Contents lists available at ScienceDirect

Chemical Geology

journal homepage: www.elsevier.com/locate/chemgeo

Two distinct mantle sources beneath the Garibaldi Volcanic Belt: Insightfrom olivine-hosted melt inclusions

Swetha Venugopala,b,*, Séverine Mouneb,c,d, Glyn Williams-Jonesa, Timothy Druittb,Nathalie Vigourouxa,e, Alexander Wilsonf, James K. Russellf

a Centre for Natural Hazards Research, Department of Earth Sciences, Simon Fraser University, BC, V5A 1S6, Canadab Laboratoire Magmas et Volcans, Université Clermont Auvergne, Campus Universitaire des Cézeaux, 63178, Aubiere Cedex, FrancecObservatoire Volcanologique et Sismologique de Guadeloupe (OVSG), Le Houëlmont, 97113, Gourbeyre, GuadeloupedUniversité de Paris, Institut de Physique du Globe de Paris, CNRS, F-75005, Paris, Francee Department of Earth and Environmental Sciences, Douglas College, BC, V3M 5Z5, CanadafDepartment of Earth, Ocean and Atmospheric Sciences, The University of British Columbia, BC, V6T 1Z4, Canada

A R T I C L E I N F O

Editor: Donald Dingwell

Keywords:Melt inclusionsGaribaldi Volcanic BeltCascadesTrace elementsTwo mantle sources

A B S T R A C T

The nature of the magmatic source beneath the Garibaldi Volcanic Belt (GVB) in NW Washington (USA) and SWBritish Columbia (Canada) has been debated both due to its classification as the northern equivalent of the HighCascades and the alkaline nature of northern basalts. Whole rock studies have shown that the GVB does not sharethe same magmatic source as the High Cascades (Mullen and Weis, 2013, 2015). Nonetheless, the presence ofalkaline basalts in this arc raises questions about the exact source of mantle enrichment and whether it is relatedto the young age of the downgoing Juan de Fuca Plate (< 10 Ma) or the presence of a slab tear at the northernend of the arc. To gain insight into the source that underlies the GVB, we sampled olivine-hosted melt inclusionsfrom each volcanic centre along the arc. Major, volatile and trace element data reveal a northward compositionaltrend from arc-typical calc-alkaline magma in the south to OIB-like melts in the north near the slab tear. Fur-thermore, contributions from the subducting slab are relatively high beneath the southern end of the arc (Cl/Nb>80) but rapidly decreases to the north (Cl/Nb<50). Finally, the significant differences in Zr/Nb fromsouth to north (80 and 9, respectively) suggest two distinct mantle sources since one source cannot producemelts with such different ratios. As such, we propose the GVB should be segmented into the Northern andSouthern groups, each having its own mantle source. Based on the geographic proximity, the enriched nature ofthe Northern group melt inclusions is likely controlled by the slab tear at the northern termination of thesubducting Juan de Fuca Plate. Melt modelling results show that 3–7 % partial melting of the primitive mantlewith a garnet lherzolite residue can reproduce the composition of the Northern group. Melt inclusions from theSouthern group, on the other hand, imply a depleted MORB mantle that has been modified by fluids derivedfrom the downgoing slab. Variability within the Southern group itself reflects the amount of hydrous fluidssupplied beneath each centre and is correlated with slab age and subsequent degree of dehydration. This studyaddresses the compositional diversity along the arc and provides evidence that the age of the downgoing plateand the presence of a slab tear exert a strong compositional control over eruptive products within one arc.

1. Introduction

The main compositional controls of eruptive products in subductionzone settings are arguably the age and tectonic context of the down-going slab (Cooper et al., 2012; Saginor et al., 2013). Dictated by theage, factors such as slab temperature, subduction angle and the amountof slab-derived fluids can cause significant variations in the composi-tion of the wedge-generated melt (Saginor et al., 2013; Weller and

Stern, 2018). The presence of faults and fractures can facilitate the riseof more primitive melt by acting as favourable pathways (e.g., Connorand Conway, 2000; Gaffney et al., 2007; Venugopal et al., 2016)whereas slab tears and slab windows can contribute melts derived fromthe deeper mantle that do not carry the arc signature (e.g., Gorring andKay, 2001).

Modelling of slab thermal structure has shown that older slabs, suchas the Pacific Plate beneath the Aleutian Islands, generally have steeper

https://doi.org/10.1016/j.chemgeo.2019.119346Received 5 February 2019; Received in revised form 11 June 2019; Accepted 19 October 2019

⁎ Corresponding author at: Centre for Natural Hazards Research, Department of Earth Sciences, Simon Fraser University, BC, V5A 1S6, Canada.E-mail address: swethav@sfu.ca (S. Venugopal).

Chemical Geology 532 (2020) 119346

Available online 21 October 20190009-2541/ © 2019 Elsevier B.V. All rights reserved.

T

subduction angles, have lower slab temperatures at a given depth, andsupply a higher flux of fluids to the sub-arc mantle (Syracuse et al.,2010; Ohtani et al., 2004; Stern, 2002; Ulmer, 2001). Younger slabs, onthe other hand, achieve higher temperatures at a given depth and lowerslab dips (Syracuse et al., 2010). Though poorly investigated, the re-lease of fluids (due to dehydration reactions) from young slabs has beenshown to be at shallower depths than in older slab settings (Ruscittoet al., 2010; Walowski et al., 2016). This suggests that ascending pri-mary derived from young, dehydrated slabs should melts should bedepleted in volatiles and slab-derived elements.

An excellent case study for studying the effect of slab age and tec-tonic framework on the resulting eruptive compositions is the Juan deFuca Plate, which is considered to be the young and hot end member inglobal subduction zone systems (Syracuse et al., 2010). The north-easterly subduction of the Juan de Fuca Plate beneath the NorthAmerican Plate produces the Cascades Volcanic Arc along the westernmargin of North America (Fig. 1; Moore and DeBari, 2012). This arc issubdivided into the High Cascades in the south and the Garibaldi Vol-canic Belt (GVB) to the north (Fig. 1). The age of the Juan de Fuca Plateat the trench decreases northward from 10 Ma beneath the northernend of the High Cascades to approximately 5 Ma beneath the GVBwhere it terminates against the Explorer Plate (Fig. 1), which has a

lower dip angle (Candie et al., 1989; Wilson, 2002). This difference ofdip produces a gap or slab tear between the two plates (Thorkelsonet al., 2011).

The relatively older Juan de Fuca Plate beneath the southernCascades is capable of supplying more subduction-derived fluids intothe mantle wedge, thereby initiating partial melting and producing arc-typical calc-alkaline magma (Green and Sinha, 2005). In terms of theGVB, however, little is known of primitive melt compositions and vo-latile contents. What is known is the coeval presence of calc-alkalinerocks in the southern GVB and alkaline rocks in the northern GVB(Green, 2006; Mullen and Weis, 2013, 2015). The GVB therefore pro-vides an opportunity to examine both the effects of slab age and slabtearing on the magmatic source composition and volatile contentswithin a single subduction zone system.

Here we present olivine-hosted melt inclusion data for the volcaniccentres within the Canadian segment of the GVB alongside melt inclu-sion data from Glacier Peak and Mount Baker compiled by Shaw(2011). Major element, volatile and trace element contents of the meltinclusions are used to decipher the dynamics of the GVB sub-arc mantleand to better understand the effect of slab age and slab tear on thecompositions of mantle melts.

Fig. 1. Regional map of the Cascade Volcanic Arc along western North America. This arc is separated into the High Cascades from northern California to northernWashington and the Garibaldi Volcanic Belt (GVB) from northwest Washington to southwest Canada. Beneath the GVB, the subducting Juan de Fuca Plate youngs inage from 10 Ma below Glacier Peak to 5 Ma beneath Mount Meager where it terminates against the Nootka Fault beneath Salal Glacier and Bridge River.

S. Venugopal, et al. Chemical Geology 532 (2020) 119346

2

2. Geological context of the GVB

The subdivision of the Cascades Arc is based on the change in strikeof the arc axis and a bend in the alignment of volcanic centres; thechange in strike is evident between Mount Rainier and Glacier Peak(Fig. 1; Mullen et al., 2017). The High Cascades, which strikes north-south, stretches between northern California and northern Washington.The GVB, on the other hand, is a Quaternary-aged glaciovolcanic arcextending between northern Washington and southwestern British Co-lumbia. This belt strikes northwest-southeast and terminates in thenorth against the Nootka Fault (Fig. 1). Volcanism within the GVB hasresulted in more than 2300 distinct vents and 22 major volcanicstructures, including Glacier Peak, Mount Baker and the volcanic fieldsof Mount Garibaldi, Garibaldi Lake, Mount Cayley, Mount MeagerVolcanic Complex, Salal Glacier, Bridge River and Silverthrone (Wilsonand Russell, 2018; Hildreth, 2007). Recently, the study by Mullen et al.(2017) classifies Glacier Peak as the northern termination of the HighCascades segment. The dominant eruptive products of the GVB are in-termediate calc-alkaline suites with basaltic eruptions comprising asmall volume fraction in the south and a large volume fraction in thenorth (Bridge River to Salal Glacier; Wilson and Russell, 2018). Thecurrent geochemical literature consists of high-precision Sr, Pb and Hfisotopic data of basaltic whole-rocks, generally one sample per volcaniccentre (Mullen and Weis, 2013, 2015, 2017). To gain further chemicalinsight into the primary mantle melt beneath the arc, samples of pri-mitive magma preserved as melt inclusions in an early-crystallizingphase such as olivine, are considered excellent proxies. Melt inclusionsare tiny pockets of melt trapped within growing crystals. Upon en-trapment, the melts are considered to be sealed in a closed-system en-vironment. In ideal situations, these inclusions are protected by the hostcrystal during magmatic ascent and eruption, at which time theyquench to glass and preserve their initial composition. However, slowascent rates and/or slow cooling following eruption can cause post-entrapment processes such as nucleation of daughter minerals, whichmodify the initial inclusion composition and should be taken into ac-count (Section 4.2). Nonetheless, analyses of melt inclusions yield va-luable information about the magmatic conditions at the time of en-trapment as well as the unique nature of the source at depth.

3. Sample description and analytical methods

Detailed sample descriptions and methods can be found as supple-mentary material. The reader is referred to Shaw (2011) for descrip-tions of Glacier Peak and Mount Baker samples. Olivine-bearing basaltsand basaltic-andesites were sampled from each centre between Gar-ibaldi Lake and Bridge River. Tephra samples were collected for Gar-ibaldi Lake and Mount Meager while lava flows were sampled fromMount Cayley. Hyaloclastite and pyroclastic breccia were sampled fromSalal Glacier and Bridge River, respectively. Samples were crushed,sieved and hand-picked for olivine crystals, which ranged in size from300 to 600 μm. In all samples, melt inclusions were 15–40 μm in dia-meter, elliptical to spherical in shape and contained a vapour bubblethat occupied less than 10 vol% of the inclusion (Supplementary FigureS1). Bubbles that occupied a larger volumetric proportion were avoidedas they most likely do not represent primary vapour bubbles (Mooreet al., 2015). Melt inclusions that showed evidence of decrepitation(i.e., loss of volatiles through fractures extending from the melt inclu-sion into the olivine host) were also avoided (Maclennan, 2017). Glassy,crystal-free melt inclusions were only present in the Mount Meagersample as well as Glacier Peak and Mount Baker (Shaw., 2011). SomeMount Meager melt inclusions contained immiscible sulphide globulesthat ranged in size between 5–10 μm. Olivines from Garibaldi Lake,Mount Cayley, Salal Glacier and Bridge River hosted melt inclusionscontaining pre-existing spinel and occasional sulphides. Based on theirirregular shape, rounded embayments and diverse volumetric propor-tions relative to their olivine hosts, the sulphides were likely co-

entrapped with the silicate melt. These melt inclusions also containeddaughter crystals, which typically form due to volatile loss, slow ascentrates and/or slow cooling following eruption (Schiano, 2003; Wallace,2005; Supplementary Figure S1). As such, these inclusions are not re-presentative of the original magma. To reverse the daughter crystal-lisation process and retrieve the original melt compositions, inclusionsfrom Bridge River, Salal Glacier, Mount Cayley and Garibaldi Lake werereheated at the Laboratoire Magmas et Volcans (LMV) in Clermont-Ferrand, France. Re-heating experiments for each inclusion were15 min in duration to reduce volatile loss (Chen et al., 2011; Portnyaginet al., 2008; Gaetani et al., 2012; Bucholz et al., 2013) and stoppedwhen the final daughter crystal disappeared (see SupplementaryMethods). The heating stage was not able to accommodate the hightemperatures required for complete melt inclusion homogenisation.Melt inclusions, olivine hosts and matrix glasses were analysed formajor elements (plus S and Cl for all glasses) at LMV using a SX-100CAMECA electron microprobe. The precision of electron microprobeanalyses (2σ) are better than 5 % for major elements, except for MnO,Na2O and K2O, which had an EMP precision<10 %. The approximate2σ precision for S and Cl is 4 % and 7 %, respectively. In-situ traceelement analyses of melt inclusions were also performed at the LMVwith a Resonetics M50 EXCIMER Laser and a 193 nm wavelengthcoupled to an Agilent 7500 cs ICP-MS (LA-ICP-MS). Analytical precisionand accuracy of measurements were stable, and most elements werebetter than 20 % at a 95 % confidence level. Sulphur speciation (S6+/Stotal) was obtained from the S Kα peak shifts relative to the peak shiftof barite and sphalerite, temperature and linear regression coefficients(Wallace and Carmichael, 1994) in order to estimate the oxygen fuga-city of the melt inclusions following the method of Jugo et al. (2005).H2O, CO2 and D/H of select melt inclusions were analysed using theCAMECA IMS 270 Ion Probe (SIMS) at the Centre de Recherches Pét-rographiques et Géochimiques (CRPG) in Nancy, France. The D/Hcontents of melt inclusions provide insight into the degree of H + dif-fusion (i.e., water loss) following entrapment. Values are commonlyreported as δD, which is the permil deviation of the D/H isotope ratiofrom the D/H ratio of Standard Mean Ocean Water (δDSMOW = 0 ‰;Hauri, 2002; Shaw et al., 2008). Errors per mil were calculated usingthe instrumental drift and are between 2–9 ‰ with one value of 19‰for Mount Meager. Finally, based on current literature estimates(Anderson and Brown, 1993; Hartley et al., 2015; Moore et al., 2015;Wallace et al., 2015; Neave et al., 2015; Aster et al., 2016 and refer-ences therein), the quantity of CO2 within the shrinkage bubble isroughly 70–90% of the total amount of CO2 in the melt inclusion. Inorder to avoid the underestimation of CO2 contents, Raman analyseswere done within the bubble for most melt inclusions; analyses forBridge River, Salal Glacier, Mount Cayley and Garibaldi Lake were postre-heating. The size and volume of bubbles and melt inclusions wereestimated using a microscope under transmitted light and Leica imagingsoftware. Melt inclusion volumes were assumed to be ellipsoidal andthe two observable axes were measured. The best estimate for the thirdaxis was approximated using the smaller ellipsoidal axis measured inthe microscope. Bubble volumes were assumed to be spherical andcalculated using measured diameters that were accurate to 2 μm. Theassociated errors for size and volume estimates is less than 10 %.

The analytical procedure for Mount Baker and Glacier Peak can befound in Shaw (2011). The main instrumental difference between Shaw(2011) and this study is their use of FTIR for H2O analyses of meltinclusions. Further details about analytical techniques can be found assupplementary material.

4. Results

4.1. Whole rock and groundmass glass

Whole rock compositions for Bridge River, Salal Glacier, MountMeager, Mount Cayley and Garibaldi Lake are presented in

S. Venugopal, et al. Chemical Geology 532 (2020) 119346

3

Supplementary Table S1. Major oxide compositions for these centresare broad with 47–55 wt% SiO2, 8–14 wt% Fe2O3, 5–11 wt% MgO and4–5 wt% total alkalis (Na2O + K2O; Fig. 2). While most samples arecalc-alkaline arc basalts, the samples from Bridge River and Salal Gla-cier are the most alkaline (Fig. 2). Mount Baker and Glacier Peak wholerock and groundmass glass compositions were not provided by Shaw(2011).

4.2. Melt inclusions

All melt inclusions experienced some degree of post-entrapmentmodification. Glassy melt inclusions from Glacier Peak, Mount Bakerand Mount Meager were corrected for olivine-rim crystallisation con-sidering the Fe-Mg equilibrium KD [= (FeO/MgO)olivine/(FeO/MgO)melt

inclusion]. Corrections were preformed using Petrolog3 software(Danyushevsky and Plechov, 2011) and the equation calibrated byToplis (2005). Overall, melt inclusions experienced minimal post en-trapment crystallization (PEC = 0 - 6.5%) and are in equilibrium withthe olivine host (KD = 0.30 – 0.36; Toplis, 2005; Supplementary TableS2). Melt inclusions from the other centres experienced daughtercrystallisation. Reversing this process through re-heating also reversesthe post-entrapment olivine rim crystallisation, thus yielding melt in-clusions in equilibrium with the olivine host (KD = 0.30 – 0.33; Sup-plementary Table S2). The temperature at which all daughter crystalsmelted within the inclusion range from 1150 to 1190 °C (Supplemen-tary Table S2). It should be noted that some re-heated Salal Glacier meltinclusions contained anomalously high FeO* contents (up to 29 wt %).This phenomenon, described by Rowe (2006), is due to the presence ofpre-entrapment magnetite, chromite and Fe-Ni sulphides. As the latterare present in three inclusions, we attribute Fe gain during re-heating tothe partial melting of Fe-Ni sulphides within Salal Glacier inclusions.These inclusions are not considered representative of the initial trapped

melt and will not be shown in diagrams nor discussed in this study.The complete dataset of corrected major, volatile and trace element

data is provided in Supplementary Tables S2 and S3.

4.2.1. Major elementsOlivine hosts from all volcanic centres have forsterite contents of 77

to 89 mol % (Supplementary Table S2). Corrected melt inclusions alongthe arc are rather variable and have 43–58 wt% SiO2, 4–15 wt% FeO*,2.5–10 wt% MgO, 0.4–2 wt% K2O and 2–8 wt% total alkalis (Fig. 2;Supplementary Table S2). The most evolved compositions, with thehighest SiO2, K2O and lowest FeO* are from Glacier Peak, Mount Bakerand Garibaldi Lake (Fig. 2). Bridge River and Salal Glacier melt inclu-sions are the most primitive with the lowest SiO2 and highest MgO andFeO* (Fig. 2). These melts also contain up to 2.8 wt% TiO2, which ishigher than typical arc values (up to 1.5 wt%; Aragón et al., 2013).

4.2.2. VolatilesVolatile contents (H2O, CO2, S and Cl) are variable along the arc.

Chlorine and sulphur contents vary significantly between100–1700 ppm and 90–4000 ppm, respectively (Supplementary TableS2). The highest Cl contents are from Garibaldi Lake and Mount Cayleywhile the highest S contents are from Mount Meager, which is con-sistent with the presence of sulphide globules and suggests saturation ofthe melt with respect to sulfide. For many samples, carbon dioxide wasmeasured both in the melt and in the shrinkage bubble. Without thebubble, the melt inclusion CO2 contents across the arc range between90–6000 ppm (Fig. 3a; Supplementary Table S2). Both Glacier Peak andMount Baker contain the lowest CO2 contents (< 850 ppm) whileMount Meager contains the highest (> 1500 ppm). Raman analysesallowed for the calculation of the CO2 density within shrinkage bubblesfrom Garibaldi Lake, Mount Cayley, Mount Meager, Salal Glacier andBridge River. Using the distance between the main CO2 peaks and the

Fig. 2. Post entrapment crystallisation-cor-rected major elements of melt inclusions in thisstudy. a) Total alkalis versus SiO2; b) MgO vsSiO2; c) K2O vs SiO2 and d) FeO* vs SiO2. Thereis a narrow range in olivine host Fo contentalong the arc, meaning the melt inclusioncompositions should be comparable. In eachdiagram, Bridge River and Salal Glacier areconsistently the most primitive along the arc.This is followed by a gradual southward in-crease in K2O and decrease in MgO and FeO*wt%. Errors are 2σ. Data from Glacier Peak andMount Baker are from Shaw (2011).

S. Venugopal, et al. Chemical Geology 532 (2020) 119346

4

equation by Wang et al. (2011), the CO2 densities range from 0.15 to0.6 g/cm3. From here, the total amount of CO2 in GVB melt inclusionswas calculated and varies between 1300 – 10,000 ppm (Fig. 3b; Steele-MacInnis et al., 2011; Supplementary Table S2); this range excludesGlacier Peak and Mount Baker (explained below). The highest total CO2

values are from Bridge River. Consistent with current literature esti-mates (Anderson and Brown, 1993; Hartley et al., 2015; Moore et al.,2015; Wallace et al., 2015; Neave et al., 2015; Aster et al., 2016;Maclennan, 2017 and references therein), approximately 40–80 % ofthe total CO2 is found in the bubble (see Supplementary Table S5 for aworked example)

Shaw (2011) did not perform Raman analyses in the bubbles ofGlacier Peak and Mount Baker samples, thus the total CO2 content ofthe melt inclusion is unknown. The amount of CO2 that partitions be-tween the glass and bubble depends on several factors including thePEC, magma ascent pathway, closure temperature and the potential fordecrepitation (Moore et al., 2015; Wallace et al., 2015; Aster et al.,2016; Maclennan, 2017 and references therein). Comparing with theglassy Mount Meager melt inclusions, Glacier Peak and Mount Bakerhave similar PEC (1–5 %; Supplementary Table S2) and entrapmenttemperatures (Section 4.3). Since none of the melt inclusions in thisstudy showed evidence of decrepitation, the only unknown parameter isthe magma ascent pathway, which we assume to be similar between the3 centres. Based on these factors, the amount of CO2 that partitions intothe bubble of Mount Meager samples can be extrapolated to GlacierPeak and Mount Baker. The average amount of the total CO2 found inthe bubble of Mount Meager inclusions is 60% (total range 38–80%).Therefore, using 60%, the total CO2 content of Glacier Peak and MountBaker increases to 400–2000 ppm and the total along arc CO2 contentbecomes 400 – 10,000 ppm (Fig. 3b; Supplementary Table S2). How-ever, we would like to emphasize that the addition of 60 % is relativelysubjective but provides a reasonable baseline for comparison of thetotal melt inclusion CO2 contents along the arc.

Water contents, measured in the melt, varies between 0.1–3 wt%(Fig. 3a, b, c). The lowest water contents are in melt inclusions fromGaribaldi Lake and Mount Cayley, while the highest values are fromMount Meager (Fig. 3; Supplementary Table S2). The D/H isotopic ratiowas measured in a select number of melt inclusions in order to assesswhether the observed H2O trends are due to post-entrapment H + loss

as opposed to trapping of variably degassed melts. Values are expressedas δD, which is the permil deviation from Standard Mean Ocean Water(δDSMOW). The values of δD from Garibaldi Lake to Bridge River rangefrom -17 to -150 ‰ (Fig. 3c; Supplementary Table S2). Values of δDbetween -20 and -80 ‰ are typical of slab fluids to the average upperMORB mantle (Fig. 3c; Shaw et al., 2008) while values closer to -150 ‰are representative of the upper limit of terrestrial mantle values(-140 ± 20 ‰; Fig. 3c; Hallis et al., 2015). Positive δD values, and anegative trend with H2O wt%, suggest loss of hydrogen, addition ofseawater and/or secondary hydrothermal alteration (Kyser and O’Neil,1984; Shaw et al., 2008; Portnyagin et al., 2008). The highest δD values(least negative) belong to melt inclusions from Garibaldi Lake andMount Cayley while melt inclusions with δD values closest to mantlevalues belong to Bridge River and Salal Glacier (Fig. 3c). Nonetheless,the δD of these melt inclusions imply primary water contents that havebeen negligibly affected by post-entrapment H + diffusion. Previousauthors have shown that long duration re-heating experiments, ontimescales of hours to days, can result in significant water loss from themelt inclusion (i.e., Portnyagin et al., 2008; Gaetani et al., 2012;Bucholz et al., 2013). δD of re-heated melt inclusions in this study donot provide any evidence for extensive water loss during experiments.

4.2.3. Trace elementsFrom Glacier Peak to Mount Meager, trace element contents of melt

inclusions show arc-like characteristics with a depletion in immobile,high field strength elements (HFSE) as well as a relative enrichment inlarge ion lithophile elements (LILE) over the light rare earth elements(LREE) when compared to normal mid-ocean ridge basalt (NMORB) andocean island basalt (OIB) magmas (Fig. 4a, b; Supplementary Table S3).Furthermore, there are notable enrichments in fluid-mobile, large ionlithophile elements (LILE), such as Ba, Pb and Sr, which suggest thepresence of a hydrous subduction component (Fig. 4a, b; Supplemen-tary Table S3; Sun and McDonough, 1989). These centres also show lowconcentrations of heavy rare earth elements (HREE) relative to primi-tive mantle values ((Sm/Yb)N = 1.1–5.7), which is indicative of garnetin the mantle source (Jennings et al., 2017; Mullen and Weis, 2015; Sunand McDonough, 1989).

Trace element contents of melt inclusions from Salal Glacier andBridge River are unique along the arc and characterized by enrichment

Fig. 3. Comparing the water contents of all melt inclusion against a) CO2 in the glass b) total melt inclusion CO2 (glass + bubble) and c) δD. Isobars calculated usingVolatileCalc (Newman and Lowenstern, 2002) are shown in a and b. a) Using only the glass measurements of CO2, the pressure of entrapment varies between 1–4kbar, with the exception of a few melt inclusions from Bridge River and Salal Glacier. b) Pressure estimates rapidly increase when considering the amount of CO2 thatpartitions into the vapor bubble. Raman analyses were performed for all samples except for Glacier Peak and Mount Baker. Using Mount Meager as a reference, theaverage amount of CO2 that entered the bubble is 60%. Therefore, we use this amount to correct the CO2 contents of Glacier Peak and Mount Baker. See text fordetails. Bridge River and Salal Glacier consistently have the highest pressure of entrapment along the arc, while Glacier Peak and Mount Baker have the lowest. c) δDof melt inclusions, normalized to SMOW, excluding Glacier Peak and Mount Baker. Most inclusions lie within the range of MORB (grey box), but with elevated H2Ocontents. MORB values from Shaw et al. (2008) and Haui (2002). More negative values (up to -150) are typical of the upper limit of the MORB mantle (Hallis et al.,2015). Positive δD of melt inclusions suggest diffusive H + loss either during magmatic ascent or during re-heating experiments. Typically, melt inclusions sufferrapid water loss during re-heating experiments, on timescales of hours to days (i.e., Portnyagin et al., 2008; Gaetani et al., 2012; Bucholz et al., 2013). However, ourexperiments were 15 minutes long and melt inclusions were hosted in large olivine hosts (e.g. Chen et al., 2011). As such, the δD of re-heated melt inclusions(Garibaldi Lake, Mount Cayley, Salal Glacier and Bridge River) are primary and not indicative of extensive water loss. Error bars are 2σ and, unless otherwise shown,are the same size as data markers. Data for Glacier Peak and Mount Baker are from Shaw (2011).

S. Venugopal, et al. Chemical Geology 532 (2020) 119346

5

in HFSE relative to primitive mantle values. These patterns do not re-semble previously discussed centres. The signatures of Bridge River andSalal Glacier are similar to enriched asthenospheric melts as opposed toarc magmas and this is mirrored in the low Ba/Ta values between140–220 for Salal Glacier and 120–250 for Bridge River (typical arcvalues are> 500; Borg et al., 2002; Fig. 5). Relative to the trend be-tween Glacier Peak and Mount Meager, there is a weaker depletion ofHREE, which suggests a reduced importance of garnet in the source(Jennings et al., 2017). Between these two volcanoes, Salal Glaciermelts tend to be slightly more enriched in HFSE and LILE. Overall, thetrend for both Salal Glacier and Bridge River closely follows that ofocean island basalts based on data by Sun and McDonough (1989)(Fig. 4c).

4.3. Magmatic conditions

Melt inclusion temperatures were calculated from olivine - meltinclusion equilibria using equation 13 from Putirka (2008). Whereapplicable, calculated values are consistent with re-heating experi-ments, which range from 1130 to 1190 °C. Using corrected melt

inclusion compositions, Garibaldi Lake melts are 1117–1178 °C, MountCayley are 1090–1149 °C, Mount Meager 1057–1142 °C, Salal Glacier1186–1303 °C and Bridge River 1241–1312 °C (Fig. 5). Errors for tem-perature values are± 50 °C. These values are also consistent withPetrolog3, which estimates the liquidus temperature of the melts(Danyushevsky and Plechov, 2011). From Shaw (2011), temperatureswere originally determined using the equations from Sugawara (2000)and Medard and Grove (2008) but are comparable to the recalculatedtemperatures using equation 13 from Putirka (2008) and Petrolog3:Glacier Peak melts are 1113–1179 °C while Mount Baker melts are1076–1116 °C (Fig. 5; Supplementary Table S2).

The pressure of inclusions was estimated using the volatile satura-tion model of Papale et al. (2006) using H2O and CO2 contents. Re-calculated CO2 contents for all volcanic centres range between 420 –10,000 ppm and subsequently yield pressure estimates between 0.4 to6.8 kbars (Fig. 5; Supplementary Table S2). These values reflect thevariably degassed nature of melt inclusions (shown previously by H2Oand δD) and suggest polybaric olivine crystallization. As such, weconsider these pressure values as a minimum. To avoid under-estimation, the maximum recorded pressure (Pmax) for each centre is

Fig. 4. Primitive mantle-normalised spider diagram of meltinclusions from a) Glacier Peak and Mount Baker (data fromShaw, 2011) b) Garibaldi Lake, Mount Cayley and MountMeager and c) Salal Glacier and Bridge River. Referencecompositions of EMORB, NMORB and OIB are from Sun andMcDonough (1989). Centres from Glacier Peak to MountMeager exhibit arc-like patterns with enriched concentrationsof fluid-mobile LILE and depletions in HFSE and LREE. Con-versely, Salal Glacier and Bridge River show depletions in LILEand elevated HFSE, particularly Nb and Ta. These enrichmentssuggest preservation of mantle-derived HFSE without the di-lution supplied by a subduction component. The trace ele-ments patterns of Salal Glacier and Bridge River closely followthat of OIB, further supporting a lack of subduction influences.

S. Venugopal, et al. Chemical Geology 532 (2020) 119346

6

considered representative of the deepest storage conditions. The lowestPmax values are from Glacier Peak and Mount Baker (2 kbars) while thehighest are from Salal Glacier (6.8 kbars).

The sulphur speciation was measured following the method of Jugoet al. (2005) and can be used to estimate the magmatic oxygen fugacitybeneath Glacier Peak, Mount Baker and Mount Meager. Glacier Peakmagmas are +0.3 < NNO<+0.82 and Mount Baker magmas areNNO + 0.29. Mount Meager magmas are fO2 NNO + 0.74 (Supple-mentary Table S2). Oxygen fugacity cannot be calculated for GaribaldiLake, Mount Cayley, Salal Glacier and Bridge River since the inclusionshave been re-heated. The sulphur speciation of these melt inclusionswill only yield fO2 values of the heating apparatus, not the magma. TheFe3+/∑Fe ratio, another useful proxy for the oxidation state of magma,was estimated using the major oxide glass composition as well astemperature and fO2 (Kress and Carmichael, 1991). Glacier Peak andMount Baker have Fe3+/∑Fe values between 0.14 to 0.26 while MountMeager melts are more oxidised with values between 0.26 and 0.28(Supplementary Table S2). For Bridge River and Salal Glacier, Fe3+/∑Fe was measured using wet chemistry of whole rocks by Mathews(1958) and Green (2006) and yield values of 0.16. From this value, thefO2 conditions can be back-calculated as the required parameters areknown (T, P and major elements of glasses). The magma beneath bothBridge River and Salal Glacier is much less oxidized than the othercentres with fO2 values of -0.37 < NNO< -1.27 and-0.64 < NNO< -1.36, respectively (Supplementary Table S2).

5. Discussion

5.1. Chemical trends along the Garibaldi Volcanic Belt

Based on the above results and calculated magmatic conditions,there are clear along-arc trends. Consistent with previous work (Mullenand Weis, 2013, 2015), there is a subtle transition from arc-typical calc-alkaline magmas in the south beneath Glacier Peak and Mount Bakertowards increasingly alkalic melts in the north (Fig. 2). Not only areBridge River and Salal Glacier the most alkaline, these melts are also theleast hydrous and contain the lowest chlorine contents along arc(Supplementary Table S2).

There is also a clear northward increase in maximum melt inclusionentrapment temperature and pressure and a decrease in the degree ofmantle oxidation (Fig. 5; Supplementary Table S2). Over the narrowrange in Fo contents of olivine hosts along the entire arc, Glacier Peakand Mount Baker yield the lowest temperatures, shallowest pressures(average 1120 °C and up to 2 kbars) and the most oxidized magmasalong the arc; Bridge River and Salal Glacier melts are significantlyhotter, trapped much deeper (average 1260 °C and up to 6.8 kbars), andare much more reduced than the other centres within the GVB.

Fig. 5. Trace element and magmatic conditions of melt inclusions with distancealong the arc. Pressure values are calculated using the volatile saturation modelof Papale et al. (2006) using H2O and CO2 contents. CO2 contents include thebubble and include the addition of 60% for Glacier Peak and Mount Baker (seetext for details). Temperatures are calculated using olivine - melt equilibriausing equation 13 from Putirka (2008). There is a marked increase in Nb anddecrease in Ba/Ta between Bridge River, Salal Glacier and the other centres. AsNb is mainly derived from the mantle, its high concentration in the northernmagmas suggest the mantle signature is preserved and not overprinted by ad-dition of any subduction component. Bridge River and Salal Glacier also reportthe highest temperatures and the deepest pressures of melt inclusion entrap-ment along the arc. Conversely, the remaining centres, from Glacier Peak toMount Meager, show an arc-typical trend: low Nb contents that become over-printed by the addition of subduction-derived components, which is shown byhigh values of Ba/Ta. The addition of subduction-derived components alsolowers the temperature of entrapment. Typical 2σ error bars are shown. Datafor Glacier Peak and Mount Baker are from Shaw (2011).

S. Venugopal, et al. Chemical Geology 532 (2020) 119346

7

5.2. Magmatic sources: northern and southern groups

Trace element patterns along the arc can provide insight into thenature of the magmatic source beneath the arc (Figs. 4 & 5). Thenorthernmost volcanoes show trace element variations that are moretypical of OIB magmas, which is consistent with the higher melt in-clusion temperatures (Figs. 4 and 5; Sun and McDonough, 1989). Theyalso show the greatest enrichment in Nb (> 20; Fig. 5) as well as allother HFS elements (Supplementary Table S3). As these HFS elementsare relatively fluid-immobile, and derived from the mantle, this sug-gests preservation of initial mantle values and thus minimal addition(i.e. dilution) of these elements by the subducting slab. This is furthershown in the depletion in fluid-mobile, slab-sourced elements such as Srrelative to typical arc magmas (maximum 836 ppm; SupplementaryTable S3). Based on LILE/HFSE ratios, the northernmost volcanoes lacka subduction or arc signature. For example, arc magmas typically haveelevated Ba/Ta values due to fluids delivered by the subducting slab.Referring to Fig. 5, Ba/Ta is the lowest beneath the northern end of thearc (up to 250) and therefore suggests minimal or lack of subduction-derived fluids (Borg et al., 2002). Centres between Glacier Peak andMount Meager show typical arc patterns with low Nb (5–12 ppm), andhigh Sr and Ba/Ta (up to 1920 ppm and 1900, respectively) (Fig. 5;Supplementary Table S3), all of which are indicative of subduction-related processes (e.g., Borg et al., 2002; Portnyagin et al., 2007;Vigouroux et al., 2012 and references therein). The variability of Ba/Tabetween Glacier Peak to Mount Meager could be explained by extensiveplagioclase fractionation prior to melt inclusion entrapment leading tolower Ba/Ta below Glacier Peak and Mount Baker. However, this wouldproduce a negative Eu anomaly, which is not seen on Fig. 4. A morelikely explanation is the variable deliverance of subduction-derivedfluids between these centres. This is discussed further in Section 5.4.1.Finally, all centres have very low Th/Yb (< 3.2) and Sr/Y (< 90),which shows that for the whole arc, there is respectively little sedimentand slab melt from the downgoing plate.

These along-arc variations are best shown by comparing Cl/Nb andSr/Nd (Fig. 6). Each respective elemental pair have similar partitioningduring magmatic processes except in the presence of hydrous fluids, forwhich Cl and Sr have high affinity (Walowski et al., 2016; Portnyaginet al., 2007). Therefore, these ratios are excellent tracers for

subduction-derived fluids. There is a clear break between Bridge Riverand Salal Glacier and the rest of the arc. The enrichment in Cl and Srfrom Glacier Peak to Mount Meager, coupled with the higher watercontents, provides strong evidence for the circulation of hydrous fluidswithin the mantle wedge. In terms of the northern centres, the deple-tion in Cl and Sr suggests negligible subduction-related fluids.

Overall, the distinct nature of the northernmost volcanoes can beexplained by the Ba and Nb contents of melt inclusions, which definetwo distinct trends (Fig. 7). This suggests the presence of two differentmantle sources beneath the GVB. One source underlies Glacier Peak toMount Meager which is capable of producing magmas with Ba/Nb>30and another beneath Bridge River and Salal Glacier that can producesignificantly lower Ba/Nb (< 13). We therefore propose that the arccan be segmented into two groups: the Northern group (Bridge Riverand Salal Glacier; Ba/Nb = 13; Fig. 7) and the Southern group (GlacierPeak, Mount Baker, Garibaldi Lake, Mount Cayley and Mount Meager;Ba/Nb = 45–60; Fig. 7).

The Zr/Nb ratio can be used to further distinguish these differentmantle sources, as these values are not affected by subduction input orfractional crystallization (Mullen and Weis, 2015; Pearce and Norry,1979). Beneath the Southern group, the average Zr/Nb is 27, which istypical of N-MORB values (∼30; Sun and McDonough, 1989) and in-dicative of partial melting of a depleted mantle plus a subductioncomponent. Conversely, beneath the Northern group, the average Zr/Nb is 7, which is more typical of an enriched mantle source (Vigourouxet al., 2012; Sun and McDonough, 1989). A single source cannot pro-duce magmas which such different Zr/Nb ratios (Mullen and Weis,2015).

5.3. Tectonic link

The segmentation of the arc strongly correlates with the presence ofthe Nootka Fault, which is a transform fault that defines the boundarybetween the Juan de Fuca and Explorer Plates and transects theNorthern group (Fig. 1; Hyndman et al., 1979). Offshore, it is a left-lateral strike slip fault between the two plates. On shore, it is inter-preted as the surficial manifestation of a slab tear or fracture zone be-tween the subducting Juan de Fuca Plate and the relatively stagnantExplorer Plate (Hyndman et al., 1979). Schmidt et al. (2008) proposedthat this fracture zone was a pathway for seawater to serpentinise the

Fig. 6. Cl/Nb vs Sr/Nd as an indicator of subduction-derived fluids beneatheach centre along the GVB, with a logarithmic y-axis. Each respective elementalpair shows similar partitioning behaviour during magmatic processes, except influids derived from the downgoing slab for which Cl and Sr have a high affinity.Bridge River and Salal Glacier show much lower values for both ratios, sup-porting the lack of subduction zone fluids while each centre between GlacierPeak and Mount Meager show elevated Cl/Nb and Sr/Nd, illustrating the im-portance of subduction-derived fluids. Typical 2σ error bars are shown. Data forGlacier Peak and Mount Baker are from Shaw (2011).

Fig. 7. Ba versus Nb for all melt inclusions. Two distinct trends clearly re-present two separate mantle sources beneath the arc. One subduction modifiedsource beneath Glacier Peak to Mount Meager (Ba/Nb = 45–60) and anotherbeneath Bridge River and Salal Glacier (Ba/Nb<13), which experiences minorsubduction influence. Typical 2σ error bars are shown. Data for Glacier Peakand Mount Baker are from Shaw (2011).

S. Venugopal, et al. Chemical Geology 532 (2020) 119346

8

subducting oceanic crust. However, this implies that the Northerngroup magmas would experience some degree of subduction influence.Conversely, Green (2006) and Mullen and Weis (2015) suggest thisfracture zone is deep enough to allow asthenospheric upwelling tooccur and exert a compositional control on the eruptive suites, whichwould explain the enriched nature of the Northern group magmas.

5.4. Modelling the magmatic source

5.4.1. Southern groupThe current whole-rock literature states that one enriched mantle

source, similar to that of intraplate basalts, underlies the GVB, and thatthe along-arc variability can be explained through variable mixingproportions of depleted mantle with slab-derived melts (Mullen et al.,2017; Mullen and Weis, 2015 and references therein). Conversely,Walowski et al. (2016) modelled the melt inclusion compositions fromthe Lassen region in the southern High Cascades through two-compo-nent mixing between enriched MORB mantle and oceanic crustal melts(i.e., sediment and slab components). This model begins with fluid-fluxmelting of the upper basaltic layer of the downgoing oceanic crustbeneath the arc and the subsequent migration of these hydrous meltsinto the overlying mantle wedge to induce melting and generate hy-drous calc-alkaline basalts (Walowski et al., 2016). This model is basedon findings that show the downgoing slab is already dehydrated once itreaches sub-arc depths. Based on the MORB-like boron contents ofLassen melt inclusions, the subducting slab is considered dry andtherefore contributes no fluids to the mantle wedge (Walowski et al.,2016). However, melt inclusion compositions across the GVB do notshow evidence for significant slab melt contribution to the mantlewedge (La/Yb<25; Defant and Drummond, 1990; Moyen, 2009). In-deed the Juan de Fuca Plate subducts beneath both the High Cascadesand the GVB, there is a distinct compositional difference between thetwo arcs, namely the transitional alkaline nature of GVB basalts and thelack of a sediment input (across arc Th/La<0.15 and based on highprecision whole rock isotope data; Mullen et al., 2017; Mullen andWeis, 2015 and references therein). Based on these factors, the mainsubduction component beneath the GVB are most likely fluids deliveredby the downgoing slab.

Considering the contrasting melt models for the GVB and the HighCascades, we initially model magma generation beneath the GVB with adepleted MORB mantle (DMM; Salters and Stracke, 2004) and add ahydrous subduction component which was calculated using the methodof Portnyagin et al. (2007) (Fig. 8a; Supplementary Table S6). De-termining the composition of the hydrous subduction component in-volves first estimating the amount of partial melting of the DMM usingelements with low solubility in slab-derived fluids (Nb, Zr and Yb). Thedegree of partial melting varies between 5–12 %. Using the degree ofmelting and the melt inclusion composition, the composition of thesubduction-modified mantle wedge is calculated with the batch meltingequation. Finally, the difference, or excess, between the DMM and thesubduction-modified mantle wedge reflects the addition by the hydroussubduction component. The amount of hydrous fluids added can beestimated by the H2O content of the melt inclusions. However, factorssuch as magma degassing prior to melt inclusion trapping, and to alesser degree post-entrapment H loss, would lead to an underestimationof the initial H2O contents and the amount of hydrous subduction fluidscontributed. Therefore, the melt inclusion H2O values are correctedusing the maximum H2O/K2O of each centre (Supplementary Table S2;following the methods of Aster et al., 2016 and Walowski et al., 2016).We perform this correction for the sole purpose of estimating theamount of hydrous fluids contributed by the downgoing slab. Re-constructed H2O values for the GVB range between 0.7–4.7 wt%. Usingonly the maximum reconstructed values (H2Omax) for each centre, thevalues vary between 1.5–4.7 wt%, which is consistent with maximumreconstructed estimates for the Lassen region (1.3–3.4 wt%; Walowskiet al., 2016). Using the reconstructed H2O contents, the modelled

amount of hydrous subduction fluids delivered to the mantle wedgevaries between 2–10 wt% (Fig. 8a). The composition of the hydroussubduction component varies slightly between the centres of theSouthern group. Interestingly, the modelled concentrations of Ba, Cland Sr in the hydrous component are the lowest beneath Mount Baker,reach a maximum beneath Garibaldi Lake and Mount Cayley, and wanebeneath Mount Meager (Supplementary Table S6). Not only is this trendmirrored in the along arc values of Ba/Ta on Fig. 5, but is also con-sistent with the Ba, Sr and Cl-rich nature of Mount Cayley and GaribaldiLake melt inclusions (up to 1700 ppm Cl). This is also seen in Fig. 6,where there is a gradual increase in Cl/Nb and Sr/Nd from MountMeager and Mount Baker towards Mount Cayley and Garibaldi Lake.

Fig. 8. Modelling the two distinct magmatic sources beneath the GVB. a) H2O/Ce vs Sr/Nd. Modelling the source beneath the Southern group starts with thedepleted MORB mantle (black diamond) and adding variable amounts of ahydrous subduction component, calculated using the method of Portnyagin et al(2007). This component is enriched in Ba, Sr, Cl, H2O and Pb. Grey area re-presents the modelling results that involve 5–12 % partial melt of DMM (blacklines) mixed with 2–10 wt% of the hydrous subduction component (dashedlines). Inclusions that lie above the grey area are assumed to have elevated slabfluid content and those that lie below are assumed to be more degassed melts.b) Modelling the enriched mantle source beneath the Northern group in termsof Dy/Yb and Nb/Y, with the Southern group melt inclusions plotted as re-ference. These ratios are optimal markers for the degree of partial melting, thepresence of garnet and heterogeneities in the mantle (Vigouroux et al., 2012).Elevated Dy/Yb and Nb/Y of the Northern group cannot be modelled by evenlow degrees of partial melting (1 %) of the DMM (black diamond). Melt in-clusions from the Northern group can be best modelled by low degrees of partialmelting (3–7 %) of the primitive mantle (Sun and McDonough, 1989) indicatedby tick marks along the model curve. The residue is garnet lherzolite (7 cpx: 25opx: 60 olv: 8 grt). Typical 2σ error bars are shown. Data for Glacier Peak andMount Baker are from Shaw (2011).

S. Venugopal, et al. Chemical Geology 532 (2020) 119346

9

5.4.2. Northern groupTo first test the hypothesis of a single mantle source beneath the

GVB, the Northern group melt inclusions are modelled using the sameprocedure as the Southern Group, i.e., depleted mantle with fluids de-livered by the subducting slab. Even with very low degrees of partialmelting (< 1%), the Northern group melt inclusion compositionscannot be recreated using this method (Fig. 8b). As discussed earlier,these melt compositions suggest an enriched mantle source that is po-tentially controlled by the Nootka Fault. It is thus useful to compare theDy/Yb and Nb/Y contents of melt inclusions as these elements are de-rived entirely from the mantle and not added by subduction-relatedcomponents (Fig. 8b). Furthermore, the Dy/Yb ratio is consideredsensitive to variations in the mantle source composition while Nb/Y is auseful proxy for the degree of partial melting and episodes of previousmelt extraction (Vigouroux et al., 2012 and references therein). TypicalMORB compositions have Dy/Yb between 1.3 and 2.0 and Nb/Y < 0.4while OIB have Dy/Yb>2 and Nb/Y > 0.8 (Fig. 8b; PetDB database;Lehnert et al., 2000). Dy/Yb can be affected by amphibole fractiona-tion, and this would be shown by a negative correlation between Dy/Yband SiO2 wt%, which is not seen in the dataset. Both Bridge River andSalal Glacier have higher Dy/Yb and Nb/Y than typical MORB melts butoverlap with OIB compositions. This suggests low degrees of partialmelting of an enriched mantle source. Modelling of the Northern groupthus begins with the mass balance equation for batch melting and withthe primitive mantle as the source (Fig. 8b; Sun and McDonough,1989). Distribution coefficients are provided in Supplementary TableS4. With melt fractions below 7 % and garnet lherzolite as a residualphase, the modelled partial melt compositions bear a strong resem-blance to the Northern group in terms of Dy/Yb and Nb/Y ratios(Fig. 8b). Modelling results are consistent with the elevated FeO* wt%(> 9), La/Ta (< 20) and low HREE contents of Northern group meltinclusions, all of which support OIB-like melt generation below thegarnet-spinel transition zone at 70 km (Gorring and Kay, 2001). Thismodel is also consistent with the whole-rock modelling results ofMullen and Weis (2013). However, the main and rather significantdifference is their incorporation of Mount Meager as having the sameenriched source as the Northern group. Although Mount Meager showstransition alkaline compositions, the trace element contents are clearlysubduction zone derived.

5.5. Magma generation beneath the GVB

The along-arc geochemical variations expressed by melt inclusionscollected from the majority of GVB volcanic centres requires partial

melting of two distinct mantle sources (Fig. 9). However, there aresignificant tectonic factors to consider. In the case of the Northerngroup, the derivation of deeply sourced primitive and enriched meltsbeneath Bridge River and Salal Glacier suggests that the Nootka Faultand slab tear is persistent through to the asthenosphere and is a fa-vourable pathway for ascending enriched melts.

Based on the glaciovolcanic overview by Wilson and Russell (2018),there has been one recorded deposit of calc-alkaline composition withinthe Salal Glacier Volcanic Field. Though one known calc-alkaline de-posit does not necessarily suggest that the Northern group experiencesoccasional subduction influence, it may however provide insight intothe nature of the ascending pathway. A likely case for melt generation isthe ascent of asthenospheric mantle through the slab tear and thesubsequent decompression-induced melting in the sub-arc mantle(Fig. 9). The ascent of asthenospheric melts through faults and fracturesdirected along the northern edge of the Juan de Fuca Plate followed bythe periodic mixing with the subduction-modified source beneath theSouthern group would account for the rare calc-alkaline signaturewithin the Salal Glacier Volcanic Field. We therefore propose that thearea of slab tear is more of a densely fractured zone between the Juande Fuca and Explorer Plates with each fault acting as a potentialpathway for ascending asthenospheric mantle (Fig. 9).

The Southern group requires a hydrous subduction component thatvaries in composition between each centre. The amount of the hydroussubduction component contributed to the mantle wedge is at a max-imum beneath Mount Cayley and Garibaldi Lake and lowest beneathGlacier Peak and Mount Baker (Fig. 9). The idea of more fluids exsolvedfrom a younger slab is contrary to the normally accepted relationshipbetween slab age and fluid exsolution. A younger slab typically implieshigher slab temperatures and thus, smaller additions of slab-derivedhydrous components to the mantle wedge (Mullen and Weis, 2015;Green and Harry, 1999). However, if we consider the slab dehydrationmodel by Walowski et al. (2016), then the slightly older plate beneathGlacier Peak and Mount Baker would have already dehydrated prior toreaching sub-GVB depths thereby supplying smaller amounts of fluids.As the slab becomes progressively younger to the north, the degree ofdehydration decreases (i.e., the slab remains hydrated) and thus theamount of fluids supplied increases. Nevertheless, the Juan de FucaPlate is one of the youngest subducting plates in the world and there-fore the total amount of fluids supplied would be very low in com-parison to global subduction zones with older slabs.

Fig. 9. Schematic diagram of magma genera-tion beneath the Northern and Southerngroups of the GVB. Starting at the slab tear inthe north, the age of the downgoing Juan deFuca Plate decreases southward from 5 Ma to10 Ma. This subsequently leads to southwardincrease in slab dip and the degree of slab de-hydration, according to the model proposed byWalowski et al. (2016). The amount of fluidsdelivered (blue arrows) to the depleted sub-arcmantle is highest beneath Mount Cayley andMount Garibaldi and lowest beneath MountBaker and Glacier Peak. Beneath the Northerngroup, the main compositional control is theslab tear. This gap between the Explorer andJuan de Fuca Plates allows for the upwelling ofenriched asthenospheric mantle (orangearrow), thus yielding the OIB-like signature ofthe Northern group melts. (For interpretationof the references to colour in this figure legend,the reader is referred to the web version of thisarticle).

S. Venugopal, et al. Chemical Geology 532 (2020) 119346

10

6. Conclusions

Olivine-hosted melt inclusions from the GVB provide new insightinto the critical factors controlling the composition of subduction zonemagmas. Factors such as the age of the downgoing slab, degree of fluidrelease into the mantle wedge as well as the tectonic framework of theGVB itself have resulted in a unique and pronounced north-south geo-chemical transition.

Beneath the northern GVB, the reduced nature of melt inclusionscoupled with higher temperatures and pressures of entrapment and theelevated contents of HFSE suggest an enriched mantle source ratherthan a subduction-modified source. We attribute these signatures to aslab tear, or dense fracture zone, between the northern edge of the Juande Fuca Plate and the neighbouring Explorer Plate. Modelling of theenriched source beneath the Northern group suggests that the slab tearis pervasive enough to reach the upper mantle. Through decompressionmelting, asthenospheric melts ascend through the tear and erupt at thesurface. In a global context, there are have been few eruptions of si-milarly enriched melts in an arc setting. Gorring and Kay (2001) showthat Patagonia plateau lavas have a distinct OIB-like signature that canbe attributed to the slab window formed by the subduction of the ChileRidge between the Nazca and Antarctic Plates. Trace element andmodelling results show that, similar to the Northern Group, melt gen-eration began near the garnet-spinel at 70 km depth (Gorring and Kay,2001).

Melt inclusions from the southern GVB, on the other hand, revealarc-typical compositions produced by a mix between the depletedMORB mantle and hydrous subduction components. The main differ-ence between the centres of the Southern group is the amount andconcentrations of Sr, Ba and Cl of the hydrous subduction component.The maximum amount of fluids with the highest Sr, Ba and Cl is de-livered to the mantle wedge beneath Mount Cayley and Garibaldi Lakewhile the lowest amounts are below Mount Baker and Glacier Peak.This pattern of fluid exsolution correlates positively with the age of thedowngoing Juan de Fuca Plate. Though Walowski et al. (2016) foundthat the downgoing Juan de Fuca Plate beneath the southern Cascadeshad significantly dehydrated prior to reaching sub-arc depths, ourmodelling results show that slab fluids are the dominant subductioncomponent beneath the southern GVB. These results suggest that thedifference in slab age between the southern Cascades and the southernGVB (3–4 Ma) is sufficient to preserve a more hydrated slab. To test theglobal applicability of younger slabs retaining their fluids, melt inclu-sion compositions from young (< 10 Ma) and hot subduction zones,such as South Chile and Mexico (Syracuse et al., 2010), would providefurther insight into the relationship between the slab age and the degreeof dehydration.

Our results show that abrupt along-arc compositional changes canbe used to identify structures in the slab that cannot otherwise beidentified by geophysical methods, either due to low-resolution ima-ging or a lack of available information. Furthermore, this work illus-trates the importance of the age and degree of dehydration of thedowngoing slab in subduction zone settings. From the delivery of slabmelts in the southern Cascades, to the exsolution of fluids in thesouthern GVB and the termination against the Nootka Fault and slabtear beneath the northern GVB, the Juan de Fuca Plate is capable ofproducing a diverse array of volcanic centres along the western marginof North America.

Acknowledgments

This work is a Clervolc contribution number 380. This work wassupported by NSERC Discovery grants to Williams-Jones and Russell(R03953) and a ClerVolc grant to Moune. Many thanks to Jean-LucDevidal and Etienne Deloule for their expertise in electron microprobeand SIMS analysis. The work and preparation of an andesitic CO2

standard by Federica Schiavi and Didier Laporte is greatly appreciated.

Estelle Rose-Koga and Nico Cluzel is thanked for the use of the re-heating stage at LMV. We would also like to thank Paul Wallace and ananonymous reviewer for many helpful comments that greatly improvedthe quality of this work.

Appendix A. Supplementary data

Supplementary material related to this article can be found, in theonline version, at doi:https://doi.org/10.1016/j.chemgeo.2019.119346.

References

Anderson, A.T., Brown, G.G., 1993. CO2 contents and formation pressures of someKilauean melt inclusions. Am. Mineral. 78, 794–803.

Aragón, E., Pinotti, L., Castro, A., Rabbia, O., Coniglio, J., Demartis, M., Hernando, I.,Cavarozzi, C., Aguilera, Y.E., 2013. The Farallon-Aluk ridge collision with SouthAmerica: implications for the geochemical changes of slab window magmas fromfore- to back-arc. Geosci. Front. 4, 377–388. https://doi.org/10.1016/j.gsf.2012.12.004.

Aster, E.M., Wallace, P.J., Moore, L.R., Watkins, J., Gazel, E., Bodnar, R.J., 2016.Reconstructing CO2 concentrations in basaltic melt inclusions using Raman analysisof vapour bubbles. J. Volcanol. Geotherm. Res. 323, 148–162. https://doi.org/10.1016/j.jvolgeores.2016.04.028.

Borg, L., Blichert-Toft, J., Clynne, M.A., 2002. Ancient and modern subduction zonecontributions to the mantle sources of lavas from the lassen region of California in-ferred from Lu–Hf isotopic systematics. J. Petrol. 43 (6), 705–723. https://doi.org/10.1093/petrology/43.4.705.

Bucholz, C.E., Gaetani, G.A., Behn, M.D., Shimizu, N., 2013. Post-entrapment modifica-tion of volatiles and oxygen fugacity in olivine-hosted melt inclusions. Earth Planet.Sci. Lett. 374, 145–155.

Candie, S.C., La Brecque, J.L., Larson, R.L., Pittman III, W.C., Golovchenko, X., Haxby,W.F., 1989. Magnetic Lineations of the World’s Ocean Basins. American Associationof Petroleum Geologists, Tulsa, Oklahoma.

Connor, C.B., Conway, F.M., 2000. Basaltic volcanic fields. Encyclopaedia of Volcanology.Academic Press, pp. 331–343.

Cooper, L.B., Ruscitto, D.M., Plank, T., Wallace, P.J., Syracuse, E.M., Manning, C.E., 2012.Global variations in H2O/Ce: 1. Slab surface temperatures beneath volcanic arcs.Geochem. Geophys. Geosystems 13 (3). https://doi.org/10.1029/2011GC003902.

Danyushevsky, L.V., Plechov, P., 2011. Petrolog3: integrated software for modellingcrystallization processes. Geochem. Geophys. Geosystems 12 (7).

Defant, M.J., Drummond, M.S., 1990. Derivation of some modern arc magmas by meltingof young subducted lithosphere. Nature 347, 662–665.

Gaffney, E.S., Damjanac, B., Valentine, G.A., 2007. Localization of volcanic activity; 2.Effects of pre-existing structure. Earth Planet. Sci. Lett. 263, 323–338.

Gorring, M.L., Kay, S.M., 2001. Mantle Processes and Sources of Neogene Slab WindowMagmas from Southern Patagonia, Argentina. J. Petrol. 42 (6), 1067–1094. https://doi.org/10.1093/petrology/42.6.1067.

Green, N.L., 2006. Influence of slab thermal structure on basalt source regions andmelting conditions: REE and HFSE constraints from the Garibaldi Volcanic Belt,northern Cascadia subduction system. Lithos 87 (1–2), 23–49. https://doi.org/10.1016/j.lithos.2005.05.003.

Green, N.L., Harry, D.L., 1999. On the relationship between subducted slab age and arcbasalt petrogenesis, Cascadia subduction system, North America. Earth Planet. Sci.Lett. 171 (3), 367–381. https://doi.org/10.1016/S0012-821X(99)00159-4.

Green, N.L., Sinha, A.K., 2005. Consequences of varied slab age and thermal structure onenrichment processes in the sub-arc mantle of the northern Cascadia subductionsystem. J. Volcanol. Geotherm. Res. 140 (1–3), 107–132. https://doi.org/10.1016/j.jvolgeores.2004.07.017.

Hallis, L.J., Huss, G.R., Nagashima, K., Taylor, G.J., Halldórsson, S.A., Hilton, D.R.,Meech, K.J., 2015. Evidence for primordial water in Earth’s deep mantle. Science 350(6262), 795–797. https://doi.org/10.1126/science.aac4834.

Hartley, M.E., Neave, D.A., Maclennan, J., Edmonds, M., Thordarson, T., 2015. Diffusiveover-hydration of olivine- hosted melt inclusions. Earth Planet. Sci. Lett. 425,168–178. https://doi.org/10.1016/j.epsl.2015.06.008.

Hauri, E., 2002. SIMS analysis of volatiles in silicate glasses, 2: isotopes and abundancesin Hawaiian melt inclusions. Chem. Geol. 183, 115–141.

Hildreth, W., 2007. Quaternary magmatism in the Cascades; geologic perspectives. U.S.Geological Survey Professional Paper 1744. pp. 125.

Hyndman, R., Riddihough, R., Herzer, R., 1979. The Nootka Fault Zone — a new plateboundary off western Canada. Geophys. J. R. Astron. Soc. 58 (3), 667–683.

Jennings, E.S., Gibson, S.A., Maclennan, J., Heinonen, J.S., 2017. Deep mixing of mantlemelts beneath continental flood basalt provinces: constraints from olivine- hostedmelt inclusions in primitive magmas. Geochim. Cosmochim. Acta 196, 36–57.

Jugo, P., Luth, R., Richards, J., 2005b. Experimental data on the speciation of sulfur as afunction of oxygen fugacity in basaltic melts. Geochim. Cosmochim. Acta 69,497–503. https://doi.org/10.1016/j.gca.2004.07.011.

Kress, V.C., Carmichael, I.S.E., 1991. The compressibility of silicate liquids containingFe2O3 and the effect of composition, temperature, oxygen fugacity and pressure ontheir redox states. Contrib. Mineral. Petrol. 108 (1–2), 82–92. https://doi.org/10.1007/BF00307328.

Kyser, T.K., O’Neil, J.R., 1984. Hydrogen isotope systematics of submarine basalts.

S. Venugopal, et al. Chemical Geology 532 (2020) 119346

11

Geochim. Cosmochim. Acta 48, 2123–2133.Lehnert, K., Su, Y., Langmuir, C.H., Sarbas, B., Nohl, U., 2000. A global geochemical

database structure for rocks. Geochem. Geophys. Geosystems 1 (5), 1012. https://doi.org/10.1029/1999GC000026.

Maclennan, J., 2017. Bubble formation and decrepitation control on the CO2 content ofolivine-hosted melt inclusions. Geochem. Geophys. Geosystems 2 (18), 597–616.

Mathews, W.H., 1958. Geology of the Mount Garibaldi map-area, southwestern BritishColumbia, Canada: part II. Geomorphology and Quaternary volcanic rocks. Geol. Soc.Am. Bull. 69, 179–198.

Medard, E., Grove, T.L., 2008. The effect of H2O on the olivine liquidus of basaltic melts:experiments and thermodynamic models. Contrib. Mineral. Petrol. 155, 417–432.

Moore, N.E., DeBari, S.M., 2012. Mafic magmas from Mount Baker in the northernCascade arc, Washington: probes into mantle and crustal processes. Contrib. Mineral.Petrol. 163 (3), 521–546. https://doi.org/10.1007/s00410-011-0686-4.

Moyen, J.-F., 2009. High Sr/Y and La/Yb ratios: the meaning of the “adakitic signature”.Lithos 112 (3–4), 556–574. https://doi.org/10.1016/j.lithos.2009.04.001.

Mullen, E., Weis, 2013. Sr-Nd-Hf-Pb isotope and trace element evidence for the origin ofalkalic basalts in the Garibaldi Belt, northern Cascade arc. Geochem. Geophys.Geosyst. 14, 3126–3155 In press.

Mullen, E.K., Weis, D., 2015. Evidence for trench-parallel mantle flow in the northernCascade Arc from basalt geochemistry. Earth Planet. Sci. Lett. 414, 100–107. https://doi.org/10.1016/j.epsl.2015.01.010.

Mullen, E.K., Weis, D., Marsh, N.B., Martindale, M., 2017. Primitive arc magma diversity:new geochemical insights in the Cascade Arc. Chem. Geol. 448, 43–70. https://doi.org/10.1016/j.chemgeo.2016.11.006.

Newman, S., Lowenstern, J.B., 2002. VolatileCalc: a silicate melt– H2O–CO2 solutionmodel written in Visual Basic for excel. Comput. Geosci. 28, 597–604.

Ohtani, E., Litasov, K.D., Hosoya, T., Kubo, T., Kondo, T., 2004. Water transport into thedeep mantle and formation of a hydrous transition zone. Phys. Earth Planet. Inter.143–144, 255–269. https://doi.org/10.1016/j.pepi.2003.09.015.

Papale, P., Moretti, R., Barbato, D., 2006. The compositional dependence of the saturationsurface of H2O + CO2fluids in silicate melts. Chem. Geol. 229 (1–3), 78–95. https://doi.org/10.1016/j.chemgeo.2006.01.013.

Pearce, J.A., Norry, M.J., 1979. Petrogenetic implications of Ti, Zr, Y, and Nb variationsin volcanic rocks. Contrib. Mineral. Petrol. 69, 33–47.

Portnyagin, M., Hoernle, K., Plechov, P., Mironov, N., Khubunaya, S., 2007. Constraintson mantle melting and composition and nature of slab components in volcanic arcsfrom volatiles (H2O, S, Cl, F) and trace elements in melt inclusions from theKamchatka Arc. Earth Planet. Sci. Lett. 255 (1–2), 53–69. https://doi.org/10.1016/j.epsl.2006.12.005.

Portnyagin, M., Almeev, R.R., Matveev, S., Holtz, F., 2008. Experimental evidence forrapid water exchange between melt inclusions in olivine and host magma. EarthPlanet. Sci. Lett. 272, 541–552. https://doi.org/10.1016/j.epsl.2008.05.020.

Putirka, K., 2008. Thermometers and barometers for volcanic systems. In: In: Putirka, K.,Tepley, F. (Eds.), Minerals, Inclusions and Volcanic Processes, vol 69. pp. 61–120Reviews in Mineralogy and Geochemistry, Mineralogical Soc. Am.

Rowe, M., 2006. Anomalously high Fe contents in rehomogenized olivine-hosted meltinclusions from oxidized magmas. Am. Mineral. 91 (1), 82–91. https://doi.org/10.2138/am.2006.1818.

Ruscitto, D.M., Wallace, P.J., Johnson, E.R., Kent, A.J.R., Bindeman, I.N., 2010. Volatilecontents of mafic magmas from cinder cones in the Central Oregon High Cascades:implications for magma formation and mantle conditions in a hot arc. Earth Planet.Sci. Lett. 298 (1–2), 153–161. https://doi.org/10.1016/j.epsl.2010.07.037.

Saginor, I., Gazel, E., Condie, C., Carr, M.J., 2013. Evolution of geochemical variationsalong the Central American volcanic front. Geochem. Geophys. Geosyst. 14. https://doi.org/10.1002/ggge.20259.

Salters, V.J.M., Stracke, A., 2004. Composition of the depleted mantle. Geochem.Geophys. Geosystems 5 (5). https://doi.org/10.1029/2003GC000597.

Schiano, P., 2003. Primitive mantle magmas recorded as silicate melt inclusions in

igneous minerals. Earth. Rev. 63, 121–144. https://doi.org/10.1016/S0012-8252(03)00034-5.

Schmidt, M.E., Grunder, A.L., Rowe, M.C., 2008. Segmentation of the Cascade Arc asindicated by Sr and Nd isotopic variation among diverse primitive basalts. EarthPlanet. Sci. Lett. 266 (1–2), 166–181. https://doi.org/10.1016/j.epsl.2007.11.013.

Shaw, S.D., 2011. H2O Contents in Olivine-Hosted Melt Inclusions From PrimitiveMagmas in the Northern Cascade. MSc thesis. University of Western Washington.

Shaw, A.M., Hauri, E.H., Fischer, T.P., Hilton, D.R., Kelley, K.A., 2008. Hydrogen isotopesin Mariana arc melt inclusions: implications for subduction dehydration and thedeep-Earth water cycle. Earth Planet. Sci. Lett. 275 (1–2), 138–145. https://doi.org/10.1016/j.epsl.2008.08.015.

Steele-MacInnis, M., Esposito, R., Bodnar, R.J., 2011. Thermodynamic model for the ef-fect of post-entrapment crystallization on the H2O-CO2 systematics of vapor-satu-rated, silicate melt inclusions. J. Petrol. 52, 2461–2482.

Stern, B., 2002. Subduction zones. Rev. Geophys. 40 (4), 1012. https://doi.org/10.1029/2001RG000108.

Sugawara, T., 2000. Thermodynamic analysis of Fe and Mg partitioning between plagi-oclase and silicate liquid. Contrib. Mineral. Petrol. 138, 101–113.

Sun, S., McDonough, W.F., 1989. Chemical and isotopic systematics of oceanic basalts:implications for mantle composition and processes. Geol. Soc. London Spec. Publ. 42(1), 313–345. https://doi.org/10.1144/GSL.SP.1989.042.01.19.

Syracuse, E.M., van Keken, P.E., Abers, G.A., Suetsugu, D., Bina, C., Inoue, T., Jellinek,M., 2010. The global range of subduction zone thermal models. Phys. Earth Planet.Inter. 183 (1–2), 73–90. https://doi.org/10.1016/j.pepi.2010.02.004.

Thorkelson, D.J., Madsen, J.K., Sluggett, C.L., 2011. Mantle flow through the NorthernCordilleran slab window revealed by volcanic geochemistry. Geology 39 (3),267–270. https://doi.org/10.1130/G31522.1.

Toplis, M.J., 2005. The thermodynamics of iron and magnesium partitioning betweenolivine and liquid: criteria for assessing and predicting equilibrium in natural andexperimental systems. Contrib. Mineral. Petrol. 149 (1), 22–39. https://doi.org/10.1007/s00410-004-0629-4.

Ulmer, P., 2001. Partial melting in the mantle wedge—the role of H2O in the genesis ofmantle-derived “arc-related” magmas. Phys. Earth Planet. Inter 127, 215–232.

Venugopal, S., Moune, S., Williams-Jones, G., 2016. Investigating the subsurface con-nection beneath Cerro Negro volcano and the El Hoyo Complex, Nicaragua. J.Volcanol. Geotherm. Res. 325, 211–224.

Vigouroux, N., Wallace, P.J., Williams-Jones, G., Kelley, K., Kent, A.J.R., Williams-Jones,A.E., 2012. The sources of volatile and fluid-mobile elements in the Sunda arc: a meltinclusion study from Kawah Ijen and Tambora volcanoes, Indonesia. Geochem.Geophys. Geosystems 13 (9), 1–22. https://doi.org/10.1029/2012GC004192.

Wallace, P.J., 2005. Volatiles in subduction zone magmas: concentrations and fluxesbased on melt inclusion and volcanic gas data. J. Volcanol. Geophys. Res. 140,217–240. https://doi.org/10.1016/j.jvolgeores.2004.07.023.

Wallace, P.J., Kamenetsky, V.S., Cervantes, P., 2015. Melt inclusion CO2 contents,pressures of olivine crystallization, and the problem of shrinkage bubbles. Am.Mineral. 100 (4), 787–794.

Walowski, K.J., Wallace, P.J., Clynne, M.A., Rasmussen, D.J., Weis, D., 2016. Slab meltingand magma formation beneath the southern Cascade arc. Earth Planet. Sci. Lett. 446,100–112. https://doi.org/10.1016/j.epsl.2016.03.044.

Weller, D.J., Stern, C.R., 2018. Along-strike variability in primitive magmas (major andvolatile elements) inferred from olivine-hosted melt inclusions from volcanoes of thesouthernmost Andean Southern Volcanic Zone, Chile. Lithos 296–299, 233–2442018.

Wilson, A., Russell, J.K., 2018. Quaternary glaciovolcanism in the Canadian Cascadevolcanic arc - paleoenvironmental implications. Special Pap. Geol. Soc. Am. 538 (6),1–26. https://doi.org/10.1130/2018.2538(06).

Wilson, D.S., 2002. The Juan de Fuca Plate and slab: isochron structure and CenozoicPlate motions. US Geological Survey Open-File Report 02-328.

S. Venugopal, et al. Chemical Geology 532 (2020) 119346

12