Statistical Treatment and Modeling of Geochemical Data of Volcanic Arcs

This contribution concerns two projects based on statistical analysis of geochemical data of lava samples from global arc systems combined with standard geochemical treatments in order to extract trends, regularities and structures not readily apparent when done by studying individual volcanoes. Part I is an article "The robustness of Sr/Y and La/Yb as proxies for crust thickness in modern arcs." This paper considers three volcanic arcs – the Aleutians, Central America, and the Andes – and applies our understanding of trace element partitioning for four key elements and their dependence on pressure (i.e., depth) to derive crustal thicknesses along arcs and arc segments. We used geochemical data from the EarthChem.org repository and combined these with recent published igneous rock compositions vs. depth studies to derive crustal thickness profiles along modern arcs. We compare our methods with geophysical surveys to assess the viability of the correlation and our techniques. That the study at least partly agrees with geophysics is a boon for the conceptual methods we devised. That disagreements exist are, one, challenges for future geophysical surveys and geochemical studies to resolve and, two, a chance to interpret our results to reimagine and to incorporate existing theory of crustal processes into a framework that is consistent with our results. Some arc's Moho may not have the sharp boundary that it is beneath continents but becomes an exchange interface with magma rising from the mantle injecting into the crust at the same time that magma cumulates (from fractionation) and crustal residues (from partial melting of the crust) founder into the mantle. The upper mantle beneath some active arc segment is then suffused with cumulates and restites separated from the lower crust such that seismic imaging of the Moho is difficult and ambiguous. Part II, "Fractionation and delamination in arc crust genesis," constrains models of fractional crystallization to generate lavas of the Alaska-Aleutian arc. The modeled dataset contains ~2,500 lava samples from ~30 volcanoes. We simulate first-order Earth processes that are heterogeneous, widely variable, and may not be in thermodynamic equilibrium, but that follow some principals that are statistically resolvable. This scheme reveals crustal processes that are plausible and produces cumulates that are similar to exposed mid and lower arc crust in Alaska and the Pakistani Himalayas. These modeled cumulates have variable compositions and densities to create a stratified crustal column as in exposed crust. Estimates of the physical characteristics and quantities of mineral assemblages can be inferred from the simulated crust to give insights into the mid and lower crust of the Alaska-Aleutian arc. For example, that the model solutions provide the quantity of the hydrous mineral amphibole in a crustal column allows me to estimate the amount of water stored in the crust, and explore the consequences of that. Furthermore, the results of fractionation modeling can be combined with approximate crustal geotherms and mineral densities to estimate the amount and rate that these dense cumulate masses are likely to sink into the upper mantle.