Mechanisms driving polymagmatic activity at a monogenetic volcano, Udo, Jeju Island, South Korea
High-resolution, stratigraphically ordered samples of the Udo tuff cone and lava shield offshore of Jeju Island, South Korea, show complex geochemical variation in the basaltic magmas that fed the eruption sequence. The eruption began explosively, producing phreatomagmatic deposits with relatively evolved alkali magma. The magma became more primitive over the course of the eruption, but the last magma to be explosively erupted had shifted back to a relatively evolved composition. A separate sub-alkali magma batch was subsequently effusively erupted to form a lava shield. Absence of weathering and only minor reworking between the tuff and overlying lava implies that there was no significant time break between the eruptions of the two magma batches. Modelling of the alkali magma suggests that it was generated from a parent melt in garnet peridotite at c. 3 to 3.5 GPa and underwent mainly clinopyroxene + olivine ± spinel fractionation at c. 1.5 to 2 GPa. The sub-alkali magma was, by contrast, generated from a chemically different peridotite with residual garnet at c. 2.5 GPa and evolved through olivine fractionation at a shallower level compared to its alkali contemporary. The continuous chemostratigraphic trend in the tuff cone, from relatively evolved to primitive and back to evolved, is interpreted to have resulted from a magma batch having risen through a single dyke and erupted the batch’s head, core and margins, respectively. The alkali magma acted as a path-opener for the sub-alkali magma. The occurrence of the two distinct batches suggests that different magmatic systems in the Jeju Island Volcanic Field have interacted throughout its history. The polymagmatic nature of this monogenetic eruption has important implications for hazard forecasting and for our understanding of basaltic field volcanism.
KeywordsMonogenetic volcanism Basalt geochemistry Jeju Island Plumbing system Alkali basalt
Basaltic volcanic fields are typically dominated by monogenetic volcanoes that have a lifespan of months to decades and that record spatially and temporally dispersed volcanism (Walker 1993). Such systems may display widely differing magma flux and eruption frequencies (Valentine and Perry 2007). Individual monogenetic (sensu stricto) volcanoes within fields are regarded as being geochemically and volcanologically simple, at least in comparison with long-lived polygenetic centres where greater degrees of magma evolution are expected.
Detailed chemical investigation into individual monogenetic volcanoes can yield insights into generation of magma in the mantle and the processes affecting the magma during ascent. The typically low magma volumes reach the surface with little fractionation or interaction with the crust (Blondes et al. 2008; Németh et al. 2003; Reiners 2002; Smith et al. 2008; Strong and Wolff 2003; Valentine and Perry 2007) due to relatively fast ascent rates (Demouchy et al. 2006; Spera 1984) and over relatively short periods of time, hence allowing small-scale investigations. By sampling eruptive sequences in great detail, it is possible to observe magmatic evolution over the short period of a monogenetic eruption, such as in the Auckland Volcanic Field, New Zealand (Smith et al. 2008), in the Big Pine Volcanic Field, California (Blondes et al. 2008) and at Parícutin and Jorullo volcanoes, Mexico (Erlund et al. 2009; Luhr 2001; Luhr and Carmichael 1985; McBirney et al. 1987). These studies have shown that single eruptive centres can be fed by magma batches that display varying degrees of chemical evolution resulting from processes within the source and plumbing/conduit system. Conversely, other centres do not display significant chemical variability during subsequent eruption stages despite varying eruption styles, such as Lathorp Wells, Nevada (Valentine et al. 2007). At Crater Flat, Nevada, separate cones and lava fields with relatively constant chemical composition and originally interpreted as derived from polygenetic activity (Bradshaw and Smith 1994) were subsequently re-interpreted to have erupted over relatively short time spans based on field relationships (Valentine et al. 2006).
Other types of magmatic variation in monogenetic volcanoes have been noted by Strong and Wolff (2003), who in the southern Cascade arc found that the chemical composition of scoria cones may differ from their related and subsequently erupted lava shields and hence related the variation to different source compositions. Chemical variability in other intraplate basaltic centres (Reiners 2002) has been attributed to mixing of different magmas derived from distinct sources. Mixing of similar magmas at various stages of evolution may also give rise to chemical variability. For example, bimodal magmatism in the Waipiata Volcanic Field, New Zealand, and the Miocene–Pliocene volcanic fields in the Pannonian Basin, western Hungary, was proposed to be controlled by fractionation and mixing processes (Németh et al. 2003). In these cases, initial eruption of tephrite was interpreted to have derived from stalling and fractionation of early-formed basanitic magma, which subsequently intersected and mixed with ascending basanite that initiated eruption of the fractionated magma. Chemical variability of monogenetic centres in the trans-Mexican Volcanic Belt instead suggests a greater involvement of the continental crust and could be linked to the early-stage evolution of composite volcano growth (Siebe et al. 2004).
These studies provide evidence for chemical complexities in monogenetic (sensu lato) volcanoes, given that eruptions may be fed by magma batches derived from different sources. They also suggest that a detailed investigation into an individual volcano within a monogenetic field can provide high-resolution information on the nature of magma generation, evolution and mixing processes that characterize the behaviour in the field as a whole. If more than one magma type is involved, it is moreover important to determine how these may interact and behave at depth, which will ultimately control the resulting eruption and final volcanic landform. Involvement of different magmas erupted from a single vent over a relatively short time interval may have repercussions on the way we perceive volcanic hazards, given that the physical behaviour of the volcano may change through the course of the eruption.
In this paper, we present a high-resolution, stratigraphically controlled geochemical investigation into the Udo tuff cone and lava shield, Jeju Island, South Korea, which is a monogenetic volcano displaying a complex chemical behaviour. This reveals insights into magma generation in the c. 8,000 km2 Jeju Island Volcanic Field, which comprises over 300 distributed monogenetic centres, plateau lavas and a concurrent central composite volcano.
Geological setting and field relationships
Udo Volcano is an example of a single vent system within the Jeju Island Volcanic Field. Udo Volcano forms an island, c. 3 × 4 km across, c. 3 km off the east coast of Jeju, consisting of a tuff cone, spatter mound and a lava shield (Fig. 1), allowing the investigation into different eruptive products unambiguously derived from a single vent. It was targeted for detailed study because the tuff cone is well exposed by wave action, making it possible to sample material representing the majority of the eruption sequence. Udo Volcano has a single main vent situated in the south-eastern part of the island (Fig. 1). A horseshoe-shaped tuff cone open towards the north-west contains a nested, rounded spatter mound and is partially filled by ponded lava. A series of flows form a lava shield to the north-west of the vent. Dating of the lava shield rocks gave a K–Ar age of 114 ± 3 ka (Koh et al. 2005), whereas dates of core samples gave 40Ar/39Ar ages of 102 ± 69 ka for a tholeiitic andesite lava flow at an elevation of 25.5 m a.s.l. and 86 ± 10 ka for an alkali basalt 7.5 m a.s.l. part of the spatter mound or associated lava (cf. cross section in Fig. 1; Koh et al. 2008). An apron of reworked tephra from the tuff cone overlies the lava shield to the north-east of the tuff cone and inside it (Fig. 1). An estimate of the eruptive volumes, taking into account the height and topographic surface of the island, is c. 0.06 km3 for the tuff stage and c. 0.65 km3 for the lava shield stage. Note that these are minimum volume estimates, as they do not take into account tephra dispersed in eruption plumes and magma frozen in the plumbing system. Despite the imprecision inherent in defining the actual thickness of all units, it is fair to assume that the volume of the lava shield stage was one order of magnitude larger than that of the tuff cone stage of this eruption.
Older basaltic lava flows from the Jeju plateau lava stage (c. 100 m thick), unconsolidated silty sediments (c. 150 m thick), as well as felsic ignimbrite and lava units underlie the volcanic complex (Sohn and Chough 1993). A depth of c. 23 m below sea level to the basal contact of the tuff in drill core (Koh et al. 2008) is probably an overestimate given that the core was taken within the tuff cone crater and hence may comprise part of the diatreme. A depth to basement of c. 15 m below sea level is more consistent with water depth around Udo Island.
The tuff cone generally comprises steeply inclined (20–30°) beds of lapilli tuff and tuff that dip radially away from the vent. A detailed sedimentological study of the tuff cone reveals that it has formed by a Surtseyan-type eruption, which became drier towards the end of the eruption (Sohn and Chough 1993). The deposition was mostly accomplished by grain flows of lapilli and blocks in addition to airfall of finer-grained tephra. Absence of marine-reworked deposits suggests that the majority of the tuff cone was constructed subaerially, although the submerged part may have formed underwater. Common inclusion of acidic volcanic rock fragments (rhyolite and welded tuff) that were most likely derived from the Cretaceous volcanic basement rocks in the eastern Jeju area suggests that the level of hydrovolcanic explosions and the depth of country-rock excavation reached more than 300 m below the present sea level (Sohn 1996).
The contact between the LTC and the upper tuff cone (UTC) is overall conformable and difficult to identify along the whole tuff cone exposures. The contact is conspicuous only at one locality where the top of the LTC consists of a chaotic deposit of contorted and brecciated strata, suggesting slumping of steep tuff cone deposits by gravitational instability (Sohn and Chough 1993) and is overlain by the well-bedded deposits of the UTC (Fig. 2d). Except for this locality, LTC and UTC can be distinguished only by subtle contrasts in deposit characteristics, the former consisting mainly of planar-bedded lapilli tuff, whereas the latter consisting of the alternations of thin-bedded tuff and discontinuous lapilli layers (Sohn and Chough 1993). A significant break in time or in eruptive activity is hence not inferred to have occurred between LTC and UTC either.
At the eastern end of the tuff cone along the beach (Figs. 1, 2d, f), the stratigraphic relationship between the tuff cone deposits and the shield-forming lavas could be identified. The tuff cone deposits here are overlain by reworked volcaniclastic deposits with near-horizontal bed attitudes (Fig. 2d). The reworked deposits are massive to cross-stratified and scour-and-fill bedded, suggesting emplacement by debris flows and stream or rill flows after the cessation of the tuff cone–forming eruption (Sohn and Chough 1992). Such reworking processes (surface run-off and gully formation) are known to begin almost simultaneously with the cessation of eruption and prior to consolidation of the pyroclastic deposits and, in some cases, during the eruption in tropical or rainy regions (Ferrucci et al. 2005; Németh and Cronin 2007). The lava flow here is intercalated between the primary tuff cone deposits and the overlying reworked volcaniclastic deposits. Only a minor part of the reworked volcaniclastic deposits are sandwiched between the tuff cone deposits and the lava flow in a wedge-form (Fig. 2d, f). This stratigraphic relationship suggests that the lava here was emplaced within days to weeks after the end of the tuff cone–forming eruption. Intra-crater, ponded lava flows are visible on the eroded western side of the tuff cone (Fig. 2e). The whole volcanic deposits of the Udo volcano can thus be considered monogenetic in that it resulted from a small-volume, short-lived single eruption without a significant break in eruptive activity.
The term monogenetic has long been used for eruptions generating scoria cones, pyroclastic cones and small lava shields (Wood 1979 and references therein). The final products of such eruptions (Valentine and Perry 2007; Verwoerd and Chevallier 1987; White 1991) are comparable in geometry and volume to the products of the Udo eruption, which can hence be considered monogenetic. These are relatively short-lived, small-volume eruptions; however, a clear definition of the chemical meaning of “monogenetic” is still lacking, and as will be discussed later, chemical variability may be complex.
Continuous, stratigraphically controlled sampling could be carried out only at the eastern margin of the tuff cone in outward-dipping flank deposits, where the accessible outcrops of the upper part of the LTC and whole UTC are available (Fig. 2c, d). Samples were collected at regular intervals of generally 1 to 3 m from lapilli- or bomb-rich beds (Fig. 2). A description of bed characteristics and interpreted emplacement mechanism for each sampled bed is reported in Table S2 of the supplementary data file. Juvenile material was selected in the field based on surface characteristics and density. Coarse lapilli and bombs were selected having cauliflower texture rather than subrounded or angular edges, and intensely mud-coated lapilli were removed. Basaltic clasts with black glassy groundmass and microvesicularity were preferred, which made them less dense compared to crystalline lava clasts (which also generally have well-formed macro vesicles). For fine lapilli layers, further discrimination and hand picking were carried out in the laboratory after sample cleaning. Samples were collected from the lava shield at different localities (Fig. 1); the coordinates are reported in Table S1 of the supplementary data file.
Rock textures and mineralogy were observed in thin sections, and mineral identifications were confirmed using an electron microprobe. The petrography of the sample suite from Udo volcano varies throughout the succession. In the lower tuff sequence, the mineral assemblage is dominated by olivine phenocrysts (3-5%) and plagioclase microphenocrysts (7-8%) in a glassy groundmass. Vesicles (25-40%) are rounded and spherical to elliptical with minor coalescence. Weak flow banding, where present, is defined by plagioclase and elliptical vesicle alignment. Olivine crystals show skeletal form, indicative of rapid growth, and some occur as polycrystalline aggregates. Olivine also occurs with intergrown chrome spinel indicating a xenocrystic origin. Microprobe data indicate that the olivine crystal cores have MgO/FeO ratios similar for most of the olivine in tuff stage samples (ratios of c. 2.5 with some cores >3), whereas rims have MgO/FeO < 2. Plagioclase needles have swallow tails form indicating rapid growth.
The phenocryst assemblage in the upper tuff sequence consists of olivine (3-4%), clinopyroxene (<2%) and orthopyroxene (<1%). Plagioclase occurs as microphenocrysts (1-5%) in a glassy groundmass and is rarest in the middle to upper part of the sequence. Vesicles (20-50%) are subrounded to subangular, subspherical and show variable degrees of coalescence. Clinopyroxene occurs as glomerocrysts and as overgrowths surrounding orthopyroxene. It shows twinning and oscillatory and sector zoning. Less often it exhibits greenish cores with colourless overgrowths. Clinopyroxene also forms rims around resorbed and sieve-textured cores. Olivine is skeletal. Orthopyroxene is rare and not observed in all samples.
The lava shield mineral assemblage consists of variable amounts of olivine (<5%), Ca-rich and Ca-poor clinopyroxenes (<1%), orthopyroxene (<2%) and plagioclase (<2%) phenocrysts to microphenocrysts in a plagioclase/clinopyroxene/oxide intergranular groundmass (cf. Koh et al. 2005). Olivine shows skeletal growth and weak alteration to orange iddingsite. Orthopyroxene has good cleavage, for which it can easily be distinguished from clinopyroxene and which makes it similar to orthopyroxene from mantle xenoliths observed in basalts from other eruptive centres on Jeju and described elsewhere (e.g. Yang 2004). Clinopyroxene crystals show lamellar twinning and also form thin rims around olivine in some cases. Plagioclase phenocrysts are normally zoned and groundmass plagioclase typically shows albite twinning. Minor (<5%) vesicles occur and are generally diktytaxitic (having angular shapes defined by crystal faces) due to the surrounding holocrystalline texture. Fe–Ti oxides in the groundmass occur as dendritic needles. Koh et al. (2005) found that in the lava shield, plagioclase in both groundmass and microphenocrysts consists of labradorite (An70-50) with rims generally less anorthitic compared to the core. Their olivine compositions were Fo80-77, and the opaque phases are mainly ilmenite.
Whole-rock chemical compositions
Mineral analyses were by electron microprobe. The instrument used was a JEOL JX-5A using a LINK systems LZ5 detector, QX-2000 pulse processor and ZAF-4/FLS matrix correction software. Standard operating conditions were an accelerating voltage of 15 kV, beam current of 0.5 nA, beam diameter of 5 μm and a live count time of 100 s for all mineral analyses. Analytical precision was estimated by replicate analyses of mineral standards as (σ) ≤ 3% for elements present in abundances >1% wt.
For the whole-rock analytical work, clean rock fragments were crushed between tungsten carbide plates and a 100-g aliquot ground to <200 mesh in a tungsten carbide ring grinder. Major and trace element concentrations were measured by X-ray fluorescence (Siemens SRS3000 spectrometer) using standard techniques on glass fusion discs prepared with SPECTRACHEM 12–22 flux. For the trace elements, a suite of 36 international standards were used for calibration, and Siemens SPECTRA 3000 software was used for data reduction. The Compton scatter of X-ray tube line RhKb1 was used to correct for mass attenuation, and appropriate corrections were used for those elements analysed at energies below the Fe absorption edge. One-sigma relative error for V is 1-3%, and detection limit is 2-5 ppm. Trace elements (apart for V) were analysed by LA-ICP-MS at the Research School of Earth Sciences, Australian National University, using an Excimer LPX120 laser and Agilent 7500 series mass spectrometer. For this work, the same fused glass discs as for XRF were used. Detection limits are <1 ppb and analytical errors <1% relative.
Representative chemical data of Udo samples
The majority of samples from the stratigraphically lowest series (LTC, U1-1 to U1-16; Fig. 4, Table 1) have a restricted MgO range between 8 and 7 wt% and do not show any systematic trends with position in the sequence (Fig. 3). They show narrow abundance ranges for all major elements except for Na2O, which has wide variability. The clustering on variation diagrams is good, apart for samples U1-5 and U1-14 which have lower MgO and samples U1-8 and U1-15 that have characteristics intermediate between the three groups (see particularly the K2O, TiO2 and P2O5 v MgO graphs in Fig. 4).
Samples from the stratigraphically intermediate series (UTC, U1-17 to U1-32 and U1-40; Fig. 4, Table 1) are relatively primitive compared to the LTC, with MgO ranging between c. 10 and c. 9 wt%. An evolutionary trend is present within the cluster from lower MgO (U1-17, 8.95 wt%) to higher MgO (U1-23, 10.05 wt%) and back to lower MgO (U1-32, 9.00 wt%) upwards through the sequence (Fig. 3). This trend is mirrored in increasing and then decreasing CaO, decreasing and then increasing Al2O3, SiO2, and to a lesser extent K2O, TiO2 and P2O5. Fe2O3 remains constant, whereas Na2O is scattered and shows only weak negative correlation with MgO. The spatter mound (U1-40) has 9.47 wt% MgO and is hence slightly more primitive than the stratigraphically youngest samples of the tuffaceous sequence.
The stratigraphically younger series (LS, U1-33 to U1-39 and U1-41 and U1-42; Fig. 4, Table 1) has a MgO range from c. 7.5 to c. 6 wt%. Major element trends are of increasing SiO2, K2O, Na2O and TiO2 (slightly) and decreasing Fe2O3 with decreasing MgO; CaO and P2O5 remain constant. Only a single gap occurs in the data between c. 6.2 and c. 6.7 wt% MgO (in both our data and Koh et al. (2005) data). This gap is not correlated with the geographical distribution of samples.
Based on CIPW normative criteria, the LTC and UTC stage samples are nepheline to olivine normative alkali basalts, whereas the LS stage samples are quartz-normative tholeiites.
The stratigraphic variation through the Udo eruptive sequence is well illustrated in the plots of normalized concentration against relative stratigraphic position (Fig. 3). The constant chemistry of the LTC contrasts with the variability of the UTC compositions and is markedly different from the LS stage compositions.
Interpretation of chemistry
The magmas forming the tuff cone stage and the lava shield stage cannot be related through fractionation of common mineral phases in basaltic magmas. No model using the phenocryst phases observed in the rocks can reproduce a c. threefold depletion in K2O and P2O5 from the tuff to the lava shield stages (Table 1, Fig. 4). For instance, MgO depletion and SiO2 enrichment may be a result of olivine fractionation; however, this would cause enrichment of K2O and P2O5 rather than the observed depletion. Conversely, crustal contamination is also not a viable model to explain the SiO2 enrichment from the tuff cone stage to the lava shield stage, because this would also result in K2O and P2O5 enrichment. Modelling of crystal fractionation within each stage, using observed or reasonable phenocryst phases for each of the two magmatic stages separately, is presented in the following discussion. Discussion of the source characteristics that may give rise to the observed chemical difference is in the following section.
Tuff cone stage
The chemical variability in the UTC cannot be the result of the mixing of two different primary magmas because of the continuity of the trend of depletion, followed by enrichment, of incompatible trace elements (Fig. 3). It also cannot be the result of mixing between the tuff primary magma and the lava shield primary magma; this is clear because the evolutionary trend from the most primitive sample of the tuff cone stage (U1-23) does not extend towards the LS stage composition. Trace element trends, such as decreasing Ni, Cr, V, Sc with increasing magmatic evolution, suggest fractionation of olivine + clinopyroxene ± spinel ± orthopyroxene. Spinel appears a more likely fractionating aluminous phase than plagioclase, because Sr is enriched with evolution in this suite. Orthopyroxene is present only in trace amounts as a phenocryst phase in the UTC samples and is generally rimmed by clinopyroxene, suggesting that it was not an equilibrium phase in the fractionating assemblage. TiO2 is not depleted with evolution, indicating the absence of a Fe–Ti oxide phase in the fractionation assemblage.
Mineral compositions used for mass balance calculations
Results of the mass balance calculation for investigation into the fractionating assemblages. See text for discussion
Subtracted amount as wt% of initial magma
Added amount as wt% of initial magma
Sum of the squares of the residuals
ol + cpx
5.0 ol, 5.8 cpx
ol + cpx + opx
11.7 ol, 7.2 cpx
ol + cpx + sp
3.6 ol, 10.6 cpx, 2.2 sp
ol + cpx + opx + sp
11.2 ol, 7.4 cpx, 0.1 sp
Ni is less well correlated with P; however, modelling suggests a bulk NiKD of c. 2.5. Comparing this to the value calculated by Smith et al. (2008), Ni appears slightly more compatible in the fractionating assemblage, suggesting the presence of olivine rather than orthopyroxene. Further discrimination between olivine and orthopyroxene is difficult. However, SiO2 is enriched with evolution (decreasing MgO). Microprobe data (Table 2), as well as representative analyses (Deer et al. 1992), indicate that both clinopyroxene and orthopyroxene have SiO2 contents similar to or higher than the starting composition (U1-23). As spinel was fractionating in very limited amounts (if at all), olivine is the only suitable phase that could generate SiO2 enrichment with fractionation.
The preferred fractionating assemblage is therefore clinopyroxene + olivine ± spinel. Orthopyroxene is present as phenocrysts; however, it does not fit crystal fractionation models and has calculated partition of FeO and MgO between crystal and liquid >0.4, which is too high for equilibrium orthopyroxene (Beattie et al. 1991). We suggest that orthopyroxene is xenocrystic, being derived from the mantle and entrained in various amounts in the fractionating magma. The presence of mainly olivine phenocrysts and plagioclase microphenocrysts in the eruptive products can be explained as in the case of Crater Hill (Smith et al. 2008), with these phases undergoing low-pressure crystallization during magma ascent in the upper plumbing system. The relative greater abundance of plagioclase microphenocrysts in the LTC compared to the UTC suggests heterogeneous nucleation of this phase in the rising magma column. Plagioclase was, however, not being fractionated given that Al2O3 and Sr are not depleted with evolution, and hence all crystallizing plagioclase was retained in the rising magma. The presence of intergrown chromian spinel in some olivine crystals suggests that these are residual cores from upper mantle olivine, with overgrown magmatic rims with lower MgO/FeO ratios. Clinopyroxene is present as a phenocryst phase just in the UTC. The absence of clinopyroxene in the LTC suggests that this phase had settled out of the upper part of the fractionating magma column but was carried to the surface by the ascent of the lower column.
Older basaltic lava flows underlie Udo Volcano (Sohn and Chough 1993), and the chemical variation of samples with intermediate composition may be due to contamination from such rocks. Alternatively, given that the intermediate samples always plot between the tuff cone stage and the lava shield stage compositions (Figs. 4, 5), mixing of approximately equal amounts of the two magma batches could also produce these intermediate compositions.
Lava shield stage
Modelling evolutionary processes in the lava shield stage is more challenging because fewer samples are available and the chemical ranges are narrower. Magma in this stage of the eruption was sub-alkali, and there was less enrichment of trace elements, lower MgO and higher SiO2 contents compared to the tuff stage (Figs. 4, 5, 6). Notable differences compared to the tuff stage trends are enrichment (rather than depletion) of CaO, Sc and V with evolution (decreasing MgO), which counts against any clinopyroxene fractionation in this stage. Cr does not show the same depletion trend as the tuff stage samples, suggesting that spinel is also not involved in the fractionation assemblage. Major and trace element variations are instead more consistent with olivine fractionation. Enrichment of K and Rb precludes involvement of residual phlogopite in the mantle at the site of fractionation, despite this phase being proposed as a residual metasomatic phase in the source of Jeju sub-alkaline magmas by Tatsumi et al. (2005). The fractionation depth would, however, be shallower for the LS stage compared to the tuff cone stage as indicated by the lack of a clinopyroxene influence in the observed chemical trends (e.g. Elthon and Scarfe 1984).
Chemical variation from basanite to olivine tholeiite has previously been recorded for the 1730-1736 eruption of Lanzarote (Carracedo and Rodriguez Badiola 1993; Carracedo et al. 1992); however, in that case, the transition was continuous, whereas in the case of Udo, the transition between alkali and sub-alkali basalts is clearly a sudden step change.
In the previous section, we showed that the two magmatic stages of Udo evolved through different fractionation processes. Next, we will investigate the chemical heterogeneity of the source in order to determine whether more than one source type was involved in magma generation and approximate a depth of magma sourcing.
The tuff cone stage and to a lesser extent the lava shield stage samples have La/Nb less than primitive mantle (Sun and McDonough 1989), and, given that the upper mantle (N-MORB source) or the continental crust are generally Nb depleted (Fitton et al. 1997), it follows that the source for the Udo tuff is Nb enriched. La also appears enriched as La/Zr and La/P are higher than primitive mantle for both stages.
Trace element systematics suggests the involvement of two different mantle sources in the generation of the Udo magmas. Notably, Pb is depleted in the tuff cone stage magma, whereas it is enriched in the lava shield magma (Fig. 6). Different trace element ratios also suggest that different mantle sources were involved in the generation of the Udo magmas rather than different degrees of partial melting of a single source (Reiners 2002; Zhi et al. 1990).
Mineral/melt distribution coefficients used for modelling of bulk partial melting of different sources
PMe 2.8 GPa
PMe 2.4 GPa
The only source material that can reasonably reproduce the trace element ratios in the Udo samples is garnet peridotite (Fig. 9). According to the bulk partial melting model, the tuff cone stage magma would have been derived from between 1 and 2% partial melting of peridotite at c. 2.8 GPa. By contrast, the lava shield magma results derived from c. 5 to 7% partial melting of peridotite at c. 2.4 GPa. The slight enrichment of Nb compared to primitive mantle, as discussed earlier, may support a shifting of the source (and the model) to a higher Nb/Zr ratio, thereby increasing the amount of partial melting required to generate the magma. Given that observed fractionation trends are towards lower Zr/Hf and the magmas plotted in Fig. 9 are not primary, it is likely that the Zr/Hf ratio of the source is also higher than primitive mantle. The aggregated fractional melting model for garnet peridotite (Fig. 9) gives a similar degree of partial melting for the tuff cone stage, but a lower degree (between 3 and 4%) of partial melting for the lava shield stage. Such degree of partial melting is very low for generation of sub-alkali magma (Frey et al. 1978) and hence the bulk partial melting model is preferred.
Garnet pyroxenite and eclogite have lower Zr/Hf ratios (e.g. Xu et al. 2009) and modelling of partial melting of these materials using KDs from Elkins et al. (2008) for garnet pyroxenite and from Klemme et al. (2002) for eclogite results in curves with Zr/Hf substantially lower than those of the Udo magma (Fig. 9). Fractionation decreased Zr/Hf, hence it is likely that neither garnet pyroxenite nor eclogite was involved in the melting process. Spinel lherzolite from mantle xenoliths found in Jeju basalts (Choi et al. 2005) has very low Nb content, and modelling using KD for spinel determined experimentally by Elkins et al. (2008) could not reproduce a melting curve fitting the data (Fig. 9, note that the curve at degrees of partial melting <10% has been omitted, to minimize distortion of the x axis; however, it is horizontal down to 1% partial melting).
Without claiming these models to be precise, it is nevertheless apparent from their results that the tuff cone stage magma was generated by a lesser degree of partial melting compared to the lava shield stage magma, and both stages were generated in garnet peridotite.
Depth of magma generation
Linear regression parameters and calculated primary magma compositions of alkali batch (tuff cone) and sub-alkali batch (lava shield) with Mg# = 70
Lava shield stage
For the lava shield stage, the equilibrium pressures based on the extrapolated CaO and Al2O3 composition is 2.9 GPa (Herzberg 1995) and between 1.7 and 2.7 GPa based on extrapolated total FeO and SiO2 (Hirose and Kushiro 1993). Due to the weak linear correlation within the sample group (Table 5), these should be treated as qualitative only.
The negative gradient of the normalized HREE patterns suggests the involvement of residual garnet in the source of both the tuff cone and the lava shield stages, which would indicate a minimum source pressure of c. 2.5 GPa (e.g. Walter 1998). Involvement of garnet peridotite rather than spinel peridotite, garnet pyroxenite or eclogite in the source is also supported by the modelling above.
The Udo tuff cone samples plot very close to samples from the Auckland Volcanic Field (Smith et al. 2008) on the trace element ratio plots (Pearce 2008; Fig. 10), suggesting that they are derived from a similar depth and degree of partial melting. The lower Th/Yb ratio of the Auckland samples, however, suggests that their source was not enriched as the Udo source is or that there was a lesser degree of crustal interaction.
By applying the partial melting trajectories of Pearce (2008; Fig. 10b), the alkali magma would be generated at 4 GPa in a melting column. However, as we have seen earlier, slightly enriched Nb would increase the Nb/Yb ratio and shift the path to the right, which would decrease the modelled depth of melting and would fit the sub-alkali batch to the curve too. For the modelled 3 GPa isobaric melting (Fig. 10b), a similar shift would increase the melt fraction.
We therefore suggest that the source for the two magmatic stages of Udo had slight chemical differences that cannot be attributed only to varying degrees of partial melting. The alkali magma source was possibly metasomatized garnet peridotite with slight Nb and Zr enrichment and Pb and Hf depletion to various degrees at c. 3 to 3.5 GPa, whereas the sub-alkali magma source was Pb-enriched garnet peridotite mantle at c. 2.5 GPa.
Model of conduit and magma batch interaction
The Crater Hill model of fractionation by flow crystallization on dyke walls (Irving 1980; Smith et al. 2008) could be reconciled with the Udo eruptive sequence if the LTC stage represents a magma batch that was fractionating at depth and subsequently erupted without time for further within-batch fractionation. The trend in the second batch, forming the UTC, is more complex. The return to greater degrees of evolution, after the trend towards more primitive composition, might represent magma that was stalled in a dyke system being squeezed out at the end of the eruption. Magmatic flow in a dyke is generally considered to be laminar (Rubin 1995), indicating that the core of the dyke rises faster than its margins and hence leaving more time to magma near the margins to fractionate by crystallization on the dyke walls (Irving 1980; Smith et al. 2008). Reducing magma pressure from the source may have led to the conduit walls compressing due to pressure from the country rock (Valentine and Gregg 2008; Valentine and Krogh 2006) and hence squeezing out the liquid portion of the crystal mush close to the margins of the dyke in a process akin to filter pressing (Anderson et al. 1984; Sinigoi et al. 1983). That the subsequent lava shield stage erupted immediately afterwards at the same location suggests that the eruptive conduit/dyke was still active, or at least its upper part, above the level of the shallower-fractionating sub-alkali magma.
The upper part of the UTC has low modal proportion of plagioclase microphenocrysts and higher vesicularity (+microvesicularity) compared to the LTC or the lower part of the UTC. This may be interpreted as indicating that the last erupted tuff magma did not have time to crystallize plagioclase (it did not spend as much time in the plagioclase stability field) or degas in the upper conduit, hence rose from the fractionation site and erupted more quickly. This is not implausible given that the early-rising magma batch would have opened a path for the later batch that could therefore rise more freely. Also, the presence of clinopyroxene phenocrysts only in the upper part of the sequence may suggest that the initial batch rose slower, allowing settling out of clinopyroxene, which was subsequently carried to the surface by faster rising magma forming the UTC.
The chemo-stratigraphic continuity (Fig. 3), especially in the UTC, suggests that the alkali magma was erupted from one single dyke/conduit. Such a trend could not have been easily achieved by a combination of several dykes carrying discrete magma batches to the surface. A single large dyke would also be associated with greater heat retention due to its smaller surface to volume ratio compared to several thin dykes, allowing the eruption of the late relatively evolved alkali magma, rather than this freezing in situ in a dispersed plumbing system.
The presence of very small quantities of fassaitic (Fe-rich) green clinopyroxene cores suggests that more evolved alkali magma was present, possibly as small-stalled bodies near the site of crystal fractionation or shallower and that minor interaction between the erupted alkali magma and these bodies took place (c.f. Duda and Schmincke 1985). These may represent earlier, smaller magma batches that did not have enough energy to reach the surface and erupt as can occur up to several years prior to monogenetic eruptions (Okada and Yamamoto 1991; Ukawa and Tsukahara 1996). Shallow crustal sills have been described in the exposed basement of small-volume basaltic volcanic fields (Diez et al. 2009; Valentine and Krogh 2006). Such features have been interpreted as having stored magma during the eruption of Parícutin, resulting in greater crustal contamination of late-erupted magmas (Erlund et al. 2009). However, at Udo, the lack of indicators of major crustal contamination throughout the eruption sequence suggests that if these shallow storage units exist, they did not play a major role in the eruption evolution and might have just frozen as they were intruded.
The most primitive samples of the LS stage were collected from the eastern part of Udo island similar to the pattern observed by Koh et al. (2005). This suggests that the initial lava flowed northward then eastward down slope from the tuff cone. This probably created a rampart that directed subsequent, more evolved flows to the north and the west. The observed sequence and fractionation trends are hence consistent with sequential eruption from a magma chamber that was undergoing olivine fractionation.
Initial interaction between the two batches is suggested by the presence in the LTC stage of samples with chemical composition intermediate between the tuff and lava shield magmas. Further, more thorough mixing was possibly avoided because of chemistry/viscosity contrast and high flux rates of the alkali magma that can reach >1 and up to 6 ms−1 (Demouchy et al. 2006; Rutherford 2008). A temperature difference due to derivation from different depths may also have created a temporary chilled boundary between the two magmas.
Under the Jeju mantle plume chemical genetic model of Tatsumi et al. (2005), the lava shield would correspond to the sub-alkalic series, whereas the tuff stage to the low-Al alkalic series. In their model, the sub-alkalic magma is thought to be derived from a shallower portion of the mantle compared to the alkalic magma, also supported by our data. This would indicate that the magma forming the tuff cone had to transit through the lava shield magma and possibly opened a path for it to follow. The tuff cone magma was derived from a relatively lower degree of partial melting compared to the lava shield magma, therefore it was probably richer in volatiles (Moore 1970), as suggested also by the greater CO2 content in alkali magmas compared to sub-alkali magmas (Holloway and Blank 1994). This would have resulted in higher propagation energy for the dyke tip to open a path to the surface (Rubin 1995). Once the vent was opened and the initial source exhausted, the sub-alkali magma could have exploited the upper section of the alkali magma’s plumbing system to reach the surface.
This raises the question of whether fertile mantle sources, giving rise to sub-alkali magma under Jeju, are continuously generating magma, which only has the possibility to erupt once a path is opened by a more active alkali magma possibly triggered by a mechanism such as that proposed by Valentine and Hirano (2010). The converse does not necessarily have to occur, given that the sub-alkali magma is derived from a shallower depth compared to the alkali magma, and hence it can erupt without disturbing the latter. Particularly, in the case of Jeju Island, a spectrum of alkali and sub-alkali magmas has occurred intercalated throughout the existence of the volcanic field. This may represent a combination of the time and volume predictability suggested by Valentine and Perry (2007). In the Udo case, the small-volume alkali magmas may relate to the tectonically controlled, time-predictable events, whereas the larger-volume sub-alkali magmas may represent magmatically controlled, volume-predictable events (Valentine and Perry 2007). The two magmatic systems, however, do not necessarily have to be acting independently or in a mutually exclusive fashion. Such behaviour, if true, has important implications for hazard prediction, because the magma flux involved in the initial eruption may not be indicative of the final eruptive volume but instead give a gross underestimate.
Detailed geochemical sampling of the low-volume basaltic eruption sequence of Udo volcano has shed light on magmatic processes in the upper mantle beneath Jeju Island, South Korea, and parts of its plumbing system.
Two distinct magmas, derived from separate sources at different depths in the mantle, were involved in this eruption. The first magma to be erupted is concluded to have been a low-volume alkali basalt magma derived from metasomatized peridotite with residual garnet at c. 3 to 3.5 GPa, which underwent clinopyroxene + olivine ± spinel fractionation at c. 1.5 to 2 GPa in the upper mantle buffered by metasomatic amphibole. Shallow crystallized plagioclase + olivine were retained in the magma as phenocrysts. During ascent, the alkali magma intersected a larger-ponded sub-alkali basalt magma batch that was fractionating olivine and derived from a chemically distinct mantle and shallower source at c. 2.5 GPa. Emptying of the deeper alkali source region caused the closure of the conduit/dyke system where fractionation had been taking place causing a subsequent squeezing out of the residual magma that had a more evolved composition, despite lower levels of shallow crystallization of olivine and plagioclase. The eruption conduit was exploited in the final stages of this polymagmatic eruption by a sub-alkali magma, which erupted to form the lava shield.
These results led us to suggest that deeply derived, low-volume alkali basaltic magma may act as a trigger or path-opener for eruption of shallower-derived, larger-volume sub-alkali basaltic magma and that the two magmatic systems are not mutually exclusive. Moreover, the two magmas used the same, single-dyke plumbing system; it is possible that this is a precondition for eruption of two magma types at a monogenetic volcano.
The contrasting nature of the two magma batches involved in the monogenetic eruption of Udo volcano shows how chemically and petrologically diverse a seemingly simple monogenetic volcano can be. It demonstrates that a high-resolution or comprehensive sample set is necessary to characterize the range in eruption chemistry and nature of eruption models for individual monogenetic vents.
From a hazard perspective, the findings of this study indicate that the course and final developments of a volcanic eruption in a monogenetic basaltic field cannot be solely predicted on the basis of the initial style of volcanism and the characteristics of the magma type involved.
Appreciation is expressed to Bob Stewart, Richard Price, Greg Valentine, Ting Wang and Mary Gee for constructive discussion and comments and to Chang Woo Kwon for able assistance in the field. Thorough review by Greg Valentine, Amanda Hintz and an anonymous reviewer greatly improved the manuscript. This project was supported by the Foundation for Research, Science and Technology International Investment Opportunities Fund Project MAUX0808 to SJC “Facing the challenge of Auckland volcanism”, by the Basic Science Research Program to YKS (2009-0079427) through the National Research Foundation of Korea funded by the Ministry of Education, Science and Technology and by a Massey University Vice-chancellor’s Scholarship to MB.
- Carracedo JC, Rodriguez Badiola E (1993) Evolución geológica y magmática de la isla de Lanzarote (Islas Canarias). Rev Acad Canaria Cien 5(4):25–58Google Scholar
- Deer WA, Howie RA, Zussman J (1992) An introduction to the rock-forming minerals. Longman, HarlowGoogle Scholar
- Elthon D, Scarfe CM (1984) High-pressure phase equilibria of high-magnesia basalt and the genesis of primary oceanic basalts. Am Mineral 69:1–15Google Scholar
- Eom Y, Yang K-H, Nam B, Hwang B-H, Kim J (2007) Gabbroic xenoliths in alkaline basalts from Jeju Island. J Mineral Soc Korea 20(2):103–114Google Scholar
- Erlund EJ, Cashman KV, Wallace PJ, Pioli L, Rosi M, Johnson E, Granados HD (2009) Compositional evolution of magma from Parícutin Volcano, Mexico: the tephra record. J Volcanol Geotherm Res In Press, Corrected ProofGoogle Scholar
- Frey FA, Green DH, Roy SD (1978) Integrated models of basalt petrogenesis: a study of quartz tholeiites to olivine melilitites from South Eastern Australia utilizing geochemical and experimental petrological data. J Petrol 19(3):463–513Google Scholar
- Fujimaki H, Tatsumoto M, Aoki K-I (1983) Partition coefficients of Hf, Zr, and REE between phenocrysts and groundmasses. J Geophys Res Supplement 89:B662–B672Google Scholar
- Holloway JR, Blank JG (1994) Application of experimental results to C-O-H species in natural melts. In: Carroll MR, Holloway JR (eds) Volatiles in magmas. Mineralogical Society of America, Washington DC, pp 187–230Google Scholar
- Irving AJ (1980) Petrology and geochemistry of composite ultramafic xenoliths in alkalic basalts and implications for magmatic processes within the mantle. In: Irving AJ, Dungan MA (eds) The Jackson volume. American Journal of Science, vol 280-A, pp 389–426Google Scholar
- Kang S (2003) Benthic foraminiferal biostratigraphy and paleoenvironments of the Seogwipo Formation, Jeju Island, Korea. J Paleontol Soc Korea 19(2):63–153Google Scholar
- Kil Y-W, Shin H-J, Yun S-H, Koh J-S, Ahn U-S (2008) Geochemical characteristics of mineral phases in the mantle xenoliths from Sunheul-ri, Jeju Island. J Mineral Soc Korea 21(4):373–382Google Scholar
- Kim I-S, Lee D (2000) Magnetostratigrphy and AMS of the Seoguipo formation and Seoguipo Trachyte of Jeju Island. J Geol Soc Korea 36:163–180Google Scholar
- Koh J-S, Yun S-H, Hyeon G-B, Lee M-W, Gil Y-W (2005) Petrology of the basalt in the Udo monogenetic volcano, Jeju Island. J Petrol Soc Korea 14(1):45–60Google Scholar
- Koh G-W, Park J-B, Park Y-S (2008) The study on geology and volcanism in Jeju Island (I): petrochemistry and 40Ar/39Ar absolute ages of the subsurface volcanic rock cores from boreholes in the eastern lowland of Jeju Island. Econ Env Geol 41(1):93–113Google Scholar
- McKenzie D, O’Nions RK (1991) Partial melt distributions from inversion of rare Earth element concentrations. J Petrol 32(1):1021–1091Google Scholar
- Németh K, Cronin SJ, Charley D, Harrison M, Garae E (2006) Exploding lakes in Vanuatu—“Surtseyan-style” eruptions witnessed on Ambae Island. Episodes 29(2):87–93Google Scholar
- Nicholson H, Condomines M, Fitton JG, Fallick AE, Grönvold K, Rogers G (1991) Geochemical and isotopic evidence for crustal assimilation beneath Krafla, Iceland. J Petrol 32(5):1005–1020Google Scholar
- Petrelli M, Poli G, Perugini D, Peccerillo A (2005) PetroGraph: a new software to visualize, model, and present geochemical data in igneous petrology. Geochem Geophys Geosys 6(Q07011)Google Scholar
- Reiners PW (2002) Temporal-compositional trends in intraplate basalt eruptions: implications for mantle heterogeneity and melting processes. Geochem Geophys Geosys 3(2)Google Scholar
- Rollinson HR (1993) Using geochemical data: evaluation, presentation, interpretation. Longman Group, Singapore, p 352Google Scholar
- Sun SS, McDonough WF (1989) Chemical and isotopic systematics of oceanic basalts: implications for mantle composition and processes. In: Saunders AD, Norry MJ (eds) Magmatism in the Ocean Basins. Geological Society, London, pp 313–345Google Scholar
- Takahashi E, Kushiro I (1983) Melting of a dry peridotite at high pressures and basalt magma genesis. Am Mineral 68:859–879Google Scholar
- Walker GPL (1993) Basaltic-volcano systems. In: Prichard HM, Alabaster T, Harris NBW, Neary CR (eds) Magmatic processes and plate tectonics. Geological Society, London, pp 3–38Google Scholar
- White JDL (1991) The depositional record of small, monogenetic volcanoes within terrestrial basins. In: Fisher RV, Smith GA (eds) Sedimentation in volcanic settings. Society for Sedimentary Geology, Tulsa, pp 155–171Google Scholar
- Wood CA (1979) Monogenetic volcanoes of the terrestrial planets. In: Proceedings of the 10th lunar planetary science conference, Pergamon Press, New York, pp 2815–2840Google Scholar
- Wood BJ (2004) Melting of fertile peridotite with variable amounts of H2O. Geophys Monog Series 150:69–80Google Scholar
- Yang K-H (2004) Fluid inclusions trapped in xenoliths from the lower crust/upper mantle beneath Jeju Island (I): a preliminary study. J Petrol Soc Korea 13(1):34–45Google Scholar
- Yi S, Yun H, Yoon S (1998) Calcareous nannoplankton from the Seoguipo Formation of Cheju Island, Korea and its paleoceanographic implications. Paleontol Res 2:253–265Google Scholar
- Yu J, Yang K, Hwang B, Lee S (2009) Textural and geochemical implications of spinel-peridotite xenoliths(Type I) in basaltic rocks from Jeju Island, South Korea. In: Eos transactions AGU, fall meeting supplement, abstract V13A-1998, vol 90(52)Google Scholar