CALIBRATION OF ANDESITIC WHOLE-ROCK COMPOSITIONS AGAINST MOHO DEPTH

Transcription

1 CALIBRATION OF ANDESITIC WHOLE-ROCK COMPOSITIONS AGAINST MOHO DEPTH Glen McIlwain October 2003 Thesis submitted as partial fulfilment for requirement of the Bachelor of Science (Honours) degree, Discipline of Geology, University of Newcastle.

2 ACKNOWLEDEMENTS Firstly, I would like to thank my supervisor Bill Collins for an exciting project and valuable assistance throughout the year. To my fellow honours students who made this year enjoyable, especially to Gavin Mantle where we formed a great team in field of New Zealand and in the lab. Finally, to the Geology department staff who made this possible.

3 i TABLE OF CONTENTS LIST OF FIGURES...iii LIST OF TABLES...vi LIST OF TABLES...vi ABSTRACT...vii CHAPTER 1 INTRODUCTION, BACKGROUND & AIMS Introduction Background Petrogenesis of Andesitic Magmas Geochemical Theory of Andesitic Magmas Solid Residue and Melt Products Formed at the Moho Pressure Controls on the Stability of Minerals H 2 O Concentration Controls on the Stability of Minerals Temperature Controls on the Stability of Minerals Trace Elements in Minerals Calcium, Sodium and Strontium Partitioning in Minerals Base Level geochemical signature of andesitic magmas Aims and Objectives...9 CHAPTER 2 APPROACH AND METHODS Approach Methods Collection of Geochemical Data and Moho Depths Geochemical Trends Andesitic Standardisation Proxy Ratios versus Moho Depth Field, Sample preparation and Lab Work...11

4 ii CHAPTER 3 VOLCANO LOCATIONS AND GENERAL PROPERTIES Arc Volcanoes of the World Moho Depths Volcano Properties Major Element Discrimination Diagrams Isotopic Ratio Correlations Temperature and H 2 O Content of Andesitic Magmas MORB Spider Diagrams for Each Volcano...22 CHAPTER 4 PROXY RATIOS VERSUS MOHO DEPTH AND CHONDRITE- NORMALISED SPIDER-DIAGRAMS Introduction Proxy Ratios versus Moho Depth Sr/Y versus Moho Depth (Figure 16) LREE/HREE or Y versus Moho Depth (Figures 13, 14 & 15) Ca/Sr versus Moho depth (Figure 20) Chondrite-Normalised REE+Sr+Y Spider-Diagrams Anomalous Groups...33 CHAPTER 5 DISCUSSION Shoshonites Effect of H 2 O H 2 O Andesite Categories Effect of P-T and H 2 O on the Residual Mineral Assemblage...39 CHAPTER 6 ADAKITES, PRODUCTS OF SLAB MELT?...43 CONCLUSIONS...45 REFERENCES...48 APPENDIX A GEOCHEMICAL RATIO PLOTS...53

5 iii LIST OF FIGURES Figure 1. Schematic cross section of an arc subduction zone, showing the dehydration of the subduction slab (grey), hydration and melting of the mantle wedge (purple), underplating of the mantle derived melts at the base of the crust and the MASH zone (yellow). Figure from Winter (2001)...3 Figure 2. P-T diagram for picritic tholeiite (anhydrous, 2.5-3% H 2 O and H 2 O saturated) showing mineral phase relations of plagioclase, hornblende and garnet. Redrawn from Loucks & Ballard (2002)...5 Figure 3. Partition Coefficients for REEs + Sr + Y between minerals and hydrous basaltic melt at 10 kbar. Data from Loucks & Ballard (2002)...7 Figure 4. Chondrite-normalised REEs+Sr+Y plots that show the different dominant residual mineral phases for andesites from three volcanoes. Mt Shasta (green) (Grove et al. 2002), Adatara (blue) (Fujinawa 1992) and Tata Sabaya (red) (De Silva et al 1993)....9 Figure 5. Map of the world showing active volcano locations (red triangles) and the locations of the studied volcanoes and Moho depth within each volcanic arc. Adapted from (Simkin & Siebert 2002)...13 Figure 6. Histogram showing the Moho depth beneath each volcano and the type of setting each one belongs to. Arc setting classification from Gill (1981)...14 Figure 7. Total alkalies-silica diagram using IUGS volcanic rock-type classifications. Symbols represent different ranges of Moho depth except for the high-k shoshonitic rocks that have been grouped together Figure 8. K 2 O-silica andesitic types diagram. Symbols represent different ranges of Moho depth except for the high-k shoshonitic rocks that have been grouped together. Divisions from Gill (1981)...18

6 iv Figure 9. Andesitic K 2 O-FeO*/MgO diagram showing dicriminations between calcalkaline and tholeiitic series, and between low-k, medium-k, high-k and shoshonitic rocks with 60% SiO 2. Discrimation lines from Gill (1981). Equations: FeO*/MgO = x SiO ; K 2 O = x SiO ; K 2 O = x SiO ; K 2 O = x SiO Figure 10. Sr-Nd isotope ratio correlation diagram. Showing mid-ocean ridge basalt (MORB) and primordial bulk silicate Earth (BSE) regions...20 Figure Sr/ 86 Sr i -silica diagram...21 Figure 12. MORB spider diagrams for White Island (Whakaari) 17.5 km, Ruapehu 22.5 km, Santorini 23 km, Campi Flegrei 25 km, Egmont (Taranaki) 30 km, Atacazo 32.5 km, Puyehue 32.5 km & Pinatubo 32.5 km...23 Figure 13. MORB spider diagrams for Parícutin 33 km, Bakening 34 km, Taal 34 km, Vesuvius 35 km, Mt Shasta 35 km, Adatara 35 km, Nekoma 35 km and Towada 35 km...24 Figure 14. MORB spider diagrams for San Pedro-Pellado 37.5 km, Ichinsky 38 km, Galeras 40 km, Chichinautzin 42.5 km, San Jeronimo 45 km, Cerro Tuzgle 45 km, Quimsacocha 50 km & Sangay III 50 km...25 Figure 15. Morb spider diagrams for Cerro Leon Muerto 55 km, Antisana 60 km, Nevado Solimana 62.5 km, Quillacas 65 km, Ollagüe 65 km, Parinacota 70 km & Tata Sabaya 70 km...26 Figure 16. Sr/Y versus Moho depth diagram for average andesite composition (60% SiO 2 ) Figure 17. La/Yb versus Moho depth diagram for average andesite composition (60% SiO 2 )....29

7 v Figure 18. La/Y versus Moho depth diagram for average andesite composition (60% SiO 2 ) Figure 19. Ce/Y versus Moho depth diagram for average andesite composition (60% SiO 2 ) Figure 20. Ca/Sr versus Moho depth diagram for average andesite composition (60% SiO 2 ) Figure 21. Chondrite-normalised REE+Sr+Y spider-diagrams divided into groups of similar shapes and slopes. (A) plagioclase + hornblende signature, (B) garnet signature, (c) plagioclase signature, (D) hornblende signature, (E) shoshonitic, (F) Garnet + Plagioclase Figure 22. Ca/Sr versus La/Yb diagram, representing changes in residual plagioclase versus hornblende and garnet with respect to Moho depth...37 Figure 23. La/Sm versus La/Yb diagram representing changes in dominant residual minerals and H 2 O Figure 24. Sr/Ce (Sr anomaly) versus La/Yb (pressure signature) diagram, showing H 2 O Andesite category divisions Figure 25. P-T diagrams for average, high and low H 2 O system showing six volcanoes that represent H 2 O andesite categories from section 5.2. P & T conditions are measured for all the volcanoes, except White Island and Ollagüe. In the White Island case, the T field represents the typical range for andesites, whereas with Ollagüe, the T field is an estimate for a thick crust, dry system...40

8 vi LIST OF TABLES Table 1. Volcano Locations, Moho depth, Properties and Classifications Table 2. Major elements, REE, Sr, Y and proxy ratios for each volcano averaged to 60% SiO

9 vii ABSTRACT An analysis of the global database of volcanoes, for which Moho depth is known, was undertaken to determine if a relation exists between andesitic composition and crustal thickness. One thousand and sixty seven samples from 31 active volcanoes from 11 volcanic arcs were examined, where the Moho depths ranged from 17 km to 70 km. With increasing Moho depth, chondrite normalised spider diagrams become progressively steeper, which is best quantified by increasing LREE/HREE ratios. Only shoshonites and some low-k calc-alkaline andesites do not follow this trend. The correlation of composition with Moho depth is best explained if it is considered andesites acquire their geochemical character at the Moho, which supports the MASH hypothesis of Hildreth & Moorbath (1988). At the Moho, the composition of andesites is controlled by the major residual minerals, plagioclase, hornblende and garnet. At shallow Moho depth (<30 km), plagioclase is the dominant mineral, at intermediate depth it is hornblende, and at greater depth (>45 km) it is garnet. The geochemical signature of andesitic magmas formed at the Moho within volcanic arc environments shows a characteristic mirror image of the dominant residual minerals, when plotted on Chondrite-normalised REE+Sr+Y graphs. Therefore, the elements that partition within these three minerals can be used as proxies for Moho depth. The Sr/Y, La/Yb, La/Y and Ce/Y ratios are particularly useful in this regard. H 2 O content affects the residual mineral assemblage. For a given pressure, increasing H 2 O content destabilises plagioclase whereas hornblende and garnet becomes increasingly stable. This effect causes the LREE/HREE ratio to be modified. Andesites with high H 2 O content (>4%) crystallise hornblende at an early stage, whereas relatively anhydrous andesites (<1%) crystallise plagioclase at an early stage. Garnet crystallises at much greater depths in anhydrous systems (40 km versus ~65 km for anhydrous versus average andesites with 1-4% H 2 O). The different andesite categories (low-, average-, high-h 2 O) can be distinguished on Sr/Ce versus La/Yb plots. This effect must be considered when applying HREE/LREE or Sr/Y to establish Moho depths. Based on this work, it appears that the composition of andesites is independent of whether it formed either by, (1) differentiation from basaltic arc magmas, (2) partial melting of under-plated basaltic arc magmas, or other crustal material, or (3) mixing of these crustal and mantle derived magmas. Adakites are formed at the Moho under unusually high-h 2 O contents and/or high-pressure conditions. Finally, the data suggest that the mantle lithosphere beneath arcs is very thin or non-existent.

10 Chapter 1 Introduction, Background & Aims 1 CHAPTER 1 INTRODUCTION, BACKGROUND & AIMS 1.1 Introduction In the past, geologists have looked at the structural evolution of the crust to qualitatively determine crustal thickness variations in orogenic systems, but the composition of andesitic magmas may also show this variation. A global compilation of active volcanoes (Plank & Langmuir 1988) shows that Mohorovičić discontinuity (Moho) depths beneath arcs varies between km. The Moho represents the boundary where the velocity of seismic P-waves increases abruptly from 7 km per second to 8 km per second for the deeper rocks. The velocity of 8 km per second indicates that the deeper rocks are probably dense ultramafic rock, peridotite, whereas the above rocks are the less dense buoyant gabbro (Best 2003). Thus, the Moho is interpreted to reflect the boundary between the lower crust and the mantle and hence the thickness of the crust. Most workers consider that andesites form at the base of the crust, either by (1) differentiation from basaltic arc magmas, (2) partial melting of under-plated basaltic arc magmas, or other crustal material, or (3) mixing of these crustal and mantle derived magmas (e.g.tatsumi & Eggins 1995). These processes occur at varying depth at the base of arcs, either within the garnet stability field (>50 km), hornblende stability field (30-50 km), or the plagioclase stability field (<30 km) (e.g. Kay & Mpodozis 2001). The Sr/Y and La/Yb ratios are particularly sensitive to the presence of these minerals, so it is possible that such ratios can be used to calibrate the Moho depth to andesitic

11 Chapter 1 Introduction, Background & Aims 2 compositions. This approach could be used to monitor the long-term crustal thickness variations in accretionary orogenic systems, which has not been attempted before. 1.2 Background Petrogenesis of Andesitic Magmas Within accretionary orogens, subduction zones are formed when two tectonic plates converge (Figure 1). Subduction zones have three main components: (1) a relatively, cold and dense subducting oceanic plate; (2) overriding buoyant continental plate; (3) a mantle wedge between the two. The crust of the oceanic plate contains a large proportion of hydrous minerals, which have formed during hydration and greenschist facies metamorphism at the mid-ocean ridge. As the oceanic slab is subducted, the crust dehydrates and releases H 2 O into the mantle wedge to form a hydrous peridotite layer, which is dragged down by the subducting slab. When a depth of ~110 km is reached, H 2 O is released from this layer by pressure-sensitive dehydration reactions (Tatsumi & Eggins 1995). The influx of free H 2 O into the mantle wedge causes partial melting and initiates rise of a melt column, which undergoes further melting during decompression in the mantle wedge peridotite. The resultant partially melted product a hot (>1100 o C), dense, hydrous basalt, which ponds at and commonly underplates the base of the overlying (more buoyant) crust, with only a small proportion of magma rising to higher crustal levels (Tatsumi & Eggins 1995). At the base of the crust (the Moho), combined processes of crustal partial melting, assimilation, storage, and homogenisation (MASH) takes place to produce andesitic magmas with a base level geochemical signature reflecting equilibration with residual

12 Chapter 1 Introduction, Background & Aims 3 minerals which form at specific Moho depths (Hildreth & Moorbath 1988). Over time, these MASH zone magmas become buoyant as they become more silica rich, and may ascend to upper levels of the crust when they experience shallow fractional crystallisation, magma mingling and wall rock contamination. Such processes will change the concentration of many elements and the magmas will evolve to more silicic compositions. For this reason, this work is restricted to examination of intermediatecomposition rocks, between 57-63% SiO 2. Figure 1. Schematic cross section of an arc subduction zone, showing the dehydration of the subduction slab (grey), hydration and melting of the mantle wedge (purple), underplating of the mantle derived melts at the base of the crust and the MASH zone (yellow). Figure from Winter (2001).

13 Chapter 1 Introduction, Background & Aims Geochemical Theory of Andesitic Magmas Solid Residue and Melt Products Formed at the Moho At the Moho within volcanic arcs, three processes may be operating to control the base level geochemical signature of andesitic magmas: partial melting of the lower crust; fractional crystallization of under-plated hydrous mantle derived mafic magmas; and magma mixing (Winter 2001). Rocks partially melt to form two components, a buoyant silicic melt and a dense solid residue. The silicic melt is formed where low temperature minerals are initially melted (eg. biotite), whereas the solid residue consists of stable high temperature minerals (eg. pyroxene). In contrast, during fractional crystallization the stable minerals drop out of the melt and become part of the solid residue. The elemental abundances within the melt and residue component are mirror images to each other, where the elements that concentrate in the residue are depleted from the melt. The mineral composition of the residue is dependent on the environment (temperature, pressure and H 2 O content), which therefore controls the base level geochemical signature within the magma Pressure Controls on the Stability of Minerals Different minerals are stable at different pressures (e.g. Kay & Abbruzzi 1996). Changes in mineral stability change the crystallization or partial melting order. For example, with increasing pressure plagioclase becomes increasingly unstable, whereas hornblende and garnet become increasingly stable. Therefore, plagioclase diminishes in modal abundance and eventually vanishes from the mineral assemblage of the solid residue with increasing pressure, whereas the modal abundances of hornblende and garnet increases. This can be seen in Figure (2). By monitoring the modal abundance of

14 Chapter 1 Introduction, Background & Aims 5 the minerals within the solid residue, through systematic changes in trace element ratios, the depth of crystallisation might be determined. Figure 2. P-T diagram for picritic tholeiite (anhydrous, 2.5-3% H 2 O and H 2 O saturated) showing mineral phase relations of plagioclase, hornblende and garnet. Redrawn from Loucks & Ballard (2002) H 2 O Concentration Controls on the Stability of Minerals Increasing the H 2 O concentrations has similar effects of increasing pressure on the crystallisation order and stability of minerals (Figure 2). High H 2 O contents (>3 wt%) (Müntener et al. 2001) suppresses plagioclase and enhances the crystallisation of hornblende and garnet, to produce a corundum normative andesitic melt. Low H 2 O (<1

15 Chapter 1 Introduction, Background & Aims 6 wt%) concentrations stabilises plagioclase earlier than garnet and hornblende, so the melt retains its quartz normative character (Müntener et al. 2001) Temperature Controls on the Stability of Minerals Changes in temperature of andesitic magmas have a minimal effect on the nature of the residual mineral assemblage when compared to pressure and H 2 O content. This is because the majority of andesites have a narrow pre-eruption temperature range between ~1000 and 1100 C (Gill 1981) Trace Elements in Minerals Different minerals concentrate different proportions of trace elements within their crystal structure. The major residual minerals in andesites (plagioclase, hornblende and garnet) all show contrasting trace elemental abundance characteristics, as shown in the partition coefficient plots (Figure 3). The plagioclase plot shows a flat gradient, large strontium (Sr) and europium (Eu 2+ ) positive anomalies, but with low overall concentrations of rare earth elements (REE) and yttrium (Y). The REE and Y are more strongly partitioned within hornblende, and a small negative Sr anomaly exists. The concave-down curved pattern maximising at the middle heavy rare earth element region, is a mirror image of the plagioclase curve. Garnet shows a steep gradient, with the heavy rare earth elements (HREE) strongly partitioned and the LREE strongly excluded from the mineral lattice. The remaining major rock forming minerals (eg. pyroxene ± olivine) have flat patterns and very low partition coefficients in andesites (Best 2003). This indicates that these other basalt-hosting minerals do not influence the behaviour of REE and Sr during andesite petrogenesis.

16 Chapter 1 Introduction, Background & Aims 7 Figure 3. Partition Coefficients for REEs + Sr + Y between minerals and hydrous basaltic melt at 10 kbar. Data from Loucks & Ballard (2002) Calcium, Sodium and Strontium Partitioning in Minerals Increasing pressure and/or H 2 O content of a magma causes crystallising plagioclase to become more calcic and less sodic (Blundy & Shimizu 1991). Sr 2+ is more strongly partitioned into sodic plagioclase than calcic plagioclase, because the relatively weak Na-Al bonds give the albite lattice the flexibility to accommodate the large Sr ion (Best 2003). The decreasing partitioning of Sr within hornblende and garnet can be explained by the following mineral reactions where the plagioclase component breaks down to form these minerals: 2((Ca,Na) Al 2 Si 2 O 8 ) + 3(MgSiO 3 ) (Na,Ca) 2 Mg 3 Al 4 SiO 6 O 22 (OH) 2 + SiO 2 Plagioclase (An > Ab) + OPX Hornblende + Quartz

17 Chapter 1 Introduction, Background & Aims 8 CaAl 2 Si 2 O 8 + (Ca,Mg) 2 Si 2 O 6 (Ca,Mg) 3 Al 2 Si 3 O 12 + SiO 2 Anorthite + CPX Garnet + Quartz Therefore, since the increasing role of calcic plagioclase in these reactions the partitioning of Sr in hornblende and garnet would decrease. This also strongly implicit on the Ca/Sr ratios of the melt, and decreasing Ca/Sr ratios in andesites would indicate higher pressure and/or H 2 O content Base Level geochemical signature of andesitic magmas The geochemical signature of andesitic magmas formed at the Moho within volcanic arc environments shows a characteristic mirror image of the dominant residual minerals. This is regarded as their base level signature and can be seen in Figure (4), where three contrasting chondrite-normalised plots display mirror images to the partition coefficient plots shown in Figure (3). Therefore, the composition of the andesitic magmas appears to reflect the presence of the pressure-dependent residual mineral assemblage. This compositional variation is well characterised by the REE, Sr and Y. It therefore might be possible to determine the stable mineral phases during formation of most andesite magmas. By using contrasting elemental ratios that are controlled by these minerals, the depth to the Moho could be determined. For example, andesites formed under thick crustal conditions (>50 km) show a mirror image of garnet (eg. high Sr/Y, La/Yb), whereas andesites formed during thin crust conditions (<30 km)

18 Chapter 1 Introduction, Background & Aims 9 show a mirror image of plagioclase or a mixture of plagioclase and hornblende (eg. low Sr/Y, La/Yb) (Kay & Mpodozis 2001). Figure 4. Chondrite-normalised REEs+Sr+Y plots that show the different dominant residual mineral phases for andesites from three volcanoes. Mt Shasta (green) (Grove et al. 2002), Adatara (blue) (Fujinawa 1992) and Tata Sabaya (red) (De Silva et al 1993). 1.3 Aims and Objectives Given the theoretical considerations outlined above, the aim of this thesis is to determine if a relation exists between andesite composition and Moho depth from active volcanoes. The objective is then to calibrate the trace element ratios of andesites with Moho depth. If successful, a proxy indicator for crustal thickness will exist, which can be applied to those orogens that contain abundant intermediatecomposition rocks.

19 Chapter 2 Approach And Methods 10 CHAPTER 2 APPROACH AND METHODS 2.1 Approach The approach is to compile a database using existing whole-rock geochemical data from active volcanoes from around the world, where Moho depths have been determined. Properties (geochemistry, isotopic characteristics, location and type) for each volcano were studied to investigate whether the volcanos can be grouped into similar petrological types. This database was used to determine which geochemical parameters vary consistently with Moho depth. 2.2 Methods Collection of Geochemical Data and Moho Depths Before any volcanic arc whole-rock trace-element geochemical data was entered into a database, the Moho depth and the age of the rocks had to be known. The age of the rocks was restricted to <1 Ma, because the thickness of the crust can rapidly change over a short period of time (Gutscher et al. 2000). Only those volcanoes with at least five analysis of andesite composition were considered, to avoid analytical bias. The majority of the data was collected from journals and a large proportion is stored in the web-based databank GEOROC ( All volcanoes that contained andesites, where Moho depths were known, were included in this study.

20 Chapter 2 Approach And Methods Geochemical Trends For each volcanic dataset, graphs of element abundance ratios were plotted against SiO 2 to investigate what petrological processes were involved in their genesis. These processes could be fractional crystallisation, partial melting of the lower crust, or mixing of both components. 143 Nd/ 144 Nd i versus 87 Sr/ 86 Sr i isotopic ratios were also plotted to determine if multiple source components were involved in magma genesis Andesitic Standardisation For the purposes of direct comparison of data, the elemental compositions for each volcano were restricted to andesitic compositions (57-63% SiO 2 ) and then averaged to 60% SiO 2. These averaged elemental compositions were plotted as chondrite-normalised spider-diagrams to compare data from each volcano Proxy Ratios versus Moho Depth Trace element ratios were plotted against Moho depth and those with the highest degree of correlation were identified for each volcanic suite. These were the values used to calibrate andesite composition against Moho depth, and thus can be considered proxy ratios of crustal thickness. Finally, proxy values were calculated for each Moho Field, Sample preparation and Lab Work A major part of my honours requirement involved collection and processing of ~150 samples of basaltic dykes from the South Island of New Zealand, with Gavin

21 Chapter 2 Approach And Methods 12 Mantle. Processing involved crushing and pulverising samples, producing XRF disks and preparing rock samples for thin section. This aspect of the work provided experience in field and laboratory work that would have been otherwise lacking in this thesis.

22 Chapter 3 Volcano Locations And General Properties 13 CHAPTER 3 VOLCANO LOCATIONS AND GENERAL PROPERTIES 3.1 Arc Volcanoes of the World Figure 5. Map of the world showing active volcano locations (red triangles) and the locations of the studied volcanoes and Moho depth within each volcanic arc. Adapted from (Simkin & Siebert 2002) Thirty-one active volcanoes from eleven different arcs around the world were studied in this thesis. It can be seen in Figure (5) that the majority of volcanoes are located within the Pacific Rim, with only three from Aeolian and Aegean arcs of Europe. The Andean continental arc system of South America is highly represented because this is where Moho depths vary most, and where a large body of geochemical data exists. The thirty one volcanoes studied is only a small proportion of the >1500 volcanoes that have erupted during the past 10,000 years (Gill 1981). This restriction is

23 Chapter 3 Volcano Locations And General Properties 14 due to the lack of geochemical analyses of rocks with an andesitic composition, where accurate Moho depth measurements have been determined. Island Arcs Continental Fragments Continental Peninsulas Continental Margins Continental Interiors Moho Depth Volc a noe s Figure 6. Histogram showing the Moho depth beneath each volcano and the type of setting each one belongs to. Arc setting classification from Gill (1981). 3.2 Moho Depths The range in Moho depths of the volcanoes studied (Figure 6) range from ~17 km at White Island (Whakaari) within the Taupo Volcanic Zone of New Zealand to ~70 km at Parinacota and Tata Sabaya within the Central Andean Volcanic Zone of South America. Gill (1981) divided each volcanic arc into geographical categories; mainland continental margins (e.g. Chile); peninsular continental margins (e.g. Kamchatka); detached continental fragments (e.g. Japan); island arcs (e.g. Tonga); or continental interiors (e.g. Eastern Colorado). These categories are used in Figure (6). This shows

24 Chapter 3 Volcano Locations And General Properties 15 that the majority of volcanoes have Moho depths between 30 and 40 km. Those volcanos located at continental margins generally have Moho depths >35 km. The deficiency of island arc volcanoes in the compilation is because these arcs do not commonly contain andesites. 3.3 Volcano Properties The general geochemical properties of each volcano are presented in table (1) and geochemical plots for each volcano exist in Appendix (A) Major Element Discrimination Diagrams The Total alkalies-silica diagram (Figure 7) shows that the majority of rocks fall into the basalt to dacite fields, with some extending to rhyolite. Those that fall in the basaltic trachyandesite to trachydacite fields are typically the high-k suites (Figure 8). A group with anomalously high K 2 O also exist, which typically have a phonolitic character (Figure 7). Such volcanoes exist in the Aeolian arc and in Eastern Central Andes (Cerro Tuzgle and San Jeronimo). These are grouped as shoshonitic in this thesis. Figures (7 & 8) also show that the majority of the rocks show a linear trend for each volcano to high K 2 O with increasing SiO 2. Some volcanoes also show a general trend from low to high K 2 O and total alkalies increasing Moho depth (Figures 7 & 8).

25 Chapter 3 Volcano Locations And General Properties 16 Table 1. Volcano Locations, Moho depth, Properties and Classifications. Volcano Volcanic Arc Latitude Longitude Moho (km) SiO 2 Range (%) 87 Sr/ 86 Sr Range Magma Processes Rock Type (60% SiO 2 ) 1 White Island (Whakaari) 2 Ruapehu New Zealand / Taupo Volcanic Zone Differentiation Andesite New Zealand / Taupo Volcanic Zone Mixing Andesite 3 Santorini Aegean Arc Mixing Trachyandesite 4 Campi Flegrei Aeolian Arc Mixing Trachyte 5 Egmont (Taranaki) 6 Atacazo 7 Puyehue New Zealand / Taupo Volcanic Zone Mixing Trachyandesite Andean Arc / Northern Andean Volcanic Zone Mixing Andesite Andean Arc / Southern Andean Volcanic Zone Differentiation Andesite 8 Pinatubo Luzon Arc Mixing Andesite 9 Parícutin Mexican Volcanic Belt Mixing Andesite 10 Bakening Kamchatka Arc Mixing Andesite 11 Taal Luzon Arc Differentiation Trachyandesite 12 Vesuvius Aeolian Arc Differentiation Trachyte 13 Mt Shasta Cascades / Southern Cascades Differentiation Andesite 14 Adatara Honshu Arc Mixing Andesite 15 Nekoma Honshu Arc Mixing Andesite 16 Towada Honshu Arc Mixing Andesite 17 San Pedro-Pellado Andean Arc / Southern Andean Volcanic Zone Mixing Andesite 18 Ichinsky Kamchatka Arc Mixing Andesite 19 Galeras Andean Arc / Northern Andean Volcanic Zone Mixing Andesite 20 Chichinautzin Mexican Volcanic Belt Differentiation Andesite 21 San Jeronimo 22 Cerro Tuzgle 23 Quimsacocha 24 Sangay III 25 Cerro Leon Muerto 26 Antisana 27 Nevado Solimana 28 Quillacas 29 Ollagüe 30 Parinacota 31 Tata Sabaya Andean Arc / Central Andean Volcanic Zone Differentiation Trachyandesite Andean Arc / Central Andean Volcanic Zone Mixing Trachyandesite Andean Arc / Northern Andean Volcanic Zone Mixing Andesite Andean Arc / Northern Andean Volcanic Zone Mixing Trachyandesite Andean Arc / Central Andean Volcanic Zone Mixing Andesite Andean Arc / Northern Andean Volcanic Zone Mixing Trachyandesite Andean Arc / Central Andean Volcanic Zone Mixing Trachyandesite Andean Arc / Central Andean Volcanic Zone Mixing Trachyandesite Andean Arc / Central Andean Volcanic Zone Mixing Trachyandesite Andean Arc / Central Andean Volcanic Zone Mixing Trachyandesite Andean Arc / Central Andean Volcanic Zone Mixing Trachyandesite

26 Chapter 3 Volcano Locations And General Properties 17 Figure 7. Total alkalies-silica diagram using IUGS volcanic rock-type classifications. Symbols represent different ranges of Moho depth except for the high-k shoshonitic rocks that have been grouped together. The K2O-FeO*/MgO diagram (Figure 9) discriminates at 60% SiO 2 whether the volcano is calc-alkaline or tholeiitic, and whether the volcano is low-k, medium-k, high-k or shoshonitic (adapted from Gill (1981)). The majority of volcanoes belong to the medium-k calc-alkaline group, although the high-k calc-alkaline group is also well represented. The few tholeiitic volcanoes are Taal (high-k), Santorini, Cerro Leon Muerto, Puyehue and Quimsacocha (medium-k) and Nekoma (low-k). One volcano is low-k, calc-alkaline (Towada) and few are shoshonitic (Campi Flegrei, Vesuvius, Cerro Tuzgle and San Jeronimo). The medium-k calc-alkaline group have Moho depths <45 km, whereas the volcanoes that belong to the high-k calc-alkaline group have Moho depths >50 km or

27 Chapter 3 Volcano Locations And General Properties 18 are located at the back-arc side of volcanic arc systems (e.g. Egmont (Taranaki) 30 km). The tholeiitic groups show no trend with Moho depth. Figure 8. K 2 O-silica andesitic types diagram. Symbols represent different ranges of Moho depth except for the high-k shoshonitic rocks that have been grouped together. Divisions from Gill (1981).

28 Chapter 3 Volcano Locations And General Properties 19 Figure 9. Andesitic K 2 O-FeO*/MgO diagram showing dicriminations between calc-alkaline and tholeiitic series, and between low-k, medium-k, high-k and shoshonitic rocks with 60% SiO 2. Discrimation lines from Gill (1981). Equations: FeO*/MgO = x SiO ; K 2 O = x SiO ; K 2 O = x SiO ; K 2 O = x SiO Isotopic Ratio Correlations Isotopic studies of arc andesites are used primarily to evaluate whether the magma source is crustal or mantle-derived or whether the magma is a crust-mantle mix (Gill 1981). The Sr-Nd isotope diagram (Figure 10) shows that the majority of volcanoes trend between mid-ocean ridge basalt (MORB) and the primordial bulk silicate Earth (BSE). These volcanoes generally have formed at low Moho depths <40 km, and show little scatter in isotope compositions. In contrast, the volcanoes formed at greater Moho depths show a large scatter towards high 87 Sr/ 86 Sr i and lower 143 Nd/ 144 Nd i, suggesting that the magma is contaminated by radiogenic crustal sources. This can be

29 Chapter 3 Volcano Locations And General Properties 20 seen also in Figure (11) where the volcanoes with low Moho depths show very little change in 87 Sr/ 86 Sr i with respect to silica, but the volcanoes with high Moho depths show a large scatter. Figure 10. Sr-Nd isotope ratio correlation diagram. Showing mid-ocean ridge basalt (MORB) and primordial bulk silicate Earth (BSE) regions. The shoshonitic volcanoes also form a distinct group with 143 Nd/ 144 Nd i ratios close to BSE, but the 87 Sr/ 86 Sr i ratios are generally elevated. However, they show a small scatter with respect to silica (Figure 11). This relation could be explained if the major source region for these magmas is enriched lithospheric mantle. Only those volcanoes formed above thick crust (i.e. those with isotopic scatter) show obvious isotopic evidence for magma mixing. Nonetheless, the wide silica range for some volcanoes, suggest they may also be crust-mantle mixes, even though they show a limited isotopic range. For these volcanoes, the crustal source must be juvenile (i.e. non-radiogenic).

30 Chapter 3 Volcano Locations And General Properties 21 Figure Sr/ 86 Sr i -silica diagram Temperature and H 2 O Content of Andesitic Magmas The temperature of andesite at eruption is a function of volatile composition, particularly water content, degree of crystallisation, and heat liberated during eruption (Gill 1981). Field measurements during the 1944, 1945 and 1946 andesitic eruptions at Parícutin (Mexican Volcanic Belt) yielded maximum temperatures of 1100 C, 1070 C and 1040 C respectively (Luhr 2001). Phase-equilibrium and mineral geothermometry experiments on Parícutin andesite have predicted that the pre-eruption temperatures were 1110±40 C. Three other volcanoes, Mt Shasta (Southern Cascades), Adatara (Honshu Arc) and Tata Sabaya (Central Andean Volcanic Zone) also show similar preeruption temperatures of C (Grove et al. 2002), C (Hunter & Blake 1995) and C (De Silva et al. 1993), respectively. Measurements from Parícutin glass inclusions within olivine indicate the H 2 O content of the melt ranged from 1.8 to 4.0 wt% (Luhr 2001). Mt Shasta and the trench-

31 Chapter 3 Volcano Locations And General Properties 22 side Honshu arc volcanoes (e.g. Adatara) show contrasting H 2 O concentrations of 4.5- >8.0 wt% (Grove et al. 2002) and <0.5 wt% (Tatsumi & Eggins 1995), respectively. The H 2 O content for Tata Sabaya is unknown. These results show that H 2 O content does not affect pre-eruption temperatures. The pre-eruption temperature for Tata Sabaya may be low due large crustal thickness (70 km) MORB Spider Diagrams for Each Volcano MORB normalised spider diagrams (Figures 12 15) are plotted for each volcano in Moho depth order. All of the plots show similar patterns, except the volcanoes with insufficient data (e.g. Santorini has only 9 out of 28 elements available). These similar patterns are positive Pb and negative Nb, Ta and Ti anomalies. Some of the volcanoes show large positive Sr anomalies (Mt Shasta, Quimsacocha and Sangay III) whereas the majority have small or no anomaly (Adatara, Puyehue, Towada and the shoshonites). The majority have large ion lithophile (LIL) element normalised abundances (particularly Rb & Ba) , except for the volcanoes with Moho depths over 60 km, which have values , the shoshonitic volcanoes with values up to 700. The high field strength (HFS) elements are more variable, with the slope generally steepening with increasing Moho Depth reflecting decreasing HREE ( ). The exceptions are the Honshu arc volcanoes (~35 km depth) that show a relatively flat HFS slope, similar to White Island (17.5 km). In addition, the shoshonitic volcanoes have a steep HFS slope compared with volcanoes with similar Moho depths. This reflects elevated LILE for Campi Flegrei and Vesuvius, but decreased HREE for Cerro Tuzgle and San Jeronimo.

32 Chapter 3 Volcano Locations And General Properties 23 Figure 12. MORB spider diagrams for White Island (Whakaari) 17.5 km, Ruapehu 22.5 km, Santorini 23 km, Campi Flegrei 25 km, Egmont (Taranaki) 30 km, Atacazo 32.5 km, Puyehue 32.5 km & Pinatubo 32.5 km

33 Chapter 3 Volcano Locations And General Properties 24 Figure 13. MORB spider diagrams for Parícutin 33 km, Bakening 34 km, Taal 34 km, Vesuvius 35 km, Mt Shasta 35 km, Adatara 35 km, Nekoma 35 km and Towada 35 km

34 Chapter 3 Volcano Locations And General Properties 25 Figure 14. MORB spider diagrams for San Pedro-Pellado 37.5 km, Ichinsky 38 km, Galeras 40 km, Chichinautzin 42.5 km, San Jeronimo 45 km, Cerro Tuzgle 45 km, Quimsacocha 50 km & Sangay III 50 km

35 Chapter 3 Volcano Locations And General Properties 26 Figure 15. Morb spider diagrams for Cerro Leon Muerto 55 km, Antisana 60 km, Nevado Solimana 62.5 km, Quillacas 65 km, Ollagüe 65 km, Parinacota 70 km & Tata Sabaya 70 km

36 Chapter 4 Proxy Ratios Versus Moho Depth And Chondrite-Normalised Spider-Diagrams 27 CHAPTER 4 PROXY RATIOS VERSUS MOHO DEPTH AND CHONDRITE-NORMALISED SPIDER-DIAGRAMS 4.1 Introduction It was suggested in chapter 1 that the best trace element ratios to discriminate changes in the Moho depth are those that represent the properties of the residual minerals of plagioclase, hornblende and garnet. These ratios are Sr/Y, LREE/(HREE or Y) and Ca/Sr. The averaged (to 60% SiO 2 ) elemental compositions are shown in Table (2) with the major elements, REE, Sr and Y. Chondrite-normalised REE+Sr+Y spiderdiagrams are also plotted for those volcanoes of similar properties and sufficient data. 4.2 Proxy Ratios versus Moho Depth Sr/Y versus Moho Depth (Figure 16) It can be seen in Figure (16) that a positive correlation exists of increasing Moho depth with increasing Sr/Y, for many volcanoes. At low Moho depths (White Island (Whakaari) ~17.5 km), average Moho depths (Parícutin ~33 km) and high Moho depths (Tata Sabaya ~70 km), Sr/Y is ~10, 25 and 70 respectively. Two main groups do not follow this trend: Mt Shasta (blue crosses) and the Adatara (red dots) groups. The Mt Shasta group (Mt Shasta, Quimsacocha and Sangay III) has very high Sr/Y values (>70) with respect to Moho depth (35 to 50 km), whereas the Adatara group (Adatara, Nekoma, Towada, Puyehue and Taal) has very low Sr/Y values (<10) at ~35 km Moho

37 Chapter 4 Proxy Ratios Versus Moho Depth And Chondrite-Normalised Spider-Diagrams 28 depths. Omitting the Mt Shasta, Adatara and the shoshonitic groups the trend can be expressed by the equation Sr/Y = x (Moho) with a R 2 value of Table 2. Major elements, REE, Sr, Y and proxy ratios for each volcano averaged to 60% SiO 2.

38 Chapter 4 Proxy Ratios Versus Moho Depth And Chondrite-Normalised Spider-Diagrams 29 Figure 16. Sr/Y versus Moho depth diagram for average andesite composition (60% SiO 2 ). Figure 17. La/Yb versus Moho depth diagram for average andesite composition (60% SiO 2 ).

39 Chapter 4 Proxy Ratios Versus Moho Depth And Chondrite-Normalised Spider-Diagrams LREE/HREE or Y versus Moho Depth (Figures 13, 14 & 15) The diagrams that are used to represent LREE/HREE ratios are La/Yb (Figure 17), La/Y (Figure 18) and Ce/Y (Figure 19). These diagrams show similar trends to each other and all three are used because certain elements are preferred with different analytical techniques (e.g. INAA=Yb and XRF=Y). Like Sr/Y ratios, LREE/HREE ratios also show a positive trend with increasing Moho depth. La/Yb, La/Y and Ce/Y values for White Island (Whakaari) (17.5 km) are 4.5, 0.5 and 1.0; Parícutin (33 km) are 11, 1.0 and 2.2; and Tata Sabaya (70 km) are 40, 7.75 and 7.0, respectively. The major difference between these diagrams and the Sr/Y diagram is that the Mt Shasta group now fits the average trend, but the Adatara group is anomalously low and displays low values similar to that of White Island, where the Moho is extremely shallow. Omitting the Adatara and shoshonitic groups each trend can be expressed by the equations: La/Yb = e x (Moho), R 2 = ; La/Y = e x (Moho), R 2 = ; Ce/Y = e x (Moho), R 2 = Ca/Sr versus Moho depth (Figure 20) It can be seen in Figure (20) that Ca/Sr has a negative correlation with Moho depth where White Island (Whakaari) has Ca/Sr values of 250, Adatara of 200 and Tata Sabaya of 40. The major difference between the Ca/Sr diagram and the previous plots is that the Adatara group now fits a linear trend (Ca/Sr = x (Moho) , R 2 = ) between White Island and Tata Sabaya, and a hyperbolic trend exists for the volcanoes that followed the normal trend in those plots (pink line).

40 Chapter 4 Proxy Ratios Versus Moho Depth And Chondrite-Normalised Spider-Diagrams 31 Figure 18. La/Y versus Moho depth diagram for average andesite composition (60% SiO 2 ). Figure 19. Ce/Y versus Moho depth diagram for average andesite composition (60% SiO 2 ).

41 Chapter 4 Proxy Ratios Versus Moho Depth And Chondrite-Normalised Spider-Diagrams 32 Figure 20. Ca/Sr versus Moho depth diagram for average andesite composition (60% SiO 2 ). 4.3 Chondrite-Normalised REE+Sr+Y Spider-Diagrams The Chondrite-normalised REE+Sr+Y spider-diagrams (Figure 21) have been divided into groups of similar shapes and slopes to help distinguish the different residual mineral phases existing at the base of the crust. To understand the data, reference should be made to Figures (3 & 4) in chapter 1. The volcanoes plotted in Figure (21A) have a moderate negative slope and a small positive Sr anomaly. This suggests that the main residual phases are plagioclase and hornblende. The plots in Figure (21B) show a very steep negative slope indicating that garnet is the main residual mineral. The plots in Figure (21C) have a very flat slope, with no or negative Sr and Eu anomalies. This suggests that plagioclase is the main residual mineral, with no or very little hornblende involved. The plots of Figure (21D)

42 Chapter 4 Proxy Ratios Versus Moho Depth And Chondrite-Normalised Spider-Diagrams 33 show the complete opposite to Figure (21C), where there is a very large positive Sr anomaly, a steep negative slope and an upward concavity shape within the HREE region. This suggests that hornblende is the main residual phase, with very little plagioclase. The plots of shoshonites (Figure 21E) show high levels of LREE, a very steep slope, and a negative Sr anomaly. They show similarity to the garnet signatures of Figure (21B), but the negative Sr anomaly shows that plagioclase was fractionating as a primary phase, not garnet. The plot within Figure (21F) shows a steep slope and high LREE, which indicates that garnet is the main residual mineral, but it also has a negative Eu anomaly, which may suggest that plagioclase was also involved, perhaps during fractionation at upper crustal levels. 4.4 Anomalous Groups The two groups that do not follow normal trends in the Sr/Y and LREE/HREE versus Moho diagrams are the same groups of volcanoes that can be separated by the Chondrite-normalised REE+Sr+Y spider-diagrams. The volcanoes that belong to the Mt Shasta group are those that contain hornblende as a residual mineral phase (Figure 21D). The volcanoes that belong to the Adatara group are those that contain plagioclase as the residual phase (Figure 21C). Therefore, when proxy ratios are being used to determine the Moho depth, Chondrite-normalised REE+Sr+Y spider-diagrams should also be used to determine what residual phases are present at some stage in the partial melting or fractionation history. This is important, for these mineral assemblages may not be present in the sampled rocks, yet their presence is as crucial for determining the shape of the REE+Sr+Y patterns. Without this information, the proxy values for Moho depths could be wrong.

43 Chapter 4 Proxy Ratios Versus Moho Depth And Chondrite-Normalised Spider-Diagrams 34 Figure 21. Chondrite-normalised REE+Sr+Y spider-diagrams divided into groups of similar shapes and slopes. (A) plagioclase + hornblende signature, (B) garnet signature, (c) plagioclase signature, (D) hornblende signature, (E) shoshonitic, (F) Garnet + Plagioclase.

44 Chapter 5 Discussion 35 CHAPTER 5 DISCUSSION It was shown in chapter 4 that a systematic trend exists between trace element ratios (Sr/Y, HREE/LREE and Ca/Sr) and Moho depth, reflecting changes in the residual mineral assemblage (plagioclase, hornblende and garnet) for the majority of andesitic volcanoes. As with most natural systems, there are usually anomalous results. In this case, the volcanoes do not follow the average trend. These include Mt Shasta, Adatara and the Shoshonitic groups. The processes that generate these anomalous groups are discussed below. 5.1 Shoshonites The shoshonitic group of volcanoes do not show many common properties with the typical calc-alkaline/tholeiitic andesites. Shoshonites can be distinguished from common calc-alkaline rocks by very high K 2 O, higher Ba, Rb, Sr, V, Cr, Ni and LREE/HREE (Rock et al. 1991). This group are located inboard from the typical arc system and their geochemical properties could result from partial melting of an enriched subcontinental lithospheric mantle (e.g. Déruelle 1991). Since shoshonitic magmas are already highly concentrated in LREE and Sr, fractionation at the Moho would have a minimal effect on the total abundance of these elements. Therefore, the volcanic products of lithospheric melting cannot be used to calculate Moho depths. 5.2 Effect of H 2 O In chapter 4, it was shown that the anomalous groups of calc-alkaline rocks (Mt Shasta and Adatara) have unusual H 2 O contents. The Mt Shasta group contains 4.5 to >8.0 wt% H 2 O, whereas the Adatara group from the trench-side Honshu arc have <0.5

45 Chapter 5 Discussion 36 wt% H 2 O. The average is 2 to 4 wt% H 2 O, such as Parícutin (Gill 1981). Since H 2 O has a similar effect to pressure on the stability of plagioclase and hornblende in andesites (Müntener et al. 2001), the H 2 O content should be also be considered in determining the Moho depth. Figure (22) shows increasing Ca/Sr, with decreasing La/Yb that reflects increasing residual plagioclase involvement, and decreasing residual hornblende and garnet involvement at shallow Moho depths. This is predicted in Chapter 1. However whereas the graph also shows a general trend from low Moho to high Moho with increasing La/Yb the increase is not systematic. For example, those volcanoes defined by open red circles (Figure 22) do not follow the pattern, yet they follow the hyperbolic trend. Because these volcanoes have similar Moho depths (pressures) and since the processes that destabilise plagioclase are increasing pressure and/or H 2 O, this variability could be due to changes in H 2 O. Evidence for this possibility is that all the volcanoes with anomalously high Ca/Sr are those with anomalously low H 2 O (Adatara group). As all volcanoes show a systematic hyperbolic trend from high La/Yb and Low Ca/Sr to low La/Yb and high Ca/Sr, this plot suggests that low H 2 O enhances plagioclase stability over that of hornblende. The La/Sm against La/Yb diagram (Figure 23) also shows a systematic trend that generally reflects the transition from residual plagioclase to hornblende to garnet, where increasing La/Sm represents an increasing role for hornblende and increasing La/Yb represents an increasing role for garnet. However, like Figure (22), a general increase in La/Yb reflects higher Moho, but the anomalous volcanoes are those with high and low H 2 O, as well as the shoshonites.

46 Chapter 5 Discussion 37 Figure 22. Ca/Sr versus La/Yb diagram, representing changes in residual plagioclase versus hornblende and garnet with respect to Moho depth. Figure 23. La/Sm versus La/Yb diagram representing changes in dominant residual minerals and H 2 O.