Emulator-Assisted Calibration of a Semi-Analytic Galaxy Formation Model for the Roman Galaxy Redshift Survey

Mock catalogues for the Roman Galaxy Redshift Survey benefit from galaxy populations whose physical properties and emission-line observables are predicted within a single galaxy-formation framework. We present an emulator-assisted calibration of the semi-analytic model Galacticus for this purpose. Because direct posterior exploration with Galacticus is computationally prohibitive, we train Gaussian-process emulators to predict the observables entering the calibration likelihood. We calibrate the model simultaneously to stellar mass functions, the star-formation-rate function, size--mass relations, the black-hole mass--velocity-dispersion relation, the mass--metallicity relation, and H-alpha luminosity functions. Combining these observables breaks parameter degeneracies present in individual fits and can reveal tensions where different observables favour different regions of parameter space. The resulting model provides a broadly successful simultaneous description of the calibration data, including the abundance and redshift evolution of H-alpha emitters relevant to the Roman Galaxy Redshift Survey. A direct Galacticus calculation at the maximum a posteriori point gives similarly good agreement with the calibration data, providing a calibrated galaxy-formation model suitable for constructing physically consistent mock catalogues for Roman.