PI: Dr Dani Leonard, co-Is: Dr Markus Rau, Dr Cora Uhlemann

Optimised modelling for precision cosmology with LSST and Euclid

Rubin Observatory’s Legacy Survey of Space and Time (LSST) and the European Space Agency’s Euclid satellite are both at the beginning of multi-year surveys which will deliver cosmological galaxy samples at unprecedented sky coverage and depth. The resulting statistical uncertainties for measurements of key cosmological observables weak gravitational lensing and galaxy clustering will be incredibly small. In principle, this will allow us to perform powerful tests of our standard cosmological model. In practice, it means that we must control the precision of our modelling to an unprecedented degree. This includes modelling not only the cosmological signal itself, but also the impact of systematic uncertainties on our measurements.

We have examined the impact of one key such systematic uncertainty: catastrophic outlier populations in photometric redshift distributions. LSST will rely on photometric data to estimate redshifts. The result is that degeneracies in galaxy spectra will create failure modes in redshift estimation, which can place whole populations of galaxies far from their true redshift. We examined this effect within a simulated LSST analysis, finding that a single un-mitigated realistic 5% outlier population resulted in significant biases in the inferred values of key cosmological parameters. However, we demonstrated that by using a composite likelihood method to mitigate this model misspecification, we can reduce this bias to less than 1-sigma.

This work considered only 2-point correlation functions as a summary statistics; we are now studying the effect of astrophysical systematic effects on 1-point statistics.

Figure: Biases in inferred cosmological parameter values for various simulated catastrophic outlier fractions, f_D, and analysis methods. From Mill, Leonard, Rau, Uhlemann, and Joudaki. Journal of Cosmology and Astroparticle Physics 2025.12 (2025): 037.