There has been much debate as to whether 210Pb-226Ra disequilibria in young volcanic rocks result from partial melting, cumulate interaction or magma degassing. Here we present new data from basalts erupted in 1978 from Ardoukoba volcano in the Asal Rift. The (210Pb/226Ra)t ratios are very low (0.2 to 0.6) and appear to correlate negatively with (226Ra/230Th). Invariant (230Th/238U) and (231Pa/235U) ratios require similar melting rates, porosities, and extents for all parental magmas. Thus, the range in (226Ra/230Th), which is negatively correlated with Th concentration, reflects fractional crystallization over millennia after the magmas were emplaced into the crust. This precludes the 210Pb deficits from resulting from partial melting. Instead, the 210Pb deficits must have formed subsequent to magma differentiation and are interpreted to reflect several decades of magma degassing. Many young basalts erupted in a variety of tectonic settings are similarly depleted in 210Pb with respect to 226Ra, suggesting that they continuously degas over a period of a few to several decades, perhaps reflecting the time required to rise to the surface from deeper reservoirs. In some of these basalts, gas accumulation leads to the shallowest, most evolved, and earliest erupting magmas having the highest (210Pb/226Ra) ratios and sometimes 210Pb excesses.
 In contrast, Rubin et al.  observed a negative correlation between (210Pb/226Ra) and (226Ra/230Th) among young oceanic basalts from the East Pacific Rise, Juan de Fuca Ridge and Loihi and concluded that this resulted from their genesis by partial melting of the mantle. By implication this would require melt extraction from the mantle and migration to the surface within decades [see also Sigmarsson, 1996; McKenzie, 2000]. Alternatively, Van Orman and Saal  have suggested that 210Pb deficits could result from diffusive processes during interaction between melts and cumulates in crustal magma chambers. Finally, fractional crystallization of phases with a greater affinity for Pb over Ra, such as a sulfide, could also potentially lower (210Pb /226Ra) in basaltic melts.
 Clearly, it is necessary to distinguish between the above possibilities before the full implications of 210Pb disequilibria for magmatic timescales and processes can be realized. A useful prerequisite is a suite of basalts erupted in a short period in a well-constrained and simple tectonic environment. Accordingly, we present here 210Pb-226Ra data for basalts from the Asal Rift and show how these provide new insights into the origins of such signals and magmatic plumbing systems.
2. The 1978 Eruption of Ardoukoba, Asal Rift
 Over the course of a week in November 1978 a series of basaltic lavas flows erupted from Ardoukoba volcano in the central Asal Rift of the African triple junction in Djibouti. This event afforded a rare opportunity to investigate very short-lived U-series disequilibria in basaltic magmas produced by decompression associated with rifting – somewhat like a subaerial analogue of a nascent mid-ocean ridge. Eruption was accompanied, and followed, by ∼2 months of significant seismic activity and normal faulting inferred to be related to magma injection in the form of dykes [Jacques et al., 1996]. Geophysical data suggest the existence of a single, southeast-dipping, magma chamber at ∼2–4 km depth beneath the Asal Rift [Van Ngoc et al., 1981]. The lavas were extruded along fissures that migrated from northwest to southeast over the week and the samples reported in Table 1 are ordered in eruption sequence from earliest to latest. All of the samples were collected within days of the eruption and are very fresh, highly vesicular, plagioclase-phyric basalts containing minor olivine and clinopyroxene. Vigier et al.  inferred that the lavas had accumulated ∼30% plagioclase.
Table 1. 226Ra and 210Pb Data for 1978 Asal Rift Samples in Order of Eruptive Sequence (Earliest to Latest)a
 Major and trace element data, along with Sr and U-Th-Pa-Ra isotope data for a suite of samples from the eruption were presented by Vigier et al. . They found that 87Sr/86Sr, (230Th/232Th), (230Th/238U), and (231Pa/235U) ratios are largely invariant (Figures 1a and 1b) but that (226Ra/230Th) ratios ranged from 1.93 to 1.35 (Figure 1c). Vigier et al.  used these data to infer minimum differentiation timescales for several related magma batches ranging from 320 to 2570 or 650 to 6730 years, using closed- and open-system fractional crystallization models, respectively. The magma residence times were related to the extent of fractional crystallization as reflected in concentrations of highly incompatible elements (e.g., Th), such that the longest residence times were inferred for the most evolved magmas which were the first erupted, whereas shorter residence times were inferred for the less evolved magmas that erupted last.
3. Analytical Techniques
 The materials analyzed here were aliquots of the same whole rock powders used by Vigier et al.  to which the reader is referred for additional details and bulk compositional data. Replicate trace element and U-series analyses performed by Vigier et al.  demonstrate the homogeneity of the powders, even between different samples from the same flow.
 In order to complete the 226Ra data reported by Vigier et al. , sample 546 was analyzed for 226Ra at Macquarie University following chemical separation and analytical methodologies described in Turner et al. , which are essentially the same as those used by Vigier et al. . For the purposes of this paper, the 226Ra concentration errors for all the samples are assumed to be no better than that quoted for the NIST 4969 226Ra standard (±1.8%) against which the 228Ra spikes were calibrated, even though within run errors were typically ≤0.5% [see also Vigier et al., 1999].
 The (210Pb) values are based on (210Po) measurements preformed at the University of Iowa in 2010 and 2011 following the methodology described in Reagan et al. . Errors quoted in Table 1 are 2σ. (210Po) measured for BCR-2 at the same time as the samples was 1.26 ± 0.05 dpm/g. Initial (210Pb/226Ra)t ratios were corrected for the 32 to 33 year interval between eruption and analysis according to the equation (210Pb/226Ra)t = 1 + [(210Pb/226Ra) − 1] ⋅ eλPbt. The age propagated error on the (210Pb/226Ra)t ratio was thus expanded to be the error on the calculated (210Pb/226Ra) ratio, calculated as ((210Pb/226Ra) × √((err Ra/Rameas∧2) + (err Po/Pomeas∧2))), multiplied by eλPbt [see also Condomines et al., 2010].
Table 1 presents our new 226Ra (sample 546) and 210Pb data to complement the data set provided by Vigier et al. . Figure 1c shows that the calculated (226Ra/230Th) ratio for sample 546 conforms to the negative correlation between (226Ra/230Th) and Th concentration originally observed by Vigier et al. . The measured 210Pb activities vary from 0.31 to 0.16 dpm/g and the resultant eruption age-corrected (210Pb/226Ra)t ratios range from 0.56 ± 0.15 to 0.20 ± 0.16. These are lower than observed in most volcanic rocks [Berlo and Turner, 2010]. Overall, the (210Pb/226Ra)t ratios largely overlap within error. However, as shown in Figure 2a, the most probable values correlate negatively with (226Ra/230Th) and overall decrease from earliest to latest erupted lava (see Table 1). SiO2 contents and the overall extent of fractional crystallization, as indicated by the abundances of incompatible elements like Th (Figures 1c and 2b) and Rb (not plotted), also decreased during the course of the eruption Vigier et al. . In the following discussion, we sequentially examine possible models for the origin of the large 210Pb deficits that are observed in all of the Asal lavas. We then consider the wider implications of the data for current models for the timescales of basaltic processes.
5. Origin Through Partial Melting?
 Although the absolute magnitude of the (210Pb/226Ra)t and (226Ra/230Th) ratios are lower and higher, respectively, than the oceanic basalt data presented by Rubin et al. , the relationship between (210Pb/226Ra)t and (226Ra/230Th) in Figure 2a is the same (excluding 2 samples, the oceanic data have (210Pb/226Ra)t = 0.86–1.1 and (226Ra/230Th) = 1.9–2.9, respectively). For a typical mantle peridotite assemblage the bulk partition coefficients for all of the elements in the U decay-series are <0.01, and the order of compatibility is Pb > U > Th ≫ Pa > Ra [e.g., Blundy and Wood, 2003] and so mantle melting should lead to deficits in 210Pb combined with excesses in 226Ra. Using parameters appropriate to the Asal Rift situation, a simple dynamic melting model [Williams and Gill, 1989] can simulate the range in (226Ra/230Th) ratios by varying porosity during melting, although it predicts slightly lower (210Pb/226Ra) ratios than observed (Figure 2a). This difference between measured and observed (210Pb/226Ra)t ratios could easily be reconciled if ∼10–20 years elapsed between partial melting and eruption (see Figure 2a).
 A similar observation was made by Rubin et al.  who concluded that their data required melt extraction from the mantle and eruption within a few decades. This would obviously place important constraints on the dynamics of melt movement (see Dosseto et al.  for a recent review). Apart from being subaerial and continental, the Asal Rift setting bears similarities with that of the East Pacific Rise and Juan de Fuca Ridge (but not Loihi) and so our new data might be seen to provide new support for the conclusions of Rubin et al. . However, while (230Th/238U) ratios vary in their oceanic basalts, a key aspect of the Asal Rift basalts is that their (230Th/238U), (230Th/232Th), and (231Pa/235U) ratios are essentially invariant (Figures 1a and 1b).
 In dynamic melting [e.g., McKenzie, 1985; Williams and Gill, 1989] or chromatographic porous-flow melting [Spiegelman and Elliott, 1993] models, (230Th/238U) ratios are most sensitive to the melting rate whereas (231Pa/235U), (226Ra/230Th) and (210Pb/226Ra) ratios are more sensitive to the porosity of the melting region [Sims et al., 1999; McKenzie, 2000; Elliott and Spiegelman, 2003]. If the 210Pb deficits of the Asal Rift basalts resulted from partial melting, the magmas must have erupted within decades of formation. As a consequence, any variation in disequilibria among the other nuclide pairs must primarily reflect the melting process. However, if we ascribe the large variation in (226Ra/230Th) ratios in Figure 2a to melting, instead of decay during fractional crystallization (see below), any successful dynamic melting simulation results in a coupled variation, and range in absolute values, of (230Th/238U) and (231Pa/235U). Figures 1a and 1b show that the variations predicted by such models are simply not observed in the data. Moreover, because Ra is more incompatible than Th, progressive melting of a mantle source should cause Th concentrations to decrease along with (226Ra/230Th), which is also the opposite of what is observed. Indeed, there is no melting model that can simultaneously reproduce the variation in the data from all four parent-daughter nuclide pairs (Figures 1 and 2a) in magmas generated from the same source. Therefore we concur with Vigier et al.  that the variation in (226Ra/230Th) post-dates partial melting and, by implication, the same must apply for the 210Pb-226Ra disequilibria. Accordingly, mantle partial melting is not the origin of the measured 210Pb deficits, at least in the case of the Asal Rift basalts. In the model illustrated in Figure 1, the combined 238U-230Th, 235U-231Pa and 226Ra-230Th disequilibria reflect a series of melts generated from a relatively homogeneous source under nearly identical conditions (e.g., at a fixed melting rate, porosity, and extent of melting) followed by fractional crystallization over millennia.
6. Fractional Crystallization
 At relatively low porosity, the dynamic melting model shown in Figure 1 can simulate the (238U-230Th) and (235U-231Pa) ratios of the Asal Rift basalts (note the highly expanded x axis scale). In Figure 1c, the same model can replicate sample (546), which has the highest (226Ra/230Th) ratio and lowest Th concentration (i.e., the least differentiated). This suggests that melt ascent from the mantle occurred sufficiently fast (<8 kyr) to preserve the 226Ra excess in this sample. As discussed and modeled in detail by Vigier et al. , the trend toward lower (226Ra/230Th) ratios and higher Th concentrations is consistent with 30–50% fractional crystallization over a timescale that was (1) insignificant relative to the half-life of 230Th (75 kyr), (2) similar to the half-life of 226Ra (1600 yr) and (3) much longer than the half-life of 210Pb (22.6 yr). The alternative explanation, of lowering the (226Ra/230Th) values and raising Th concentrations in magmas by bulk assimilation of crustal materials, is considered not to be viable because Th and Sr isotopic compositions are homogeneous in the Asal lavas [Vigier et al., 1999]. Thus, although the mantle-melting model predicts both 226Ra excesses and 210Pb deficits in the primary melts, any 210Pb disequilibria would probably have decayed away during ascent and clearly would have been fully erased during the time scales inferred for fractional crystallization. In other words, the 210Pb deficits were generated by a process that post-dates both melt ascent from the mantle and subsequent fractional crystallization. This also precludes the origin of the 210Pb deficits by sulfide fractionation. In any case, sulfide saturation is likely to occur in response to redox changes caused by the onset of magnetite fractionation [Jenner et al., 2010]. Increases in TiO2 and Au with increasing extent of fractional crystallization in the Asal Rift basalts (not shown) independently indicate that neither magnetite nor sulfide fractionation occurred in these magmas.
7. Interaction With Cumulates?
Van Orman and Saal  suggested that a possible explanation of 210Pb deficits in basaltic magmas lies in diffusive interaction with cumulates in crustal magma chambers. In this model it is the rapid decay of 210Pb and its greater compatibility and more rapid diffusion in plagioclase and clinopyroxene, relative to 226Ra, that results in the potential for cumulates to become an internal sink for 210Pb. Numerical calculations suggest that such a process could, in principle, develop (210Pb/226Ra) ratios <0.5 within a few decades [Van Orman and Saal, 2009]. In practice, such models remain hard to test in individual magmas not least because chromatographic effects during porous flow through crystals with which they are in equilibrium have little effect on overall magma compositions [e.g., Navon and Stolper, 1987]. For the Asal Rift lavas, we consider it highly unlikely that their 210Pb deficits result from magma – cumulate interaction for the following reasons. First, different levels of crystal fractionation from nearly identical parent magmas explains the variation in the bulk compositions of the Asal basalts, and most if not all of this crystal fractionation and crystal liquid separation occurred prior to the generation of the measured 210Pb deficits. Second, interaction between magmas migrating through interconnected pores in plagioclase-rich cumulates derived from similar magmas should produce (210Pb) < (226Ra) < (230Th) rather than the observed (210Pb) < (226Ra) > (230Th). Clinopyroxene rich cumulates could produce the correct sequence of relative activities, but would not have the leverage to produce the large 210Pb deficits seen in the Asal lavas. Third, the greatest 210Pb deficits are found in the Asal lavas with the greatest enrichments in Sr and Eu and thus the greatest amount (∼30%) of accumulated plagioclase [Vigier et al., 1999]. According to the Van Orman and Saal  model, plagioclase should have significant 210Pb excess such that plagioclase accumulation would mitigate any original 210Pb deficits in these lavas and potentially impart 210Pb excesses. We observe the opposite.
 The decrease of (226Ra/230Th) toward secular equilibrium with increasing fractional crystallization in the Asal Rift basalts in Figure 1c is similar to many other intravolcano U-series data sets [e.g., Blake and Rogers, 2005]. Figure 2b shows that (210Pb/226Ra)t broadly increases toward secular equilibrium with increasing fractional crystallization in these rocks (cf. the most evolved magmas were erupted first). As discussed above, the very different half-lives of these two nuclide pairs requires that these reflect different processes. Since neither partial melting or interaction with cumulates can readily explain the Asal rift 210Pb-226Ra disequilibria and the latter must have formed after millennial scale fractional crystallization, we infer that the 210Pb deficits result from magma degassing subsequent to differentiation. It is possible that differentiation occurred at depth and degassing occurred in the magma chamber imaged 2–4 km beneath the rift.
 Using the formulation of Gauthier and Condomines , we estimate that the Asal Rift basalts underwent degassing for at least 20–50 years prior to eruption (Table 1). Note that these calculations assume 100% efficient, continuous degassing of 222Rn and that less efficient degassing would increase these estimated durations. Figure 2c shows that the inferred duration of degassing may decrease with increasing fractional crystallization. Unfortunately, the errors on the (210Pb/226Ra)t values in the basalts propagated for the age of eruption (cf. Figures 2a and 2b) preclude a definitive statement about the relative degassing timescales or degassing efficiencies for the more and less differentiated lavas. Nevertheless, when combined with other observations there appear to be some generalities emerging from 210Pb systematics, as we now discuss.
9. Wider Implications
 The best explanation for the 210Pb deficits with respect to 226Ra in the Asal Rift basalts is that their magmas degassed for two to five decades before they erupted. Similar deficits have been documented in basaltic lavas from continental [e.g., Chakrabarti et al., 2009] and oceanic [Rubin et al., 2005] rift environments and even in some subduction related lavas [e.g., Turner et al., 2004]. Thus, there is a growing body of literature suggesting that basalts can degas persistently for decades before they erupt. This degassing could be related to depressurization and slow segregation of CO2, H2O, SO2 and other volatile species as the magmas rise toward the surface. By implication, the same might be true of the oceanic basalts discussed in Rubin et al.  and those from the East Pacific Rise appear to be CO2 super-saturated and may therefore have been undergoing prolonged degassing.
 It has been suggested that even highly voluminous magmas may assemble only decades prior to eruption [e.g., Druitt et al., 2012]. Since degassing will inevitably lead to crystallization [e.g., Cashman and Blundy, 2000; Brophy, 2009] it is, in principle, also possible that degassing and fractional crystallization might be linked. Our new data from the Asal basalts are similar to results from previous studies of Vestmannaeyjar volcano in Iceland [Sigmarsson, 1996] and Sangeang Api in Indonesia [Turner et al., 2004], in that it is the early erupted ± more differentiated magmas from single eruptions that have the higher (210Pb/226Ra) ratios. This is the opposite of what would be expected if degassing promotes crystallization. The evidence at Asal is that fractional crystallization occurred prior to degassing and such observations are not restricted to extensional settings. For example, Turner et al.  reported 210Pb deficits from Sangeang Api volcano in the Sunda arc where differentiation is inferred to have occurred over millennia [Turner et al., 2003].
 In this scenario, inferred correlations between the calculated duration of degassing and the degree of fractional crystallization (e.g., Figures 2b and 2c) do not require that these two processes are coupled. Rather, it may be that a common result of magmatic plumbing systems is that gas accumulates in the more differentiated magmas that reside at the top of magma reservoirs and are the first to erupt. In some instances, such gas accumulation can even lead to these magmas having 210Pb excesses like those observed at Arenal volcano [Reagan et al., 2006]. We infer that the Asal Rift magmas rose toward the surface and ponded for 1–3 decades [e.g., Gauthier and Condomines, 1999] before eruption. The uppermost, more differentiated magmas were fluxed by gas from below, reducing their 210Pb deficits. Another possibility is that the more differentiated magmas degassed less efficiently. Either way, the degassing timescales are clearly longer than, and therefore unrelated to, the week-long period of eruption and subsequent months of seismicity. Both of these magmatic processes occurred on significantly shorter timescales than the development of the bulk compositional variation within the magmas that was established by fractional crystallization over the preceding millennia.
 We are grateful to Kim Berlo and heather Handley for many discussions about 210Pb disequilibria and to Jim Gill and Jim Van Orman for their constructive reviews. This work was directly funded by Australian Research Council Discovery Proposal DP0988658 to S.T. and M.R. and used instrumentation funded by ARC LIEF and DEST Systemic Infrastructure Grants, Macquarie University and Industry. M.R. also thanks NSF grant EAR 0738776 for support. This is GEMOC publication 188/831.