KIPAC Tea Talk: How Much Can We Trust Cosmology from Galaxy Surveys? From Pixels to Cosmological Inference

Abstract: The statistical power of galaxy surveys is growing rapidly, but our confidence in their cosmological constraints increasingly depends on controlling systematics. I will discuss this challenge through two complementary probes: photometric galaxy clustering and weak gravitational lensing.

Photometric clustering measures the angular distribution of galaxies projected over broad redshift kernels. Because observing conditions vary across a survey footprint, they alter detection, photometry, and sample selection, and therefore both the apparent galaxy density and its redshift distribution. I will show how the resulting anisotropic redshift kernels modify clustering correlations, and particularly, how that makes combining clustering and lensing a problem. 

Weak lensing traces matter through small, coherent distortions of galaxy images. As surveys become deeper, overlapping sources at different redshifts, i.e., blending, complicate both shear measurement and lensing kernel through which the signal is interpreted. I will show how this redshift mixing can be modeled efficiently using simulation-based emulators. Finally, I will present a practical simulation-based shear-inference framework that incorporates shear-dependent detection, selection, and blending within a unified likelihood.