Delineating the Exmouth mantle plume (NW Australia) from denudation and magmatic addition estimates

Volcanic margins are rifted margins characterized by massive igneous activity. These features have been commonly explained as a result of extension and decompression melting over a thermal mantle anomaly of 100–150 °C (e.g., White and McKenzie, 1989; White et al., 2008; Rooney et al., 2011), which is generally explained as plume induced (e.g., Courtillot et al., 1999; Montelli et al., 2004). Other explanations include small-scale convection or state an absence of significant thermal anomalies due to fertile mantle (e.g., Korenaga et al., 2002; Anderson, 2005). Crucial factors in modern popular volcanic rifting models are the prerift history and timing of the thermal anomaly (e.g., Bown and White, 1995; Armitage et al., 2010). These models assume that lithospheric thinning can only be achieved by extension. Traditionally, actively upwelling mantle or plume material was thought to flatten below stiff lithosphere (e.g., Farnetani and Richards, 1994). However, recent studies suggest that lithosphere can be eroded by plume-generated convective currents or gravitational instabilities (e.g., Moore et al., 1999; Ribe, 2004; Agrusta et al., 2013; Brune et al., 2013).

One of the defining tests for the presence of a mantle plume or active upwelling has been the evaluation of domal surface uplift predicted by theoretical models (e.g., Farnetani and Richards, 1994). However, recently, it has been proposed that the uplift pattern might be more complex and show higher-frequency uplift superimposed on a lower-frequency pattern as a result of the stress state and lithospheric structure (e.g., Burov and Cloetingh, 2009; Burov and Gerya, 2014). Moreover, it has been suggested that there might be no uplift at all for a thermo-chemical plume (Sobolev et al., 2011). Nevertheless, domal uplift is still characteristic of many large igneous provinces (LIPs), as indicated by abundant geological and geophysical data (e.g., Saunders et al., 2007). Two types of uplift are generated by a plume; the first is transient uplift due to an upwelling thermal anomaly. The second is permanent uplift, owing to plume-induced partial melting of the lithosphere, generating a magmatic underplate or high-velocity body at lower-crustal levels (e.g., Tiley et al., 2004). Moreover, underplating of high-density mantle melts can lead to fractionation and shallower magmatic activity, evidenced by regional dolerite intrusion complexes and flood basalt extrusions (Ridley and Richards, 2010).

Rifted margin uplift at magmatic rifts is commonly characterized by linear uplift paralleling the rifted margin (e.g., Menzies et al., 2002), as a result of decompression melting and subsequent magmatic addition at the breakup margin. Furthermore, rift flank uplift is also influenced by the flexural rigidity or the equivalent elastic thickness of the lithosphere. Part of the flank uplift is transient, as a result of the thermal anomaly or plume temporarily elevating the rift margin above sea level and allowing initial flood basalts to flow downhill, before subsiding and forming seaward-dipping reflector series (e.g., White et al., 2008). These features are often difficult to unravel in outcrop studies where large flood basalt outpourings have obscured earlier vertical motions and erosion events. Alternatively, later erosion might have stripped away any evidence of earlier uplift phases. However, better horizontal and vertical resolution can be obtained on magmatic rifted margins from offshore seismic profiles tied in with hydrocarbon exploration wells.

The NW Australian offshore area has long been recognized as a volcanic margin (e.g., White and McKenzie, 1989; Coffin and Eldholm, 1994) characterized by minor postrift subsidence and relatively limited magmatic activity (Symonds et al., 1998), with a large intrusive component sourced by an underlying high-velocity body (Rohrman, 2013). Theories on the origin of the NW Australian LIP (Fig. 1) have been hampered by poorly constrained spatial and temporal evidence. Legacy models for the LIP propose either a loosely constrained mantle plume without a clear hotspot track, or a rifting-related mechanism (e.g., Mihut and Müller, 1998; Karner and Driscoll, 1999; Mutter et al., 1989; Hopper et al., 1992). However, due to extensive seismic-reflection coverage and various exploration wells drilled by the hydrocarbon industry, this region provides an excellent natural laboratory in which to study the relationships among magmatism, uplift, sedimentation, and breakup. In this article, I focus on: (1) estimating the spatial extent and relative amount of uplift from observations on denudation (erosion), magmatic addition, and sedimentation, and (2) explaining the results with a simple isostatic mass balance and a one-dimensional melting model. The findings suggest the presence of a mantle plume before margin breakup in the NW Australian region. Moreover, the results explain most of the currently available data.

The NW Australian margin (Fig. 1) consists of a number of basins and continental fragments that formed as a response to several extensional episodes (e.g., Longley et al., 2002; Gibbons et al., 2012). The south is characterized by Paleozoic basins such as the Bernier Platform. Further to the north, there are the Mesozoic Exmouth subbasin, Barrow subbasin, and Exmouth Plateau, while to the west, there is the fragmented and poorly known Wallaby Plateau (Fig. 1; Sayers et al., 2002; Stagg et al., 2004).

The main basin-forming event took place in the Late Permian to accommodate up to 7 km of deltaic Mungaroo Formation (Fig. 2). In the Exmouth subbasin and Barrow subbasin, Late Triassic and Middle Jurassic extension provided accommodation for a thick Jurassic section (Fig. 2), while on the Exmouth Plateau, extension was limited, and the Early to mid-Jurassic interval is a highly condensed section, bounded at the top by the Callovian regional unconformity (Longley et al., 2002).

The southern Exmouth subbasin records extensive Early Cretaceous erosion. However, the northern Exmouth subbasin, Barrow subbasin, and southern Exmouth Plateau record deposition of the south-sourced Berriasian–Valanginian Lower Barrow Group delta (Figs. 1 and 2), indicating major erosion of a largely missing source area. Lower Barrow sediments are separated from the overlying Birdrong Group by the Valanginian unconformity, marking continental breakup between Australia and Greater India along the Cuvier margin, which is linked to the Gascoyne margin by the Cape Range fracture zone transform (Robb et al., 2005).

Magmatic sills and dikes of tholeiitic affinity (Ludden and Dionne, 1992) intruded the Mungaroo and younger formations and were traditionally assigned a Late Jurassic to Early Cretaceous age (Symonds et al., 1998), predating and synchronous with breakup (Fig. 2). Sills and dikes have been sampled in only two wells, ODP 766 and YE-1 (Fig. 1), and most of the evidence on the presence of intrusives comes from seismic-reflection data. Recently, indirect dating of vent structures associated with sill intrusion on the Exmouth Plateau and Exmouth subbasin yielded Late Jurassic ages (Rohrman, 2013; Magee et al., 2013, 2015), predating breakup, whereas McClay et al. (2013) inferred Early Cretaceous ages for sills at the Cuvier breakup margin. The origin of the Exmouth Plateau intrusive sheets is most likely from the underlying high-velocity body (Fig. 3A; Rohrman, 2013). Mutter et al. (1989) and Lorenzo et al. (1991) first identified this high-velocity body in expanded spread profile (ESP) seismic data below the Exmouth Plateau, interpreted as underplated magmatic material at the base of the crust (Hopper et al., 1992) with seismic velocities of ∼7 km/s. Overall, intrusive activity seems more intense near the intersection of the Cape Range fracture zone and Cuvier margin, as evidenced by large gabbroic intrusions (Müller et al., 2002), dike-dominated complexes (Fig. 1), and anomalously thick sills originating from the Cape Range fracture zone region. Features synchronous with breakup are the extrusion of seaward-dipping reflectors and flood basalts from the Gascoyne margin, Cuvier margin rift axis, Wallaby Saddle, and in the Galah Province (Fig. 1; Symonds et al., 1998; Rey et al., 2008). Extrusives are relatively scarce in the Exmouth region compared with other volcanic margins, with many of the extrusive rocks in the South Exmouth subbasin probably eroded, as observed by truncated sills on seismic data (Mihut and Müller, 1998).

Finally, the area experienced postrift sedimentation and subsidence throughout the Cretaceous and Tertiary, while the Exmouth Plateau was affected by a Campanian inversion event (Bradshaw et al., 1998), followed by regional Neogene compression as a result of plate reorganization (e.g., Longley et al., 2002).

Lorenzo et al. (1991) viewed the Cape Range fracture zone as a leaky transform as a result of a hot ocean spreading ridge juxtaposed with stretched continental crust of the Exmouth Plateau, generating an ∼10-km-thick underplated wedge below the Exmouth Plateau and ∼3 km of denudation. Lorenzo et al. (1991) argued that the lack of a hotspot trail is evidence for the absence of a mantle plume and therefore supported an interpretation of secondary convection (Hopper et al., 1992). However, shallow intrusions in the Exmouth Plateau sourced by the underlying underplate are Late Jurassic (Rohrman, 2013) and older than the Hauterivian (magnetic anomaly M10N) ocean spreading ridge in the Cuvier Abyssal Plain (Robb et al., 2005), indicating that the Late Jurassic underplate must have another origin. Karner and Driscoll (1999) proposed a four-phased extension and breakup model for the Exmouth Plateau and Exmouth subbasin to explain the current basin fill and crustal configuration. Estimated pure shear stretching factors for the Permian are around 1.4 on the Exmouth Plateau, decreasing eastward, while in the Exmouth subbasin and Barrow subbasin, Late Triassic and Middle Jurassic extension with a pure shear stretching factor of ∼1.2–1.5 provided accommodation for a thick Jurassic section (Fig. 2). Ongoing extension in the Callovian and Kimmeridgian decreased stretching to values below 1.2. Meanwhile, the Exmouth Plateau experienced limited extension in the Late Jurassic (β ∼1). Tithonian–Valanginian lower-crustal and mantle thinning factors on the Exmouth Plateau are around 2.65–2.8, while upper-crustal extension is limited (<1.2). This discrepancy was interpreted as a crustal detachment by Karner and Driscoll (1999). Closer to the Gascoyne margin, crustal stretching increases to ∼7, while lower-crustal and mantle stretching are around 10, resulting in breakup (Driscoll and Karner, 1998). Initial oceanic half-spreading rates for the Cuvier Abyssal Plain and Gascoyne Abyssal Plain are ∼35 mm/yr (Fig. 1; Gibbons et al., 2012). Karner and Driscoll’s (1999) model predicts partial melting using ambient mantle temperatures (T_p ∼1330 °C), generating an underplate thickness of ∼7 km near the Gascoyne breakup margin to 600 m over the central Exmouth Plateau during Tithonian–Valanginian extension. However, ocean bottom seismometer lines and potential field modeling (Goncharov et al., 2006; Direen et al., 2008) suggest that the maximum underplate thickness is offset from the COT (Continental Ocean Transition) toward the central Exmouth Plateau and thins near the Gascoyne margin (Fig. 3B).

Finally, Mihut and Müller (1998) and Müller et al. (2002) proposed a mantle plume of ∼400 km diameter for the region, centered on the eroded Bernier Platform (Fig. 1), with a poorly constrained hotspot track moving northwest along the Wallaby Zenith fracture zone (Fig. 1). The Wallaby Plateau has been recognized to consist mainly of continental crust, and only the northern part, known as the Quokka Rise (Fig. 1) is still considered to be igneous (Symonds et al., 1998; Sayers et al., 2002); thus, this theory no longer holds. The proposed plume is also located away from main magmatism on the Exmouth Plateau, Galah Province, and Cape Range fracture zone. In the next section, I will introduce seismic and well data and use these data as a basis for modeling efforts to establish a spatial denudation pattern and discriminate between rifting- or mantle plume–related uplift.

Various two- and three-dimensional (2-D and 3-D) seismic-reflection data sets were used in this study. However, the primary data set was the 2D NWS07 seismic survey acquired by PGS in 2007 using a shot interval of 37.5 m, a streamer length of 8 km, and a nominal fold of 106. Total record length is 12 s. Figure 3 shows two perpendicular seismic lines on the Exmouth Plateau., across the Cape Range fracture zone transform margin (A) and Gascoyne passive margin (B), respectively. Profile A depicts a magmatic underplate (purple) derived from ESP and gravity data (Fig. 1; Mutter et al., 1989; Lorenzo et al., 1991), estimated at ∼10 km thick near the Cape Range fracture zone (Figs. 1 and 2) and thinning toward the northeast and sourcing the shallower sill complex (Rohrman, 2013). These intrusive sheets are visible as high-amplitude (hard) reflectors cutting the Triassic Mungaroo stratigraphy (Fig. 3). The blue-green marker is an intra-Triassic (intra-Mungaroo) reflector indicative of 2–3 km of denudation near the Cape Range fracture zone, consistent with previous estimates (Lorenzo et al., 1991). The overlying Lower Barrow Group (yellow) shows a distinctive prograding pattern near the Cape Range fracture zone, developing into a basinward aggrading sequence. Profile B displays a thinning high-velocity body (underplate) toward the Gascoyne margin breakup axis, with seismic velocities between 7.1 and 7.8 km/s obtained from ocean bottom seismometers (Fig. 1; Goncharov et al., 2006) and a basinward arch where the underplate is thickest. The Lower Barrow Group thickens toward the east and shows a prominent inversion in the center of the profile indicative of a later Campanian inversion event (Bradshaw et al., 1998).

Here, I use an Airy isostatic approach to estimate denudation and uplift. Denudation is taken to be equal to erosion and is estimated from seismic-reflection and well data as well as backstripping exploration wells, and it is calculated as the amount of missing section relative to a reference point (either seismic data and/or offset well data). Erosion is assumed to be subaerial. Uplift is defined as tectonic uplift (e.g., Rowley and White, 1998).

Two profiles were used to obtain denudation estimates, the first, C–C′ (Figs. 4 and 5), starts at the Cape Range fracture zone, runs SW-NE, and uses denudation data from Lorenzo et al. (1991) and seismic-reflection data calibrated to offset well data (Fig. 3) up to ∼60 km from the Cape Range fracture zone, while denudation at larger distances from the Cape Range fracture zone is based on backstripped exploration wells V1 (Vinck-1), I1 (Investigator-1), and J1 (Jupiter-1; Cloetingh et al., 1992). Relatively little denudation seems to have taken place at distances from 60 km to 200 km from the Cape Range fracture zone (V1 to J1; Fig. 4), which is typical for large tracts of the Exmouth Plateau. Errors from seismic-reflection data are assumed to be ±100 m within 60 km of the Cape Range fracture zone, since most denudation has taken place on the Triassic Upper Mungaroo Formation, the stratigraphy of which is well defined from seismic stratigraphy and well data. However, these error estimates are up to ±500 m when there are large eroded regions, and there are no suitable reference sections. This applies to profile D–D′, where denudation is not well defined, mainly owing to poorer-quality seismic data and rapidly changing geology, so we have to rely on well data. Figure 6 shows a well correlation panel from the Exmouth Plateau into the Exmouth subbasin. The section is flattened on top Jurassic (top gray zone) and suggests significant erosion at the Jurassic-Cretaceous boundary in the Exmouth subbasin (wells H1, YE1, Ju1). Furthermore, these wells have large sections of Upper Jurassic Dingo Formation (Fig. 2) missing. Erosional products were deposited as the sand-rich Lower Barrow Group (well I1). However, since the Dingo Formation consists of mainly claystones, the origin of most of the Lower Barrow Group must have been elsewhere, most likely to the southwest of the Cape Range fracture zone, as evidenced from Figure 3A, depicting prograding deltaic sequences from the south in the Lower Barrow Group. Denudation at H1 is between 500 m to 1.5 km, while at YE-1, it is ∼1 km, decreasing to ∼500 m at Ju1 (Fig. 5B). This is consistent with the study of Müller et al. (2002), who quoted ∼1.5 km of denudation between SP1 and Pa1 (Fig. 5B). Overall, the denudation pattern defined in Figure 5B follows an ellipsoid or domal shape, with maximum denudation around 2.6 km at the Cuvier margin and Cape Range fracture zone and about one third of the dome missing owing to plate tectonics. Moreover, the intersection between the Cape Range fracture zone and Cuvier margin also appears to have the highest amount of magmatic intrusions. However, there could be a component of rifting-related uplift along the Cuvier margin due to potentially large differences in flexural rigidity between the margin and the cratonic hinterland. Magmatic intrusion and denudation patterns along the Gascoyne margin with small flexural rigidity differences are more likely generated by rifting owing to its better-defined rift-parallel nature and underplating distribution (Fig. 5).

To calibrate the model, I use profile C–C′ (Figs. 4 and 5), since the Exmouth Plateau has reasonably well-documented underplate thicknesses derived from several ESP and ocean bottom seismometer data (e.g., Mutter et al., 1989; Lorenzo et al., 1991; Fomin et al., 2000; Figs. 1, 3, and 5A), although there is a wide range in seismic vintages. However, using these in Equation 1 and assuming no contribution from lithospheric thinning yields denudation values that are too high compared to seismic- and exploration well–derived denudation.

Using the observed values of D from C–C′ (Fig. 4), it is possible to calculate X and L. A reasonable fit is obtained when ΔT = 100–150 °C and maximum T_p = 1430–1480 °C; lower T_p values would require higher lithospheric thinning estimates and lateral temperature gradients between the Cape Range fracture zone and J1 location, while lower L values would require a higher T_p or the addition of eclogite. Using the 50/50 mid-ocean-ridge basalt (MORB)–peridotite mix solidus (Fig. 8; e.g., Spandler et al., 2008), mimicking addition of 50% eclogite to peridotite mantle, it is possible to make some crude estimates about melt thicknesses, since experimental data on the MORB-peridotite mix liquidus are not well constrained (e.g., Pertermann and Hirschmann, 2003). However, the lowest ΔT would still be ∼75 °C. The high T_p is consistent with elevated temperatures inferred from Early Cretaceous tholeiitic sill geochemistry at Ocean Drilling Program (ODP) Site 766 (Ludden and Dionne, 1992) (Fig. 1) and elevated heat flow from paleothermal indicators in various Exmouth Plateau and Exmouth subbasin wells (J1, I1, Ju1) suggesting a heat flow increase from ∼55 mW/m² to ∼70 mW/m² at the Jurassic-Cretaceous boundary (He and Middleton, 2002) up to 250 km away from the Cape Range fracture zone in a subcircular pattern and mimicking the underplating extent (Fig. 5). A comparison of the X/L relationship obtained from Figure 8 with measured denudation (Figs. 4 and 5), using Equation 1, suggests that the true or effective underplate thicknesses are ∼40% of the seismically measured value. This implies that measured seismic underplate (high-velocity body) thicknesses do not consist of 100% crystallized melt and, instead, are a mix of intruded melt and lower-crustal host rock (White et al., 2008), since seismic velocities for lower crust and underplate overlap (Mooney et al., 1998). High-velocity body thicknesses below oceanic crust (Fig. 5A) were not considered in the analysis, since they are most likely affected by later extension and decompression-related magmatism.

The denudation curve in profile C–C′ (Fig. 4) was calculated using Equation 1 with a conversion factor for X of 40% and decreases from ∼4 km near the Cape Range fracture zone to 440 m at J1. L is estimated around 30–40 km (Cape Range fracture zone) to 10−15 km (J1). Late Jurassic paleo–water depths are presumed to drop from 0 near the Cape Range fracture zone to 100 m over the Exmouth Plateau. We can now use the denudation/underplate relationship from profile C–C′ and apply it to profile D–D′, where underplate thicknesses are missing (Fig. 5A), using Equation 1 (Fig. 4). Measured denudation in the west of D–D′ suggests a calculated effective underplate of 2.5 km and L = 17.5–23.3 km, but higher values are possible closer to the COT (Fig. 4). Thus, I propose that this region is close to where a plume/upwelling hit the lithosphere, consistent with the magmatic record.

Figure 9 depicts tectonic subsidence versus time for two locations in Figure 4 (Cape Range fracture zone and J1). While J1 is constrained by well analysis using standard backstripping techniques (e.g., Cloetingh et al., 1992), the Cape Range fracture zone location lacks a well but has similar Triassic seismic stratigraphy and tectonic subsidence history to J1. Tectonic uplift (U) at the Cape Range fracture zone is estimated from the amount of eroded Triassic section from Figure 4 (D = 2.6 km), calculated back into the tectonic subsidence curve (U = 545 m) using the relation from Chadwick (1985): U = D(ρ_m – ρ_s)/ρ_m (Fig. 9). Timing of earliest plume uplift is from recognition of the Callovian unconformity, followed by intrusive magmatism in the Tithonian (Rohrman, 2013) on the Exmouth Plateau. Anomalous subsidence after the plume event (Fig. 9) can be explained by the flow of hot plume material toward the breakup axes during extension (e.g., Sleep, 1997; Buck, 2004; Armitage et al., 2009). Following plume exit from the region, cooling further contributes to thermal subsidence.

Although the Cape Range fracture zone–Cuvier margin area suffered maximum denudation, thinning, and underplating, most of it must have taken place on a continental fragment that filled the current Cuvier Abyssal Plain gap, before breakup. Where did this fragment go? One solution is that this sliver drifted away with Greater India and was subsequently integrated in the Himalayan orogeny. However, Gibbons et al. (2012) proposed that the Sonja Ridge and Sonne Ridge region (Fig. 1) are continental crust fragments that rifted away from the Cuvier margin. To test if the Sonne and Sonja Ridge fragments could be the missing pieces that were originally located to the south of the Cape Range fracture zone, I use the present-day Exmouth and Wallaby Plateau crustal thicknesses as an analog (Symonds et al., 1998). This suggests that the Sonja-Sonne Ridge fragment’s initial Late Jurassic crustal thickness must have been similar at ∼15–20 km, as a result of earlier extension episodes. Denudation analysis from profile C–C′ (Fig. 4) suggests ∼3 km denudation for the Sonja-Sonne Ridge fragment, consisting of Triassic Mungaroo Formation and Early Cretaceous mafic rocks, and requires an effective underplate of ∼4 km for the Sonja-Sonne Ridge fragments (Fig. 1) and 30–40 km of lithospheric thinning. Palynology and zircon provenance studies of the Lower Barrow Group (Fig. 2) on the Exmouth Plateau suggest mainly erosion products from the Mungaroo Formation (Fig. 5B; Exon and Buffler, 1992; Lewis and Sircombe, 2013) from a southern direction, consistent with major denudation of the Sonja-Sonne Ridge fragments, while Lower Barrow Group shale analysis points to an additional mafic volcanic component (Fig. 1; Exon and Buffler, 1992). Further evidence for a southern source area comes from conglomerates in the Lower Barrow Group, sampled by wells in the northern part of the Exmouth subbasin, indicating a proximal source in the south. Sediment mass balance calculations for the eroded region (∼70,000 km³) compare with the Lower Barrow delta (∼90,000 km³), although approximately a fifth of the delta was derived from Greater India or Antarctica (Fig. 1; Lewis and Sircombe, 2013).

Subsequent stretching of the lithosphere with a stretching factor of 2–3 would thin the Sonja-Sonne Ridge fragment crust to 5–10 km, resembling oceanic crustal thickness, but without coherent magnetic striping (Ebinger and Casey, 2001). This also explains the similar present-day water depth of the underplated Sonja-Sonne Ridge fragment compared with normal-thickness Cuvier Abyssal Plain oceanic crust (Sayers et al., 2002; Stagg et al., 2004).

I argue that the timing for plume impact was between ca. 165 and 136 Ma, causing magmatism and uplift (Figs. 2 and 9). The origin of the plume or active upwelling can possibly be tracked down from below the Pilbara craton (Fig. 1), undergoing an uplift phase in the Jurassic (Ghori et al., 2005) and deflecting the upwelling toward the thinner lithosphere of the Exmouth region where partial melting took place. This is consistent with the recognition of an large low-shear-velocity province edge below Western Australia at 160 Ma, assumed to be the birthplace of mantle plumes (Torsvik et al., 2010). Underplating, magmatism, and denudation were focused on the intersection of the Cuvier margin and Cape Range fracture zone (Fig. 1), characterized by a remnant positive Bouguer gravity anomaly (Rey et al., 2008). On impact, the plume head spread below the rest of the Exmouth region (Fig. 9A), generating thinned lithosphere, underplating, elevated heat flow, and shallow intrusives. The limit of underplating from ESP and ocean bottom seismometer data is depicted in Figure 5 and is based, apart from aforementioned data, on the presence of a thermal spike at the Jurassic-Cretaceous boundary from vitrinite reflectance/T_max modeling (He and Middleton, 2002). While wells within the underplate extent region (Fig. 5; e.g., I1, J1, Ju1, H1) do have this modeled heat spike, wells outside this area (e.g., NG1, A1) do not. This upwelling event was separate from later rift-related magmatism, as evidenced by an offset in the magnetic and seismic signature of the underplate from rifting-related magmatism (Goncharov et al., 2006; Direen et al., 2008), as well as the postulated Late Jurassic age of the underplate (Rohrman, 2013). It is likely that the region with hot thin asthenosphere exceeds that depicted for underplating, although the evidence is sparse. I speculate that the small sill complex in the northeast of the Exmouth Plateau, outside the underplated region (Fig. 1), consists of low-Ti tholeiites or alkaline basalts, directly sourced from the hot thin asthenosphere, similar to the Afar hotspot (Beccaluva et al., 2009), while the main sill complexes are sourced by the underplate closer to the center of the plume.

Having established the initial plume or upwelling location, we can now concentrate on the surface expression of plume activity using the plate model of Gibbons et al. (2012) (Fig. 10). This model has the Australian continent relatively stable during the Mesozoic. Ocean spreading of normal-thickness oceanic crust at the Cuvier and Gascoyne margins commenced in the Hauterivian (Robb et al., 2005; Rey et al., 2008). At this point, the expression of plume activity or hotspot track left the area in a relative western direction (Fig. 10) to arrive at the igneous Quokka Rise (Figs. 1 and 10B; Sayers et al., 2002), reflecting eastward movement of the Australian continent and assuming a stationary plume. Finally, the hotspot moved to the Zenith Plateau, just before Quokka Rise and Zenith Plateau were separated by rifting ca, 120 Ma (Fig. 10C; Gibbons, et al., 2012), having been active for 30–45 m.y. There is a possibility that the hotspot could have moved to the Greater India plate at that time; evidence exists in Tibet, where the Comei magmatic province records mafic sills and dikes dated between 150 and 130 Ma in strongly folded Late Triassic to Cretaceous sedimentary rocks (Zhu et al., 2008, 2009). Interestingly, the recent plate reconstructions of Gibbons et al. (2012) place the Comei province very close to the NW Australian margin (Fig. 10), suggesting that the Comei province could be a missing piece of the Exmouth mantle plume.

The model and data integration presented here delineate a Late Jurassic mantle plume in the greater Exmouth region, feeding underplate and shallower intrusive sheets and influencing subsequent continental breakup. The novel approach in this isostatic assessment is that plume-derived hot mantle with a T_p of 100−150 °C above ambient mantle was directly linked to lithospheric thinning, subsequent melting, and underplate generation, resulting in the observed denudation. The plume interpretation is consistent with recently obtained T_p ∼1560 °C from olivines found in 128 Ma picrites of the Comei region (Fig. 10; Xia et al., 2014)), suggesting even higher excess mantle temperatures (160–270 °C), although this is dependent on the choice of ambient mantle temperatures (Putirka, 2005). An alternative to the present model is emplacement of the underplate laterally like a giant sill from above a plume conduit. In this case, all the denudation at J1 is due to the sill, ∼1.5 times thicker than in the plume head case, since the lithospheric thinning component in Equation 1 is zero. However this model seems less likely, because seismic-reflection data below the Exmouth Plateau suggest evidence for the high-velocity body to be sourced by presumed ultramafic sills/dikes (Rohrman, 2013) derived from the thinned lithosphere. Fertile mantle and delamination (Elkins Tanton and Hager, 2000) were used by Sobolev et al. (2011) to explain the Siberian flood basalts. However, in the case of the Exmouth region, there is no evidence for such an extensive melting event from the geological record. Furthermore, eclogite would reduce denudation, owing to its high density. Thus, it seems unlikely that significant amounts of eclogite melting (>15%) were involved in the Exmouth mantle plume, even when considering higher values for the thermal expansion coefficient (α) (Schutt and Lesher, 2006). Limited sampled magmatism at ODP Site 766 provides evidence for low-pressure fractionation of plagioclase and clinopyroxene (cumulate gabbro; Ludden and Dionne, 1992), consistent with the underplating theory (Farnetani et al., 1996), though a normal oceanic source cannot be completely ruled out (Ishiwatari, 1992). Other dredge samples from the region are highly weathered (Sayers et al., 2002) and unsuitable for detailed geochemistry interpretation.

Mantle plumes are sometimes linked to driving continental breakup through the viscous normal stress of a mantle diapir initiating breakup (e.g., Hill et al., 1992). However, dynamic models tend to overestimate viscous normal stress and subsequent uplift (e.g., Farnetani and Richards, 1994), and plume-related faults are seldom observed, suggesting normal stresses are relatively minor. Highly eroded Precambrian LIPs display radial dike swarms attributed to plume activity (Ernst et al., 2005), but these are most likely fracture related and consist of fractionated tholeiitic dolerite, suggesting an underplate origin (Cox, 1980). A decade ago, Burov and Guillou-Frottier (2005) proposed a dynamic mantle plume model using a more realistic rheology for the lithosphere, implying that the uplift pattern might be more complex and mimic features that are commonly associated with plate tectonics. However, if the plumes are relatively weak, hot convective features, then the associated magmatism and circulation of hot volatiles are more likely to generate fault weaknesses and perhaps décollements (Heinson et al., 2005), which are typical of the Exmouth Plateau. Hence, plumes could enable breakup, instead of driving it (e.g., Buck, 2004; Armitage et al., 2009; Brune et al., 2013). These observations are confirmed in this study and seem consistent with other volcanic margins, such as the Deccan Traps and North Atlantic (Saunders et al., 2007), where the sequence of events is more obscured by magmatism. However, Armitage et al. (2010) stressed the importance of rift history, stating that prior extension is crucial for the amount of generated magmatism along the margins. NW Australia is, however, complex and fragmented, making it difficult to verify this theory.

Figure 11 depicts a cartoon of the evolution of the Exmouth region along a N-S profile crossing the Exmouth Plateau and terminating on the Bernier Platform, using concepts outlined in this paper (e.g., Fig. 2), starting with the preplume extensional setting in the Early–Middle Jurassic (A), followed by plume activity, uplift, underplating, and sill intrusion in the Late Jurassic (B). This process leads to passive flow of the hot asthenospheric mantle to regions where active extension takes place (e.g., Sleep, 1997), with faults being weakened by intruding magma, leading to renewed or continuing magmatism and breakup in the Valanginian (C). By that time, the plume (now a hotspot) has left the area, and extension-related flood basalts (future seaward-dipping reflectors) are being generated while the asthenosphere is still anomalously hot during the Valanginian. This is evidenced by linear sill complexes close to the Gascoyne and Cuvier passive margin, possibly related to breakup (McClay et al., 2013). After the hot plume material has thermally decayed, regional tilting generates seaward-dipping reflectors, followed by normal ocean spreading commencing in the Hauterivian (D).

Although the present model successfully explains currently available data, there are alternatives. Essentially, the proposed model is very similar to White and McKenzie’s (1989) original model, without a crustal thinning component. It can be argued that the configuration and denudation pattern at the start of rifting (Fig. 9) can be explained by simply placing a 100-km-thick hot asthenosphere layer at the base of the lithosphere, tapering toward the edge, with a similar configuration as in Figure 10 and an isostatic compensation level at 200 km. This eliminates the prerequisite of a hot upwelling being able to thermally thin the lithosphere, but it would generate no underplating, unless fertile mantle (eclogite) is added (Fig. 8). However, the amount of added eclogite has to be very high (up to 50%) to generate the underplate thicknesses required. This seems in contradiction with geochemical studies (Sobolev et al., 2007) that quote eclogite percentages of <20% for the Siberian LIP and selected ocean island basalts. Thus, it seems that this scenario is less plausible than the one originally proposed.

A rifting and decompression melting origin (Nielsen and Hopper, 2004; Armitage et al., 2009) seems unlikely due to the denudation and magmatic addition distribution pattern, with higher denudation and magmatic thickening toward the transfer zone margin (Cape Range fracture zone), instead of the passive margin, where relatively minor denudation and magmatic addition are observed. The underplated region along the Cape Range fracture zone (Lorenzo et al., 1991) could be described as part of transform margin generation (e.g., Robb et al., 2005; Gregg et al., 2009; Hebert and Montesi, 2011). However, the scale and spatial extent of magmatism would be more difficult to explain. Moreover, deposition of the Lower Barrow Group indicates denudation of a much larger area than that defined by the Cape Range fracture zone and south Exmouth subbasin. Depth-dependent stretching (e.g., Karner and Driscoll, 1999) is theoretically almost identical to the model presented here. Both models are characterized by mantle thinning in the absence of, or with limited, upper-crustal stretching. However, the Karner and Driscoll (1999) model predicts lower-mantle thinning factors (2.65) close to the Cape Range fracture zone, increasing northeastward to 2.8, while underplate thicknesses decrease in the opposite direction, requiring a decrease in stretching. Figure 8 depicts the stretching scenario. In order to make this work, higher T_p near the Cape Range fracture zone (+50 ° C hotter) is required, while at the J1 location, T_p needs to be –50 °C cooler compared to ambient mantle. This is theoretically possible, if the age of the Cape Range fracture zone underplate is similar to the timing of seafloor spreading in the Cuvier Abyssal Plain. However, sills and underplate are older than seafloor spreading. Hence, the stretching scenario does not work, and the plume model provides a better and more elegant solution.

Some uncertainties remain. The main issues concern the large uncertainty in denudation estimates, up to ±500 m and paleo–water depths (e.g., van Hinte, 1978), which can account for denudation errors of several hundred meters. This makes it difficult to validate the proposed model, when modeled denudation is equal or less than the error estimate. However, by locating the region of plume impact from maximum denudation and increased magmatism, these issues can be circumvented. Moreover, a recent assessment of paleo–water depths at deep-water margins (Crosby et al., 2011) indicated that uncertainties can be greatly reduced by sticking to a few rules: (1) using negligible bathymetry at the start of rifting, (2) using geometry of seismic reflectors, such as clinoforms, indicative of potential changes in paleobathymetry, and (3) relating the predicted time-depth cooling of nearby oceanic crust to deep-water margins.

The Exmouth region has undergone multiple additional denudation events since breakup, although their magnitudes have been limited. The most important is the global Valanginian lowstand event (Fig. 2), coinciding with breakup and generating a few hundred meters of erosion and subsequent redeposition (Ghori et al., 2005) postdating plume-related uplift (Fig. 2). This event is identified off the Peedamullah Shelf (PS) in O1, an area unaffected by underplating (Fig. 1). Another, Campanian uplift event inverted the Lower Barrow Group sediments on the Exmouth Plateau, with minimal erosion (Bradshaw et al., 1998). Finally, a more localized uplift phase took place in the Neogene as a result of plate reorganization (e.g., Longley et al., 2002). Taking these uncertainties into account, calculated plume-induced denudation values most likely represent maximum values.

This paper proposes that a mantle plume event took place in the Exmouth region in the Late Jurassic, predating breakup and generating lithospheric thinning, magmatism, and tectonic uplift as delineated by isostatic denudation analysis. The locus of uplift was located below a now highly extended and subsided continental fragment known as the Sonne Ridge–Sonja Ridge region, with erosional evidence recorded in the sedimentary record of the Exmouth Plateau as the Barremian to Valanginian Lower Barrow delta. Elevated heat flow and extensive intrusive magmatism further bear evidence to the active upwelling event. A crucial factor in this analysis is the recognition and spatial extent of a high-velocity body below continental crust from ESP and ocean bottom seismometer data, recognized as the source of shallow sills and dike complexes observed in seismic-reflection data from the Exmouth Plateau, Exmouth subbasin, and Wallaby Plateau (Rohrman, 2013). High-velocity body thicknesses were used as input to calculate mantle thinning and subsequent denudation analysis. Where high-velocity body data are absent, available denudation data were used to calculate high-velocity body thickness and mantle thinning, filling in the gaps. It transpires that only ∼40% of the high-velocity body in the Exmouth region is directly related to the upwelling event, with the remaining percentage consisting of preexisting lower crust. Denudation is explained by a simple thermal model, linking plume-induced lithospheric thinning with melting and underplate formation. The T_p of the plume is estimated at 75–150 °C above ambient mantle temperatures (∼1330 °C) and does not require addition of eclogite, although the presence of small amounts (<15%) remains possible. The thermal plume eroded up to 30–40 km lithosphere of the prebreakup Sonne Ridge–Sonja Ridge–Cape Range fracture zone–Cuvier margin region and generated an ∼4-km-thick effective underplate and ∼500 m of tectonic uplift. Meanwhile, distant regions like the Exmouth Plateau and Wallaby Plateau only underwent 10–15 km of lithospheric thinning by the plume head, with effective underplating less than 500 m thick and denudation less than a few hundred meters. Therefore, it might be difficult to delineate a plume event below sedimentary basins when lithospheric thinning factors are low (<20 km), due to error margins in measurements equaling or exceeding modeled values. However, in this case, the region of highest denudation and observed intrusive magmatism delineates the plume. Widespread plume-induced lithospheric thinning set the boundary conditions for subsequent extension-related magmatism and breakup. After initial upwelling, the thermal plume can be traced as a hotspot to the Quokka Rise and Zenith Plateau in the mid-Cretaceous and possibly to Greater India.

I would like to thank PGS for permission to use seismic images from the New Dawn two-dimensional seismic survey. John Armitage and an anonymous reviewer are thanked for useful and constructive reviews. Furthermore, I would like to thank Lithosphere Editor Arlo B. Weil for editorial handling, as well as various anonymous reviewers for thorough comments on previous versions of the manuscript.