Abstract

Contributed Talk - Splinter LargeScale   (MW-2235)

Forward-modelling galaxy surveys for next-generation cosmological measurements

Luca Tortorelli
LMU Munich

Stage IV experiments will provide the most stringent constraints ever achieved on the cosmological parameters that describe the growth of the large scale structure (LSS) of the Universe by measuring positions, magnitudes and shapes for billions of galaxies. The unprecedented statistical power is however limited by systematics, whose error budget dominates over the statistical uncertainties. The accurate determination of galaxy redshift distributions is possibly the dominant observational systematic in LSS cosmology as recent results from KiDS-Legacy, DES Y6 and re-analysis of HSC Y3 have shown. The forward-modelling of photometric and spectroscopic galaxy surveys, a method that bridges cosmology with galaxy evolution, is one of the most promising approaches to solve the problem of accurate galaxy redshift distribution estimates. This method generates physically motivated galaxy populations, maps them to multi-band observables, and produces realistic mock survey images (and spectra) that can be passed through the same detection, measurement, and selection steps as the data. In this talk, I will discuss the past (Tortorelli+18,20,21) and on-going efforts (Tortorelli+24, Tortorelli+25) in forward-modelling galaxy surveys, from the modelling of the galaxy population (GalSBI-SPS) to the simulation of images (UFig) and spectra (USpec), as well as the data required to calibrate this model. The model is currently being constrained against HSC Deep observations, whose depth and blending regime make them a valuable precursor of Stage-IV surveys. This allows us to test whether a physically grounded population model can reproduce the observed colour, magnitude, and size distributions precisely in the regime where conventional redshift calibration is most challenging. A central goal is to turn this constrained forward model into a robust prediction for galaxy redshift distributions, propagating selection effects and measurement biases self-consistently rather than treating them as external corrections. This will thus enable improved cosmological analyses of existing and upcoming datasets where we will be able to control the systematic related to accurate redshift distributions determination.