Field-level simulation-based inference with galaxy catalogs: the impact of systematic effects

It has been recently shown that a powerful way to constrain cosmological parameters from galaxy redshift surveys is to train graph neural networks to perform field-level likelihood-free inference without imposing cuts on scale. In particular, de Santi et al. [58] developed models that could accurately infer the value of Ω_m from catalogs that only contain the positions and radial velocities of galaxies that are robust to different astrophysics and subgrid models. However, observations are affected by many effects, including (1) masking, (2) uncertainties in peculiar velocities and radial distances, and (3) different galaxy population selections. Moreover, observations only allow us to measure redshift, which entangles the galaxy radial positions and velocities. In this paper we train and test our models on galaxy catalogs, created from thousands of state-of-the-art hydrodynamic simulations run with different codes from the CAMELS project, that incorporate these observational effects. We find that while such effects degrade the precision and accuracy of the models, the fraction of galaxy catalogs for which the models retain high performance and robustness is over 90%, demonstrating the potential for applying them to real data.