Evolution of Mafic Tungnárhraun Lavas: Transcrustal Magma Storage and Ascent Beneath the Bárðarbunga Volcanic System

Abstract

The Tungnárhraun basalts in southern Iceland record a transcrustal magma system formed during Holocene deglaciation. These large-volume (>1 km³) Early through Mid-Holocene lavas contain ubiquitous plagioclase feldspar macrocrysts that are too primitive to have grown from the host lavas. Thermobarometry based on plagioclase melt and clinopyroxene melt equilibrium reveals a transcrustal structure with at least three distinct storage regions. A lower-crustal mush zone at ~14–30 km is fed by primitive, low ⁸⁷Sr/⁸⁶Sr magmas with diverse Ti/K and Al/Ti signatures. Plagioclase feldspar growth is controlled by an experimentally determined pseudoazeotrope where crystals develop inversely correlated An and Mg contents. The rapid ascent of magmas to mid-crustal levels (~8–9 km) allows the feldspar system to revert to conventional thermodynamic phase constraints. Continued plagioclase growth releases heat, causing olivine and pyroxene to be resorbed and giving the magmas their characteristic high CaO/Al₂O₃ values (~0.8–1.0) and Sc contents (~52 ppm in matrix material). Mid-Holocene MgO-rich lavas with abundant plagioclase feldspar macrocrysts erupted directly from this depth, but both older and younger magmas ascended to a shallow-crustal storage chamber (~5 km) where they crystallized olivine, clinopyroxene, and plagioclase feldspar and evolved to lower MgO contents. The Sr isotope differences between the plagioclase macrocrysts and their carrier melts suggest that the fractionation involves the minor assimilation of country rock. This model does not require the physical disruption of an established and long-lived gabbroic cumulate mush. The transcrustal structures documented here existed in south Iceland at least throughout the Holocene and likely influenced much of Icelandic magmatism.

1. Introduction

Icelandic basalt volcanism is supported by a series of stacked and often-interconnected sills or magma mush regions ranging in depth from the local Moho through the entire crustal section, e.g., [1,2,3,4,5,6,7,8,9]. The deepest storage domains are supplied by mafic melts with contributions from the heterogeneous mantle of the Mid-Atlantic Ridge and the Iceland mantle plume, e.g., [10,11,12,13,14,15,16,17,18]. Petrological and geochemical investigations of recent eruptions on the Reykjanes Peninsula at Fagradalsfjall and the Sundhnúksgígar crater row, as well as Holocene basalts in the Eastern Volcanic Zone, have shaped our understanding of these transcrustal magma systems [6,8,9,19,20,21,22,23,24,25]. Individual reservoirs may be connected intermittently or over long periods of time by conduits through which magma is transported vertically or laterally, and lavas often carry crystal cargo (e.g., glomerocrysts, macrocrysts, and cognate inclusions) interpreted as disrupted crystal mush frameworks remobilized by magmas traveling within or between reservoirs [3,5,8,26,27,28,29,30,31,32]. The geochemical features of the crystal cargo provide thermobarometric constraints on the crustal storage profile but detailed information on the long-term stability of transcrustal systems is broadly lacking.

We focus on Early to Mid-Holocene (~7900–2000 ybp) Tungnárhraun basalts from south Iceland because they fill an important temporal and compositional gap in the work that has sought to capture the transition from Early Holocene subglacial eruptions (i.e., Ljósufjöll, <100 ka; [8]) through the historic Veidivötn flow of 1477 AD [8,27,32,33,34,35]. Together, they span a critical period of enhanced eruptive activity that occurred in conjunction with the most recent deglaciation event beginning ~12 ka, e.g., [14]. Several Tungnárhraun flows contain up to ~20 vol.% anorthitic (An_≥85) plagioclase feldspar macrocrysts and are classified as plagioclase ultraphyric basalts (PUBs; [36]). PUBs occur in most mafic igneous environments and their plagioclase macrocrysts are often significantly more anorthitic than compositions in equilibrium with carrier lavas [8,27,34,36,37,38]. They are commonly associated with mid-ocean ridge segments and off-axis seamounts in areas of ultra-slow through intermediate spreading rates [36,38,39,40,41,42,43,44,45,46,47,48] and at ocean islands where they are also associated with periods of low magma flux (e.g., La Reunion, [49]). The high-volume Tungnárhraun PUBs that erupted during the period of deglaciation appear to defy this consistent pattern of occurrence.

Holocene deglaciation and isostatic rebound across Iceland are associated with increased volcanic activity and flow volumes [14,50,51,52,53,54,55], but their relationship to abundant anorthitic plagioclase macrocrysts remains unresolved. In some cases, the macrocrysts are shown to be cogenetic (e.g., [3]) but in others they are clearly xenocrysts as indicated by their An-rich compositions (e.g., [8,27]) and by the Sr isotopic differences between the plagioclase feldspars and surrounding basalt matrix [34,56,57]. These compositional differences have been interpreted to represent interactions among melts derived from the Iceland mantle plume and the geochemically heterogeneous mantle underlying the Mid-Atlantic Ridge, with magma transport through gabbroic crystal mushes at one or more crustal depths (e.g., [8,27,35]).

The Tungnárhraun basalts studied here are part of the Bárðarbunga volcanic system, which has been the most active eruptive region in Iceland since the last glacial maximum e.g., [32,35]. Bárðarbunga is one of several en echelon and parallel fissure systems of the Eastern Volcanic Zone; it consists of a centrally located 80 km² caldera positioned above the Iceland mantle plume and buried beneath the Vatnajökull glacier, and basaltic fissure swarms which emanate from the edge of Vatnajökull to the north and southwest [33,58,59,60]. This system is estimated to have erupted nearly 70 km³ of volcanic material since the beginning of the Holocene. Field and geochemical data from the distal portions of each of the Bárðarbunga fissures reveal the mixing of magmas with products of the Askja and Torfajökull volcanoes to the north and southwest, respectively [61,62,63,64,65]. The plagioclase-rich lavas of Tungnárhraun likely emanated from the Veiðivötn and/or Vatnaöldur fissures southwest of Bárðarbunga (Figure 1). The lava flows broadly follow the Tungná river from central Iceland to the southern coast. Individual flow volumes range from 0.1–25 km³; the total volume of Tungnárhraun is estimated at ~45 km³, just over half of the Holocene lava produced by the Bárðarbunga system [60,66,67]. Several Bárðarbunga eruptive units, including individual Tungnárhraun flows, are well characterized geochemically and/or petrographically [5,8,24,25,26,27,28,29,30,31,32,68,69,70]. Earlier work [8] suggested a clear temporal evolution from more primitive to more evolved carrier melts across the deglaciation. Our new data indicate a more complex scenario in which a series of primitive magmas erupted stratigraphically between units with lower MgO content during a period of low magma flux.

Our goal is to use the geochemical signatures of these lavas and their associated crystal cargo in conjunction with new and published thermobarometric constraints to (1) evaluate the conditions of magma storage, (2) reconstruct the transcrustal magma system of Bárðarbunga across the period of deglaciation during a period of varying magma flux. Experimental results on the pressure-controlled shape of the plagioclase pseudobinary phase diagram [72] help document the storage and ascent pathway of mafic lavas in southern Iceland over the past ~100,000 years. Specifically, we use the major and minor element distributions and Sr isotope signatures of the basaltic rocks and plagioclase feldspar macrocrysts, diffusion profiles in olivine, and mineral–melt thermobarometry to document the relationships between the crystals, their carrier melts, and their eruptive conditions. Our results indicate the long-lived availability of multiple basaltic parent magmas and a consistent pattern where the growth of plagioclase feldspar macrocrysts at pressures between 5 and 10 kbar occurs in concert with the resorption of clinopyroxene. This process appears to take place across all conditions of magma influx, with subsequent mid-crustal storage enhanced at times of higher flux. Results from Bárðarbunga are significant because the volcanic system has a long and well-documented eruptive history and provides potential insight into the effects of reducing glacial overburden pressure on eruptions from transcrustal magma systems. Our findings help constrain the evolution of the crustal structure and plumbing system beneath the Bárðarbunga volcanic system, which has been the most productive in Iceland throughout the Holocene. The Tungnárhraun PUBs are analogous to lavas found among Tertiary through recent basalts of Iceland [27,73,74,75,76] and thereby may constrain very long-lived transcrustal magmatic processes.

2. Materials and Methods

Roughly 90 samples with minimal visual weathering were collected from outcrops of the Tungnárhraun lava flows (Figure 1). Outcrops were located using Orkustofnun (National Energy Authority of Iceland) surface deposit and geological maps [77,78,79]; sample locations were georeferenced in ArcGIS using Sentinel 2 imagery and imported into the Avenza mapping app for iPhone [33,77,78,79]. Outcrops of interest were scanned using the 3D scanning program Scaniverse app for iPhone (Niantic, Inc., San Francisco, CA, USA). Sampling focused on lavas from units THd (formerly units THd and THe), THf, and THi (formerly units THh, THi, and Drekahraun) erupted between ~7900 and 3200 ybp [67], augmented by a small number of samples from well-studied older (THb or Thjórsá; [34]) and younger (THj, formerly THj, and THk) units for comparison.

Thin sections prepared by Spectrum Petrographic for petrographic observation and electron probe microanalysis (EPMA) were scanned for image analysis with a Zeiss AxioScan.Z1 (Oberkochen, Germany). Mineral compositions were collected using a Cameca SX Five Electron Probe Micro Analyzer (EPMA) (Gennevilliers, France) at the Pennsylvania State University Materials Characterization Laboratory (Table S1). Analyses were performed with an accelerating voltage of 15 kV and a probe current of 30 nA. The instrument was standardized for Na, Al, Si, and Ca on plagioclase, Mg and Cr on pyrope, K on orthoclase, Ti on synthetic sphene, and Mn on almandine. Three spot analyses, on average, were obtained from each area of interest in the macrocrysts; 3–8 plagioclase feldspar macrocrysts were analyzed for core and rim compositions in each sample. Plagioclase feldspar microlite (groundmass) compositions were determined for 10–15 individual crystals. Rim to core transect resolution for 16 olivine crystals was fixed at 5 µm spacing between points for the first 150 µm and then between 10 and 30 µm for the remainder of the transect. Olivine microlite compositions were determined for 2–8 individual crystals. Clinopyroxene macrocrysts were rare; 3 were analyzed for compositional variations. Microlite compositions were determined for 2–8 crystals.

Single mineral spot analyses of trace elements were measured using a Thermo Scientific iCAP RQ-ICP-MS (Waltham, MA, USA) attached to a Teledyne/Photon Machines Analyte G2 Excimer Laser Ablation System at the Pennsylvania State University LionChron Laboratory. The laser was operated at 20 Hz with a 40 μm ablation spot size. Data reduction was accomplished using Iolite v.4 software with ²⁹Si as an internal standard [80,81]. Natural and synthetic external standards reproduced measured elements to within 5% error. Analyses were carried out on 4–5 macrocrysts and 1–4 microlites in each sample. Compositions discussed in this paper are in Table S1.

Samples were prepared for whole rock analysis by cutting ~5 cm slabs with a ceramic tile saw. After polishing to remove saw marks, the slabs were reduced to ~5 mm pieces using an alumina ceramic jaw crusher and powdered in a tungsten carbide disk mill. Analysis focused on sparsely phyric portions of samples and large plagioclase feldspars were removed manually prior to powdering.

Analyses of major element abundances and Sc concentrations were performed at the Pennsylvania State University Laboratory for Isotopes and Metals in the Environment (LIME) on a Thermo iCAP 7400 Inductively Coupled Plasma Emission Spectrometer and a Thermo iCAP RQ Inductively Coupled Plasma Mass Spectrometer, respectively. Prior to major element analyses, powdered samples were prepared via lithium metaborate fusion [82]. A total of 50–100 mg of powdered sample was mixed with 400 mg of lithium metaborate powder and transferred to graphite crucibles. The mixtures were heated to 900 °C for 10 min and the melt beads were added to 100 mL of 5% HNO₃ solution in Teflon beakers for 20 min. Samples were further diluted by combining a 2.5 mL aliquot of the sample with 10 mL each of 2% HNO₃ and Lutetium internal standard solutions. Prior to Sc analysis, powdered basalt samples and hand-picked plagioclase-free glassy matrix materials (50 mg) were weighed into clean Teflon vials for acid digestion. Concentrated acid was used to digest the samples. Three milliliters of HF, one milliliter of HClO₄, and one milliliter of HNO₃ were added to the powder initially and the vial was sealed and heated at 100 °C for 24 h. Then, the samples were dried at 120 °C to drive off the HNO₃ and HF, and 0.5 mL of HClO₄ was added before heating at 120 °C overnight. Upon cooling, reverse aqua regia was added to the vial (3 mL of HNO₃ and 1 mL of HCl). Samples were reacted at room temperature for one hour before being sealed and heated overnight at 150 °C. Samples were then completely dried down at 160 °C and resuspended in 4 mL of 4N HNO₃ for analysis. After acid digestion, the samples were diluted by combining a 0.1 mL aliquot of digested sample with 5.9 mL of 2% HNO₃. Samples were run with standards BHVO-1, BCR-1, and BIR-1 and the instrument was calibrated with 21 natural and synthetic standards. Compositions discussed in this paper are in Tables S2 (major elements) and S3 (Sc abundances).

Sr isotopic analyses were performed at the Pennsylvania State University LIME facility. Sr was separated from the matrix using a two-step ion exchange column procedure: major cations were removed using AG50W-X8 (200–400 mesh) resin (Bio-Rad, Hercules, CA, USA) and Rb was separated from Sr using Sr Spec resin. Collected Sr was loaded onto a single Ta filament in 1 μL of 10% HNO₃. The filament was also loaded with 1 μL of Ta gel and 1 μL of phosphoric acid. Sr was analyzed on a Triton Plus Thermal Ionization Mass Spectrometer (TIMS) using ⁸⁶Sr/⁸⁸Sr = 8.37520938 as a normalization value. SRM 987 was used as an external standard at the beginning and end of each sample turret. Long term mean SRM 987 is ⁸⁷Sr/⁸⁶Sr = 0.710250 ± 0.000002. These data are in Table S4.

We calculated magma storage conditions based on clinopyroxene–melt pairs following [83], which has a standard error of estimate of ±1.4 kbar. Mineral–melt equilibrium was evaluated first through Fe–Mg exchange (i.e., K_D(Fe-Mg)^cpx-liq = 0.27 ± 0.03 [84]) and subsequently by comparing observed and calculated pyroxene components; we accepted values within 10% of a 1:1 correspondence (see experimental calibrations of [85,86]) for the CaTs component as it displays the greatest variability in our dataset. Crystallization temperatures for the plagioclase macrocrysts were calculated following [87].

3. Results

3.1. Petrography and Mineral Chemistry

The Tungnárhraun lavas are characterized by heterogeneous porphyritic textures with regions containing up to ~20 vol.% plagioclase feldspar macrocrysts (>1 cm; Figure 2) and sparse olivine + plagioclase glomerocrysts (~1 vol.%) in a glassy to microcrystalline matrix. The plagioclase macrocrysts are generally most abundant in the Middle Holocene units THd and THf (~14–20 vol.%), and less abundant in the older (Early Holocene Thjórsárhraun and THb, <5 vol.%) and younger units (Mid-Late Holocene, THi, and THj, ~5–10 vol.%) although many samples are heterogeneous in appearance at the outcrop, hand sample, and thin section scale.

The macrocrysts range from euhedral to subhedral and have normal zoning from ubiquitous An_85-91 cores to rims of lower and more variable An content (An_58-86) (Figure 3; Table S1); some crystals display fine-scale oscillatory zoning but remain overall normally zoned. The macrocrysts from the post-Early Holocene flow units commonly display several thin (~10–70 µm) compositionally distinct zones (rim, mantle, and core) identified on backscattered electron images. The Tungnárhraun plagioclase feldspar compositions overlap with published data from the Early Holocene Thjórsárhraun [34] and Mid-Holocene Drekahraun and Thjórsárdalshraun flows [8], and all samples include a population of highly calcic macrocrysts. Anorthite content decreases abruptly from core (An_~85-91) to mantle (An_~70-80) and typically decreases again at the crystal rim (An_~50-70) in all studied feldspars. The microcrysts, which occur in all lavas and appear as fractured pieces of larger feldspars, are compositionally similar to the cores and mantles of the macrocrysts (An_75-90). Groundmass plagioclase feldspars are commonly intergrown with macrocryst rims; their compositions in the Mid-Holocene units THf and THi range from An_91-82, while those from earlier and later more evolved units extend to lower values of An_91-76 (Figure 3; Table S1).

The plagioclase Ti and K contents (Table S1) display a strong negative correlation with An in all crystals, increasing progressively from macrocryst core to mantle to rim and groundmass crystals (Figure 4a,c). The Mg contents increase with decreasing An content in macrocryst cores (An_~93-85) and then decrease in rims and groundmass crystals (Figure 4b,d; Table S1) as observed in mid-ocean ridge basalts [88,89] and global PUBs [72].

The Tungnárhraun lavas contain up to ~1% olivine crystals which are euhedral to subhedral and are typically found near plagioclase macrocrysts. Olivine cores range from Fo_82-87 and are surrounded by narrow rims of Fo_75-80 (Figure 5; Table S1). Olivine microcrysts with composition Fo_69-84 are present in all samples; groundmass olivine is Fo_65-76. Representative olivine transects and backscattered electron images of the corresponding crystals are given in Figure 5. The CaO contents of olivine decrease slightly with decreasing Fo content in macrocrysts (0.35–0.41 wt.%) and microcrysts (0.32–0.42 wt.%) but extend to higher values in groundmass crystals (0.36–0.83 wt.%). Ni contents decrease from ~1450 to 700 ppm as olivine composition changes from Fo_~87-70; these values fall within the range of most Icelandic basalts [90] and mid-ocean ridge basalts derived from peridotitic source material [91].

The Tungnárhraun clinopyroxenes are augitic in composition (Table S1). Clinopyroxene macrocrysts are rare (<0.1%) and are compositionally Wo_42-49En_42-48Fs_9-17, with Mg# 83–85, 0.5–0.8 wt.% TiO₂, and 4.2–4.9 wt.% Al₂O₃. Microcrysts of pyroxene are present in all flows and have compositions of Wo_~40En_~47Fs_~13, with Mg# 75–85, 0.2–1.00 wt.% TiO₂, and 0.7–4.5 wt.% Al₂O₃. Groundmass pyroxene is mostly interstitial between the plagioclase laths and olivine and has a composition of Wo_31-41En_44-49Fs_10-25, with Mg# <80, 0.60–1.22 wt.% TiO₂, and 1.97–3.21 wt.% Al₂O₃. As an indicator of magmatic compositional heterogeneity, we calculated the Al₂O₃/TiO₂ values of the pyroxenes (Figure 6; Table S1). Values of this parameter decrease overall with decreasing Mg# and individual macrocrysts have wide ranges. Unlike the plagioclase feldspar macrocryst compositions, the pyroxenes from the mixed unit THj have a lower Mg# and Al₂O₃/TiO₂ values than those measured in the pyroxenes from unit THb samples of overlapping bulk rock MgO content (Figure 6).

3.2. Whole Rock Geochemistry

All Tungnárhraun mafic lavas are tholeiitic; their total alkalis and silica contents categorize them as basalts (Figure 7). The older lavas (units THa through THi) have ≤6 wt.% normative olivine and samples from the youngest unit (THj) are mildly quartz normative (<1 wt.%). Previous work [8] suggested that lava MgO contents decrease across the entire Tungnárhraun sequence, but our data (Table S2) are consistent with the results of tephra analysis [35] that indicate an episode of more primitive basalt eruptions between ~6400 and 3200 ybp [67]: flow units THb (Early Holocene), THd, and THj (Mid-Holocene) have lower MgO (5.6–6.9 wt.%) than samples from THf and THi (7.5–7.9 wt.% MgO; Figure 8). Basaltic lavas mapped as THe [33] are subsumed within THd in a more recent compilation [67] although their MgO contents fall within the range of THf and THi lavas (7.5–7.7 wt.% MgO; Table S2); we distinguish these two groups herein as THdd and THde.

All bulk rock analyses were made on portions of samples that contain ~<5 vol.% plagioclase feldspar megacrysts. Nonetheless, the plagioclase-phyric samples from units THdd, THde, and THf have slightly lower FeO, TiO₂, P₂O₅, and CaO/Al₂O₃ than the nearly aphyric lavas with a similar MgO content (Figure 8). These differences are minor and do not affect our findings in meaningful ways. The unit THj samples have anomalously high K₂O contents and encompass products of magma mixing between Veidivötn- (Tungnárhraun) and Torfajökull-type basalts [61,62,63,64,92] but we include them in this work because the plagioclase feldspar macrocrysts they contain are important to developing a coherent view of the regional crustal structure.

Figure 7. Total alkalis against silica for southeast Iceland lava suites [93]. Light orange field THb encircles data from [34]) on Thjórsárhraun. Gray field of tephras encloses data for Ljósufjöll, Fontur, Brandur, and Saxi cinder cones, Drekahraun, Thjórsárdalshraun, and the 1477 Veiðivötn lava [8], and for the 871 Vatnaöldur and 1477 Veiðivötn flows [35]. Other sources of data: [5,30,58,94,95,96,97,98,99,100].

Figure 7. Total alkalis against silica for southeast Iceland lava suites [93]. Light orange field THb encircles data from [34]) on Thjórsárhraun. Gray field of tephras encloses data for Ljósufjöll, Fontur, Brandur, and Saxi cinder cones, Drekahraun, Thjórsárdalshraun, and the 1477 Veiðivötn lava [8], and for the 871 Vatnaöldur and 1477 Veiðivötn flows [35]. Other sources of data: [5,30,58,94,95,96,97,98,99,100].

Figure 8. Major element variations against MgO for bulk Tungnáhraun lavas (data points) and groundmass glass compositions (shaded fields). (a) TiO₂, (b) K₂O, (c) P₂O₅. (d) TiO₂/K₂O variations show wide ranges among Holocene Tungnáhraun mafic lavas despite limited heterogeneity in K₂O contents. Drekahraun lavas include several samples with anomalously low K₂O (see inset for full range; dotted field encloses all data shown in the larger panel). (e) CaO/Al₂O₃. (f) CaO/Al₂O₃ where shaded fields enclose melt inclusions within plagioclase feldspar. Colors, symbols, and sources of data are as in Figure 7; melt inclusion data from [8,27].

Figure 8. Major element variations against MgO for bulk Tungnáhraun lavas (data points) and groundmass glass compositions (shaded fields). (a) TiO₂, (b) K₂O, (c) P₂O₅. (d) TiO₂/K₂O variations show wide ranges among Holocene Tungnáhraun mafic lavas despite limited heterogeneity in K₂O contents. Drekahraun lavas include several samples with anomalously low K₂O (see inset for full range; dotted field encloses all data shown in the larger panel). (e) CaO/Al₂O₃. (f) CaO/Al₂O₃ where shaded fields enclose melt inclusions within plagioclase feldspar. Colors, symbols, and sources of data are as in Figure 7; melt inclusion data from [8,27].

The Tungnárhraun mafic lavas (excluding unit THj) have CaO/Al₂O₃ values ~0.80–0.87 which are higher than those of typical mid-ocean ridge basalts (Figure 8e); CaO/Al₂O₃ decreases with decreasing MgO and mafic samples with abundant plagioclase feldspar macrocrysts have lower values than nearly aphyric samples with a comparable MgO content. Bulk lavas from all compositional groups are moderately rich in Sc (29–37 ppm; Table S2) while the plagioclase-free matrix materials of MgO-rich lavas studied here contain up to 52 ppm Sc (Table S3), which is substantially higher than the values observed in global MORB (average ~40 ppm; [101]). The matrix material of the lower MgO lavas analyzed here contains 48–52 ppm Sc, while that of the mixed magmas in unit THj have lower abundances (33–46 ppm; Table S3).

3.3. Sr Isotope Geochemistry of Plagioclase and Matrix

The ⁸⁷Sr/⁸⁶Sr values of Tungnárhraun plagioclase feldspar macrocrysts extend to values that are less radiogenic than those of their host basalts (Figure 9; Table S4). The samples from units THb and THdd with the lowest host magma MgO contents have matrix values between 0.703234 and 0.03250 and their corresponding plagioclase feldspars have ⁸⁷Sr/⁸⁶Sr values between 0.703085 and 0.703121. The MgO-rich lavas from units THde, THf, and THi have tightly clustered ⁸⁷Sr/⁸⁶Sr values for both plagioclase feldspars (0.703103–0.703138) and matrix material (0.703135–0.703160). The most radiogenic Sr isotope values are measured in the mixed magmas of unit THj: plagioclase feldspar ⁸⁷Sr/⁸⁶Sr values range from 0.703131 to 0.703211 and matrix materials from 0.703214 to 0.703333.

4. Discussion

4.1. Evolution of Tungnárhraun Eruptive Units

The goal of this work is to constrain the structure of the Bárðarbunga transcrustal magma storage system and infer how it evolved during the Holocene period. Our samples fill an important gap in the published data that enables the investigation of the full Tungnáhraun compositional spectrum: The magnesian basalts (7.5–7.9 wt.% MgO) are the only samples that overlap with the range of primitive glasses from the <100 ka subglacial Ljósufjöll system (7.5–9.4 wt.% MgO), also associated with Bárðarbunga volcano [8]. A direct comparison between these samples is not inappropriate in that the THi lavas are crystal-poor so bulk rock analyses can be considered reasonable approximations of liquid compositions. The bulk rock compositions for the THde and THf lavas corrected for a ~5–7 vol.% accumulation of An₈₈ plagioclase feldspar fall within this same range. Groundmass glass from the early Holocene (~8600 ybp) cinder cones Brandur, Fontur, and Saxi as well as the mid-Holocene Thjórsárdalshraun (~3000–4000 ybp) and Drekahraun (~3200 ybp) flows contain 5.6–7.5 wt.% MgO, thereby falling within the compositional gap observed between the mafic and more evolved Tungnáhraun lavas of our suite (Figure 8). Melt inclusions within plagioclase and olivine from Brandur, Fontur, and Saxi extend to even more MgO-rich compositions (~10 wt.%; [8]), indicating that highly magnesian liquids were present within the system but did not accumulate to form eruptible magmas.

The major element oxide variations (Figure 8) among the Tungnáhraun lavas and tephras preclude a unified origin through either the progressive fractionation of a common parental basalt or variable degrees of accumulation of the observed crystal phases olivine, plagioclase feldspar, and clinopyroxene (see also [8]). The heterogeneous TiO₂/K₂O values measured in our lavas and published glasses and melt inclusions require parental mafic liquids with internally variable but clearly defined compositional ranges (Figure 8). Glasses from the oldest Ljósufjöll samples and the mid-Holocene Thjórsárdalshraun and Drekahraun flows overlap our THi lavas (TiO₂/K₂O ~10–14, with a few Thjórsárdalshraun values ~20 and a handful of extremely K₂O-poor Drekahraun glasses extending to values approaching 90; [8]). In contrast, the THf and THde mafic lavas have lower TiO₂/K₂O values (~6–10) that overlap with glasses from the Early Holocene Brandur, Fontur, and Saxi cones and Thjórsá flow as well as the historical 1477 Veidivötn and 871 Vatnaöldur tephras [8,34,35]. The overall variation in this parameter decreases with decreasing MgO and thus does not correlate with eruptive age.

Tungnáhraun eruptives of all ages—from the <100 ka Ljósufjöll [8] through the 2014–2015 Holuhraun flow [5]—are characterized by CaO/Al₂O₃ values between 0.8 and 1.0, which is substantially higher than those expected for melts of peridotite [102] (Figure 8). This feature, coupled with the unusually high Sc contents of Tungnáhraun and Holuhraun mafic lavas, suggests the accumulation of abundant clinopyroxene. We note, however, that the lavas and tephras are at most sparsely clinopyroxene-phyric, and high CaO/Al₂O₃ values measured for melt inclusions within olivine and plagioclase feldspar [8] cannot be explained in this manner. As described below we attribute this feature to the resorption of early-formed clinopyroxene crystals during the establishment of the transcrustal magma system of Bárðabunga (see also [103]).

4.2. Origin of the Crystal Cargo

It is well established that the olivine, clinopyroxene, and especially the plagioclase feldspar macrocrysts in Tungnáhraun lavas and many global PUBs are xenocrystic in origin (e.g., [8,27,34,46,48,72,104]). Several lines of evidence support this interpretation:

• The macrocrysts are too primitive to have formed from their carrier melts. As noted previously [8] the most magnesian melt inclusions from the <100 ka Ljósufjöll tephras are appropriate in composition to have crystallized the most forsteritic olivine and high-Mg# clinopyroxene macrocrysts, and approach compositions with sufficiently high Mg# and Ca# (molar Ca/Ca + Na) to have been in equilibrium with the plagioclase feldspar cores.
• All basalts studied here—as well as those from older and younger Bárðarbunga eruptives—contain a population of plagioclase feldspar macrocrysts with core compositions An_85-91. In contrast, the macrocryst rims and groundmass crystal compositions reflect consistently the MgO content of the carrier lavas (Figure 3).
• Plagioclase feldspar macrocryst abundance does not correlate with carrier liquid composition: the volume of plagioclase increases from Early Holocene THb to Mid-Holocene units THdd, THde, and THf and then decreases and becomes markedly heterogeneous in the mixed lavas of units THi and THj (Section 3.1; Figure 2).
• The Sr isotope compositions of the host lavas and feldspar macrocrysts have distinct ranges (Figure 9). The plagioclase macrocrysts have consistently lower ⁸⁷Sr/⁸⁶Sr values than those measured in the glassy groundmass matrix of their host basalts; the more primitive lavas and their macrocrysts show narrower isotopic ranges and less separation between macrocrysts and glassy matrix material.

4.3. Timescales of Magmatic Processes in the Tungnárhraun Basalts

The geochemical features of the xenocrysts allow us to constrain the P–T conditions of magma storage (Section 4.4 below) and the timescales of melt transport from storage to eruption. Olivine diffusion modeling can reveal crystal–liquid equilibration processes acting over timescales from hours up to tens of years [105,106]. We modeled Fe–Mg interdiffusion on 14 olivine crystals with thin, normally zoned rims (Figure 5) to establish the time scales of diffusive relaxation and cooling, and to explore the changes in these parameters over the history of the Tungnárhraun sequence. The modeling was performed using a modified version of the finite difference method presented in [106]. The original Mathematica code was rewritten in R and tested with the data of [105] to ensure replication of results. The diffusion temperatures for individual flows (1137–1190 °C, Table S5) were taken from calculated clinopyroxene crystallization temperatures. Oxygen fugacity was calculated as per the original code [106]. The initial composition of the profile is taken as the average core composition, which was observed to be nearly constant for several hundred µm into the crystal (Figure 5). The boundary condition (i.e., final composition) was the lowest forsterite content measured at the crystal rim; these values match the groundmass olivine compositions from the respective host lavas in all profiles. The code was then applied to transects of olivine macrocrysts from the Tungnárhraun flows. The model employed a pressure of 2 kbar (the corresponding value for the clinopyroxene crystallization temperatures employed here), with diffusion on the c-axis to constrain the minimum timescale of diffusion in the absence of crystallographic information. We chose a time step of 60 s over a period of 3600 days (~10 years) and a grid spacing of 5 µm to satisfy the stability criterion for the finite difference method (dt/dx²*D < 0.5), where D is the diffusion coefficient (see Figure S1 for representative model profiles).

The Tungnárhraun olivine macrocrysts preserve a diverse suite of zoning profiles (Figure 5), some of which are not suitable for simple diffusion modeling and indicate a more complex mixing and transport history. We are mindful that picking only the cleanest profiles results in the omission of important information regarding magma mixing and recognize that our models are qualitative. Many of the olivine composition profiles record at least one magma recharge that partially equilibrated with the crystal cores prior to rim growth; these crystals have a higher forsterite zone (Fo_≥76) around an intermediate core (Fo₇₆), which diffused to the final equilibrium boundary condition during mixing and transport (Figure 5). Profiles showing this feature are from subhedral olivine found alongside highly resorbed plagioclase; both are evidence of reworking by fresh magma. In contrast, profiles with a classic error function shape (e.g., COB-2 macro 1; Figure 5) are found in euhedral olivine–plagioclase glomerocrysts which form the basis of our models.

Our modeling suggests a lower limit of equilibration timescales less than 300 days (Table S5), which corresponds well to those calculated previously for Bárðarbunga lavas and tephras: ~150–400 days for Early Holocene tephras and <100 days for the 1477 Veidivötn lava [20]. These results seem typical for Icelandic systems, as we note they are in good accordance with olivine mush remobilization time scales calculated for the 2021 Fagradalsfjall eruption [19]. While emphasizing the qualitative nature of our results, we note that the fastest time scales are those calculated for the most crystal-poor MgO-rich lava (unit THi), with ascent times less than one month. It is interesting that this period of rapid ascent of primitive basalts coincides with a regional episode of low eruptive volumes relative to those in the Early and Late Holocene [35]. This inference suggests that the flux of primitive magma has not decreased gradually across the period of Holocene deglaciation but rather that the crustal storage and transport processes have varied.

4.4. Evolution of the Tungnárhraun Transcrustal Magma System

4.4.1. Formation of Plagioclase Feldspar Macrocrysts

The compositional transects across the Tungnárhraun plagioclase macrocrysts reveal a complex distribution of Mg between their rims and cores (Figure 4) that is also observed in mid-ocean basalts and PUBs (e.g., [89,104]). Recent experimental studies have suggested two mechanisms that may operate independently or in concert to produce this feature. An evaluation of an extensive database of calculated mineral–melt partition coefficients from phase equilibria experiments and natural samples found that the partitioning of Mg into plagioclase feldspar is dependent upon the conditions of temperature and pressure in addition to the melt composition [89]. This mechanism is consistent with the distribution of Mg within the Tungnárhraun plagioclase feldspars, but does not provide additional insights into magma formation, transport, and storage.

Following [72] we recognize that this behavior may be driven by the thermodynamics of the plagioclase feldspar pseudobinary system which has been shown experimentally to contain an azeotrope (more correctly termed a pseudoazeotrope, as the natural system is not binary), or reversal, at plagioclase feldspar compositions of An_80-85, suggesting that a magma with a sufficiently high Ca# (defined as molar Ca/[Ca + Na]) that evolves or fractionates at 5–10+ kbar will produce plagioclase feldspars with progressively higher anorthite as the MgO content of the liquid decreases (Figure 10). Experiments on both normal and enriched mid-ocean ridge basalt starting compositions (N- and E-MORB) define this feature with liquids in equilibrium with plagioclase feldspar, olivine, clinopyroxene, and orthopyroxene at both 5 and 10 kbar [72]. Experimental work on lunar basalts [107,108] and mantle analog systems [109] find the same feature for multiply saturated liquids.

If the system is moved to lower pressures, e.g., through the rapid ascent of the melt from depths below ~5–10 kbar to <5 kbar within the crust, the plagioclase feldspar pseudoazeotrope will evolve towards a normal topology, leading to eutectic-like behavior in which the feldspar is stable but the mafic phases undergo resorption (Figure 10). Petrographically this change is manifested in the growth of large homogeneous plagioclase feldspars, the heat of fusion from which leads to the loss of olivine and pyroxene. Critically, the growing plagioclase feldspars will display the expected decrease in Mg with decreasing An content as seen towards the crystal rims (Figure 4).

The resorption of clinopyroxene can explain two significant and perplexing features of the Tungnárhraun lavas, tephras, and melt inclusions: unusually high CaO/Al₂O₃ values (Figure 8) and Sc contents in the basalts over 50 ppm in the absence of clinopyroxene crystals (Figure S2). Both olivine and clinopyroxene macrocryst populations within the lavas include crystals with resorption features, consistent with this interpretation. We emphasize that the samples have not accumulated clinopyroxene crystals, but rather record the effects of resorption at pressures less than ~5 kbar. These observations suggest both a common pathway of evolution for the plagioclase macrocrysts and a mechanism to develop the characteristic geochemical features of their carrier liquids. Furthermore, they suggest that the plagioclase macrocrysts and the compositionally diverse parental basalts are integrally related by the process of melt evolution involving the pseudoazeotrope. Finally, this finding enables us to constrain the pressure conditions of crystal core formation to between 5 and 10 kbar, or depths of ~14–30 km in the Icelandic crust.

4.4.2. Physical Evolution of the Bárðarbunga Volcanic System

Icelandic PUBs are thought to reflect the entrainment of large plagioclase crystals and subordinate plagioclase + olivine glomerocrysts from magma mush zones at mid- to deep-crustal levels beneath the volcanic systems [3,8,27,34]. A comprehensive model for the crustal storage and transport pathways for the Tungnárhraun basalts must explain three critical petrographic and geochemical observations: (1) Lavas contain ubiquitous and variable amounts of plagioclase feldspar (and subordinate olivine ± clinopyroxene) macrocrysts that did not grow in their carrier magmas. (2) Geochemical variations in erupted lavas and plagioclase feldspar macrocrysts require the involvement of multiple parental melts. (3) Plagioclase feldspar macrocrysts have slightly but consistently less radiogenic Sr isotopic values than their carrier melts.

The thermobarometric investigation of Tungnárhraun eruptives provides a picture of magma storage and crystallization conditions within the Bárðarbunga volcanic system. The individual clinopyroxene microcrysts from all units studied here record equilibrium with their host liquids (Figure S3) and give crystallization pressures of ~1.6–2.1 kbar at temperatures between 1170 and 1230 °C (note that crystals that fail our equilibrium criteria yield overlapping P–T conditions) (Table S5); sparse macrocrysts from unit THb indicate a crystallization pressure of ~2.3 ± 0.7 kbar at comparable temperatures (equations of [83,86]).

Our new results align well with the pyroxene thermobarometry on samples from <100 ka through historic lavas (average 2.2 ± 0.7 kbar, ~1170–1205 °C [8]) and on samples from the 2014–2015 Holuhraun event (~2.3 ± 1.4 kbar, ~1170 °C [5]). The pyroxene thermobarometry results (Figure S4) are consistent with published equilibration conditions calculated from olivine-hosted melt inclusions that record olivine–plagioclase–augite–melt (OPAM) equilibrium in all studied Bárðarbunga lavas and tephras (1.9 ± 0.8 kbar, [8]; 2.6 ± 0.4 kbar, [35]; 1125–1225 °C and ~1–3 kbar [110,111]). Mafic macrocrysts typically record higher pressures and temperatures of melt entrapment (~3.1–4.4 kbar [8,35]), extending to ~1225–1250 °C and ~5 kbar [111].

The plagioclase feldspar macrocryst compositions record growth at higher temperatures than those represented by the clinopyroxene microcrysts (Table S5). The crystallization temperatures for An_85-90 plagioclase feldspar macrocryst cores (calculated following [87]) yield consistent temperatures between 1277 and 1367 °C throughout Tungnárhraun. Macrocryst rims and groundmass feldspars show greater variability: samples with <7 wt.% MgO in bulk lava have macrocryst rims that yield temperatures from 1014 to 1220 °C, whereas the rims on plagioclase feldspars in the more primitive samples yield generally hotter temperatures between 1157 and 1318 °C.

Using a crustal density of 2.86 g/cm³ [112], thermobarometry indicates that most Tungnárhraun lavas equilibrated prior to eruption in the middle crust at one or more depths between ~5 and 8 km. The larger value is consistent with regional geodetic observations from the Holuhraun event suggesting a deflating source at 8–12 km beneath Bárðarbunga central volcano [4] as well as the depth to the associated seismic swarm [113,114]. In contrast, the Early Holocene tephras and olivine, clinopyroxene, and plagioclase feldspar macrocrysts of all ages record a period of crystallization at greater depth, loosely constrained between 9 and 30 km.

These results lead us to propose a transcrustal magma system for Tungnárhraun as shown schematically in Figure 11. While greater complexity is certainly possible and even likely (e.g., [6]) we find that this simple representation explains the petrological and geochemical characteristics of the lavas and their crystal cargo.

Stage 1: Beginning at least in the Early Holocene, the crust beneath Bárðarbunga contains storage region(s) located between ~5 and 10 kbar (~14 and 30 km depth). This lower-crustal region is fed by primitive, low ⁸⁷Sr/⁸⁶Sr melts from the Icelandic sublithosphere that have diverse minor element, trace element, and isotopic characteristics. The experimentally determined plagioclase feldspar pseudoazeotrope restricts the depth of this storage zone to depths between ~14 and 30 km, consistent with the storage depth of ~17.5 km determined from primitive melt inclusions in olivine and plagioclase feldspar crystals [8]. This depth is roughly equivalent to the crust–mantle boundary beneath the Eastern Volcanic Zone [115], suggesting melts pond at this level. The melts crystallize primitive plagioclase feldspar, olivine, and pyroxene, as recorded in diverse melt inclusions [8]. These multiply saturated liquids evolve along the plagioclase pseudoazeotrope, generating the observed pattern of Mg enrichment in plagioclase.

Stage 2: The melts—carrying their crystalline cargo—ascend rapidly from ≥5 kbar to shallower levels in the crust. Thermobarometric evidence from glasses, melt inclusions, and mineral analyses suggest this second storage depth is roughly 3.2 kb or 8–9 km. This depth is compatible with regional observations of magma storage based on geophysical, geodetic, and microseismic observations at Holuhraun before and subsequent to the 2014–2015 eruption [4,113,114,116,117,118]. The dimensions of the Stage 2 chamber are not well constrained; we estimate a volume on the order of a few cubic kilometers based on caldera subsidence and the eruptive volume of the Holuhraun event [110,116,117,119].

The Stage 2 magma chamber is critical to the development of the Bárðarbunga and Tungnárhraun PUB lavas. At this depth the pseudoazeotrope no longer exists, and plagioclase feldspar An content and melt MgO evolution are controlled by the more familiar basaltic phase constraints. Plagioclase feldspar continues to grow, releasing heat that causes olivine and pyroxene to be resorbed into the melt. This step gives the PUBs their characteristic high CaO/Al₂O₃ and Sc character, which cannot be attained without the resorption of clinopyroxene. It is significant that no new carrier melt needs to be introduced into the system in order to produce the resorption features seen in mafic phases or the xenocrystic nature of the plagioclase feldspars. The most prevalent liquid at this stage likely contains ~8 wt.% MgO.

The Stage 2 storage chamber may be tapped shortly after the resorption of the clinopyroxene, which we infer results in eruptive units such as THi with ~8 wt.% MgO and a small cargo of plagioclase macrocrysts. Convective overturn within the chamber that may be triggered by the injection of replenishing mafic liquids could result in eruptives such as THde and THf, which have ~8 wt.% MgO but more abundant plagioclase which were mobilized from a crystal mush.

Stage 3: If not erupted directly, melts—bearing variable amounts of plagioclase macrocrysts—may ascend to a yet shallower storage chamber where they crystallize equilibrium olivine, clinopyroxene, and plagioclase feldspar as they evolve to lower MgO contents. The depth of this storage region is constrained by clinopyroxene thermobarometry to have been consistently stable at ~5 km since at least the Middle Holocene (e.g., [8]). Prolonged residence in the shallow Icelandic crust permits the assimilation of somewhat more radiogenic and hydrothermally altered materials (e.g., [8,19,105]) that results in extensive oxygen isotope exchange and gives the carrier magmas a more radiogenic Sr isotope signature than the plagioclase macrocrysts which are slow to undergo diffusive equilibration. The accumulation of multiple magmas at this depth leads to the observed reduction in minor element heterogeneity (e.g., TiO₂/K₂O) in the mafic lavas, while extensive storage results in the generation of more evolved melts through fractionation.

Note that even these lower-MgO carrier liquids retain the signature of pyroxene resorption: high CaO/Al₂O₃ values and Sc contents. We emphasize that the ascent rate to develop PUB lavas needs only to have an upward velocity greater than or equal to the settling velocity of the plagioclase in order to counter their negative buoyancy in basalt (0.5–1 cm/s) and carry them to the surface [28,46,48], because the denser mafic phases have disappeared. In this scenario, magma ascent from 18 km (6 kbar) at 1.0 cm/s would take ~20 days. These values are consistent with the lower range of eruptive time scales calculated by Mg–Fe diffusion modeling of olivine macrocrysts (Table S6). The olivine crystals that preserve diffusive profiles likely were entrained at the same time as the plagioclase, as they commonly form glomerocrysts and are zoned from highly forsteritic (Fo_82-86) cores to equilibrium compositions at the rim (Fo_<75) that match those of the groundmass crystals.

The lavas from the Late Holocene unit THj that reflect mixing between Tungnárhraun magmas and basalts from the neighboring Torfajökull complex provide additional insight into the scale of magmatic processes in southern Iceland. These lavas are characterized by higher silica and alkali contents (Figure 7) than the Tungnárhraun lavas, and by plagioclase feldspar macrocrysts and groundmass that overlap with the low-MgO lavas reported here. In contrast, their clinopyroxene macrocrysts have substantially lower Al₂O₃/TiO₂ values and Mg#s than those of low-MgO Tungnárhraun flows while the groundmass clinopyroxenes have higher and more variable Al₂O₃/TiO₂ values than Tungnárhraun lavas of comparable MgO content (Figure 6). These characteristics suggest that the mixed lavas contain plagioclase feldspar macrocrysts which originated in a Stage 1 storage region but were subsequently carried to the surface by distinct melts from the Torfajökull system. There are two feasible interpretations: (1) Stage 1 chambers are ubiquitous in southern Iceland and form a fundamental part of the magmatic architecture. (2) Stage 1 chambers are characteristic only of the Bárðarbunga volcanic system and the mixing between Tungnárhraun and Torfajökull magmas takes place at very shallow crustal levels. Our data do not allow us to distinguish between these two possibilities.

The petrological and geochemical features described herein are not unique to the Tungnárhraun flows and require our model to be generalizable across the Eastern Volcanic Zone and broadly throughout Iceland. It is estimated that basaltic lavas with ~25% modal plagioclase feldspar (± minor olivine and/or clinopyroxene) comprise 12–15% of the Tertiary lava piles in the eastern fjords [31,73,120] and up to 29% of the lavas in the northwestern region [121,122] in addition to their frequent occurrence in Holocene lavas (e.g., [27,34]. The Tertiary Grænavatn Group in eastern Iceland was suggested [31] to be underlain by either an isolated mush zone around a pipe-like conduit, or an elongated mush region beneath most or all the volcanic lineament. Our model would obviate the need for a widespread mush zone, requiring only a finite magma chamber in the mid- to lower crust where conditions support the development of large plagioclase feldspars at the expense of olivine and pyroxene.

The majority of Holocene Icelandic PUBs are products of the Bárðarbunga system, e.g., the Kinnarhraun and Bárðardalshraun flows of the Bárðardalur Valley and Gígöldur on the northern side [32]. The chemistry and ⁸⁷Sr/⁸⁶Sr ratios of plagioclase macrocrysts from these flows overlap with values measured in Early Holocene lavas [34]. These Bárðardalur lavas, like the Tungnáhraun, include more primitive (8–9 wt.% MgO) and less primitive (7–8 wt.% MgO) mafic lavas [123]. It is significant that the MgO-rich lavas of Kviahraun and Dyngjuhals have CaO/Al₂O₃ values between 0.8 and 1.0 and contain 40–51 ppm Sc. These geochemical observations indicate that the magmatic processes we infer for the southern Bárðadbunga fissures operated in the northern portion of this system as well.

4.4.3. Temporal Evolution of the Bárðarbunga Volcanic System

The processes to which we attribute the geochemical and petrographic features of the Holocene Tungnárhraun lavas are a consistent feature of the volcanic system throughout its history. With the availability of these melts widespread in space and time, the question becomes one of eruptability and the interplay between magma supply and crustal extension. We suggest that the low-MgO mafic lavas are fed from Stage 3 magma chambers, with eruption triggered by an influx of mafic magma from storage areas deeper in the crust. Plagioclase feldspar macrocrysts in these lavas are often complexly zoned with cores, mantles, and rims of strikingly different compositions. In contrast, the high-MgO mafic lavas erupt directly from Stage 2 magma chambers; their crystal cargo does not indicate residence in the shallow crust. We suggest that these eruptions are triggered by extension rather than by magmatic overpressure. The most MgO-rich Tungnárhraun lavas (units THde, THf, and THi) erupted during the mid-Holocene eruptive low [35] between roughly 2000 and 5000 ybp. The first two flows contain abundant plagioclase feldspar macrocrysts, but the third is sparsely phyric. Similarly, the amount of plagioclase in the Bárðardalur lavas declines as volcanic output began to slow ~4.0 ka [109]. Around this time, both lava sequences have a ~1 km³ transition lava, THi in the Tungnárhraun and Kvíahraun in the Bárðardalur, with high MgO but notably low volumes of plagioclase (<10%).

The crustal structure we propose is uncomplicated and yet accounts for all of the features observed in the eruptive products (lavas, tephras, and melt inclusions) and their crystalline cargo (olivine, clinopyroxene, and plagioclase feldspar) across the entire history of Bárðabunga. It provides a mechanism for diverse parental melts to be trapped within individual melt inclusions prior to becoming homogenized during Stage 1 storage. It also renders extensive mush zones that cut across the region unnecessary, as melt remobilization of crystal cumulates by subsequent magmas is not required. Finally, it avoids the challenge of requiring moderate-MgO melts to sample primitive crystal mushes from below.

This pattern suggests that the entrainment of abundant plagioclase occurs preferentially in primitive magmas with low crustal residence times and hence rapid transport, and also that the crystal mush zone may become depleted in plagioclase feldspar macrocrysts at times of low magma flux. Detailed analysis [7,124] found a relationship between magma production and storage depth: with increased production the final depth of storage is shallow crust, while with low magma production the basaltic magma is stored in the lower crust or at the CMB. This result is consistent with our results and proposed crustal structure, although the trigger for eruption during times of low magma flux remains unclear. We conclude that during the Mid-Holocene eruptive low, the pathway of magma ascent rerouted away from the mush zone, either as a result of the changing stress field after deglaciation or in response to a rifting episode unrelated to magma supply (e.g., [13,14,27]).

5. Conclusions

Early through Mid-Holocene Tungnárhraun basalts with ~5.5–6 wt.% MgO and ~8 wt.% often contain up to 20 vol.% plagioclase feldspar macrocrysts with sparse olivine ± clinopyroxene that represent xenocrystic cargo in their carrier lavas. The plagioclase macrocrysts are consistently too An-rich and the mafic phases have higher Mg#s than crystals grown under equilibrium conditions in their host lavas, and the ⁸⁷Sr/⁸⁶Sr values of plagioclase macrocrysts and host basalts define distinct populations. Multiple parental melts are also indicated by variations in Ti/K and Al/Ti in high-An plagioclase feldspar macrocryst cores.

Mineral–melt thermobarometry indicates a transcrustal structure and transport system with at least three distinct storage regions. A lower-crustal mush zone at the base of the Icelandic crust (~14–30 km) is fed by primitive, low ⁸⁷Sr/⁸⁶Sr magmas with diverse trace element signatures. At this depth, an experimentally determined plagioclase feldspar pseudoazeotrope causes the melt to grow crystals in which the An and Mg contents are correlated inversely. The rapid ascent of magmas to mid-crustal levels (~8–9 km) removes the pseudoazeotrope and allows the thermodynamics of the feldspar system to revert to well-known phase constraints. Continued plagioclase feldspar growth releases heat, causing olivine and pyroxene to be resorbed into the melt. This step gives the lavas their characteristic high CaO/Al₂O₃ values (~0.8–1.0) and Sc contents (up to 52 ppm), which cannot be attained without the resorption of clinopyroxene. Mid-Holocene MgO-rich lavas with abundant plagioclase feldspar macrocrysts erupted directly from this depth, but older and younger magmas ascended to a shallow-crustal storage chamber (~5 km) where they crystallize olivine, clinopyroxene, and plagioclase feldspar as they evolve to lower MgO contents. Minor crustal assimilation during this stage gives the carrier magmas more radiogenic ⁸⁷Sr/⁸⁶Sr values than their plagioclase feldspar cargo.

This model predicts the prolonged growth of feldspar macrocrysts during the resorption of mafic phases and does not require the physical disruption of an established and long-lived gabbroic cumulate in the Icelandic crust. The processes and transcrustal structures documented here occurred in south Iceland at least throughout the Holocene and are likely to have taken place throughout the history of Icelandic magmatism. Periods of low magma flux, i.e., in the mid-Holocene, are characterized by the eruption of more primitive basalts from the mid-crust, suggesting that eruptions were triggered by physical rifting or an isostatic response to deglaciation rather than requiring an influx of melt from below.