9:00 AM - 9:15 AM
[MGI29-01] Temporal variations of the magma plumbing system of Fuji Volcano, Central Japan, decoded by multivariate statistical analyses of lava geochemistry
Keywords:subduction related igneous activity, chemical composition, unsupervised machine learning, Fuji volcano
Fuji Volcano, central Japan, has erupted basaltic lavas throughout its history of one hundred thousand years, with a few exceptions of andesitic to dacitic activities. However, the cause of the high eruption rate with a limited range in chemical composition is poorly understood. Regarding geochemistry, the limited compositional range brings some difficulties in resolving the magmatic processes inherent in the compositional data of the lavas. In such a case, Independent Component Analysis (ICA) is a useful method to explore the hidden structure in the multi-dimensional compositional data in spite of the limited range since ICA does not depend on the amplitude of data signals or the variances. In this study, ICA is applied to the major element compositional data of 997 lava samples from Fuji Volcano, to extract the independent components (ICs), i.e., a statistically independent set of compositional vectors, from the lava chemistry. A thermodynamic simulator (MELTS) is then used to interpret the ICs and the corresponding geological-magmatic processes, particularly the processes in magma plumbing system in the crust. Based on the interpretation of ICs, magmatic processes and their temporal changes through the history of the volcano can be discussed efficiently.
As a result, four ICs (IC1 to IC4) are obtained with ICA. The three of these are accounted for by crustal processes that are independent but overlapping. IC3 is a vector with increasing Ti, Fe, Na, K, & P and decreasing Al, Mg, & Ca, while constant Si. IC3 is interpreted as an indicator for the degree of crystallization of tholeiitic magma with the fractionation of a pair of pyroxenes and plagioclase crystals. IC2 is a vector with increasing Si & Al against decreasing Ti, Fe, Mg, & Ca. IC2 can be explained by a difference in water content that causes a variation in fractionated mineral assemblages. Pyroxenes and plagioclase are the main fractionated phases, and olivine and Fe-Ti oxides appear with higher water contents, leading to increasing Si & Al contents associated with suppression of plagioclase fractionation. IC4 is a vector with increasing Al against decreasing Mg. This trend can be explained by the addition of plagioclase and the subtraction of mafic minerals. Plagioclase modal composition correlates with the IC4 score of whole rock chemistry. According to the MELTS simulation results under various pressure-temperature-H2O conditions, the difference in crystallization pressure may produce a compositional vector close to IC4 independent of other ICs, via controlling the timing of plagioclase crystallization relative to mafic minerals Therefore, IC4 can be taken as a pressure indicator of crystallization.
As the effects of the P-T-H2O condition are separated by ICs, temporal changes in the condition of the magma plumbing system can be estimated in detail solely by lava chemistry. The pressure condition estimated from the IC4 score implies that the final depth of crystallization is around 300±200 MPa throughout the history of the Fuji volcano. However, in the Subashiri stage (8ka-), some lavas show a shallower pressure index, which was mainly derived from flank eruptions possibly associated with low-pressure crystal fractionation. Water content estimated from the IC2 score illustrates differences among the eruption stages. In the Hoshiyama stage (100-17 ka) and Subashiri stage, the average water content is greater than 1.5 %, but the water content is about 1% or less in the Fujinomiya stage (17-8 ka).
IC1 is a vector with increasing Ti, Na, K, & P against decreasing Mg & Ca while nearly constant Si. IC1 cannot be explained by simple crystal fractionation in the crust. The average IC1 score of each eruption stage is different between the Hoshiyama stage and the later stages. Whether IC1 requires variations in deeper processes and/or sources in the mantle will be tested using more constraints including trace elements and radiogenic isotope studies.
As a result, four ICs (IC1 to IC4) are obtained with ICA. The three of these are accounted for by crustal processes that are independent but overlapping. IC3 is a vector with increasing Ti, Fe, Na, K, & P and decreasing Al, Mg, & Ca, while constant Si. IC3 is interpreted as an indicator for the degree of crystallization of tholeiitic magma with the fractionation of a pair of pyroxenes and plagioclase crystals. IC2 is a vector with increasing Si & Al against decreasing Ti, Fe, Mg, & Ca. IC2 can be explained by a difference in water content that causes a variation in fractionated mineral assemblages. Pyroxenes and plagioclase are the main fractionated phases, and olivine and Fe-Ti oxides appear with higher water contents, leading to increasing Si & Al contents associated with suppression of plagioclase fractionation. IC4 is a vector with increasing Al against decreasing Mg. This trend can be explained by the addition of plagioclase and the subtraction of mafic minerals. Plagioclase modal composition correlates with the IC4 score of whole rock chemistry. According to the MELTS simulation results under various pressure-temperature-H2O conditions, the difference in crystallization pressure may produce a compositional vector close to IC4 independent of other ICs, via controlling the timing of plagioclase crystallization relative to mafic minerals Therefore, IC4 can be taken as a pressure indicator of crystallization.
As the effects of the P-T-H2O condition are separated by ICs, temporal changes in the condition of the magma plumbing system can be estimated in detail solely by lava chemistry. The pressure condition estimated from the IC4 score implies that the final depth of crystallization is around 300±200 MPa throughout the history of the Fuji volcano. However, in the Subashiri stage (8ka-), some lavas show a shallower pressure index, which was mainly derived from flank eruptions possibly associated with low-pressure crystal fractionation. Water content estimated from the IC2 score illustrates differences among the eruption stages. In the Hoshiyama stage (100-17 ka) and Subashiri stage, the average water content is greater than 1.5 %, but the water content is about 1% or less in the Fujinomiya stage (17-8 ka).
IC1 is a vector with increasing Ti, Na, K, & P against decreasing Mg & Ca while nearly constant Si. IC1 cannot be explained by simple crystal fractionation in the crust. The average IC1 score of each eruption stage is different between the Hoshiyama stage and the later stages. Whether IC1 requires variations in deeper processes and/or sources in the mantle will be tested using more constraints including trace elements and radiogenic isotope studies.