Barometers Behaving Badly II: a Critical Evaluation of Cpx-Only and Cpx-Liq Thermobarometry in Variably-Hydrous Arc Magmas

ABSTRACT The chemistry of erupted clinopyroxene crystals (±equilibrium liquids) have been widely used to deduce the pressures and temperatures of magma storage in volcanic arcs. However, the large number of different equations parameterizing the relationship between mineral and melt compositions and intensive variables such as pressure and temperature yield vastly different results, with implications for our interpretation of magma storage conditions. We use a new test dataset composed of the average Clinopyroxene-Liquid (Cpx-Liq) compositions from N = 543 variably hydrous experiments at crustal conditions (1 bar to 17 kbar) to assess the performance of different thermobarometers and identify the most accurate and precise expressions for application to subduction zone magmas. First, we assess different equilibrium tests, finding that comparing the measured and predicted Enstatite-Ferrosillite and KD (using Fet in both phases) are the most useful tests in arc magmas, whereas CaTs, CaTi and Jd tests have limited utility. We then apply further quality filters based on cation sums (3.95–4.05), number of analyses (N > 5) and the presence of reported H2O data in the quenched experimental glass (hereafter ‘liquid’) to obtain a filtered dataset (N = 214). We use this filtered dataset to compare calculated versus experimental pressures and temperatures for different combinations of thermobarometers. A number of Cpx-Liq thermometers perform very well when liquid H2O contents are known, although the Cpx composition contributes little to the calculated temperature relative to the liquid composition. Most Cpx-only thermometers perform very badly, greatly overestimating temperatures for hydrous experiments. These two findings demonstrate that the Cpx chemistry alone holds very little temperature information in hydrous systems. Most Cpx-Liq and Cpx-only barometers show similar performance to one another (mostly yielding root mean square errors [RMSEs] of 2–3.5 kbar), although the best Cpx-only barometers currently outperform the best Cpx-Liq barometers. We also assess the sensitivity of different equations to melt H2O contents, which are poorly constrained in many natural systems. Overall, this work demonstrates that Cpx-based barometry on individual Cpx only provides sufficient resolution to distinguish broad storage regions in continental arcs (e.g. upper, mid, lower crust). Significant averaging of Cpx compositions from experiments reported at similar pressures can re