Speaker
Description
Stage-IV galaxy surveys will soon deliver data of unprecedented depth and volume. Extracting the maximum information from these data while keeping systematics under control poses new challenges for the field. In this talk, I will present GalSBI, an open-source framework that addresses both simultaneously by constructing highly realistic synthetic galaxy catalogs through simulation-based inference.
GalSBI combines parametric luminosity functions, morphologies, and SEDs with SHAM-OT — a novel subhalo abundance matching scheme based on optimal transport — to jointly model the photometric and spatial properties of observed galaxy populations. Model parameters are inferred by comparing realistic forward-modelled image simulations directly to HSC and DES data. The constrained framework simultaneously reproduces observed magnitudes, colors, sizes, redshift distributions, and clustering statistics with high fidelity.
Because GalSBI operates at the pixel level, selection effects and calibration systematics propagate naturally into any downstream data product, making it applicable to a wide range of analyses, such as photometric redshift estimation, blending and shape measurement calibration, or augmenting N-body simulations with realistic galaxy populations within simulation-based inference pipelines.