A reassessment of strong line metallicity conversions in the machine learning era

ABSTRACT Strong line metallicity calibrations are widely used to determine the gas phase metallicities of individual H ii regions and entire galaxies. Over a decade ago, based on Sloan Digital Sky Survey Data Release 4, Kewley & Ellison published coefficients third-order polynomials that can be convert between different strong for global galaxy spectra. Here, we update work in three ways. First, by using newer data release, approximately double number galaxies polynomial fits, providing statistically improved coefficients. Second, include calibration suite five additional diagnostics have been proposed last were not included Ellison. Finally, develop new machine learning approach converting calibrations. The random forest (RF) algorithm is non-parametric therefore more flexible than conversions, due its ability capture non-linear behaviour data. RF method yields same accuracy as (updated) but has significant advantage single model applied over wide range metallicities, without need distinguish upper lower branches R23 trained made publicly available use community.