Contrasting He^C Relationships in Nicaragua and Costa Rica: Insights Into C Cycling Through Subduction Zones§

Available online at www.sciencedirect.com

Earth and Planetary Science Letters 214 (2003) 499^513 www.elsevier.com/locate/epsl

Contrasting He^C relationships in Nicaragua and Costa Rica: insights into C cycling through subduction zones§

Alison M. Shaw a;, David R. Hilton a, Tobias P. Fischer b, James A. Walker c, Guillermo E. Alvarado d

a Fluids and Volatiles Laboratory, Scripps Institution of Oceanography, La Jolla, CA 92093-0244, USA b Department of Earth and Planetary Sciences, University of New Mexico, Albuquerque, NM 87131-1116, USA c Northern Illinois University, DeKalb, IL 60115, USA d AŁ rea de Amenaza y Auscultacio¤nS|¤smica y Volca¤nica, ICE, Apdo. 10032-1000, San Jose¤, Costa Rica

Received 18 March 2003; received in revised form 24 June 2003; accepted 10 July 2003

Abstract

3 4 13 We report He/ He ratios, relative He, Ne, and CO2 abundances as well as N C values for volatiles from the volcanic output along the Costa Rica and Nicaragua segments of the Central American arc utilising fumaroles, 3 geothermal wells, water springs and bubbling hot springs. CO2/ He ratios are relatively constant throughout Costa Rica (av. 2.1U1010) and Nicaragua (av. 2.5U1010) and similar to arcs worldwide (V1.5U1010). N13C values range from 36.8x (MORB-like) to 30.1x (similar to marine carbonate (0x)). 3He/4He ratios are essentially MORB- like (8 þ 1 RA) with some samples showing evidence of crustal He additions ^ water spring samples are particularly susceptible to modification. The He^CO2 relationships are consistent with an enhanced input of slab-derived C to magma sources in Nicaragua ((L+S)/M = 16; where L, M and S represent the fraction of CO2 derived from limestone and/or marine carbonate (L), the mantle (M) and sedimentary organic C (S) sources) relative to Costa Rica ((L+S)/ M = 10). This is consistent with prior studies showing a higher sedimentary flux to the arc volcanics in Nicaragua (as traced by Ba/La, 10Be and La/Yb). Possible explanations include: (1) offscraping of the uppermost sediments in the Costa Rica forearc, and (2) a cooler thermal regime in the Nicaragua subduction zone, preserving a higher proportion of melt-inducing fluids to subarc depths, leading to a higher degree of sediment transfer to the subarc mantle. The 10 absolute flux of CO2 from the Central American arc as determined by correlation spectrometry methods (5.8U10 3 10 mol/yr) and CO2/ He ratios (7.1U10 mol/yr) represents approximately 14^18% of the amount of CO2 input at the trench from the various slab contributors (carbonate sediments, organic C, and altered oceanic crust). Although the absolute flux is comparable to other arcs, the efficiency of CO2 recycling through the Central American arc is surprisingly low (14^18% vs. a global average of V50%). This may be attributed to either significant C loss in the forearc region, or incomplete decarbonation of carbonate sediments at subarc depths. The implication of the latter

* Corresponding author. Present address: Department of Terrestrial Magnetism, Carnegie Institution of Washington, Washing- ton, DC 20015, USA. E-mail address: [email protected] (A.M. Shaw).

§ Supplementary data associated with this article can be found at doi:10.1016/S0012-821X(03)00401-1

EPSL 6782 2-9-03 500 A.M. Shaw et al. / Earth and Planetary Science Letters 214 (2003) 499^513 case is that a large fraction of C (up to 86%) may be transferred to the deep mantle (depths beyond the source of arc magmas). ß 2003 Elsevier B.V. All rights reserved.

3 Keywords: volatile recycling; Central America; CO2/ He; helium isotopes; carbon isotopes

1. Introduction goal is to assess whether a balance exists between CO2 input at the trench and CO2 emanating from Understanding how physical features of sub- Central American volcanoes ^ as has been postu- duction zones in£uence the chemistry of arc mag- lated for N2 [6]. mas is a fundamental issue in subduction-related A ¢nal aim is to consider the nature of the studies. Prior studies have focused on the Central volcanic output of carbon in relation to its prov- American volcanic arc because it is characterised enance on the subducting slab. By coupling CO2 by dramatic variations in the angle of subduction, and He measurements (isotopes and relative the thickness of the arc crust through which mag- abundances), we can identify and quantitatively mas erupt and the faulting style of the subducting assess the various contributors to the magmatic plate [1^3]. Several geochemical parameters, in- output from the various subduction zone reser- cluding Ba/La, 10Be, B, La/Yb and N15N, have voirs: (a) the mantle wedge, (b) the overlying been shown to e¡ectively track changes along arc crust through which the magmas erupt and the arc due to variations in these subduction forc- (c) the subducting slab ^ both the oceanic base- ing functions [1,4^8]. Of particular interest is the ment and sedimentary veneer of the slab. Helium Nicaragua^Costa Rica section of the arc, since the isotopes are sensitive indicators of mantle deriva- volcanic outputs in these adjacent segments show tion and crustal in£uences, whereas N13C values remarkable geochemical di¡erences. For example, can be used to identify mantle (M), marine car- particularly low 10Be and B concentrations are bonate/limestone (L) and organic sedimentary (S) observed in the volcanic output in Costa Rica components [11]. A previous study investigating versus extremely high values measured in Nicara- He^C relationships in the North Fiji back-arc ba- guan volcanics [7^9]. These geochemical data pro- sin [12] suggested that signi¢cant fractionation of vide evidence for a high slab £ux and recycling of L and S components occurs during the subduc- sediments beneath Nicaragua, but with a minimal tion process. Since the sedimentary input on the slab contribution to the volcanic output in Costa Cocos plate is well characterised in terms of L and Rica, despite similar sedimentary inputs at the S components, our study of the He^C relation- trench. Therefore, a major aim of this study is ships along the volcanic front in Central America to assess how regional di¡erences in subduction will be able to test this conclusion. style a¡ect the volatile systematics ^ particularly He^C isotopes and relative abundances ^ of the arc output in Nicaragua and Costa Rica. 2. Geologic setting and background The Central American arc also presents a unique opportunity to consider C cycling through Volcanoes in Costa Rica and Nicaragua form subduction zones since the subducting sedimenta- the southern segment of the Central American ry sequence on the incoming Cocos plate is espe- volcanic arc which extends from central Costa cially carbonate-rich (26.55 wt% CO2 [10]), result- Rica to western Guatemala (see Fig. 1). Volca- ing in an extremely high in£ux of carbon into the nism along this active margin is the consequence subduction zone. Determining whether this man- of Cocos plate subduction beneath the Caribbean ifests itself in an enhanced volcanic CO2 £ux plate. This region has been the focus of numerous along the arc relative to other arcs worldwide is geochemical studies (e.g. [1,6,9,13,14]) since there an important objective of this study. A related are striking changes in subduction forcing param-

EPSL 6782 2-9-03 A.M. Shaw et al. / Earth and Planetary Science Letters 214 (2003) 499^513 501 eters (e.g. faulting style of incoming plate, slab these lavas has been attributed to either sediment dip, crustal thickness) whose in£uence can be as- underplating, whereby sediments are e¡ectively sessed at a relatively ¢ne scale given the narrow scraped o¡ the down-going plate before reaching spacing of volcanoes (V25 km) compared with the zone of magma generation [15], or subduction that found in other arc settings (40^70 km) [1]. erosion [9,16,17]. In the former case, the process Despite comparable subduction rates (V8^9 of sediment underplating may be related to fault- cm/yr) and incoming sediment compositions (in- ing style. A recent study [3] looking at the rela- ferred from similar lithologies in Ocean Drilling tionship between outer rise faulting and geochem- Project (ODP) site 1039 and Deep Sea Drilling ical variability along the arc showed how large Project (DSDP) site 495 o¡shore Costa Rica o¡sets in reactivated trench-parallel faults on the and Guatemala, respectively), remarkable geo- plate subducting beneath Nicaragua could protect chemical di¡erences are observed in the volcanic sediment from o¡scraping, resulting in a high output in Nicaragua and Costa Rica. For exam- sediment £ux being transferred into the zone of ple, Morris et al. [7] measured exceptionally high magma generation. In contrast, the relatively 10Be enrichments in lavas from Nicaragua (10Be/ small displacement of faults (new) in the slab sub- Be up to 59.9U10311) while anomalously low val- ducting beneath Costa Rica would favour o¡- ues were found in neighbouring Costa Rica (10Be/ scraping and thus limit the sediment £ux in the 311 10 Be = 1.4U10 ). Be is a short-lived (t1=2V1.5 volcanic output, as observed. An alternative ex- Myr) cosmogenic radionuclide that becomes con- planation for the discrepancy in 10Be and other centrated in the uppermost oceanic sediments. Be- slab signals is the idea of subduction erosion, cause postulated mantle 10Be contents are low, where any slab signature is essentially diluted by any enrichments observed in erupted arc lavas material eroded from the underside of the upper are indicative of contributions from recently sub- plate during subduction [9,16,17]. Seismic evi- ducted sediment. Likewise, Ba/La ratios, often dence reveals that as seamounts are subducted, used to trace a slab-derived £uid phase, are high they can e⁄ciently erode the upper plate by a in Nicaragua and low in Costa Rica [4,5]. Addi- process referred to as seamount tunnelling [17]. tionally, low La/Yb ratios, generally attributed to Seamounts associated with the Galapagos hotspot high degrees of partial melting, are found in Nic- are common features observed on the Cocos plate aragua [4]. It has been suggested that a higher subducting beneath Costa Rica, but they are vir- slab £ux could be responsible for enhanced melt- tually non-existent on the downgoing plate in ing of the wedge beneath Nicaragua [4]. The re- Nicaragua. The reason why subduction erosion sults of these ¢ndings suggest that Nicaraguan is enhanced in Costa Rica relative to Nicaragua volcanics are more strongly in£uenced by a slab- may be related to the di¡erence in sea£oor mor- sediment signal than volcanoes in Costa Rica. phology on the subducting plate [16^18]. The contrasting geochemical characteristics observed along the arc have been attributed to variations in physical subduction parameters such as: 3. Analytical techniques (1) crustal thickness, (2) angle of dip, (3) sediment underplating, (4) faulting style and (5) subduction Geothermal £uid samples were collected from erosion [1,3^5,7,9,15]. Relatively thinner crust (32 fumaroles, bubbling hot springs, water springs km) and a higher angle of slab dip (65^75‡ below and geothermal wells using evacuated AR-glass the volcanic front) in Nicaragua a¡ord a geome- £asks and/or copper tubes (see [19] for further try conducive to a high £ux of slab £uids being details on collection procedures). Samples were transferred to a relatively small volume of asthe- extracted on a vacuum line using a series of traps nosphere, leading to a high degree of melting at liquid nitrogen and acetone/dry ice tempera- [1,4]. Although Costa Rica is characterised by tures to isolate any water vapour and to separate shallower dips (35^65‡) and slightly thicker crust the condensable gas fraction (mainly CO2) from (40 km), the virtual absence of a 10Be signal in the non-condensable fraction. The condensable

EPSL 6782 2-9-03 502 A.M. Shaw et al. / Earth and Planetary Science Letters 214 (2003) 499^513

Fig. 1. Regional map of Central America showing the location of the major volcanic centres in Costa Rica and Nicaragua. The Cocos plate is presently subducting beneath the Caribbean plate at a rate of 82^88 mm/yr [2]. Deep-sea drill site 1039 from ODP leg 67 [33] is shown for reference. fraction was frozen into a Pyrex0 breakseal, and 21 (Nicaragua). Sample locations are given in whereas the non-condensable fraction was puri- Fig. 1, and the complete data set is plotted as a ¢ed using a hot Ti getter and activated charcoal function of latitude in Fig. 2. traps. A calibrated split of this gas was aliquoted into an AR-glass breakseal. Helium and neon 4.1. 3He/4He ratios abundance measurements, as well as the 3He/ 4 3 4 He ratios, were measured on a MAP 215 noble All measured He/ He ratios (reported as (Rm/ 3 4 gas mass spectrometer. The CO2 fraction of the RA) where Rm = measured He/ He of sample and 3 4 condensable gas was further puri¢ed using a var- RA = He/ He of air) have been corrected for the iable temperature trap, and abundances were e¡ects of atmospheric contamination (to RC/RA). measured manometrically. The N13C values were Agreement between air-corrected duplicate sam- measured from an aliquot of the CO2 gas using a ples generally falls within 0.2 RA with the excep- VG Prism stable isotope mass spectrometer. tion of samples from Poa¤s and San Jacinto where duplicates vary by 0.38 and 0.36 RA, respectively. Air-corrected 3He/4He ratios span a wide range 4. Results in values (from 0.74 RA to 8.1 RA) indicating the 3 4 in£uence of both crustal ( He/ HeW0.05 RA) and 3 4 3 We report He/ He ratios, CO2/ He ratios, He abundances (for water spring samples) and C isotope results for samples in Tables 11 (Costa Rica) 1 See Tables 1 and 2 in the online version of this paper.

EPSL 6782 2-9-03 A.M. Shaw et al. / Earth and Planetary Science Letters 214 (2003) 499^513 503

Turrialba (8.1 RA), Poa¤s (7.6 RA) in Costa Rica and at Mombacho (7.6 RA) in Nicaragua. Indeed, most of fumarole samples, irrespective of location, lie within the typical MORB range (8 þ 1 RA), consistent with values observed at other arcs worldwide (e.g. [19,22,23]). The principal exception to this observation is a £ank fumarole sample from Masaya, where air contamination is a major problem (the measured He/Ne is exceptionally low). The lowest 3He/4He values (0.7^3.4 RA) are found in water spring samples, and point to a signi¢cant radiogenic He contribution. Based on prior arguments against He being subducted into the mantle [24], the presence of radiogenic He likely re£ects contamination processes in the uppermost crust (see also Section 5.1). The bubbling hot spring and geothermal well samples show varying degrees of crustal interaction with 3 4 He/ He ratios ranging from 2 RA to MORB-like values.

Fig. 2. Along-arc variations in CO /3He, N13C (CO ), and 13 2 2 4.2. N C (CO2) 3He/4He results for geothermal £uids from Costa Rica and Nicaragua. The shaded boxes show: (1) the CO /3He range 2 N13 for arcs worldwide: 1.5 þ 1.1U1010 [27], (2) the MORB range The C (CO2) values of most samples (Fig. 2) of N13C values: 36.5 þ 2.5x [11], and (3) the 3He/4He range are s 34x, or higher than the nominal range for MORB: 8 þ 1 RA [19]. found in MORB (36.5 þ 2.5x [11]). This includes all sample types (fumaroles, geothermal wells, spring waters and bubbling hot springs) from both Nicaragua and Costa Rica sections of (upper) mantle (8 þ 1 RA) contributions [19] to the arc. The highest value observed in this study is Central American volcanism. Although other 30.1x at Rinco¤n de la Vieja in northern Costa 13 studies (see [20,21]) have argued for a plume-like Rica. N C (CO2) values lower than 34x are mantle beneath Costa Rica (associated with the observed in fumaroles from Poa¤s(36.8x), and Galapagos hotspot), we ¢nd no indication of in various spring water samples from both Costa 13 plume-like He isotope values anywhere along the Rica and Nicaragua. The lowest N C (CO2) val- arc. MORB (mid-ocean ridge basalt)-like and pre- ues (313.4x) were found in water samples from dominantly radiogenic 3He/4He ratios are found Arenal (Costa Rica). The signi¢cance of the large 13 in both Costa Rica and Nicaragua. In addition, range in N C (CO2), in terms of crust versus slab/ there is no evidence for systematic variations in mantle wedge inputs to the carbon inventory, is He isotopes along the strike of the arc (Fig. 2), considered in Section 5.1. despite a decrease in crustal thickness towards the 3 north. Based on our observations, we note that 4.3. CO2/ He much of the observed variation is related to the 3 type of geothermal media sampled (i.e. fumaroles, The CO2/ He ratio can be used as a sensitive bubbling hot springs, water springs and geother- indicator of carbon provenance since the Earth’s mal wells). various reservoirs (crust/mantle/atmosphere) have The highest 3He/4He values are found in fuma- diagnostic values [25,26]. In addition, it is an im- role samples from both segments of the arc ^ at portant parameter to constrain in order to calcu-

EPSL 6782 2-9-03 504 A.M. Shaw et al. / Earth and Planetary Science Letters 214 (2003) 499^513

3 plot CO2/ He ratios as a function of dissolved He concentration for all water spring samples (except for samples from San Francisco Libre (N-15, I-118) ^ discussed below). Most samples are en- riched in He relative to air-saturated water (i.e. 3 44.1 ncm He/g H2O at 25‡C [29]). However, thermal waters from Telica and Recreo Verde have exceptionally low values, 16.0 and 14.2 ncm3 He/g H2O respectively. These low values represent highly degassed samples, where He loss has likely occurred during formation of a vapour phase. 3 The trend of increasing CO2/ He ratios with decreasing He contents for other samples can also 3 be explained by the preferential partitioning of He Fig. 3. Water spring CO2/ He ratios plotted as a function of dissolved He contents. Note the consistent trend of increas- into a vapour phase and subsequent loss since its 3 ing CO2/ He ratios with decreasing He contents. Samples solubility in water is much lower than that of from Recreo Verde and Telica have He contents signi¢cantly CO . 3 2 lower that the air-saturated water (ASW) value of 44.1 ncm Duplicate samples from San Francisco Libre STP/g H2O (shown with a dashed line). Two samples from San Francisco Libre are not plotted since they have excep- show particularly high He concentrations (2970 3 tionally high He contents: they also have low 3He/4He values and 3230 ncm He/g H2O), and exceptionally 3 8 8 indicating severe crustal contamination. The solid line, low CO2/ He values (2.2U10 and 3.6U10 ). bounded by the dotted lines representing the 95% con¢dence This locality, however, is relatively far ( s 20 limits, is a linear regression through all water spring samples km) from the volcanic centre of Momotombo except those mentioned above. The shaded region represents 3 and thus the observed He^C characteristics may the worldwide average CO2/ He ratio for arc-related geothermal £uids. be dominated by crustal in£uences. This is consistent with the low 3He/4He ratios (1.79 and 1.88 RA), indicating addition of radiogenic He.

3 late volcanic CO2 £uxes [27]. The CO2/ He ratios of all samples (with the exception of samples from 5. Discussion Tipitapa (N-32, N-33) and San Francisco Libre (N-15, I-118)) are signi¢cantly higher than values In order to evaluate the integrity of our results, 3 9 found at mid-ocean ridges (CO2/ He = 2U10 we must ¢rst determine which samples may have [27]), presumably due to the addition of either been a¡ected by crustal contamination and/or slab-derived or crustal carbon (see Fig. 2). We near-surface processes. Once compromised sam- 3 ¢nd that most samples have CO2/ He ratios that ples have been identi¢ed and removed from fur- lie within the range found at arcs worldwide ther consideration, we can use combined CO2^He (1.5 þ 1.1U1010 [28]). We note, however, that all systematics to determine the provenance of the C water spring samples have particularly high values [11]. In this way, we can determine the relative 3 13 (CO2/ He up to 6U10 ). In addition, bubbling contributions from the various C reservoirs (man- hot spring samples from Telica and geothermal tle, sediments, slab) and evaluate whether subduc- well samples at Miravalles show ratios which are tion of the carbonate-rich Cocos plate sediments elevated with respect to average arc values. results in an enhanced C output along the Central American arc, as compared with subduction zones 4.4. Dissolved He contents in water spring samples dominated by siliceous sedimentary inputs. By identifying the source of volatiles, we aim to esti- We observe a wide range in He concentrations mate the amount of C which is recycled via the 3 from 14.2 to 3230 ncm He/g H2O. In Fig. 3 we subduction zone factory.

EPSL 6782 2-9-03 A.M. Shaw et al. / Earth and Planetary Science Letters 214 (2003) 499^513 505

4 HeV0.05 RA) relative to mantle values (8 þ 1 RA). A compilation of arc-related volcanic and geothermal £uids at arcs worldwide yields an 3 4 average He/ He value of 5.37 þ 1.87 RA [19]; however, it is important to note that this database includes samples which have been a¡ected by 3 4 crustal additions ( He/ He as low as 0.01 RA). In the subsequent sections, we will only consider samples with air-corrected 3He/4He values higher than 5.4 RA, thereby avoiding samples in£uenced by signi¢cant crustal inputs. 3 In the case of CO2/ He ratios, values in crustal material can vary widely depending on rock li- thology; however, given the virtual absence of 3 3 primordial He in all types of crust, CO2/ He ratios are predominantly high (1011^1013) [26].In 4 3 Fig. 4. Ternary CO2^ He^ He plot of all geothermal samples Fig. 4 we note that water spring samples along illustrating the e¡ects of air and crustal contamination. with some of the geothermal well samples trend Dashed lines bound the upper mantle range of 3He/4He ra- 3 4 towards a high CO2/ He (and CO2/ He) crustal tios (8 þ 1 RA). Samples in£uenced by crustal contamination plot within the shaded region of the diagram, bounded by endmember. 3 calculated mixing lines between typical arc (CO2/ He = In addition to crustal contamination, atmo- 9 3 4 1.5 þ 0.6U10 , He/ He=8þ1 RA) and crustal endmembers spheric gases can in¢ltrate magmatic systems via 3 U 12 3 4 (CO2/ He = 5^50 10 , He/ He = 0.05 RA). We note that all air-saturated recharge waters, thereby masking water spring samples (open triangles) show signi¢cant crustal source characteristics. Samples which have low contamination. Some samples show contamination from an air-derived component and plot close to mixing lines between X values have su¡ered signi¢cant air contamina- an arc endmember and air (dotted line) or air-saturated tion and as such are unreliable tracers of mantle water (ASW; dash-dotted line). The N2/Ar ratios of samples processes. Samples N-7, N-6, N-14 and CR-14 in this region of the diagram (with the exception of La Mari- have particularly low X values (2.5^3.1) and their na) con¢rm the presence of an air-derived contaminant (N2/ He^C systematics are overwhelmed by an air Ar = 52^96, where N2/ArairW80 and N2/ArASWW40). The 3 component (see Fig. 4). Although samples from CO2/ He ratio of air is calculated using [He] = 5.24 ppm, 3 4 36 He/ He = 1.4U10 [48] and [CO2] = 370.9 ppm [49]. The Tipitapa and San Francisco Libre plot near the 3 CO2/ He of air-saturated water is calculated using air con- air-contaminated samples, they have high X val- centrations and Ostwald coe⁄cients at 15‡C (He = 0.009325, ues (112.9^240.1) and thus have been in£uenced CO2 = 1.071 [50]). 3 by a low CO2/ He crustal component rather than atmospheric gases. 5.1. Integrity of results Given the variety of sample types collected during this study, the present data set allows for a 4 3 In Fig. 4 we plot samples on a CO2^ He^ He realistic evaluation of the integrity of di¡erent ternary diagram to identify the e¡ects of crustal sampling media for tracing mantle processes. A contamination. There are a number of criteria clear conclusion is that all water spring samples which may be used to recognise contaminated show evidence of crustal contamination: low 3He/ samples, i.e. samples deemed unrepresentative of 4He values, low X values, extreme N13C values 3 4 3 the underlying magma source: (1) low He/ He (313.4 to 31x) and generally high CO2/ He 3 values, (2) high CO2/ He ratios and (3) low X ratios. Such samples are thus considered highly values (where X = air-normalised He/Ne ratio; susceptible to near-surface fractionating process- see Table 1 caption). Helium isotopes are highly es, which have modi¢ed C isotopes to extreme sensitive indicators of crustal in£uences since values (see [14] for further details). crustal production values are low (3He/ It has been argued that geothermal well sam-

EPSL 6782 2-9-03 506 A.M. Shaw et al. / Earth and Planetary Science Letters 214 (2003) 499^513 ples are the most pristine samples since they orig- by in¢ltration of surrounding groundwaters into inate from deep reservoirs (1^2 km depth) and are the geothermal system. Based on our observations less prone to near-surface boiling and degassing at two sites, we conclude that although young [14]. To test this hypothesis we sampled at two geothermal wells can be exploited for volcanic di¡erent geothermal well sites on the £anks of gases, results from older wells should be treated active volcanoes ^ Momotombo and Miravalles with caution. ^ and compared the results to other sampling media (fumaroles in Momotombo’s crater and bub- 5.2. Sources of CO2 bling hot spring gas on Miravalles’ £ank). Results are given in Table 3. At Momotombo volcano, we One approach used to assess carbon prove- sampled two di¡erent wells (OM-53 and MT-43) nance at subduction zones is the three-component and found that a 2-month-old well (sample N-16) mixing model of Sano and Marty [11]. Based on yielded results which were essentially indistin- He^C characteristics, samples can be described in guishable from the summit fumarole samples terms of carbon mixtures derived from three end- (N-11, N-31); however, an older well that had members: limestone and/or marine carbonate (L), been in production for over 2 years showed evi- the mantle (M) and sedimentary organic C (S). dence of crustal contamination (lower 3He/4He, Using the following equations, the relative contri- slightly lower X values and higher N13C values ^ butions (expressed as fractions, f) from the vari- see Table 2). Likewise, geothermal well samples ous sources can be determined: from the power plant at Miravalles (CR-10, CR- ð13C=12CÞ ¼ 20 and CR-21) showed signi¢cantly lower 3He/ obs 4He ratios than bubbling hot spring samples on 13 12 13 12 13 12 fMð C= CÞM þ fLð C= CÞL þ fSð C= CÞS Miravalles’ £ank (CRT-9). A possible explanation 12 3 is that air-derived volatiles may be introduced 1=ð C= HeÞobs ¼ into geothermal wells either by injection wells or 12 3 12 3 12 3 fM=ð C= HeÞM þ fL=ð C= HeÞL þ fS=ð C= HeÞS Table 3 Comparison of geothermal well gas samples to other sam- fM þ fL þ fS ¼ 1 pling media at Momotombo and Miravalles volcanoes 3 N13 Site Medium RC/RA CO2/ He C Endmember compositions used in prior studies x ( ) [11,28] are M: N13C=36.5x,C/3He = 1.5U109 ; Momotombo volcano (Nicaragua) L: N13C=0x,C/3He = 1013 ;S:N13C=330x, U 10 3 fumarole 7.0 2.7 10 2.6 C/3He = 1013. fumarole dup a ^^ 32.6 well MT-43b 6.4 3.1U1010 31.3 There are two potential problems with this ap- well MT-43 dup 6.4 3.3U1010 31.1 proach [19,22]: (1) the e¡ect of crustal volatiles is new well OM-53c 6.9 3.6U1010 32.6 not considered and (2) the N13C value of organic Miravalles volcano (Costa Rica) matter (330x) may be unrepresentative since it U 10 3 bubbling hot spring 6.8 1.3 10 3.5 may be fractionated to heavier values upon sub- well PTM-45 5.2 6.6U1010 31.2 well PTM-45 dup 5.3 7.8U1010 31.2 duction (see [30^32]). To circumvent these prob- well PGM-8 3.6 14.8U1010 31.6 lems, we have removed samples showing crustal All errors are given in Tables 1 and 2. in£uences from our dataset (Section 5.1) and cal- a Samples indicated with a dup represent duplicates of pre- culated L, M and S fractions using an extreme ceding sample. estimate of the equilibrium N13C value of organic b MT-43 was sampled 22 months earlier by Snyder et al. matter, N13C=312x (see [30]), in addition to 3 4 [14], yielding a He/ He of 6.7 RA ; thus we note a drop in the above endmembers (for comparison). The re- the isotopic composition since then (to 6.4 RA) possibly due to dilution by re-injection wells ^ see text for discussion. sults of these calculations are shown in Table 4. c Geothermal well OM-53 was 2 months old at the time of An important parameter used to compare out- sampling, whereas site MT-43 was over 2 years old. put £uxes in di¡erent arcs is the L/S ratio (i.e. the

EPSL 6782 2-9-03 A.M. Shaw et al. / Earth and Planetary Science Letters 214 (2003) 499^513 507 fraction of C derived from a limestone or marine m of pelagic carbonates overlain by V175 m of carbonate source to that from a sedimentary or- hemipelagic diatom-rich mud [33,34]. Since there ganic component). The worldwide average arc L/S is no di¡erence in the subducting lithologies in value (from [28]) is 6.0 þ 3.8 (calculated assuming Costa Rica and Nicaragua, it may not be surpris- S=330x) whereas the mean values calculated ing that their L/S ratios are similar. However, for Costa Rica and Nicaragua are 9.6 þ 4.3 and prior studies have emphasised striking di¡erences 11.1 þ 2.1, respectively (Table 4). Such high values in the geochemical characteristics (Be, B, La/Yb, presumably re£ect the carbonate-rich nature of Ba/La, etc.) of lavas from Costa Rica and Nicara- the sedimentary sequences being subducted. gua, so we would have anticipated a higher pro- Deep-sea drill sites o¡ Guatemala (DSDP 495) portion of sedimentary-derived C in Nicaragua (a and Costa Rica (ODP 1039) show little variation lower L/S). For example, assuming that the hemi- in sedimentary composition, consisting of V250 pelagic portion of the sediment column in Costa

Table 4 L-M-S calculations for Costa Rica and Nicaragua geothermal £uids a a a a b Volcano locality Sample IDType L M S L/S L/Smod Costa Rica Turrialba CR-3 FC 78.1 10.5 11.4 6.9 2.1 CR2-1 FC 73.4 15.3 11.4 6.5 2.0 CR2-2 FC 74.8 15.9 9.2 8.1 2.6 CR3-2 FC 78.8 9.3 11.9 6.6 2.0 Irazu¤ CR-5 FC 88.0 5.5 6.4 13.6 4.8 Poa¤s CR2-14 FC 65.8 15.1 19.0 3.5 0.8 CR3-17 FC 73.1 9.8 17.1 4.3 1.1 Poco Sol CRT-2 BF 84.1 10.5 5.4 15.6 5.6 Quebrada Naranja CRT-5 BF 83.7 6.4 9.9 8.4 2.8 CRT-6 BF 81.9 9.5 8.6 9.5 3.2 Miravalles CRT-9 BF 79.3 11.5 9.2 8.6 2.9 Rinco¤n de la Vieja CRT-10 BF 97.2 2.3 0.5 197.4 78.4 CR2-9 BF 95.2 3.9 0.8 116.9 46.2 Averagec 82.3 9.1 8.6 9.6 3.3 Nicaragua Mombacho N-35 FC 81.4 8.4 10.2 8.0 2.6 N-20 MF 84.6 8.2 7.2 11.7 4.1 Xiloa N-12 BF 83.7 10.1 6.1 13.6 4.9 N-13 BF 83.3 10.6 6.0 13.8 4.9 Momotombo N-11 FC 87.1 5.5 7.5 11.6 4.1 N-16 GF 88.1 4.1 7.8 11.3 3.9 N-34 GF 91.9 4.8 3.3 27.9 10.6 N-17 GF 92.8 4.5 2.7 34.5 13.2 Cerro Negro N-4 FC 88.6 4.8 6.6 13.4 4.7 Telica N-8 BF 89.4 0.3 10.3 8.7 2.9 N-9 BF 88.8 0.3 10.9 8.1 2.6 San Cristobal N-2 FC 87.2 4.0 8.8 9.9 3.4 N-5 FC 85.4 5.9 8.7 9.8 3.3 Averagec 86.4 5.6 8.1 11.1 3.8 Worldwide averaged 74.6 12.8 12.5 6.0 a L, M and S proportions (in %) are calculated using the same endmembers as in [11]. b 13 L/Smod is calculated using a sedimentary organic endmember of N C=312x (as opposed to 330x). c Averages for Nicaragua and Costa Rica are calculated using an average value for each locality (in cases where there is more than one sample at a given locality). Four samples are excluded from average calculations: CRT-10 and CR2-9 (possible out- liers), as well as N-34 and N-17 (crustal contamination). d 3 4 Calculated from [11] considering only samples with He/ He ratios = v 5.4 RA.

EPSL 6782 2-9-03 508 A.M. Shaw et al. / Earth and Planetary Science Letters 214 (2003) 499^513

Rica (containing a higher proportion of sedimentary organic C) is being underplated before reaching the zone of magma generation, as invoked by Valentine et al. [15], we would expect a lower S contribution (i.e. a higher L/S) in the volcanic output in Costa Rica. However, we note that the L/S ratios are virtually indistinguishable irrespective of whether 330x or 312x is selected as 13 the N CS endmember (Table 4, L/S and L/Smod). 13 In the subsequent sections we select a N CS endmember value of S = 330x in order to be consistent with prior studies, allowing direct compar- Fig. 5. Variations in (L+S)/M as a function of latitude. The isons to be made. average value for Nicaragua (16) is higher than Costa Rica’s The ¢rst-order implication of uniform L/S ra- average (10) and we note a general increase in relative slab tios would be that there is no signi¢cant di¡erence contributions to the north. Average values (dashed lines) are between the composition of sediment-derived C shown relative to the worldwide average for arcs (shaded that makes it into the magma generation zone in area) = 6.0 þ 2.4 (calculated from [11], considering only samples with 3He/4He ratios v 5.4 R ). Costa Rica versus Nicaragua. One possible expla- A nation is that the uniformly high values re£ect loss of organic C by thermal decomposition in the forearc (see [35]). This would be consistent and we note that the average ratio in Nicaragua is with the occurrence of methane-rich seeps, o¡- 16.0 versus a mean of 10.0 in Costa Rica. Such an shore mud volcanoes and methane hydrate zones enhanced slab contribution to the volcanic output in the forearc region of the Central American arc, in Nicaragua agrees with prior ¢ndings (i.e. 10Be as well as at arcs worldwide [36]. An argument [7], B/La [8], Ba/La [4,5], etc.). There are two pos- against this possibility is that subduction-zone sible explanations for this observation: (1) o¡- metamorphic rock studies have shown that signif- scraping of sediments in Costa Rica [15] and (2) icant loss of organic C at temperatures below a warmer thermal regime in Costa Rica possibly V500‡C does not occur [32]. These studies even associated with subduction of the Cocos ridge (a suggest that s 75% of subducted organic C is hotspot-related oceanic plateau) and/or the trace retained to great depths. An alternative explana- of the Galapagos hotspot [20]. tion for the constant L/S ratios may lie in the fact The idea of o¡scraping of sediments was put that sedimentary carbon is found not only in the forth by Valentine et al. [15] to explain low 10Be shallow hemipelagic muds but also in the lower contents of lavas in Costa Rican arc volcanics. If carbonate sequences. ODP drilling has revealed the water-rich upper hemipelagic portion of the that underlying carbonates bear carbonaceous or- sediment column was lost, then only the older ganic matter at a level up to V1% (as opposed to (10Be-poor due to decay) and more refractory V3% for the muds) [33,34]). Whereas o¡scraping sediments would be available to contribute to and underplating of the hemipelagic sediments o¡ the volcanic output. In this way, loss of the hemi- Costa Rica may account for the low 10Be signal pelagic sediments in Costa Rica could lead to low- (see below), a striking di¡erence in L/S ratios er (L+S)/M ratios. Isotopically light N isotopes in along the arc may not necessarily be anticipated geothermal £uids from Costa Rica [6] con¢rm since both sedimentary units contain organic C. minimal sedimentary-derived N contributions to Although there are no discernible along-strike the volcanic output. variations in the L/S ratio, there is a marked dif- Alternatively, the enhanced slab contribution ference in the relative slab contribution (L+S) vs. observed in Nicaragua may be related to the ther- mantle contributions between Costa Rica and mal structure of the subduction zones in Costa Nicaragua. In Fig. 5 we plot (L+S)/M vs. latitude Rica and Nicaragua. Crucial parameters control-

EPSL 6782 2-9-03 A.M. Shaw et al. / Earth and Planetary Science Letters 214 (2003) 499^513 509 ling the release and transfer of sediments to the set) and a global 3He arc £ux of 92.4 mol/yr [19] mantle wedge are the thermal conditions at subarc scaled to the length of the Central American arc depths and the availability of melt inducing £uids (using a trench length of 1450 km [38], relative to [37]. The contrasting plate geometries of the sub- the global trench length of 43 400 km [39]), we 10 duction zones in Costa Rica and Nicaragua could derive a CO2 £ux of 7.1U10 mol/yr for the have important consequences for their thermal Central American arc. structures. Steep subduction in Nicaragua implies An alternative method to derive CO2 £uxes for a much cooler slab at subarc depths, which would a particular volcano is to use COSPEC measure- not dehydrate until greater depths. The warmer, ments of SO2 £uxes in conjunction with measured shallower subduction along the arc in Costa Rica CO2/SO2 values. Since COSPEC measurements could promote signi¢cant £uid loss from the slab can only be obtained for large £ux volcanoes (along with £uid mobile elements) prior to reach- (those with a plume), it is necessary to account ing the magma generation zone (i.e. in the forearc for the smaller £ux volcanoes to estimate a total region). Such a model was invoked to explain CO2 £ux for the Central American arc. One ap- variations in B (and 10Be) concentrations along proach assumes that the distribution of volcanic the strike of the arc [20]. The lower slab temper- emissions for any arc segment follows a power atures associated with subduction in Nicaragua law [40] and using this relationship a total £ux would result in deeper initiation of dehydration can be calculated. Hilton et al. [19] used this reactions, enhancing the transfer of £uids to sub- method to estimate a total CO2 £ux of arc depths which would trigger melting. The oc- 5.8U1010 mol/yr for the Central American arc. currence of low La/Yb values (generally indicative Despite potential caveats such as the large varia- of high degrees of partial melting) in Nicaraguan tion in measured CO2/SO2 ratios for a given arc arc volcanics is consistent with this model [4].We and the error associated with individual COSPEC suggest that higher degrees of melting in the Nic- measurements (estimated at 10^40% [41]), this araguan subarc (relative to Costa Rica) could re- value agrees remarkably well with the estimate 3 10 £ect enhanced sediment transfer to the volcanic derived using CO2/ He relationships (7.1U10 output in Nicaragua, resulting in higher (L+S)/ mol CO2/yr). M values. If we consider that global subaerial CO2 £uxes are estimated at 2.5U1012 mol/yr (see [42] and 5.3. Absolute £ux calculations references therein), then the Central American £ux represents 2.3^2.8% of the entire CO2 output We use two di¡erent techniques to calculate the from subaerial volcanoes (both along arcs and at £ux of CO2 from volcanoes along the strike of the hotspots). However, most of the global CO2 £ux Central America arc. In the ¢rst case, we use mea- is attributed to only two volcanoes: Mt. Etna 3 3 sured CO2/ He ratios, along with an assumed He (V23%) and Popocatepetl (13.6%) [42]. The glob- 12 £ux (scaled to length of arc) to derive a CO2 £ux al CO2 £ux along arcs is estimated at 1.6U10 estimate. Secondly, we use correlation spectrom- mol/yr [19]. Using this value, we calculate a global eter (COSPEC) measurements of SO2 £uxes at average (including Central America) output £ux 7 individual volcanoes, coupled with measurements of 3.7U10 mol CO2/yr per km of arc (or trench), of CO2/SO2 ratios to calculate a CO2 £ux for the as compared to the average for Central America entire Central American arc. (4.0^4.7U107 mol/yr/km). Despite the potentially 3 Average CO2/ He ratios for Nicaragua and large errors associated with these calculations, we Costa Rica are 2.5 þ 0.4U1010 (n = 9) and note that the Central American output £ux is sur- 2.1 þ 0.4U1010 (n = 13), respectively (from Tables prisingly low (only slightly higher than the aver- 1 and 2; excluding samples which show crustal age) given the high in£ux of CO2-bearing sedi- contamination ^ see Section 5.1). If we assume ments on the down-going plate relative to other 3 10 an average CO2/ He ratio of 2.3U10 for the arcs. For example, the Central American margin entire Central American arc (based on this data has an annual sedimentary input £ux of 176 mol

EPSL 6782 2-9-03 510 A.M. Shaw et al. / Earth and Planetary Science Letters 214 (2003) 499^513

CO2/km arc, more than ¢ve times the average of since it is based on core measurements (ODP site 31 mol CO2/km arc subducted at arcs worldwide 1039) in this region and is consistent with the (calculations based on sedimentary CO2 £uxes in notion that carbonate sediments have poorer [19]). In Section 5.4, we consider the total C mass preservation of organic matter than hemipelagic balance along the Central American margin to muds [43]. evaluate the e⁄ciency of C recycling through the Oceanic crustal basement is also a potentially subduction zone factory. signi¢cant source of C since carbonate veining and/or calcite precipitation associated with hydro- 5.4. Mass balance at subduction zones thermal alteration a¡ects sea£oor crust progres- sively with age [44]. We have calculated the oce- In order to assess how much C is cycled anic crustal C contribution using two methods: through the Central American subduction zone (1) an average altered crust value of 0.214 wt% versus the amount which is ultimately transferred CO2 for 7 km thick crust [44] and (2) a value of to the deep mantle, we must quantitatively deter- 2.95 wt% CO2 for the upper 500 m, where most of mine the C inputs from the various sources: (1) the alteration is concentrated [45]. In both cases subducting carbonate sediments, (2) organic mat- we assume a crustal density of 2.89 g/cm3. In the 5 ter and (3) altered oceanic crust. A critical as- ¢rst case we calculate a CO2 input of 1.10U10 sumption in any mass balance calculation is that Mmol/yr, which is essentially indistinguishable the present-day input at the trench is representa- from a value of 1.08U105 Mmol/yr calculated in tive of the material that is producing the arc mag- the second case. Since the age of the subducting mas today. Sediments on the subducting Cocos crust is 23 þ 5 Myr and precipitation of carbon- plate underwent a major transition from carbonates within the crust may persist for up to 100 ate-rich sediments to hemipelagic muds due to the Myr [44], these estimates represent maximum val- so-called carbonate crash 10 Ma associated with a ues. major change in ocean circulation patterns (see Based on the above calculations, the relative [43]). Since then, however, sediment compositions proportion of limestone and marine carbonate C have remained relatively homogeneous and thus to sedimentary organic C being input at the arc the assumption that present-day sediment compo- from both sediments and oceanic crust is 10:1. sitions at the trench are representative is reason- This value is remarkably similar to the L/S values able. observed in the volcanic output (Costa Rica aver- The £ux of sedimentary material into the Cen- age L/S = 9.6 and Nicaragua average L/S = 11.1). tral American subduction zone can be estimated The implication of this ¢nding is that organic car- from the subduction rate (77 mm/yr), sediment bon is not signi¢cantly fractionated from carbon- thickness (425 m), trench length (1450 km) and ate during the subduction and magma generation the bulk density (1.62 g/cm3) [38]. Taking bulk processes. This is in contrast to results reported in sediment water content (48.69%) and the dry a study on submarine basalt glasses in the North weight CO2 concentration (26.55 wt%) into ac- Fiji back-arc basin [12], where considerable frac- count, we calculate a carbonate-derived CO2 in- tionation of the L/S ratio was calculated. The put of 2.38U105 Mmol/yr (as in [19]). proportion of L to S measured in the volatile out- The organic component can be estimated in put in the back-arc (L:S = 7:3) was signi¢cantly two ways: (1) assuming an average of 1 wt% or- di¡erent from proportions being input at the 4 ganic CO2 [32] to yield 1.75U10 Mmol CO2/yr trench (L:S = 20:1). The most reasonable explana- (as in [19]) or (2) using a calculated average of tion for the discrepancy between our results and 0.98 wt% C for the upper hemipelagic portion, those of Nishio et al. is that they did not consider and 0.26 wt% C for the lower carbonate sequence, C losses from the arc front, which are likely to be based on drill report measurements [33], resulting signi¢cant (based on this work and observations 4 3 in an organic CO2 input £ux of 3.56U10 Mmol/ of high CO2/ He ratios in arcs worldwide [28]). In yr. The latter value is deemed more representative addition, a back-arc setting would have a di¡erent

EPSL 6782 2-9-03 A.M. Shaw et al. / Earth and Planetary Science Letters 214 (2003) 499^513 511 thermal regime and availability of £uids, poten- slab C (relative to other arcs), resulting in the low tially resulting in less e⁄cient carbonate devolati- e⁄ciency (14^18%) of C cycling. lisation (see discussion below). Considering the CO2 contributions from the slab components (crust+sediments), the total 6. Conclusions CO2 input along the Central American margin is of the order of 3.8U1011 mol/yr, as compared to a 1. Most geothermal £uids from Costa Rica and volcanic output of 5.8^7.1U1010 mol/yr. If we Nicaragua have MORB-like 3He/4He ratios, subtract the mantle C contribution to the volcanic N13C values ranging from 36.8 to 30.1x, 3 output (9.1% in Costa Rica and 5.6% in Nicara- and elevated CO2/ He values relative to gua), we can determine the percentage of material MORB (2.1U1010 and 2.5U1010, for Costa which is recycled through the arc (i.e. [(L+S)out/ Rica and Nicaragua respectively), consistent (L+S)in]U100). Based on these calculations, we with prior studies of arc-related domains (e.g. ¢nd that 14^18% of the material input at the [27]). arc is lost via arc volcanism. Consequently, as 2. Water springs and old geothermal wells show 13 much as 86% of the CO2 input along the arc evidence for crustal additions (variable N C, 3 4 3 could be transferred to the deeper mantle. Note, low He/ He and high CO2/ He ratios) and however, that it is likely that some fraction is lost are considered unreliable tracers of mantle pro- in the forearc. The percentage of CO2 emitted via cesses. the Central American arc is exceptionally low rel- 3. Using the approach of Sano and Marty [11], ative to global estimates of the output/input, we ¢nd that the geothermal £uid volcanic out- V0.50 (calculated from global £ux estimates in put is dominated by a limestone/marine car- [46] and [19]). Kerrick and Connelly [46] attrib- bonate component (82^86%), with relatively uted low CO2 arc emissions along arcs to two low sedimentary organic and mantle C contri- processes: (1) strong devolatilisation of clay-rich butions. marls in the forearc region, provided the geo- 4. The L/S ratios observed in the volcanic output therm is su⁄ciently high, and (2) retention of in Costa Rica (9.6) and Nicaragua (11.1) are CO2 in carbonate-bearing marine sediments to high relative to arcs worldwide (6.0). This depths greater than subarc depths. It is possible likely re£ects the carbonate-rich nature of the that both of these processes occur in the Central sediments on the downgoing Cocos plate. The American subduction zone to limit CO2 loss via L/S input ratio is 10, implying that signi¢cant arc volcanism. In addition, given the carbonate- fractionation of organic matter C from carbon- rich nature of the lower sediments, it is possible ate C does not occur during subduction and that there are insu⁄cient £uids available at subarc magma generation. depths to promote e⁄cient decarbonation of the 5. The (L+S)/M ratios in Costa Rica ( = 10) and carbonate sediments [47].H2O-rich £uid availabil- Nicaragua ( = 16) support the suggestion of an ity is potentially the limiting factor controlling enhanced slab £ux to the volcanic output in whether marine carbonates will devolatilise at Nicaragua, consistent with prior ¢ndings subarc depths or be transferred to the deep man- [1,4,5,7,8,13]. tle (e.g. wet solidus versus dry solidus). As dis- 6. Similar £uxes of CO2 for the entire Central cussed, the warmer subduction in Costa Rica American arc are derived using COSPEC would likely initiate £uid loss in the forearc, re- methods (5.8U1010 mol/yr) and measured 3 3 sulting in limited availability of H2O-rich £uids CO2/ He ratios combined with global He necessary to trigger devolatilisation at subarc £uxes scaled to arc length (7.1U1010 mol/yr). depths. Likewise, in Nicaragua, despite a greater These values, however, are low (14^18%) rela- transfer of slab C to the arc, it is possible that tive to the amount of CO2 being input at the insu⁄cient melt-inducing £uids are available to trench. This imbalance can be attributed to accommodate the overwhelmingly high in£ux of three processes: (i) C losses in the forearc re-

EPSL 6782 2-9-03 512 A.M. Shaw et al. / Earth and Planetary Science Letters 214 (2003) 499^513

gion, (ii) incomplete decarbonation of lime- element and isotopic evidence for tectonic control of stone/marine carbonate and (iii) limited avail- source mixing and melt extraction along the Central American arc, Contrib. Mineral. Petrol. 105 (1990) 369^ ability of melt-inducing £uids to accommodate 380. the in£ux of slab C. [5] L.C. Patino, M.J. Carr, M.D. Feigenson, Local and re- 7. Up to 86% of CO2 input at the trench is po- gional variations in Central American arc lavas controlled tentially subducted back into the mantle ^ by variations in subducted sediment input, Contrib. Min- however, the forearc may be a signi¢cant res- eral. Petrol. 138 (2000) 265^283. [6] T.P. Fischer, D.R. Hilton, M.M. Zimmer, A.M. Shaw, ervoir to consider; constraining this £ux Z.D. Sharp, J.A. Walker, Subduction and recycling of should be the focus of future studies. nitrogen along the Central American margin, Science 297 (2002) 1154^1157. [7] J.D. Morris, W.P. Leeman, F. Tera, The subducted com- Acknowledgements ponent in island arc lavas: constraints from Be isotopes and B-Be systematics, Nature 344 (1990) 31^36. [8] W.P. Leeman, M.J. Carr, J.D. Morris, Boron geochemis- M. Carr and L. Patino are thanked for con- try of the Central American Volcanic Arc ^ constraints on structive comments. We acknowledge M. Wahlen the genesis of subduction-related magmas, Geochim. Cos- and B. Deck (S.I.O.) for assistance with stable mochim. Acta 58 (1994) 149^168. 10 isotope measurements. Logistical support was [9] J. Morris, R. Valentine, T. Harrison, Be imaging of sediment accretion and subduction along the northeast provided by the Instituto Costarricense de Elec- Japan and Costa Rica convergent margins, Geology 30 tricidad (Costa Rica) and W. Strauch at INETER (2002) 59^62. (Nicaragua). We thank C. Ram|¤rez, F. Ar|¤as, W. [10] T. Plank, C.H. Langmuir, Tracing trace elements from Suiter, J. Hlebica, M. Zimmer, P. Pe¤rez, S. sediment input to volcanic output at subduction zones, McKnight, L. Elkins and J. Albrecht for assis- Nature 362 (1993) 739^743. [11] Y. Sano, B. Marty, Origin of carbon in fumarolic gas tance in the ¢eld. In addition, we are grateful to from island arcs, Chem. Geol. 119 (1995) 265^274. personnel at Ormat and Miravalles power plants [12] Y. Nishio, S. Sasaki, T. Gamo, H. Hiyagon, Y. Sano, for access to geothermal wells. This project was Carbon and helium isotope systematics of North Fiji Ba- funded by the following NSF grants: EAR- sin basalt glasses: carbon geochemical cycle in the sub- 0003628 (D.R.H.), EAR-003668 (T.P.F.), EAR- duction zone, Earth Planet. Sci. Lett. 154 (1998) 127^ 138. 0003664 (J.A.W.) and EAR-0079402 MAR- [13] G. Snyder, U. Fehn, I-129 in volcanic £uids: Testing for GINS.[SK] the presence of marine sediments in the Central American volcanic arc, Nucl. Instrum. Methods Phys. Res. B 172 (2000) 568^573. [14] G. Snyder, R. Poreda, A. Hunt, U. Fehn, Regional var- References iations in volatile composition: Isotopic evidence for carbonate recycling in the Central American volcanic arc, [1] M.J. Carr, Symmetrical and segmented variation of phys- Geochem. Geophys. Geosys. 2 (2001) U1^U32. ical and geochemical characteristics of the Central Amer- [15] R.B. Valentine, J.D. Morris, D. Duncan Jr., O.S.P.L. 170, ican volcanic front, J. Volcanol. Geotherm. Res. 20 (1984) Sediment subduction, accretion, underplating and arc vol- 231^252. canism along the margin of Costa Rica: Constraints from [2] M. Protti, F. Guendel, K. McNally, Correlations between Ba, Zn, Ni, and 10Be concentrations, EOS Trans. AGU 78 the age of the subducted Cocos plate and the geometry of (1997) 673. the Wadati-Benio¡ zone under Nicaragua and Costa [16] R. von Huene, D.W. Scholl, Observations at convergent Rica, in: P. Mann (Ed.), Geologic and Tectonic Develop- margins concerning sediment subduction, subduction ero- ment of the Caribbean Plate Boundary in Southern Cen- sion and the growth of continental crust, Rev. Geophys. tral America, Geol. Soc. Am. Spec. Pap. 295 (1995) 380. 29 (1991) 279^316. [3] R.K. Kelly, K.D. McIntosh, E.A. Silver, N.W. Driscoll, [17] C.R. Ranero, R. von Huene, Subduction erosion along J.A. Go¡, C.R. Ranero, R. von Huene, A quantitative the Middle America convergent margin, Nature 404 analysis of £exural faulting in the Cocos Plate at the Mid- (2000) 748^752. dle America Trench from Nicaragua to Costa Rica, EOS [18] P. Vannucchi, D.W. Scholl, M. Meschede, K. McDougall- Trans. AGU 82 (Fall Meet. Suppl.) (2001) Abstract Reid, Tectonic erosion and consequent collapse of the F1148. Paci¢c margin of Costa Rica: Combined implications [4] M.J. Carr, M.D. Feigenson, E.A. Bennett, Incompatible from ODP Leg 170, seismic o¡shore data, and regional

EPSL 6782 2-9-03 A.M. Shaw et al. / Earth and Planetary Science Letters 214 (2003) 499^513 513

geology of the Nicoya Peninsula, Tectonics 20 (2001) 649^ [36] G. Bohrmann, K. Heeschen, C. Jung, W. Weinrebe, B. 668. Baranov, B. Cailleau, R. Heath, V. Huhnerbach, M. [19] D.R. Hilton, T.P. Fischer, B. Marty, Noble gases and Hort, D. Masson, I. Trummer, Widespread £uid expul- volatile recycling at subduction zones, in: D. Porcelli, sion along the sea£oor of the Costa Rica convergent mar- C.J. Ballentine, R. Wieler (Eds.), Noble Gases in Cosmo- gin, Terra Nova 14 (2002) 69^79. chemistry and Geochemistry, Mineral. Soc. Am. 47 (2002) [37] S.M. Peacock, Fluid processes in subduction zones, Sci- 319^370. ence 248 (1990) 329^337. [20] W.P. Leeman, M.J. Carr, Geochemical constraints on [38] T. Plank, C.H. Langmuir, The chemical composition of subduction processes in the Central American Volcanic subducting sediment and its consequences for the crust Arc: Implications of boron geochemistry, in: P. Mann and mantle, Chem. Geol. 145 (1998) 325^394. (Ed.), Geologic and Tectonic Development of the Carib- [39] D.K. Rea, L.J. Ru¡, Composition and mass £ux of sedi- bean Plate Boundary in Southern Central America, Geol. ment entering the worlds subduction zones ^ implications Soc. Am. Spec. Pap. 295 (1995) 57^73. for global sediment budgets, great earthquakes, and vol- [21] M.J. Carr, M.D. Feigenson, L.C. Patino, J.A. Walker, canism, Earth Planet. Sci. Lett. 140 (1996) 1^12. Volcanism and geochemistry in Central America: Prog- [40] S.L. Brantley, K.W. Koepenick, Measured carbon dioxide ress and problems, AGU Spec. Publ. (in press). emissions from Oldoinyo Lengai and the skewed distribu- [22] M.C. van Soest, D.R. Hilton, R. Kreulen, Tracing crustal tion of passive volcanic £uxes, Geology 23 (1995) 933^ and slab contributions to arc magmatism in the Lesser 936. Antilles island arc using helium and carbon relationships [41] R.E. Stoiber, L.L.J. Malinconico, S.N. Williams, Use of in geothermal £uids, Geochim. Cosmochim. Acta 62 the correlation spectrometer at volcanoes, in: H. Tazie¡, (1998) 3323^3335. J. Sabroux (Eds.), Forecasting Volcanic Events, vol. 1, [23] B. Marty, A. Jambon, Y. Sano, Helium isotopes and CO2 Elsevier, New York, 1983, pp. 425^444. in volcanic gases of Japan, Chem. Geol. 76 (1989) 25^40. [42] D.M. Kerrick, Present and past nonanthropogenic CO2 [24] D.R. Hilton, J.A. Hoogewer¡, M.J. van Bergen, K. Ham- degassing from the solid Earth, Rev. Geophys. 39 (2001) merschmidt, Mapping magma sources in the east Sunda- 565^585. Banda arcs, Indonesia: constraints from helium isotopes, [43] T. Plank, V. Balzer, M. Carr, Nicaraguan volcanoes rec- Geochim. Cosmochim. Acta 56 (1992) 851^859. ord paleoceanographic changes accompanying closure of [25] B. Marty, A. Jambon, C/3He in volatile £uxes from the the Panama gateway, Geology 30 (2002) 1087^1090. solid Earth: Implication for carbon geodynamics, Earth [44] J.C. Alt, D.A.H. Teagle, The uptake of carbon during Planet. Sci. Lett. 83 (1987) 16^26. alteration of ocean crust, Geochim. Cosmochim. Acta [26] R.K. O’Nions, E.R. Oxburgh, Helium, volatile £uxes and 63 (1999) 1527^1535. the development of continental crust, Earth Planet. Sci. [45] H. Staudigel, T. Plank, B. White, H.-U. Schmincke, Geo- Lett. 90 (1988) 331^347. chemical £uxes during sea£oor alteration of the upper [27] B. Marty, I.N. Tolstikhin, CO2 £uxes from mid ocean oceanic crust: DSDP sites 417^418, in: G.E. Bebout, ridges, arcs and plumes, Chem. Geol. 145 (1998) 233^248. D.W. Scholl, S.H. Kirby, J.P. Platt (Eds.), Subduction [28] Y. Sano, S.N. Williams, Fluxes of mantle and subducted Top to Bottom, American Geophysical Union, Washing- carbon along convergent plate boundaries, Geophys. Res. ton, DC, 1996, pp. 19^38. Lett. 23 (1996) 2749^2752. [46] D.M. Kerrick, J.A.D. Connolly, Metamorphic devolatili- [29] M. Ozima, F.A. Podosek, Noble Gas Geochemistry, zation of subducted marine sediments and the transport Cambridge University Press, Cambridge, 2002, 286 pp. of volatiles into the Earth’s mantle, Nature 411 (2001) [30] J. Hoefs, M. Frey, The isotopic composition of carbona- 293^296. ceous matter in a metamorphic pro¢le from the Swiss [47] D.M. Kerrick, J.A.D. Connolly, Metamorphic devolatili- Alps, Geochim. Cosmochim. Acta 40 (1976) 945^951. zation of subducted oceanic metabasalts: implications for [31] B. Cramer, B.M. Krooss, R. Littke, Modelling isotope seismicity, arc magmatism and volatile recycling, Earth fractionation during primary cracking of natural gas: a Planet. Sci. Lett. 189 (2001) 19^29. reaction kinetic approach, Chem. Geol. 149 (1998) 235^ [48] R.F. Weiss, Solubility of helium and neon in water and 250. seawater, J. Chem. Eng. Data 16 (1971) 235^241. [32] G.E. Bebout, The impact of subduction-zone metamor- [49] C.D. Keeling, T.P. Whorf, Atmospheric CO2 records phism on mantle-ocean chemical cycling, Chem. Geol. from sites in the SIO air sampling network, in: Trends: 126 (1995) 191^218. A Compendium of Data on Global Change, Carbon Di- [33] G. Kiruma, Shipboard Scienti¢c Party, Proc. ODP Init. oxide Information Analysis Center, Oak Ridge National Rep. 170 (1997) 45. Laboratory, U.S. Department of Energy, Oak Ridge, TN, [34] J. Auboin, Shipboard Scienti¢c Party, Init. Rep. DSDP 67 2002. (1982) 79. [50] E. Wilhelm, R. Battino, J. Wilcock, Low pressure solubil- [35] J.C. Moore, P. Vrolijk, Fluids in accretionary prisms, ity of gases in liquid water, Chem. Rev. 77 (1977) 219^ Rev. Geophys. 30 (1992) 113^135. 262.

EPSL 6782 2-9-03