and bathymetric data were then plotted along track on a Mercator projection using the program "MERCPROF".
The next step was to select magnetic and bathymetric profiles for projection using the program "BTMAGPRJ". This program allows for projection of profiles along track, to a specified azimuth, or to a small circle through a specified point based on a given pole of rotation. In the projection process, magnetic anomaly profiles are zero-meaned, and both magnetic and bathymetric profiles are smoothed by a cubic spline fitting technique after interpolation to a 0.5 km spacing. Initial identifications of magnetic anomalies can then be made, along with calculations of spreading rates. Comparisons with the corresponding bathymetric profiles can also be made. This is followed by magnetic block modeling ("MDLMAG") to verify magnetic anomaly identifications, to further quantify spreading rates, and to summarize the recent spreading history. The completed two-dimensional block model consists of a continuous string of rectangles with a constant specified magnetization and layer thickness that are magnetized parallel or antiparallel to the earth's present-day magnetic field at that latitude. Paleofield directions can be used if the age of the model warrants it. The vertical boundaries of these rectangles are calculated from the chronology of geomagnetic field reversals at specific times (e.g., LaBrecque et al., 1977)  The plate motion inversion program used in this study to obtain the Euler poles was written by Richard Gordon and Seth Stein (Northwestern university) using the Minster et al. [1974] relative plate motion inversion algorithm. The basic assumption made is that the specified plates are rigid. The fact that Minster et al. [1974] and Minster and Jordan [1978] produced rigid plate models that satisfactorily explained all the data (RMl and RM2, respectively) argues in favor of this basic assumption. Symmetric and orthogonal spreading is also assumed. The data used were only from well-defined plate boundaries and included relative spreading rates calculated from magnetic anomaly profiles and directions of relative motions derived from transform fault trends and earthquake slip vectors.
According to the Euler theorem, the instantaneous motion of a rigid plate on the surface of a sphere can be completely and uniquely described by an axial rotation, or angular velocity vector . In the forward problem, the components of relative velocity between two plates at any point on their common boundary can be computed if the paramenters of the angular velocity vector are known. These parameters are rate of rotation and latitude and longitude of the rotation pole. In the inversion technique (or inverse problem) of Minster et al. [1974] and Minster and Jordan [1978] that was used in this study, we use observations of the components of relative velocities at plate boundaries to obtain the best representation of instantaneous motions in the form of angular velocity vectors. The procedure involves iterative perturbation of a chosen starting model until convergence is attained.
It is actually a least squares fitting technique in which the solution converges until a change in the model no longer makes the error function smaller. Minster et al. [1974] adapted the linear theory of maximum likelihood for use in their inversions which allowed them to statistically weight the data based on the level of uncertainty (a subjective evaluation of data quality), and to estimate the uncertainty attached to the model induced by errors in the data. They also developed the concept of data importances which depends on the nature and distribution of the data, and on the data uncertainties.
Since a linear technique is being applied to a nonlinear problem, it is important to construct a reasonably accurate starting model so the iterative procedure does not converge to a local minimum.  demonstrated that the proximity of the final model to the starting model justified the use of a linear theory. They did local studies before constructing their global model (RM!), starting with published angular velocity vectors, and found that poles for the global model did not differ much from those obtained from the local studies. Minster and Jordan [1978] constructed their relative plate motion model, RM2, using a revised data set and RM! as the starting model.
In the first part of this study (Manuscript I), we determined best fit and closure poles for the respective plate pairs in the Juan Fernandez four-plate system, and used these poles (angular velocity vectors) as the starting model in the four plate inversions. As described in Manuscript I, new data from this study and Engeln [1985] were used in addition to RM2 data (Table 3). A present day tectonic model could then be constructed for the Juan Fernandez microplate. In