The data presented here sample only a fraction of the total magmatism in the Galápagos islands, but do offer a current and dynamic picture of the plume and cover much of the compositional spectrum observed in the islands, from depleted (Wolf) to enriched (Sierra Negra/Cerro Azul). The restricted nature of the data set necessarily places limitations on the depth of analysis and absolute model solutions. However, it is noted that Wolf Volcano, Fernandina, and Sierra Negra are each described as geochemically “monotonous” (internally) [e.g., Geist and Teasdale, 2001; Geist et al., 2005], supporting the attempts at spatial investigations using a restricted data set.
4.1. Controls on U-Th Disequilibria
The 230Th excesses observed in most OIB (and MORB) are usually attributed to the presence of residual garnet during mantle melting [e.g., Beattie, 1993; Prytulak and Elliot, 2009]. However, it is noted that high-pressure residual clinopyroxene may also create significant 230Th excess in the melt [Wood et al., 1999]. The 230Th excesses and steep REE patterns of Galápagos volcanic rocks (Figure 2) provide support for a garnet-bearing source residue beneath the archipelago. REE constraints on the degree of melting (La/Yb) and amount of residual garnet (Tb/Yb) are shown in Figure 7. Assuming a similar starting composition for all volcanoes, partial melting of a primitive mantle (garnet-peridotite) source requires only a small amount of residual garnet (2%–4%) and small degree of melting (∼3%–5%) to fit the data (noting that the magma source for Wolf Volcano may be more depleted than primitive mantle). These values lie within the calculated range for the degree of partial melting (2.5% to ∼10%) and amount of residual garnet (2%–9%) proposed by Harpp and White  for the same volcanic centers, and the small degree of melting of MORB-source with ∼5% garnet proposed by Geist et al.  for Wolf Volcano. The tight cluster of the Galápagos data on La/Yb-Tb/Yb suggests that despite source compositional heterogeneities (see below), relatively similar degrees of melting and modes of residual garnet are involved at each volcano.
Correlations of (238U/232Th) and (230Th/238U) activity ratios with Nb/Zr (Figures 4c and 4f) and long-lived radiogenic isotopes (Figures 4b and 4e) may be accidental (if the fundamental behavior of each volcano is unique), but suggest that mantle source heterogeneity affects melting processes and the generation of U-series disequilibria in the Galápagos volcanic rocks. A relationship between Sr-Nd-Pb isotopes, element ratios and U-series disequilibria is also reported for GSC samples and is interpreted as mixing between an enriched (Central Galápagos plume) and depleted (upper mantle and/or Eastern Galápagos plume) component [Kokfelt et al., 2005]. The significantly more primitive Sr-Nd isotope ratios, lower LREE/MREE ratio, lower Nb/Zr (but still > MORB), and lower Th/U ratios of Wolf Volcano in the north of Isabela island may signify a larger contribution of depleted upper mantle/plume component compared to volcanoes further south [e.g., Harpp and White, 2001; Geist et al., 2005]. The question then arises whether the U-series disequilibria yield information on melting dynamics of the plume (e.g., melting rate and upwelling velocity), are primarily controlled by source heterogeneity and mixing, or a combination of the two.
Based on geophysical [Hooft et al., 2003] and He isotope [Kurz and Geist, 1999] data, the center of the Galápagos plume is thought to lie beneath or slightly southwest of Fernandina. Therefore, assuming similar source fertility, if melting rate were the primary control on (230Th/238U) disequilibria in Galápagos volcanic rocks, we might expect the highest melting rate, fastest upwelling and therefore, least disequilibria (lowest 230Th excess) at Fernandina. However, this is not what is observed; the lowest excess is observed at Wolf Volcano, north of Fernandina and toward the margin of the geochemically enriched plume trace. On the other hand, at 120–150 km depth the focus of a low velocity anomaly (interpreted as the locus of the upwelling plume) is centered beneath northern Isabela, i.e., Wolf Volcano [Villagómez et al., 2007]. Between 150 and 100 km depth, the plume conduit inclines northward and is located offshore between Wolf Volcano (Isabela) and Pinta [Toomey et al., 2002; Villagómez et al., 2007]. This depth range is compatible with melting of fertile/depleted garnet peridotite [Robinson and Wood, 1998], therefore it is actually reasonable to expect the fastest upwelling rate, and therefore, the lowest 230Th excess at Wolf Volcano.
Variations in the length of the melting column may contribute to the spatial variations observed in U-Th disequilibria. A longer melting column (corresponding to either a higher mantle temperature and/or more fertile source) should correlate with increased residence time in the melting column, and therefore higher 230Th excess. S-wave velocity data suggests that the calculated variation in lithospheric thickness beneath the four volcanoes of this study is relatively small (50–56 km, [Villagómez et al., 2007]) and REE inversion modeling and parameterization suggest that the top of the melting column is between 57 and 58 km beneath all the volcanoes of interest [Gibson and Geist, 2010]. Therefore, the effects of lithospheric thickness variation [Feighner and Richards, 1994; Villagómez et al., 2007] do not seem to be a significant control on the melting regime in this part of the archipelago. Consequently, in order to create a longer melting column, variable depths to the base of the melting column would be required. Assuming a similar source lithology and mineralogy, the volcano with the highest MREE/HREE ratio (indicating the greatest garnet involvement) may correspond to the deepest melt and therefore, the longest melting column. As a result, it would have the longest time for in-growth and so the largest 230Th excess [e.g., Bourdon and Sims, 2003]. However, this is not the case in the Galápagos rocks. We see an inverse relationship between Tb/Yb and (230Th/238U) activity ratios (cf. GSC lavas where a positive correlation is observed [Kokfelt et al., 2005]). Note that the trace element variations induced by differences in degree/depth of melting are likely to be small compared to the large variability in Nb/Zr ratios and radiogenic isotope ratios inherited from the source.
The extent to which variations in source lithology, such as varying proportions of mafic pyroxenite/eclogite and garnet peridotite contribute to the extent of U-series disequilibria is not easy to assess (e.g., see the discussion by Stracke et al.  and references therein). In a simple scenario, assuming comparable trace element fractionations (bulk DU/DTh ratios) for pyroxenite/eclogite and garnet peridotite during partial melting [Pertermann et al., 2004], a greater contribution of eclogite (and possibly pyroxenite) should correspond with relatively lower (230Th/238U) activity ratios (due to their higher average melt productivity) and generally higher SiO2 contents [Yaxley and Green, 1998; Walter, 1998; Pertermann and Hirschmann, 2003a; Hirschmann et al., 2003]. This general trend is observed in our limited Galápagos data set (Figure 4h), although we acknowledge that small variations in SiO2 can be caused by variations in pressure of melting. However, if eclogite/pyroxenite represent previously recycled oceanic crust or metasomatized material, their 87Sr/86Sr isotopic ratios would be expected to be higher than that of “normal” mantle peridotite. Clearly, this does not fit with the observed trend of increasing (230Th/238U) with increasing 87Sr/86Sr isotope ratio (Figure 4e). Also, the REE models (Figure 7) require only a small amount of residual garnet (2%–4%) to fit the Galápagos data, and may preclude an important role for eclogite/garnet-pyroxenite in the source, as these would be expected to have a significantly higher garnet mode. Furthermore, the highest SiO2 content in the new data set is still relatively low (48 wt%) compared to that expected in basaltic eclogite melts (>57 wt% SiO2 at ∼3 GPa [Yaxley and Green, 1998; Pertermann and Hirschmann, 2003a]).
The preceding discussion, argues against a significant contribution from eclogite in the source. Nevertheless, melting of garnet-pyroxenite has been inferred to be able to create 230Th excesses [Pertermann and Hirschmann, 2003b] and therefore, as an alternative interpretation, the Sierra Negra samples, which display generally the lowest SiO2, highest Sr isotope ratios, lowest U/Th ratio and highest 230Th excess, could be interpreted to reflect a higher contribution of a low-SiO2 pyroxenite melt, produced deeper (∼35–50 km) than that of peridotite (due to its lower solidus temperature [Pertermann and Hirschmann, 2003b]), resulting in a greater residence time in the melting column, and therefore higher resultant 230Th excess.
In summary, the correlation of U and Th isotopes with other geochemical parameters indicates a significant link between mantle source heterogeneity and melting dynamics. Based on the above discussion and present knowledge it is difficult to isolate a simple and geochemically consistent explanation for the correlations observed between long-lived radiogenic isotopes, trace element ratios and U-Th disequilibria. This is, perhaps, not surprising considering the range of possible source component and process combinations and the limited nature of the data set.
4.2. Observations From Shorter-Lived U-Series Isotopes
In contrast to U-Th disequilibria, the 226Ra excesses in Galápagos volcanic rocks do not correlate with volcano location, long-lived radiogenic isotopes, trace element ratios or other U-series activity ratios (e.g., Figure 5). If U-Th disequilibria is primarily controlled by source heterogeneity, this may suggest that storage and shallow-level effects (e.g., interaction with cumulates [Saal and Van Orman, 2004]) overprint those of the source on a timescale relative to the half-life of 226Ra (1599 years). Intriguingly, and to our knowledge for the first time in OIBs, initial (210Pb/226Ra) activity ratios of Galápagos volcanic rocks appear to correlate with volcano location, Nb/Zr and radiogenic isotopes (Figure 6). The apparent correlation of short-lived (210Pb/226Ra)0 disequilibria with long-lived Sr-Nd isotope variation is difficult to explain as it is usually assumed to reflect fractionation of 210Pb (or 222Rn) from the 226Ra parent within the last 100 years.
(210Pb/226Ra)0 activity ratios may be affected by several processes, such as: partial melting, sulphide fractionation, plagioclase accumulation, magma degassing or interaction with crustal cumulates [e.g., Rubin et al., 2005; Van Orman and Saal, 2009; Berlo and Turner, 2010, and references therein; Condomines et al., 2010]. 210Pb excesses are relatively uncommon in OIBs [Berlo and Turner, 2010]. 210Pb excesses in other volcanic settings are usually explained by 222Rn (a precursor of 210Pb) gas accumulation and fluxing preceding and/or during eruption [Berlo et al., 2004; Turner et al., 2004; Berlo and Turner, 2010; Condomines et al., 2010; cf. Kayzar et al., 2009]. In the Galápagos data set, the highest 210Pb excesses are found in 2 tephra samples, consistent with these being erupted during more gas rich phases of the eruption, but these samples also have higher (226Ra/230Th), and higher total Pb (Figure 5b and Table 1), suggestive of contamination, for example by Pb sublimates [Gauthier and Condomines, 1999]. Nevertheless, excluding these 2 samples, the correlations between 210Pb excesses and other isotope systems in the Galápagos data suggests that other processes capable of fractionating these systems may play a role.
In a recent review of 210Pb-226Ra disequilibria in volcanic rocks, sulphide fraction and accumulation of 210Pb-rich plagioclase were deemed unlikely protagonists as the primary cause of (210Pb/226Ra)0 variations [Berlo and Turner, 2010]. The latter case is supported by the unrealistic amount of plagioclase crystals required to create the observed excesses (e.g., ∼30% to create (210Pb/226Ra)0 of ∼1.4 [Berlo and Turner, 2010]). Up to 20% plagioclase accumulation is proposed in lavas of Wolf Volcano [Geist et al., 2005] owing to the large quantity of plagioclase phenocrysts in the lavas, their zonation patterns and anomalously high Sr and Eu concentrations in high Al2O3 content (>15 wt%) rocks. Plagioclase accumulation is also thought to be important in controlling major element contents at Fernandina [Allan and Simkin, 2000] where plagioclase phenocrysts occasionally comprise up to 40% of the lava. Despite Al2O3 contents >15 wt% in 3 samples of this study (Table 1), none of the samples show particularly anomalous Eu behavior (Figure 2). For plagioclase accumulation to explain the 210Pb-226Ra disequilibria, the plagioclase phenocrysts would have had to be very young (decades or less), require extreme initial (210Pb/226Ra), and for Wolf Volcano, much more than 20% accumulation of zero age crystals would be required to produce the observed (210Pb/226Ra) of 3.8. Therefore, plagioclase accumulation is not our preferred model to account for (210Pb/226Ra)0 disequilibria in Galápagos volcanic rocks.
Pb is more compatible than Ra in all likely mantle minerals [Blundy and Wood, 2003]. Therefore, it is extremely difficult to explain 210Pb excess by partial melting in the mantle [e.g., Rubin et al., 2005]. This would require a residual phase with DRa > DPb, such as phlogopite, which has so far only been implicated as a possible residual phase at Floreana [Bow and Geist, 1992], southeast of the study region.
Saal and Van Orman  suggest that high 226Ra excesses in oceanic basalts are produced by diffusive exchange between new batches of magma and previously crystallized cumulate minerals (plagioclase and/or clinopyroxene) in the crust However, this process should produce 210Pb deficits, not excesses in the melt and therefore, is inconsistent with the Galápagos Ra-Pb isotope data.
At face value, the correlation of (210Pb/226Ra)0 with Zr/Nb and long-lived radiogenic isotope suggests that 210Pb excess is related to melting and source variations (similar to the conclusion reached by Kurz and Geist  for helium isotopes of Galápagos basalts) rather than degassing and/or shallow level processes. However, as shown above, partial melting of typical mantle mineralogies is not consistent with 210Pb excesses. If 210Pb excesses are related to 222Rn gas accumulation at shallow levels in a closed system model [e.g., Condomines et al., 2010] for example, via the stalling of ascending magma as it encounters a more viscous mush layer, it is not obvious why 210Pb-226Ra disequilibria might correlate with source heterogeneity geochemical indicators. Perhaps volatile content variability is related to source heterogeneity? It is also not clear why the most primitive lavas possess the highest 210Pb excesses. Furthermore, for 210Pb excess to correlate with long-lived geochemical features, it requires decoupling from Ra (the parent nuclide), suggesting possible segregation of Pb or an intermediate daughter (e.g., Rn) from Ra in the melt prior to modification of Ra-Th disequilibria at shallow levels. More data is clearly required to understand the fascinating relationships between long-lived and short-lived U-series isotopes and other geochemical data.