210Pb-226Ra disequilibria in young gas-laden magmas

We present new 238U-230Th-226Ra-210Pb and supporting data for young lavas from southwest Pacific island arcs, Eyjafjallajökull, Iceland, and Terceira, Azores. The arc lavas have significant 238U and 226Ra excesses, whereas those from the ocean islands have moderate 230Th and 226Ra excesses, reflecting mantle melting in the presence of a water-rich fluid in the former and mantle melting by decompression in the latter. Differentiation to erupted compositions in both settings appears to have taken no longer than a few millennia. Variations in the (210Pb/226Ra)0 values in all settings largely result from degassing processes rather than mineral-melt partitioning. Like most other ocean island basalts, the Terceira basalt has a 210Pb deficit, which we attribute to ~8.5 years of steady 222Rn loss to a CO2-rich volatile phase while it traversed the crust. Lavas erupted from water-laden magma systems, including those investigated here, commonly have near equilibrium (210Pb/226Ra)0 values. Maintaining these equilibrium values requires minimal persistent loss or accumulation of 222Rn in a gas phase. We infer that degassing during decompression of water-saturated magmas either causes these magmas to crystallize and stall in reservoirs where they reside under conditions of near stasis, or to quickly rise towards the surface and erupt.

This sample has about 2% phenocrysts of augite and plagioclase, which form rare glomerocrysts with magnetite. The matrix is highly vesicular colorless glass. Sample F0805 is a crystal poor, vesicular dacite from the summit of Fonualei volcano that is inferred to have erupted between 1979 and 1990 15 . Our sample of rhyolite from Havre volcano erupted in July 2012 just west of the Kermadec volcanic front at about 1000 meters water depth. The volume of the eruption is estimated at about 1.5 km 3 , making it the largest and deepest recorded silicic submarine eruption 16 . A pumice raft resulting from this eruption spread across the southwest Pacific and Tasman Sea over the ensuing months 17 . Our sample is a spherical homogenous light gray pumice clast of about 900 cm 3 collected from a high-tide accumulation of Havre pumice on a Sydney Harbour beach in February 2014. It consists of frothy fresh glass with about 1% plagioclase and trace augite.
Van A1 is a vesicular lava bomb with abundant mm-scale plagioclase collected shortly after eruption from Yasur volcano, Vanuatu in August 2008 18 .
The Serreta sample is a glassy quenched margin of one of the gas-inflated basaltic pillows that breached the sea surface off of Terceira island in the Azores in February 1999 19 .
Finally, sample EJ-1 is ash deposited during the April 15, 2010 eruption of Eyjafjallajökull in Iceland. It is a portion of sample 2 from Gislason et al. 20 , where it is described in detail. Sigmarsson et al. 21,22 analyzed a separate fraction of this sample. All tephras from this eruption are more than 85% juvenile 23 . The benmoreite ash is glassy and relatively crystalline, with abundant bimodally zoned crystals of olivine, plagioclase, augite, and magnetite 24 .
We undertook full compositional characterization of these samples (i.e. major and trace elements, Sr-Nd isotopes and 238 U-230 Th-226 Ra-210 Pb disequilibria; Table 1) to distinguish between origins of ( 210 Pb/ 226 Ra) disequilibria by magma generation and differentiation in different tectonic settings and for different magma types.

Results
With the exception of the Serreta sample from the Azores, all of the samples analyzed here have < 3.6 wt.% MgO and < 50 ppm Cr indicating that they are significantly differentiated from their respective mantle-derived parental magmas ( Table 1). The sample from Yasur, Vanuatu is a basaltic trachyandesite (Fig. 1) with a bulk-rock major and trace element composition similar to those of other lavas erupted from this steady-state volcano 25,26 . Tonga samples from Hunga are low-K andesites with SiO 2 between 57.5 and 58 wt.%. Tonga samples from Home Reef and Fonualei samples are low-and medium-K dacites with 64.5 to 66 wt.% SiO 2 respectively. The glass in our sample of the Havre, Kermadec pumice is a medium-K rhyolite with 73 wt.% SiO 2 and an overall composition matching those of other pumice glass from this eruption 27 (Table 1) Table 1 from the island arcs and ocean islands illustrating the wide-ranging compositions analyzed here. See Fig. 1 and the text for sample locations. rhyolite has a greater offset, with relatively high 143 Nd/ 144 Nd and 87 Sr/ 86 Sr ratios (0.513115, 0.703705). Thorium isotopic compositions of the arc samples vary from 1.70 to 0.93, in line with regional variations 28,29 . Similarly, the ( 230 Th/ 232 Th) ratio in the benmoreite from Eyjafjallajökull is comparable to those of other silicic Icelandic lavas 22 .
In contrast, the Serreta basalt has ( 230 Th/ 232 Th) = 1.40, which is the highest ratio thus far measured for lavas from the Azores 30,31 . Uranium, Th, and Ra concentrations for the samples analyzed here vary over wide ranges (0.13 to 1.68 ppm, 0.13 to 5.05 ppm, and 116 to 654 fg/g respectively; Table 2). Uranium isotope ratios are within error of secular equilibrium for most samples. The exception is the Fonualei dacite with ( 234 U/ 238 U) = 1.016, which is attributed to seawater alteration. The arc samples have significant 238 U and 226 Ra excesses over 230 Th, with ( 230 Th/ 238 U) ranging from 0.54 to 0.93 and ( 226 Ra/ 230 Th) from 1.38 to 4.80 (Fig. 3). The Havre rhyolite has the least U-Th-Ra disequilibria of these samples. Sample Van A1 from Yasur also has relatively low ( 238 U/ 230 Th) and ( 230 Th/ 232 Th) values (0.83 and 1.46 respectively). These values are similar to those reported for lavas erupted from Yasur in 1975 and 1993 29 .
In contrast with the arc lavas, but like ocean island basalts in general, the Serreta basalt has 230 Th and 226 Ra excesses (( 230 Th/  Measured 210 Po activities in leached samples range from 0.10 to 1.4 dpm/g (see Supplementary Information  Table S1). This contrasts with the 3 to 14 fold higher activities in the leachates of the youngest samples (1.5 to 6.5 dpm/g) resulting from significant Po in sublimates adhered to vesicle walls and ash particles during eruption 37 . The inferred ( 210 Pb/ 226 Ra) 0 ratios of leached samples are all within analytical error of secular equilibrium with the exception of a 14% 210 Pb excess in the Home Reef sample and a significant 210 Pb deficit in the Serreta sample (( 210 Pb/ 226 Ra) 0 = 0.71; Table 2).
Our ( 210 Pb/ 226 Ra) 0 data for Iceland sample EJ-1 is within analytical error of the value published in Sigmarsson et al. 22 , although our measured concentrations of U, Th and Ra are lower by 6.5-15%, suggesting our fraction of the bulk tephra sample had higher proportions of mineral particles versus glass.  Table 2

Discussion
The data for young samples presented above are discussed in terms of geographical area and tectonic setting below. We go on to compare these data to the global dataset and address wider implications of short-lived radionuclide disequilibria in lavas.
Southwest Pacific arc samples. The samples from Hunga, Home Reef, and Fonualei in the Tonga arc have large 238 U and 226 Ra excesses that are typical for, indeed only found in, oceanic island arcs of the western Pacific 27,28,31,32 . These disequilibria, as well as the trace element patterns and elevated Sr isotope values, are inferred to reflect fluid addition of U, Ra, and Sr to the mantle wedge source from the subducting Pacific plate, in keeping with previous interpretations of U-series data for Tonga arc lavas 28,32 . The subducting Pacific plate beneath the Tonga arc is relatively cold 38 and the high H 2 O/Ce 39 , and low Th/La ratios for these lavas (Fig. 4) imply that fluids released from the Pacific plate are water rich and the subducting plate minimally melts. The Havre rhyolite marks an end-member isotopic composition for the Tonga-Kermadec system, with its unusual pairing of low ( 230 Th/ 232 Th) and high 143 Nd /144 Nd compared to oceanic basalts 34 and typical subducted lithologies 40,41 . These observations are consistent with derivation of most Nd from the local mantle, and most Th from subducting sediment.
The tight negative and positive correlations respectively between ( 226 Ra/ 230 Th) and ( 230 Th/ 238 U) versus Th/La for our arc samples (Fig. 4) suggests that much of the variation in 226 Ra-230 Th-238 U disequilibria relates to the degree to which Th is mobilized from the subducting plate rather than bulk assimilation of crust or time-scale of fractionation. The Home Reef dacite and Hunga andesites have similarly high ( 226 Ra/ 230 Th) ratios, consistent with similar century to millennium time frames of fractionation for these lavas 32 . The smaller 226 Ra excess but similar U-Th systematics for the Fonualei dacite compared with the other Tonga samples imply a modestly extended time-frame of fractionation, perhaps a few thousand years. The Havre rhyolite has comparatively low 238 U and 226 Ra excesses, a low ( 230 Th/ 232 Th) value, and a comparatively high Th/La value with respect to the Tonga lavas. We  Table 2. The black bars represent the range of Th/La in MORB glasses 42 . For samples from western Pacific arcs, the low ( 230 Th/ 238 U) and high ( 226 Ra/ 230 Th) values associated with samples with high Th/La suggests that these ratios largely reflect the degree to which Th is mobilized from the subducting plate. Symbols as in Fig. 1. Error bars are illustrated for values larger than the sample symbols. attribute all of these observations to a more significant transfer of Th from subducting sediment 28 , which resulted in less initial 238 U-232 Th-226 Ra disequilibria in the parental magmas for the Havre rhyolite compared with the Tonga arc, and a time frame of fractionation that also did not exceed several thousand years. Similarly, the incompatible trace element enriched basaltic trachyandesite from Yasur has moderate excesses of 238 U and 226 Ra over 230 Th compared with the Tonga arc samples, which might also reflect transfer of Th from subducted sediment 29 . Most lavas from the Tonga-Kermadec and Vanuatu arcs have ( 210 Pb/ 226 Ra) 0 ratios within analytical error of 1 ( Table 2, Fig. 5). The U-Th-Ra and trace element systematics outlined above indicate that the parental magmas for these lavas were water rich. The equilibrium ( 210 Pb/ 226 Ra) 0 values indicate that disequilibrium created by any process including fluid transfer from the subducting plate, melting, and deep degassing decayed away before eruption 9 . Moreover, the last stage of 222 Rn degassing must have occurred over a time period short enough to be undetectable using 210 Pb-226 Ra disequilibria 12 . The small 210 Pb excess in the Home Reef pumice is can be explained by Rn transfer in a gas phase to the dacite from un-erupted, more mafic recharge magma that may have triggered the eruption 14,48 . Atlantic Ocean island samples. The Serreta alkaline basalt from Terceira is characterized by an unusually high ( 230 Th/ 232 Th), but 230 Th and 226 Ra excesses that are similar in scale to those generally observed for alkaline ocean island basalts elsewhere, which we interpret to reflect the dynamics of melting garnet-bearing mantle [49][50][51] . The enrichment of U over Th in the source of this basalt that lead to the high Th isotopic composition, therefore, likely took place more than 350,000 years before magma genesis. This ancient U enrichment could have resulted from mantle carbonation 52 .
The benmoreite we analyzed from Eyjafjallajökull has a major element, trace element, and isotopic composition that is consistent with it representing a mixture between an alkaline basalt and an alkali rhyolite or trachyte 21,24 . Its low ( 226 Ra/ 230 Th) value (1.066) compared to a basalt erupted in 2010 from Fimmvörðuháls on Eyjafjallajökull's flank (1.368) 22 is most readily explained by several thousand years of aging during differentiation and storage of the silicic mixing end-member.
The initial ( 210 Pb/ 226 Ra) 0 ratio of 0.71 for the Serreta sample is similar to the values observed for most 10,34,35 but not all 48 ocean islands and basalts erupted in continental rift settings 45,46 . Basalts from Surtsey, Eldfell, and Fimmvörðuháls in Iceland have comparable 210 Pb deficits 22,46 . The origin of these 210 Pb deficits is a focus of the next section.
In contrast with Icelandic basalts, the benmoreite from Eyjafjallajökull has a ( 210 Pb/ 226 Ra) 0 ratio within error of equilibrium, whereas both the alkaline basalt and trachyte erupted at Eyjafjallajökull have 210 Pb deficits 22 , and mixing between these particular magmas could not have resulted in the equilibrium ( 210 Pb/ 226 Ra) 0 value of the benmoreite. Therefore, like the arc samples discussed above, the benmoreite erupted in 2010 appears to have resided in the crust long enough for any original 210 Pb-226 Ra disequilibrium to decay away. These data also suggest that the basalt that triggered the 2010 eruption was different than the one inferred to have mixed to generate the benmoreite (see also ref. 24). Pb excesses. We attribute these variations to differences and volatile profiles and modes of degassing in these different tectonic settings. See text for further explanation.
SCieNTiFiC REPORTS | 7:45186 | DOI: 10.1038/srep45186 210 Pb in gas-laden magma. Magma migration and degassing can strongly affect the ( 210 Pb/ 226 Ra) values of magmas because of the volatility of Rn. This volatility is illustrated by the near absence of 222 Rn in freshly erupted lavas from ocean island and arc settings 53 . In contrast, Pb is only weakly volatile 5 . Thus, to produce measurable 210 Pb deficits, magmas must continually lose Rn for a minimum of about two years. After a century of degassing, magmas will have ( 210 Pb/ 226 Ra) values close to zero. Comparatively longer durations of degassing are required to produce 210 Pb deficits of a particular magnitude if Rn loss is less than 100% efficient 5 or is discontinuous 14 . Excesses of 210 Pb over 226 Ra can be generated relatively rapidly by transfer of Rn-bearing gasses or fluids from greater volumes of magma and into smaller volumes of magma 48 . In this section, we explore whether Rn-Ra fractionation is the principal cause of variations in ( 210 Pb/ 226 Ra) values for magmas from both ocean island and arc settings.
The relative concentrations of the volatile species CO 2 and H 2 O in parental magma impact how and when a magma degasses and the consequences of this degassing on its physical properties. Both CO 2 and H 2 O concentrations in primary ocean-island basalts have been estimated to be from several tenths of a percent up to 1 wt.% 54,55 . Volatile concentrations in basalts from the Azores and Iceland appear to be in keeping with these values. Olivine-hosted melt inclusions from the basalt lava balloons erupted in 2001 from Terceira, Azores preserve water contents of up to 0.9 wt.% H 2 O and 1500 ppm CO 2 19 . Olivine-hosted glass inclusions in basalts from Pico island near Terceira have as much as 0.4 wt.% CO 2 , implying even higher concentrations in parental magmas 52  Primitive basalts with several tenths of a percent CO 2 in both ocean island and arc settings would typically saturate in a CO 2 -rich vapor phase and potentially begin open-system degassing at deep crust or upper mantle depths 55,58 . CO 2 dominated degassing has little effect on mineral stability and magma viscosity 61,62 . H 2 O largely remains dissolved in magmas until the upper several km of crust, where it becomes considerably less soluble 63 .
Open-system loss of H 2 O from silicate magmas in the upper crust strongly affects viscosity, because it lowers magma temperatures, enhances polymerization, and causes bubble and mineral growth. Latent heat released by crystallization can mitigate the viscosity increase by raising the temperature of a magma 64 , but overall, the loss of a water-rich volatile phase from magmas typically increases viscosity by orders of magnitude 65 .
Volatile phase saturation also pressurizes the system by bubble growth, creating a driving force towards eruption that is moderated by open system loss of the gas phase 65 . It is the unsteady interplay between changes in viscosity, pressurization, and conduit/reservoir geometry that determines whether water loss will cause a magma to freeze and stall out in the crust, or to migrate rapidly to the surface and erupt [66][67][68] .
Magmas with relatively low H 2 O/CO 2 ratios, such as those found in most ocean islands, typically arrive to the surface environment sparsely crystalline and only modestly differentiated. The subaqueously erupted Serreta basalt is an example of such a magma. If its 210 Pb deficit resulted from a steady loss of Rn in a CO 2 -rich gas phase, then its ( 210 Pb/ 226 Ra) 0 ratio of 0.71 suggests a rise time from the depth of CO 2 -saturation to the surface of 11 years assuming that Rn degassed with perfect efficiency 5 . The required maximum average magma velocity would depend on the initial CO 2 content of the magma. Assuming 1 wt.% H 2 O, 0.5 wt.% CO 2 and a crustal density of 3000 kg/m 3 , saturation of the gas phase would occur at about 20 km 69 . The maximum average velocity of Serreta magma in this case would have been about 1.8 km/yr. The ( 210 Pb/ 226 Ra) 0 value for the Serreta basalt is nearly identical to the median value for ocean island basalts in general 10 , suggesting that this magnitude of rise time and magma velocity is typical for ocean island basalts in general.
An alternative explanation for 210 Pb deficits in ocean island basalts and MORB is Ra-Pb fractionation during melting. This explanation is based on weak correlations between ( 210 Pb/ 226 Ra) and geochemical parameters that vary with changes in degree of melting, such as ( 226 Ra/ 230 Th) 3,10 . However, recent studies of ( 210 Pb/ 226 Ra) variations in lavas from individual ocean islands 35,36 , continental rifts 44,45 , and mid-ocean ridges 44 have not supported melting as the primary cause of ( 210 Pb/ 226 Ra) variations. These studies have illustrated, for example, that although mineral-melt partitioning during magma generation should only produce 210 Pb deficits, some lavas from all tectonic settings have 210 Pb excesses (Fig. 5). Such excesses can result from crystal fractionation of K-feldspar and amphibole, but only in highly differentiated magmas such as phonolite 8 (Fig. 5), not in magmas that crystallize olivine, augite, and plagioclase like the Serreta basalt 19 .
Only rarely does shallow degassing of water from an ocean island magma cause it to stall out in shallow magma reservoirs for periods of time long enough to equilibrate ( 210 Pb/ 226 Ra) values. Portions of Iceland offer a physical setting where such magma stagnation and differentiation is common because of its rather water rich parental magmas, glaciation, thicker and somewhat older crust 70 . The benmoreite from Eyjafjallajökull appears to have stagnated at a depth in the crust of about 1.7-5.0 km 71 without persistent Rn degassing as evidenced by its equilibrium ( 210 Pb/ 226 Ra) ratio.
As is the case for ocean island basalts, primitive arc magmas must typically begin to saturate in a CO 2 -rich volatile phase at deep crust or mantle depths. The rare arc basalts that erupt with 210 Pb deficits 10 reflect this process. However, most arc magmas stall in the crust and differentiate significantly. For parental magmas with 2-6 wt.% water, open-system degassing of water-rich vapor phases begins at depths of 1.5-12 km, which match the observed storage depths for magmas in volcanic arcs 60 . The Yasur lava is likely at the shallow end of this range based on the probable lower water contents 25 for its parental magmas compared with those from Tonga.
Once arc magmas stall, they must typically maintain ( 222 Rn/ 226 Ra) equilibrium for most of the time spent in the reservoir system, which means these magmas must not undergo persistent open-system degassing. Pulses of 222 Rn loss by degassing as magmas resided in the crust are allowed, so long as these pulses were spaced in time long enough to prevent measurable decay of 210 Pb. Numerical modeling of steady-state systems undergoing periods of complete 222 Rn loss followed by periods of repose with ( 222 Rn) = ( 226 Ra) indicates that such systems will have: ( 210 Pb/ 226 Ra) = 1/(1 + t d /t r ) where t d is the time the system spends degassing and t r is the time spent in repose. In this circumstance, values of ( 210 Pb/ 226 Ra) remain within analytical error of equilibrium only if t d /t r is less than about 0.03, indicating that any magma with equilibrium ( 210 Pb/ 226 Ra) must not lose Rn for the vast majority of its residence time in the crust.
An alternative explanation is that the fluxes of Rn in and out of the magma are equal. Of these two scenarios, we favor the former because a fortuitous balance between 222 Rn lost and gained seems unlikely to have occurred in multiple systems. Another alternative is that that 222 Rn supplied from recharge magmas mutes any 210 Pb deficit that exists in a resident magma due to degassing. However, this explanation requires magma renewal rates approaching 1 magma volume/year 5 , which basically just explains why equilibrium ( 210 Pb/ 226 Ra) values are maintained in magmas as they flow rapidly through conduits to the surface. Thus, we conclude that the near equilibrium ( 210 Pb/ 226 Ra) values for eruptives from Yasur, Hunga Ha'apai, Home Reef, Havre, Fonualei and Eyjafjallajökull require that their parental magmas stalled and differentiated in the crust without persistent open-system degassing for more than a century before eruption. For all of these systems, any open-system loss of 222 Rn to a gas phase during the final rise toward the surface must have occurred over a period of less than two years to prevent ( 210 Pb) from being measurably less than ( 226 Ra).

Conclusions
Parental magmas in arc and ocean island settings typically have concentrations of 0.3-1.3 wt.% CO 2 , which leads to the onset of degassing of magmas in the upper mantle or deep crust. This degassing leads to a steady rise of these magmas and a persistent Rn loss that results in the common 210 Pb deficits in ocean island settings. The Serreta basalt analyzed here has a near-median ( 210 Pb/ 226 Ra) 0 value for an ocean island basalt implying a minimum of 8.5 years of rise to the surface at a velocity as high as about 2.4 km/year after it first saturates in a gas phase. For ocean island basalts, significant H 2 O degassing typically occurs too shallowly to significantly impede their progress toward the surface. An exception appears to be Eyjafjallajökull, where magmas ponded and differentiated to trachytic to rhyolitic compositions within the relatively thick and glaciated crust. Its system of magma chambers must be relatively complex, because the equilibrium ( 210 Pb/ 226 Ra) 0 value for the benmoreite suggests that basaltic and silicic endmembers mixed more than a century before it erupted.
Arc magmas typically have significant subducted H 2 O 58,60 , and open-system degassing of this water in the middle to upper crust typically will cause magmas to freeze and stall 65 . The commonplace equilibrium ( 210 Pb/ 226 Ra) values in arc magmas suggest that once these magmas stall, they must not persistently lose or gain Rn for long enough periods or at high enough rates to affect ( 210 Pb) values. The further implication is that water-saturated magmas do not generally undergo persistent open-system degassing. Remobilization of these stalled magmas might occur by melt-crystal separation or by pressurization caused by magma recharge 72 .
Water-rich magmas must rise to the surface while degassing over periods of less than 2 years from their final staging reservoir to maintain ( 210 Pb/ 226 Ra) values within analytical error of secular equilibrium. Rise times for water-saturated magmas between staging reservoirs could be similarly rapid. In this circumstance, equilibrium ( 210 Pb/ 226 Ra) values will be maintained throughout magma systems that are saturated in a water-rich volatile phase. Quiescent final staging reservoirs and rapid rise times to the surface implied by equilibrium ( 210 Pb/ 226 Ra) values in water-laden magmas suggests an inherent lack of long-term predictability of explosive eruptions.

Methods
Major elements. Hand specimens were crushed using a stainless steel jaw press and the chips were then ultrasonically washed in MilliQ H 2 O to remove any seawater contamination. The Havre rhyolite was washed with weak HCl and water. The washed chips were subsequently dried and then powdered in an agate mill. Major element concentrations were determined on a Siemens ® SRS300 XRF at the University of Auckland, following standard techniques 73 . Fresh glass was analyzed for the Havre rhyolite. A polished thin section was carbon coated and analyzed by a Cameca SX100 electron microprobe at Macquarie University. A defocused beam was used with an accelerating voltage of 15 kV and a beam current of 15 nA. Counting times of 10 s were used for both peak and background measurements. The analysis in Table 1 represents the average of 5 spots. Spectrometer calibration was achieved using the following standards: Jadeite (Na), Fayalite (Fe), kyanite (Al), olivine (Mg), chromite (Cr), spessartine garnet (Mn), orthoclase (K), wollastonite (Ca, Si) and TiO 2 (Ti).
Trace elements. Sample powders were dissolved in a HF-HNO 3 mixture for trace element analysis. Most samples were analyzed at Macquarie University using an Agilent ® 7500CS ICP-MS following procedures outlined in Eggins et al. 74 . Detection limits during the period of study are listed in the Supplemental Materials Table S2. Sample EJ-1 was analyzed using a Thermo X-series II ICP-MS at the University of Iowa using methods described in Peate et al. 75 .
U-series. Samples were spiked with 236 U-229 Th and 228 Ra tracers and dissolved using an HF-HNO 3 -HCl mix in heated teflon pressure bombs. Samples were brought into solution, dried, and U, Th and Ra were separated chromatographically 76 . U and Th concentrations and isotope ratios were measured in dynamic mode on a Nu Instruments ® MC-ICP-MS at Macquarie University. Mass bias was determined assuming 238 U/ 235 U = 137.88 and the IC0 gain was determined during interspersed dynamic analyses of CRM145 assuming a 234 U/ 238 U ratio of 5.286 × 10 −5 77,78 . Radium aliquots were loaded onto degassed Re filaments using a Ta-HF-H 3 PO 4 activator solution 78 and 228 Ra/ 226 Ra ratios were measured to a precision typically ~0.5% in dynamic ion counting mode on the TIMS at Macquarie University. Most of the samples were analyzed with TML and BCR-2 standards 79,80 between 2008 and 2011 (see Supplementary Information Table S2). For the Serreta and Havre Ra analyses that were performed in 2015, the 228 Ra concentration of the spike was determined at the same time as the unknowns by analyzing the TML rock standard and assuming secular equilibrium. Analysis of BCR-2 performed at the same time yielded values that were within 3% of secular equilibrium (see Supplementary Information Table S2). Estimated total analytical errors (2σ ) for activity ratios listed in Table 2 Table 1 are based on ( 210 Po) and assume secular equilibrium between 210 Pb and 210 Po (t 1/2 = 138 days) for samples analyzed more than 2 years after eruption. For samples Van-A1 and HH09-01, multiple measurements of 210 Po in-growth over about 3 years allowed for extrapolation to equilibrium ( 210 Pb) o values (see Supplementary Information Table S1). Errors quoted in Table 2 are 2σ . ( 210 Po) values measured for USGS standards BCR-2 and RGM-2 during sample analysis are in Supplementary Information Table S2.