Combining spectroscopic and photometric surveys using angular cross-correlations III: Galaxy bias and stochastisity

TL;DR

This paper investigates how galaxy bias modeling and stochasticity influence the accuracy of cosmological measurements from combined spectroscopic and photometric surveys, emphasizing the benefits of overlapping survey strategies.

Contribution

It provides a detailed analysis of the impact of bias modeling, stochasticity, and survey overlap on cosmological parameter constraints in combined survey forecasts.

Findings

Bias modeling significantly affects forecasted constraints.

Overlapping surveys reduce the impact of bias uncertainties.

Adding counts-shear information stabilizes bias error estimates.

Abstract

In the first paper of this series, we studied the effect of baryon acoustic oscillations (BAO), redshift space distortions (RSD) and weak lensing (WL) on measurements of angular cross-correlations in narrow redshift bins. Paper-II presented a multitracer forecast as Figures of Merit (FoM), combining a photometric and spectroscopic stage-IV survey. The uncertainties from galaxy bias, the way light traces mass, is an important ingredient in the forecast. Fixing the bias would increase our FoM equivalent to 3.3 times larger area for the combined constraints. This paper focus on how the modelling of bias affect these results. In the combined forecast, lensing both help and benefit from the improved bias measurements in overlapping surveys after marginalizing over the cosmological parameters. Adding a second lens population in counts-shear does not have a large impact on bias error, but removing all counts-shear information increases the bias error in a significant way. We also discuss the relative impact of WL, magnification, RSD and BAO, and how results change as a function of bias amplitude, photo-z error and sample density. By default we use one bias parameter per bin (with 72 narrow bins), but we show that the results do not change much when we use other parameterizations, with at least 3 parameters in total. Bias stochasticity, even when added as one new free parameter per bin, only produce moderate decrease in the FoM. In general, we find that the degradation in the figure of merit caused by the uncertainties in the knowledge of bias is significantly smaller for overlapping surveys.