Open-source Python3 tools for Thermobarometry: Revealing the good, the bad and the ugly of determining P-T-X conditions in igneous systems

The chemistry of erupted minerals and melts are commonly used to determine the pressures, temperatures and H₂O contents of magma storage regions beneath volcanic centres. In turn, these estimates are vital for hazard assessment, to understand the formation of critical metal deposits, and to inform models of continental crust formation. In the last few decades, more than 100 empirical and thermodynamic expressions have been calibrated using measurements of phases in experimental studies where these intensive parameters are known. By collating these different models into a computationally-efficient, open-source Python3 package, Thermobar, we can critically assess the performance of thermobarometers in igneous systems, and propagate analytical errors. When we apply published models for different mineral equilibrium to a new experimental dataset not used in model calibration, we find that stated errors vastly underestimate the true uncertainty when these workflows are applied to natural systems.

Specifically, we find that realistic calculation workflows involving Clinopyroxene (Cpx) equilibrium (e.g., iterating pressure and temperature) have uncertainties spanning the entire crust in most tectonic settings. Using Thermobar functions to propagate analytical error using Monte Carlo simulations, we suggest that these large errors result from imprecise analyses of minor elements such as Na in experimental (and natural) Cpx. Common analytical conditions used for Cpx yield highly correlated pressure-temperature arrays spanning the entire crust, which have been incorrectly interpreted as trancrustal storage in natural systems. Insuffucient analyses of each phase in experimental products means that this analytical error is not sufficiently mediated by averaging, so reported mineral compositions deviate from the true phase composition. This impacts thermobarometer calibration, as well as assessment of these methods using test experimental datasets.

Overall, we demonstrate that the development of Python3 infrastructure for common quantitative workflows in volcanology is vital to allow rigorous error assessment and model intercomparison; such assessments simply aren’t feasible using traditional tools (e.g., Excel workbooks). Specific changes to analytical, experimental and model calibration workflows (e.g., higher beam currents and count times in Na) will be essential to produce a more robust dataset to calibrate and test the next generation of more precise and accurate Cpx-based barometers. In turn, this will enable more rigorous investigation of magma storage geometries in a variety of tectonic settings (e.g., distinguishing true transcrustal storage vs. storage in discrete reservoirs).