High-precision 40 Ar/ 39 Ar dating of pleistocene tuffs and temporal anchoring of the Matuyama-Brunhes boundary Quaternary Geochronology

High-precision 40 Ar/ 39 Ar ages for a series of proximal tuffs from the Toba super-volcano in Indonesia, and the Bishop Tuff and Lava Creek Tuff B in North America have been obtained. Core from Ocean Drilling Project Site 758 in the eastern equatorial Indian Ocean contains discrete tephra layers that we have geochemically correlated to the Young Toba Tuff (73.7 ± 0.3 ka), Middle


Introduction
The Earth's magnetic field alternates between periods of normal polarity, in which the mean polarity of the field was the same as the present, and reverse polarity, in which the polarity was the opposite. Reversals in geomagnetic field polarity have occurred episodically throughout much of geologic time. To the extent that these polarity reversals are globally synchronous they can be used as tick marks whose ages, when calibrated, are invaluable components of the Geological Time Scale. Two challenges limit the utility of geomagnetic polarity reversals as time stamps in the geologic record: (1) the unambiguous correlation of magnetic polarity records recovered from rocks or sediments with a global record of such events, and (2) the accuracy of age calibrations. In this paper we address both issues in the case of the reversal between the Matuyama and Brunhes geomagnetic polarity epochs, also known as the Matuyama-Brunhes boundary (MBB), whose age is of fundamental importance to many topics in the Earth Sciences yet has been controversial.
The MBB was the most recent full reversal of the Earth's magnetic field, and serves as a Global Boundary Stratotype Section and Point (GSSP), selected by the International Commission on Stratigraphy as a marker for the beginning of the Middle Pleistocene. An age of 780 ka for the MBB (Shackleton et al., 1990) was determined by orbital tuning of benthic and planktic v 18 O records from Ocean Drilling Program (ODP) Site 677 in the eastern equatorial Pacific (Fig. 1). The tuning was calibrated to an ice volume model (Imbrie and Imbrie, 1980), which was based on a series of orbital solutions (Berger and Loutre, 1988). The specific location of the MBB within the ODP 677 core was unknown and hence extrapolation of its location from Deep Sea Drilling Project Site 607 was required (Fig. 1). Other orbital tuning ages for the MBB range from 730 ka (Imbrie et al., 1984;Ruddiman et al., 1989) to 790 ka (Johnson, 1982).
An extremely precise orbitally-tuned age of 773 ± 1 ka was recently proposed for the MBB (Channell et al., 2010). Five North Atlantic records placed in isotope age models that were constructed by correlation of the v 18 O record directly or indirectly to an ice volume model were used to place the MBB consistently at the young end of Marine Isotope Stage (MIS) 19. The orbitally-tuned MBB age inferred by Channell et al. (2010) was stated to be consistent with an 40 Ar/ 39 Ar age (776 ± 2 ka, 1 sigma, analytical uncertainty only) (Coe et al., 2004), from Hawaiian lavas, but only if the age of Fish Canyon sanidine (FCs), a secondary mineral standard for the 40 Ar/ 39 Ar radio-isotopic dating system, is adjusted to an age of 27.93 Ma e an age known from numerous studies to be too young (e.g., Kuiper et al., 2008;Renne et al., 2011;Rivera et al., 2011;Wotzlaw et al., 2013;Morgan et al., 2014). Remarkably, Singer (2014) subsequently reanalysed the same Hawaiian lavas and obtained the identical result of 776 ± 1 ka (1 sigma, analytical uncertainty only) but using a different 40 Ar/ 39 Ar calibration (decay constants of Min et al., 2000; the FCs age of 28.201 Ma, Kuiper et al., 2008). The perfect congruence of these two ages is spurious, however, as applying the same calibration to both 40 Ar/ 39 Ar ages (Coe et al., 2004;Singer, 2014) indicates that they differ by 5 ± 3 ka.
Direct radio-isotopic ages have been determined for the MBB through 40 Ar/ 39 Ar dating of various lava flows with transitional directions or known relationships to the MBB. Baksi et al. (1992) dated lavas from Maui with transitional paleomagnetic directions related to the MBB to yield an 40 Ar/ 39 Ar age of 783 ± 11 ka (1 sigma, analytical uncertainty only) using the decay constants of Steiger and Jager (1977) and the SB3 biotite standard at 162.9 Ma, which is equivalent to Fish Canyon sanidine (FCs) at 27.5 Ma (Lanphere and Baadsgaard, 2001). Recalculated 1 this age is 795 ± 11/12 ka. Singer and Pringle (1996) determined an 40 Ar/ 39 Ar weighted mean age for 8 basaltic to andesitic lava flows inferred to have erupted during the MB-reversal from Chile, Tahiti, La Palma and Maui. They calculated an age (779 ± 2 ka, 1 sigma, analytical) using the decay constants of Steiger and Jager (1977) and the Taylor Creek sanidine (TCs) standard with an age of 27.9 Ma, which is also equivalent to FCs at 27.5 Ma. Recalculated the MBB age of Singer and Pringle (1996) is 791 ± 2/3 ka. Singer et al. (2005) incorporating the data of Coe et al. (2004) proposed that there were two age clusters for MBB-related lava flows: (1) 776 ± 2 ka (1 sigma, analytical uncertainty), and (2) 793 ± 3 ka (1 sigma, analytical uncertainty). These ages were calculated using the decay constants of Steiger and Jager (1977) and TCs with an age of 28.34 Ma, which is equivalent to an FCs age of 28.02 Ma (Renne et al., 1998). Recalculated these ages are 773 ± 2/3 ka and 790 ± 3/3 ka, respectively. It was proposed that the older age was related to an initial demise of the axial dipole, onset of geodynamo instability, and non-dipolar field behaviour e a precursor to reversal of field polarity. A MBB precursor event with low field intensity has been noted in some (Kent and Schneider, 1995;Hartl and Tauxe, 1996;Channell et al., 2009Channell et al., , 2010 but not all marine records (e.g., Suganuma et al., 2015). Importantly, there is no direct palaeomagnetic evidence linking the older age of Singer et al. (2005) to such a precursor event. As discussed (above), Singer (2014) made new measurements on old MBB-related samples (Coe et al., 2004), giving a recalculated age of 779 ± 1/1 ka.
The dating of silicic tuffs that straddle the MBB has vast potential for determination of accurate and precise event timings with robust, fully quantifiable uncertainties, especially if high-K phases such as sanidine/anorthoclase are present for 40 Ar/ 39 Ar analyses. The sanidine-and zircon-bearing Bishop Tuff (BT) deposited below the Lava Creek Tuff Member B (LCTB) but above the MBB, has been the focus of much interest. Briefly, Sarna-Wojcicki et al. (2000) calculated sedimentation rates in terrestrial sections throughout western North America making the simple (but geologically tenuous) assumption of constant sedimentation rate between the LCTB and the BT. Employing the inferred sedimentation rates, they calculated the duration between the BT and the MBB represented by the intervening sediment. These results, that are independent of which calibration of the 40 Ar/ 39 Ar system is used, imply that the MBB is 15 ± 2 ka older than the BT if the assumption of uniform sedimentation is correct. Unfortunately, the age of the Bishop Tuff has remained controversial and hence its reliability for determination of an age for the MBB has been questioned (e.g., Channell et al., 2010).
The age, character and tempo of the MBB was clarified by recent high-precision 40 Ar/ 39 Ar dating of sanidine from tephra layers that bracket the boundary within the Sulmona Basin paleolake in Central Italy (Sagnotti et al., 2014;Giaccio et al., 2015). The lacustrine sediments within which the tuffs are intercalated are characterized by biogenic magnetite and were sampled at high resolution, allowing the reconstruction of the MBB in very fine detail. The Sulmona results show that the MBB is significantly older than 773 ± 1 ka (Channell et al., 2010) with a recalculated age of 783 ± 1/ 1 ka. Using the same parameters but the weighted mean astronomical ACs age rather than the Optimisation Model ACs age (Niespolo et al., 2016), the Sulmona data yield a MBB age of 780 ± 1/ 1 ka (Sagnotti et al., 2014), both ages resolvably older (10 ± 2 ka and 7 ± 2 ka, respectively) than 773 ± 1 ka.
Most recently U-Pb SHRIMP-II ages have been determined for a tephra associated with the MBB in Japan (Suganuma et al., 2015). However, primarily for reasons highlighted by Ickert et al. (2015) and discussed by us in detail below, we do not consider either the U/Pb data accurate or appropriate for use in determining an age for the MBB (i.e., poor characterisation of Th/U melt from which the zircons grew). Further, Suganuma et al. (2015) failed to disequilibrium-correct their Tera-Wasserburg relations, the oxygen isotope data are of inadequately low resolution to precisely define the boundaries of MIS 19, and there are no quantitative constraints for the sedimentation rate.
Owing to large degrees of scatter in the data that attempt to temporally anchor the MBB there has been the suggestion that the reversal was diachronous on a global-scale (e.g., Rivera et al., 2011). Until recently the MB-reversal, as with other geomagnetic polarity reversals, has been considered isochronous, certainly at the relatively poor levels of temporal resolution (accuracy and precision) attained using the 40 Ar/ 39 Ar geochronometer between 1990 and 2010. However, computer modeling of the MB-reversal has highlighted potential for a millennial-scale (±1 ka) offset in the onset of the polarity reversal for sites in the Atlantic and Pacific Oceans (Leonhardt and Fabian, 2007). The study also proposed reversal durations of between 2 and 10 ka, which exceeds the MBB duration recorded by the high-resolution record at Sulmona by more than an order of magnitude (Sagnotti et al., 2016).

Ocean Drilling Project (ODP) Leg 758
Ocean Drilling Project (ODP) Site 758 resides on the crest of Ninetyeast Ridge (5 23.05 0 N, 90 21.67 0 E) in a water depth of 2924 m (Fig. 1). Three holes were cored at Site 758 (A, B, C). Within the timeframe of interest (Holocene-Pleistocene) stratigraphic analysis showed good recovery (Shipboard Scientific Party, 1989). Due to the possibility of gaps occurring between successive Advanced Piston Corer (APC) core sections, sections in ODP 758A and ODP 758B were staggered in depth relative to each other. It was possible to provide high-resolution between-hole correlation by using a combination of paleomagnetic remanence, magnetic susceptibility and distinct lithological and tephra markers (Shipboard Scientific Party, 1989;Dehn et al., 1991;Farrell and Janecek, 1991;Gee et al., 1991).
There were two defined scientific aims for drilling Site 758: (1) to study the tephrochronology of the Indonesian volcanic arc relative to a changing climate signal, and (2) to study in detail the behaviour of Earth's magnetic field during polarity transitions. As such the ODP 758 deep sea core contains: (1) records of distal tephras (volcanic ash layers) from the Indonesian volcanic arc above and below the MBB (Dehn et al., 1991), (2) cm-scale resolution v 18 O records from benthic and planktic foraminifera (Farrell and Janecek, 1991;Chen et al., 1995), and (3) a detailed paleomagnetic stratigraphy that shows the precise and well-defined location of the MBB, as well as the onset and termination of the Jaramillo Geomagnetic Excursion (JGE) (Shipboard Scientific Party, 1989;Gee et al., 1991).

ODP 758 tephrochronology
The tephra layers documented within Site 758 provide a unique record of explosive volcanism for the North Indian Ocean. Over 200 visible tephra (ash) layers have been documented from the site ranging in thickness from millimetres to decimetres. Many of the tephras are present in one hole but not in the neighbouring holes. The tephras also display variable forms, ranging from discrete to diffuse layers and patches/pods of ash. The local absence and variable physical characteristics of a given tephra is due to the variety of depositional processes operating on the crest of the Ninetyeast Ridge (Dehn et al., 1991) and hence not unexpected (Carey, 1997). Given the distance (Fig. 1) between Site 758 and the Indonesian volcanic arc (the most proximal and therefore probable volcanic source) the tephra found in the three holes are distal in nature, finegrained and dominated by glass shards rather than mineral grains. As such the tephras are unsuitable for direct 40 Ar/ 39 Ar dating due to paucity of required mineral phases.
Two of the distal tephras located in ODP 758 have been robustly correlated using glass and mineral chemistry to the Young Toba Tuff (YTT, Ash A, c. 74 ka; Mark et al., 2014;Storey et al., 2012) and Middle Toba Tuff (MTT, Ash C, c. 500 ka;  eruptions of the Toba super-volcano (Dehn et al., 1991) (Fig. 1). Older ash units (E, d and D) have been the focus of much attention with the question raised as to which, if any, correlates with the oldest super-eruption (Old Toba Tuff, OTT, c. 800 ka;  of Toba (Shane et al., 1995). Lee et al. (2004) linked Ash D with OTT but this correlation was subsequently questioned Shane et al., 2004). Further geochemical work is required to confidently link these ashes to proximal deposits, and to determine whether or not their source was indeed Toba. Note, the age of Ash D was calibrated by astronomically tuned oxygen isotope stratigraphy to 788.0 ± 2.2 ka .

Paleomagnetic data of ODP 758
In addition to containing discrete tephra layers, ODP 758 contains records of geomagnetic reversals and excursions e the relationships between the tephra and geomagnetic are quantifiable. For the purpose of this discussion/study we have focussed on the composite ODP 758 record (ODP 758A, 758B, 758C) as presented by Farrell and Janecek (1991) (Fig. 2). By constructing a composite depth section from Holes 758A and 758B it was possible to splice across recovery gaps with the result being an undisturbed, continuous sedimentary section that extends from 0 to 116 mbsf, which is equivalent to the past c. 7 Ma (well beyond the time interval of interest for this study). The continuity of this composite section was checked with several independent stratigraphies (e.g., Farrell and Janecek, 1991). The paleomagnetic data for ODP 758 are not ideal in that they were not collected from discrete samples; the U-channel and on-board measurements have in all probability smoothed the paleomagnetic signal, but the core still retains clear transition zones and paleomagnetic directional changes (Shipboard Scientific Party, 1989;Dehn et al., 1991;Farrell and Janecek, 1991;Gee et al., 1991).
1.5. The location of the MBB at site 758 Chen et al. (1995) determined an astronomical age of c. 784 ka for the MBB in ODP 758. Although similar to the MBB age of Shackleton et al. (1990) (780 ka) there is no uncertainty associated with these ages and as such, it is unclear whether the offset is 'real' (i.e., geological or methodological). However, it has been suggested that any astronomically tuned age for this period of time cannot be defined better than to within ± 5 ka due to uncertainties in the phase relationship between insolation and climate (Martinson et al., 1987;Imbrie and Imbrie, 1980). The tuning approach of Chen et al. (1995) did vary from that of Shackleton et al. (1990), whereas the latter tuned their v 18 O data to the model of Imbrie and Imbrie (1980) from 0 to 1.6 Ma, and to the obliquity cycles from 1.6 to 2.6 Ma, Chen et al. (1995) tuned their entire record to the ice volume simulation based on the model of Imbrie and Imbrie (1980). This approach allowed for a fine tuning approach of both the 41 and 23 ka cycles simultaneously.
The ODP 758 v 18 O data (also a composite of v 18 O measurements from ODP 758A and ODP 758B; Farrell and Janecek, 1991;Chen et al., 1995) show the location of the MBB in the early part of Marine Isotope Stage (MIS) 19 (Fig. 2), earlier than the position of the MBB as noted by Channell et al. (2010) in the North Atlantic. This is not uncommon; the MBB has been identified in the middle of MIS 19 (Suganuma et al., 2015), at the start of MIS 19 (Horng et al., 2002) and even within MIS 20 (Langereis et al., 1997), although the latter was attributed to delayed acquisition of the Earth's magnetic signal in the sediment (i.e., lock-in of paleomagnetic remanence) (Kent, 1973). As a consequence, two important questions emerge: (1) should we expect the MBB to be globally located at the exact same location within a MIS? And (2) if so, why do we see such variation in the location of the MBB in different v 18 O records? We consider these to be questions of temporal and spatial resolution within records -we revisit this discussion below.
1.6. The Jaramillo geomagnetic excursion at site 758 The Jaramillo geomagnetic excursion (JGE) was a reversal of the geomagnetic field that occurred c. 1000 ka (Singer, 2014;Kissel et al., 2014) and is also preserved within ODP 758. It was a shortterm reversal in the Matuyama reversed magnetic chronozone. Within ODP 758 the JGE onset is dated at c. 1070 ka and the JGEtermination at c. 997 ka (Chen et al., 1995). The Toba tephra are positioned above the JGE within ODP 758.

Australasian microtektites
Australasian microtektites have been found below the MBB within sediment cores from throughout the Indian Ocean, western equatorial Pacific Ocean, Philippine, Sulu and Celebes Seas, and most recently the South China Sea (Hyodo et al., 2011 and references within). The Australasian tektites have previously been dated by 40 Ar/ 39 Ar at 799.2 ± 3.4/3.8 ka (Smit et al., 1991) with the data of Yamei (2000) reproducing this age but suggesting the presence of excess 40 Ar (noted from isochron analysis of the data, Yamei 2000). As such, the current 40 Ar/ 39 Ar age for the Australasian microtektites should be considered as a maximum age constraint.
Within Site 758 the peak abundance of Australasian microtektites occurs 8 cm below a tephra horizon labelled as Ash 'D' and immediately prior to Termination IX . We know from other Pacific and Indian Ocean records that the Australasian Tektite peak concentration is located immediately prior to Termination IX (e.g., Glass and Koeberl, 2006;Valet et al., 2014), the transition from Marine Isotope Stage (MIS) 19e20. Despite a large degree of dispersion throughout the core, the Australasian Tektites main concentration peak is in the correct stratigraphic position relative to other cores from throughout the region.

ODP 758 and LR04
LR04 is a 5300 ka stack of v 18 O records from 57 globally distributed sites that have been aligned using an automated graphic correlation algorithm (Lisiecki and Raymo, 2005). This was the first Pliocene-Pleistocene stack to contain more than three records that extend back beyond 850 ka. The LR04 stack contains the composite ODP 758 core data. As an automated graphic correlation algorithm was used to construct the stack, its stratigraphic features are therefore independent of any time scale. An age model was subsequently constructed by aligning the benthic v 18 O stack to a simple model of ice volume whilst taking into consideration the average stacked sedimentation rate of the individual sediment cores. The LR04 stack places the MBB at c. 780 ka (following Shackleton et al., 1990).

Study scope
Using a combination of tephrochronology and high-precision 40 Ar/ 39 Ar geochronology we aim to examine the temporal relationship between the Toba tephra layers preserved in ODP 758 to primarily constrain the age of the MBB, the Australasian Tektites and Termination IX within the Indian Ocean. Furthermore, we examine the relationship of the MBB to the Bishop Tuff and Lava Creek Tuff B (LCTB) in North America. Our findings agree perfectly with those of Sagnotti et al. (2014) and further question the accuracy of the Channell et al. (2010) MBB age estimation, unless an 'extremely' diachronous reversal is invoked, an event that our data discount.

Field relations, proximal tuffs: Sumatra
Fieldwork was conducted on Sumatra to directly sample the relevant proximal (crystal-rich) Toba deposits. Samples were collected from roadside exposures from Siguragura and to the north of Haranggoal. Sample locations shown in Fig. 3 relative to the accepted stratigraphy of those sequences . Correlation of proximal deposits across the caldera is extremely difficult because of poor exposure and abundant rainforest vegetation, but has been attempted previously by Knight et al. (1986) and . Toba eruptive units are intermittently exposed along the road that climbs the caldera wall to the northwest of Haranggoal. Andesitic lavas, dated at c. 1.3 Ma , are exposed at the present lake level. The c. 1.2 Ma Haranggoal dacite tuff (HDT) overlies the andesitic lava flows . The HDT is a brown, densely welded and often-jointed (radial and columnar) ignimbrite with large, lightly coloured, flattened pumices that reach 1 m in length  and is exposed in the roadside to elvations of~1060 m. Above the HDT are three further eruptive units (see Fig. 3). At elevations above about~1250 m near Haranggoal, the YTT caps the eruptive sequence, as it does for almost all of the area surrounding Lake Toba. Below this,   40 Ar/ 39 Ar dated a middle unit, 140 m-thick, to 501 ± 5 ka, noting that this had a normal paleomagnetic polarity, and ascribed this unit to the MTT. The light gray upper parts of the MTT (containing generally black and glassy fiamme), are underlain by a dark gray to black, often columnarjointed vitrophyric tuff (up to 100 m thick) also normally magnetised (Knight et al., 1986). This vitrophyre has been described as the basal component of MTT by . The vitrophyric unit underlying the light gray MTT does not resemble OTT deposits reported from elsewhere around the lake (e.g. Siguragura, see below) and for this reason previous authors have assumed that OTT is absent from this part of the caldera wall near Haranggoal. We now know that the dark vitrophyric underlying unit is not part of the MTT sequence, but is an older eruption product (discussed below).
3.2. Siguragura (2 31 0 16.7 00 N, 99 16 0 29.2 00 E) Chesner and Rose (1991) described the Oldest Toba Tuff at Siguragura. The OTT is well exposed on the Uluan block and along the course of the Asahan River, SE of the lake, where it is a very thick (> 300 m in most locations) densely welded ignimbrite that is brown to light grey in colour. On the Uluan block OTT it is often columnar jointed and numerous flow units can be recognised, and contains fiamme (up to 30 cm, . The OTT sample used in this study was collected from Siguragura, in the Asahan River valley, and material from this area had been previously 40 Ar/ 39 Ar dated by Diehl (1987) (c. 840 ka) and noted to have a reversed paleomagnetic polarity.

Lava Creek Tuff B
Similarly, the Lava Creek Tuff B (LCTB) has been described in detail by many other studies (e.g., Wotzlaw et al., 2015) and as such, details of just the sampling site are described. The LCTB ignimbrite was sampled from the location described by Christiansen (2001) at the quarry near the east end of the dam at Grassy Lake Reservoir, just south of Yellowstone National Park (N44 13.074 0 , W110 81.417 0 ). The sample was from the relatively crystal-rich densely welded basal vitrophyre of the ignimbrite. Bulk composition of this sample is identical to compositions of LCTB reported by Christiansen (2001).

Electron microprobe glass and biotite geochemistry
Major element compositions of glass and biotite from both the proximal Toba deposits (YTT, MTT, OTT) and distal deposits (Ash layers A, C, d, D, E) were determined using a wavelength-dispersive JEOL 8600 electron microprobe (EMP) at the Research Laboratory for Archaeology and the History of Art, University of Oxford. The instrument was calibrated at 15 kV using a range of mineral standards. A low beam current (6 nA), and defocused (10 mm) beam were used to analyse individual glass shards. Single biotite crystals were analysed with a beam current of 15 nA, and a 5 mm beam. Peak counting times were 30 s for all elements except Na (10 s in glass and 20 s in biotite). The EMP calibration was verified using a range of reference glasses from the Max Planck Institut (Jochum et al., 2006) and minerals from the Smithsonian (Jarosewich et al., 1980). Totals of glass analyses were mostly >95% and normalized to 100% to account for variable secondary hydration. Biotite analytical totals were typically >92 wt.%. All raw analyses of the glass and biotite, and the reference materials are included in appendix file 'SF#1' (.pdf).

40 Ar/ 39 Ar dating
A detailed sample preparation routine is discussed by Mark et al. (2010) but briefly: feldspars (sanidine) were separated after disaggregating, washing and sieving followed by magnetic and density separations and finally ultrasonic cleaning in 5% hydrofluoric acid for 5 min. Feldspars were handpicked under binocular microscope for analysis. Samples were irradiated in the CLICIT facility of the Oregon State University TRIGA reactor using the Alder Creek sanidine (Nomade et al., 2005) as a neutron fluence monitor. 40 Ar/ 39 Ar analyses were conducted at the NERC Argon Isotope Facility, Scottish Universities Environmental Research Centre (SUERC) and the Berkeley Geochronology Center (BGC). Samples analyzed at BGC were run and reported blindly, without knowledge of the SUERC results (and vice versa). Details of irradiation durations, J measurements, discrimination corrections are provided in appendix file SF#3b (.pdf). Irradiation correction parameters are shown below. For J determinations three bracketing standard positions surrounding each unknown were used to monitor the neutron fluence. Ten measurements were made for each bracketing standard position. The weighted average 40 Ar*/ 39 Ar K was calculated for each well, and the arithmetic mean and standard deviation of these three values was used to characterize the neutron fluence for the unknowns. This approach was deemed sufficient as, due to the relatively short irradiation durations, there was no significant variation between the three positions in a single level of the irradiation holder. This also facilitated high-precision measurement of the J-parameter. Note that for all J-measurements no data were rejected. Samples were analyzed in several batches; backgrounds and mass discrimination measurements (via automated analysis of multiple air pipettes) specific to each batch are summarized in appendix file 'SF#1' (.pdf). Air pipettes were run (on average) after every 5 analyses. Backgrounds subtracted from ion beam measurements were arithmetic averages and standard deviations. Mass discrimination was computed based on a power law relationship (Renne et al., 2009) using the isotopic composition of atmospheric Ar reported (Lee et al., 2006) that has been independently confirmed . Corrections for radioactive decay of 39 Ar and 37 Ar were made using the decay constants reported by Stoenner et al. (1965) and Renne and Norman (2001), respectively. Ingrowth of 36 Ar from decay of 36 Cl was corrected using the 36 Cl/ 38 Cl production ratio and methods of Renne et al. (2008) and was determined to be negligible. Argon isotope data corrected for backgrounds, mass discrimination, and radioactive decay and ingrowth are given in the appendix file 'SF#1' (.pdf).
At SUERC the samples were analyzed by total fusion and stepheating with a CO 2 laser and measurements made using a MAP 215-50 (MAP2) noble gas mass spectrometer. The mass spectrometer is equipped with a Nier-type ion source and analogue electron multiplier detector. Mass spectrometry utilized peak-hopping by magnetic field switching on a single detector in 10 cycles.
At BGC the samples were analyzed by total fusion with CO 2 lasers on two different extraction systems mated to MAP 215 mass spectrometers (MAP1 and MAP3). MAP1 is a 215C and MAP3 is a 215-50. Both have Nier-type ion sources and analog electron multiplier detectors. Mass spectrometry utilized peak-hopping by magnetic field switching on a single detector in 10e15 cycles.
For some of the age comparisons made herein, contributions from sources of systematic uncertainty (i.e., uncertainties in 40 Ar/ 4 K of the standard and 4 K decay constants) are neglected and only analytical uncertainties in isotope measurements of samples and standards are included. These uncertainties are referred to herein as "analytical precision". For the purposes of this study analytical uncertainties include contributions from uncertainties in the interference corrections because these interference corrections have variable effects due to the slight variable chemistry of the samples considered. Where not otherwise distinguished, uncertainties are stated as X ± Y/Z, where Y is the analytical uncertainty as defined above, and Z is the full external precision considering both analytical and systematic sources of uncertainty (e.g., decay constant).
Age computation uses the weighted (by inverse variance) mean of 40 Ar*/ 39 Ar K values for the sample and standard, combined as R-values and computed using the method of Renne et al. (2010). Outliers in both single-crystal samples and standards were discriminated using a 3-sigma filter applied iteratively until all samples counted are within 3 standard deviations of the weighted mean ± one standard error. This procedure screened older crystals that are logically interpreted as xenocrysts. No younger outliers were recorded during analysis of all samples. Processing of the data using the nMAD approach of Kuiper et al. (2008) has no impact on the probability distribution plots for each sample.
The analytical approach adopted was to initially analyze at all times single crystals of sanidine by total fusion (SUERC & BGC). Following the initial analyses if no xenocrystic contamination was observed, samples of small crystal populations (n3) were stepheated (SUERC). The purpose of the step heating was to verify that initial trapped 40 Ar/ 36 Ar compositions overlap with accepted atmospheric values (Lee et al., 2006;Mark et al., 2011).

Fieldwork & tephra geochemistry
There are limited accessible outcrops around Lake Toba that preserve a full volcanic stratigraphy. Hence the eruption history of Toba has been pieced together from distal locations (e.g., marine cores) and a couple of non-correlated sites located proximal to the caldera. As discussed above, we sampled the deposits at two of these key localities, Haranggoal at Siguragura (Fig. 3), as the volcanic units preserved at these localities have been previously ascribed to the three different Toba eruptions (OTT, MTT and YTT). At Siguragura the OTT unit is unconformably overlain by the YTT. The YTT is also preserved at Harrangoal and is underlain by two units ascribed to the MTT, which lie above the Harrangoal Dacitic Tuff (HDT, Shane et al., 1995;Lee et al., 2004, Chesner and. Although the lowermost vitrophyric portion of the unit between YTT and HDT lies within the same stratigraphic position as OTT, it was assumed by  to be the MTT, as it appears texturally different from the OTT at Siguragura. Interestingly, within ODP 758 there are two tephras of the approximate age of the OTT, Ash d and Ash D. We hypothesized that the OTT deposits at Siguragurra and Harrangoal are products of two different eruptive events that occurred c. 800 ka, which correspond to Ash D (termed OTTB, Haranggoal) and Ash d (termed OTTA, Siguragura) in the distal ODP 758 core. Our geochemical data on both glass and biotite (Fig. 4) show that despite a different appearance, this lower presumed MTT unit at Haranggoal is compositionally indistinguishable from the other OTT proximal deposits and that indeed Ash d and Ash D correspond to two different eruptions that occurred about the same time as 'the OTT' (hence labelling OTTA and OTTB). Based upon the previously published estimated sedimentation rate in ODP 758 of 1.7 cm/ka (Farrell and Janecek, 1991), there would be c. 6 ka between Ash d and Ash D.

40 Ar/ 39 Ar dating results
All of our new 40 Ar/ 39 Ar data are presented in Figs. 5e8. Pooled ages for the two samples dated by both SUERC and BGC are calculated from the weighted mean R-values (Renne et al., 1998) corresponding to the single crystal total fusion measurements for each sample. All data are presented at the 1 sigma confidence level and are summarized in Table 1.
OTTA total fusion (pooled) : SUERC analysed 61 crystals and BGC analysed 40 crystals (sample NP1). One data point was rejected due to the analysis of a plagioclase grain rather than a sanidine crystal. Data define a weighted mean pooled age of 792.5 ± 0.5/0.6 ka (Fig. 5).
OTTA step-heating : SUERC performed four step-heating experiments on four different aliquots of sanidine crystal populations (n3) (sample NP1). All experiments yielded 100% 39 Ar plateaux with initial 40 Ar/ 36 Ar trapped components that overlapped with atmospheric values (Lee et al., 2006). The weighted mean plateau age (791.9 ± 1.0/1.1 ka) and inverse isochron age ( Fig. 6) are indistinguishable at the 1 sigma level from the total fusion weighted mean age. OTTB total fusion : SUERC analysed 62 crystals of OTTB (sample NP9). Data define a weighted mean age of 785.7 ± 0.5/0.7 ka (Fig. 5). No data points were rejected from the age calculation.
OTTB step-heating : SUERC performed four step-heating experiments on four different aliquots of sanidine crystal populations (n3) (sample NP9). All experiments yielded 100% 39 Ar plateaux with initial 40 Ar/ 36 Ar trapped components that overlapped with atmospheric values (Lee et al., 2006). The weighted mean plateau age (785.3 ± 0.8/1.0 ka) and inverse isochron age (Fig. 6) are indistinguishable at the 1 sigma level from the total fusion weighted mean age.
MTT total fusion : SUERC analysed 83 crystals of MTT sanidine (sample NP2). The data show a bimodal distribution with an older age population of c. 800 ka (n33) and a juvenile population defining an age of 502.0 ± 0.6/0.7 ka (n50) (Fig. 5). We consider the statistically significant juvenile population to define the age of the MTT eruption. Owing to the presence of xenocrysts we did not perform  incremental step-heating on small crystal populations. Note the xenocryst age population is consistent with the approximate age of the OTT.
BT total fusion : SUERC analysed 225 crystals of BT sanidine using a single mass spectrometer and BGC analysed 94 crystals using two different mass spectrometers with the samples irradiated in two separate batches. The data yielded a weighted pooled mean of 766.8 ± 0.4/0.6 ka (n319), with no data points rejected (Fig. 7).
BT step-heating : SUERC performed eight step-heating experiments on eight different aliquots of sanidine crystal populations (n3). All  experiments yielded 100% 39 Ar plateaux with initial 40 Ar/ 36 Ar trapped components that overlapped with atmospheric values (Lee et al., 2006). The weighted mean plateau age (766.1 ± 0.6/0.8 ka) and inverse isochron age (Fig. 7) are indistinguishable at the 1 sigma level from the total fusion weighted mean age.
LCTB total fusion : SUERC analysed 34 sanidine crystals of LCTB. The data show a dominantly bimodal distribution with an older age population of c. 700 ka (n5) and a single old crystal of c. 960 ka (n1). 28 sanidine crystals defined a juvenile age population with weighted mean age of 627.0 ± 1.5/1.7 ka (Fig. 8). We consider this juvenile age population to define the age of the LCTB eruption.
Owing to the presence of xenocrysts we did not perform incremental step-heating experiments.
The R-values and their corresponding ages (calculated weighted averages for samples that have both fusion age and step-heating age data) are shown in Table 1 and appendix file 'SF#1' (.pdf). It is these ages (Table 1) and associated R-values that are discussed throughout the remainder of the text.

40 Ar/ 39 Ar dating of the proximal Toba Tuffs
We performed 40 Ar/ 39 Ar dating on sanidine separated from the proximal YTT, MTT, OTTA (Siguragura) and OTTB (Harrangoal) that have been correlated geochemically to tephra layers preserved within the marine core record to construct a high-precision radioisotopic chronology for the Pleistocene of ODP 758 that is independent of the astronomical age model.
The 40 Ar/ 39 Ar age data for the YTT (Ash A) have been published previously and define a robust inverse isochron age, the data being reported relative to the ACs standard age of 1.2056 Ma (Renne et al., 2011). This age for the YTT was indistinguishable from the YTT age of Storey et al. (2012), relative to the same 40 Mark et al. (2013Mark et al. ( , 2014 and Storey et al. (2012). This yields the most accurate and precise age for the YTT, integrating the data from both laboratories to give an age of 73.7 ± 0.3/0.4 ka (R YTT ACs : 0.06196 ± 0.00025). Note the recalculated data is available in appendix file 'MBB data summary' (.pdf). The age of the MTT (502.0 ± 0.6/0.7 ka, R MTT ACs : 0.42219 ± 0.00050) is in agreement with the 40 Ar/ 39 Ar age of  allowing for differences in the 40 Ar/ 39 Ar calibration used. Our data for the OTT show that at c. 800 ka there were two temporally distinct eruptions from the Toba volcano: OTTA and OTTB, 792.4 ± 0.5/0.6 ka (R OTTA ACs : 0.6646 ± 0.00042) and 785.6 ± 0.7/0.8 ka (R OTTB ACs : 0.66075 ± 0.00059), respectively. There is a c. 6 ka temporal offset between these two eruptions, which broadly agrees with the temporal offset suggested by the application of the average ODP 758 sedimentation rate between Ash d and Ash D (discussed above). The age for Ash D is also indistinguishable from the astronomically tuned age presented by Lee et al. (2004) of 788.0 ± 2.2 ka. These two ages (our 40 Ar/ 39 Ar age and the astronomically calibrated age) are in good agreement with the previous OTT/Ash D 40 Ar/ 39 Ar age of Hall and Farrell (1995) but younger than the OTT 40 Ar/ 39 Ar age of Diehl (1987).
The geochronological data and reinterpretation of the field geology show that the proximal Toba stratigraphy requires further detailed mapping, geochemistry, and eruption volume estimates so that we can develop a revised understanding of the old Toba eruption cycle. Ash E in ODP 758 is geochemically distinct from any of the Toba eruption products that we have analysed and, at present, the tephra provenance is unknown.

A Bayesian 40 Ar/ 39 Ar age-depth model for the Pleistocene of ODP 758
Bayesian age-depth modelling of ODP 758 was performed using the OxCal (ver. 4.2) software of Bronk Ramsey (2013) and Bronk . A 'P_Sequence' (i.e., Poisson process) deposition model was applied (Bronk Ramsey, 2008), whereby the deposition rate of the sediment sequence is allowed to vary from that of a constant deposition rate through time (i.e., a uniform 'U_Sequence' in OxCal) according to the additional constraint of a parameter, 'k' (a higher value of k gives an increasingly linear deposition rate; lower values of k allow increasing flexibility away from a uniform deposition rate). In the context of sediment deposition, the P_Sequence model provides a realistic representation of sediment accumulation, with the complexity (randomness) of the underlying deposition modelled according to a Poisson process. The k parameter is not fixed a priori, allowing the program itself to determine an unbiased measure of the rigidity of the deposition rate, based upon the dating ('likelihood') information combined within the P_Sequence model prior (Bronk Ramsey and Lee, 2013).
Since the four tephra layers within ODP 758 represent macroscopic, instantaneous deposits ('instantaneous' in the context of the timescales considered here, at least), their respective thicknesses (Ash A, YTT 34 cm; Ash V, MTT 23 cm; Ash d, OTTB 13 cm; and Ash D, OTTA 2 cm thick) were excluded to provide an 'event-free depth' scale (e.g., Katsuta et al., 2007;Schlolaut et al., 2012) so that the regular, 'background' deposition rate could be effectively modelled.
In addition to the four 40 Ar/ 39 Ar dated tephra units, 'Date' functions were also inserted within the model to provide posterior age distributions for the depths of the MBB, Australasian tektite layer, and Jaramillo event (onset and termination) within ODP 758 (unlike the tephra units, these latter Date functions were included without any prior chronological information associated with them). The top and bottom of the sediment sequence (at 0 and 18.9 m depth) were additionally constrained by 'Boundary' functions, with  the upper boundary defined as the date of core extraction, AD 1988. The lower boundary was somewhat arbitrary, but represents the subsequent break between core sections below the Jaramillo event, the base of section 758B-2H (Shipboard Scientific Party, 1989). As there was no other sedimentological evidence within the stratigraphy for abrupt changes in the mode of sediment deposition, no further 'Boundary' functions were inserted within the P_Sequence model.
To assess whether two age distributions are statistically different within OxCal, the 'Difference' function (which simply subtracts one age distribution from another) is applied. Here, Difference queries were applied between the modelled ODP 758 ages and published ages for both the MBB and Australasian tektite layer. If the calculated probability range for the Difference query does not include zero at a given confidence level (typically, 95.4% confidence), a null hypothesis (that the two age distributions are consistent) can be rejected, and the ages can be described as being statistically significantly different (Macken et al., 2013;Wood et al., 2014).
Due to the limited number of likelihood data (i.e., four 40 Ar/ 39 Ar ages plus date of core extraction) within the model, there is no contradictory information (given the model prior) to 'pull' the modelled age-depth profile away from the raw, un-modelled data. Accordingly, all of the individual modelled data points exhibit excellent agreement indices of 100% (i.e., there is no evidence of stratigraphic inversions, or unreliable 40 Ar/ 39 Ar measurements.) In order for the Poisson process (P_Sequence) age-depth profile to pass through these data points (which fall well away from linear sediment deposition), however, the modelled k parameter must be fairly low. The result of this is that the chronological precision of the interpolated (and extrapolated) depths is lower than if the deposition rate were more linear (i.e., if OxCal had determined a higher value for k). This reduction of modelled chronological precision becomes more pronounced further away from the 40 Ar/ 39 Ar-dated core depths, as illustrated in Fig. 9. Table 2 provides a summary of unit/event depth information (both mbsl and 'event-free' depth) used in the model construction. Note that interpretation of the positions of geomagnetic events within marine cores can be subjective -we have used the depths for the tephra layers and geomagnetic polarity reversals that have been published previously (Dehn et al., 1991;Farrell and Janecek, 1991). We note that, unfortunately, paleomagnetic intensity data are not available for ODP 758 but, due to the sensitivity of the model to the 40 Ar/ 39 Ar data and large uncertainties as we move away from these tie points, small changes (i.e., cm-scale changes) in the location of the MBB will not significantly impact the age ranges reported relative to the uncertainty associated with each modelled age (median age shifts approximately 0.7 ka/cm).

ODP 758 defined MBB age
The modelled age for the MBB (all modelled ages below reported at the 68.2% confidence level) is 784 ± 2 ka (Fig. 9). The MBB age as defined by ODP 758 is statistically older (at the 95.4% confidence level) than the proposed age of Channell et al. (2010) for the North Atlantic, but in complete agreement with the age proposed by Sagnotti et al. (2014) for samples from the Sulmona Basin. This age is indistinguishable from the ODP 758 astronomically tuned age of c. 784 ka.

ODP 758 defined JGE age
The modelled age range for the MBB is more precise than the extrapolated ages derived for the JGE onset (median: 1082 ka, 1001e1159 ka) and JGE termination (median: 1002 ka, 933-1064 ka) ( Fig. 9). This is due to the lack of an age constraint stratigraphically below the JGE within ODP 758 to anchor the Bayesian age-depth model. We are surprised however that even with absence of this constraint, that the median ages are relatively close to the reported ages for the onset (1070 ka) and termination (997 ka) of the JGE, respectively (Chen et al., 1995). With respect to current discussions in the literature (e.g., Singer, 2014) our data, owing to this low precision output from the model (i.e., lack of a temporal marker below the JGE onset ), do not advance understanding of the timing of the JGE.

The age of the Bishop Tuff
A robust age for the BT will allow determination of a North American MBB age that provides an independent comparison with the ODP 758 constraint. New data collected here and re-calculation of previously published data (e.g., Simon et al., 2014) shows that there is temporal alignment between the 40 Ar/ 39 Ar and 238 U-206 Pb geochronometers for the BT. However, we suggest caution in the (over)interpretation of the high precision zircon ID-TIMS 238 U-206 Pb ages for dating of Pleistocene volcanic eruptions. A 'high-precision' BT zircon ID-TIMS 238 U-206 Pb age of 767.1 ± 0.5 ka (1 sigma, full uncertainty, Crowley et al., 2007) is significantly younger than the BT zircon ion microprobe (SIMS) ages (Reid and Coath, 2000;Simon and Reid, 2005) that suggest a mean pre-eruptive zircon magma residence time greater than 50 ka. Recently, Ickert et al. (2015) collected new SIMS and ID-TIMS 238 U-206 Pb age data and demonstrated both inter-and intra-grain variability in apparent U-Pb ages of BT zircon crystals. The new data support the forward modelling of Simon et al. (2008) and explain the discrepancy when interpreting the previous SIMS and ID-TIMS U-Pb age data (Reid and Coath, 2000;Simon and Reid, 2005;Crowley et al., 2007), but highlight that the single coherent population of juvenile crystals dated by Crowley et al. (2007) was not an 'eruption age' as previously implied, but a result of the strong correlation of the uncertainty in one component of the 230 Th disequilibrium correction. If the correlation is accounted for correctly then there exists substantial variability in the ages of the zircon crystals that precludes determination of a meaningful weighted mean crystallisation age.
The dating of BT zircon by U/Pb methods is highly challenging because the concentration of radiogenic Pb is low and the correction required for disequilibrium in the intermediate daughter products is large. These corrections for young zircon are significant, for example, a correction of greater than 80 ka was employed previously for BT zircon (Crowley et al., 2007). Moreover, the corrections are often based on best-case scenarios and assumptions (models) that are difficult to validate, e.g., that the host magma 'Th/ U melt ' composition employed accurately represents the melt composition from which the zircon grew and requires that the magma itself was in U-series equilibrium prior to zircon growth. Hence the utmost caution must always be employed when interpreting geologically young ID-TIMS 238 U-206 Pb BT zircon data with respect to 'eruption ages' and we suggest that in general, such ages reported with relative uncertainties at the permil level are suspect of being unduly optimistic.
Several studies have now shown that young ID-TIMS 238 U-206 Pb zircon ages 'approach' eruption ages (e.g., Rivera et al., 2013), but as highlighted by Ickert et al. (2015) -when interpreting such data (i.e., zircon that forms/closes over a continuum rather than in response to a specific geological event, e.g., eruption) one should appreciate that employment of a weighted mean to geologically young high-precision zircon ages requires an assumption that necessitates a geologically implausible event. A further, important point to note is how the Th/U melt uncertainty is propagated in the correction (e.g., Crowley et al., 2007). This also affects the reported uncertainty of the new zircon rim 'eruption age' measured by SIMS (Chamberlain et al., 2014; discussed in-depth by Ickert et al., 2015). For the uncertainty propagation two approaches are prevalent. The first is similar to that of Crowley et al. (2007). The Th/U melt correction is applied to each individual zircon analysis and then the weighted mean of the population is determined. In the second the Th/U melt is treated as a systematic variable, and so this uncertainty is propagated following determination of a weighted mean age, applying it to the weighted mean 206 Pb*/ 238 U. The latter leads to a larger and we contend a more realistic age uncertainty. Note that the uncertainty on the corrected age has an inverse relationship with the Th/U melt value, which is clearly demonstrated in Fig. 10. Sarna-Wojcicki et al. (2000) presented the first 'high-precision' 40 Ar/ 39 Ar age measurements from the BT but recently the BT has been extensively studied. Rivera et al. (2011) reported an 40 Ar/ 39 Ar sanidine age of 767.4 ± 1.1 ka (R BT FCs : 0.02706 ± 0.00005), relative to their proposed astronomically tuned age for Fish Canyon sanidine (FCs). Zeeden et al. (2014) made measurements in the same laboratory and reported a BT 40 Ar/ 39 Ar age that is identical to the age reported by Rivera et al. (2011) but note they rejected more than 30% of their data culling the MSWD to 0.3 to improve analytical precision. Although the data were trimmed symmetrically around the determined mean (not impacting accuracy), this is an approach that is not 'best practice' with respect to statistical assessment of geochronological data. Relative to the highest precision attainable using the 40 Ar/ 39 Ar technique, the data of Sarna- Wojcicki et al. (2000), Rivera et al. (2011) and Zeeden et al. (2014) can be improved on. Thus we collected new high-precision 40 Ar/ 39 Ar age data to better define R BT ACs (Fig. 7). Exhaustive new analyses validate the results of Rivera et al. (2011) and Zeeden et al. (2014). However, for calculation of 40 Ar/ 39 Ar ages we do not favour the use of the FCs calibration presented by Rivera et al. (2011); we provide our reasoning below. 40 Ar/ 39 Ar data (Renne et al., 2013) for the Cretaceous-Palaeogene (K-Pg) boundary show that the orbitally-tuned FCs calibration of Rivera et al. (2011) places the K-Pg boundary exactly intermediate between two possible choices of 405 ka orbital eccentricity cycles. The implication is that the astronomically tuned age for FCs (Rivera et al., 2011) is paradoxically inconsistent with any astronomically tuned age for the K-Pg boundary. Rather the 40 Ar/ 39 Ar calibration (FCs age) of Renne et al. (2011) is proven to be the most consistent with the orbitally-tuned age (Kuiper et al., 2008) for the K-Pg boundary (Renne et al., 2013). It is this calibration with robust and quantifiable uncertainties that we favour. Using the updated R ACs FCs (0.041707 ± 0.000011) reported by Niespolo et al. (2016) as a parameter in the optimized calibration of Renne et al. (2010), along with the decay constant from Renne et al. (2011), our 40 Ar/ 39 Ar data define a BT eruption age of 766.6 ± 0.4/ 0.5 ka (R BT ACs : 0.64473 ± 0.00034) (Fig. 7). In view of our comprehensive data set, from multiple eruptive BT units (n3), laboratories (n2), mass spectrometers (n3), and operators (n3), we regard this as the most precise and accurate age for the BT. The 40 Ar/ 39 Ar age of Rivera et al. (2011) recalculated relative to the same parameters as our data is 767.6 ± 1.0/1.1 ka, the 40 Ar/ 39 Ar age of Sarna- Wojcicki et al. (2000) is 768.7 ± 3.2/3.3 and the 40 Ar/ 39 Ar age of Simon et al. (2014) is 769.0 ± 3.1/3.2 ka. These data are all indistinguishable relative to each other and consistent with a relatively imprecise astronomically tuned age for the Bishop Tuff of 765 ± 8 ka (Zeeden et al., 2014). The data are also consistent with the interpretation of Ickert et al. (2015) that the BT zircon ID-TIMS 238 U-206 Pb eruption age is < 775 ka. Fig. 11 shows a summary of these data against the 'weighted mean' BT zircon ID-TIMS 238 U-206 Pb age (Crowley et al., 2007) and the BT zircon ID-TIMS 238 U-206 Pb age distribution of Ickert et al. (2015). The 238 U-206 Pb data, 40 Ar/ 39 Ar and astronomical ages have all converged for the BT, to provide a robust temporal marker for the Pleistocene Time Scale.

The MBB and the Bishop Tuff
To define the most accurate age for the North American MBB relative to the BT we adopted the same approach as Sarna- Wojcicki et al. (2000), but built in extra uncertainty to our calculations (as detailed below). We highlight that due to the nature of the terrestrial sections used by Sarna-Wojcicki et al. (2000) (potential for unknown hiatuses in the stratigraphy) this approach is not going to yield a high-precision age constraint, just a useful comparison with the ODP 758 and Channell et al. (2010) MBB ages.
We have made new 40 Ar/ 39 Ar age determinations on sanidine from the LCTB (Fig. 8), the tuff that postdates the BT in several North American sections that also contain a record of the MBB position (Fig. 12). Single crystal analyses show a LCTB juvenile age population (n28) with a robust 40 Ar/ 39 Ar age of 627.0 ± 1.5/1.7 ka (R LCTB ACs : 0.52734 ± 0.00126). Note that our age for the LCTB is indistinguishable from the 40 Ar/ 39 Ar age of Matthews et al. (2015) when both are calculated relative to the same calibration (627.4 ± 1.5/1.7 ka, R LCTB ACs 0.52754 ± 0.00124) and is also indistinguishable at the 2 sigma confidence level from the ID-TIMS 206 Pb/ 238 U LCTB age (629.2 ± 4.3 ka, 2 sigma full uncertainty) reported by Wotzlaw et al. (2015).
As the LCTB postdates the BT we then, following Sarna- Wojcicki et al. (2000), simply calculated the sedimentation rate between LCTB and BT, and subsequently extrapolated Dt from the BT to the MBB in each (n5) individual section (Fig. 13, Table 3). As expected, there is considerable scatter in the extrapolated Dt and MBB ages (MSWD 54) relative to what precision would predict, due most likely to the presence of hiatuses in the stratigraphic sections, either between the LCTB-BT and/or BT-MBB. Therefore, we deem it inappropriate to use the standard error of the mean as a representative uncertainty for the BT-defined MBB age constraint. We choose to use the SEM Â SQRT(MSWD) of all five measurements as the most appropriate method for determining a robust uncertainty. Note that we assigned 20% uncertainty to the stratigraphic distances between LCTB, BT and the MBB to also account for potential hiatuses in the stratigraphy at each site (Table 3). The approach defines a MBB age of 789.1 ± 5.6 ka (68.2% confidence). Simply using the average age ± standard deviation of the stratigraphic and age measurements for each sites yields an MBB age of between 776 and 802 ka (Table 3). Both age ranges are resolvable from the proposed MBB age of Channell et al. (2010).

The age of the MBB
In proposing paradigm-changing shifts in the age of key events within the Geological Time Scale the burden of proof is high. We feel that the data presented here, when considered with respect to the study of Sagnotti et al. (2014), pose serious questions concerning the accuracy of the approaches used previously to date the MBB (for example : Channell et al., 2010). There are now three independent robust and accurate 40 Ar/ 39 Ar age constraints indicating that current estimations for the age of the MBB are too young. As such, the revision to the Geological Time Scale that we propose has far reaching implications for Quaternary science and other dating techniques.
We consider this to be currently the most accurate MBB age and the most robust temporal anchor for the Pleistocene Geomagnetic Time Scale. It is c. 10.4 ± 1.5 ka older than the MBB age proposed by Channell et al. (2010), and consequently c. 3.4 ± 0.7 ka older than the age of the MBB in the LR04 stack (Lisiecki and Raymo, 2005) (or as defined by Shackleton et al., 1990).
We note that our absolute age for the MBB is dependent on  Ickert et al. (2015). The data are corrected for Th/U melt using a value of 2.81 (Anderson et al., 2000) e the data cannot be interpreted with respect to a mean age as each bulk zircon age is a function of integrating a time-series of crystallisation. Ickert et al. (2015) highlight that with a Th/U melt value of 2.81 the BT probably erupted post-775 ka (blue line in figure). Thus it is inescapable that there are discrepancies between the most precise 40 Ar/ 39 Ar ages and some (but not all, see discussion below) orbital tuning ages for the MBB.

Magnetic lock-in delay
It may be questioned whether the ODP 758 MBB age that we present (or the approach we have taken with respect to OPD 758) is biased by delayed acquisition of magnetic remanence (i.e., lock-in) in the sediment (Kent, 1973;Suganuma et al., 2011), owing to a relatively low sedimentation rate. The only possible evidence for this is that the MBB in ODP 758 is not located at the same position within MIS 19 as it is in the high-resolution North Atlantic cores (Channell et al., 2010). There is no doubt that magnetic lock-in delay is a real issue for interpretation of paleomagnetic data from some sediment cores (e.g., Tauxe et al., 1996) but not others (e.g., Valet et al., 2014) and that when assessing such phenomena, we have to consider the level of precision one is achieving when temporally resolving paleomagnetic events relative to any potential lock-in offset.
Typically, the degree of magnetic lock-in delay in marine sediments at the depth of the MBB is minimal (e.g., Tauxe et al., 1996;Bleil and von Dobeneck, 1999;Horng et al., 2002) but in ODP 758 we do not consider lock-in of paleomagnetic remanence as significant, certainly at the depth of the MBB. Our reasoning for this is that there are now three unrelated 40 Ar/ 39 Ar age constraints for the MBB from three geographically distal sites representing three distinct depositional environments (Sulmona Basin, North America, South China Sea) that are indistinguishable from each other at the 68.2% confidence interval. They yield a weighted mean MBB age with a MSWD of 0.8, revealing no excess scatter in the data as there would be, for example, if paleomagnetic lock-in delay was significant in one of the records. To make an argument for our ODP 758 data being affected significantly by delayed magnetic lock-in would require for the North America MBB record and the Sulmona Basin record to be as equally offset from a younger MBB by exactly the same amount of time. As a consequence, our data now raise two important questions: (1) why is there an offset in the location of the MBB relative to MIS 19 in different records? And (2) why are some orbitally tuned ages for the MBB consistently younger than the most precise and accurate 40 Ar/ 39 Ar ages? 7.9. Why is there an offset in the location of the MBB relative to MIS 19?
As paleomagnetic lock-in delay in ODP 758 is not significant with respect to the temporal resolution we have achieved, we are left with two possibilities to explain the differences in the ODP 758 MBB location within MIS 19 relative to the high-resolution North   Table 3.
Atlantic cores (Channell et al., 2010): (i) it is the position of the MBB that has changed within rock archives across the globe -diachronous onset as suggested by previous modeling of the MBreversal (Leonhardt and Fabian, 2007;Olson, 2011); or (ii) it is not the position of the MBB that has changed, it is the onset and termination of MIS 19 within marine records that are different across the globe (relative to a fixed position for the geomagnetic reversal).
(i) As Earth's magnetic field intensity drops to low values during polarity reversals the field direction progresses through a 180 change while the field is weak. The time it takes for this process to happen is uncertain. Modeling of the MBB event has suggested reversal durations of between 2 and 10 ka and highlighted potential for a millennial-scale variability in onset of the MB-reversal for sites in the Atlantic and Pacific Oceans (Leonhardt and Fabian, 2007). Such age offsets have not been reported in the literature, although few existing age data are sufficiently precise to resolve diachrony at this scale. A diachronous MBB is supported by the work of Olson (2011) who compared the MBB paleomagnetic trajectories to a complex dynamo model depicting a polarity reversal. Both were initiated by gradual reductions in dipole intensity leading to a reversal precursor event (intensity low) and subsequent transient polarity recovery. Following this was rapid dipole collapse and final directional reversal that began with reverse flux generation in one hemisphere. Virtual geomagnetic poles (VGPs) from sites located proximal to the reverse flux follow complex paths crossing the equator several thousand years prior to the simpler VGP paths from the more distal sites e the magnetic intensity variations produced by the dynamo model reversal correlate with intensity variations inferred for the MBB (Olson, 2011). These theoretical/model data suggest diachronous onset of the MB-reversal on a time scale that should be resolvable using highprecision radio-isotopic dating. However, our data show temporal coincidence of the MBB at the 68.2% confidence interval for three sites that vary with respect to latitude and longitude. Clement (2004) noted that polarity reversal durations vary with site latitude; low latitude sites have shorter reversal durations than mid-to high-latitude sites. As such, the physical data (analytical measurements) do not appear to support a diachronous MB-reversal model (Leonhardt and Fabian, 2007;Olson, 2011) in which the reversal timing is a systematic function of latitude. Instead, the data support the interpretation that the MB-reversal was a globally isochronous event at the millennial scale.
(ii) Benthic v 18 O (e.g., Fig. 14) is used to align marine records from across the globe. Simplistic tuning of records, or wiggle matching, requires one to make the assumption that the global climate system responds uniformly over millennial time scales. v 18 O change is thought generally to be globally synchronous to within 1 ka e the approximate mixing time of an ocean. This is the fundamental assumption in construction and utillisation of global marine stacks such as LR04 (Lisiecki and Raymo, 2005). Radiocarbon data support this supposition for recent geological time with 14 C ages of the Last Glacial Maxima (c. 24 ka) as identified in v 18 O data agree to within 1 ka (Duplessy et al., 1991). If v 18 O changes are not synchronous to within the time it takes the oceans to mix, then any age model that is based on alignment of v 18 O signals would contain significant errors (several ka).
However, the further back in time one goes the greater degrees of offset in v 18 O we can observe, with respect to mixing of the oceans. High-resolution records from the Iberian Margin that chart the last glacial termination (Termination I) provide direct evidence for diachronous benthic v 18 O response with 14 C age models showing the Atlantic was leading the Pacific by c. 4 ka (Skinner and Shackleton, 2005). The c. 4 ka offset would result in an erroneous age model if a stack was constructed or if benthic v 18 O was used as a proxy for ice volume. The mixing of v 18 O throughout the oceans is complicated further by changes in water depth. Studies have demonstrated that it can take an extra 1.5 ka for changes in v 18 O to reach deep-water sites within the same ocean (Labeyrie et al., 2005;Waelbroeck et al., 2006), let alone for v 18 O to be transmitted between shallow and deep water at sites more distal or isolated from the Atlantic. Lisiecki and Raymo (2009) compared v 18 O records from both sites in the Atlantic and Pacific to assess the respective leads and lags in benthic v 18 O. They concluded that v 18 O data show a statistically significant Atlantic lead relative to the Pacific v 18 O. For Terminations I-IV a Pacific benthic v 18 O lag of 1.6 ka was estimated and at 128 ka and 330 ka, a c. 4 ka lag for the Pacific was determined. It was concluded that such leads-lags, probably generated by diachronous temperature changes (without the requirement for slower circulation), will lead to uncertainties of several ka during glacial terminations and this must be taken in to account when using benthic v 18 O records as a proxy for the timing of ice volume change. Lisiecki and Raymo (2009) note that for different terminations the v 18 O lag could vary dramatically due to the differences in ice volume at the glacial maximum and/or the insolation forcing (Parrenin and Paillard, 2003;Parrenin et al., 2007). Given the increase in lag times between the Atlantic and Pacific Oceans from 1.6 to 4 ka between Terminations I-IV (20e330 ka) it is currently unclear what the lag time would have been by Termination IX, the termination that pre-dates the MBB (Fig. 14). These data (Skinner and Shackleton, 2005;Lisiecki and Raymo, 2009) show that the onset and termination of Marine Isotope Stages across the globe cannot, and should not, be considered as synchronous at the level of temporal resolution now attainable using radioisotopic dating. It is also unclear whether we can expect the duration of MIS within different oceans to be the same or whether contraction-expansion of MIS can occur, especially for sites that are proximal or distal to the sites of ice melting (the poles). It is currently impossible to determine accurate age offsets for the MIS times scale between the Atlantic and Pacific, let alone temporally constrain interactions between the Atlantic and other oceans (e.g., Indian Ocean).
Therefore, we contend that within v 18 O records that are obtainable for the Pleistocene there is no requirement for the position of the MBB to be located at the same point in MIS 19. The fact that ODP 758 v 18 O data show the MBB at an earlier position (Fig. 2) in MIS 19 relative to the high-resolution records of the North Atlantic (Channell et al., 2010) should not result in immediate dismissal as evidence of a paleomagnetic 'lock in' delay (Roberts and Winklhofer, 2004;Suganuma et al., 2011). Further, such v 18 O offsets between marine records should not be used as an assessment of the degrees of paleomagnetic lock-in delay (e.g., Horng et al., 2002).
The EU-funded INTIMATE (INTegrating Ice core, MArine and TErrestrial records) network has recognized previously problems of assuming synchronous global response in climate systems and as such, has devised protocols to avoid making such assumptions, which can introduce unquantifiable uncertainties in age models (http://cost-es0907.geoenvi.org). INTIMATE correlations are based on independent tie-points (temporal anchors), which are coupled and compared through the use of either tephra markers (tephrochronology) (e.g., Smith et al., 2013) or accurate/precise chronologies (e.g., Smith et al., 2011;Staff et al., 2013). This approach has led to construction of robust 'event stratigraphies' that have allowed testing of leads and lags in response to climate forcing (Bjorck et al., 1998;Alloway et al., 2007). For 'absolute' dating of processes and events within the Geological Time Scale the INTI-MATE approach is more robust than wiggle matching that, at best, allows for relative assessments of time.

Integration of the astronomical and 40 Ar/ 39 Ar MBB age
The new age for the MBB is consistent with the astronomical age reported by Chen et al. (1995) for ODP 758. At first sight there appears to be an offset between the MBB age obtained by us using the 40 Ar/ 39 Ar dating technique and the LR04 stack MBB age (Lisiecki and Raymo, 2005; consequently the astronomical age of Shackleton et al., 1990). However, we do not consider this to be the case -with application of appropriate uncertainties (±5 ka, Martinson et al., 1987) these ages are indistinguishable. We therefore propose our MBB age could now be used as a high-precision tie point in the model of Lisiecki and Raymo (2005) to line up the occurrence of the MBB in marine records (that do not exhibit significant magnetic lock-in effects).
However, the offset with the proposed MBB age of c. 773 ka is real and we suggest there must be uncertainties (beyond reported precision) or errors in the approach of Channell et al. (2010). A detailed review and assessment of astronomical dating and its inherent uncertainties is beyond the scope of this contribution, but we can make some first order observations and pose questions for consideration.
We have already highlighted that previous work has suggested that any astronomically tuned age for the Pleistocene cannot be defined better than to within ± 5 ka (Martinson et al., 1987;Imbrie and Imbrie, 1980). Astronomical ages are derived by wiggle matching climate proxy cycles in sedimentary sequences to either: (1) astronomic solutions (Laskar et al., 2004) for orbital cycles, or (2) calculated solar insolation for a specific latitude and time of year (Milankovitch, 1930;Hays et al., 1976). For example, Fig. 14 shows the planktonic and benthic v 18 O records for ODP 983 relative to ice volume models based on midsummer and integrated summer insolation forcing. If current levels of uncertainties associated with the astronomical dating approach are robust (Martinson et al., 1987), the revision we are proposing (c. 10 ka) to the astronomical MBB age (773 ± 1 ka) of Channell et al. (2010) (relative to our 40 Ar/ 39 Ar MBB age, 783.4 ± 0.6 ka) is too large to be accounted for by uncertainties in phasing assumptions. Whilst it is true that there are inherent uncertainties associated with the orbital calculations themselves; the orbital solution being sensitive to both shifts in tidal dissipation, and changes in global ice volume that may potentially alter the Earth's dynamical ellipticity (Laskar et al., 1993), other astronomical tuning studies (e.g., Shackleton et al.,  Channell et al., 2010). Also shown is Virtual Geomagnetic Polar (VGP) latitude (black) and relative intensity proxy (grey). Ice volume models based on midsummer (orange) and integrated summer (blue) insolation forcing are also plotted. The blue line shows the position of the MBB in the records and the blue box the duration of the MB-reversal at the site of ODP 983 as defined by the benthic v 18 O. The lower x-axis shows the astronomical timescale as discussed by Channell et al. (2010 whereas the upper x-axis shows the revised chronology based on the high precision 40 Ar/ 39 Ar age constraints presented here. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) 1990; Chen et al., 1995) that report MBB ages indistinguishable from our MBB age, use the same orbital calculations which must have the same intrinsic uncertainties/inaccuracies. The remaining possibilities are: (1) orbital sediment cycles may have been mismapped onto orbital forcing by Channell et al. (2010), or (2) the level of precision attained by Channell et al. (2010) is grossly underestimated. It should be noted that although Channell et al. (2010) proposed an error of just ± 1 ka for their MBB age, this represents the standard deviation of the midpoint of the MBpolarity transition for multiple marine records, and not a realistic uncertainty associated with the astronomical dating approach to deriving the MBB age in the Atlantic marine cores.
7.11. The MBB and transitionally magnetized lava flows As discussed above, 40 Ar/ 39 Ar ages for transitionally magnetised lava flows have been cited (Baksi et al., 1992;Singer and Pringle, 1996;Coe et al., 2004;Singer et al., 2005;Singer, 2014) as supporting evidence for the astronomical age of the MBB (Channell et al., 2010) and as evidence for a MBB precursor event. There are relatively large age corrections associated with the analysis of low radiogenic 40 Ar basaltic-andesitic groundmass that can significantly impact the accuracy of 40 Ar/ 39 Ar ages (McDougall and Harrison, 1999;Barfod et al., 2014). In comparison to the levels of precision achieved by Sagnotti et al. (2014) and here by the targeting of K-rich sanidine, the level of accuracy and precision attained for dating of young lavas is typically poor, especially if relying on high background isotope extraction techniques such as furnace step-heating (e.g., Singer et al., 2005). Fig. 15 shows the MBB lava data (Baksi et al., 1992;Singer and Pringle, 1996;Coe et al., 2004;Singer et al., 2005) relative to the proposed ages for the MBB. Note that none of the data are consistent with a MBB age of c. 773 ka (Channell et al., 2010) e this observation is independent of which 40 Ar/ 39 Ar calibration is utilised. Further, whereas it was previously concluded that the data from Chile, Tahiti and La Palma were dating a MBB precursor event, Fig. 15 shows that these data are consistent with our ages for the MBB and the data of Sagnotti et al. (2014). In actual fact, we suggest that the data of Singer et al. (2005) are not dating two events (a MBB precursor event and the MB-reversal) as invoked previously to explain the excess scatter in the data; the data are only dating the MB-reversal, albeit at relatively low accuracy and precision.
The large degree of scatter in data is probably associated with difficulty in dating basaltic-andesitic lava flows. For example, just by examining the data presented from Tahiti (Singer et al., 2005) atmospheric argon contamination ( 40 Ar ATM ) was accounting for (typically) more than 70% of the total 40 Ar budget and thus any small error in this correction would impact age accuracy (but not necessarily age precision). In comparison to sanidine of similar age that we have dated, crystals typically contained c. 10% 40 Ar ATM . Hence the corrections are much smaller and easier to make. Most of the single data points of Singer et al. (2005), with the exception of some of the Maui data, are indistinguishable at 95% confidence from the MBB ages at 783.4 ± 0.6 ka when normalized to the same calibration (Fig. 15). Although it is possible that the Maui data are dating a prolonged reversal event (Fig. 15) the work of Sagnotti et al. (2014) suggests the MB-reversal was not of this duration and we consider this scenario unlikely given the reproducible MBB ages we have collected for sites of different latitudes and longitudes -we suggest the data are most likely problematic (inaccurate), certainly at the levels of accuracy and precision that have been reported previously (Singer et al., 2005). Taking this in to consideration we have calculated a weighted average for all the ages for the MBB related lavas (including data from Maui) presented by Singer et al. (2005). We have determined the uncertainty using SEM Â SQRT(MSWD) as there is significant scatter in the data  (Baksi et al., 1992;Singer and Pringle, 1996;Singer et al., 2005) relative to Renne et al. (2011) and the MBB ages for ODP 758, North America, Sulmona Basin (Sagnotti et al., 2014) and the estimated MBB age of Channell et al. (2010).
It is important to note that we have not stated that there was not a MBB precursor event, as indeed different records highlight a geomagnetic intensity low prior to the MBB (Kent and Schneider, 1995;Hartl and Tauxe, 1996;Channell et al., 2009Channell et al., , 2010. We are simply stating that the 40 Ar/ 39 Ar data of Singer et al. (2005) are most likely associated with and dating the MBB. Apart from the age discrepancy reported by Singer et al. (2005) and the explanation invoked to explain this discrepancy, there is no paleomagnetic evidence that the lavas from these sites are related to a MB-reversal precursor event. In fact, the sites were specifically targeted in the first place as they were thought to contain detailed records of the MBB.

Implications for quaternary age models
Decoupling the age of the MBB from the astronomical time scale and assumptions concerning alignment of v 18 O isotope records allows for independent testing of different Quaternary age models. It is important to realise that our results do not relocate the MBB within the different paleoclimate records; they simply question the robustness/accuracy of the time scales associated with the paleoclimate records.
Termination IX within ODP 758 is coincident with OTTB at 785.6 ± 0.7/0.8 ka (Fig. 2). Valet et al. (2014) using an astrochronological model determined a Termination IX age of 788e789 (±5) ka from high-resolution Be records across the equatorial Indian Ocean, which allowing for the uncertainties associated with the astronomical tuning approach, is indistinguishable from the ODP 758 age. Further, as Valet et al. (2014) conducted both Be and paleointensity measurements on the same samples we can compare the age of the MBB in ODP 758 with the onset of the MBB from their study. Based on the astronomical age model presented by Valet et al. (2014) the relative paleointensity drop and recovery associated with the MBB occurred at 784 (±5) ka while the cosmogenic Be data indicates reversal onset at 780 (±5) ka -both these ages for the MBB are indistinguishable from our reported MBB age of 783.4 ± 0.6 ka. These data are also commensurate with the age for Termination IX in the LR04 stack (788e789 (±5) ka, Lisiecki and Raymo, 2005).
In addition to now having three independent radioisotopic age constraints (ODP 758, North America sections, Sagnotti et al., 2014) placing the MBB at 783 ± 0.6 ka, the MBB age reported by us is indistinguishable with the astronomical MBB age reported by Shackleton et al. (1990), Chen et al. (1995) Lisiecki and Raymo (2005) and Valet et al. (2014). The implication is that the MB reversal was, at the current levels of temporal resolution, isochronous. There is also agreement between all these records for the age of Termination IX. Our Termination IX age (785.6 ± 0.7/0.8) is also indistinguishable from the age for Termination IX as determined from the Sulmona Basin record (c. 787 ± 2/2 ka) (Giaccio et al., 2015). Fig. 14 (lower x-axes) shows that the offset between the onset of the MB reversal in the Atlantic Ocean record ODP 983 (Channell et al., 2010) and Termination IX is c. 15 ka (Channell et al., 2010). Accepting that the MBB age of Channell et al. (2010) is inaccurate and that the age presented here is correct, as well as assuming that the location of the MBB in the Atlantic records is accurate, then relative to a MBB age of c. 783 ka Termination IX in the Atlantic Ocean should be positioned at c. 798 ka. Does this observation suggest that with respect to the onset of Termination IX, the Atlantic Ocean was leading the Indian Ocean (and terrestrial records) in v 18 O response by c. 12 ka?
As discussed previously, Lisiecki and Raymo (2009) did note that for different terminations the v 18 O lag does vary dramatically due to the differences in ice volume at the glacial maximum and/or insolation forcing (Parrenin and Paillard, 2003) with differences of ± 4 ka by Termination IV at 330 ka. If so, then this study and such a large lead-lag between the Atlantic and Indian Oceans further highlights that the dangers of 'wiggle matching' approaches to comparing climate records from across the globe.
We can also use the position of the 'isochronous' MB reversal within multiple paleoclimate archives to correlate between records. Raisbeck et al. (2007) concluded that the enhanced 10 Be flux in the EPICA Dome C ice core is a product of low dipole intensity during the MB-transition. Fig. 16 shows the MBB tie point in different records at c. 783 ka allowing for correlation between the LR04 stack and EPICA Dome C, as well as correlation to Northern Hemisphere July insolation. The horizontal displacement (along the x-axes) of these tie points shows the inaccuracies in the different time scales currently in use. Our interpretation does not impact the relative temporal offset within a single record, but does highlight that without high-precision independently dated tie points, it is currently not possible to directly compare climatic records from different sources throughout the Pleistocene.
The MBB has also been identified within Chinese loess and red clay sections (Zhou et al., 2014;Wang et al., 2014) but the apparent timing and duration of the MBB remain controversial due to inconsistencies in stratigraphic location. This Chinese record however is of key importance, as it would allow for cryospheric-marine-land correlation of climate records and paleoclimatic reconstruction across reservoirs. If as suggested (Zhou et al., 2014), S7 and S8 within the Chinese loess sections (Fig. 16) correspond to MIS 19 and 21, respectively, then the MBB in the loess significantly pre-dates the MBB elsewhere. Complex postdepositional processes have been invoked to explain the massive downward shift of the MBB in the loess (Suganuma et al., 2010(Suganuma et al., , 2011 but there is another possibility to consider. Several studies have previously proposed that S8 correlates to MIS 19 (Liu et al., 2008;Yang et al., 2010;Wang et al., 2006;Jin and Liu, 2011) and not MIS 21, which then places the MBB in the Chinese loess sections close to the location of the MBB in both the marine records and ice cores. Although this interpretation causes issues for stratigraphic correlation between the loess sections in China (Zhou et al., 2014), we consider it to be the most plausible with respect to the evidence at hand. Fig. 16 shows the proposed correlation and the linkage of S8 to MIS 19.

ODP 758 defined Australasian Tektite age
The modelled mean age for the Australasian tektite layer (main concentration interval in ODP 758) is 786 ± 2 ka (Fig. 9). The stratigraphic position pre-dates Termination IX, which is positioned at 785.6 ± 0.7/0.8 ka and we consider this modelled age to currently be the most accurate age for the tektites. As the Australasian tektites are found in Indochina, southern China, the Philippines, Malaysia, Indonesia and Australia an accurate age could be used as an isochronous marker horizon across continents (Smith et al., 2011. In the absence of a crater location, it has been suggested that the impact event that produced the tektites is located in Indochina, probably in close proximity to ODP Hole 1144A (Glass and Koeberl, 2006), but this remains supposition.

Conclusions
The present study has (1) provided a robust chronology for the multiple eruptions of the Toba super-volcano, (2) identified a multiple Toba eruption scenario at approximately 800 ka, (3) provided a robust and accurate age for the Australasian tektites, (4) defined robust high-precision ages for the BT and LCTB, (5) allowed for determination of an accurate and precise MBB age of 783.4 ± 0.6 ka, (6) shown at the level of temporal resolution attainable using radioisotopic dating the MB reversal can be considered isochronous, and (6) dated Termination IX in the Indian Ocean. We highlight issues that pose significant challenges to the accuracy of U/Pb zircon dating in the Quaternary and suggest that relative uncertainties at the permil level are unduly optimistic. Finally, at the level of resolution now attainable for Pleistocene climate archives using radioisotopic dating, it is not valid to assume that response to changing v 18 O can be considered synchronous.
As ODP 758 features in the LR04 marine stack, the highprecision 40 Ar/ 39 Ar ages for the YTT, MTT, OTTA and OTTB, as well as the age for the MBB and Australasian tektites, can be used as temporally accurate and precise anchors. These anchors allow for global-correlation of the geological record, synchronisation of v 18 O climate archives (e.g., ice cores, lake records and speleothems) (e.g., Mark et al., 2014), and for testing of the inter-hemispheric phasing of climate (Shulmeister et al., 2006;Broecker, 1998;Stocker and Johnsen, 2003;Mark et al., 2014). If the misalignment of the Chinese loess sequences is, as suspected, responsible for placing the MBB relative to MIS 19 in the wrong place, then the MBB tie point can, for the first time, allow for climatic reconstruction and correlation within different paleoclimate archives.