Animated reconstructions of the Late Cretaceous to Cenozoic northward migration of Australia, and implications for the generation of east Australian mafic magmatism

The generation of basalts is frequently ascribed to the processes of rifting and hotspot activity (e.g., Wilson, 1963; White and McKenzie, 1989). The Late Cretaceous through Cenozoic magmatic history of eastern Australia is possibly somewhat unique in that it includes periods of what have been previously described as both rift-related and hotspot-related basaltic magmatism that overlapped in time and space (e.g., Fig. 1; Wellman and McDougall, 1974a). Cenozoic rift-related mafic volcanism in eastern Australia has been attributed to the Paleogene opening of the Tasman and Coral Seas (Fig. 2; Ewart et al., 1988). In apparent contrast, a series of onshore and offshore, age-progressive, bimodal shield volcanoes that decrease, at least in part, in both age and volume to the south have been attributed to the northward motion of Australia over stationary and irregularly active (but primarily waning) mantle plumes (e.g., Wellman and McDougall, 1974a; Cohen et al., 2007; Sutherland et al., 2012; Jones and Verdel, 2015). Detailed ⁴⁰Ar/³⁹Ar geochronological studies of these age-progressive volcanoes have previously been used to quantify plate velocity and, in doing so, have revealed Cenozoic periods of anomalously slow plate motion (Knesel et al., 2008; Cohen et al., 2013). In particular, an apparent late Oligocene to early Miocene period of significant plate slowdown has previously been linked to docking of the Ontong Java Plateau (OJP; Fig. 2) with the Solomon Islands (Knesel et al., 2008).

Several previous studies have used paleomagnetic data to reconstruct the Cenozoic motion of Australia. From these studies, two general types of apparent polar wander paths (APWPs) have emerged: a linear path (i.e., an APWP with latitudinal changes but little variation in longitude; Fig. 3A; Idnurm, 1985; Musgrave, 1989) and a competing path that includes significant variation in longitude (Fig. 3B; Embleton and McElhinny, 1982). The linear APWP is largely based on paleomagnetic data from sedimentary basins and laterites, and it is consistent with the northward trajectory of India (Idnurm, 1985). However, paleomagnetic data from sedimentary and lateritic profiles can be affected by inclination shallowing and (particularly in the case of laterites) a complex history of inheritance and crystallization of magnetic minerals that could result in oversmoothing of APWPs (Butler, 2004; Vasconcelos et al., 2008). In contrast, the longitudinal APWP is based largely on paleomagnetic data from a combination of basalts and laterites (Embleton and McElhinny, 1982). The longitudinal APWP includes a significant westward diversion during Miocene time (Fig. 3B), a possible artifact that could have arisen from the failure of some previous paleomagnetic studies to account for paleosecular variation of the Earth’s magnetic field (Idnurm, 1985). Alternatively, this diversion in the APWP may be accurate, and it may correspond with the eastward offset of the Lord Howe Seamount Chain and Tasmantid seamounts (McDougall and Duncan, 1988; Knesel et al., 2008; Fig. 1), as well as the initial soft docking, which refers to a collisional event without deformation, of the OJP with the Solomon Islands (Knesel et al., 2008).

The generation of east Australian Cenozoic basalts, as well as the conflicting APWPs of the Australian plate described here, are thus fundamental components of the overall Cenozoic tectonic history of Australia and the southwest Pacific. Cenozoic plate reconstructions are therefore powerful tools for clarifying the relationship between tectonic and magmatic events that shaped the Australian continent, Papua New Guinea (PNG), the Solomon Islands, the OJP, and the Tasman and Coral Seas (Fig. 2). In this study we use a combination of plate reconstructions and ⁴⁰Ar/³⁹Ar geochronology to test both the competing Cenozoic APWPs and the relationship of volcanism to plate motion. We used G-PLATES software (https://www.gplates.org/) to create two Late Cretaceous through Cenozoic plate reconstructions that correspond with the linear and longitudinal APWPs. We assess these two reconstructions, and thus the APWPs on which they are founded, by evaluating whether they reproduce two key geologic and tectonic observations: (1) an eastward offset in the Lord Howe Seamount Chain and Tasmantid seamounts (Knesel et al., 2008); and (2) a reduction in the velocity of Australia between 26 and 23 Ma, which coincides with anomalously high volume and eruption rates of the Tweed shield volcano (Fig. 1; e.g., Knesel et al., 2008; Cohen et al., 2013). In addition, we evaluate the agreement between these APWPs and the global moving hotspot reference frame (GMHRF) model of Doubrovine et al. (2012).

East Australian Late Cretaceous–Cenozoic volcanoes have traditionally been divided into two broad groups: central volcanoes and lava fields (Fig. 1; Wellman and McDougall, 1974a). The key attribute used to distinguish the two groups is that central volcanoes, although dominantly basaltic, include a component of felsic flows and/or intrusions, whereas lava fields are nearly entirely basaltic, with a few exceptions of limited felsic flows (Wellman and McDougall, 1974a; Sutherland et al., 2014). The central volcanoes form a series of north-south chains of bimodal shield volcanoes along the east coast of Australia (Fig. 1; Wellman and McDougall, 1974a, 1974b; Sutherland et al., 2012; Davies et al., 2015). The oldest of the central volcanoes in the Bowen Basin region of northeastern Australia erupted at Hillsborough ca. 34 Ma, and the northern portion of the chain extends south to the Buckland province, which erupted ca. 27 Ma (Fig. 1; e.g., Wellman and McDougall, 1974a; Sutherland et al., 2012). At the latitude of Buckland this inland volcanic track appears to end, although it may resurface in central New South Wales as a series of leucite-bearing basaltic provinces of Miocene age, and it possibly continues south through Victoria to the Bass Strait (Sutherland, 1981; Cohen et al., 2008; Davies et al., 2015).

Approximately 500 km east of Buckland, a second, coastal track of central volcanoes is exposed at Fraser Island (Fig. 1), which is characterized by 30.4 ± 0.2 Ma mafic magmatism. The coastal track continues south from Fraser Island through New South Wales, although the number of tracks involved is debatable (Sutherland et al., 2012). The youngest exposure of any central volcano is the ca. 6 Ma Macedon-Trentham province in Victoria (Fig. 1; Sutherland et al., 2014). Based on ⁴⁰Ar/³⁹Ar ages of the final felsic products erupted from each central volcano, the overall north-south chains form age-progressive tracks that become younger to the south (Cohen et al., 2013). The age progression of the central volcanoes is mirrored by the offshore Tasmantid seamounts and Lord Howe Seamount Chain (Fig. 1), and all of the east Australian volcanic tracks have been previously ascribed to the Cenozoic northward passage of the Australian plate over multiple stationary thermal anomalies (Wellman and McDougall, 1974a; McDougall and Duncan, 1988; Cohen et al., 2007; Sutherland et al., 2012; Davies et al., 2015).

Lava fields are generally considered to contrast with central volcanoes in that lava fields are not bimodal, shield building, or age progressive, but nevertheless they often overlap temporally with central volcanoes or are adjacent to central volcanoes (Wellman and McDougall, 1974a, 1974b; Cohen et al., 2007). These previously described distinctions notwithstanding, several lava fields include evolved lithologies ranging from trachytes (Newer Victorian Volcanics) and mugearites (e.g., Liverpool, Walcha, and New England) to phonolites (McBride and Maybole; part of New England). In addition, some lava fields (e.g., Barrington) are remnants of shield volcanoes (Johnson et al., 1989). Eruption of the lava fields has been linked with opening of the Tasman and Coral Seas (Ewart et al., 1988; O’Reilly and Zhang, 1995), although many lava field provinces were active long after rifting of those ocean basins ceased (e.g., the North Queensland lava fields erupted from ca. 8 Ma to recent time; Fig. 1; Wyatt and Webb, 1970; Griffin and McDougall, 1975; Stephenson et al., 1980). Previous explanations for lava-field generation include decompression melting associated with rifting, as well as the ascent of diapers and mafic underplating from an anomalously hot asthenosphere (Johnson et al., 1989). However, no conclusive evidence of Cenozoic underplating has been recognized (Johnson et al., 1989), and neither of these explanations accounts for the absence of conjugate magmatism in eastern Gondwana rifted fragments that are now to the east of the Coral and Tasman Seas.

A series of lava fields extend in an east-west alignment between the inland and coastal central volcanic chains in the region between Buckland and Fraser Island (Fig. 1). The spatial and temporal associations of these particular lava fields with the neighboring central volcanoes blur the distinctions between rift-related and plume-related magmatism in eastern Australia. In this study we contribute new ⁴⁰Ar/³⁹Ar data from the largest of these lava fields (Bauhinia and Monto; Fig. 1), as well as from basalts of the nearby, but putatively distinct, central volcanoes of Peak Range and Springsure (Fig. 1). Similar lava field provinces, such as the western portion of the New England (formally central) Province, exhibit no time-space correlation, but they share isotopic characteristics with nearby central volcanoes (Vickery et al., 2007) and have since been included in some depictions of age-progressive tracks (Sutherland et al., 2012).

The past motion of the OJP may be directly relevant to the Cenozoic motion of the Australian plate and the distribution of some of the east Australian central volcanoes (e.g., Knesel et al., 2008). The OJP is to the north of the Solomon Islands (Fig. 2) and was originally part of the larger Ontong Java–Manihiki–Hikurangi Plateau (e.g., Taylor, 2006). Break-up of this earlier plateau and subsequent northwest drift of the OJP to its current position occurred during the Cretaceous and Cenozoic (Taylor 2006), although the timing of its collision with the Solomon Islands remains contentious (e.g., Yan and Kroenke, 1993; Petterson et al., 1999; Mann and Taira, 2004; Holm et al., 2016). A hiatus of arc magmatism in the Solomon Islands suggests that docking of the OJP and subsequent blocking of the subduction zone occurred between 25 and 20 Ma (Petterson et al., 1999), leading to subduction on the northern side of the OJP by 12 Ma (Yan and Kroenke, 1993). Plate reconstructions, coupled with an eruption hiatus in the Solomon Islands arc, suggest early Miocene docking (Yan and Kroenke, 1993; Hall, 2002), but, because the original extent of OJP is unknown, it is possible that its leading edge was subducted before 20 Ma (Hall, 2002; Holm et al., 2013). However, other studies have concluded that there was no cessation in subduction (and therefore no contact between the Solomon Islands and OJP) until ca. 5 Ma (e.g., Mann and Taira, 2004). Even accounting for subduction of the leading edge of the OJP, paleomagnetic evidence from both the OJP and the accreted Malaita terrane at the Solomon Islands constrains the collision to ca. 20 Ma (Musgrave, 2013), and there is no substantial geological evidence of a ca. 25–20 Ma collision in the Solomon Islands (Mann and Taira, 2004; Holm et al., 2013; Musgrave, 2013).

Changes in the Cenozoic motion of Australia may have also been related to interactions between PNG and tectonic elements to the north. During the early Miocene, the PNG-Australian plate boundary terminated where the Medial Mountains System is currently located (Fig. 2; Hamilton, 1979). A north-dipping Medial Mountains subduction zone and an active island arc system were north of the boundary. A complex history of oblique continent-arc collisions is recorded on the New Guinea peninsula, including several collisions in the Eocene, Oligocene, and perhaps early Miocene (Hall, 2002; Baldwin et al., 2012). The timing and number of continent-arc collisions are matters of debate: some studies suggest that composite terranes were formed before collision, and other studies argue for separate collisional events for each arc (Fig. 4; van Ufford and Cloos, 2005; Baldwin et al., 2012). For simplicity, we consider the collision of the composite terranes, rather than individual arcs. The Sepik terrane accreted in the early Oligocene (Hill and Hall, 2003), and the Philippines-Halmahera arc accreted in the late Oligocene (ca. 25 Ma; Hill and Hall, 2003). These terranes form part of what we refer to as the North New Guinea terranes (Fig. 4; Pigram and Davies, 1987; Crowhurst et al., 1996; Davies, 2012; Holm, 2013). During the late Oligocene to middle Miocene, the continental crust of PNG collided with the East Papua composite terrane (Pigram and Davies, 1987) and a series of Paleogene island arcs (Jaques and Robinson, 1977; Gaina and Müller, 2007; Davies, 2012). The timing of the Finnisterre arc collision is possibly the most contentious. Some studies suggest a single collision ca. 5 Ma (Abbott et al., 1994), while other studies propose two separate collisional events: the collision of the Finnisterre arc with PNG in the middle Miocene, and a subsequent Pliocene collision between the Finnisterre arc and Bismark arc (Pigram and Davies, 1987). Collision of PNG with these exotic terranes progressively led to cessation and subsequent reversal of the subduction zone (Jaques and Robinson, 1977; Baldwin et al., 2012).

The detachment of the Loyalty slab may have also affected the pattern of Cenozoic mafic magmatism in eastern Australia (Sutherland et al., 2012; Cohen et al., 2013). Previous studies have described an extinct northeast- to east-dipping subduction zone to the north of New Caledonia (e.g., Schellart et al., 2006, 2009). The subduction zone extended from the transform boundary at d’Entrecasteaux past New Caledonia to Northland (Fig. 2). During the late Eocene–early Oligocene, a series of ophiolites and terranes were obducted onto New Caledonia (Aitchison et al., 1995), followed by obduction onto Northland between 24 and 21 Ma (Mortimer et al., 2003). Toward the end of the subduction period, subhorizontal tearing of the slab is believed to have taken place, resulting in slab detachment at 30 Ma in the north and 24 Ma in the south (Schellart et al., 2009). Detachment of the slab induced magmatism along the subduction zone, resulting in volcanism in the Norfolk Basin, New Caledonia, and Northland (Schellart et al., 2009).

The timing of proposed slab tearing roughly coincides with a major period of Cenozoic volcanic activity in eastern Australia. Sutherland et al. (2012) suggested that east Australian volcanism was related to perturbed mantle flow from progressive tearing of the slab during detachment; they argued that opening of a slab window during detachment could cause asthenospheric mantle convective swells and related volcanism. Cohen et al. (2013) came to a different conclusion, arguing that greater plate velocities during the Oligocene could be due to the initiation of slab tearing between 29 and 28 Ma.

In order to include east Australian volcanism in our tectonic reconstructions, we compiled a data set of previous geochronological results (Supplemental Material¹) from east Australian central volcanoes and lava fields, largely assembled by Vasconcelos et al. (2008). We incorporated additional published U-Pb (Sutherland et al., 2012, 2014), K-Ar (Wyborn and Owen, 1986; Gray and McDougall, 2009), and ⁴⁰Ar/³⁹Ar geochronology results (Matchan and Phillips, 2011; Cohen et al., 2013; Sutherland et al., 2014) and, as described in detail in the following, we added new ⁴⁰Ar/³⁹Ar data from four key east Australian volcanic provinces (Peak Range, Springsure, Bauhinia, and Monto; Fig. 5). U-Pb ages that were in agreement with corresponding zircon fission track and K-Ar ages from the same volcanoes were included in the data set, unless the authors noted Pb contamination, loss of Pb, or if there was evidence of zircon inheritance (Sutherland and Fanning, 2001). K-Ar data were used only when there were insufficient ⁴⁰Ar/³⁹Ar results for comparison. K-Ar data that were considered unreliable by the original authors were excluded, as were K-Ar results that did not agree with ⁴⁰Ar/³⁹Ar ages from the same formation or sample, particularly if argon loss, recoil, or excess argon had been identified.

We selected 18 mafic samples from the Peak Range (7 samples), Springsure (6 samples), Bauhinia (4 samples), and Monto (1 sample) volcanic provinces for ⁴⁰Ar/³⁹Ar geochronology (Fig. 5). The samples were collected from roadcuts and other outcrops in an approximately east-west transect across the Bowen Basin. Samples were crushed in a tungsten-carbide percussion mill, and whole-rock crushed fragments (∼1 mm³) were cleaned using distilled water and ethanol in an ultrasonic bath and hand-picked under a binocular microscope. Approximately five fragments from each sample were loaded into a 21-pit aluminum disk, as per Vasconcelos et al. (2002). Sanidine from the Fish Canyon Tuff (age of 28.201 ± 0.04 Ma; Kuiper et al., 2008) and GM1550 (biotite, 98.79 ± 0.96; Renne et al., 1998) were also loaded as secondary standards. The disks were irradiated for 14 h in the CLICIT (cadmium-lined in-core irradiation tube) facility TRIGA-type reactor at Oregon State University. Two whole-rock aliquots were analyzed for each sample on a MAP 215–50 mass spectrometer in the University of Queensland Argon Geochronology in Earth Sciences Laboratory (UQ-AGES). Samples were continuously heated with an Ar-ion laser with a defocused beam, as outlined in Vasconcelos (1999) and Vasconcelos et al. (2002). To constrain the accuracy of the ⁴⁰Ar/³⁹Ar results of the whole-rock aliquots, two grains of GA1550 biotite from each irradiation disk were also analyzed by the incremental-heating method. Additional details of the ⁴⁰Ar/³⁹Ar procedure and numerical data are available in sections A and B of the Supplemental Material (see ^{footnote 1}).

Two animated reconstructions (Animations 1 and 2) were created using G-PLATES software (version 1.5.0). The linear reconstruction used the base reconstruction rotation file compiled by Seton et al. (2012) that consists of a database of the past 200 m.y. of rotation. Continents within the region, the Cenozoic volcanic provinces, the New Guinea composite terranes, and the OJP were digitized in their present-day locations, linked to the overall rotation of their respective tectonic plates, and reconstructed to their previous positions. Active central volcanoes are shown in orange in these reconstructions, and lava fields are shown in blue. Inactive provinces change to black. Red dots signify the locations and approximate ages of dated samples and give an idea of overall volcanic activity. The PNG terranes remain dark blue until their contact with continental New Guinea. Magnetic anomaly lines change from red to green as they increase in age, and inactive rift-zones are blue. While the initial positions of the North New Guinea terranes and East Papua composite terrane are unknown, the collisional timing of these terranes with PNG is well constrained (Pigram and Davies, 1987; Davies, 2012). Both terranes have been reconstructed relative to the Caroline plate (Fig. 2; Gaina and Müller, 2007), but their reconstruction is only approximate (Pigram and Davies, 1987; Hill and Hall, 2003; Davies, 2012).

A second reconstruction, the longitudinal reconstruction, was created using a paleomagnetic path for the Australian continent after 60 Ma based on data produced and compiled by Embleton and McElhinny (1982) after Wellman et al. (1969), Rahman (1971), and Wellman (1975). The model included virtual geomagnetic poles (VGPs) at 4.5, 15, 22, 25, 30, 50, and 60 Ma. The 15, 25, 30, 50, and 60 Ma poles are from a corrected lateritic profile (Embleton, 1981). A plate reconstruction pole was generated using these VGPs. To prevent the violation of relative plate boundaries that are assumed fixed, the reconstruction tree was arranged so that all other plates move relative to Australia.

The samples selected for ⁴⁰Ar/³⁹Ar geochronology contained euhedral to subhedral olivine (15%–30%), clinopyroxene (5%–10%), and plagioclase (10%–20%) phenocrysts in a groundmass (30%–60%) of olivine, clinopyroxene, and plagioclase ± orthopyroxene (Fig. 6). In several samples, olivine was altered to iddingsite and plagioclase was altered to sericite. Because olivine and clinopyroxene phenocrysts are sparse in these samples, their influence on the ⁴⁰Ar/³⁹Ar results was minimal. The results of the 34 whole-rock analyses are presented in Table 1 and Figures 7–9. Analytical results are reported in section B of the Supplemental Material (see ^{footnote 1}).

Seven samples from the Peak Range Province were analyzed (Fig. 7). The combined isochron is the preferred age for these samples, and the isochrons and step-heating plateaus of the samples were generally in agreement. Each sample produced a plateau that encompassed ∼80% of the released gas (Table 1), with no obvious signs of recoil or argon loss. The Peak Range province yields two main clusters of ages: samples PR01a, PR02, PR05, and PR07 produced ages of 26.0 ± 0.2 Ma, 33.3 ± 0.4 Ma, 31.6 ± 0.3 Ma, and 34.7 ± 0.4 Ma, respectively, while samples PR04, PR08, and PR09 produced significantly older ages of 47.2 ± 0.5 Ma, 44.7 ± 0.4 Ma, and 45.0 ± 0.3 Ma, respectively.

Six samples from Springsure were analyzed, producing a restricted age range of 27.8 ± 0.3–28.2 ± 0.2 Ma. The combined isochron is also the preferred age for these samples (Fig. 8). Sample IJS07 showed signs of extensive alteration, with initial ⁴⁰Ar/³⁶Ar outside the error of the present atmospheric ratio (298.56 ± 0.31; Lee et al., 2006, Table 1). Samples IJS08b, IJS11b, CVLF01, and CIK produced well-constrained combined isochrons of 28.0 ± 0.4 Ma, 27.8 ± 0.3 Ma, 28.2 ± 0.3 Ma, and 28.2 ± 0.2 Ma, respectively. A single grain isochron is the preferred age for sample CIS. The combined isochron of CIS exceeds the present atmospheric ratio, but the second fragment produced an isochron with no excess Ar and an age of 28.0 ± 0.3 Ma.

Four samples from the Bauhinia province and one from the Monto province were analyzed, and the combined isochrons yield the most reliable ages for these samples. Of the samples from the Bauhinia region, TRLF01, TRLF04, TRLF05, and CIB produced isochrons with ages of 22.5 ± 1.8 Ma, 28.0 ± 0.3 Ma, 27.0 ± 2.8 Ma, and 26.8 ± 0.3 Ma, respectively, an overall range from ca. 28 to 23 Ma (Fig. 9). Samples TRLF01 and TRLF05 produced combined isochron ages of 22.5 ± 1.8 Ma and 27.0 ± 2.8 Ma, respectively. In these latter samples, excess Ar (⁴⁰Ar/³⁶Ar intercept above 298.6 ± 0.31) led to relatively large errors. As a result of alteration, only one fragment from TRLF05 (27.8 ± 1.5 Ma), and neither fragment from TRLF01, produced a plateau age. Despite the excess argon, the single plateau produced for the TRLF05 fragment was concordant with the age produced by the combined isochron. Without a plateau, the best estimate for the age of TRLF01 comes from the combined isochron.

The single sample from Monto (AA2) produced a combined isochron with an age of 29.0 ± 0.3 Ma, which is within error of the plateau age (29.1 ± 0.4 Ma) from a single fragment, but its ⁴⁰Ar/³⁶Ar of 330 ± 11 exceeded the present atmospheric argon ratio. As a result, the two fragments analyzed from sample AA2 produced plateau ages that are not within error: 29.1 ± 0.4 Ma and 30.2 ± 0.4 Ma. The plateau age of the second grain includes the low-temperature steps, which may reflect atmospheric argon. The preferred age is from the single isochron, which produces an age of 29.1 ± 0.4 Ma and an ⁴⁰Ar/³⁶Ar intercept at 296 ± 13.

The new ⁴⁰Ar/³⁹Ar results were added to the geochronology data compilation and incorporated into our two reconstructions. The reconstructions show alternate versions of the northward migration of the Australian plate during the past 100 m.y. (Animations 1 and 2). The most notable difference between the linear and longitudinal paths is between 50 and 14 Ma. While in the linear path the longitude of the Australian continent has been similar over the past 100 m.y. to the present-day longitude, in the longitudinal path Australia was farther to the east prior to 25–22 Ma, before diverting west and subsequently continuing on a northward trajectory (Fig. 3).

An inherent shortcoming of paleomagnetic poles is that they do not constrain absolute longitude (e.g., Butler, 2004), but plate motions described relative to the mantle can resolve longitudinal position (Müller et al., 1993). The GMHRF uses the position of seamount chains, which are considered to be the surface expression of plumes, to determine a best-fit rotation relative to their hotspot source. The motion of plates is approximated by fitting the age progression of the seamounts to their geometry. Many studies have used only hotspots from a single hemisphere (e.g., O’Neill et al., 2005; Torsvik et al., 2008), which may not represent the entirety of the mantle, but Doubrovine et al. (2012) produced a multiplate, multihemisphere GMHRF model that compares well with other hotspot reference frame models (e.g., O’Neill et al., 2005; Torsvik et al., 2008). We utilize the Doubrovine et al. (2012) model here because it is more likely to represent the entire mantle, rather than just the Northern Hemisphere.

We compared the linear APWP of Idnurm (1985), the longitudinal APWP of Embleton and McElhinny (1982), and synthetic VGPs derived from the GMHRF model of Doubrovine et al. (2012; Fig. 10). We used the APWP of Idnurm (1985) instead of the linear model used in Seton et al. (2012) because our goal is to compare only the APWPs. The latitude of the 60 Ma pole from the GMHRF (59.9°S) is slightly more similar to the 60 Ma pole from the longitudinal APWP (58.6°S) than the linear APWP (61.7°S). At 30 Ma, the latitude from the linear APWP more closely matches the GMHRF, although the latitudes from the longitudinal model are better matches to the GMHRF from 20 Ma onward. Notably, the pronounced apparent shift in longitude at 20 Ma in the longitudinal APWP is not obvious in the GMHRF, which includes a slight change in longitude (<0.4°) at that time, but otherwise the temporal resolution of the GMHRF is too coarse to resolve subtle shifts in plate motion. Although the longitudes of the synthetic paleopoles from the GMHRF are more similar to those of the linear APWP, the linear APWP and GMHRF have significantly different age-latitude relationships. For this reason, the longitudinal APWP seems to be a slightly closer match to the GMHRF, although, as noted here, the temporal resolution of the GMHRF is too coarse for detailed comparison with some of the most important and characteristic features of the longitudinal APWP.

As another way of comparing the various estimates for the Cenozoic motion of Australia, the northward component of Cenozoic velocity of the Australian plate was approximated by reconstructing the geographical position of an arbitrary point (27°S, 150°E) at 10 m.y. increments back to 60 Ma using the GMHRF, the linear APWP, and the longitudinal APWP (Fig. 11A). Model errors were calculated after Doubrovine et al. (2012). Northward plate velocity was relatively slow from 60 to 40 Ma in each of the three models (Fig. 11B), although the northward velocity of Australia during this period based on the longitudinal APWP (∼40 mm/yr) is closer to the velocity predicted by the GMHRF (∼30 mm/yr) than is the linear APWP (∼15 mm/yr). The GMHRF model indicates an increase in northward plate velocity to ∼60 mm/yr after ca. 40 Ma that remains virtually constant. The longitudinal and linear APWPs also indicate a marked increase in velocity to ∼60 mm/yr and ∼45 mm/yr, respectively, ca. 40 Ma. According to the longitudinal APWP, plate velocity decreased between 30 and 20 Ma to ∼30 mm/yr, before resuming relatively fast northward velocity of ∼55 mm/yr. Unlike the GMHRF and the longitudinal APWP, which produce plate velocities similar to modern rates (∼60–70 mm/yr; Tregoning, 2002; DeMets et al., 2010), the linear APWP indicates extremely fast plate velocities of ∼100 mm/yr from 10 Ma onward. Based on the overall fit of the GMHRF synthetic VGPs to the APWPs and the approximate relative plate speeds, we find that the GMHRF is more similar to the longitudinal APWP than the linear APWP.

Our animated reconstructions (Animations 1 and 2) illustrate the temporal and spatial development of east Australian Late Cretaceous to Cenozoic mafic magmatism in the context of the tectonic history of the southwest Pacific. The animations show that there was relatively limited east Australian volcanism from Cretaceous to mid Cenozoic time, including during the initial rifting of the Tasman and Coral Seas. Magmatism became widespread along much of eastern Australia during the latest Eocene to Oligocene, peaking during the late Oligocene to earliest Miocene. Volcanism dissipated in northeastern Australia ca. 20 Ma (early Miocene), although it continued along the southeast coast. By the middle to late Miocene (ca. 10 Ma), southeast Australian volcanism had largely ceased as well, such that there was little volcanic activity occurring in eastern Australia at that time. Volcanism was rejuvenated during Pliocene time in the Newer Volcanics Province of southeast Australia, as well as the North Queensland lava fields in northeastern Australia.

As noted above, the relationship between central volcanoes and lava fields is particularly acute in the region separating the inland and coastal central volcanic chains. In this region, ⁴⁰Ar/³⁹Ar ages from the Peak Range province, which is part of the central volcanic track of eastern Australia, plot into two groups. Four relatively young samples (PR01a, PR02, PR05, and PR07) range in age from ca. 35 to 26 Ma, consistent with previous geochronological results from the region (Cohen, 2007). Results from this young group suggest several distinct flows in the Peak Range volcano and peak activity between ca. 35 and 30 Ma, although a relatively young age of 26 Ma (sample PR01a) was obtained from the northern margin of the Peak Range volcano. The 26 Ma basalt postdates the late-stage felsic eruption used to define the age progression of the inland volcanic track, and it is consistent with relatively young ages from the adjacent Nebo province (ca. 21 Ma; Sutherland et al., 1977). The other group, which consists of samples PR04, PR08, and PR09 from the northwestern margin of the Peak Range volcano province, is significantly older (47.2 ± 0.5–44.65 ± 0.4 Ma). These mid-Eocene ages coincide with a prior ⁴⁰Ar/³⁹Ar result from the same area (Cohen, 2007), but are older than many previously published results from the Peak Range (Duncan and McDougall, 1989). The age range of the older group is also far outside the proposed 34–6 Ma time frame for the overall passage of eastern Australia over stationary hotspots (Cohen et al., 2007). The ⁴⁰Ar/³⁹Ar results from this old group of Peak Range basalts therefore imply that either some Peak Range hotspot-related magmas were produced far earlier than previously described, or the Peak Range central volcano overlies an earlier lava field. The relatively long duration of volcanism of Peak Range may constitute evidence that magmatism there originated from multiple processes (Wellman and McDougall, 1974a). These results imply a complex history of eruption of the Peak Range volcano that is difficult to reconcile with simple northward passage over a fixed hotspot.

Animations 1 and 2 were used to investigate the relationship between eruption of the lava fields and rifting in the Tasman and Coral Seas. Volcanic activity in the lava field provinces spans from the Late Cretaceous to Holocene, with periods of peak activity at 90 Ma (Embleton et al., 1985), 60–40 Ma (e.g., Duncan and McDougall, 1989), 34–15 Ma (e.g., Wellman and McDougall, 1974a; Vickery et al., 2007), and 5 Ma to recent (e.g., Henley and Webb, 1990; Gray and McDougall, 2009). From 34 to 6 Ma, mafic volcanism was widespread in eastern Australia but was not restricted to the central volcanoes. Many provinces, including the Monaro, Older Victorian Volcanics, New England, Liverpool, Hoy, and the Tasmanian, were active during this period but were distinctly unrelated to age-progressive volcanism. Peak activity of the relatively old (60–40 Ma) lava fields of Monaro, Hoy, Monto, the Older Victorian Volcanics, and Barrington (Duncan and McDougall, 1989) correspond with the opening of the Tasman and Coral Seas (73–52 Ma; Gaina et al., 1998a, 1998b), a temporal relationship that is consistent with previous suggestions that the lava fields are vaguely related to Tasman and Coral Sea rifting (Ewart et al., 1988; O’Reilly and Zhang, 1995). However, at the time of their eruption, our reconstruction places the distance between the spreading ridge in the Tasman and Coral Seas and the aforementioned volcanoes at >200 km (White and McKenzie, 1989; van den Bogaard, 2013). In many examples from around the world, the inception of rifting is associated with volcanism in close proximity to either side of the rift (e.g., White et al., 1987; White and McKenzie, 1989; Franke, 2013); however, in the case of Tasman and Coral Seas rifting, volcanism was restricted to the east Australian coast at a considerable distance from the rift. The lack of correspondence between the onset of rifting in the Tasman and Coral Seas and eruption of the lava fields raises some doubts about whether lava fields are truly rift related. Furthermore, subsequent Oligocene to Miocene (34–15 Ma) lava field magmatism, such as at Central and Doughboy, Liverpool, Bauhinia, Monto, and Mitchell, as well as the later pulses of volcanic activity in the Hoy and Monto provinces, occurred long after the cessation of rifting, and at a considerable distance from spreading ridges in the Tasman and Coral Seas (Wellman and McDougall, 1974a; McDougall and Roksandic, 1974; Vickery et al., 2007). Eruption of these lava fields was broadly coincident, however, with the period of age-progressive volcanism of the central volcanoes. As explained in more detail in the following, an alternative interpretation is that eruption of the lava fields was related to edge-driven convection (EDC). The formation of the present-day continental margin of eastern Australia via the opening of the Tasman and Coral Seas may have been a necessary condition for east Australian EDC, and, in that respect, eruption of the lava fields might be considered tangentially rift related. However, we find little evidence linking their eruption directly with rifting.

The distribution of new ⁴⁰Ar/³⁹Ar ages from Peak Range, Springsure, Bauhinia, and Monto illustrates a close temporal relationship in the region between central volcanoes and lava fields. Volcanic activity at Peak Range, a central volcano, occurred during three periods (47–44 Ma, 35–28, and a pulse at 26 Ma), although the older period may be more accurately described as lava field volcanism. The ⁴⁰Ar/³⁹Ar results from mafic and felsic parts of the nearby Springsure central volcano indicate activity from at least 29–27 Ma, a time frame that overlaps with both the second phase of volcanism at Peak Range and eruption of the adjacent Buckland central volcano (Fig. 1; Cohen et al., 2007). Immediately to the east of Springsure, our data reveal a protracted history of volcanism at the Bauhinia lava field between 28 and 23 Ma, and, farther to the east, the Monto lava field was erupting at 29 Ma. In other words, during a time interval from ca. 30–25 Ma, both central volcano and lava field basaltic magmatism occurred within a small region of southeast Queensland. Given that these provinces are largely coeval, geographically adjacent, and seem to grade into each other, we find it doubtful that there is a substantive difference in origin, in this region, between central volcanoes and neighboring lava fields. An implication of this conclusion is that previous clear-cut distinctions between plume-related central volcanoes and rift-related lava fields are likely inaccurate. Perhaps a more plausible scenario is one in which there were parts of eastern Australia, the Bauhinia region being a leading example, where the generation of Cenozoic basaltic magmatism of the fields and central volcanoes were intrinsically linked.

The most prominent alternative to the traditional plume-related and rift-related models of east Australian Cenozoic basaltic magmatism is EDC, which refers to convective cells of magma that arise from a lateral thermal gradient across a large change in lithospheric thickness (King and Anderson, 1998). EDC is generally believed to form at boundaries between hot asthenospheric mantle and cold lithospheric mantle. The juxtaposition forms a region of convective melting that circulates against the edge of the contact at sufficient plate speeds (Farrington et al., 2010).

The sporadic eruption pattern of Late Cretaceous to Cenozoic mafic magmas in eastern Australia includes a broad and irregular distribution of volcanism, multiple late-stage eruptions (e.g., a pulse of magmatism at Peak Range; 26 Ma, this study; and Nebo; ca. 20 Ma; Sutherland et al., 1977), and eruption interruptions (Jones and Verdel, 2015), attributes that deviate significantly from a classic plume model (e.g., Wilson, 1963). A useful analogue for east Australian mafic magmatism might be the Canary Islands seamount province, where older volcanism was dominant in the south, both older and younger volcanism occurred in the middle, and rocks of intermediate age are found in the north (van den Bogaard. 2013). The pattern of sporadic volcanism at the Canary Islands seamount province, which nevertheless forms a broad age progression, has been attributed to plume interaction with an EDC cell (Geldmacher et al., 2005).

In the case of eastern Australia, EDC may arise from steps in lithospheric thickness. Farrington et al. (2010) concluded that at current plate velocities, the temperature variation and shear flows at the east Australian stepped lithospheric boundary are within the expected range to support EDC. In fact, EDC has been proposed as a mechanism for formation of the lava fields of the Newer Volcanics Province (Demidjuk et al., 2007; Holt et al., 2013; Davies and Rawlinson, 2014). It is important that, while moderately high sublithospheric heat flow along eastern Australia is consistent with EDC, the absence of north to south gradients in heat flow is difficult to reconcile with proposed Cenozoic passage of eastern Australia over mantle plumes (Hasterok and Gard, 2016). We suggest that, similar to the Canary Islands, EDC may have played an important role in producing the Late Cretaceous to Cenozoic mafic magmatism in eastern Australia. EDC is less likely, however, to account for the age progression of the offshore Tasmantid seamounts and Lord Howe Seamount Chain, as discussed in the following.

The Lord Howe Seamount Chain and Tasmantid seamounts are thought to have formed as a result of the northward migration of the Australian plate over stationary hotspots (Fig. 1; e.g., Wellman and McDougall, 1974a), and both seamount chains have east-west offsets. Geochronology and robust regression models based on Cenozoic plate velocities place both offsets between 26 and 23 Ma, during a period of anomalously slow plate velocity (discussed in the following) (Knesel et al., 2008), although ⁴⁰Ar/³⁹Ar dating of Middleton Reef in the Lord Howe Seamount Chain (Mortimer et al., 2010) is ∼2 m.y. older than that predicted by robust regression plate models (Sutherland et al., 2012). These offsets coincide with a westward divergence of the Australian plate between 25 and 22 Ma in the longitudinal APWP (Fig. 12). There is no such early Miocene divergence in the linear reconstruction, so, in the case of this criterion, the longitudinal reconstruction is more closely aligned with readily observable physical features. A westward bend in age-progressive tracks is also present onshore after 17 Ma, implying an additional change in plate motion, although it is not evident in either reconstruction. It is likely that the precision of the APWP is insufficient to reflect this subtle shift in plate motion, which may be artificially intensified between Nandewar and Canobolas by proximity to the Great Dividing Range (Fig. 1).

An alternative explanation for the offset in the offshore volcanic chains is that they could be related to the structure of the east Australian lithosphere, as opposed to changes in plate motion (Fishwick et al., 2008). However, while offsets in the onshore and Tasmantid chains (Fig. 1) may correspond with a bend in the stepped-lithospheric boundary of eastern Australia (Fishwick et al., 2008), the offset in the Lord Howe seamounts does not (Knesel et al., 2008). Therefore a stepped lithosphere probably did not produce the same pattern of volcanism in both the Tasmantid and Lord Howe chains, nor could such a pattern be produced by purely northward motion of Australia, as illustrated in the linear reconstruction. Movement of the plumes is a possibility, although this explanation would require that both the Lord Howe and Tasmantid plumes shifted simultaneously and abruptly between 26 and 23 Ma (Knesel et al., 2008).

We used Animations 1 and 2 to evaluate the Cenozoic motion of a point in eastern Australia, which was somewhat arbitrarily chosen as Stradbroke Island (Fig. 1). The past latitude of Stradbroke Island is significantly different in the two reconstructions and from this difference we estimated changes in the magnitude of the Cenozoic velocity of eastern Australia (Fig. 13). For the longitudinal reconstruction, there is a reduction in the north-component of velocity to ∼30 mm/yr between 26 and 22 Ma. According to the linear reconstruction there is no late Oligocene to early Miocene reduction in plate velocity, although there is a reduction in the north component of velocity from 61.9 mm/yr to 42.5 mm/yr after 21 Ma. We then compared these velocity estimates with similar estimates from Knesel et al. (2008) and Cohen et al. (2013), and found correspondence with the estimates of the longitudinal reconstruction for the late Oligocene to early Miocene (Fig. 13). This correspondence is significant because it highlights agreement between two independent estimates of the Cenozoic motion of Australia: the longitudinal reconstruction is fundamentally based on the paleomagnetic data of Embleton and McElhinny (1982), while the estimates of Knesel et al. (2008) and Cohen et al. (2013) are based purely on high resolution ⁴⁰Ar/³⁹Ar geochronology from a number of east Australian central volcanoes.

As described previously, there are two likely causes for the late Oligocene to early Miocene slowdown of the Australian plate and contemporaneous shift in plate trajectory: soft docking of the OJP and the Solomon Islands, and collision of PNG with the East Papuan composite terrane. Although the longitudinal reconstruction seems to illustrate the effects on the Australian plate of one or both of these collisions, we are unable to distinguish the respective effects of the two because they were essentially coeval. Likewise, our results do not bear directly on the possible existence of detached slabs and slab windows beneath the Loyalty arc during middle Cenozoic time (Sutherland et al., 2012), although the general lack of a clear age progression in the overall Cenozoic volcanism in eastern Australia argues against an interpretation that links onshore volcanism with tectonic interactions in the Loyalty arc region.

Two competing Late Cretaceous–Cenozoic reconstructions of the motion of Australia illustrate several key points relevant to the overall tectonic development of the southwest Pacific. First, while the temporal resolution of the Cenozoic GMHRF (Doubrovine et al., 2012) is not sufficient to resolve an apparent late Oligocene–early Miocene slowdown or offset in the motion of Australia, the apparent latitudes and plate speeds for Australia before 40 Ma and after 30 Ma based on the GMHRF are more consistent with the longitudinal APWP of Embleton and McElhinny (1982). Second, this APWP, which incorporates paleomagnetic data from Cenozoic volcanic rocks of the Australian continent, includes a westward shift of the Australian plate at 26–22 Ma. The westward divergence coincides with a longitudinal offset in the Tasmantid seamounts and Lord Howe Seamount Chain. The linear APWPs of Idnurm (1985) and Musgrave (1989), which include no significant longitudinal movement, do not account for the offset in the seamount chains. The longitudinal APWP includes a late Oligocene to early Miocene period of reduced northward velocity of Australia that is detectable from high-resolution geochronology of age-progressive volcanoes in eastern Australia (Knesel et al., 2008). Therefore, comparison of the linear and longitudinal reconstructions suggests that, on balance, the longitudinal reconstruction is more compatible with observable geological features in the southwest Pacific.

Onshore east Australian age-progressive volcanism, which has been used previously as a basis for reconstructing the Cenozoic motion of Australia, seems to reflect a complex interplay between magmatic processes. We find no clear spatial or temporal distinctions between lava fields and central volcanoes, and no obvious correlations between even the oldest lava fields and the opening of the Tasman and Coral Seas. The widespread distribution of Cenozoic volcanism in eastern Australia, eruption hiatuses at individual volcanoes, and late-stage volcanism are not simply explained by a framework consisting of clearly distinct, plume-related and rift-related Cenozoic volcanoes. This period of east Australian magmatism is more likely fundamentally related to EDC arising from the stepped structure of the east Australian lithosphere.

We thank David Thiede for his assistance in the laboratory. The ⁴⁰Ar/³⁹Ar geochronology was funded by the Australian Coal Industry ACARP project C22023. Prof. Lin Sutherland and Prof. Dietmer Müller provided valuable comments and suggestions. In particular we thank Prof. D. Müller for his advice with GPLATES (https://www.gplates.org/).