Transient fluvial incision and active surface uplift in the Woodlark Rift of eastern Papua New Guinea

The Woodlark Rift of southeastern Papua New Guinea (Fig. 1A) forms part of the boundary between the Australian and Woodlark plates and is rapidly extending (20–40 mm/yr) (Abers, 2001; Taylor et al., 1999; Tregoning et al., 1998; Wallace et al., 2004), with extension accommodated in part by seismogenic low-angle normal faults (Abers et al., 1997). Metamorphic core complexes occur within this zone of extension (Daczko et al., 2009, 2011; Davies and Warren, 1988; Hill, 1994; Hill et al., 1992; Little et al., 2007; Ollier and Pain, 1980; Speckbacher et al., 2011; Spencer, 2010; Spencer and Ohara, 2008), where they form domal islands (D'Entrecasteaux Islands) in the center of the rift and domes on the southern margin of the rift (the Suckling-Dayman massif on the Papuan Peninsula) (Figs. 1 and 2). The domes are topographically prominent, reaching elevations as high as 3.7 km above sea level (asl). Seismic and gravity data indicate that the crust beneath the D'Entrecasteaux Islands is presently relatively thin (20–30 km) and that the metamorphic core complexes are isostatically compensated by low-velocity, buoyant mantle (Abers et al., 2002).

In the D'Entrecasteaux Islands, some metamorphic core complexes host the youngest known high- and ultrahigh-pressure rocks, including coesite eclogite exhumed from depths ≥90 km since 8 Ma at plate-tectonic rates (Baldwin et al., 2004, 2008). There, buoyancy-driven exhumation has been recently proposed to play a key role in exhuming high- and ultrahigh-pressure rocks as crustal diapirs from mantle depths to the surface, forming gneiss domes that were subsequently exhumed by normal faulting in the upper crust (Little et al., 2011; Martinez et al., 2001). Other metamorphic core complexes on the Papuan Peninsula and in the Prevost Range of Normanby Island expose high-pressure schists, and it is unknown whether, or to what extent, buoyancy forces have impacted their geologic evolution. Since metamorphic core complexes can be viewed as a subset of extensional gneiss domes (Yin, 2004), we use in this paper the term metamorphic core complex to describe the extensional gneiss domes of the Woodlark Rift.

Despite rapid exhumation, high topography, and active extension in the Woodlark Rift, it is debated whether all metamorphic core complexes are actively rising, or whether footwall uplift and crustal extension have migrated from the D'Entrecasteaux Islands to the Papuan Peninsula (Mann et al., 2009; Ollier and Pain, 1980). Clarification of the region's landscape evolution will be crucial in order to bridge the gap between geophysical and geodetic observations (Abers et al., 2002, 1997; Tregoning et al., 1998; Wallace et al., 2004) and longer-term structural (Daczko et al., 2009; Hill, 1994; Little et al., 2007, 2011) and thermochronological data (Baldwin et al., 1993, 2004, 2008; Fitzgerald et al., 2008).

The D'Entrecasteaux Islands emerged above sea level during the Pliocene–Pleistocene (Baldwin and Ireland, 1995; Tjhin, 1976), while the Papuan Peninsula has been above sea level since the Miocene (Smith, 1970). Younger constraints on the Woodlark Rift's topographic evolution are relatively sparse or imprecise (Table 1). For example, apatite fission-track ages from the D'Entrecasteaux Islands are as young as 0.4–1.2 Ma (Baldwin et al., 1993; Fitzgerald et al., 2008) and provide information on rock exhumation from shallow crustal levels, but they lack precision due to low spontaneous track densities. Apart from structural observations (Daczko et al., 2011; Little et al., 2007, 2011), most data relevant to Quaternary tectonics on the D'Entrecasteaux Islands and Papuan Peninsula come from landforms and sedimentary deposits. On both the D'Entrecasteaux Islands and Owen Stanley Range, faceted mountain fronts and extensive corrugated dip slopes coincide with fault-related lithologies (Fig. 2) and have been cited as evidence for recent fault slip (Hill et al., 1992; Little et al., 2007; Ollier and Pain, 1980; Spencer, 2010; Spencer and Ohara, 2008). However, facets may form by nontectonic processes such as landslides (Ellis et al., 1999). Moreover, some dip slopes in the study area are parallel to shear-zone foliation, and dip slightly less than range-bounding faults, and may therefore be partly structurally controlled (Little et al., 2011). While qualitative patterns in drainage density and valley beheading on dip slopes on the Suckling-Dayman massif are clear evidence for active fault slip, rock uplift, and rolling-hinge flexure (Ollier and Pain, 1980; Spencer, 2010; Spencer and Ohara, 2008), such observations do not quantify deformation and are not widespread.

Raised Pliocene–Pleistocene deltaic deposits and Pliocene–Pleistocene to Holocene coral terraces directly constrain neotectonic vertical motions on parts of the Papuan Peninsula and northeastern Normanby Island (Little et al., 2007; Mann et al., 2009; Mann and Taylor, 2002; Smith, 1970). In contrast, marine sedimentary rocks and Quaternary terraces have not been identified on Fergusson and Goodenough Islands, suggesting that the coastlines on these two islands have been stable or subsiding over the Holocene (Mann and Taylor, 2002). However, it is clear that coral terraces only quantify rock uplift near the coastline, the coastline is not everywhere formed on lower-plate or footwall rocks, and so the absence of terraces is inconclusive. Thus, large gaps remain in our understanding of Quaternary tectonics in subaerial parts of the Woodlark Rift.

In this paper, we present quantitative results of stream profile analyses that define patterns of active erosion and surface uplift in the D'Entrecasteaux Islands and the eastern Papuan Peninsula. Stream profile analyses are widely used to evaluate regional patterns in rock uplift rate (e.g., Cyr et al., 2010; Hodges et al., 2004; Kirby and Whipple, 2001; Kirby et al., 2003; Snyder et al., 2000; Wobus et al., 2003), drawing from a theoretical basis (e.g., Whipple and Tucker, 1999) and empirical support (e.g., Cyr et al., 2010; DiBiase et al., 2009; Harkins et al., 2007; Ouimet et al., 2009; Safran et al., 2005; Snyder et al., 2000) for a relationship between steady-state stream profile form and erosion rate over 10–100 k.y. time scales. Non-steady-state stream profiles may record additional information about the spatial and temporal distribution of crustal deformation (e.g., Clark et al., 2005, 2006; Dorsey and Roering, 2006; Kirby et al., 2007; Schoenbohm et al., 2004; Whittaker et al., 2008).

We show that the D'Entrecasteaux Islands and eastern Papuan Peninsula are undergoing transient erosion resulting from a regional increase in rock uplift rate, probably in the late Quaternary, and as a consequence they are undergoing surface uplift. We then document quantitative spatial patterns in transient incision and stream channel slopes, showing that these are consistent with geologic evidence for patterns of rock uplift. Finally, we assess whether the D'Entrecasteaux Islands are still undergoing rock uplift (on the 10–100 k.y. time scale) versus the alternative hypothesis that rock uplift has ceased on the D'Entrecasteaux Islands and has shifted south to the Papuan Peninsula. Results are used to examine whether geophysically imaged mantle structures beneath the rift (Abers et al., 2002) correspond to currently rising, stable, or sinking topography.

The Woodlark Rift is an east-west–trending, ∼120-km-wide zone of active normal faulting, seismicity, and volcanism, straddling a complex plate boundary ahead of the westward-propagating Woodlark basin spreading center (Fig. 1). Northward subduction of the Australian passive continental margin beneath an island arc and oceanic lithosphere occurred in the Late Cretaceous–Paleocene, as indicated by the southward-obducted Papuan ultramafic belt (Davies and Jaques, 1984; Lus et al., 2004). Subducted sediments and basalts experienced peak metamorphism at ca. 8–7 Ma to form high- and ultrahigh-pressure rocks that have since been exhumed at rates of ∼2 cm/yr (Baldwin et al., 2008; Monteleone et al., 2007; Zirakparvar et al., 2011). By the late Miocene, northward subduction of the Woodlark microplate at the New Britain and San Cristobal Trenches (Fig. 1) initiated rifting in the Woodlark basin (Weissel et al., 1982). Seafloor spreading in the Woodlark basin followed rifting and has propagated westward in a series of jumps, with the spreading tip currently located just east of Moresby Seamount, a metamorphic core complex ∼25 km northeast of Normanby Island (Speckbacher et al., 2011; Taylor et al., 1999).

Five metamorphic core complexes, spanning a distance of 265 km, have been identified above sea level in the Woodlark Rift west of the Woodlark basin spreading center (Daczko et al., 2009; Davies and Warren, 1988; Hill et al., 1992; Little et al., 2007). The metamorphic core complexes occur in two approximately parallel NW-SE–trending bands along the axis and southern flank of the rift (Fig. 3). Four metamorphic core complexes make up the prominent topographic domes on the D'Entrecasteaux Islands and rise broadly in elevation from east to west: the Prevost Range (1200 m asl) on Normanby Island, Oiatabu dome (1864m asl) and Mailolo dome (1665 m asl) on Fergusson Island, and Goodenough dome (2536 m) on Goodenough Island. Lower-plate rocks also crop out on southern Fergusson Island (Morima massif) and northwestern Normanby Island. The fifth metamorphic core complex is the Suckling-Dayman massif in the Owen Stanley Range, which includes the prominent dome of Mount Dayman (a.k.a. Dayman dome, 2896 m asl) and the dissected peak of Mount Suckling (3676 m asl).

Structural styles and metamorphic conditions in lower-plate rocks vary. Lower-plate rocks on the Papuan Peninsula and the Prevost Range consist of schist (Daczko et al., 2009; Little et al., 2007), and their exhumation and domal shape are attributed to low-angle normal faulting and flexural uplift (Spencer, 2010). Lower-plate rocks on Goodenough Island, Fergusson Island, and northwestern Normanby Island consist of eclogite-bearing migmatitic gneisses recording high- and ultrahigh-pressure metamorphism (Baldwin et al., 2004, 2008; Davies and Warren, 1988); their exhumation, domal shapes, and internal deformation may be the result of a more vertical trajectory and diapiric flow (Little et al., 2011). Except in the Prevost Range, lower-plate rocks are intruded by synextensional granitoids formed during decompression melting at ca. 4–2 Ma (Baldwin and Ireland, 1995; Baldwin et al., 1993; Daczko et al., 2009; Monteleone, 2006). Upper-plate rocks consist of ophiolite (Papuan ultramafic belt) and Miocene–Pliocene terrigenous sedimentary rocks overlain by younger marine and alluvial sediments and volcanics (Davies and Warren, 1988).

Since ca. 8 Ma, the D'Entrecasteaux lower-plate rocks have been exhumed at rates of ∼10 km/m.y. until at least 0.4 Ma (Baldwin et al., 1993, 2004, 2008). By the time the D'Entrecasteaux Islands emerged above sea level in the Pliocene–Pleistocene and shed gneissic clasts into the Trobriand basin, seismic-reflection data indicate that distributed normal faulting in the Trobriand basin north of the islands had ceased (Stewart et al., 1986; Tjhin, 1976), while faulting continued in the Goodenough basin to the south (Fang, 2000). Presently, subsidence continues in both basins, though faster in Goodenough basin (Fang, 2000).

Based on Euler pole rotations, average extension rates at the longitude of Moresby Seamount have been 37 mm/yr since ca. 2.1 Ma (Taylor and Huchon, 2002). Global positioning system (GPS) data show that extension at the longitude of Moresby Seamount is currently ∼20 mm/yr (Tregoning et al., 1998; Wallace et al., 2004). To the west, relative plate motion is accommodated by the 1400-km-long transtensional Owen Stanley fault zone, which forms the northern margin of the Owen Stanley Range (Wallace et al., 2004).

Topographically, metamorphic core complexes in the Woodlark Rift are distinguished by their domal morphology, weakly dissected triangular facets or dip slopes (similar to turtlebacks), and commonly parallel drainage (Fig. 2). Facets and dip slopes coincide with faults and shear zones (Little et al., 2007; Ollier and Pain, 1980; Spencer, 2010; Spencer and Ohara, 2008). Parallel drainage patterns tend to align with the upper-plate motion, and, in places, streams occupy fault corrugations (Daczko et al., 2011; Little et al., 2007; Spencer and Ohara, 2008). Detachment faults are observed where upper-plate rocks are in contact with the lower plate. Many fault surfaces are eroded, leaving underlying mylonitic shear zones, 500–1000 m thick, that generally dip radially outward 0° to 45° from dome centers (Daczko et al., 2009; Davies and Warren, 1988; Hill, 1994; Hill et al., 1992; Little et al., 2007, 2011). Mylonites record dominantly top-to-the-north shear (Daczko et al., 2009; Hill, 1994; Little et al., 2007). In general, mylonitic shear zones reflecting the final stages of deep-seated exhumation are consistent with offshore seismic-reflection evidence and earthquake focal mechanisms for north-dipping, low-angle, normal faults dipping 25°–35°N (Abers, 1991; Abers et al., 1997) (Fig. 3). Dome-bounding faults have been mapped in the field (Davies and Warren, 1988; Little et al., 2007, 2011). On the northern flanks of Goodenough and Fergusson Islands, dome-bounding faults dipping 30°–40°N offset Holocene sediments (Little et al., 2011). Other faults are not exposed onshore but are inferred on the basis of bathymetry and seismicity (Davies and Warren, 1988; Little et al., 2007).

The present-day regional climate is tropical-rainy (Köppen classification), with a mean annual rainfall 2–4 m/yr based on historical rain gauge measurements (McAlpine, 1983) and satellite-derived rainfall estimates for 1997–2006 from the Tropical Rainfall Measuring Mission (TRMM) 2B31 data set (Mulligan, 2006). There is no evidence that any peaks in the region were glaciated in the Pleistocene (Löffler, 1970, 1977). The lowest elevation evidence for glaciation in the Owen Stanley Range is ∼225 km west of Mount Suckling at 3400–3450 m, higher than anywhere in the study area with the exception of Mount Suckling at 3676 m asl.

It is crucial to distinguish steady state from transient landscapes, in part to distinguish between streamwise variations in steepness index that arise from: (1) steady-state erosion in equilibrium with nonuniform rock uplift rates, and (2) transient erosion amid uniform rock uplift rates (Whittaker et al., 2008). Transient erosion occurs as streams adjust to changes in rock uplift rate, base level, or climate, partly by a change in channel slope that transmits headward through the channel network. In detachment-limited stream channels, such adjustments are exhibited on slope-area plots as distinct breaks in power-law scaling and on longitudinal profiles as distinct convexities, or knickpoints (Whipple, 2004). Knickpoints thus separate areas of the landscape where streams and hillslopes have responded to new boundary conditions from areas where streams and hillslopes have not responded but may be adjusted to earlier boundary conditions (Crosby and Whipple, 2006). In the wake of migrating knickpoints, steepness indices measured below knickpoints correlate in a positive fashion with rock uplift rate (Kirby et al., 2007), so long as those reaches are not oversteepened, where sediment-dependent erosion declines (Crosby et al., 2007; Gasparini et al., 2006). Oversteepened reaches tend to be very concave (θ > 1) (Schoenbohm et al., 2004) and do not adhere to the functional relationship in Equation 2. Even where oversteepening is present, transient landscapes are useful because it is possible to quantify incision, yielding information about magnitudes of erosional responses to base-level changes and rock uplift (e.g., Clark et al., 2005; Harkins et al., 2007). Such methods have also been used to establish patterns of fault throw on individual faults (e.g., Schoenbohm et al., 2004; Whittaker et al., 2008).

Longitudinal profiles were extracted from a hole-filled 3 arc-s (∼90 m) Shuttle Radar Topography Mission (SRTM) digital elevation model (DEM) (Jarvis et al., 2008). Holes (areas of no data) in original SRTM coverage do not occur in most valley bottoms (see GSA Data Repository Fig. DR1¹), but where they do, the streams were either not analyzed, or the holes are sufficiently small that they did not affect the analysis. DEM data were processed and analyzed using established quantitative techniques (Wobus et al., 2006). Channels were selected with drainage areas >10⁵ m² (cf. Kirby et al., 2007; Kirby and Whipple, 2001), although we acknowledge that nonfluvial processes such as debris flows may periodically erode channels in mountain settings with drainage areas up to ∼10⁶ m² (Stock and Dietrich, 2003). To reduce noise and artifacts, we removed spikes in longitudinal profile elevations and then smoothed elevations with a 1000 m streamwise moving average filter, retaining information at long (>1 km) length scales. At every 20 m elevation interval, we calculated channel slope and upstream drainage area. Steepness (k_s) and concavity (θ) indices were estimated from ordinary least-squares regression of the logarithms of channel slope and area. For comparison among streams and to circumvent the dependency of k_s on θ, we estimated a normalized steepness index, k_sn, fitted to the slope and area data using a fixed reference concavity, θ_ref (e.g., Wobus et al., 2006). Typically, θ_ref is assigned the study area's mean value (θ_average = 0.49 ± 0.02), but results are not very sensitive to choice of θ_ref (Schoenbohm et al., 2004). For consistency with other studies, we used θ_ref = 0.45 (e.g., Cyr et al., 2010; DiBiase et al., 2009; Ouimet et al., 2009; Schoenbohm et al., 2004; Snyder et al., 2000; Wobus et al., 2003). To illustrate regional patterns, we estimated k_sn in a 0.5 km moving window along trunk streams and tributaries selected at DEM cells with A_j > 5 km² and A_c > 0.5 km², where A_j is the contributing area above tributary junctions, and A_c is the contributing area at the channel head. This automated analysis was performed regardless of profile shape or likely bed material.

In order to recognize where bedrock channels may be undergoing transient erosion, as opposed to steady-state erosion, we also analyzed individual profiles and slope-area plots from streams distributed across footwall (or lower-plate) and hanging-wall (or upper-plate) rocks in the D'Entrecasteaux Islands and Owen Stanley Range. While most of the streambeds upstream of the mountain fronts are bedrock or mixed bedrock-alluvial, a few apparently alluvial reaches occur at high elevations in the Owen Stanley Range, as suggested by Landsat images and aerial photographs, and these reaches were excluded from the analysis. We also excluded reaches flowing across Quaternary alluvium downstream of mountain fronts.

Finally, knickpoints were located on longitudinal profiles and slope-area plots, and mapped at the DEM cell immediately above a downstream increase in slope. To assess the role, if any, of lithology or structure on profile parameters and knickpoint locations, results were compared against digitized geologic maps (Bain et al., 1972; Davies, 1972; Davies and Smith, 1974; Hill, 1991; Little et al., 2007).

Where knickpoints were identified and judged to not be structurally or lithologically controlled, but possibly migratory, we estimated incision by reconstructing the stream profile above the knickpoint in the downstream direction toward the coastline or mountain front, and subtracting the modern profile elevation at the mountain front. Profiles were reconstructed by extrapolating the profile in the downstream direction using their normalized steepness index, the reference concavity (θ_ref = 0.45), and modern flow paths (cf. Berlin and Anderson, 2007; Clark et al., 2005; Harkins et al., 2007; Hoke et al., 2007; Schoenbohm et al., 2004). Reaches chosen for reconstruction have concavity indices similar to the regional average and steepness indices that are locally representative. Errors in reconstructed profile elevations were estimated by propagating errors in normalized steepness index and uncertainties in measured knickpoint elevations. Knickpoint elevation uncertainties were not directly estimated; rather we assigned an error of ±30 m, as determined in other studies using the same resolution DEM in similar high-relief settings (e.g., Harkins et al., 2007).

From the study area, 171 stream profiles were analyzed individually (GSA Data Repository Table DR1; Fig. DR1 [see ^{footnote 1}]). The simplest characterization of stream profiles is based on the presence or absence of knickpoints. In total, 140 knickpoints were identified (GSA Data Repository Table DR2 [see ^{footnote 1}]). The majority (77%) of streams on metamorphic core complexes and fault blocks in the study area have between one and three prominent knickpoints, whereas a minority (23%) of streams are fully concave, with no knickpoints.

Fully concave stream profiles (Fig. 4) occur mostly in upper-plate watersheds of northwestern Normanby Island, on tectonically stable or subsiding Pleistocene volcanoes such as Mt. Trafalgar, and in isolated occurrences in lower-plate or footwall watersheds (e.g., generally the southern sides of Goodenough dome, Mailolo dome, Prevost Range, and Owen Stanley Range). Fully concave streams exhibit slope-area scaling at drainage areas generally larger than 2 × 10⁵–4 × 10⁵ m² (Fig. 4). In these streams, mean θ is 0.49 ± 0.02, with a range of 0.38–0.66 (GSA Data Repository Table DR3 [see ^{footnote 1}]), similar to values expected for equilibrium profiles (Whipple and Tucker, 1999). In fully concave streams, k_sn ranges over an order of magnitude from 13 m^0.9 to 261 m^0.9, with the lowest values occurring on Mt. Trafalgar volcano, northwestern Normanby Island, and the easternmost Owen Stanley Range. This is about half of the range observed across the entire study area in all streams (10–599 m^0.9 in individually analyzed streams; 2 to ∼800 m^0.9 in automatically analyzed streams).

In contrast, streams with knickpoints are nearly ubiquitous across the region. They occur on all of the D'Entrecasteaux Islands and across most of the eastern Owen Stanley Range, regardless of lithologies and structures. The observed knickpoints share a number of characteristics, many of which are visible in the representative profiles (Fig. 5).

First, stream gradients steepen across every knickpoint in the downstream direction by a factor of ∼2–3. This steepening contributes to most of the variability in k_sn observed across the study area.

Second, most knickpoints do not overlie mapped lithologic contacts or faults but are commonly upstream of mapped or inferred faults, as seen on long profiles, suggesting that knickpoints did not form because of spatial contrasts in erodibility. In fact, there is no clear relationship between steepness index and rock type, with low and high values occurring in gneiss, weakly deformed granodiorite, and schist.

Third, drainage area upstream of knickpoints spans nearly three orders of magnitude (0.2–100 km²). The relationship between drainage area and streamwise distance from the drainage divide can be fitted by a power law with an exponent of 0.56 (Fig. 6). Because knickpoints have such large contributing areas (generally >1 km²), that furthermore show scaling behavior following Hack's law for stream networks (Dodds and Rothman, 2000), it is clear that knickpoints lie within channel networks rather than at the hillslope-channel transition (Crosby and Whipple, 2006; Harkins et al., 2007).

Fourth, knickpoints range in elevation across the study area from 59 to 3083 m, but their elevations are spatially autocorrelated at length scales <10 km: That is, knickpoints in adjacent watersheds tend to occur at similar elevations, varying less than a few 100 m. Examples of this from three parts of the study area are shown in Figure 7. Most streams have a single large knickpoint. Streams with multiple knickpoints include either a knickpoint that lies in a higher or lower elevation band (relative to the most prominent knickpoint), or both. These higher and lower knickpoints are less common, respectively constituting 26% and 14% of all knickpoints mapped.

Finally, many knickpoints occur at the upstream ends of inner gorges. Inner gorges are characterized by cross-valley slopes that are steeper than cross-valley slopes measured above the knickpoints, indicating the landscape below knickpoints is more deeply incised (Fig. 8). Some inner gorges transition to gentler hillslopes in the upslope direction. On the Mai'iu River tributary on Mt. Suckling, this transition to gentler hillslopes is associated with terrace-like surfaces observed in DEMs and satellite images. These surfaces can also be traced ∼5 km downstream of the knickpoint, from which they vary in elevation by <100 m (Fig. 5B). The corollary to the observation that steep hillslopes adjoin steep channels in deep gorges below knickpoints is that topography above knickpoints tends to be gentler and have lower local relief.

The observation that knickpoints occur in relatively narrow elevation bands is consistent with field studies (Berlin and Anderson, 2007; Crosby and Whipple, 2006; Harkins et al., 2007; Hoke et al., 2007) and model predictions of transient detachment-limited stream profile evolution (Niemann et al., 2001). Such knickpoints are migratory rather than static, and they typically migrate vertically at the same rate where rock uplift rates and stream erosion processes are uniform. Overall, the presence of knickpoints in a large range of drainage areas within the fluvial network, the lack of evidence for structural or lithologic control, the clustering of knickpoints within narrow elevation bands, and the occurrence of knickpoints above inner gorges is most simply explained by transient fluvial incision. Importantly, transient incision appears to be a regional phenomenon.

Transient erosion presents a special challenge for quantitative analysis (Whittaker et al., 2008). As opposed to equilibrium stream profiles, transient stream profiles are actively evolving, with k_sn and θ varying as a function of time and distance along streams, in addition to responding to external forcing factors (e.g., climate and tectonics). In transient streams, k_sn and θ can only be meaningfully compared between stream reaches that formed concurrently (Whittaker et al., 2008) and that show evidence for adjustment to former or modern boundary conditions (e.g., through slope-area scaling) (Kirby et al., 2007). In this section, we establish a correlation among knickpoints and stream segments in different drainage basins that permits comparison of normalized steepness index and amounts of transient incision across the study area.

Transient knickpoints and the reaches above/below them can be compared with knickpoints and reaches in adjacent watersheds (e.g., Schoenbohm et al., 2004; Whittaker et al., 2008). Correlative parts of stream profiles are assumed to have formed at roughly the same time, though the exact timing may vary slightly, and correlative knickpoints are expected to occur at similar elevations except for systematic variations due to differential rock uplift.

The analyzed streams have as many as three knickpoints, but only one of these knickpoints occurs locally within a narrow elevation range across neighboring watersheds, and can be traced along the entire length of the Owen Stanley Range east of the Suckling-Dayman massif as well as around each of the D'Entrecasteaux Islands. This is the lowest elevation knickpoint in 86% of streams. These knickpoints are the benchmark of our correlation, and we designate them as K₁ or stage-one knickpoints (Figs. 9–11). While there is no direct evidence that K₁ knickpoints were initiated at the same time, particularly when considering different mountain range segments, fault segments, metamorphic core complex domes, or islands where tectonic histories may differ, this correlation scheme is the most straightforward given the available data. Any temporal correlation is most likely to be valid within these individual areas, and although it becomes more tentative with increasing distance, the traceable progression of K₁ knickpoints from dome to dome, island to island, and islands to peninsula supports this correlation.

In the process of correlating K₁ knickpoints, we also correlated higher (K₀) and lower (K₂) knickpoints. Like K₁ knickpoints, these occur over locally restricted elevation ranges (Figs. 9–11). K₀ knickpoints are rare, most likely because they have propagated through all but the highest-elevation watersheds; K₂ knickpoints are rare, likely because they are the result of recent, localized throw on faults bordering a limited number of drainage basins.

For clarity, knickpoints and stream reaches below knickpoints are assigned to the same stage. Thus, the stream reaches immediately below the ubiquitous K₁ knickpoints are designated S₁, and so on. Fully concave streams (i.e., those without knickpoints) are assumed to correlate to S₁. This is consistent with the observation that fully concave streams have similar steepness indices to S₁ reaches in adjacent watersheds (Figs. 9–11), and do not occur adjacent to watersheds with S₂ reaches. Like knickpoint elevations, k_sn in S₁ and S₀ reaches tends to vary little (<100 m^0.9) over short length scales (<10 km), though there is more scatter in the Owen Stanley Range than in the D'Entrecasteaux Islands. Concavity indices vary mostly as a function of correlation stage, similar to what others have noted elsewhere (Schoenbohm et al., 2004). Throughout the study area, S₀ is commonly concave in profile, with median θ = 0.48 (see GSA Data Repository Table DR4 [^{footnote 1}]; Fig. 5). S₁ profiles tend to be more variable in concavity, with some streams on Suckling-Dayman massif being convex (θ < 1), but across the region, these S₁ profiles have a median θ = 0.68, which is within the range observed for adjusted streams.

Transient incision below the relict landscape was estimated in 59 streams by reconstructing the former S₀ profiles (i.e., prior to S₁ development) downstream of K₁ knickpoints to the coastline, and then subtracting the modern profile elevations at the mountain front, as depicted in Figure 5. Estimates of S₁ incision (i.e., that incision below K₁ knickpoints) range across the study area from 40 to 1270 m (GSA Data Repository Table DR1 [see ^{footnote 1}]). On a local scale (<10 km), incision estimates are similar to one another. On the larger scale of individual domes or islands, incision varies with K₁ knickpoint elevation (Figs. 9B–11B). This relationship is also evident at the regional scale in a comparison of spatially averaged values (Fig. 12; GSA Data Repository Table DR5 [see ^{footnote 1}]). Regionally, the least average incision is on Normanby Island (∼170 m) and the greatest is on Suckling-Dayman massif, in the saddle between Mount Suckling and Mount Dayman (∼780 m).

Plotted together, incision and k_sn of S₁ reaches show a positive and statistically significant correlation in both the D'Entrecasteaux Islands (Fig. 13A) and the Owen Stanley Range (Fig. 13B). Visible positive relationships also exist for individual domes and mountain range segments. Even though there may be small differences in the exact trends defined by individual domes, there is large overlap in trends at the regional scale such that it is difficult to identify a systematic difference between the Papuan Peninsula and the D'Entrecasteaux Islands (Fig. 13C). Considering empirical studies that show a functional association between k_sn and erosion rate (e.g., Ouimet et al., 2009), a relationship between k_sn and erosion magnitude would be expected if K₁ knickpoints were initiated at approximately the same time and erosion rates in S₁ streams were mostly equilibrated with rock uplift rates. It is therefore likely that k_sn in transient S₁ reaches records information about relative rates of erosion as well as magnitudes of transient erosion. The large overlap in trends at the regional scale also implies that erosional efficiency (coefficient K in Eq. 2) is approximately spatially uniform.

The estimates of incision given here run counter to the casual observation that dip slopes appear weakly incised, particularly on the Suckling-Dayman massif. However, the lack of apparent incision in some cases can be explained as a consequence of the trajectory of bedrock motion relative to the fault trace. Where mountain fronts are bounded by low-angle normal faults and bedrock motion has a significant nonvertical component (e.g., Dayman dome), the apparent incision at the base of the exposed fault surface integrates erosion over a shorter amount of time, and is therefore less than the true vertical component of incision downstream of the knickpoint (for a graphical explanation, see GSA Data Repository Fig. DR2 [^{footnote 1}]). Incision estimated from projected profiles may also underestimate the total vertical component of incision, but not as much (see GSA Data Repository for a derivation of this error [^{footnote 1}]). Given that most faults in the study area dip <25° and projected streams slope <8 × 10⁻², this error is less than 20%. Even acknowledging that continued erosion above knickpoints also reduces apparent values of incision over time, results in the study area appear to underestimate vertical incision and rock uplift by only a small amount. Supporting this, projected S₀ reaches are only slightly lower (<100 m) than apparent fluvial terraces on Mount Suckling and raised delta topsets along Goodenough Bay (Figs. 5B and 5D). Thus, we conclude that any bias arising from erosion above knickpoints and horizontal bedrock motion is small overall.

The widespread occurrences of relatively gentle topography above high-elevation knickpoints and incision below them are strong evidence for recent rock uplift. This inference is drawn for the following reasons: (1) reconstructed stream profiles project above modern base level at the mountain front, (2) mountain fronts are faulted and generally close to the coast, and (3) present-day sea level is close to its highest elevation for the past 3 m.y. (Miller et al., 2005). These same points are made graphically in swath profiles across metamorphic core complexes (Fig. 14) that capture the general form of stream profiles in their minimum elevations, total relief in their maximum elevations, and present a comparison to reconstructed S₀ profile elevations.

In addition, the inference that stream profiles in the study area are not steady state implies that basinwide erosion rates lag behind rock uplift rates (cf. Kirby et al., 2007). Thus, stream profiles as well as their entire watersheds have increased in mean elevation. Areas upstream of knickpoints represent relict, relatively low-relief landscapes that are undergoing surface uplift. In this fashion, these relict landscapes are analogous to the erosion surfaces on the Papuan Peninsula south of Goodenough Bay, which have presumably risen from lower elevations (Smith, 1970; Smith and Simpson, 1972), but have not yet been erased by erosion. Moreover, the relict stream networks show that transient erosion is widespread across the entire study region rather than just confined to the area around these previously identified erosion surfaces.

While we have not specifically estimated the change in spatially averaged elevations, it is apparent that the relict portions of drainage basins have been elevated by approximately the same amounts as the local transient incision estimated at mountain fronts. Thus, the upper portions of drainage basins have been uplifted as little as ∼170 m on Normanby Island, on average, to as much as ∼780 m in the Suckling-Dayman massif. This surface uplift constructed a significant fraction of the present-day relief. By calculating the percentage change in basin relief using reconstructed S₀ and modern stream profiles (Clark et al., 2005), average basin relief has increased by 40% in the D'Entrecasteaux Islands and 80% in the Owen Stanley Range in the late Quaternary (GSA Data Repository Table DR5 [see ^{footnote 1}]).

The presence of knickpoints and the elevated reconstructed profiles are strong evidence for recent rock uplift and net surface uplift, but it is less obvious to ascertain to which type of perturbation the transient stream profiles are responding. Inner gorges below knickpoints indicate increases in stream erosion that are commonly attributed to either rapid base-level fall (Densmore et al., 1997) or an increase in rock uplift rate (Kirby et al., 2007; Stock et al., 2005). Although a change in climate toward greater precipitation or higher-magnitude storms could increase erosion rates as well (Molnar, 2004), an increase in erosivity would likely drive channel slopes to become systematically gentler rather than steeper (Whipple and Tucker, 1999; Wobus et al., 2010), which is not the behavior observed below knickpoints.

Possible sources of relative base-level fall include eustatic sea-level fall, stream capture, and rock uplift. Numerical models suggest that eustatic sea-level fall may cause transient subaerial erosion if the exposed seafloor (below modern sea level) is steeper than the upstream stream channel (above modern sea level) (Snyder et al., 2002). In this manner, gentle offshore slopes and wide continental shelves may have insulated streams on a number of active margins from transient erosion during eustatic sea level fall (∼120 m) during the Quaternary (e.g., Duvall et al., 2004; Snyder et al., 2002; VanLaningham et al., 2006). In our study area, offshore slopes above the 100 m isobath (Fig. 3) range from >0.1 to <0.002, whereas subaerial slopes along the coast are rarely <0.01. Although some offshore slopes are steeper than onshore slopes and some are gentler, knickpoints form along all coasts in the study area, not only those with steep offshore bathymetry. On land, stream capture may shorten and steepen drainage courses to the coast, thus driving transient erosion in individual watersheds (Prince et al., 2011). However, the widespread and common occurrence of knickpoints suggests that capture is not the primary cause. We conclude that transient erosion in the study area most likely initiated from an increase in rock uplift rate rather than from stream capture or eustatic sea-level change.

Age constraints on the history of landscape evolution are relatively poor, but they confirm that transient erosion is young. On the Papuan Peninsula, the reconstructed S₀ profile of the Uga River (Fig. 5D) intersects the top of the raised Pliocene–Pleistocene Uga paleodelta at ∼500 m above sea level (Smith, 1970), indicating that S₁ incision is likely Pleistocene in age. S₀ reaches farther east are incised <200 m into relict erosion surfaces of inferred late Pliocene to Pleistocene age (Smith, 1970). Together, these observations suggest that S₁ incision is Pleistocene in age and S₀ is Pliocene–Pleistocene, but younger than those erosion surfaces of inferred late Pliocene to Pleistocene age.

On the D'Entrecasteaux Islands, thermochronologic data indicate rapid exhumation (on the order of 10 km/m.y.) from several kilometers depth since 0.4 Ma, depending on the paleogeothermal gradient, which is sure to be high given these high exhumation rates. On Goodenough and Fergusson Islands, apatite fission-track ages in lower-plate rocks collected near sea level are 0.4–1.2 Ma, within error of 0 Ma at the 2σ level (Baldwin et al., 1993; Fitzgerald et al., 2008). Thus, uplift of S₀ profiles on these islands (40–890 m) represents a small fraction of rock uplift and exhumation in the late Pleistocene, indicating that S₁ incision is very young. Additional research is required to more accurately constrain rates of Quaternary erosion and dates for the onset of transient incision.

Transient incision and channel steepening reflect fluvial responses to rock uplift. Next, we consider spatial patterns in the S₁ incision amounts, K₁ knickpoint elevations, and S₁ normalized steepness indices. Patterns over individual domes and mountain ranges reflect patterns in differential rock uplift that can be associated with activity on specific mapped faults and that can be compared to independent geologic and thermochronologic data. Regional-scale patterns may provide information about larger geodynamic processes.

Morima massif (Fig. 9C) is formed in the footwall of the east-west–striking Morima Coast fault, an inferred normal fault on the southern side of Fergusson Island (Davies and Warren, 1988; Little et al., 2011). On a map of k_sn, streams draining south from the range crest are visibly steeper than those draining north. Comparing S₁ streams alone, mean values of k_sn are, in fact, statistically significantly greater south of the range crest (199 ± 51 m^0.9) than to the north (92 ± 31 m^0.9) at the 95% confidence interval (GSA Data Repository Table DR6 [see ^{footnote 1}]). These contrasting stream gradients are probably the response to greater footwall uplift close to the Morima Coast fault. Comparing only south-flowing streams along strike (Figs. 9A and 9B), transient incision, knickpoint elevation, and k_sn all vary in an arcuate pattern, with greatest values in the middle of the range (Fig. 9), implying an arcuate pattern in Quaternary fault throw (cf. Cowie and Roberts, 2001). This pattern is consistent with independent evidence for differential rock uplift. Andesitic tuff currently near sea level at the western tip of the massif was deposited at or above sea level at 0.8 Ma (Fig. 9B; Baldwin et al., 1993). At the same time, lower-plate gneiss currently near sea level in the middle of the range, with apatite fission-track ages of 0.4–0.8 Ma, was still at a depth of several kilometers, implying the center of the massif has experienced more late Quaternary rock uplift than its western tip.

Streams are steeper (but not significantly) on the northeastern flank of Mailolo dome than on the southwestern flank, probably due to ongoing slip on the northeast-dipping Mwadeia fault (Fig. 9C; GSA Data Repository Table DR6 [see ^{footnote 1}]). Compared along strike, incision and k_sn are the greatest in the center of the range, as for Morima massif, although knickpoint elevations do not clearly show this pattern (Figs. 9A and 9B). Similarly on Goodenough dome, streams draining the northeastern flank, near the Wakonai fault, are significantly steeper than on the opposite flank. Compared along strike of the Wakonai fault, northeastward-flowing streams show weak evidence for greater uplift toward the northwest.

The differences in apparent rock uplift on the northeastern and southwestern sides of Goodenough and Mailolo domes are consistent with the sense of tilting inferred from attitudes of foliation formed by diapiric ascent of lower-plate gneiss (Little et al., 2011). Foliation dips more steeply on the southwestern sides of the domes, suggesting tilting of these domes ∼20° down to the southwest due to active slip on Wakonai and Mwadeia faults.

Although uplift may be dominated by faulting on the northeastern flanks of the domes, there is evidence, at least on Goodenough dome, for active faulting on their southern sides as well. For example, in an unnamed stream (number 13) on the southwestern side of Goodenough dome, k_sn sharply and anomalously decreases by a factor of three where it crosses the Fakwakwa fault (Davies and Warren, 1988), even though the hanging wall and footwall are the same core-zone gneiss (Figs. 9C and 10C). Various previous studies have inferred the fault is active (Ollier and Pain, 1980) or inactive (Little et al., 2011). In the present study, the abrupt downstream reduction in k_sn suggests an abrupt downstream reduction in rock uplift rate from down-to-the-southwest throw.

On Oiatabu dome, steepness indices in map view (Fig. 9C) increase toward the north. S₁ streams draining the eastern flank along the Elologea fault are steeper (but not statistically significantly) than those draining its western flank (GSA Data Repository Table DR6 [see ^{footnote 1}]). Young apatite fission-track ages (1.0 ± 0.6 Ma) along the northern coastline are consistent with recent rapid exhumation of this part of the dome (Baldwin et al., 1993). On Oiatabu as well as Goodenough and Mailolo domes, geomorphic evidence for ongoing rock uplift is consistent with range-bounding faults that cut Quaternary alluvium (Little et al., 2011).

On Normanby Island (Fig. 10), stream profiles show evidence for greatest incision and uplift east of the Trobriand fault, particularly in the Prevost Range. Uplifted knickpoints on the northern side of the Prevost Range may indicate displacement along the Prevost detachment or other north-dipping normal faults, but there is greater geomorphic evidence that Quaternary rock uplift increases systematically from west to east across the Prevost Range metamorphic core complex due to faulting east of Normanby Island. The trend in incision and k_sn is consistent with the sense of tilt in a Quaternary coral terrace along the northern coast, which has been attributed to throw on an east-dipping normal fault within the Normanby transfer fault system (Little et al., 2007).

On the Papuan Peninsula, S₁ streams draining the northern side of the Owen Stanley Range are steeper than those on its southern side (GSA Data Repository Table DR6 [see ^{footnote 1}]). Footwall or lower-plate uplift is dominated by throw on the north-dipping Goodenough fault and Mai'iu detachment fault (Mann et al., 2009; Spencer, 2010). Extensive slip on the Mai'iu fault has driven flexural lower-plate uplift and doming, which in turn have driven north-flowing streams to capture south-flowing streams and locally reverse drainage directions (Spencer, 2010).

On the southern side of the Suckling-Dayman massif, there is evidence to suggest down-to-the-south throw on the western Onuam fault (see map in Fig. 11): Streams that flow southward abruptly decrease in slope by a factor of two where those streams cross the fault. There, the fault corresponds with the topographic margin of the dome. Meanwhile, it is not apparent that the fault is active farther east where it crosses the crest of the Owen Stanley Range and is traversed by north-flowing streams.

Overall, the Suckling-Dayman massif shows the greatest evidence for recent rock uplift in the eastern Owen Stanley Range based on patterns in incision and k_sn (Fig. 11). Steepness indices are generally greater for a given amount of incision on the Suckling-Dayman massif than elsewhere, possibly indicating slightly faster, more recent incision and uplift (Fig. 13B). An inferred increase in Quaternary rock uplift toward the west is consistent with a westward increase in Pliocene erosion surface elevations south of Goodenough Bay (Smith, 1970) and with a westward increase in metamorphic grade of the footwall (Daczko et al., 2009; Davies, 1980). Also, the region of maximum inferred rock uplift coincides with the most expansive, least dissected dip slopes on the exhumed Mai'iu detachment fault and the area of stream capture and drainage reversal (Spencer, 2010).

On the regional scale, steepness indices and incision in S₁ streams also vary systematically. Grouped into geographical bins representing individual domes and mountain range segments (demarcated in Figs. 9B, 10B, and 11B), average values increase by a factor of two to three from east to west across the D'Entrecasteaux Islands and the eastern Owen Stanley Range (Fig. 12). Regional patterns are roughly similar to the east-west trend in peak elevations (Fig. 12). In the D'Entrecasteaux Islands, geomorphic evidence for a westward increase in rock uplift is consistent with a general westward decrease in Pliocene–Pleistocene low-temperature thermochronological ages on lower-plate rocks. Although time constraints on incision and uplift are limited, these results suggest similar patterns of regional-scale differential rock uplift on both the Papuan Peninsula and D'Entrecasteaux Islands.

One objective of this study was to investigate the pattern of active subaerial deformation across the Woodlark Rift. Some previous studies suggested that the D'Entrecasteaux Islands are currently stable or subsiding, and thus the focus of extension and footwall uplift has moved from the center of the Woodlark Rift to its southern margin on the Papuan Peninsula (Mann et al., 2009). Our analyses indicate significant Quaternary surface uplift on both the D'Entrecasteaux Islands and Papuan Peninsula, providing quantitative support for earlier qualitative inferences based on mountain front morphology (Hill et al., 1992; Little et al., 2007, 2011; Ollier and Pain, 1980). The greatest ongoing rock uplift and surface uplift occur in lower-plate rocks. Continued lower-plate uplift in the center of the Woodlark Rift is consistent with focal mechanisms from active low-angle normal faults north of the D'Entrecasteaux Islands shown in Figure 15 (Abers, 1991; Abers et al., 1997).

While we cannot exclude the possibility that rock uplift in the islands has abruptly ceased in the Holocene (Mann et al., 2009), this seems unlikely. Stream profiles in the study area actually indicate a regionwide increase in rock uplift rates during the Pleistocene. This change may be linked to a large shift in the Australia-Woodlark Euler pole at ca. 0.52 Ma (Taylor et al., 1999), when relative plate velocity vectors rotated counterclockwise from 30° near Normanby Island to ∼60° near the Suckling-Dayman massif, and slowed ∼40%. It is likely that this shift in plate kinematics subsequently caused spreading center segments in the Woodlark basin to rotate counterclockwise between 170 and 80 ka (Goodliffe et al., 1997; Taylor et al., 1995; Tregoning et al., 1998). On land, exhumed mylonites from the Prevost Range metamorphic core complex on Normanby Island transition from top-to-the-NNE shear, parallel to the 3.6–0.52 Ma extension direction, to top-to-the-SSE shear, parallel to the modern extension direction (Wallace et al., 2004), and slip probably increases on NNE-striking normal faults (Little et al., 2007). It is likely that the geomorphic evidence presented here for an eastward increase in uplift in the Prevost Range is a consequence of ongoing throw on a NNE-striking normal fault that is part of the Normanby transfer fault system east of the island (Little et al., 2007). At the present resolution, the relationship of the geomorphology to this profound change in relative plate motion remains unclear. However, it is possible that such a plate motion change led to a regionwide increase in rock uplift rate, through increasing fault linkages or the thermal evolution of the crust (Cowie and Roberts, 2001), a mantle response in the rift zone due to plate motion change, or a combination thereof.

In the D'Entrecasteaux Islands, similar patterns between active rock uplift and low-temperature cooling ages suggest that the processes driving present-day rock uplift and erosion are likely related to the processes that have driven exhumation in the region over the past ∼4 m.y. It has been proposed that feedbacks between focused erosion and isostasy (Fletcher and Hallet, 2004; Zeitler et al., 2001) have enhanced rock uplift and exhumation in the D'Entrecasteaux metamorphic core complexes (Davies and Warren, 1988). Lower-plate gneisses underwent partial melting and appear to have ponded and flowed westward at the base of the crust from 4 to 2 Ma, and then deformed into domal structures by 2.0–1.8 Ma as they were incorporated into the upper crust (Little et al., 2011), at about the time the D'Entrecasteaux Islands emerged above sea level in the Pliocene–Pleistocene (Baldwin and Ireland, 1995). Even though focused erosion may have enhanced rock uplift, particularly once domes breached the ocean surface, our observations indicate net surface uplift of these domes, requiring a mechanism other than erosion-driven isostasy. Hill et al. (1995) considered that the high elevations of the D'Entrecasteaux metamorphic core complex are supported by thick crust inflated by pluton emplacement. More recently, geophysical studies (Abers et al., 2002) have shown that the crust is thin (20–30 km) beneath the D'Entrecasteaux metamorphic core complexes, and it is 10–15 km thicker to the north and south (Fig. 15). Mantle extension has also occurred, with slower P-wave velocities beneath the D'Entrecasteaux Islands implying complete replacement of subcontinental mantle by asthenosphere (Abers et al., 2002). Bouguer gravity anomalies over the D'Entrecasteaux Islands indicate isostatic compensation (Abers et al., 2002). Buoyant asthenosphere, rather than lower-crustal flow, likely maintains this compensation and supports the high elevations at a regional scale. This area of positive buoyancy is almost certainly related to the thermal evolution of the rift and partial melting ahead of the westward-propagating Woodlark basin spreading center, as is manifest by active volcanism found in close association with the metamorphic core complexes (Fig. 15).

Thus, the landscape apparently records tectonic processes at two scales. Local (dome-scale) variability in uplift on land and basin subsidence offshore is the likely result of crustal extension, faulting, and lower-crustal flow. This pattern is convolved with a regional (rift-scale) pattern arising from mantle flow, possibly driven toward the west by seafloor spreading in the Woodlark basin, and buoyancy forces. Whereas large-scale mantle flow generally produces dynamic topography with amplitudes <1 km and rates of change measured in tens of meters per million years (Braun, 2010), plumes impinging on the base of the lithosphere may briefly raise Earth's surface as much as ∼1 km at rates up to 3 km/m.y. (Hartley et al., 2011). Mantle processes beneath the Woodlark Rift may be driving surface uplift intermediate to these two extremes.

We used stream profile analysis to investigate the patterns of recent topographic evolution of the D'Entrecasteaux Islands and the eastern Papuan Peninsula, along the axis and southern margin, respectively, of the active Woodlark Rift. Stream profiles across the study area are characterized by well-defined knickpoints that most likely record a transient erosional response in channel networks to an increase in rock uplift rates during the late Quaternary. Overall, topography is not in steady state: Relict topography above knickpoints has risen from ∼200 to 800 m during the late Quaternary, and mean basin elevations have by consequence also increased. Over this time, basin relief has increased by 40% in the D'Entrecasteaux Islands and 80% increase on the Papuan Peninsula.

Geomorphic attributes, including an estimate of transient stream incision and the normalized channel steepness index, were correlated locally and regionally to facilitate quantitative spatial comparison. Over local and regional scales, these attributes correlate with available geologic and thermochronologic evidence for differential rock uplift. We showed that transient incision and normalized steepness index are useful guides for inferring neotectonic displacements on faults.

Spatially, stream profiles indicate that late Quaternary rock uplift and surface uplift increase from east to west across metamorphic core complexes and fault blocks in both the D'Entrecasteaux Islands and Owen Stanley Range. Although the Woodlark Rift is clearly a manifestation of lithospheric extension in a region of complex plate motions, geophysical data show that the high present-day elevations of the D'Entrecasteaux Islands are supported by buoyant asthenosphere beneath thinned crust. Large-scale geomorphic patterns drawn among actively rising metamorphic core complexes suggest broad linkages connecting surface, crustal, and mantle processes along the Australian-Woodlark plate boundary. In the future, improved dating of landscape changes, including more accurate constraints on the timing of transient stream incision, will further clarify the relationship between tectonic and surface processes as well as the geodynamic evolution of the rapidly evolving Woodlark Rift.

We are grateful for many discussions on eastern Papuan geology and tectonic evolution with colleagues, notably, Tim Little, Laura Webb, Laura Wallace, Paul Mann, Hugh Davies, Ian Smith, Geoff Abers, and Roger Buck. We are indebted to our late colleague, Alec Waggoner, for especially spirited discussions leading up to the writing of this paper. We thank Kelin Whipple and colleagues for the development of the stream profiler tool (www.geomorphtools.org), and Eric Kirby and Peter Sak for useful discussions. This project was supported by the National Science Foundation EAR Continental Dynamics Program (EAR-0709054). Alex Whittaker and an anonymous reviewer provided insightful comments that substantially improved the paper.