Ilopango Caldera was formed ∼1810 years ago by eruption of the Tierra Blanca Joven (TBJ) dacite (70 km3) in central El Salvador. A subsequent eruption in 1880 produced a cluster of dacite domes in the center of Lago Ilopango that contain olive-bearing enclaves of basaltic andesite. The purpose of this study is to use trace element, isotope, and U-series data from the TBJ and 1880 eruptions to assess petrogenesis and the timescale of magma storage. We find that although the range of trace element data in the TBJ dacite can be reproduced by simple crystal fractionation of a plagioclase- and amphibole-rich mineral assemblage, the87Sr/86Sr and 207Pb/204Pb data suggest that the 1880 basaltic andesite enclave has a different source than the dacites. This is consistent with U-series data that show the TBJ dacites have lower (230Th/232Th) than the 1880 basaltic andesite enclave (1.5 versus 1.6, respectively). All Ilopango rocks have 230Th excesses, and the range in (238U/232Th) of the TBJ dacites can be modeled by crystal fractionation of a mineral assemblage including accessory zircon and allanite from a magma that is similar in composition to the enclave. Mineral isochrons yield crystallization ages of <10,000 years, and disequilibrium (226Ra/230Th) values suggest a similarly short residence time. Our data suggest that the large volumes of erupted dacite at Ilopango Caldera are generated very rapidly.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 In this paper we present new geochemical data from the Ilopango TBJ dacites and the post-caldera lavas and mafic enclaves, including U-series data on whole rocks and mineral and glass separates. The Ilopango eruptive products are ideal for238U-230Th and 226Ra-230Th disequilibria studies as they are <1800 years old and allow for robust interpretation of the measured disequilibria, they range from basaltic andesite to rhyolite and the deposits are moderately crystal-rich. We show that, despite a large volume of erupted dacite, the minerals do not record a protracted storage time, but instead record generation of the silicic magmas only a short time (<10,000 years) before eruption, which was most likely triggered by an influx of mafic magma that represents a different magmatic lineage.
2. Geologic Background
 The Central American Volcanic Arc (CAVA) results from subduction of the Cocos Plate beneath the Caribbean Plate [DeMets, 2001]. The dip of the slab beneath El Salvador is 55°, and it gradually increases to 75° beneath Nicaragua before progressively flattening as the trench approaches the subducting Cocos Ridge in Panama [Carr et al., 1990; MacKenzie et al., 2010]. The convergence rate in El Salvador is 70 mm yr−1 [DeMets, 2001; DeMets et al., 1990], and the crust is estimated to be 32–38 km thick [Carr, 1984]. The caldera-forming eruption of Ilopango produced 70.6 km3 of dacite (18 km3 onshore, 52.6 km3 offshore [Kutterolf et al., 2008b]) and resulted in formation of the 8 × 11 km caldera that is now the location of Lago Ilopango [Rose et al., 1999]. The TBJ ignimbrites from this cataclysmic eruption blanket much of El Salvador and have been traced as far away as the Greenland ice sheet [Zielinski, 2000]. An initial 14C eruption age for TBJ of 1605 ± 20 years BP was established by Dull et al. , however a revised 14C age of 1810 ± 30 years BP was determined by Mehringer et al. . The TBJ eruption is considered to be a major archeological event that altered local cultures [Dull et al., 2001] and is frequently used as a time-stratigraphic marker [Mehringer et al., 2005]. Outcrops of the TBJ ignimbrites near Lago Ilopango are up to 30 m thick and contain white pumice clasts that are up to 10 cm in diameter. Individual pulses of this ignimbrite are subdivided into six units referred to as TBJ-A (oldest) – TBJ-F (youngest). Where observed, the contact between these units is a well-defined but uneven erosional boundary. The A unit is recognized as a crossbedded basal surge deposit (<1 m thick) that contains a small percent of pumice and abundant lithics. The B - F units represent the main phase of explosive activity [Mehringer et al., 2005]. Older eruptions are referred to as TB-2 through TB-4 [Rose et al., 1999]; the 36.3 km3TB-4 ash has been dated at 36 ka and the TB-2 has been dated at 15 ka [Lexa et al., 2011; Kutterolf et al., 2008b]. An eruption of a lava dome in 1880 produced a cluster of rocky islands located in the center of Lago Ilopango, called Cerros Quemadas. The dacite from this rocky outcrop contains olivine-bearing mafic enclaves of basaltic andesite. Another island in Lago Ilopango, Il Pat, is composed entirely of flow banded rhyolite and its eruption age is only constrained to post-date the TBJ eruption.
 In general, all pumice clasts from the TBJ units are highly vesicular, moderately crystal-rich (10–15%) and contain plagioclase and hornblende as the primary mineral assemblage. Trace amounts of clinopyroxene, apatite and oxides are observed in the matrix and as inclusions in the crystals. The mineralogy of the TBJ-B unit consists predominantly of plagioclase (80%) and hornblende (20%), with trace amounts of clinopyroxene and oxides. The plagioclase crystals are 2–3 mm in length, euhedral to subhedral and comprise two populations, those that have sieve texture cores (Figure 2a) and those that are pristine. The sieve texture cores have abundant inclusions of glass and minerals, including apatite, oxides and clinopyroxene and comprise 2–3% of the crystals. The pristine plagioclase crystals contain 1% apatite inclusions that are generally parallel to the growth zones (Figure 2b). The hornblende crystals (1 mm in length) are euhedral to subhedral and also comprise two populations; those that have pristine rims with abundant apatite, orthopyroxene inclusions (Figures 2c and 2d) and oscillatory zoning, and those that have reaction rims and are nearly completely dissolved. Both the plagioclase and the hornblende crystals in TBJ-B occur as individual crystals, not intergrowths. The mineralogy of the TBJ-E unit is very similar to the TBJ-B unit in terms of texture and crystal populations, except that there are intergrowths of plagioclase and hornblende, and the percentage of hornblende is slightly higher. The mineralogy and texture of the TBJ-F unit is identical to the TBJ E unit, and includes abundant intergrowths of plagioclase and hornblende. Overall, the TBJ dacite assemblage comprises plagioclase (75%), hornblende (20%), orthopyroxene (3%) and 2% accessory clinopyroxene + magnetite + ilmenite + apatite.
 Unlike the TBJ dacites, the 1880 Cerros Quemadas dacite dome is crystal rich (40% crystals, 60% matrix) and contains phenocrysts of plagioclase (2 mm) and olivine (1 mm) in a fine grained matrix of lath-shaped plagioclase (0.2–0.4 mm), hornblende and clinopyroxene. The plagioclase phenocrysts include three populations, those that are euhedral, those that have euhedral rims with resorbed, glass-inclusion-rich cores (Figure 2e), and those that have euhedral cores with resorbed rims. This dacite also has compact clots of fine-grained (0.1–0.5 mm) hornblende and plagioclase intergrowths. The mafic enclave is petrographically similar to the dacite dome, although it contains a higher percentage of hornblende in the matrix along with a trace amount of orthopyroxene. In many cases the hornblende has been partially dissolved (Figure 2f).
4. Analytical Methods
 Six samples were analyzed from Ilopango caldera, including units B, C and F of the TBJ tephra, the mafic enclave the host dacite from Cerros Quemadas, and the flow-banded rhyolite from Il Pat dome. Samples were washed in ultrapure water before methodical removal of weathered rims. A fraction of crushed rock was powdered in a Spex Al-ceramic shatterbox and sent to Michigan State University (MSU) for ICP and XRF analyses; the remainder was used for mineral separations. Sample powders were analyzed at MSU for trace elements using the Platform Inductively Coupled Plasma-Mass Spectrometer (ICP-MS) following procedures outlined byJohnson et al. , and for major elements using the Bruker S4 PIONEER 4 kW wavelength dispersive X-ray fluorescence spectrometer following procedures outlined byJohnson et al. , Knaak et al. , and Deering et al. . Rocks were dissolved in ultra-pure HF-HNO3-HCl acid using high pressure vessels, and then separated by aliquots for Sr, Nd, Pb, U, Th and Ra analyses. In order to determine if zircon had dissolved during bombing, several samples were also digested in 15 ml Teflon beakers. Chemical separation of Sr, Nd and Pb were conducted in the University of Iowa clean lab using Eichrom Sr-spec (Sr and Pb) and Ln-spec (Nd) resins. Separation of U-Th and Ra isotopes was conducted in the clean lab at Woods Hole Oceanographic Institute (WHOI) using established protocols [Sims et al., 2008a, 2008b]. Glass and mineral separates (amphibole, pyroxene, plagioclase and magnetite) from Ilopango units B, E and F were separated at the University of Iowa using a combination of magnetic separation and hand-picking. Silicate mineral separates were hand-picked to 99% purity, leached in a 10% HF solution before being crushed to a fine powder. All mineral separates were dissolved using HF-HNO3-HCl acids followed by fuming with boric and perchloric acids. Aliquots of whole rock, glass and mineral solutions for isotope dilution (ID) measurements were spiked with229Th and 233U, allowed to equilibrate for 48 hours then put through anion resin columns in order to separate Ra from U and Th. The Ra aliquot was spiked with 228Ra and put through several cation resin columns. Immediately before analysis, anion resin was used as the final clean-up column following the procedures established bySims et al. [2008b] in order to separate ingrown 228Th. Ra, Th, U, Nd, Sr, and Pb isotopes were measured at WHOI using the Thermo-Finnigan NEPTUNE Plasma ionization multicollector mass spectrometer following procedures outlined bySims et al. [2008a]. For Sr and Nd the internal precision is 5–10 ppm (2σ); external precision after adjusting to 0.71024 (NBS SRM 987) and 0.511847 (La Jolla Nd standard), respectively, is estimated to be 15–30 ppm. Pb analyses have internal precisions of 15–30 ppm; SRM 997 Tl was used as an internal standard, and then normalized to NBS 981 using the values of Todt and Cliff . External reproducibility (including full chemistry) for 208Pb/204Pb, 207Pb/204Pb, and 206Pb/204Pb are 150–200 ppm (2σ) and 90 ppm (2σ) for 208Pb/206Pb. 234U/238U was measured following the procedures established by Ball et al.  and Sims et al. [2008a, 2008b]. 230Th/232Th was also measured using WHOI's ThermoFisher NEPTUNE and procedures by Ball et al.  and Sims et al. [2008b, 2008c]. Reproducibility is within propagated analytical errors, which takes into account uncertainties in spike calibrations and tail corrections and decay constants (less than 2σ). 238U and 232Th concentrations were measured by isotope dilution on a single collector sector ICPMS (ThermoFisher Element) at WHOI. 226Ra concentrations were measured by isotope dilution mass spectrometry using WHOI's ThermoFisher NEPTUNE using techniques developed Sims et al. [2008b, 2008c]. Minerals from the dacite units E, F and B were separated by hand-picking, mounted in epoxy and analyzed for major and minor elements using the Cameca JXA-8200 Super-probe at UCLA. Mineral data are given in Table S1 in theauxiliary material.
 Major and trace element data are presented in Table 1. All samples fall into the medium K2O sub-alkaline group based on the classification system ofLe Maitre et al.  (Figure 3a), and are similar to the overall composition of volcanic rocks from El Salvador (shaded field in Figure 3a). Variations in major element compositions for the Ilopango samples are considered typical of subduction zone volcanic rocks, including decreasing CaO, Al2O3, FeO and MgO with increasing SiO2 (Figure 3b) that often is attributed to plagioclase + pyroxene + magnetite +/− olivine +/− amphibole fractionation during differentiation [e.g., Gill, 1981]. The 1880 Cerros Quemadas dacite is similar in major element composition to the TBJ dacites. The mafic enclaves within the dacite are basaltic andesites. The only true rhyolite in this study is the sample from the Il Pat dome.
Table 1. Major and Trace Element Data From Rocks of Ilopango Caldera
Trace elements were measured using the Platform (ICP-MS) at MSU following procedures outlined inDeering et al. . Major elements were analyzed using a Bruker S4 PIONEER-wavelength dispersive XRF.
SiO2 (wt %)
TiO2 (wt %)
Al2O3 (wt %)
Fe2O3 (wt %)
MnO (wt %)
MgO (wt %)
CaO (wt %)
Na2O (wt %)
K2O (wt %)
P2O5 (wt %)
Total (wt %)
 Primitive-mantle normalized trace element data are shown on incompatible trace element (ITE) and rare earth element (REE) diagrams (Figure 4). Samples show typical subduction zone characteristics, including relative depletions in Ta, Nb and Ti, and elevated concentrations of large ion lithophile elements (LILE) with respect to rare earth elements (REE). Primitive-mantle normalized REE patterns are U-shaped with strong enrichments in light REE and weak enrichments in heavy REE over middle REE for the dacites. The REE pattern of the mafic enclave is similar to that of the dacites, but has lower concentrations of light REE, and higher concentrations of middle REE.
 Long-lived isotope data are listed inTable 3 and shown in Figure 5. Our data are consistent with previously reported values for El Salvador, showing a positive correlation between 87Sr/86Sr and 143Nd/144Nd that is a feature of CAVA rocks (Figure 5a). This unusual positive correlation in CAVA rocks has been explained by variable degrees of melting of mantle associated with a flux of fluid derived from marine sediment (model shown in Figure 5a) [Feigenson and Carr, 1986; Carr et al., 1990]. Ilopango samples have 206Pb/204Pb and 207Pb/204Pb values that range from of 18.619–18.942 and 15.557–15.624, and plot within previously established data fields for Central American rocks (Figure 5b). The mafic enclave from Ilopango has the highest 206Pb/204Pb and 207Pb/204Pb values that trend toward the metamorphic basement value of Feigenson and Carr .
 U-series whole rock analyses are given inTable 2 and shown in Figure 6a. All samples in this study are characterized by (230Th/238U) >1.0, with 230Th excesses ranging from 1.6% (unit F) to 9.0% (dacite dome). The basaltic andesite enclave has (230Th/238U) = 1.04. None of the observed excesses correlate with SiO2 or MgO concentrations. The (230Th/232Th) values range from 1.5 for the TBJ dacites to 1.6 in the mafic enclave. Mineral-separate data from the three Ilopango TBJ units (B, E and F) are shown inFigure 6b. The data form linear arrays that overlap the range of whole rock values, except for the F-unit amphibole that has (238U/232Th) of 2.8 and an age corrected (230Th/232Th) 1.49, lower than the whole rock and mineral average of 1.51. Age corrected data and the modified York regression method [Mahon, 1996] were used to calculate the isochron ages for the mineral arrays. The averaged isochron age for units E, F and B is <4000 years, however given the scatter of the data, the linear array can only be reasonably constrained to an age range consistent with crystal growth of less than 10,000 years.
Table 2. U-Series Data for Ilopango Whole Rocks, Mineral Separates and Glass
(230Th/232Th) Age Corr.
(226Ra/230Th) Age Corr.
Whole Rock Data
TML (n = 5)
AThO (n = 3)
Mineral Separates and Glass
 Age-corrected226Ra values for Ilopango whole rocks and glasses are shown in Table 3 and Figure 7. The mafic enclave has the highest age-corrected (226Ra/230Th) of 1.12, and the dacite whole rocks and glasses have (226Ra/230Th) ranging from 1.02 to 0.82 (Figure 7a). The most evolved (i.e., highest SiO2) samples have the greatest 226Ra depletion with respect to 230Th (Figure 7b).
Table 3. Isotope Data for Ilopango Caldera, El Salvador
 Electron microprobe data from the Ilopango minerals show that the plagioclase crystals from all three units range from An49-An-77 (Figure S1) and have less than 0.3 weight % K2O (Table S1). There is no consistent trend between core and rim compositions, and crystals from all three units vary among normal, reverse and oscillatory zonation. All analyzed pyroxene are orthopyroxene (Di2En50–58) (Figure S1); orthopyroxene and hornblende have abundant inclusions of apatite and Fe-Ti oxides (Figure S2). The composition of the matrix glass ranges from 79 to 81 wt % SiO2, consistent with the composition of glass inclusions in the plagioclase and hornblende, which are range from 79 to 80 wt % SiO2. Glass inclusions in the orthopyroxene and hornblende range from 56 to 58 wt % SiO2.
6.1. Magma Differentiation at Ilopango Caldera
 The trace element variation observed in the Ilopango rocks is typical of arc rocks. In general, the Ta-Nb depletion in arc lavas is presumed to be a characteristic of the source and reflects the sequestering of Ta and Nb in a residual mantle phase [Tiepolo et al., 2001; Reagan and Gill, 1989], or alternatively the relatively low mobility of Ta and Nb in fluids from the subducting slab [Schmidt et al., 2004]. The enrichment in the LILE and LREE concentrations is attributed to the slab fluid component as well as to the relative incompatibility of these elements during crystallization [Hawkesworth et al., 1997; Davidson, 1996]. The TBJ samples have greater relative depletions in the MREE than the 1880 samples; a trait that is commonly attributed to amphibole fractionation [Davidson et al., 2007]. The geochemical similarities in the Ilopango rocks suggest a common lineage, in which the trace element variation can be modeled without assimilation, using an assemblage of plagioclase (70%) + amphibole (20%) + orthopyroxene (5%) + olivine (5%) + magnetite (0.01%) + zircon (0.004%) + apatite (0.001%) +/− allanite (0.001%) at 50% crystallization (Figure 4a). However, either the basaltic andesite does not represent a parental magma for the TBJ dacites, or crustal assimilation contributes to differentiation (AFC). This is indicated by the differences in 87Sr/86Sr, 143Nd/144Nd and 207Pb/204Pb values between the basaltic andesites and dacites (Figure 5), as well as the positive correlation between the isotope values and weight % SiO2 (Figure 8).
 The TBJ dacites and the dacite dome have higher 143Nd/144Nd than the basaltic andesite (Figure 5a), and in the CAVA, high-Nd isotope values are typically attributed to a depleted mantle source [Carr et al., 1990] or to assimilation of the ultramafic high-Nd forearc crust of the Chortis Block [Geldmacher et al., 2008]. Figure 5a shows a mixing curve between a 5% melt of Chortis forearc sediment and the Ilopango basalt andesite. This curve shows that 20–30% of the sediment melt is required to account for increase in 87Sr/86Sr and 143Nd/144Nd in the dacites. The Carr model requires much lower percent of high-Nd slab fluid. Either of these models can account for the variations in Sr and Nd isotope values. However, the207Pb/204Pb and 206Pb/204Pb of the basaltic andesite is consistent with neither sediment fluid nor Chortis block sediment melt, but trends toward metamorphic basement (Figure 5b). This inconsistency points toward our initial hypothesis that the basaltic andesite is not a parental magma for the TBJ dacites, but likely triggered the 1880 lava dome activity by injection into the base of caldera.
 The 50–60% fractional crystallization that is sufficient to generate the observed variation without significant crustal assimilation (Figure 4a) is also consistent with the proportions of minerals observed in thin section as well as the trace element ratios, including Sr concentrations, U/Th, La/Dy and La/Yb (Figure 9). The Sr and La/Dy values (controlled by plagioclase and hornblende, respectively) can be reproduced using a composition similar to the basaltic andesite using the same mineral proportions. The minerals that provide the most leverage over the whole-rock (238U)/(230Th) concentrations are accessory phases like allanite, apatite, monazite and zircon. Due to the high crystal/magma partition coefficients, even small amounts of these minerals provide ample leverage on the U/Th of igneous rocks [Schmitt, 2011] without significantly changing the major element composition of the magma. For example, U/Th in the Ilopango samples is poorly correlated with weight % MgO and P2O5 and Zr concentrations (Figure 10). However, the U/Th values cannot be explained without crystallization of accessory zircon and allanite (Figure 9b). If the 1880 dacite is excluded from Figures 10a–10c, there are indeed consistent trends between U/Th on the MgO and Zr ppm. The weak colinearity on Figure 10 might reflect the slightly different composition of the 1880 dacite compared to the TBJ tephras, suggesting that neither of the 1880s lavas is directly linked to the TBJ tephras. Nonetheless, the overall similarity in compositions suggests that they were generated from similar sources and followed similar liquid lines of descent.
6.2. U-Series Evidence for Magma Genesis at Ilopango
 In terms of (230Th/238U), most Central American lavas have values typically within 10% of equilibrium; the greatest observed 238U excesses are about 15% at Cerro Negro. On the other hand, 230Th excesses are typically of greater magnitude, and a significant proportion of the basalts and basaltic andesites have 230Th excesses rather than 238U excesses (e.g., Concepción and high-Nb Nejapa lavas in Nicaragua [Reagan et al., 1994]). Excess 230Th is particularly pronounced in mafic lavas from Central Costa Rica (e.g., Turrialba, Irazu) that have parental lavas generated by low-degrees of mantle melting in the presence of garnet and little slab flux [e.g.,Reagan and Gill, 1989]. 230Th excesses in basalts from Nicaragua have been attributed to melting of hornblende-bearing crystal cumulates [Reagan et al., 1994], and Walker et al.  concluded that 230Th excesses in mafic rocks from Guatemala are the result of crustal contamination. Thomas et al.  explained 230Th excesses in Nicaragua and Costa Rica using a continuous flux-ingrowth melting model. The 20%226Ra excess (with respect to 230Th) in the Ilopango 1880 basaltic andesite enclave is similar in scale to those found in lavas from Costa Rica [Thomas et al., 2002] and Guatemala [Walker et al., 2007], unlike the 50–250% 226Ra excesses found in the strongly subduction fluid affected lavas from Nicaragua [Reagan et al., 1994; Thomas et al., 2002]. The 230Th and 226Ra excesses in the Ilopango samples are most consistent with the flux-ingrowth melting model ofThomas et al. , which simultaneously explains the 4% 230Th excesses and the 20% 226Ra excesses in the Ilopango basaltic andesite using an average extent of melting of 3–4% and a mantle composition similar to that beneath Nicaragua (Figure 12a).
6.3. U-Series Evidence for Magma Differentiation at Ilopango
 If, like the basaltic andesite, the 230Th excesses of the TBJ dacites are inherited from the mantle, this would require that the time elapsed since mantle melting, including ascent and differentiation to dacite is less than 5,000 years; a short time for continental arcs [Turner et al., 2000]. It seems more likely that the increase in (230Th/238U) from the basaltic andesite to the 1880 dacite and rhyolite domes is more likely a differentiation effect. Most of the compositional differences between the basaltic andesite and the TBJ tephras can be modeled using crystal fractionation of accessory phases (Figures 9b and 12b). It has been shown that crystallization of accessory phases allanite, apatite and monazite can sequester Th and create significant disequilibria [Schmitt, 2011; Garrison et al., 2006], but zircon is the only accessory mineral capable of sequestering significant U to create 230Th excesses, perhaps with the exception of magnetite. The TBJ mineral separate data show that magnetite from all three units have (238U) excesses of 4%. The high (238U/230Th) of 2.7 in the TBJ-F amphibole mineral separate (Figure 6b) is likely the result of zircon inclusions. The correlation between decreasing magnitude of the 226Ra excess and weight percent SiO2 (Figure 7b) strongly suggests that the 226Ra variation is related to crystallization of a Ra-bearing phase within the last few thousand years, although aging might have equilibrated the (226Ra/230Th) values for some dacites. The minerals that provide the greatest leverage over (226Ra)/(230Th) are K-bearing phases, specifically amphibole and the alkali feldspars [Sims et al., 2002] including anorthoclase [Cooper and Donnelly, 2008; Reagan et al., 1994]. Ra partitions preferentially into these minerals, and less effectively into plagioclase [Fabbrizio et al., 2009]. The composition of the Ilopango feldspar (Table S1) ranges An49-An53 with K2O<0.3 wt %, and this composition of plagioclase (labradorite) does not significantly partition Ra [Fabbrizio et al., 2009, 2010]. On the other hand, elastic strain models suggest that amphibole does preferentially retain Ra relative to Th [Feineman and DePaolo, 2003]. Amphibole also retains MREE, and fractionation is commonly represented as a decrease in Dy/Yb during differentiation. There is a positive correlation between (226Ra)/(230Th) and Dy/Yb in the Ilopango samples (Figure 13), and one way to explain the Ra-Ba-Th trend is fractionation from basaltic andesite to dacite involving a hornblende-rich assemblage that lowers Dy/Yb and (226Ra)/(230Th). In the Ilopango basaltic andesite Dy/Yb = 1.6, however in the dacite, rhyolite and TBJ dacites Dy/Yb ranges from 1.35 to 1.22 (Figure 13). This 9% decrease correlates with a concomitant decrease in (226Ra/230Th). We propose that the (238U/232Th) and (226Ra/230Th) variation in the Ilopango dacites reflects crystallization of hornblende, followed by accessory phase fractionation of allanite and zircon in the magma chamber prior to eruption of the TBJ dacites (model curves shown on Figures 7a, 9b, and 12b). This model is also consistent with the variations in REE patterns for Ilopango tephras (Figure 4a).
6.4. U-Series Evidence for Magma Storage Times at Ilopango Caldera
 The Ilopango whole-rock data form a sub-horizontal array on the equiline diagram (Figure 6) that is consistent with a <10,000 year time frame since U-Th fractionation for all the samples, including the rhyolite dome, which, on the basis of these data is likely to have the same eruption age as the 1880 dacite domes. The mineral separate data yield isochrons that were calculated using age-corrected data for the TBJ-B, E and F units. By using age-corrected data, the calculated isochron ages represent the time elapsed between the average age of mineral growth and magma eruption. This reflects the average storage time of the TBJ dacite, assuming that the crystals grew in conjunction with crystal fractionation. The averaged isochron ages for units E, F and B, using the York regression [Mahon, 1996] is <4,000 years, however given the considerable scatter in the (230Th/232Th) values, it is most reasonable to interpret the array as being consistent with fractionation from basaltic andesite to dacite in less than 8,000 years. This a short enough time span to preserve (226Ra)/(230Th) disequilibrium in some samples (TBJ-B and the rhyolite dome), although other dacites had a long enough residence time in the crust for226Ra to attain equilibrium with 230Th. This is quite remarkable considering that this timeframe must allow for the generation of 70 km3 of dacite. Using the calculated 50% crystallization from the basaltic andesite, this equates to a continuous crystallization rate on the order of 5.0 × 10−3 yr−1, similar to trachytes from the Azores [Snyder et al., 2007].
 Several recent studies have raised the issue of mixing among crystal populations that may mask the true isochrons in volcanic rocks [Fabbrizio et al., 2009; Cooper et al., 2003; Cooper and Donnelly, 2008]. These studies demonstrate that mixing of multiple crystal populations, or the presence of melt inclusions in minerals can yield a linear array that is actually a mixing line. Indeed, the three populations of plagioclase crystals in the TBJ units cannot have an entirely shared history. We point out, however that the mineral assemblages among the TBJ tephras have identical (230Th/232Th), whereas scavenged minerals with significant concentrations of Th and U concentrations, if they are significantly older than the Ilopango dacite affect the nearly horizontal U-Th array. This further illustrates the lack of evidence for the involvement of a significant volume of older magmas or crystals in the TBJ eruption. The three Ilopango units (B, E, F) could represent mixtures of unsampled end-member magmas since they all shared the same magma reservoir. Nevertheless, the isotope and trace element data indicate that the magmas are cogenetic, and cogenetic magmas that define a horizontal isochron must be essentially “zero-aged” even if they mix with each other. Similarly, melt inclusions can also have a large impact on the isotope compositions of the crystals because glass inclusions may contain the majority of U, Ra and Th in the plagioclase and amphibole. At Ilopango, however, the impact of contamination by glass inclusions is likely to be low for these samples, since none of the mineral separates have (230Th/232Th) above the whole rock value of 1.5, and the glass composition is similar to the whole rock values.
 In the broader context of magmagenesis, our results are consistent with studies that show fractional crystallization and melt extraction to be the primary mechanisms for rhyolite formation in arc systems [Bachmann and Bergantz, 2003, 2004; Brophy, 1991]. Calculated segregation times for the Ilopango dacites, using models of hindered settling and filter pressing (compaction) from Bachmann and Bergantz  range from 104-106years for large magmatic systems. Our results suggest that crystal fractionation at Ilopango is at the short end of this range. The previous significant eruption from Ilopango Caldera, the TB-2 eruption is dated at 15,000 years old, and the TB-4 eruption is dated at 36,000 years. On the basis of these ages, it appears that the repose time between eruptions at Ilopango Caldera is on the order of 15–20 ka. Our U-series data are consistent with an average crustal residence time for magmas involved with the TBJ eruption that is well within this time-frame.
 Dacite and rhyolite lavas are commonly believed to represent the degassed residual magma after large eruptions [Streck and Grunder, 2007], and the Cerros Quemados dacite domes are consistent with other examples of post-caldera magmatism (e.g., Incapillo Caldera [Goss et al., 2009], Valles Caldera [Wolff and Gardner, 1995], and Mount St Helens [Rowe et al., 2008]). In some cases, dome lavas are believed to represent magma that was co-genetic with the primary eruption (Long Valley Caldera [Heumann and Davies, 1997; Higgins and Meilleur, 2009]); in other cases there is evidence to indicate a component of juvenile magma (Yellowstone [Vazquez et al., 2009]). At Ilopango caldera, the basaltic enclaves clearly postdate the development of the TBJ dacites, and it is likely that input of this juvenile magma into the system triggered the 1880 eruption. The rhyolite dome (Il Pat), which must postdate the TBJ eruption, has a composition indicating that is was generated by crystal fractionation of residual TBJ magma (Figures 7 and 9a).
 In terms of timescales, repose times between rhyolite eruptions range from 105-106 years for large caldera systems (i.e., Yellowstone [Reid et al., 1997] and Toba [Vazquez and Reid, 2004; de Silva, 2008]), to 103 years for smaller systems (e.g., Okataina Volcanic Complex [Klemetti et al., 2011] and Cotopaxi Volcano [Hall and Mothes, 2008; Garrison et al., 2011]). On the basis of tephra chronology, repose times have been established for many calderas in the CAVA (e.g., Chiltepe, Coatepeque, Apoyo, Los Chocoyos, Apoyeque [Kutterolf et al., 2007, 2008a, 2008b]), however the magma residence times for these systems are unknown. In fact, despite the large number of explosive rhyolite calderas in the CAVA there are surprisingly few studies on magma residence times. Future studies of residence times in other CAVA caldera systems may be useful to assess if rapid differentiation and short storage times are typical of other Central American calderas.
 The U-series and long-lived isotope data from TBJ dacites of Ilopango Caldera indicate that neither the mineral isochrons nor whole rock U-series data record a protracted storage time for the Ilopango TBJ dacite. Crustal residence times of magmas at Ilopango are on the order of 103-104 years, and magma production rate at Ilopango must be on the order of 10−4 km3/km/yr in order to explain the U-series disequilibria in this system. The (238U)/(230Th) variation in the Ilopango TBJ dacites is related to accessory phase crystallization during differentiation, and the (226Ra)/(230Th) variation is related to amphibole fractionation. The 1880 eruption was likely triggered by an intrusion of an isotopically distinct basaltic andesite into the base of the system. In fact, the eruption of rhyolite and dacite domes in Lago Ilopango illustrates that magmatic activity is ongoing, and that this rapidly developing system has a robust supply of magma. However, the 15–20 ka repose time between large eruptions, coupled with the observation that the most recent eruption was of a degassed, relatively mafic dacite suggests that a large body of more differentiated dacite does not presently exist beneath Ilopango Caldera.
 We thank NSF for funding Garrison under the MARGINS post-doctoral fellowship program (OCE 0405262). We also thank Shan deSilva, TomVogel and the editor James Tyburczy for thoughtful reviews that strengthened our final manuscript. We are grateful to D. Peate and M. Wortel at the University of Iowa, as well as L. Ball, J. Blusztajn and Dave Schneider at Woods Hole Oceanographic Institute for assistance with analyses. We also thank the lab staff at MSU for conducting the XRF/ICP analyses, and Frank Kyte and UCLA for help with the electron microprobe analyses. We are also grateful to the staff at SNET (Servicio Nacional de Estudios Territoriales) in El Salvador for their help with sample locations and field work.