Tectonics of the Papua‐Woodlark Region

The Papua‐Woodlark region exemplifies many plate tectonic processes, from active continental breakup to double plunging subduction zones. We present a synthesis of geological and geophysical data from the region and interpret these data to characterize the present plate boundaries. A subducted slab shown to underlie the Trobriand forearc is associated with the Papuan volcanic arc and the diapiric emplacement of felsic gneissic domes in the Woodlark rift from previously subducted continental materials in the mantle wedge. In contrast, the Suckling‐Dayman metamorphic core complex formed in the footwall of the Mai'iu extensional detachment, augmented by calc‐alkaline and high‐K magmatic intrusion and uplift. Utilizing Global Positioning System relative velocities, transform fault azimuths, ridge axis azimuths, and two fault slip directions, we derive a plate kinematic model for the Australia, Woodlark, Trobriand, and Solomon Sea plates to describe the neotectonics and provide insight to the past tectonics. We present two cases—with and without subduction at the Trobriand Trough—and favor the former based on our characterization of the plate boundaries. Our model estimates current Trobriand subduction rates of 3–5 mm/yr with obliquity increasing westwards, and full spreading rates, increasing eastwards along the Woodlark Basin spreading center, of 19–35 mm/yr. Based on seafloor morphology, we further estimate the time of the change in the pole of the Woodlark Basin opening at 450 Ka followed by a decrease in spreading rate and complete reorientation of the spreading axis fabric at 225 Ka.

Toward answering these and related questions, in this study, we revisit the geological, seismological, marine geophysical, and geodetic information to derive an improved plate kinematic description of the tectonics of the Papua-Woodlark region. We find that many of the disparate views can be reconciled by a better understanding of both the tectonic inheritance and recent changes in the configurations/velocities of plate boundary processes.

Gravity
The Papua New Guinea-Solomon Islands region of the western Pacific displays a rapidly evolving mosaic of microplates between the major Pacific and Australia plates that are obliquely converging at ∼110 mm/ yr (Tregoning et al., 1998;Wallace et al., 2004). The present and former subduction trenches in the region are marked by large negative free air gravity anomalies (FAAs, Figure 2). These include the New Britain Trench, accommodating subduction of the Solomon Sea lithosphere to the north beneath New Britain and Bougainville, and the San Cristobal Trench, accommodating subduction of the Australia Plate beneath Guadalcanal and San Cristobal. In between, where the young Woodlark Basin is being subducted northeast beneath the New Georgia Group of the Solomon Islands, there is a deformation front but it lacks a flexed outer rise, bathymetric trench, and associated negative FAA . The Trobriand Trough also has an associated negative FAA as well as a large positive (>200 mGal) FAA along the Trobriand outer forearc high, paralleled by a small negative FAA associated with the Trobriand forearc sedimentary basin ( Figure 2; Fitz & Mann, 2013a, 2013bFrancis et al., 1987). Notably, there is no negative FAA associated with the northern Woodlark Rise. The Trobriand Trough, outer high, and forearc basin all terminate eastwards at the Nubara transform fault. The implication is that the Solomon Sea Plate is or has been subducted on three sides, but not on the fourth along the Nubara Fault, and hence that the initial opening of the eastern Woodlark Basin was not "back-arc" Weissel et al., 1982). The pronounced Trobriand outer forearc gravity high can be explained by this part of the forearc being out of isostatic equilibrium, resting directly on the Solomon Sea slab. Gravity anomalies that trend just south of east in the Solomon Sea ( Figure 2) likely reflect its relict spreading center. The NW-trending negative FAA near 147°E, 13°S (Figure 2) marks the buried rift valley of the former spreading axis of the Coral Sea Basin (Weissel & Watts, 1979).

10.1029/2020GC009209
The thickly sedimented Aure Trough, Pocklington Trough and Rennell Trough, remnants of Paleogene northward subduction, are marked by negative FAA; likewise the North Solomon Trough, the site of southward subduction of the Pacific Plate prior to ca. 10 Ma (e.g., Cooper & Taylor, 1985, 1987bHamilton, 1979;van Ufford & Cloos, 2005). East-trending rift basins in the Woodlark and Pocklington rises are marked by less-positive FAA, as is the stair-step axis of the spreading center in the eastern Woodlark Basin (Figure 2; Martinez et al., 1999). Other negative FAAs are associated with rift basins, such as in Goodenough Bay, as well as with sedimentary basins among the Solomon Islands ( Figure 2). The northern Australia plate boundary is traced by river valleys and a relative gravity low down the axis of the Papuan Peninsula.
Subduction at the New Britain and San Cristobal trenches is characterized by normal faulting on the incoming plate, shallow thrust faulting beneath the forearc, and steeply dipping Wadati-Benioff zones extending to depths exceeding 200 km, overlain by active arc volcanoes (Figures 3 and 4).  Depth-coded colored circles show teleseismic earthquakes of at least 40 km depth from the International Seismological Center (ISC)-EHB catalog (Engdahl et al., 1998(Engdahl et al., , 2020Weston et al., 2018). Focal mechanisms of shallow (<40 km depth) earthquakes from the Centroid Moment Tensor catalog (Dziewonski et al., 1981;Ekström et al., 2012) are plotted at their ISC-EHB locations. Purple dots show microseismicity recorded during two PASSCAL experiments in SE Papua Eilon et al., 2015;Ferris et al., 2006). Active volcanoes from the Smithsonian catalog (Global Volcanism Program & Venzke, 2013) are located with yellow triangles. Cross-sections of seismicity and topography along profiles A, B, C, and N are shown in   Figure 3 and inset map. Seismicity is projected from within a zone 50 km wide on either side of the profiles, except for profile A which is from within a zone 100 km wide on the NW side only and for profile D which is from within a zone 80 km wide on either side. Bottom two panels (e and f): cross-sections of seismicity and topography along profiles C and N, such that all the seismicity within the rectangular box delimited by C-C' and N-N' is projected onto the two orthogonal profiles. Inset map (g) locates the seismicity cross-sectional profiles and the Wadati-Benioff zone contours at 50 km depth intervals on the top of the subducted lithospheric slab. NBT, New Britain Trench; RMF, Ramu Markham Fault; TT, Trobriand Trough. such as the M w 8.1 earthquake of April 1, 2007 (Furlong et al., 2009;F. Taylor et al., 2008) and the M w 7.1 earthquake of January 3, 2010 (Newman et al., 2011), both of which generated tsunamis.
There are also two clusters of east-trending strike-slip earthquakes east of Simbo Transform and south of the San Cristobal "Trench" that suggest active deformation, possibly by left-lateral bookshelf faulting reactivating ∼E-W abyssal hill faults. It is likely that this easternmost Woodlark Basin is a picoplate or deformation zone such that the north-trending Simbo Transform reflects motion between the Woodlark Plate and this picoplate rather than the current NNW motion relative to the Australia Plate. This area has been rotated counterclockwise, not unlike the rotation of the Magdalena microplates associated with attempted ridge subduction off Baja California (Michaud et al., 2006). Adjacent to the ridge subduction slab window a south-dipping remnant Pacific slab is seismically active to depths of 200 km and may source Savo arc volcano (Cooper & Taylor, 1985, 1987b. But the dominant arc volcanism is in the New Georgia Group of the Solomon Islands, including in the forearc (e.g., on Rendova and Kavachi) and even south of the trench on the incoming plate (on Simbo Island and Coleman and Kana Keoki seamounts) (Chadwick et al., 2009;Crook & Taylor, 1994;R. W. Johnson et al., 1987;Konig et al., 2007;Taylor & Exon, 1987). Seismicity beneath these volcanoes is shallow and diffuse-not unexpected given how readily thin young oceanic lithosphere may be resorbed into the mantle. It is likely that this arc volcanism owes its geochemistry and location to both ridge subduction and arc reversal.

Woodlark Basin Spreading Center
Marine geophysical studies over the past 4 decades have characterized seafloor spreading in the Woodlark Basin with increasing resolution-both on and off axis-such that it is one of the few ocean basins that is completely and systematically mapped with swath bathymetry, acoustic imagery, magnetic, and gravity data (Goodliffe & Taylor, 2007;Goodliffe et al., 1997Goodliffe et al., , 1999Martinez et al., 1999;Taylor, 1987;Taylor et al., 1995Taylor et al., , 1999Taylor et al., , 2009Taylor & Huchon, 2002;Weissel et al., 1982). As the oceanic crust is also thinly sedimented, the seafloor fabric and magnetic anomaly age distribution are very well known. The basin opening from ∼0.5 to 3.6 Ma west of Simbo transform, where both conjugates of the spreading are preserved and have not been rotated approaching the Solomon subduction zone, is well described by a single Euler pole at 147°E, 9.3°S, near Port Moresby . A complexity in the basin evolution, however, is a recent (within the Brunhes Chron) change in spreading direction and rate, manifest in reoriented spreading segments and transforms as well as nontransform offsets ( Figure 5, Goodliffe et al., 1997;Martinez et al., 1999;Taylor et al., 1995Taylor et al., , 1999. Land GPS studies suggest that the present Australia-Woodlark (A-W) opening pole is somewhat further south and that the spreading rate may have slowed substantially (Tregoning et al., 1998;Wallace et al., 2004Wallace et al., , 2014. The recognition that the Nubara Fault continues SW of Woodlark Island to a triple junction with the western spreading segments (Figures 1, 2, and 5) also means that the Trobriand Plate is separate from the Woodlark Plate Taylor et al., 2009;Wallace et al., 2014). Furthermore, if the Trobriand Trough remains an active, albeit slow, site of subduction (see below), then the Trobriand Plate is also separate from the Solomon Sea Plate.
In this and the next three subsections, therefore, we describe the current plate boundaries in the region, together with their fault and fabric parameters that we have measured, in order to solve, with the GPS data, the best fitting multiplate kinematic description of the present tectonics.
The opening of the Woodlark Basin west of Simbo Transform has been organized in five spreading segments, with overlapping nontransform boundaries between segments 1 and 2, transform faults between segments 2-3 (Moresby) and 3-4 (Davies), and an eastward propagating spreading center that lengthened segment 4 at the expense of segment 5 ( Figure 5; Goodliffe et al., 1997;Taylor et al., 1999). Following the reorientation of spreading direction during the Brunhes Chron, the Moresby and Davies transforms as well as the propagator developed short intratransform spreading segments ( Figure 5). The axial rift valley in the eastern Woodlark Basin subdivided into several reoriented rift segments which are the locus of many normal and strike-slip earthquakes. In contrast, there is almost no teleseismicity nor focal mechanisms on the relatively inflated but slower spreading segments 1 and 2 in the western Woodlark Basin, which have overlapping spreading centers but lack transform faults (Figures 3 and 5). This difference in spreading morphology is accompanied by the eastern basin having lower magnetization, thinner crust, and being regionally ∼500 m deeper than the western basin. Martinez et al. (1999) examined and modeled these characteristics as being the result of rift-induced secondary mantle convection in the western basin, with its thicker continental margins, but not in the east, giving the western basin characteristics elsewhere associated with faster (rather than slower, in this case) spreading.
Noting that the spreading fabric of segments 1 and 2 cannot be copolar with those of segments 3, 4, and 5, and given the observations above, we infer that spreading on segments 1 and 2 is somewhat oblique. We therefore use the abyssal hill and transform fault azimuths associated with segments 3, 4, and 5 as input to our inversion for the A-W opening pole ( Figure 5, Table 1).

Nubara Transform Fault
A cluster of strike-slip earthquakes with aligned NE-trending nodal planes marks the Nubara Fault on the Woodlark Rise ( Figure 6). The right-lateral focal mechanism for the largest of these, a magnitude 7 event in 1974, was first published in Weissel et al. (1982). Since that time, the seafloor bathymetry has been fully swath mapped, revealing three subparallel strike-slip fault strands (with small right-stepping releasing bends) on the central of which the earthquakes cluster ( Figure 6, Taylor et al., 2009). The Nubara Fault continues to the SW past the southern end of the Trobriand Trough, across the Woodlark Rise to the southern side of the Egum Graben (Taylor & Huchon, 2002;Taylor et al., 2009) where there is a cluster of right-lateral strike slip and normal fault earthquakes (Figures 3, 5, and 6). We have measured the azimuths of the trace of the Nubara Fault at points along its length that are not releasing bends ( Figure 6, Table 2). We fit small circles to these data to determine the location of the Euler pole between the Woodlark and Trobriand/Solomon Sea plates.

Trobriand Trough and Papuan Arc
Trench flexure accompanied by negative FAA gravity and normal faulting occurs on three sides of the Solomon Sea, at the Trobriand Trough as well as the New Britain Trench (Figures 2 and 3 (Table 1). White lines bound the Brunhes magnetization crust. Transform faults are labeled M (Moresby), D (Davies), S (Simbo), and N (Nubara). Focal mechanisms of earthquakes from the Centroid Moment Tensor catalog (Dziewonski et al., 1981;Ekström et al., 2012) are plotted at their International Seismological Center-EHB locations. Plate triple junctions are circled in aqua. earthquake may not be as well located as Letz and coworkers surmise, there is no question that it occurred beneath the Trobriand forearc, given where the tsunami was and was not reported. The Trobriand Trough is thickly sedimented and has recently undergone convergence, as evidenced by large thrust sheets, spaced 5-7 km apart, forming the lower landward slope of the trough, ponding sediments behind them (Davies et al., 1987;Silver et al., 1991; Figure 7). Our bathymetry swath mapping on R/V Kilo Moana cruise 0418 traced the active deformation front of the Trobriand accretionary prism along its whole length west of 153°E ( Figure 7). In addition to these features, subduction at the Trobriand Trough has produced the characteristic architecture of an arc-trench system, including an outer-arc structural and FAA gravity high, up to 5-7-km thick forearc sedimentary basin, volcanic front, and behind-the-front volcanoes, with arc volcanics dated from 15 Ma to present (e.g., Fitz & Mann, 2013a, 2013bFrancis et al., 1987;Pinchin & Bembrick, 1985;Smith & Milsom, 1984;Taylor, 1999).
The seismogenic zone under the Trobriand forearc is contiguous westwards with the southern limb of the doubly plunging seismogenic slab under the collision of the Huon Peninsula and Adelbert-Finisterre range with the New Guinea mobile belt of the north Australian margin (Figure 4; Hayes et al., 2018, their Figure  3; Davies, 2012). As Abers and Roecker (1991) pointed out, however, the seismicity deeper than 40 km and possibly illuminating a slab east of 148°E is quite patchy. We reproduce their stacked seismicity plots across and along strike (Figures 4d and 4e) with the benefit of 3 decades more teleseismic data, plus the microseismicity recorded during two PASSCAL experiments in SE Papua Eilon et al., 2015;Ferris et al., 2006). Critically, for over ∼500 km, the various tele-and microseismicity patches south of the Trobriand Trough (Figure 4e) project in cross section onto one dipping seismogenic zone from crustal depths to ∼125 km (Figures 4d and 8). Furthermore, the depths to Moho under the Trobriand outer forearc, derived from the receiver functions of seismic stations on the low-relief islands, are so deep (42-49 km, Figure 8) that the deepest are most likely to come from the Moho of the subducted plate (Abers et al., 2002). The mantle P-wave velocity variations of Abers et al. (2002) are compatible with the slab evidenced by the seismicity, whereas the mantle S-wave velocity variations of Eilon et al. (2016), that incorporated 3-D seismic anisotropy, inexplicably do not image the subducted slab evidenced by the microseismicity that was located by the same experiments ( Figure 8). The teleseismic and microseismic data together define a Wadati-Benioff zone associated with a south-dipping subducted slab, the top of which we have contoured at 50-km depth intervals, which is seismogenic to successively greater depths westwards (Figures 3 and 4) Note. The last three columns present Case 2 azimuthal predictions, the difference from observations, and full spreading rate predictions at each point.  Taylor et al., 1999)

and the Estimated Standard Error
Although we conclude that the Trobriand Trough and Papuan arc represents a Neogene subduction zone that remains active, the current subduction rates may be no more than a few millimeters per year (Kirchoff-Stein, 1992;Reed et al., 1988), which also may explain the lack of a cosmogenic Be 10 signature in the 1951 Mount Lamington andesite (Gill et al., 1993). As a means to bracket the possible rates, we develop plate kinematic solutions for two cases: with and without current subduction at the Trobriand Trough.

Papuan Peninsula and Woodlark Rift
For much of its length the Australia-Trobriand (A-T) plate boundary trends SE along the axis of the Papuan Peninsula as the double-stranded Owen Stanley Fault, separating the Owen Stanley Metamorphics (Kagi Metamorphics, Emo Metamorphics) from the Papuan Ultramafic Belt (Figures 1 and 2; Davies & Jaques, 1984;Davies & Smith, 1971;Pieters, 1978;Smith & Davies, 1976 (Dziewonski et al., 1981;Ekström et al., 2012) plotted at their International Seismological Center-EHB locations-except for the magnitude 7 earthquake near 154°E, whose focal mechanism and location is from Weissel et al. (1982). Azimuths of Nubara Transform Fault segments that are not releasing bends bounding the Woodlark Plate are shown in dark blue, with dark blue tadpole fill for the Trobriand segment and white tadpole fill for the Solomon Sea segment (Table 2). Plate triple junctions are circled in aqua. EG, Egum Graben; MS, Moresby Seamount; TT, Trobriand Trough. Enlarged images of the bathymetry in the two boxed areas are shown in the middle and bottom panels, revealing the detailed trace of the right-lateral Nubara Transform Fault and its releasing bends.
Although geomorphology and other data indicate that these faults localize the principal plate boundary motion along the Peninsula (e.g., Webber et al., 2018), there are subparallel normal faults and microseismicity further south, including those that border the Milne Bay graben (Figures 7-9; Jongsma, 1972).
Marine geophysical and earthquake data indicate that there is another set of normal faults overlapping to the north, that extend west from the Australia-Trobriand-Woodlark (A-T-W) triple junction to the D'Entrecasteaux Islands, including from east to west, Normanby, Fergusson and Goodenough islands, beneath which the crust is regionally thin (Figure 9). These normal faults include grabens to the north and south of Moresby Seamount, and three subparallel asymmetric graben north of Normanby Island that extend west to Fergusson Island (Figures 1, 7, and 9; Goodliffe & Taylor, 2007;Taylor & Huchon, 2002), in what Abers (2001) and others subsequently have collectively termed the Woodlark Rift. A cluster of microseismicity continues WNW along the north coast of Fergusson Island (Figures 3 and 9; Abers et al., 2016;Eilon et al., 2015). The three gneissic domes of the D'Entrecasteaux Islands are sinistrally offset along trend by NE-striking transfer faults (e.g., Davies & Jaques, 1984;Figures 1, 7, and 9). Another microseismicity cluster marks a NNE-striking transfer fault off the east coast of Normanby Island, and yet another trends WSW from Goodenough Island to the Cape Vogel Peninsula (Figures 3 and 9; Abers et al., 2016;Eilon et al., 2015). Together, these observations indicate that there are overlapping zones of extension (north and south) and transfer faults (east and west) that bound tectonic blocks that comprise the D'Entrecasteaux Islands and Goodenough Basin (Wallace et al., 2014). There is a tongue of thicker crust that trends from beneath the Cape Vogel Peninsula ESE under the Goodenough Basin (Figure 9), which is itself being internally deformed, as evidenced by active normal faults imaged on seismic reflection data there (Fitz & Mann, 2013a, 2013bMutter et al., 1996). The thicker crust is evidenced by a greater Moho depth (measured by receiver functions, Abers et al., 2016) that is structurally below the hanging wall rollover of the listric Goodenough Basin Fault and thus not caused by it. Although it has been associated (Fitz & Mann, 2013a, 2013b with lower crustal flow accompanying roll-back of the Trobriand slab, as proposed by Kington and Goodliffe (2008), that is an ad hoc rather than emergent characteristic of their model.
The easternmost A-T plate boundary segment is also the westernmost Woodlark Basin spreading segment 1a (Goodliffe et al., 1997;Taylor et al., 1999). The A-T-W triple junction, which we locate at 151.83°E, 9.78°S, is formed between spreading segment 1a, the Nubara Fault (along the southern edge of the Egum Graben), with the third arm being a nontransform offset between overlapping Woodlark spreading segments 1a and 1b (see Figure 2 of Goodliffe & Taylor, 2007, for details). Note that there is also a cluster of mostly strike-slip but also normal fault earthquakes on the southern margin, ahead of the dueling/propagating spreading BENYSHEK AND TAYLOR Note. The last three columns present Case 1 and Case 2 azimuthal predictions, the difference from observations, and rate predictions at each point.  (Figures 3 and 7). A campaign GPS station on Strathord Island in that region is moving ∼5 mm/ yr to the NNE relative to Australia (STRA, Figure 2 of Wallace et al., 2014) also suggesting that not all extension is focused on the spreading segments but is partly distributed in the margins.

Plate Kinematic Models
Given the Australia, Woodlark and Trobriand plate boundaries described above, we seek the best fitting multiplate kinematic description of the present tectonics. Among other things, our goal is to determine whether the Papua-Trobriand GPS data and Woodlark Basin opening history can be better reconciled and to analyze the kinematics to improve our understanding of the tectonic processes. Given that the current subduction rates at the Trobriand Trough may be no more than a few millimeters per year (Kirchoff-Stein, 1992; BENYSHEK AND TAYLOR 10.1029/2020GC009209 12 of 26  (Dziewonski et al., 1981;Ekström et al., 2012) plotted at their International Seismological Center-EHB locations. H = location of 1895 magnitude 7.3 tsunamigenic earthquake (Letz et al., 2016). Active volcanoes from the Smithsonian catalog (Global Volcanism Program & Venzke, 2013) are located with yellow triangles. GPS vectors relative to Australia in the ITRF2008 reference frame (Altamimi et al., 2011) are shown with arrows and error ellipses at the labeled campaign sites of Wallace et al. (2014) and also at the continuous station at Port Moresby (Willis et al., 2016). Stars locate the current A-T pole (this study) and the former Australia-Woodlark pole .  Reed et al., 1988), as a means to bracket the possible rates we develop plate kinematic solutions for two cases: with the Trobriand Trough being (1) inactive and (2) active-in which latter case, the Trobriand Plate is separate from the Solomon Sea Plate.
Our kinematic models are constrained by the following data: GPS relative velocities, transform fault azimuths, ridge axis azimuths, and two fault slip directions. Data utilized in our inversion are listed in Tables 1-3.  Table 1 presents data utilized for the A-W plate boundary. These data include both transform fault and ridge axis azimuths from spreading segments 3, 4, and 5. We exclude azimuths from segments 1 and 2 due to their inferred spreading obliquity and overlap with possible continued extension of the rifted continental margins. Data utilized for the Woodlark-Trobriand (W-T) plate boundary are listed in Table 2 and consist only of fault azimuths along the Nubara Fault, exempting the releasing bends.
The A-T plate pair has the most extensive data set (  (Figure 7, Table 3). One slip direction we use is that of the mylonitic lineations on the north flank of the Dayman dome (Daczkco et al., 2011), the other is of the megascopic grooves on the lower north flank of the Moresby Seamount shallow-angle normal fault (Speckbacher et al., 2011 Figure 7). Hence, the westwards transfer of extension from the north side of the D'Entrecasteaux Islands to the south side of Goodenough Bay appears complete by the Barrier Islands transfer fault (Davies & Jaques, 1984) that trends SW between Fergusson and Goodenough islands (Figure 7). On the other hand, the three GPS sites closest to the Trobriand Trough on the outer forearc high (KAWA, LOS2, and GUA1) display components of westward motion that differ from predictions of the best fitting Trobriand Plate motion (by 2.4, 3.6, and 7.9 mm/yr, respectively, Table 3, Figure 7). They may result from greater coupling to the subducted Solomon Sea plate, and/or differential motion within the Trobriand subduction earthquake cycle, and/or for LOS2 and GUA1 incomplete removal of coseismic displacements from BENYSHEK AND TAYLOR 10.1029/2020GC009209 13 of 26 In the middle panel, the seismicity is overlain on a projected cross-section from 150.8°E of mantle P-wave velocity variations (Abers et al., 2002). Also shown on that section are Moho depths (in red) and depth ranges (in green) determined from receiver functions, as well as the forearc basin sediments (in yellow) (ibid) and the inferred top (dark blue) of the subducted slab (with positive mantle velocity anomalies) descending from the Trobriand Tough (TT). In the bottom panel, the same seismicity is overlain on a projected cross-section from 150.4°E of mantle S-wave velocity variations (Eilon et al., 2015) but that do not image the subducted slab evidenced by the microseismicity located by the same experiments.

Methodology
A grid search of the region (longitude-latitude) was performed to determine the statistically best fit pole location for each plate pair given the input data (Tables 1-3). We utilized a bootstrap method (Efron & Tibshirani, 1986) involving random resampling with replacement of data inputs and, for each of 5,000 iterations, calculated chi-square statistics (sum of squares of the difference between observation and prediction divided by the square of the standard deviation) for each grid point for each of three datasets separately. For the A-T GPS data, an additional grid search in omega (angular rotation) space was performed at 0.1°/Myr and 0.01°/Myr then refined to 0.001°/Myr to calculate predictions of the expected linear velocity (in eastings and northings) at each GPS station to locate the statistically best fit rotation rate for any longitude-latitude pair result. We then took each plate pair's 5,000 results as a reasonable representation of possible pole error and calculated (95 percentile) contours of confidence using a density function and methodology described by Wilson (1993). Next, chi-square statistics for each grid point within the 95 percentile confidence regions were recalculated without resampling and summed across each possible triplet solution.
BENYSHEK AND TAYLOR 10.1029/2020GC009209 14 of 26 Figure 9. Microseismicity (black dots), crustal thickness, and mantle S-velocity anomaly at 60 km depth (between the orange dotted lines), determined from regional seismic experiments WOODSEIS and CDPAPUA (modified after Abers et al., 2002Abers et al., , 2016Eilon et al., 2014Eilon et al., , 2015Eilon et al., , 2016Ferris et al., 2006). Circles, colored by the best fitting Moho depth, are scaled to P(dVm>0), the fraction of successful velocity inversion models for which the Moho velocity step is positive; triangles show stations for which P(dVm) is less than 0.5 (yellow) or no receiver functions could be determined (open) . Yellow and green lines approximate 30 and 40 km Moho depth contours. The thinnest crust is beneath the DʻEntrecasteaux Islands. Note the tongue of thicker crust extending from the Cape Vogel Peninsula ESE under the Goodenough Basin. The thicker crust in the north is likely associated with the subducted Trobriand slab (Figure 8; Abers et al., 2002 Notes. Velocity columns ending with "A08" are with respect to the Australia Plate in the ITRF08 reference frame (Altamimi et al., 2011). At the bottom, two input azimuthal data points from Dayman Dome (Daczkco et al., 2011) and Moresby Seamount (Speckbacher et al., 2011) and their associated errors. The last three columns present Case 2 predictions, the difference from observations, and rate predictions at the two azimuthal points.

Table 3
Global Positioning System (GPS) Site Locations, Velocity Eastings and Northings, and Estimated Error From Wallace et al. (2014 ) where AωT represents the vector describing motion between the Australia and Trobriand plates, AωW the Australia and Woodlark plates, and WωT the Woodlark and Trobriand plates, each represented by a rotation pole (longitude, latitude, and angular rate).

Results
Euler pole locations are plotted on Figure 10 in longitude-latitude space and angular rotation rate-latitude space with their associated 95 percentile confidence regions (Table 4). Predictions, differences from observations and estimated rates at input data locations for the three plate pairs are listed in Tables 1-3. Rates of motion along the plate boundaries are presented in Figure 11. The A-T plate pair resulted in the same Euler pole for both Case 1 and Case 2, likewise for the A-W plate pair at least up to the 100th place in angular rate. Therefore, whether or not the Trobriand Trough is active does not change the relative plate motion vectors at the A-T-W triple junction (Figure 11 insert). However, the difference between 1.988°/Myr and 1.983°/Myr in the angular rate for the A-W Euler pole is associated with a shift of the W-T Euler pole by half a degree in longitude and a 10th in latitude between Case 1 and Case 2.

A-T Plate Boundary
The most robust kinematic data set and cornerstone to our model comes from the A-T plate pair, whose GPS data provide the only rate constraint in our inversions. Using GPS rates and two lineations, we calculate the best fit current A-T Euler pole and angular rotation rate (147.6°E, 9.7°S, 2.560°/Myr) and confidence regions ( Figure 10). The error of the A-T Euler pole is tightly constrained and the results of Wallace et al. (2014) are within it (Figure 10). Model predictions and differences relative to the A-T GPS data are presented in the same format as the campaign measurements (eastings and northings, Table 3). The average differences are −0.27 and 0.60 mm/yr in easting and northing, respectively. Vectors along the plate boundary change from compressional to extensional from west to east given the location of the pole ( Figure 11) and range from −10 to 20 mm/yr. The resultant A-T Euler pole is also close to the Taylor et al. (1999) A-W Euler pole for 0.52-3.6 Ma. Further reflection on this is presented in the Discussion.

A-W Plate Boundary
Bathymetry swath mapping data of the Nubara Fault crossing the Woodlark Rise (Figure 6, Taylor et al., 2009) and of the eastern Woodlark Basin volcanic axis ( Figure 5) collected since Taylor et al. (1999) allowed us to slightly refine picks of the current axial seafloor fabric and transform fault azimuths for spreading segments 3, 4, and 5 as compared to previous studies. Our Case 1 (147.4°E, 12°S, 1.988°/Myr) and Case 2 (147.4°E, 12°S, 1.983°/Myr) results differ only slightly in angular rotation rate but the difference allows for greater change between cases for the W-T plate boundary ( Figure 10, Table 4). With the short spreading history of the current Euler pole, there is limited distance between azimuthal observations resulting in a stretched confidence region that extends along the trend of the spreading center. Note that the Taylor et al. (1999) and Wallace et al. (2014) A-W Euler poles lie along the same trend since they utilized the same azimuth data originally published by Taylor et al. (1999), which also included the maximum horizontal compressive stress on Mount Victory as a constraint ( Figure 10, Table 4). The modifications to the spreading center azimuths combined with the GPS constraints on the A-T Euler pole resulted in a slower angular rotation rate for the A-W pole than Taylor et al. (1999), similar to the rotation rate of Wallace et al. (2014) (Figure 10). Our favored result (Case 2) produces spreading rates along the Woodlark Basin spreading center that systematically increase from 19 mm/yr in the west to 35 mm/yr in the east (similar to the results of Wallace et al., 2014; Figure 11).

Case 1: Three-Plate Solution
The recognition of a separate Trobriand Plate from the Woodlark Plate (F. Taylor et al., 2008;Wallace et al., 2014) accompanied the recognition of the Nubara Fault as a transform fault with short right-stepping releasing bends. This allowed us to fit a small circle to the fault to identify a possible Euler pole. In the case where there is no active subduction at the Trobriand Trough, the Nubara transform fault would be the W-T plate boundary along its whole length. Observations of the Nubara fault azimuth over that ∼450 km length reasonably well constrain the azimuth to the Case 1 W-T Euler pole (148.3°E, 1.7°S, 0.573°/Myr) and confidence regions (Figure 10). For comparison, the gold star in Figure 10 is the W-T pole calculated from the A-T and A-W poles of Wallace et al. (2014; Wallace et al. (2014). Our resulting W-T Euler pole for Case 1 produces rates along the Nubara Fault that range from 9.8 to 10 mm/yr ( Figure 11, in which the depicted Case 2 W-SS rates northeast of the intersection with the Trobriand Trough would double, from 5 to 10 mm/yr for Case 1).

Case 2: Four-Plate Solution
Given the evidence, summarized above, that the Trobriand Trough is accommodating active subduction, albeit slowly, the Nubara Fault is a divided plate boundary with the Trobriand Trough separating the Trobriand and the Solomon Sea plates. This is our preferred Case 2 solution, for which the sum of squares of W-T residual azimuths (predicted vs. measured) is 101. The portion of the Nubara Fault to the east of the Trobriand-Woodlark-Solomon Sea triple junction (153.49°E, 9.04°S) represents the Woodlark-Solomon Sea (W-SS) plate boundary and the portion to the west the W-T plate boundary. This four-plate solution has the relative motion between the Solomon Sea, Trobriand, and Woodlark plates being all nearly in the same direction at the triple junction (Figure 11 insert). Although the W-T Euler pole locations are similarly or better constrained than the A-W pole (Figure 10), we lack independent rate measurements for the Solomon Sea Plate, so we arbitrarily partition the W-T rate of motion along the Nubara Fault equally between the other two vectors (W-SS and Solomon Sea-Trobriand [SS-T]). The result for the Case 2 W-T Euler pole (148.3°E, 1.8°S, 0.579°/Myr) is presented in Figure 10 by the northernmost blue stars and confidence regions. The light blue stars and confidence regions symbolize the W-SS Euler pole (148.8°E, 2.2°S, 0.311°/Myr). Rates along the Nubara Fault between the Trobriand and Woodlark plates are the same as Case 1 (9.8-10 mm/ yr) and are about half that (4.9-5 mm/yr) between the Woodlark and Solomon Sea plates. The SS-T Euler pole (147.8°E, 1.4°S, 0.271°/Myr) is calculated from these results and predicts rates of subduction along the Trobriand Trough that range from ∼3 to 5 mm/yr ( Figure 11). The SS-T vectors are increasingly oblique to the west along the Trough, such that their orthogonal convergence rates would be even less, consistent with the expectations of Reed et al. (1988) and Kirchoff-Stein (1992  to 6-10 mm/yr, inversely proportional to the W-SS rates varying from 9.8 to 10 mm/yr down to zero, if the W-SS and SS-T rates were partitioned differently.

Partitioning Spreading During the Brunhes Chron
Older seafloor fabric within the Brunhes Chron represents a primarily left-stepping ridge system in the eastern part of the basin until counterclockwise rotation of the spreading direction produced a transtensional system that resulted in more and shorter ridge segments and offsets ( Figure 5). West of Moresby Transform, the ridge segments are spreading obliquely and are right-stepping, so the transtensional change resulted in increased nontransform offsets. This obliquely spreading part of the basin (segments 1 and 2), without transform faults, is not amenable to the following calculations.
To quantitatively address the changes in spreading rate and direction within the Brunhes Chron, we first selected conjugate features on either side of the spreading ridges (segments 3, 4, and 5) that represent the youngest fabric formed about the older pole described by Taylor et al. (1999). These pseudo-isochrons are represented in green in Figure 12. We found that our calculated A-W pole location (147.4°E, 12°S) could be used to reconstruct the pseudo-isochrons back together; however, our calculated modern-day angular rotation rate (1.98°/Myr) was not sufficient within the time constraints. To determine the average age of our pseudo-isochrons, we calculated the total distances between conjugate points and the Brunhes/Matayama boundary ( Figure 12, green tadpoles) and, using the older pole, derived a best fit age of 0.45 Ma. We also measured the width of reoriented seafloor fabric associated with the axial valley and, using the current pole, derived the oldest age of the current spreading rate to be about 0.225 Ma (Figure 12, yellow bars). We then calculated the remaining distance between our two known Euler poles (distance from end of yellow lines to 450 Ka pseudo-isochrons). From this distance and the time constraints, we calculated the necessary spreading rates and derived the angular rate of 4.234°/Myr to explain earlier spreading rates. This angular rate is also the angular rate of the Taylor  Euler pole changed location at 450 Ka, while maintaining the faster angular rate, and then slowed to the current angular rate at 225 Ka. This is the simplest three-fold partitioning of spreading during the Brunhes Chron. It uses an instantaneous change in Euler pole locations and a subsequent instantaneous change in Euler pole angular rates. More gradual and/or complex changes between 450 and 225 Ka are possible, summing to the same results.

Discussion
While our model for the neotectonics of the region provides a good explanation of the current day plate motions, there are several questions still to be addressed about the past tectonics. For instance, what is the age of the A-T-W triple junction? Also, what caused the spreading changes in the Woodlark Basin? We calculate that the most recent change in spreading direction of the Woodlark Basin spreading center occurred at 0.45 Ma and we know that spreading was consistent about one pole prior to that back to at least 3.6 Ma . Therefore, we expect the boundary forces to also be consistent during that time. The question to be answered is: What initiated the late reorientation and slowing of the seafloor spreading? We consider four main driving forces that could result in the reorientation of the Woodlark Basin spreading center, some of which have previously been considered by other authors (i.e., Ott & Mann, 2015;Wallace et al., 2014;Weissel et al., 1982): 1. Slab pull from the New Britain (western) section of the New Britain Trench 2. Increased resistance to convergence west of the Woodlark Basin pole of opening 3. Perpendicularity of ridge subduction beneath the Solomon Islands 4. Propagation of the Nubara Fault to form the boundary between the Trobriand and Woodlark plates All or any one of these possibilities are plausible given the known constraints. Pull from the subduction at the New Britain Trench has long been hypothesized to be the primary reason for the opening of the Woodlark Basin (Weissel et al., 1982) but the history of subduction there is not very well known other than that it has produced a seismogenic slab to depths exceeding 600 km and that there is an eastward propagating arc-continent collision (Cooper & Taylor, 1987a;T. Johnson & Molnar, 1972;Silver et al., 1991). A change in that system could contribute to the change in spreading direction of the Woodlark Basin spreading center at 0.45 Ma. This is similar to the system in the Red Sea-Gulf of Aden where the Tethys subduction under Zagros and Makran, coupled with the eastward propagating continental collision of Africa with Eurasia, has been modeled in 3-D lab experiments as the primary driving force for spreading, albeit that its localization was augmented by a zone of lithospheric weakness generated by the Afar plume (Bellahsen et al., 2003). In their model, it is the contrast between collisional locking of the western portion of the trench and the continued slab pull of the eastern active portion of the trench that causes the separation and rotation of the Arabian plate from Africa-which is quite analogous to the New Guinea collision-New Britain Trench situation opening the Woodlark Basin.
The opening of the Woodlark Basin and Rift has been matched across its pole of opening (near Port Moresby, Figure 1, TL99 Figure 10, Taylor et al., 1999), scissor-like, by compression of the NW Papuan Peninsula region (Figures 3 and 11). This has been expressed by the inversion of forearc normal faults there into thrusts (Figure 1, Pinchin & Bembrick, 1985) and by the formation of the rear-arc Aure-Moresby foreland fold and thrust belt (Bulois et al., 2017;Ott & Mann, 2015). Changes to the resistance to this convergence, due for example to finite shortening and crustal thickening, could be an additional cause for changing the kinematics on the opening side of the pole.
Alternatively, the reorientation of the Woodlark spreading axes may position them more perpendicular to subduction at the (eastern) New Britain-San Cristobal Trench, as occurred on the ridge system subducting beneath Baja California (Michaud et al., 2006). It is also consistent with the rotation of the seafloor fabric east of the Simbo Transform, and of the southern Simbo Transform itself, that dates to ∼2 Ma (i.e., prior to magnetic anomaly 2; Figures 5 and 12 W-T plate boundary and the A-T-W triple junction also allowed the Trobriand Plate to resume its prior to 0.45 Ma northward motion relative to Australia (Figure 11 inset), perhaps driven by trench suction. This is reflected in the similarity between the locations of the Taylor et al. (1999) 0.52-3.6 Ma A-W pole and the current A-T pole ( Figure 10). That the Trobriand Trough convergence and Nubara transform faulting are both directed orthogonal to the slab pull from beneath the western New Britain Trench may help explain how such slab pull is transmitted across both plate boundaries to cause the AW Euler pole and the NW-vectors of Woodlark-Australia opening ( Figure 11). Unquestionably there is at least one active plate boundary (the Nubara Transform Fault), and we suggest two (the Trobriand Trough), between the New Britain Trench and the Woodlark spreading center and rifts (Figure 11, also see Biemiller et al., 2019, their Figure 5 and/or Figure 1 of Wallace et al., 2014). Either slab pull can be transmitted across the near-orthogonal motion on this/these boundaries without producing extension there, or New Britain slab pull is not a primary driver of the Woodlark Basin opening, and one or more of the three other possible causes discussed above dominates.

Conclusions
1. Subduction at the Trobriand Trough is occurring, very slowly, based on the following evidence: trench flexure accompanied by negative FAA gravity anomaly and normal faulting, correlative Papuan arc volcanism, presence of a south-dipping patchily seismogenic slab to depths of 50-125 km beneath the Trobriand forearc contiguous westwards with the southern limb to 200 km depth of the doubly plunging seismogenic slab under the collision of the Huon Peninsula with the New Guinea mobile belt, a tsunami-generating historic earthquake in the outer forearc, large thrust sheets, spaced 5-7 km apart, forming the lower landward slope of the trough and ponding sediments behind them, and high-resolution bathymetric mapping of the active deformation front along the length of the trough (Figures 2-4, 7, and 8) 2. Our plate kinematic model estimates current spreading rates along the Woodlark Basin spreading center to range from 19 to 35 mm/yr, subduction at the Trobriand Trough to range from 3 to 5 mm/yr, right-lateral motion along the Nubara Fault at up to 10 mm/yr, and motion along the Owen-Stanley Fault and Woodlark Rift to range from −10 mm/yr (compressional) to 3-21 mm/yr (extensional), all reported from west to east (consistent with previous studies; Figure 11) 3. The Papua-Trobriand GPS data provide the recent temporal (decadal) resolution for this plate kinematic model. Given the substantially faster seafloor spreading rates in the Woodlark Basin, measured from conjugate magnetic anomalies back to 3.6 Ma, we estimate a change in pole location at 450 Ka (within the Brunhes Chron) followed by, based on abyssal hill and fracture zone trends, a decrease in spreading rate at 225 Ka (Figure 12) 4. The emplacement of gneissic domes in the continental rift is along the volcanic front and, like the proximal arc magmatism, is dominantly by diapiric processes originating in the mantle wedge above the south-dipping lithospheric slab subducted at the Trobriand Trough ( Figure 8) (that mantle having been preconditioned by continental materials introduced from the south during Paleogene subduction). In contrast, the Suckling-Dayman MCC formed in the footwall of the Mai'iu extensional detachment, aided, and abetted by rear-arc calc-alkaline and high-K magmatic intrusion and uplift

Data Availability Statement
The additional swath bathymetry data used in this study are available for download from the Rolling Deck to Repository (R2R; https://www.rvdata.us/search/cruise/KM0418) and Pangaea (https://doi.pangaea. de/10.1594/PANGAEA.900113?format=html#download). The microseismicity data are previously published in Ferris et al. (2006), and Eilon et al. (2015), and are available through the IRIS DMC.

Acknowledgments
Our thanks to Andrew Goodliffe and Fernando Martinez for helping to collect (on cruise KM0418 and previously) and archive the marine geophysical data, Colin Devey for providing high-resolution spreading axis bathymetry of segments 1-3, and Geoff Abers for the microseismicity data set. Reviewer 1 and Laura Wallace made suggestions that helped us to improve the manuscript. Funding for this study was provided by the University of Hawaii at Manoa and the National Science Foundation to BT. This is SOEST contribution number 11139.