Evolution of geochemical variations along the Central American volcanic front



[1] New geochemical analyses of volcanic rocks in El Salvador add to existing data from Nicaragua and Costa Rica to create a comprehensive set of geochemical data for Central American volcanics. These data coupled with previously published 40Ar/39Ar ages covering the past 30 Ma show that Costa Rica and Nicaragua had similar U/Th and Ba/La values until 10 Ma when the region developed the distinctive along arc variations that made this margin famous. U/Th values increased in Nicaragua since the Miocene, while remaining unchanged along the rest of the volcanic front. This coincides temporally with the Carbonate Crash, which caused a transition in Cocos plate sediments from low-U carbonates to high-U, organic rich hemipelagic muds. Increases in uranium are not observed in Costa Rica because its lower slab dip produces a more diffuse zone of partial melting and because of the contribution from Galapagos-derived tracks dilutes this signal. Ba/La has been used as a geochemical proxy for contributions from the subducting slab; however, our analyses indicate that the Ba concentrations do not vary significantly along strike either in the subducting sediment or the volcanic front. Along-arc variation is controlled by changes in La, an indicator of the degree of partial melting or source enrichment. Trace element models of five segments of the volcanic front suggest that a subducting sediment component is more important to magmas produced in El Salvador and Nicaragua than in Costa Rica, where the geochemistry is controlled by recent (<10 Ma) recycling of Galapagos tracks.

1. Introduction

[2] We report here new geochemical data on Oligocene-Mid-Miocene as well as Early Quaternary rocks from El Salvador, which allow us to extend the along the arc profiles in Ba/La and U/Th for the youngest Oligocene-Mid-Miocene arc. These new geochemical data (major and trace elements) for volcanic rocks in El Salvador provide us with the most complete data of Central American volcanics published to date. Along with 40Ar/39Ar ages and geochemical modeling, these data extend our knowledge over the past 30 my of the evolution of this arc and help us better define the space-time variations in Ba/La and U/Th and understand their origin.

1.1 Geologic Setting

[3] The Central American volcanic front extends 1,100 km from the border between Mexico and Guatemala to Costa Rica (Figure 1) and is generated by the northeasterly subduction of the Cocos Plate underneath the Caribbean Plate at a rate that ranges from 6 cm/yr off southern Guatemala to 9 cm/yr off southern Costa Rica [DeMets, 2001]. One of the obvious features in Central America is the physical segmentation of the volcanic front into seven right stepping lines that vary in length from 100 to 300 km [Carr, 1984].

Figure 1.

Map of the evolution of the Central American volcanic front. Triangles are active volcanoes. Inset on the lower left shows locations of core sites mentioned in this study. Location of Miocene to present volcanic front after Bundschuh and Alvarado [2007]. Gray shaded box is the Cocos Plate peak fault relief from Kelly [2003].

[4] Geochronological and geological data define an approximate age for the volcanic front of Costa Rica and Nicaragua of 600 and 330 ka, respectively [Carr et al., 2007]. Based on this age constraint, the estimated extrusive volcanic flux ranges from 1.3 × 1010 kg/m/Ma in western Nicaragua to 2.4 × 1010 kg/m/Ma in central Costa Rica, but overlap within the calculated error [Carr et al., 2007]. The similar fluxes and similar Ba contents (Figure 2) along the margin suggest that regional Ba/La variations are not related to the total amount of released slab component but the mechanism of fluid delivery into the mantle wedge. A focused fluid mechanism that produces high degree melts (low La/Yb) is preferred for Nicaragua; meanwhile a diffuse mechanism that will produce relatively lower degree of partial melting (high La/Yb) is more likely in Costa Rica and El Salvador [Carr et al., 2007].

Figure 2.

Variation of Ba, La, Th, and U along the volcanic front. Colored data symbols are volcanic front data from Carr et al. [2013] plotted against distance along arc. Gray circles are Miocene samples from Plank et al. [2002], Gazel et al. [2009], and Saginor et al. [2011]. Gray crosses are modeled values at different degrees of partial melting from Table 2. Note the uniform concentrations of fluid-mobile Ba all along the volcanic front but high concentrations of La, U, and Th in central Costa Rica.

1.2. Stratigraphy of the Subducting Cocos Plate

[5] The subducting Cocos Plate has a simple sedimentary stratigraphy consisting primarily of an upper hemipelagic layer overlying a carbonate sedimentary layer [Patino et al., 2000]. It is generally thought that these sediments are not accreted and that the complete subduction of the Cocos Plate allows comparison of the subducted input and volcanic output [Patino et al., 2000; Plank et al., 2002; Solomon et al., 2006]. However, significant 10Be enrichment in Nicaraguan front volcanoes may suggest incomplete subduction of the hemipelagic sediments toward the edges of the margin in Guatemala and Costa Rica [Leeman et al., 1994].

[6] Ocean Drilling Program (ODP) sites 844 and 845 [Plank et al., 2002], Deep Sea Drilling Program (DSDP) Site 495 [Patino et al., 2000], and ODP Site 1039 [Solomon et al., 2006] provided the sedimentological and chemical stratigraphy for the subducting Cocos Plate (Figure 3, core locations in Figure 1). The two sites closest to the trench are Site 1039 off the Nicoya Peninsula of Costa Rica and Site 495 off Guatemala. Despite being 600 km apart, these two sites have very similar lithologies and barium distributions throughout both the carbonate and hemipelagic layers. Uranium values of sites 495, 844, and 845 increase sharply in the upper hemipelagic layer. The increase occurs earliest at Site 495, which is closest to the trench. If we calculate the time needed for sites 844 and 845 to reach the near trench position of Site 495 and add that time to their age profiles, the corrected age versus U concentration profiles (Figure 3b) agree within error. Although the elevated U concentrations appear to be related to both time and proximity to the Central American Volcanic Front as volcanic ash could be a possible contributor to the elevated U concentrations, a more likely cause is elevated organic carbon concentration related to upwelling near the coast, coupled with organic carbon's ability to precipitate U from the water column [Plank et al., 2002]. In contrast to U, both sites 844 (600 km offshore Nicaragua) and 845 (435 km offshore Guatemala) have similar Ba profiles as 1039 and 495 [Patino et al., 2000; Plank et al., 2002; Solomon et al., 2006]. The similar barium values in these four cores suggest that barium levels are relatively constant throughout sedimentary layers on the Cocos Plate and are not controlled solely by distance from the margin.

Figure 3.

(a) U concentration in core samples vs. age [after Plank et al., 2002]. (b) U concentration in core samples versus adjusted age. Note the increase of U after 5 Ma.

[7] In the carbonate section, most incompatible elements are in low concentration or vary little with depth (e.g., Sr, Ba). In the hemipelagic section, there are some strong gradients: Ba and Pb increase and U decreases with depth. Th, K, Rb, Cs, and Sr show little variation with depth. The mean values of Ba, La, Y, and Pb are only slightly higher in the hemipelagic section than in the carbonate section [Patino et al., 2000; Solomon et al., 2006]. However, the mean values of U, Cs, Th, K, and Rb are much higher in the hemipelagic section.

1.2. Along Arc Geochemical Variation

[8] Large geochemical variations occur in Central American front lavas [Figure 4, Carr et al., 1990; Morris et al., 1990; Leeman et al., 1994] that made this arc a focus site for the National Science Foundation (NSF) Margins Program. Based on Na2O content [Plank and Langmuir, 1993] and La/Yb [Carr et al., 1990], these authors suggested that degree of melting is highest beneath Nicaragua (where the highest Ba/La is also reported) and that the extent of melting decreases toward the northwest and southeast. Similarly, the element ratios that were thought to trace slab input, Ba/La, U/Th, and Ba/Th are highest in Nicaragua and decrease toward the northwest and southeast (Figure 4).

Figure 4.

Geochemical and geophysical variation along the volcanic front in Central America. Colored symbols are from the active front [Carr et al., 2013] and gray circles are Oligocene-Mid-Miocene samples [Plank et al., 2002; Gazel et al., 2009; Saginor et al., 2011]. Balsamo samples are represented by the gray circles within the Salvadoran segment. Colored symbols for geophysical measures refer to present day. Slab age and crustal thickness are from Carr et al. [1990]. Slab dip is from Syracuse and Abers [2006].

[9] Oxygen isotopes (δ18Oolivine) from phenocrysts in basalts and basaltic-andesites from the Central American volcanic front vary from a minimum of 4.6‰ centered in western Nicaragua to a maximum of 5.7‰ in Guatemala [Eiler et al., 2005]. This variation correlates with major and trace elements, and Sr and Nd isotope values of the host lavas. Eiler et al. [2005] interpreted this trend as variations in δ18O slab component contribution to the mantle. Models suggest that these variations are produced by two end members, a water rich flux with low δ18O produced by dewatering of serpentine from the altered oceanic crust of the Cocos Plate and a relatively water-poor sediment melt with high δ18O. The first end member dominates the slab component in the center of the arc, the second is more evident in the northwestern part of the arc. In central Costa Rica, both fluxes are small or even insignificant [Eiler et al., 2005].

[10] Leeman et al. [1994] found that B/La ratios correlated with 10Be/9Be, which allowed the authors to estimate that the flux of subducted sediment is at least twice as high in Nicaragua than in Costa Rica. They proposed that the higher flux of sediments is due to subduction of older, cooler, and steeper slab below Nicaragua.

[11] The slab contribution for volcanic front originates from both the sedimentary layers and the altered oceanic crust subducting along the Middle American Trench [Patino et al., 2000; Hoernle et al., 2008; Gazel et al., 2009, 2012]. Patino et al. [2000] show that the sediment signal is a mixture between the upper organic rich hemipelagic section of the Cocos Plate (with low Ba/Th and high U/La) and the lower pelagic carbonate section (with high Ba/Th and low U/La) of the Cocos Plate.

[12] The increase in U/Th values in Nicaraguan volcanic material since the Miocene has been attributed to changes in the subducting sediments following the Carbonate Crash [Lyle et al., 1995; Plank et al., 2002] at 10–12 Ma. At that time, the Central American isthmus began to close, which shut thermohaline circulation from the Pacific to the Caribbean [Coates et al., 1992; Lyle et al., 1995; Farrell et al., 1995; Montes et al., 2012]. This caused the carbonate compensation depth to rise and sediments to become enriched in organic carbon at the expense of carbonate [Hoffmann et al., 1981]. The result is that Cocos Plate sediments consist mainly of a lower carbonate layer and an upper hemipelagic layer, with the latter unit showing enrichment in U [Patino et al., 2000]. This increase in uranium content of subducted sediments is reflected in Nicaraguan lavas in the form of a substantial increase in U/Th values between the Miocene and active volcanic front [Patino et al., 2000; Plank et al., 2002]. Recently, Saginor et al. [2011] found that the transition to high U/Th in the volcanic output of Nicaragua after the Central American gateway closure was more gradual than previously thought.

1.3. Study Area

[13] Reynolds [1980] summarized the Oligocene-Mid-Miocene volcanic stratigraphy of Northern Central America using many of the units defined by Wiesemann [1975]. An elongated Oligocene-Mid-Miocene volcanic belt strikes N 70° W, roughly parallel to the Pacific Coast for 800 km through Guatemala, El Salvador, and into Honduras. In central and eastern Honduras and northern Nicaragua, the Oligocene-Mid-Miocene belt widens significantly. Reynolds [1980] divided the Oligocene-Mid-Miocene volcanic sequence into three lithostratigraphic formations that roughly parallel the Pacific Coastline: the Chalatenango Formation (Middle to Upper Miocene) composed of rhyolitic tuffs and lavas, the Balsamo Formation (Upper Miocene to Pliocene) composed of andesitic lavas, tuffs, and lahars, and the Cuscatlán Formation (Pliocene) composed of rhyolitic tuffs and basaltic lavas. The only ages ever published for the Balsamo Formation were from Saginor et al. [2011] and they were both found to be approximately 1 Ma, significantly younger than suggested by Reynolds [1980].

[14] The Chalatenango Formation occurs inland from the presently active volcanic belt. The Balsamo Formation is coincident with or on the Pacific coastal side of the currently active volcanic belt. In eastern and central El Salvador, the Cuscatlán Formation overlies the Balsamo Formation on the coastal side of the volcanic belt. In western El Salvador, the Cuscatlán Formation occurs on the northern side of the Oligocene-Mid-Miocene volcanic belt, where it overlies the Chalatenango Formation [Reynolds, 1980]. The characteristics of the volcanics changed from dominantly silicic tuffs to andesitic flows to basaltic flows during the Late Pliocene to Quaternary [Reynolds, 1980]. A secondary type of volcanism, composed of bimodal basalt-rhyolite suites called Cuscatlán occurred since the Late Pliocene in the region behind the volcanic front. The extensive flows and tuffs of the Chalatenango Formation in the Middle Miocene occurred inland of the present volcanic front and along the central and northern parts of the Oligocene-Mid-Miocene volcanic belt. These extensive, predominantly silicic deposits represent a substantially more productive, broader, and longer lasting volcanic episode.

[15] In Nicaragua, the Oligocene-Mid-Miocene was dominated by basaltic and andesitic lavas of the Coyol and Tamarindo formations, with the latter being the only Miocene Nicaraguan volcanism to the southwest of the active front [Plank et al., 2002; Saginor et al., 2011]. The Coyol is often found associated with Miocene ignimbrites of the Matagalpa Formation [McBirney and Williams, 1965]. Costa Rica was similarly active during this time period with the exception of 10–7 Ma, when a gap in volcanism was associated with pluton emplacement [Macmillan et al., 2004] and pulses of slab detachment following the collision with Galapagos tracks [Gazel et al., 2012].

[16] It is now clear that the volcanic front has shifted through time [Plank et al., 2002; Carr et al., 2007; Saginor et al., 2011]. Figure 1 shows the approximate past positions of the volcanic front for El Salvador, Costa Rica, and Nicaragua, which appear to have shifted location over time. In general, the volcanic belt has migrated toward the trench, implying either a steeper slab dip over time or slab rollback; however, there are substantial differences along the margin. In Nicaragua, the modern front has been moving trenchward for at least 10 Ma [Plank et al., 2002] and based on the available geochronology data there is a gap in volcanism between 7 and 3.6 Ma [Saginor et al., 2011]. In western Costa Rica, the strike of the volcanic front has rotated 30 degrees counterclockwise since about 12 Ma [Macmillan et al., 2004].

[17] Available geochronological data indicate that the currently active volcanic front started at approximately 600 ka in the same location as the earlier front that was active from 2 to 1 Ma [Carr et al., 2007]. For the most part, the Oligocene-Mid-Miocene volcanic belt is located inland from the Quaternary to modern (<2 Ma) volcanic front with the exception of Costa Rica where the modern front is underlain by volcanism of Early Miocene to Pliocene age (Figure 1).

[18] Over the last 20 Ma, the volcanic front of Central America has migrated, usually in the trenchward direction, allowing the geochemical history to be sampled [Ehrenborg, 1996; Plank et al., 2002; Carr et al., 2007; Gazel et al., 2009; Saginor et al., 2011]. Plank et al. [2002] provided the first geochemical reconstruction of Nicaraguan volcanism from the Miocene to present. Their Ba/La profile for the Oligocene-Mid-Miocene arc is essentially the same as that of the modern volcanics, there is some variability through time (section 4.3), while the U/Th profile along the Oligocene-Mid-Miocene arc in Nicaragua is very different from the profile derived from the presently active volcanoes.

2. Data and Analytical Methods

[19] To address the lack of analyses of nonactive segments of the arc from the northern segment of the volcanic front, we collected 20 samples the units mapped as Balsamo Formation across El Salvador. Sample locations documented with GPS are reported in Table 1. Major element (wt%) data for the Balsamo samples were obtained using by X-ray fluorescence (XRF) at Michigan State University following procedures outlined in Hannah et al. [2002]. Data are reported in Table 1.

Table 1. New XRF and ICP-MS Geochemical Data for El Salvadorian Balsamo Formation
Age         Age         Age  
Easting288.194 296.705296.705315.114321.397325.39330.822 Easting400.638400.638 378.724378.389231.205235.336211.067 Easting222.816212.772
Northing1518.3 1518.0241518.0241510.3031504.7461504.2681506.846 Northing1461.251461.25 1460.1571464.9271538.5771532.1331518.507 Northing1516.6111515.563

[20] Trace element data (ppm) were collected from rock sample solutions, digested by concentrated HF-HNO3, and analyzed on a Finnigan MAT Element, high-resolution inductively coupled plasma mass spectrometer (HR-ICP-MS) (Table 1) at the Institute of Marine and Costal Sciences at Rutgers University. The analytical run consisted of 20 rocks sample solutions, two digestion blanks, two USGS rock standards (BIR-1a, BHVO-1), five duplicate samples, and three standard additions (consisting of nine samples). Standard additions were distributed evenly through the run. Indium was added to all the samples during solution preparation for drift monitoring. Drift was reduced by normalizing the indium intensities of all the measurements in the run to the first measurement in the run. Indium drift was about 20% during the run. The relative percent difference between the calculated and given reference values for BHVO-1 (USGS) were below 6% with the exception of Cs (26%), Eu (11%), Ta (9%), W (15%), Pb (19%), Th (16%), and Y (10%). Pb and Th agree within 6% of the ICP-MS data reported by Jenner et al. [1990]. Precision determined by duplicate analyses was better than 7% RDS. All the details of the method (standard values used, blanks, etc.) can be found in Bolge et al. [2009].

3. Results

[21] The new samples from the Balsamo formation range from basalts to andesites, dominated by basaltic-andesites (Figure 5). In part, this reflects a sampling bias in favor of mafic lavas. The major element data from Balsamo Formation are within the range of modern lavas from the volcanic front in El Salvador and Nicaragua, belonging to the calc-alkaline series (Figure 5). The new trace element data from the Balsamo formation in El Salvador support the results presented by Plank et al. [2002] and Saginor et al. [2011] for the Oligocene-Mid-Miocene lavas of Nicaragua. Ba/La values for the Balsamo samples are lower than in NW Nicaragua and similar to the profile for the modern volcanic front (Figure 4). The volcanic front segment of El Salvador is ∼symmetrical to the values of SW Nicaragua and northern Costa Rica.

Figure 5.

Arc rock type classification from Peccerillo and Taylor [1976]. Note that the new data from El Salvador Balsamo Formation belonging to the calc-alkaline series overlaps modern El Salvador and Nicaragua volcanic fronts.

4. Discussion

4.1. Trace Element Modeling Along the Volcanic Front

[22] Previous Central American trace element models were limited by the lack of estimates of the composition of magma sources (subducting sediments, mantle wedge composition, etc.). For the different arc models presented here, we used the complete magma source compositions and model constraints summarized by Gazel et al. [2009]. Furthermore, Hoernle et al. [2008] and Gazel et al. [2009, 2012] showed that a component from subducting Galapagos-derived tracks is necessary for modeling isotopic and trace element data in southern Central America.

[23] Trace element concentrations for the modern volcanic front were modeled for five segments along the volcanic front (Figure 6, model parameters and results in Table 2): El Salvador, Northwest Nicaragua, Southeast Nicaragua, Northwest Costa Rica, and Central Costa Rica. The northern segment (Guatemala) of the volcanic front was not modeled because the volcanic material from that region shows evidence of crustal contamination, most likely due to relatively thicker Paleozoic continental crust basement [Feigenson and Carr, 1986].

Figure 6.

Modeled trace element concentrations for El Salvador, NW Nicaragua, SE Nicaragua, NW Costa Rica, and central Costa Rica (black crosses and dashed lines). High-precision volcanic front data from Carr et al. [2013]. Details about modeling and modeled data in Table 2.

Table 2. Trace-Element Modeling Results Following the Methodology Described in Gazel et al. [2009]
SampleMetasomatized MantleTiO[2]K[2]OP[2]O[5]RbSrYZrNbBaLaCePrNdSmEuDyYbLuTaPbThUDistance (km)
Central CR_5%FDM+3%Galapagos+0.1 Sed 11.472.901.4064.591154.9527.50189.7223.481009.1473.58139.7118.2647.426.151.835.582.560.361.275.227.644.021050.00
Central CR_10%FDM+3%Galapagos+0.1 Sed 11.211.450.7932.30618.9023.83115.1111.78504.5736.8571.349.8928.024.181.314.452.210.320.642.643.822.011050.00
Central CR_15%FDM+3%Galapagos+0.1 Sed 11.010.970.5421.53414.3220.7179.377.85336.3824.5747.576.6319.153.040.983.611.910.280.421.762.551.341050.00
Central CR_20%FDM+3%Galapagos+0.1 Sed 10.850.730.4016.15310.8118.0659.875.89252.2818.4335.684.9714.422.340.762.971.650.250.321.321.911.001050.00
NW CR_5%DM+0.6%Galapgos+0.6 Sed.11.411.680.5334.07708.5528.88136.6110.641620.5332.5661.138.2924.904.661.525.392.710.390.614.983.522.44910.00
NW CR_10%DM+0.6%Galapgos+0.6 Sed.11.160.840.3017.03380.5424.6082.365.34810.2716.3131.244.5014.743.
NW CR_15%DM+0.6%Galapgos+0.6 Sed.10.960.560.2011.36254.8221.0656.663.56540.1810.8720.843.0210.092.300.813.431.950.290.201.681.170.81910.00
NW CR_20%DM+0.6%Galapgos+0.6 Sed.10.810.420.158.52191.1618.1442.722.67405.138.1515.632.267.591.770.632.821.670.
SE Nicaragua_5%FDM+0.4%Galapagos+0.4 Sed.11.391.000.3621.20528.0728.83122.926.011458.7218.9036.415.4218.924.341.455.352.710.390.374.261.971.92750.00
SE Nicaragua_10%FDM+0.4%Galapagos+0.4 Sed.11.150.500.2110.60283.6124.5774.103.02729.369.4718.612.9411.202.951.
SE Nicaragua_15% FDM+0.4%Galapagos+0.4 Sed.10.950.330.147.07189.9121.0350.992.01486.246.3112.411.977.662.140.783.411.950.290.121.440.660.64750.00
SE Nicaragua_20%FDM+0.4%Galapagos+0.4 Sed.10.800.250.105.30142.4718.1138.441.51364.684.739.311.485.771.650.612.791.670.
NW Nicaragua_10%FDM+0.2%Galapagos+0.6 Sed11.140.553.5911.57316.3124.5372.082.561047.648.6916.162.6610.572.901.
NW Nicaragua_15%FDM+0.2%Galapagos+0.6 Sed10.950.362.437.71211.8121.0049.591.70698.435.7910.781.787.232.110.773.401.950.290.111.840.640.79650.00
NW Nicaragua_20%FDM+0.2%Galapagos+0.6 Sed10.800.271.825.78158.9018.0937.391.28523.824.358.081.345.441.620.602.791.670.250.081.380.480.59650.00
NW Nicaragua_25%FDM+0.2%Galapagos+0.6 Sed10.680.221.464.63127.1215.6829.941.02419.063.486.471.074.361.310.492.331.440.
El Salvador_5%FDM+ 1%Sed21.461.460.2631.04799.3036.36122.454.641143.9418.8330.735.0619.194.701.636.253.420.500.323.492.300.82450.00
El Salvador_10%FDM+ 1%Sed21.190.730.1515.52416.4629.2072.072.33571.979.4215.532.6610.843.031.104.692.720.400.161.751.150.41450.00
El Salvador_15%FDM+ 1%Sed20.970.490.1010.35278.0423.7949.211.55381.316.2810.351.787.322.140.803.632.200.330.111.170.770.27450.00
El Salvador_20%FDM+ 1%Sed20.810.370.087.76208.5419.7037.031.17285.994.717.771.345.501.630.612.891.810.280.080.880.580.21450.00

[24] The melting model used in this study was aggregated fractional melting [Shaw, 1970] described by the following equation:

display math(1)

where CL is the average concentration of the element in the melt, C0 is the initial concentration of the element in the source, F is the melt fraction, and D0 is the initial bulk partition coefficient. Equation (1) is derived from the mass balance equation C0 = F × CL+(1 − F)CS, and the bulk partition coefficient D = CS/CL [Shaw, 1970]. CS is the concentration of the element in the solid phase. The partition coefficients used in our modeling (peridotite and eclogite sources) were from the compilation of Kelemen et al. [2003]. The DM composition was inverted from sample SO-144-1 from Werner et al. [2003]. Modeled melts in eclogite facies from the subducting Seamount Province and Cocos/Coiba Ridge were based on the average calculation of the data published by Hoernle et al. [2000] and Werner et al. [2003]. The subducting sediments were obtained from Patino et al. [2000].

[25] Our models were produced by metasomatism of depleted mantle (DM) by variable contributions of partial melts from the subducting slab. The modeling followed an optimization approach until a match was obtained for the trace element composition of the volcanic front lavas. The Salvadorian source was modeled with the highest sediment component of the four segments (1%) and no Galapagos component, while central Costa Rica was modeled with the lowest sediment component (0.1%) and the highest Galapagos component (3%). The metasomatized mantle (MM) was then melted (by aggregated fractional melting, equation (1) over a range of partial melts chosen to bracket the range of compositions found within each of these four segments (Figure 6). Trace element concentrations in all the segments are in an overall agreement with modeled values. To avoid fractionation correction, we restricted our samples to basalts and basaltic andesites (SiO2 < 55 wt %). Therefore, the values we are reporting here should not be considered compositions of primary magmas but as a test of the different components necessary to reproduce the volcanic front data. Modeled Ba/La and U/Th (Figure 6) values also reflect the overall decrease toward Costa Rica as seen in the data, although model values for U/Th are slightly elevated in the Northwest Nicaraguan segment. Nevertheless, the overall pattern was reproduced and modeled values for Ba, La, U, and Th closely match the actual data (Figures 2 and 6).

[26] According to the modeling results, the sediment component in El Salvador and Nicaragua plays a more important role than in the generation of Costa Rican magmas; however, this does not necessarily mean that more sediment is entering the subduction zone offshore Nicaragua. It is also possible that the sediment component makes up a smaller percentage of the MM source in Costa Rica because of the additional Galapagos component. This supports the conclusion of Carr et al. [2007] that the flux of incompatible elements to the volcanic front does not change significantly along strike. The model also shows that the subducting Galapagos tracks component decreases from Costa Rica to Nicaragua, which is expected because the Galapagos hotspot tracks collide with the margin in central Costa Rica [see details in Hoernle et al., 2008; Gazel et al., 2009, 2012].

4.2. Geochemical Variations Along and Across the Volcanic Front

[27] Central American volcanoes have a roughly symmetrical distribution of Ba/La, with a peak centered in western Nicaragua and lower values towards both ends of the volcanic front (Guatemala and central Costa Rica) (Figure 3). This ratio has been used as a geochemical proxy for contributions from the subducting slab, as Ba is a well-known fluid mobile element in subduction systems [e.g., Pearce and Parkinson, 1993; Plank and Langmuir, 1993; Pearce and Peate, 1995]. The high Ba content of Cocos Plate sediments is the cause of the high Ba/La in Central America relative to other arcs [Plank and Langmuir, 1993].

[28] Carr et al. [1990] suggested that high Ba/La occurs in Nicaragua because the slab's steep dip causes a more focused flux of material from the subducting slab. Figure 2, which contains data from both Nicaragua and Costa Rica [Carr et al., 2013], shows the along arc profile for Ba/La and also separates the two elements into their own plots, so that the effect of each on along arc variation of Ba/La can be better understood. This figure makes clear that Ba does not change significantly along strike. Instead, the variation is controlled by variations in La (Figure 2), a proxy for degree of partial melting or source enrichment. Furthermore, data from ODP sites 844, 845, 1039, and 495 reveal that the Ba profile in offshore sediments are remarkably similar throughout the Cocos Plate [Patino et al., 2000]. Carr et al. [2007] determined Ba fluxes from the volcanoes of Nicaragua and Costa Rica and found that there is no significant variation between these two segments of the volcanic front.

[29] 10Be/9Be has been used as a proxy for the subduction of the upper portion of the sediment column [Leeman et al., 1994] and a positive correlation has been observed between 10Be/9Be and Ba/La. This correlation is one of the reasons Ba/La was used as a tracer of the sediment signal [Carr et al., 2003]. We agree with Leeman et al. [1994] that 10Be/9Be is high in Nicaragua and low in Costa Rica because more sediment is subducted in Nicaragua, but this is only one reason why Costa Rica has such a low sediment signal. The other is that, as our current modeling shows (Figure 6), the influx of incompatible elements from the Galapagos hotspot tracks dilutes the signal from whatever small amount of sediment is able to enter the subduction zone.

[30] La/Yb coupled with Pb-isotopes are good discriminators between source enrichment and degree of partial melting in Central America, as this particular arc is not controlled by sediments [Feigenson et al., 2004; Hoernle et al., 2008; Gazel et al., 2009, 2012]. In the case of NW Nicaragua, La/Yb values are very low (<5) and the Pb isotopes are consistent with a DM mantle (206Pb/204Pb <18.6) [Feigenson et al., 2004; Hoernle et al., 2008; Gazel et al., 2009]. Thus, in NW Nicaragua the Ba/La variations can be explained by high melt fraction (controlled by flux melting) of a depleted mantle source that is only enriched in fluid mobile elements (Ba, B, Sr, etc.) by subduction processes. On the other hand, in southern Central America (central Costa Rica segment) highly radiogenic Pb isotopes (206Pb/204Pb > 18.8) correlate with high La/Yb (>10). Thus, an enriched component is necessary to explain the low Ba/La in this segment of the arc.

[31] Different explanations have been proposed for the higher degrees of partial melting in Nicaragua. Carr et al. [1990] first suggested that a steep slab dip in Nicaragua forced the slab-derived fluids into a smaller volume of mantle wedge that subsequently melted it to a higher degree of partial melting. However, determining slab dip beneath the volcanic front is difficult especially at margins like Central America, where the dip increases with depth. Syracuse and Abers [2006] compiled a global set of dips at arcs by measuring the average dip between 50 and 250 km. Using those depth contours, we estimated a dip beneath the volcanic front by fitting a second-order polynomial to the three contours closest to the volcano and evaluating the gradient at the volcano. The result is a broad region of maximum dip across all of Nicaragua and shallower dips to the NW and SE. The peak of slab dip in Central America is actually a relatively flat plateau encompassing all of Nicaragua. In contrast, the geochemical data have generally sharper peaks that all lie within this region of high and constant slab dip, which is expected if slab dip was a controlling factor. Ba/La peaks roughly at Cerro Negro (660 km in Figure 4), Sr isotopes peak at Masaya (740 km), U/Th peaks at Nejapa (720 km), and La/Yb has its minimum at Cerro Negro.

[32] Figure 7 shows three across arc profiles of the subducting slab [Syracuse et al., 2008] together with the overlying Caribbean Plate that demonstrates that the maximum possible height of the melting column varies from 35 km in central Costa Rica to 150 km in northern Nicaragua. Thus, fluid focusing coupled with the higher melting potential (extended melting column) in Nicaragua suggests that the degree of partial melting would be higher regardless of variations in slab hydration. Nevertheless, Ranero et al. [2003] discovered pervasive bending related faulting that reactivated ridge parallel faults now oriented nearly parallel to the trench. These faults allow the downgoing Nicaraguan slab to establish a steeper dip and may lead to increased serpentinization that can penetrate into the upper mantle of the subducting Cocos Plate. Alternatively, these cracks could simply act as more efficient pathways to deliver water from the slab into the overlying mantle wedge. Since this faulting is most pronounced in Nicaragua, it is also possible that the Nicaraguan slab carries more water into the subduction zone than in Costa Rica. Kelly [2003] placed the peak in fault relief at a distance along the arc between 670 and 720 km, using the same scale as Figure 4. Thus, the peak in fluid release should occur between volcanoes Momotombo and Nejapa or at the right step separating the western and eastern Nicaraguan segments. This is consistent with the location of the geochemical peaks and the substantial variation in the fault roughness height from 20 to 80 m is consistent with the sharp peaks in the geochemical ratios. Seismic refraction data suggests that the amount of water stored in the Cocos Plate may be 2.5 times higher offshore Nicaragua than the Nicoya Peninsula [Van Avendonk et al., 2011].

Figure 7.

Variation of the slab dip and depth along Central America. (a) Location of the seismic profiles along the volcanic front, (b) Northern Nicaragua segment, (c) Northern Costa Rica segment, and (d) Central Costa Rica segment. Note the systematic decrease in slab dip and depth from north to south. Slab structure from Syracuse et al. [2008] and crustal thickness from Carr [1984] and Hayes et al. [2013].

[33] The relatively high age of the subducting slab in NW Nicaragua can help explain the higher degrees of partial melting. The age of the slab currently subducting underneath Nicaragua is approximately 3 my older than the crust entering southern Costa Rica [Carr et al., 1990], which suggests it should be more highly serpentinized and carry additional fluids into the subduction zone [Rüpke et al., 2004], although this effect is likely to be minor. This is because the Nicaraguan slab has spent more time sitting on the ocean floor and also because it is cooler, which promotes increased serpentinization.

[34] In summary, all these mechanisms suggest that subducting slab beneath Nicaragua should be releasing more fluids, and thus, increasing the degree of partial melting within the mantle wedge. The measured water contents of melt inclusions along the volcanic front in Central America [Sadofsky et al., 2008] have a limited range and no large peak in Nicaragua. It is likely that the water contents of magmas are buffered to some extent as high degrees of melting lower the water content in the melt [Plank et al., 2013]. High degree of partial melting explains the depletions in La, and thus controlling the high Ba/La in Nicaragua since Ba does not change along the arc strike.

[35] Neither U nor Th concentrations vary significantly along the arc with the exception of Costa Rica, which is elevated in both (Figure 2). The incoming sediments do not vary significantly along strike in either element, which suggests that high levels in Costa Rica are controlled by other factors. Radiogenic isotopes [Gazel et al., 2009] together with trace-element models (Figure 6) suggest that Costa Rica is affected by a smaller percentage of sediment than Nicaragua and that the influence of the Galapagos derived hotspot tracks overwhelms the signal coming from the sediments themselves [Gazel et al., 2009]. While in Nicaragua high U/Th values are likely due to the higher U/Th values in the subducting sediments since the Carbonate Crash [Plank et al., 2002; Saginor et al., 2011].

4.3. Temporal Controls on Ba/La and U/Th in Central America

[36] In this section, we evaluate the variation of U/Th and Ba/La along volcanic front through time. 40Ar/39Ar ages [Plank et al., 2002; Saginor et al., 2011; Alvarado et al., 2012] in Costa Rica and Nicaragua have been compiled to create a ∼30 my record of geochemical evolution in the region (Figure 8). El Salvador does not have enough dated samples to include in this figure and two samples thought to be Oligocene-Mid-Miocene from the Balsamo Formation were instead found to be from only around 1Ma [Saginor et al., 2011].

Figure 8.

Evolution for key trace elements (Ba, La, U, and Th) in Costa Rica (blue) and Nicaragua (red) in the last 30 Ma [data from Plank et al., 2002; Carr et al., 2007; Gazel et al., 2009; Saginor et al., 2011; Carr et al., 2013]. Note how the Ba/La and U/Th ratios, used to trace the sediment component, clearly changed after 10 Ma, separating the trends of Costa Rica and Nicaragua. Although there is an overall increase in Ba in Costa Rica over time, the separation from Nicaragua is found in La, U, and Th, all of which dramatically increased in Costa Rica in the last 10 Ma following the interaction of the arc with Galapagos tracks [Gazel et al., 2009].

[37] Previous diagrams showing along arc variations in Ba/La and U/Th grouped samples into either “modern” or “Tertiary” volcanism [Carr et al., 1990; Patino et al., 2000; Plank et al., 2002; Carr et al., 2007; Saginor et al., 2011]. Figure 8 provides details into the evolution of Central American volcanism through time by including available ages for volcanism prior to the advent of the modern front. It is well documented that modern Ba/La values are higher in Nicaragua than in Costa Rica; however, Figure 8 demonstrates that was not always the case. Instead, along arc Ba/La was remarkably similar until regional variations appear higher in Nicaragua, lower in Costa Rica at approximately the time of the collision of the Coiba Ridge [∼12–8 Ma, Macmillan et al., 2004; Gazel et al., 2009] and during the Carbonate Crash. By looking at Ba and La independently (Figure 8), we see that this change in Nicaragua was driven by an increase in Ba, while in Costa Rica, the change was driven by an increase in La.

[38] In Nicaragua, the volcanic front has migrated trenchward since the Miocene [Plank et al., 2002, Saginor et al., 2011], which means that it is now closer to the major source of Ba, the subducting sediments. This may explain the Ba increase in Nicaragua through time. Figure 8 also shows that La has not changed in Nicaragua since the Miocene. Tamarindo samples from Nicaragua reinforce the case that changes in Ba are driving local changes in Ba/La. Several of these samples appear distinctly elevated (they are circled in Figure 8) in both Ba/La and Ba when compared to other Nicaraguan samples of similar ages. The Tamarindo Formation is coeval with Middle Miocene volcanism in the Nicaraguan Highlands; however, it was emplaced 80 km closer to the trench [Saginor et al., 2011]. Another possibility for the Ba increase in both Nicaragua and Costa Rica is an increase in primary productivity due to upwelling following the initial closure of the isthmus in response to the collision of the Coiba and then the Cocos Ridge. Cocos Plate sediments do show a modest increase in Ba following this transition; however, this change affected the entire region, not just Nicaragua. Also, as previously discussed, Ba/La levels are far more sensitive to changes in the degree of partial melting and therefore changes in La.

[39] In Costa Rica, the decrease in Ba/La is controlled not by Ba (which increases with time along with Nicaragua), but rather by a sharp increase in La. The change in La occurs because the mantle source is enriched by metasomatic interaction with Galapagos-derived tracks [Gazel et al., 2009, 2012]. Although melt fraction may also be lower, source enrichment is more consistent with the change of La/Yb together with contemporaneous changes to Galapagos-derived isotopic signature [Hoernle et al., 2008; Gazel et al., 2009, 2012].

[40] The modest Ba enrichment in Costa Rica over time may be due to the high Ba concentrations in the Galapagos tracks; however, since this increase is also seen in Nicaragua as well, this explanation may be insufficient. One consequence of the arc changing shape over time is that the slab dip and the arc-trench distance may well have changed as well. For the present arc, there is some evidence that Ba/La decreases as arc-trench distance increases. The clearest example occurs in the volcanoes around the Gulf of Fonseca, which is bounded by El Salvador, Honduras, and Nicaragua (Figure 9). There is a right step in the volcanic front northwest of Cosigüina Volcano and there is an unusually wide distribution of composite cones. Cosigüina is 165 km from the trench and Zacate Grande is 30 km further back. Yet across these 30 km, Ba/La decreases from about 100 to about 50. Most of the cross-arc variation in Ba/La occurs in the first 30 km behind the volcanic front where the depth to the seismic zone increases from 116 km at Cosigüina to 200 km at Zacate Grande. Zacate Grande also marks the furthest landward extent of the Wadati-Beniof Zone. The wide variation within individual volcanic centers makes it doubtful that cross-arc gradients in Ba/La can be reliably recognized across the short distance within a volcanic center.

Figure 9.

Across arc variation in Ba/La, Ba, and La. Conchagua, Conchaguita, Meanguera, and Cosiguina are part of the active volcanic front. El Tigre appears morphologically younger than Conchagua but has no historic activity. Zacate Grande is a mix of older andesites and young basalts, although there are no available ages. In the backarc region, the gradient in Ba/La continues [Patino et al., 1997]. Esteli and Tegucigalpa are lava fields of subalkaline basalts and basaltic andesites erupted in grabens well behind the volcanic front. Yojoa is a very young alkaline basalt complex adjacent to the strike-slip Caribbean-North American plate boundary.

[41] Figure 9 shows that Ba/La increases toward the trench; however, this could be either due to increasing Ba flux from the subducted sediments or an increase in La, and therefore the degree of partial melting (assuming a constant mantle composition). When these elements are examined separately, it is clear that Ba does not change systematically across the arc, while La decreases steadily towards the trench.

[42] U/Th values have also shifted through time. Figure 8 shows that U/Th was similar along the arc until the collision of the Coiba Ridge at 10 Ma, when the ratio decreased in Costa Rica and increased in Nicaragua. U shows no along arc variation until 10 Ma when it increased in both countries, but to a greater extent in Costa Rica. Th has no along arc variation until 10 Ma when Th increased in Costa Rica, but remained the same in Nicaragua. In Costa Rica, a modest increase in U coupled with a greater increase in Th produced the temporal change in U/Th. The measured increase in U in the subducting sediments since the Carbonate Crash is a possible explanation; however, that does not explain why Costa Rican volcanoes currently have more U than Nicaraguan volcanoes. We suggest that the increase in Costa Rican U since the Miocene is also influenced by the subduction of trace-element enriched Galapagos tracks, which are absent in Nicaragua. The increase in U/Th in Nicaragua is more straightforward and can be explained by the sediment transition from low U carbonates to high U hemipelagics.

5. Conclusions

[43] Variations in Ba/La along the arc and through time are primarily controlled by La, a proxy for the degree of partial melting or source enrichment. This is true for both Miocene and modern volcanics. Changes in Ba/La in Central America through time can be controlled by changes in Ba only if the degree of partial melting does not change (Ex: Nicaragua since the Miocene).

[44] Nicaragua has the highest degree of partial melting along the Central American volcanic front. A number of factors can help explain the high melt fractions in Nicaragua. First, a steeper slab dip under Nicaragua creates a higher melting column, which focuses the fluid flux from the subducting slab. Second, the distribution of trench parallel faulting related to slab bending results in deep faults that appear to be allowing an added influx of water in Nicaragua and the water leads to a higher degree of partial melting. Third, variations in the age of the subducting slab can affect the degree of serpentinization of subducting ocean crust and the upper mantle.

[45] Plank et al. [2002] identified a long-term high level (>70) of Ba/La along the Nicaraguan volcanic arcs. New age data and geochemical data better define the origin of the very high Ba/La values found in Nicaragua. The first occurrence of Ba/La >100 was about 15 Ma in the Tamarindo Formation at the very front of the then active volcanic belt. Consistently high values of Ba/La began sometimes after 9 Ma and after the collision of Galapagos tracks and Carbonate Crash. Overall, the Ba concentrations of lavas appear to have increased gradually from the earliest samples at around 25 Ma to the present.

[46] U/Th values in Nicaragua experienced a significant increase since the Miocene due to increases in U-bearing sediment following the Carbonate Crash. While U also increased in Costa Rica, U/Th remained low due to increases in Th derived from the subducting Galapagos tracks.


[47] This study was supported by NSF Tectonics Program grant EAR- 1221414 to Gazel, and through the NSF Margins Program, grants EAR0203388 and OCE 0505924 to Carr, Saginor, and Condie. We would like to thank Editor Cin-Ty Lee and two anonymous reviewers, as well as C. Saginor and N. Sou whose comments and revisions greatly improved this manuscript.