Should we believe the results of ultraviolet-millimetre galaxy spectral energy distribution modelling?

Galaxy spectral energy distribution (SED) modelling is a powerful tool, but constraining how well it is able to infer the true values for galaxy properties (e.g. the star formation rate) is difficult because independent determinations are often not available. However, galaxy simulations can provide a means of testing SED modelling techniques. Here, we present a numerical experiment in which we apply the SED modelling code MAGPHYS to ultraviolet-millimetre synthetic photometry generated from hydrodynamical simulations of an isolated disc galaxy and a major galaxy merger by performing three-dimensional dust radiative transfer. We compare the properties inferred from the SED modelling with the true values and find that MAGPHYS recovers most physical parameters of the simulated galaxies well. In particular, it recovers consistent parameters irrespective of the viewing angle, with smoothly varying results for neighbouring time steps of the simulation, even though each viewing angle and time step is modelled independently. The notable exception to this rule occurs when we use a Small Magellanic Cloud-type intrinsic dust extinction curve in the radiative transfer calculations. In this case, the two-component dust model used by MAGPHYS is unable to effectively correct for the attenuation of the simulated galaxies, which leads to potentially significant errors (although we obtain only marginally acceptable fits in this case). Overall, our results give confidence in the ability of SED modelling to infer physical properties of galaxies, albeit with some caveats.