Corresponding author: G. Fitz, Hess Corporation, Hess Tower, 1501 McKinney, Houston, TX 77010, USA. (email@example.com)
 The D'Entrecasteaux Island (DEI) gneiss domes are fault-bounded domes with ~2.5 km of relief exposing ultrahigh-pressure (UHP) and high-pressure (HP) metamorphic gneisses and migmatites exhumed in an Oligocene-Miocene arc-continent collision and subduction zone subject to late Miocene to recent continental extension. Multichannel seismic reflection data and well data show the Trobriand basin formed as a fore-arc basin caused by southward Miocene subduction at the Trobriand trench. Subduction slowed at ~8 Ma as the margin transitioned to an extensional tectonic environment. Since then, the Trobriand basin has subsided 1–2.5 km as a broad sag basin with few normal faults deforming the basin fill. South of the DEI, the Goodenough rift basin developed after extension began (~8 Ma) as the hanging wall of the north-dipping Owen-Stanley normal fault that bounds the basin's southern margin. The lack of upper crustal extension accompanying subsidence in the Trobriand and Goodenough basins suggests depth-dependent lithospheric extension since 8 Ma has accompanied uplift of the DEI gneiss domes. Structural reconstructions of seismic profiles show 2.3–13.4 km of basin extension in the upper crust, while syn-rift basin subsidence values indicate at least 20.7–23.6 km of extension occurred in the entire crust since ~8 Ma. Results indicating thinning is preferentially accommodated in the lower crust surrounding the DEI are used to constrain a schematic model of uplift of the DEI domes involving vertical exhumation of buoyant, postorogenic lower crust, far-field extension from slab rollback, and an inverted two-layer crustal density structure.
 The D'Entrecasteaux Islands (DEI) of eastern Papua New Guinea are an ~150 km long, northwest-southeast trending chain of three fault-bounded topographic domes with up to 2.5 km of relief exposing ultrahigh-pressure (UHP) and high-pressure (HP) metamorphic gneisses and migmatites. The DEI gneiss domes have been exhumed in a zone of continental extension ahead of the westward propagating Woodlark seafloor spreading system that initiated in the late Miocene (~6–8 Ma) [Taylor et al., 1999] (Figure 1). Plate motion predictions based on magnetic isochron anomalies in Woodlark basin show that seafloor spreading rapidly propagated ~500 km westward at ~25 mm/yr after a uniform degree of ~200 ± 40 km of north-south continental extension [Taylor et al., 1999] (Figure 1). Previous onland mapping and radiometric dating studies interpret the DEI exposures of middle and lower crust rocks as metamorphic core complexes (MCCs) exhumed by significant amounts (>175 km) of continental extension preceding oceanic spreading in eastern Papua New Guinea [Davies and Warren, 1992; Hill, 1994; Baldwin et al., 2004; Webb et al., 2008; Baldwin et al., 2008a, 2008b] (Figure 1). However, interpretation of multichannel seismic (MCS) reflection data and well data from the Trobriand basin and the Goodenough basin adjacent to the Goodenough and Fergusson Islands of the DEI showed no evidence for the large amount of upper crustal extension predicted by the Woodlark basin opening models of [Taylor et al., 1999]. The Trobriand and Goodenough basins have subsided 1–2.5 km since ~8 Ma, contemporaneous with the uplift of Goodenough and Fergusson Islands while demonstrating little upper crustal normal faulting or any syn-rift sedimentary wedges controlled by faults bounding the gneiss domes.
 Large discrepancies between measurements of upper crustal extension accommodated by brittle faulting and estimates of total crustal and lithosphere extension are observed at many rifted margins [Driscoll and Karner, 1998; Davis and Kusznir, 2004; Lavier and Manatschal, 2006; Reston, 2009; Huismans and Beaumont, 2011]. Explanations of these discrepancies include (1) limitations due to the resolution of seismic reflection data preventing accurate quantitative extensional measurements [Marrett and Allmendinger, 1992]; (2) complex polyphase fault geometries have led to underestimates of total upper crustal extension [Reston, 2009]; and (3) the occurrence of depth-dependent extension of the lithosphere where the lower crust and mantle lithosphere accommodate more extension than the upper crust. The process of depth-dependent extension is predicted to play an important role in several models of lithospheric extension including (1) the simple shear lithosphere extensional detachment model [Wernicke, 1985; Driscoll and Karner, 1998; Davis and Kusznir, 2004]; (2) the decoupled pure shear model [Lavier and Manatschal, 2006]; and (3) the upwelling divergent flow within extending continental lithosphere model [Kusznir and Karner, 2007].
 The motivation of this work is to understand the relationship in this area between upper crustal extension, lower crustal extension, and mantle lithosphere extension, as well as the respective roles they play in gneiss dome exhumation. Seismic receiver studies have determined the crust thins from 32–43 km beneath the Papuan Peninsula to 20–26 km beneath the DEI and thickens again to >35 km in the Trobriand basin fore-arc high area [Abers et al., 2002] (Figure 1). Crustal thinning and elevation of the Moho suggest the current high topography of the Goodenough and Fergusson Islands gneiss domes is isostatically supported by mantle upwelling. This study seeks to understand the mechanism of uplift of the Goodenough and Fergusson Islands gneiss domes by comparing upper crustal extension to lower crustal extension in offshore basins surrounding the islands. Structural reconstructions of MCS seismic line interpretations in the Trobriand and Goodenough basins and related onland geology from ~8 Ma to recent constrain estimates of upper crustal extension. Subsidence analysis determined by flexural backstripping of ~8 Ma to recent sediments in the Trobriand and Goodenough basins provides lower crustal extension estimates. The assessment of depth-dependent crustal extension here is used to quantify crustal behavior during the uplift of the Goodenough and Fergusson Islands gneiss domes.
2 Tectonic Setting of Eastern Papua New Guinea
 The tectonics of eastern Papua New Guinea are controlled by the west-southwest oblique convergence of the Pacific and Australian plates and the resulting interactions of small microplates within this broad zone of interplate convergence [Hall, 2002] (Figure 1). Microplate interaction began in the Paleogene with collision between the northeast Australian continental margin and the northern island arc system of Papua New Guinea [Davies and Warren, 1988]. Orogenic thickening occurred as the partially subducted Australian continental fragment, the Papuan Plateau, collided with the island arc. Arc-continent collision also led to southward obduction of the Papuan ultramafic body (PUB) ophiolite body over the Papuan continental plateau [Davies and Jaques, 1984; Lus et al., 2004]. The Paleocene Papuan Orogen forms the central core of the southeastern Papuan Peninsula that is 1–3 km above sea level and underlain by continental crust up to 35 km thick at the eastern end of the peninsula [Abers et al., 2002; Ferris et al., 2006]. Farther to the west, seismic refraction studies show unrifted Papuan Peninsula crust to be up to 50 km thick [Finlayson et al., 1976].
 The Owen-Stanley fault zone (OSFZ) is the active plate boundary juxtaposing continental rocks of the Papuan orogeny (including the Owen-Stanley metamorphic belt) from the overlying PUB to the north on the Woodlark plate [Davies and Jaques, 1984; Daczko et al., 2011] (Figure 1). The Owen-Stanley plate boundary fault is linked to the Woodlark spreading ridge by the Trobriand transfer fault [Little et al., 2007]. GPS data show that the OSFZ becomes increasingly strike-slip and transpresssional in the western part of the Papuan Peninsula increasingly extensional in this study area and on the Woodlark spreading ridge [Wallace et al., 2004] (Figure 1). The PUB outcrops on the Papuan Peninsula north and east of the Owen-Stanley fault zone from ~147°E to ~149.5°E and on Goodenough and Fergusson Islands over high-grade metamorphic rocks [Davies, 1971; Davies and Warren, 1988]. Offshore, the PUB extends at least as far east as the Moresby Seamount area based on the presence of recovered Paleocene gabbro and diabase recovered by ODP Leg 180 drilling [Monteleone et al., 2001] (Figure 2).
 A reversal of subduction polarity has been proposed in eastern Papua New Guinea in the early Miocene (possibly late Oligocene) as northward subduction of the Australian continental margin [Pigram et al., 1989] stalled and continued convergence of the Australia and Pacific plates became accommodated primarily by southward subduction of Solomon Sea oceanic crust along the Trobriand trench [Davies and Smith, 1971; Davies et al., 1987; Honza et al., 1987., Davies and Warren, 1988; Quarles van Ufford and Cloos, 2005] (Figure 1). Subduction along the Trobriand trench has been cited as a possible source for Miocene arc magmatism throughout the Papuan Peninsula that intruded continental basement and the PUB [Taylor and Huchon, 2002; Quarles van Ufford and Cloos, 2005]. Southward subduction at the Trobriand trench also generated the Trobriand fore-arc basin, a 3500 km2 extensional fore arc that initiated in the early/middle Miocene and filled with 5–7 km of clastic and carbonate sedimentary rocks [Taylor and Huchon, 2002; Francis et al., 1987] (Figure 2). Available geophysical evidence shows post-Miocene subduction slowed at the Trobriand trench, and presently only a few millimeters of convergence per year is occurring there [Davies and Jaques, 1984]. Seismicity beneath the Trobriand trench also lacks an organized pattern indicating southward subduction [Abers et al., 2002]; however, active volcanism that occurs here has been attributed to either southward subduction of the Solomon Sea lithosphere at the Trobriand trench [Smith and Davies, 1976; Martinez and Taylor, 1996] or rift-related decompression melting of subduction-modified mantle [Johnson et al., 1978]. The modern volcanic front trends east from Mt. Lamington, which erupted in 1951 [Taylor, 1958], to Mt. Victory, which erupted ~1870 [Taylor, 1957], followed by the D'Entrecasteaux Islands that are skirted by Pleistocene andesitic volcanic cones, the Amphlett Islands, and the Egum Atoll [Taylor and Huchon, 2002] (Figure 2).
 The exhumed high-pressure (HP) and ultrahigh-pressure (UHP) lower plate quartzo-feldspathic gneiss domes of Goodenough and Fergusson Islands breached an upper plate composed of the PUB, volcanic, and sedimentary cover rocks along kilometer-scale mylonitic shear zones dipping ~20°–40° northeast and 24°–56° to the northwest [Davies and Warren, 1988; Davies and Warren, 1992; Hill, 1994] (Figure 2). The domes contain abundant migmatites and two phases of granodiorite pluton intrusions at ~4.6 and ~2.1 Ma [Hill et al., 1995]. Pods and dikes of mafic eclogites in the gneiss domes have in situ zircon U-Pb ion microprobe ages of ~8–2 Ma show and show younging from an east to west direction from Fergusson to Goodenough Islands [Baldwin et al., 2004; Monteleone et al., 2007]. HP and UHP mineral assemblages from continental rocks of Australian affinity indicate the Goodenough and Fergusson Islands metamorphic rocks exhumed from 50 km to possibly >75 km at plate tectonic rates of >1 cm/yr [Monteleone et al., 2007; Baldwin et al., 2008a, 2008b]. The DEI are ringed by small Pleistocene to recent monogenetic volcanic cones and active geothermal pools that indicate a high geothermal gradient beneath the islands that decreases rapidly toward the surrounding basins [Martinez et al., 2001; Mann and Taylor, 2002].
 Low-temperature thermochronologic dating of the domes indicating rapid cooling from ~550°C to 100°C since <4 Ma have been cited as evidence for continued exhumation [Baldwin et al., 1993]. On the other hand, a study of late Quaternary coral reefs indicate the DEI have not been uplifting over the past 10 ka but instead have been stable or subsiding [Mann and Taylor, 2002].
 The Goodenough basin south of the Goodenough and Fergusson Islands developed during the same late Miocene to recent period as the unroofing of the gneiss domes from 8 to 2 Ma [Monteleone et al., 2007]. Upper crustal extension in the Goodenough basin is primarily focused on the Owen-Stanley fault zone along the southern edge of the basin (Figures 2 and 4). Normal displacement on the Owen-Stanley fault, which dips northward beneath the offshore Goodenough basin at a shallow angle at 20°–30° [Spencer, 2010; Daczko et al., 2011] has caused modest hanging wall deformation of upper Miocene to Pleistocene marine sedimentary rocks in the Goodenough basin. The hanging wall has flexed downward in the Goodenough basin, expressed in bathymetry by water depth increases from 1000 to 1400 m (Figure 2). The Papuan Peninsula is rapidly uplifting at ~2 mm/yr as part of the footwall of the Owen-Stanley fault [Mann and Taylor, 2004].
2.1 Late Miocene to Recent Extensional Tectonics of Eastern Papua New Guinea
 Microplate formation due to oblique convergence of the Pacific and Australia plates has led to counterclockwise rotation of the Woodlark plate relative to the Australian plate and westward propagation of extension in the Woodlark basin [Weissel et al., 1982; Wallace et al., 2004]. Seafloor spreading in the eastern Woodlark basin transitions to continental extension in the D'Entrecasteaux Island/eastern Papuan Peninsula margin, then to convergence between the Woodlark and Australian plates in the western Papuan Peninsula (Figure 1).
 A late Miocene regional unconformity dated at 8.4 Ma constrains the onset of Woodlark rifting and provides a paleo-sea level surface to track subsequent, post–8.4 Ma subsidence [Taylor and Huchon, 2002] (Figure 2). Euler pole solutions of Woodlark plate motion relative to the Australia plate determined by magnetic isochron anomalies and fracture zone traces show that the majority of the basin opening from ~6 Ma to 0.54 Ma has occurred as rotation at a rate of 4.234° Ma−1 around a single Euler pole (147° ± 1°E–2°E, 9.3°S ± 0.2°S) [Taylor et al., 1999]. Rifting has propagated ~500 km westward at ~25 mm/yr, proceeding along zones of weakness determined by the suture between the collided continental crust of Australia, and ultramafic and arc material [Taylor et al., 1999; Taylor and Huchon, 2002].
 Following the onset of rifting, the northern Woodlark rise margin is characterized by 2–3 km of subsidence with little visible faulting, while the southern Pocklington rise margin has subsided similar amounts while being deformed by diffuse upper crustal extensional faulting [Goodliffe and Taylor, 2007]. Sometime after 1.6 Ma, extension focused west of the seafloor spreading tip on the Moresby fault, an active, north dipping low-angle fault bounding the Moresby Seamount [Taylor and Huchon, 2002]. The spreading center underwent a 22° counterclockwise reorientation at in the Pleistocene (0.52 Ma) that shifted the Euler pole location and resulted in a more northeast-southwest opening direction [Goodliffe, 1998; Wallace et al., 2004].
 Unmetamorphosed to high-pressure (HP) metamorphic rocks that formed due to subduction of the Australian continental margin have been exhumed during the late Miocene (8.4 Ma) to recent extension along the Australia-Woodlark plate boundary [Webb et al., 2008]. Reactivation of the shallowly north dipping Owen-Stanley fault zone facilitated exhumation of unmetamorphosed to low-grade metamorphic rocks along top-to-the-north shear zones bounding the Suckling-Dayman massif [Daczko et al., 2009], Misima Island [Baldwin et al., 2008a, 2008b], the Provost Range massif of Normanby Island, and the Moresby Seamount [Little et al., 2007] (Figures 1 and 2). Mineral lineations and fault surface megacorrugations trend roughly parallel to the 0.5–6 Ma direction of opening between the Australia and Woodlark plate on the Dayman Dome [Dazcko et al., 2011], Misima Island [Baldwin et al., 2008], and the Prevost Range MCC [Little et al., 2007].
 The Goodenough and Fergusson Islands gneiss domes are located ~80 km north of the OSFZ (Figure 2). The offshore Trobriand basin north and west of the DEI contains Miocene-recent strata that have not been significantly deformed by normal faults, suggesting there has been no reorganization of the plate boundary here since Woodlark-related extension began in the late Miocene (8.4 Ma). A systematic variation to lineations and foliations of the Goodenough and Fergusson Islands gneisses show flow occurred orthogonal to the plate boundary direction and was mechanically decoupled from plate motion in the rift [Little et al., 2011]. Opening of the Woodlark basin primarily around a single Euler pole based on a rigid plate model implies north-south rifting initiated simultaneously at ~8.4 Ma along the length of the plate boundary from the eastern border of Woodlark basin to ~149°E [Goodliffe, 1998]. A second implication is the transition from rifting to seafloor spreading occurred after a uniform degree of 200 ± 40 km north-south extension [Taylor et al., 1999]. Previous studies of continental extension ahead of Woodlark seafloor spreading have identified significantly less upper crustal extension, both in the Trobriand and Goodenough basin/DEI area [Fang, 2000], and the Moresby Seamount area including the Pocklington and Woodlark rise margins [Goodliffe, 1998; Fang, 2000; Kington and Goodliffe, 2008].
3 Data and Methods
 Upper crustal extension determined by structural reconstructions and total crustal extension predicted from subsidence values are presented in this study based on geologic interpretations of four MCS reflection lines from 1992 leg 3 cruise (cruise EW9203) of the R/V Maurice Ewing and one previously published exploration MCS reflection line [Tjhin, 1976]. The Goodenough #1 well was used to correlate the ages of units interpreted from MCS line 1-112 including the 8.4 Ma surface previously discussed by Taylor and Huchon  (Figure 2). Upper crustal extension and subsidence values were obtained by performing structural reconstructions and flexural backstripping of sediments to the 8.4 Ma unconformity surface that marks the onset of rifting and provides a datum for the paleo-sea level surface [Taylor and Huchon, 2002]. The selected lines form three north-south profiles, roughly parallel to the direction of extension from the late Miocene (~6) to the Pleistocene (0.5 Ma) (Figure 2).
 Profile 1 is based on an interpretation of MCS line 1181 in the Trobriand basin 28 km west of Goodenough Island (Figure 2). Profile 2 is based on interpretations of MCS line 1168 in Trobriand basin 6 km west of Goodenough Island and MCS line 1190 in Goodenough basin (Figure 2). Profile 3 is based on interpretations of MCS line 1-112 in Trobriand basin 6 km north of Goodenough Island and MCS line 1193 in Goodenough basin (Figure 2).
3.1 Measuring Upper Crustal Extension
 Upper crustal extension was measured using structural reconstructions based on geologic and structural interpretations along the three profiles (Figure 2). A technique of sequential removal of sedimentary units was used combining 2-D flexural backstripping and palinspastic reconstruction. This result allowed the isostatic response of the lithosphere to sedimentary loading and unloading to be included in the construction of balanced cross sections [Lavier et al., 2000].
 Midland Valley's 2D Move v. 2011.1 software program was used to depth convert cross-section interpretations and generate balanced structural restorations. Cross sections were converted from two-way travel time to depth using interval velocities determined from velocity analysis of the EW9203 MCS data. All cross sections maintain a 1:1 aspect ratio before modeling.
 Sequential 2-D backstripping involves the removal of each sedimentary layer while accounting for the physical and geologic processes responsible for deformation [Lavier et al., 2000]. First, the sedimentary unit above the horizon of interest is removed, correcting for compaction of the sediments beneath it and the flexural isostatic response of the crust due to the weight of the overlying sediments. Next, deformation due to faulting is reconstructed using the inclined shear method described by White et al.  that restore hanging wall and footwall surfaces to their original positions that are assumed to be smooth surfaces. Finally, the profile is corrected for subsidence and sea level change based on the time-appropriate depositional settings found in the Goodenough #1 well and seismic interpretation. This workflow is repeated for each interval from the present to the start of rifting at 8.4 Ma, providing total extensional measurements along each profile from which rates of tensile shear and a horizontal stretching factor [β = final width/original width] can be determined.
 Midland Valley's 2DMove v. 2011.1 software program was also used to model the downdip geometry of the Owen-Stanley fault that bounds the southern Goodenough basin. The Owen-Stanley fault is a major plate boundary fault that accommodates most of the extension and strain in this part of eastern Papua New Guinea [Wallace et al., 2004]. In the south of Goodenough Bay, the Owen-Stanley fault is mainly in an offshore, coastal setting and has not been imaged by MCS reflection data.
 MCS reflection line 1190 across the Goodenough basin images sediment wedges and normal faults of the hanging wall of the Owen-Stanley fault that underlies the offshore Goodenough basin. The Owen-Stanley fault controls the southward basement flexure of the Goodenough basin and is modeled as having a listric geometry here due to the presence of syn-rift growth wedges in Goodenough basin, a rollover anticline with crestal normal faults (the Goshen fault zone), and accompanying synthetic and antithetic brittle faults in the Goodenough basin hanging wall. These structures are similar to those produced in analog experiments of hanging walls of listric fault models [McClay and Scott, 1991; Erikson et al., 2000].
 I have used the inclined shear method to model the geometry and the depth to detachment of the Owen-Stanley fault [White et al., 1986; Fossen et al., 2003]. Modeling parameters required an initial fault dip at the surface to be selected for the Owen-Stanley fault; however, because the Owen-Stanley fault is not seismically imaged, a fault dip was selected based on the dip of the Dayman dome shear surface ~35 km northwest and along strike of the OSFZ at Goodenough Bay. In this study, the Dayman dome is interpreted to have been exhumed due to footwall uplift along the same normal fault that bounds the south of Goodenough Basin. A fault dip of 35° was used based on studies of the Daymon dome shear surface [Spencer, 2010; Daczko et al., 2011], and an inclined antithetic shear angle of 55° was used based on the dip of antithetic faults in Goodenough basin.
3.2 Estimating Crustal Thinning and Extension From Subsidence
 Syn-rift subsidence along the three 2-D profiles provides a second input used to predict crustal thinning and extension in the DEI area. The sequential backstripping and palinspastic reconstruction technique described above was reapplied to the three profiles but did not correct for subsidence. Syn-rift sedimentary units were removed in order of youngest to oldest while correcting for compaction, flexural loading, sea level change, and structural deformation due to brittle upper crust faulting. This approach provides a 2-D profile of water-loaded subsidence for each cross section, assuming the 8.4 Ma surface represents paleo-sea level [Taylor and Huchon, 2002]. Because tectonic subsidence due to upper crustal extension is accounted for in the 2-D profiles of water-loaded subsidence, it is reasonably assumed that observed crustal extension measured from these subsidence values predominantly occurs in the middle and lower crust.
 Crustal thinning can be determined from syn-rift subsidence values by assuming thinning and thermal expansion of the mantle lithosphere during rifting, where crustal thinning produces subsidence while lithospheric thinning inhibits subsidence [McKenzie, 1978; Turcotte and Schubert, 2002].
 Equation ((1)) calculates the crustal thinning factor β by using backstripped water-loaded subsidence values assuming Airy isostatic principles and instantaneous extension [McKenzie, 1978]. Although the subsidence values used in equation ((1)) were obtained through flexural backstripping that take into account the strength of the lithosphere, calculations of crustal thinning assuming Airy isostasy due not consider lithospheric strength.
= Crustal thinning factor
= Thickness of sedimentary basin
= Density of mantle
= Density of continental crust
= Density of basin infill
= Pre-rift crustal thickness
= Thermal expansion coefficient of mantle
= Asthenosphere temperature
= Surface temperature
= Pre-rift thermal lithospheric thickness
 Use of flexural isostatic principles to determine crustal thickness depends critically on knowing the depth of strength maxima in the crust, an unknown in this area [Watts, 2001]. However, any significant flexural rigidity not accounted for in the flexural backstripping of sediments would result in underestimates of crustal extension when applying Airy isostasy. Additionally, assuming instantaneous extension implies that the crust and mantle lithosphere were stretched at a short time interval relative to the time needed for thermal cooling of the lithosphere [McKenzie, 1978]. Jarvis and McKenzie  conclude that the instantaneous stretching model is a reasonable approximation if the duration of stretching is less than 20 Myr and that giving the lithosphere sufficient time to cool and thermally contract would reduce crustal extension measurements.
 The crustal thinning factor is equivalent to the horizontal stretching factor assuming crustal volume is conserved during extension, allowing estimates of total horizontal extension to be made by summing the stretching factor determined from subsidence over the extended margin [Kington and Goodliffe, 2008]. This assumption ignores the effects of lower crustal flow or thermal expansion that are both likely occurring in the Woodlark and DEI area [Taylor and Huchon, 2002; Goodliffe and Taylor, 2007; Kington and Goodliffe, 2008; Little et al., 2007]. Flow of the middle or lower crust from beneath Trobriand and Goodenough basins toward focused extensional areas along the Owen-Stanley fault zone and/or the DEI would remove lower crust material from the cross-sectional areas. Hence, estimates based on this calculation are the minimum observable amounts of extension as opposed to total extension estimates that would include the effects of lower crustal flow.
 Estimates of extension also depend heavily on estimating pre-rift crustal thickness. A value of 50 km was used based on refraction seismic profiles of the unrifted, western part of the Papuan Peninsula [Finlayson et al., 1976], while adopting lower values would reduce extensional estimates. However, the HP and UHP mineral assemblages from Goodenough and Fergusson Islands indicate the underlying crust was at least 50 km thick in this area prior to rifting [Monteleone et al., 2007] and possibly greater than 75 km, hence a pre-rift crustal thickness value of 50 km likely would not overestimate the amount of extension. Other parameters used in the analysis are standard values that include a crustal density of 2800 kg/m3, a mantle density of 3300 kg/m3, a thermal expansion coefficient of the mantle of 3 × 10ˉ5 Kˉ1, and a lithospheric thickness of 160 km to the 1300°C isotherm.
3.3 Accounting for Thermal Subsidence Due to Miocene Fore-Arc Extension in the Trobriand Basin
 One-dimensional backstripping analysis [Steckler and Watts, 1978] of sedimentary units from the Goodenough #1 well (Figure 2) provides a means to account for possible ongoing thermal subsidence occurring after Miocene fore-arc extension in the Trobriand basin. Backstripping analysis was performed using the ZetaWare, Inc. Trinity basin modeling software, where syn-rift and post-rift sedimentary units were removed in order of youngest to oldest while correcting for compaction, flexural loading, and sea level change. Water depth, lithology, and age control for the Goodenough #1 well were obtained from Harris et al.  and Taylor and Huchon .
 Backstripping the fore-arc syn-rift and post-rift sequences deposited prior to the 8.4 Ma unconformity provided a method to first determine total Miocene subsidence, which was then entered into equation ((1)) to estimate the crustal thinning factor β for Miocene fore-arc extension. Equation ((2)) [McKenzie, 1978] was then used to calculate the expected thermal subsidence that occurred during the period from 8.4 Ma to the present. The thermal subsidence value from the Goodenough #1 well calculated using equation ((2)) was then compared to the syn-rift subsidence values along the three 2-D profiles.
= Thermal Subsidence
= Age since rifting
= Thermal decay time constant for the lithosphere
 An age since rifting of 11 Ma or 3.47 × 1014 s was used to calculate expected thermal subsidence since the end of Miocene rifting. The thermal decay time constant for the lithosphere τ was determined by equation (3).
= Thermal diffusivity.
 A thermal diffusivity value of 10− 6 m2s− 1 was used, resulting in a thermal decay time constant of 2.59 × 1015 s or 82 Ma.
4 Structural Restorations
4.1 Profile 1
 Figure 3 shows a depth-converted 98.2 km long north-south cross section interpreted from MCS line 1181 in Trobriand basin that is located 26 km west of Goodenough Island. Extension in this area has had a primarily north-south orientation from at least 6 Ma to 0.5 Ma [Taylor et al., 1999], indicating the cross section is roughly orthogonal to the strike of the faults.
 MCS line 1168 contains five interpreted stratigraphic sedimentary sequences in the Trobriand basin which have been labeled sequences 1–5 in stratigraphic order. Sequences 1–3 were deposited in the fore-arc setting of Trobriand basin from the early to late Miocene [Fang, 2000]. The top surface of sequence 3 is interpreted to represent the 8.4 Ma erosional unconformity identified in the Goodenough #1 well [Taylor and Huchon, 2002]. Above the unconformity, sequences 4 and 5 deposited from 8.4 Ma to recent, representing “syn-rift” deposits in this area. The increased thickness of sequence 5 compared to sequence 4 over a shorter time period indicates subsidence has accelerated since 1.8 Ma.
 In the north of the cross section, an array of normal faults is present that formed in the early/middle Miocene and was inverted in the late Miocene during uplift of the fore-arc high (Figure 3). Interpretation of these faults shows they were not active in this basin since the late Miocene and show no observable displacement due to normal faulting. For this reason, they were not reconstructed and included in measurements of upper crustal extension.
 In the central area of the cross section, the Goodenough structural high bounded by faults dipping 55° to the north and 47° to the south locally deforms overlying and directly adjacent sediments from sequences 1–4. The structural high is aligned with the DEI and Cape Victory volcano that is parallel to the Trobriand trench (Figure 1). Sequence 5 is undeformed above the Goodenough structural high. Flexural backstripping of sequence 5 and restoration of faults deforming sequence 5 yields 0.6 km of upper crustal extension in the past 1.8 Myr (Figure 3). High-angle normal faults south of the offshore structural high predominantly accommodated extension here. Backstripping and restoration of faults in sequence 4 shows an additional 1.7 km of extension accommodated by restoring the faults that bound the structural high (Figure 3). A total 2.3 km of extension since the start of Woodlark plate extension 8.4 Ma gives an upper crust horizontal stretching factor β = 1.02 (Table 1).
Table 1. Comparison of Upper Crustal to Middle and Lower Crustal Extension Values from the Trobriand and Goodenough Basins
4.2 Profile 2
 Profile 2 (Figure 4) includes depth-converted geologic cross sections interpreted from MCS line 1168 6 km west of Goodenough Island in Trobriand basin, MCS line 1190 in Goodenough basin, and a projected geometry of the Owen-Stanley fault based on deformation in the hanging wall. The Trobriand basin and the Goodenough basin are separated by the Cape Vogel structural high between the Cape Vogel Peninsula and Fergusson Islands (Figure 2).
 The offshore Goodenough structural high is ~34 km wide in the north-south direction in MCS line 1168 (Figure 4). Above the high, sequence 3 is thinned by erosion, and sequence 4 is also thinned because the high remained elevated from the late Miocene to the Pleistocene (~8.4 to ~1.8 Ma). The high began to subside after the Pleistocene (~1.8 Ma), and sequence 5 was deposited unconformably.
 The Goodenough basin contains distinct Plio-Pleistocene sedimentary sequences because the Goodenough basin became separated from the Trobriand basin in the late Miocene by uplift of the DEI and the Cape Vogel structural high. The basin geometry is controlled by extension on the shallowly dipping Owen-Stanley fault [Spencer, 2010; Daczko et al., 2011], leading to basement flexure, syn-rift growth wedges, a rollover anticline, and synthetic and antithetic brittle faults. Extension here also led to uplift of the Papuan Peninsula footwall.
 Figure 4 shows backstripping of sequence 5 and fault reconstruction of syn-depositional deformation. 1.5 km of extension from the Pleistocene (~1.8 Ma) to recent was accommodated primarily in a small graben south of the offshore Goodenough structural high that appears to be an isolated rift (Figure 4). The faults bounding the Goodenough high are also responsible for minor amount of extension in this area.
 In profile 2, more extension has been accommodated in the Goodenough basin, with removal of strata in Goodenough basin resulting in 4.4. km of stretching. Reconstruction of profile 2 in Figure 5 shows the proposed margin configuration in the late Miocene (~8.4 Ma), where the Trobriand basin is approximately at sea level and the Goodenough basin is subaerially exposed. These paleo-elevation estimates are confirmed in the Trobriand basin by the presence of late Miocene fluvial-deltaic deposits in Goodenough #1 well and in the Goodenough basin by a prominent, jagged basement reflector characteristic of subaerial exposed surfaces. The Papuan Peninsula at this time was likely a topographically elevated orogeny [Taylor and Huchon, 2002], so it is possible Goodenough basin was also elevated at this time and would result underestimates of its extension. The offshore Goodenough structural high was also elevated at this time in the Pliocene (1.8 Ma) as a result of 2.8 km of extension.
 In the Goodenough basin, 9.1 km of extension has occurred in the past 8.4 Ma, resulting in a horizontal β = 1.21, while in Trobriand basin, 4.3 km of extension results in a horizontal β = 1.05 (Table 1).
4.3 Profile 3
 Profile 3 (Figure 5) includes depth-converted geologic cross sections interpreted from MCS line 1-112 6 km north of Goodenough Island in Trobriand basin, MCS line 1193 in Goodenough basin, and a projected geometry of the Owen-Stanley fault. MCS line 1-112 contains the same sedimentary sequence as line 1181 and 11868, while a second sequence appears in MCS line 1193 in Goodenough basin. The yellow sequence 1 is part of a series of backstepping delta clinoforms that deposited mostly east of MCS line 1193 due to the unroofing of the Fergusson Island gneiss dome. Although the structure of the Goodenough dome is not known at depth along profile 3, it is assumed to be similar to the structure of the offshore high in MCS line 1168 (Figure 4).
 For this profile, separate reconstructions from the Trobriand and Goodenough basin are combined at 8.4 Ma. Figure 5 shows the backstripping of a young, thin unit of pelagic sediments deposited in Pleistocene to recent time as a result of increased subsidence in the Goodenough basin. Reconstruction of faults that occurred during this period of subsidence resulted in 1.3 km of extension. In the Goodenough basin, an array of normal faults that occurred after deposition of sequence 1 accounts for 1.8 km of extension. In the Trobriand basin, backstripping of sequence 5 and reconstruction of extension from 1.8 Ma to recent results in 0.3 km of extension due to a paucity of recent, active normal faults.
 Figure 5E shows the proposed basin configuration of the Trobriand basin at 8.4 Ma, while also accounting for ~32 km of extension occurring in the Goodenough Island area. At this point, Trobriand basin is elevated to sea level. Reconstruction of faults from 8.4 Ma to recent shows that 0.4 km of extension has occurred along the MCS line 1-112 profile resulting in a horizontal β = 1.006. Restoring the basement surface of Goodenough basin to above sea level results in minimum estimate of 5.9 km of extension and provides a horizontal β = 1.1 (Table 1).
5 Crustal Thinning and Extension Measured From Water-Loaded Subsidence Values
 2-D profiles of water-loaded subsidence determined by flexural backstripping and structural reconstruction provide another method by which to measure crustal extension. Because upper crustal extension was corrected for when obtaining subsidence values, it is assumed that crustal extension measured from subsidence predominantly occurs in the middle and lower crust.
 Profile 1 (Figure 6) of MCS Line 1181 shows β increasing to ~1.2 on the northern side of the cross section to nearly 1.4 in the central area of the basin over the structural high before decreasing to 1.25 on the southern side of the cross section. This yields a total of ~20.5 km of middle/lower crustal extension across profile 1, with a tensile strain of 21.6%.
 Profile 2 shows the variability of crustal stretching and extension from MCS line 1168 in the Trobriand basin to MCS line 1190 in the Goodenough basin (Figure 6). In the Trobriand basin, stretching factor values are lower than in profile one, with an average β of 1.16 in the north decreasing to 1.1 over the structural high due to less subsidence closer to Goodenough Island. MCS line 1190 has higher overall stretching factors, increasing from ~1.1 in the north to nearly 1.9 on the south side of the basin. Summing the north-south values of estimated extension predicts 9.2 km of extension in the area of MCS line 1190 (tensile strain = 16.5%) and 13.9 km of extension in the area of MCS line 1168 (tensile strain = 14.5%).
 In profile 3 north of Goodenough Island, MCS line 1-112 shows subsidence increasing from the north and south sides of Trobriand basin where β is equal to 1.1–1.2 to the center of the basin where β approaches a value of 1.4. MCS line 1-112 has an average β = 1.21. The total extension predicted from subsidence for this MCS line 1-112 equals 12.5 km (tensile strain = 18.1%). South of Goodenough Island in the Goodenough basin, MCS line 1193 shows a more variable β value, although the average β = 1.24 is similar to MCS line 1-112. 11.1 km of extension is predicted from subsidence providing a tensile strain of 19.1%. In the central area of profile 3 where Goodenough Island is located, the crustal thinning value is unavailable from this offshore seismic data set; however, seismic receiver function studies indicate the crust is 26–29 km beneath Goodenough Island [Abers et al., 2002]. Using the same original crustal thickness value of 50 km results in a crustal thinning β ≈ 1.8. This β value may be higher than 1.8 if crustal thickening focused below the islands during the Paleogene Papuan orogeny. The high crustal thinning values at Goodenough Island compared to the Trobriand and Goodenough basins show strain has focused in a small area ~30–35 km wide since the onset of extension 8.4 Ma.
6 Backstripping Goodenough #1 Well Miocene Sedimentary Units to Calculate Thermal Subsidence due to Miocene Fore-Arc Extension in the Trobriand Basin
 Figure 7 shows a subsidence curve for the Goodenough #1 well from the Trobriand Basin. The Goodenough #1 penetrates Holocene through Miocene sediment and encounters lower Miocene/upper Oligocene volcanic basement rock [Taylor and Huchon, 2002]. Total subsidence remained at ~1100 m from the end of fore-arc extensional rifting ~11 Ma to the 8.4 Ma unconformity (Figure 1). Using 1110 m as a value for Miocene syn-rift subsidence in equation ((1)) results in a thinning factor of β = 1.08.
 Once the Miocene syn-rift β was determined, thermal subsidence since the end of rifting 11 Ma could then be predicted using the time-dependent subsidence equation ((2) [McKenzie, 1978]. Using β = 1.08, a pre-rift thermal lithospheric thickness (γLO) of 160 km, a time since rifting of 11 Ma, and a thermal decay time constant of 2.59 × 1014 s, or 82 Ma predicts only 25 m of thermal subsidence would be expected due to Miocene fore-arc extension 11 Ma after the end of rifting. This number is small in part because not much extension occurred in the Miocene fore arc, as indicated by the Miocene syn-rift β of 1.08, and also because of the relatively short time since rifting, 11 Ma, of a basin with a thermal decay constant of 82 Ma. The thermal decay constant is mainly dependent on the pre-rift thermal lithospheric thickness. Overall, this small thermal subsidence value shows that a lithosphere 160 km thick with continental crust 50 km thick would not be subject to significant post-rift thermal subsidence after such a small amount of extension.
7.1 Crustal extension estimates in the Trobriand and Goodenough basins
 In the Trobriand basin, structural reconstructions of geologic interpretations of MCS lines 1181, 1168, and 1-112 show little upper crust brittle extension since ~8.4 Ma, totaling 2.3, 4.3, and 0.4 km of extension, respectively (Table 1). In contrast, extension estimates based on subsidence values from the same lines are much larger, reaching 20.7 km in MCS line 1181, 13.9 km in MCS line 1168, and 12.5 km in MCS line 1-112. The larger estimates from the subsidence analysis indicate that extension in the middle and lower crust is greater, supporting the possibility of depth-dependent lithospheric extension where the lower crust and mantle lithosphere accommodate more extension than the upper crust. Although each of these results are subject to a degree of uncertainty, error analysis was not performed due to the consistent trend of extension estimates along multiple transects.
 In the Goodenough basin, there is less of a crustal extension discrepancy between extension estimates based on the two methods. Structural reconstruction of MCS line 1190 (Figure 4) shows 9.1 km of upper crustal extension affected this part of the Goodenough basin largely as the result of movement on the listric Owen-Stanley fault zone; offset from high-angle normal faults in the hanging wall of the fault account for a small percentage (10%–15%) of extension. Extension predicted from subsidence on MCS line 1190 is a similar value of 9.2 km (Figure 6) (Table 1). On MCS line 1193, 5.9 km of upper crustal extension is measured while subsidence analysis from the same line yields 11.1 km of middle and lower crustal extension (Table 1).
 The upper crust versus lower crust discrepancy is less in the Goodenough basin than the Trobriand basin possibly because the serpentinized base of the PUB [Little et al., 2007] at the OSFZ is weaker here than a crustal detachment in the Trobriand basin. Along-strike changes in the active rate of slip on the OSFZ may also play a role in the observed discrepancies between extension amounts observed from fault versus subsidence analysis (Table 1). MCS line 1190 has little extensional discrepancy between the fault (9.1 km) and subsidence (9.2 km) method, while MCS line 1193 to the east does show a discrepancy between the fault (5.9 km) and subsidence (11.1 km) method.
 Studies of Holocene footwall uplift on the Papuan Peninsula [Mann and Taylor, 2004] show that uplift rates are highest in the southwestern end of Goodenough Bay reaching 3 mm/yr and steadily decrease to the east. This variation in Holocene uplift rates indicates the majority of slip on the Owen-Stanley fault is happening at the southwest corner of the basin where MCS line 1190 is located (Figure 2). This focusing of upper crustal extension on the Owen-Stanley fault also supports the interpretation by Wallace et al.  that the OSFZ is acting as the present-day Australia-Woodlark plate boundary in this area of eastern PNG (Figure 2).
7.2 Causes of Discrepancy in Extension Estimates
 The measurements of crustal extension shown in Table 1 show depth-dependent crustal extension is occurring beneath the Trobriand basin. The extension discrepancy in the Trobriand basin cannot be explained by underestimates of upper crustal extension due to misinterpretation of complex polyphase fault geometries [Reston, 2009] because the seismic lines establish that there are very few normal faults present since extension began at 8.4 Ma. Additionally, subseismic resolution faulting is unlikely to make up for this extension discrepancy. Because upper crustal extension measurements are so low, subseismic faulting resulted in an additional 35%–50% of extension in addition to visible extension due to faulting [Yielding et al., 1996], the extension discrepancy in the Trobriand basin would still be substantial. Moreover, extension estimates based on subsidence values most likely underestimate the true lower crustal extension values, and therefore, the discrepancy is likely higher to be higher.
 Because of the significant subsidence (1–2.5 km) in the absence of upper crustal brittle faulting, the late Miocene to recent history of the Trobriand basin is better characterized as a broad sag basin as opposed to a rift basin. This is unusual because sag basins typically form due to post-rift thermal subsidence [Reston, 2009]; however, backstripping and subsidence analysis of the Goodenough #1 well shows the Trobriand basin has not been subject to much thermal subsidence due to Miocene fore-arc extension. The Trobriand basin is clearly in an extensional environment as shown by Euler pole solutions [Taylor et al., 1999], plate motion vectors from GPS data [Wallace et al., 2004] and its position ahead of the westward-propagating Woodlark seafloor spreading system, hence thermal cooling of the lithosphere is not causing subsidence in the Trobriand basin.
 For depth-dependent crustal extension of the crust to occur, the rigid upper crust must be decoupled from the lower crust and the mantle lithosphere. The two-layer, density inverted crustal structure beneath the Trobriand and Goodenough basins provides an ideal medium for this to occur. The weak, serpentinized base of the upper crust Papuan ultramafic body likely acts as a midcrustal detachment in this extensional environment [Little et al., 2007]. Furthermore, Paleogene crustal thickening during the Papuan Orogen to a depth of at least 50 km (Finlayson et al., 1976] likely heated and weakened the felsic lower crust, preconditioning it to flow at the onset of late Miocene extension.
 For the lower crust to thin without deforming the upper crust above it, a mechanism must exist to balance the preferential extension in the lower crust and mantle lithosphere. Kington and Goodliffe  proposed this is accomplished by a steepening of the angle of subduction of the Trobriand slab that is postulated to have previously underlain the DEI region (Figure 9). Because subduction at the Trobriand trench appears to be very slow or have ceased north of the Trobriand basin, the subducted slab can be expected to sink, pulling the slab to a steeper angle beneath the DEI region [Kington and Goodliffe, 2008]. This rollback of the slab would then produce space for the hot, weak lower crust to flow while the strong, upper crust composed mainly of PUB remains undeformed. The mantle lithosphere weakened by volatiles released from the subducting Trobriand slab can be expected to also be thinned [Kington and Goodliffe, 2008]. Although there is little direct evidence for this model, it does explain how preferential lower crustal thinning in the absence of upper crust deformation has occurred in the Trobriand basin since 8.4 Ma.
7.3 Ductile Flow of the Lower Crust as a Mechanism For Diapiric Intrusion and Uplift of the Goodenough and Fergusson Islands Gneiss Domes
 The discrepancy between upper crustal and lower crustal extension beneath the Trobriand and Goodenough basins suggests the lower crust here has thinned by flowing toward the Goodenough and Fergusson Islands area (Figure 8). The lack of significant upper crustal extension by brittle normal faults in the adjacent Trobriand and Goodenough basins suggests that a narrow, local zone of upper crust thinning focused the upward penetration of buoyant, post-orogenic lower crust in a manner similar to that proposed by Martinez et al. . Thermomechanical modeling of extensional gneiss domes has shown that highly stretched, local grabens in the upper crust can initiate the upward flow of lower crust [Tirel et al., 2004]. Localized necking of upper crust ahead of DEI dome emplacement, along with infilling of lower crustal flow in an east-west orientation was argued by Little et al.  and modeled in 3-D by Ellis et al. .
 Beneath the DEI, the Moho is elevated by 10–15 km compared to the Papuan Peninsula and northern fore-arc high area, showing thinning is not confined to the crust and the elevated DEI are isostatically compensated in part by low-density mantle [Abers et al., 2002]. Traditionally, lower crustal flow is assumed to support the high elevation of MCCs by creating a smooth Moho topography beneath the domes, as is the case in the Basin and Range province in the western United States [Klemperer et al., 1986]. However, crustal flow is not always successful in removing thickness contrast, as elevated Moho topographies have been preserved for many millions of years in extensional provinces like the North Sea and the Aegean Sea [McKenzie et al., 2000].
 The persistence of mantle relief beneath the DEI has been explained as a result of an inferred underlying Trobriand subducted slab that has released volatiles and heat that have weakened upper mantle and crust underlying the DEI [Martinez et al., 2001]. Another explanation may be variations in mantle and lower crust viscosities and lower crust channel thickness requires a longer time scale for the lower crust to smooth Moho topography [McKenzie et al., 2000; Abers et al., 2002]. Thermomechanical modeling of extensional gneiss domes by Tirel et al. , however, shows the rate of boundary displacement has a stronger influence on Moho response than temperature, with the amount of Moho arching beneath gneiss domes increasing with extension rates.
 Extension of continental lithosphere in eastern Papua New Guinea is primarily controlled by a “centerline” of rheologically weak continental crust [Taylor et al., 1999]. The crustal heterogeneity defining the weak “centerline” coincides with a collisional suture zone in the lithosphere that marks the site of a Paleogene collision and ophiolite obduction event marked by the OSFZ [Little et al., 2007] (Figure 9). The gently north dipping planar base of the Papuan ultramafic body has been reactivated since Neogene by north-south extension related to the opening between the Australian plate and the Woodlark microplate [Little et al., 2007; Webb et al., 2008] (Figure 9). Extensional reactivation, beginning in the late Miocene (~8.4 Ma) of the regionally widespread, gently north dipping serpentinized based of the PUB at the Owen-Stanley fault zone therefore controls the main rift in the Goodenough basin. The same process of extensional reactivation of this surface is described along different segments of the OSFZ ranging from the Dayman dome in the east [Dazcko et al., 2011, Spencer, 2010], the Prevost Range MCC on Normanby Island [Little et al., 2007], and the Moresby seamount in the east [Goodliffe et al., 1999].
 Subsidence analysis predicting crustal extension values supports the claim that the dominant mechanism for exhumation of UP and UHP metamorphic rocks in the Goodenough and Fergusson Islands gneiss domes occurred by a process of diapiric upwelling due to lower crustal flow and convergence upon a restricted area of upper crust thinning (Figure 9), a key conclusion of Little et al. . The localized zone of uplift in the DEI area is possibly defined by the preexisting Trobriand arc or Miocene faults in the fore-arc basin of the Trobriand subduction system reactivated during extension. A minimum of 10 km of north-south extension along the low-angle (~30°–35°) Owen-Stanley fault zone south of Goodenough Bay has also led to exhumation low-grade metamorphic and unmetamorphosed rocks in the footwall block of this fault on the Papua Peninsula and blueschist grade metamorphic rocks in the Dayman dome area ~35 km to the west [Daczko et al., 2009]. The DEI and Papuan Peninsula, which are separated by only 80 km, therefore show two distinct styles of structure and depths of exhumation of lower crustal material.