A new 40Ar/39Ar eruption age for the Mount Widderin volcano, Newer Volcanic Province, Australia, with implications for eruption frequency in the region

ABSTRACT The Mount Widderin shield volcano is located near Skipton, western Victoria, in the Western Plains subprovince of the monogenetic Pliocene–Holocene Newer Volcanic Province (NVP). Radiometric ages for lavas in the Hamilton–Skipton–Derrinallum area are few, owing to limited suitable outcrop for K–Ar or 40Ar/39Ar geochronology studies. Existing age constraints for flows in this area have been inferred from Regolith Landform Units (RLUs), complemented by a small number of K–Ar studies on ≥1 Ma flows. Although the RLU approach provides a valuable overview of relative eruption ages across the NVP, it is of limited use in eruption frequency studies. Additional radio-isotopic ages are required to refine age ranges for individual RLUs, and to validate previous assignment of individual flows to specific RLUs. We report a new, high-precision 40Ar/39Ar age of 389 ± 8 ka (2σ) for a Mount Widderin basalt sample. Based on this age and geomorphic observations, we propose that both the Widderin and Elephant lava flows be reassigned from the Eccles RLU to the Rouse RLU. We use the 389 ± 8 ka (2σ) age for Widderin, along with published K–Ar ages, to anchor a stratigraphic sequence of 15 individual flows in the Hamilton–Skipton–Derrinallum area, demonstrating that intermittent volcanism has occurred in this area from ≥3 Ma to ≤0.389 Ma. Within the limits of available data for the NVP, this time span of volcanic activity is second only to that of the Melbourne area. We consider the significance of the Widderin eruption age, in conjunction with published age constraints for maars and scoria cones of the Western Plains subprovince, building on previous studies that have focused solely on lava flow ages. The inclusion of the additional data weakens the argument for a decrease in volcanic activity after ca 0.9 Ma as implied by published ages for lava flows only. Additional detailed combined geochronology–geomorphology studies of lavas, scoria cones and maars in strategically selected small areas are advocated to better understand eruption frequency across the NVP.

The NVP has been subdivided into three subprovinces based on geomorphological characteristics (Joyce, 1975;Ollier & Joyce, 1964): Central Highlands, Western Plains and Mount Gambier ( Figure 1). In the Central Highlands subprovince, volcanoes erupted within an already uplifted and dissected terrain (Boyce, 2013;Holdgate et al., 2006;Joyce, 1992Joyce, , 1999. In the Western Plains subprovince, volcanic centres are more sparsely distributed. Here, thin but extensive lava flows mantle a low-profile gently undulating surface developed on weathered Cretaceous to Neogene sediments (Joyce, 1999). The Mount Gambier subprovince is separated from the Western Plains subprovince by some 50 km. It contains a cluster of volcanoes in the northwest (Mount Burr Range) that are stratigraphically constrained to be of Pleistocene age (Holt, Holford, & Foden, 2014), and the Holocene eruption centres of Mount Gambier and Mount Schank in the southeast.
As reviewed elsewhere (e.g. Gray & McDougall, 2009;Matchan & Phillips, 2011), available age data indicate that volcanism initiated at ca 4.5 Ma in the Central Highlands subprovince, with the most recent activity in the Mount Gambier subprovince; however, there is no apparent age progression across the NVP. Based on a comprehensive compilation of KÀAr ages for lava samples from the Western Plains subprovince (n D 44), together with a small number of volcanoes with inferred <0.1 Ma ages (n D 5), Gray and McDougall (2009) proposed a peak in volcanic activity at 3.0À1.8 Ma. Although this interval could represent a peak in effusive volcanism, it may also be an artefact of relatively low sampling density in the Western Plains subprovince. Furthermore, maars and scoria cones were excluded from consideration, presumably due to a paucity of age constraints at the time of publication. This is an important observation because if the relative frequencies of different eruption styles have varied over the lifetime of the NVP, this could lead to significant bias in eruption frequency studies based solely on the ages of lava flows.
There are at least 175 eruption centres in the Western Plains alone, of which approximately half are lava shields or complex centres that experienced some period(s) of effusive eruption activity (Boyce, 2013). The remainder are scoria cones or maars that did not produce any significant amount of lava (Boyce, 2013). Most maars and scoria cones do not have any published age constraints. Improving the geochronological record of the NVP is important for understanding the petrogenetic evolution of the province, testing models for its origin, as well as establishing eruption frequencies for geohazard evaluation. On a local scale, improved age constraints for key volcanic units provide a stratigraphic reference framework for Quaternary landscape evolution and paleoclimatic variation.
Considering the number of eruption centres and thicknesses of basalt sequences (>100 m in places; e.g. Hare, Cas, Musgrave, & Phillips, 2005), the existing geochronological record for the NVP is relatively sparse. This is primarily due to limited availability of suitable material for dating (unaltered holocrystalline outcrop/drillcore), and the difficulty in attributing flows to specific eruption points, many of which may have been fissures/low-profile shields that have been subsequently buried. The existing NVP eruption record is overwhelmingly constructed from KÀAr dating studies on lava flows aged 0.5 Ma (e.g. Aziz-ur-Rahman & McDougall, 1972;Gray & McDougall, 2009;McDougall, Allsop, & Chamalaun, 1966). These data are complemented by a small number of relatively recent 40 Ar/ 39 Ar geochronology studies (Hare et al., 2005;Ismail, Phillips, & Birch, 2013;Matchan & Phillips, 2011. In the case of the very young (100 ka) scoria cones and maars in the Western Plains subprovince, published age constraints are largely derived from radiocarbon analyses of underlying swamp material or crater lake sediments (e.g. Builth et al., 2008), complemented by several cosmogenic nuclide exposure dating studies (Gillen et al., 2010;Stone, Peterson, Fifield, & Cresswell, 1997), luminescence studies (Sherwood, Oyston, & Kershaw, 2004;Smith & Prescott, 1987), and 40 Ar/ 39 Ar dating of anorthoclase megacrysts entrained by basalt melts (Ismail et al., 2013). 40 Ar/ 39 Ar (and KÀAr) dating of basalts younger than ca 0.5 Ma has historically been analytically challenging, owing to the difficulty in resolving extremely small radiogenic 40 Ar signals (typically <5À20% of total 40 Ar) from atmospheric argon. However, recent advances in noble-gas mass spectrometry have improved the precision achievable with the 40 Ar/ 39 Ar dating technique by an order of magnitude (e.g. Matchan & Phillips, 2014;Phillips & Matchan, 2013). Sample quality plays a critical role in the precision of individual ages, as any minor deficiencies in the key criteria (i.e. purity, lack of alteration and weathering, crystallinity, grainsize) have a measurable effect on the gas release profile during step-heating, and impact age precision and accuracy. Therefore, detailed  Matchan and Phillips (2011). Approximate boundary of the Western Plains subprovince as redefined by Boyce (2013) is indicated by dashed line. Study area is indicated by a black square. sample characterisation and careful evaluation of argon isotopic data are critical when evaluating the accuracy of eruption ages.
Adopting a similar approach to the KÀAr dating study of Gray and McDougall (2009), this study demonstrates how geomorphological observations can be used in tandem with radio-isotopic dating to unravel the eruption history of a complex region. We consider lava flows in the poorly documented SkiptonÀDerrinallum area, focusing on Mount Widderin ( Figure 2).

Geological setting and existing age constraints
Basalt flows in the SkiptonÀDerrinallum area have been previously mapped as two units: undifferentiated Plains Basalts and Stony Rises (e.g. VandenBerg, 1997; Figure 2). Although this simple classification system is useful (and necessary) from a regional geology perspective, it masks the complex eruption history of this area. MacInnes (1985) identified and mapped 14 separate lava flow units in the HamiltonÀSkipton area using a combination of aerial photograph interpretation, field observations, petrography and major-element chemistry. Relative ages were assigned on the basis of lava surface preservation, levels of soil development and field relationships, with a focus on flows to the west of Mount Emu Creek. The portion of the resultant map relevant to the SkiptonÀDerrinallum study area is reproduced in Figure 3, showing the nine basalt units identified in this area by MacInnes (1985)  profiles (MacInnes, 1985). The Moorallah and Hamilton basalt units in the west of the study area exhibit degraded stony rises. The Moorallah basalt is considered to have diverted the Mount Emu Creek to its present location ( Figure 3). On the basis of geomorphology alone, there is no obvious age difference apparent between the Moorallah and Hamilton basalts. The Fyans (#1) basalt exhibits stony rises and post-dates both the Moorallah and Hamilton basalts, having dammed the drainage network established on the southern margin of the Hamilton basalt (MacInnes, 1985). Minimal drainage pathways have developed on the Fyans (#1) basalt. The ages of the Elephant and Widderin flows, which both exhibit stony rises, relative to one another and the Fyans (#1) basalt, are unconstrained by stratigraphy.
The work of MacInnes (1985) was incorporated into the regolith landform map of Joyce (1999), refining earlier soil and regolith mapping schemes (Gibbons & Gill, 1964;Ollier & Joyce, 1986). In this classification system, the lava flows from Mounts Widderin, Elephant and Fyans are included in the Eccles Regolith Landform Unit (RLU) and assigned a nominal age of 200À0 ka (Joyce, 1999). The Mount Hamilton flows are assigned to the Rouse RLU (1À0.2 Ma), Mount Vite Vite flows to the Dunkeld RLU (3À1 Ma), and the remaining flows to the Hamilton RLU (5À4 Ma; Joyce, 1999). Gray and McDougall (2009) obtained the first KÀAr ages for basalts in this region, determining KÀAr ages of 3À2 Ma for basalts outcropping in the vicinity of Darlington (17 km SW of Derrinallum), previously assigned to the 5À4 Ma Hamilton RLU. However, Gray and McDougall (2009) noted that their reported ages are imprecise owing to high levels of atmospheric argon contamination resulting from analysis of weathered samples collected from the scant outcrop present in this area. These authors also obtained a single KÀAr age of 2.95 § 0.16 Ma (2s) for a tholeiitic basalt outcropping between Darlington and Derrinallum, corresponding to the 'Terrinallum basalt' unit of MacInnes (1985). Based on the stratigraphic age sequence of flows summarised above, an age of ca 3 Ma is regarded as a probable upper age estimate for volcanism in the SkiptonÀDerrinallum area, although we note that the undated Westmere and Streatham basalts may be older. Aside from a weighted mean KÀAr age of 0.99 § 0.04 Ma (2s) for the stony rises of the Stockyard Hill basalt (Gray & McDougall, 2009) 16 km NNW of Skipton, there are no further published age constraints for basalts in this region. Based on the flow stratigraphy outlined above, the Widderin, Elephant and Fyans (#1) basalts are classified as the youngest flows in the SkiptonÀDerrinallum area. No published age constraints exist for these flows.

Mount Widderin
The Mount Widderin shield volcano is located 6 km south of Skipton (Figure 2). A system of lava tubes near the summit of Mount Widderin (360 m elevation), commonly referred to as the Skipton Caves or Widderin Caves, includes one of the largest known chambers in the NVP (Ollier, 1963). These caves were once host to rare secondary phosphate minerals formed from guano, but these deposits have since been destroyed by human activity (Birch & Henry, 1993;Pilkington & Segnit, 1980). The Widderin basalt forms a gently undulating landscape ( Figure 4a) that is typical of degraded stony rises in the Western Plains (Ollier & Joyce, 1964;Skeats & James 1937). The local topographic highs (stony rises) comprise rounded, vesicular lava blocks with development of shallow red soil (Figures 4a,5). Soil is considerably thicker in the depressions between these stony rises, typically completely covering underlying basalt.
The Widderin lavas represent a considerable modification to the landscape and would have significantly reconfigured the local drainage system. The northwestern boundary of the basalt is now traced by the Mount Emu Creek, a major drainage feature in the BolacÀSkipton region ( Figure 2). The Widderin lavas flowed predominantly to the southwest, channelled for approximately 25 km by a paleovalley (Figures 2, 3). To our knowledge, the contact between the Widderin and Elephant lava flows has not been documented previously.

Comparison of Widderin and Elephant lava landforms and age implications
Sediments in the vicinity of Lake Logan somewhat obscure the contact between the Widderin and Elephant lava flows. However, field observations reveal that the morphology and preservation of stony rises from these two volcanoes are significantly different, and allow a contact to be mapped to an accuracy of »100 m, coincident with the location of Vite Vite Road (Figure 3; Supplementary Papers Figure B2). We observe that the flows from Mount Elephant commonly form steepsided mounds covered with angular lava blocks and thinly developed red soil (Figure 4b). The soil horizon is poorly developed and shallow in depressions, with small pieces of lava commonly breaking the surface. This contrasts with the rounded lava blocks of the undulating Widderin stony rises described above. In isolation, the difference in morphology between the Elephant and Widderin stony rises may be interpreted to chiefly reflect differences in lava viscosity, rather than significantly different erosional histories. However, the steep-sided and angular nature of the rises, taken together with the relatively limited soil development, suggests that the Elephant flows may be significantly younger than the Widderin flows. In turn, the Mount Elephant stony rises appear significantly more degraded than those of Mount Eccles/Budj Bim (Gillen et al., 2010), which have a reported cosmogenic 21 Ne exposure age of 36 § 6 ka (2s; Gillen et al., 2010), in agreement with radiocarbon ages of 32À28 ka for Mount Eccles crater lake sediments and post-basaltic swamp deposits (e.g. Builth et al., 2008).

Sample collection and characterisation
Determination of robust 40 Ar/ 39 Ar ages requires holocrystalline, unweathered basalt samples with negligible glass in order to minimise excess argon contents, radiogenic argon loss and irradiation-induced recoil issues. Owing to the weathered and highly vesicular nature of the stony rises, lack of road-cuttings through the Widderin lava profile and the absence of significant quarrying activity, accessing material suitable for 40 Ar/ 39 Ar geochronology proved challenging. For the current study, a 1 kg piece of apparently unweathered basalt with minimal vesicle content was carefully selected from a pile of rock extracted during construction of a »6 mdeep well on the Booriyalloak Homestead (37 44ʹ25.2ʺS; 143 19ʹ49.9ʺE; 204 m el.).
Petrographic inspection reveals plagioclase (laths up to 4 mm) and olivine phenocrysts set in a finely crystalline groundmass of interlocking plagioclase, clinopyroxene, olivine, magnetite and apatite ( Figure 6); olivine is commonly altered to iddingsite, but plagioclase is unaltered. Late-stage feldspathic pools contain apatite and magnetite. Irving and Green (1976) classified the Widderin lava as an olivine tholeiite transitional to olivine basalt and reported a K 2 O content of 1.09 wt%.

Sample preparation and irradiation
Mount Widderin sample NVP26 was crushed into gravel-sized fragments using a steel jaw crusher. Individual chips were then screened for alteration or large vesicles, and acceptable chips were crushed manually using a steel mortar. Crushed material was washed to remove dust and sieved to 180À250 mm. Following initial concentration by magnetic separation, approximately 500 mg of groundmass material was hand-picked under a binocular microscope. Grains were cleaned ultrasonically using the following procedure: 5% HNO 3 (10 min), 2% HF (1 min), deionised water (10 min) and acetone (5 min). The sample was then loaded into an aluminium foil packet, placed in a quartz tube (UM#61), bracketed by packets containing flux monitor standard Alder Creek Rhyolite sanidine [1.1811 § 0.0011 Ma (2s); Phillips, Matchan, Honda, & Kuiper, 2016]. Can UM#61 was irradiated for 25 min in the Cd-shielded CLICIT facility of the Oregon State University TRIGA reactor.

Analytical procedures
Individual groundmass aliquants of either »70 or 100 mg were loaded into a custom-made copper sample holder, covered with a ZnS glass disc and loaded into the sample chamber of a gas-handling system connected to a new-generation multi-collector Thermo Fisher ARGUSVI mass spectrometer. This system has been described in detail by Phillips and Matchan (2013), but the line has since been modified to include two additional SAES ZrÀAl getters for improved clean-up of gas released by hydrous phases. Following preliminary low-temperature out-gassing to remove the bulk of atmospheric argon, groundmass aliquants were incrementally heated in five to seven steps over the range of 4À30% laser power using a Photon Machines CO 2 laser following procedures described by Matchan and Phillips (2014).
Argon isotopic results (Supplementary Paper Table A1) are corrected for system blanks, mass discrimination, radioactive decay and reactor-induced interference reactions. Isotopic results that exclude the interference corrections (as per the recommendation of Renne et al., 2009) are presented in the electronic appendix, alongside blank correction values (Supplementary Paper Table A2). Line blanks were measured after every third or fourth sample analysis. Mass  discrimination and detector bias were characterised via automated analysis of air pipette aliquots prior to the first analysis. Owing to the short irradiation time, it was not feasible to include Ca/K/Cl salts/glasses in the same package. Therefore, correction factors determined for K-glass and Ca-salts contained in another recent package irradiated in the CLICIT facility (UM#58) were used: ( 36 Ar/ 37 Ar) Ca D (2.5713 § 0.0023) £ 10 À4 ; ( 39 Ar/ 37 Ar) Ca D (6.6200 § 0.0801) £ 10 À4 ; ( 40 Ar/ 39 Ar) K D (1.00 § 0.05) £ 10 À10 ; ( 38 Ar/ 39 Ar) K D (1.2136 § 0.0016) £ 10 À2 . The J-value for sample NVP26 [0.0001070135 § 0.0000000648 (0.061%; 1s)] was calculated based on an age of 1.1811 § 0.0011 Ma (2s) for AC sanidine (Phillips et al. 2016) using the decay constants of Steiger and J€ ager (1977) and the Lee et al. (2006) atmospheric argon composition [ 40 Ar/ 36 Ar D 298.56 § 0.62 (2s)]. Calculated uncertainties associated with weighted mean and plateau ages include uncertainties in the J-value, but exclude errors associated with the age of the flux monitor and the decay constant. Unless otherwise stated, uncertainties are reported at the 2s or 95% confidence level. Plateau ages are defined as including >50% of the total 39 Ar, from at least three contiguous steps, with 40 Ar Ã / 39 Ar ratios within error of the mean at the 95% confidence level (e.g. McDougall & Harrison, 1999). 40 Ar/ 39 Ar age results 40 Ar/ 39 Ar age results are summarised in Table 1 and Figure 7, with the full analytical dataset reported as supplementary information (Supplementary Papers Table A1).
Step-heating spectra and isochron plots were generated using ISOPLOT/ Ex v.3.75 (Ludwig, 2012). Errors associated with individual apparent ages are 0.5À2% (2s). The age spectra for five aliquants show a general trend of decreasing apparent ages with increasing temperature, with calculated ages ranging from 403À394 ka for the initial step to 375À365 ka at fusion (Figure 7a; Supplementary Papers Figure A1). The exception is aliquant NVP26-2, which shows an initial decrease in apparent age from 402.6 § 3.6 ka to 388.1 § 3.8 ka, similar to the other step-heating experiments, before an increase towards 442 § 17 ka at fusion. Radiogenic 40 Ar ( 40 Ar Ã ) yields were typically 35À45% of total 40 Ar measured in each experiment. The bulk of the 39 Ar (90À95%) was released during the low-to mid-temperature heating steps (4À14% laser power).
Plateau ages were calculated for three of the five aliquants: 391.1 § 2.4 ka (NVP26-1), 394.4 § 1.7 ka (NVP26-2), 387.1 § 1.9 ka (NVP26-3). Calculated total-gas ages are identical, and indistinguishable from corresponding plateau ages at the 2s level (Table 1). Inverse isochron data reveal a high degree of scatter and yield an apparent trapped 40 Ar/ 36 Ar ratio of subatmospheric composition [ 40 Ar/ 36 Ar i D 295.6 § 1.6 (95% CI; MSWD D 3.6); Figure 7b; Table 1]. Inverse isochron ages tend to be slightly older than corresponding plateau and total-gas ages but, owing to relatively large associated errors, are not significantly different at the 2s level (Table 1). All uncertainties include uncertainty in the J-value and are reported at the 2s level unless otherwise stated.

Age of Mount Widderin basalt
Step-heating results were analogous for four of five aliquants, reflecting a homogenous groundmass composition. The exception to this was aliquant NVP26-2, where anomalously old apparent ages determined for the high temperature steps likely reflect the release of excess argon from inclusions in phenocrysts out-gassed at high temperature (e.g. olivine/ pyroxene). Aliquant NVP26-2 is therefore excluded from subsequent discussion. The monotonic decrease in apparent ages from low to high temperature observed for all other step-heating experiments is characteristic of fine-grained samples that have experienced irradiation-induced recoil of 39 Ar and 37 Ar between phases with differing K-and Ca-contents (e.g. Jourdan & Renne, 2014;Koppers, Staudigel, & Wijbrans, 2000). This isotopic disturbance is reflected by the scatter of data in argon three-isotope space ( Figure 7b; Table 1, MSWD values typically >>1) that precludes calculation of robust inverse isochron results.
Irradiation-induced argon recoil and its implications for the 40 Ar/ 39 Ar dating technique has long been recognised (e.g. Huneke & Smith, 1976;Onstott, Miller, Ewing, Arnold, & Walsh, 1995;Turner & Cadogan, 1974). The reaction that produces 39 Ar ( 39 K(n,p) 39 Ar) involves neutrons of sufficiently high energy that the daughter 39 Ar atom may be recoiled on the order of 0.1À0.2 mm from the parent 39 K atom position (Onstott et al., 1995). The average recoil distance for 37 Ar is slightly longer (»0.5 mm) owing to the greater activation energy required for the 40 Ca(n,a) 37 Ar reaction to proceed (Onstott et al., 1995). Although these distances are small, the recoil loss/redistribution of 39 Ar K and 37 Ar Ca in fine-grained samples can have a significant effect on 40 Ar/ 39 Ar apparent ages, especially in the case of Ca-rich samples (e.g. Jourdan & Renne, 2014;Koppers et al., 2000).
The impact of irradiation-induced recoil on argon isotope systematics in basalt groundmass has been studied in detail by Koppers et al. (2000). The redistribution of 39 Ar from relatively K-rich phases (e.g. plagioclase/feldspathic mesostatis), which outgas at low temperatures, into K-poor Ca-rich phases (e.g. clinopyroxene), which outgas at higher temperatures, results in lower 39 Ar/ 40 Ar ratios for low-temperature steps and concomitantly higher 39 Ar/ 40 Ar ratios for high-temperature steps (Koppers et al., 2000). Similarly, the recoil redistribution of 37 Ar from Ca-rich to Ca-poor sites results in elevated 39 Ar/ 40 Ar and 36 Ar/ 40 Ar ratios owing to amplified interference corrections (Koppers et al., 2000).
Taking the simple case of a sample with a uniform trapped component, these competing effects essentially serve to decrease concordancy between data for high and low-temperature steps, affecting calculation of 40 Ar/ 36 Ar and 40 Ar Ã / 39 Ar isochron intercept values. It is noted that subatmospheric 40 Ar/ 36 Ar values have been reported for some basalts and attributed to isotopic mass fractionation during lava emplacement (e.g. Matsumoto & Kobayashi, 1995;Ozawa, Tagami, & Kamata, 2006). However, in the case of NVP26, the shape of the age spectrum, together with the degree of scatter apparent in argon three-isotope space, implicates recoilinduced isotopic disturbance, such that the absolute ( 40 Ar/ 36 Ar) i values calculated by inverse isochron analysis are unreliable. Importantly, the inverse isochron data show no evidence for excess argon aside from NVP26-2, such that assumption of a trapped atmospheric composition is valid for calculation of total gas ages.
Assuming that all 39 Ar and 37 Ar have been retained in the sample, but redistributed between phases (i.e. no net loss), the total gas age should approximate the groundmass crystallisation age (e.g. Jourdan & Renne, 2014). We note that if a slightly subatmospheric ( 40 Ar/ 36 Ar) i ratio of 295 is assumed, this results in a decrease in apparent ages of 1À3% for the low-to mid-temperature steps ( 14% laser power, containing 90À95% of the 39 Ar released) and 5À10% for the highertemperature steps. Excluding aliquant NVP26-2, a weighted mean total-gas age of 389.0 § 2.1 ka (0.5% 2s; MSWD D 1.1) is calculated. Given the uncertainty surrounding the true composition of the trapped component, which may be slightly subatmospheric, we propose that the error associated with the total-gas age be expanded to 2% to accommodate this uncertainty. Therefore, we propose an age of 389 § 8 ka (2%; 2s) for the Widderin basalt.

Implications for the timing of other volcanism in the region
The Widderin basalt was previously assigned to the Eccles RLU (Joyce, 1999 (Joyce, 1999), which range from ca 60 to 30 ka (Gillen et al., 2010;Stone et al., 1997). The stony rises from Mount Elephant appear better preserved than those in the Widderin lava landform, suggesting Mount Elephant significantly younger. Therefore, the Widderin basalt is interpreted to have dammed the northward flowing Elephant basalt, consistent with a topographic crosssection constructed along the WidderinÀElephant basalt outcrop (Supplementary Papers Figure B1). On the basis of field observations, the eruption age of Mount Elephant can be loosely constrained to 390À40 ka (i.e. older than Mount Eccles and younger than Mount Widderin lavas), but future direct 40 Ar/ 39 Ar dating of ejecta is warranted to better constrain the eruption age of Mount Elephant lavas.
As mentioned earlier, proposed peaks in volcanic activity in the NVP (e.g. Gray & McDougall, 2009) may be an artefact of relatively low sampling density and/or bias owing to consideration of only lava-producing eruptions. We propose that if the eruption histories of specific regions are considered in detail, with attention given to numbers of eruptive units and provenance, together with stratigraphic constraints (e.g. Hare et al., 2005), this may improve understanding of eruption frequency across the province as a whole. We establish a relative age framework for flows in the HamiltonÀ SkiptonÀDerrinallum area (Figure 8) using the flow stratigraphy summarised above, referenced to the 40 Ar/ 39 Ar age of 389 § 8 ka for Mount Widderin reported here, and the KÀAr ages of 0.99 § 0.04 Ma and 2.95 § 0.16 Ma reported previously for Stockyard Hill and the Terrinallum flow, respectively (Gray & McDougall, 2009). It is readily apparent that in this relatively small area volcanism has been sporadic since 3 Ma to 0.389 Ma. This duration of intermittent activity rivals that of the Melbourne area where basalt ages span 4.6À0.7 Ma (Gray & McDougall, 2009).

Eruption frequency in the Western Plains subprovince
The age compilation of Gray and McDougall (2009) (Gray & McDougall, 2009;Matchan & Phillips, 2011McDougall & Gill, 1975), with the remainder assigned ages of <0.1 Ma, owing to preservation of lava landforms, an interpretation supported by cosmogenic nuclide exposure dating studies (Gillen et al., 2010;Stone et al., 1997). Interpreting eruption frequency information solely from histograms (e.g. Gray & McDougall, 2009) is not ideal, as slightly altering bin values or adding a small number of data points can dramatically affect the apparent age distribution, especially if the number of data points in each bin is low. We suggest that a more meaningful approach may involve consideration of kernel density estimation (KDE, e.g. Vermeesch, 2012) in conjunction with a corresponding probability density plot (PDP). A KDE gives essentially the same frequency summary information as a histogram, but is continuous and thus avoids the under-smoothing issues inherent to histograms (Vermeesch, 2012). However, like the histogram, the KDE does not take into account the precision of each datum, and may in fact over-smooth the true age distribution. In contrast, the PDP incorporates age uncertainties and is routinely used for evaluating unimodal geochronological data. However, interpreting a PDP constructed with data from multiple age populations must be done with caution as individual high-precision ages may obscure the true age distribution. Therefore, we propose that the KDE and PDP be considered in tandem to obtain an accurate sense of eruption frequency. Figure 9 shows the KDE and PDP for eruption age dataset reported by Gray and McDougall (2009) for lava flows in the Western Plains subprovince. From the KDE, a peak in activity from ca 3 to 1.4 Ma is apparent, consistent with the conclusion of Gray and McDougall (2009), with an apparent waning in activity after ca 0.9 Ma. However, the addition of a small number (n D 9) of published age constraints for maars and scoria cones, together with the new age for Mount Widderin, results in a significant change to the age distribution pattern (Figure 9). Within the limits of the updated dataset, eruption frequency appears to have increased since ca 0.5 Ma (Figure 9). However, the proportion of maars and scoria cones with published age constraints is small, and the age distribution may be affected by a bias from studies generally focusing on better-preserved, younger eruption centres. Regardless, we stress that consideration of ages for lava flows alone may give a skewed impression of true age distribution if the dominant eruption style has varied over time (i.e. from effusive to more explosive) as has been suggested in previous studies (e.g. Price, Gray, & Frey, 1997).

Use of volcanic chronology in Quaternary landscape evolution studies
There is scope to extend early ideas based on regolith and drainage mapping undertaken in the 1960s and 1970s (Gibbons & Gill, 1964;Ollier & Joyce, 1964) in the context of neo tectonism and volcanic chronology in the Western Plains.
Although the historical seismicity of western Victoria is relatively subdued (e.g. Clarke & McCue, 2003), there is evidence for significant neo tectonic activity (e.g. Joyce, 1975;Paine, Bennetts, Webb, & Morand, 2004;Sandiford, 2003). While most of this activity is considered to have occurred prior to the onset of NVP volcanism (e.g. Paine et al., 2004), there is geomorphic evidence in southwestern Victoria for »200 m of fault-related uplift of Pliocene shoreline deposits during the Late Neogene, with subsequent faulting between 2 and 1 Ma, as constrained by KÀAr dated lava flows (Sandiford, 2003). There is also evidence for several phases of vertical movement on the northÀsouth-trending Rowsley Scarp (near the Anakies), warping of basalts by monoclinal movement (e.g. Lovely Banks Monocline), and local uplift/faulting related to individual eruption points (Joyce, 1975). Therefore, refining age constraints for <2 Ma eruption centres may improve understanding of the timing of neo tectonic activity in western Victoria. Furthermore, as NVP basalts significantly modified regional drainage patterns (e.g. Raiber & Webb, 2008;Taylor & Gentle, 2002), determining well-constrained eruption ages for key flows may provide useful stratigraphic markers for changes in paleo-environment (e.g. Baker, 2008).