KiDS-1000 cosmic shear reanalysis using MetaCalibration

TL;DR

The paper addresses potential biases in cosmic shear measurements by applying a MetaCalibration-based pipeline to KiDS-1000 and comparing results to lensfit. By leveraging SKiLLS simulations for validation and PSF-circularization to control additive bias, the authors obtain smaller multiplicative biases and a higher effective source density, yielding a tighter $S_8$ constraint of $0.789^{+0.020}_{-0.024}$, while remaining in mild tension with Planck. The analysis demonstrates robustness across PSF leakage corrections, blending effects, and morphology sensitivity, with consistent COSEBI and two-point statistics and a ~28% improvement in cosmological constraints relative to previous KiDS-1000 work. These results suggest the Planck–KiDS tension is not driven by shear calibration and highlight MetaCalibration as a promising path for KiDS-Legacy analyses with larger data sets and refined redshift calibrations.

Abstract

A number of cosmic shear studies have reported results that are in mild tension with the Planck cosmic microwave measurement. To explore if this can be caused by biases in the shear estimation, we revisit the analysis of data from the Kilo-Degree Survey (KiDS) using an alternative shape measurement pipeline that is more robust to uncertainties in the calibration. To this end, we present an implementation of MetaCalibration, and compare its performance to that of lensfit, which has been used in previous analyses of these data. We find that the multiplicative bias is reduced, especially for the most distant redshifts, as derived from multi-band image simulations designed to match the KiDS data (SURFS-based KiDS-Legacy-Like Simulations: SKiLLS). For all tomographic bins we obtain a multiplicative bias $|m|<0.017$, with negligible additive bias. Importantly, the calibration has a negligible sensitivity to key galaxy properties. The resulting robust shear estimates were used to obtain cosmological parameter constraints. We find that the parameter $S_8\equiv σ_8 \sqrt{Ω_\mathrm{m}/0.3} =0.789_{-0.024}^{+0.020}$ is consistent with the previous KiDS-1000 lensfit constraint of $S_8=0.776^{+0.029 +0.002}_{-0.027-0.003}$ (statistical + systematic errors). Thanks to the higher effective source density, the constraining power is improved by about 28%. The difference in $S_8$ with the Planck value remains at a similar level, 1.8$σ$, implying that it is not caused by the shear measurements.

Figures (30)

• Figure 1: Comparison of source features between KiDS-1000 and SKiLLS selected for the metacal catalogue for each tomo-bin (any weighting is not applied): r-band magnitude (top), size (middle), and ellipticity (bottom). The sixth bin shows the most deviation for r-band magnitude and ellipticity. Regarding size, the deviation gradually increases for high redshift bins.
• Figure 2: Schematic work flow of the MetaCalibration pipeline, which starts with the same input catalogue that was used in previous KiDS-1000 analyses.
• Figure 3: Performance of the circularizing procedure applied to a representative KiDS-1000 pointing (at RA=9.4$^\circ$, Dec =-32.1$^\circ$) with the ellipticities shown as the length of bars (see the reference on top: blue bar) and the ratio of $e_1$/$e_2$ as their angle. The model PSFs, which were evaluated at the positions of the detected galaxies (whose average is shown on the grid), show some variation across the field of view, although the overall level of PSF anisotropy is small ($|e| \lesssim 0.02$: left panel). After circularising, the PSF shapes are consistent with zero ($|e| \lesssim 10^{-4}$: right panel). Please note that the ellipticity scale of the right panel is 100 times larger than the left panel.
• Figure 4: The mean shear response, $R^\gamma$, as a function of apparent magnitude for the six tomographic bins of simulated SKiLLS galaxies (left) and real KiDS-1000 galaxies (right). We used ten equi-populated subsamples for each tomographic bin to compute the averages.
• Figure 5: Multiplicative bias as a function of tomographic bin estimated from fixed-shear simulations. We find that the biases for MetaCalibration (blue points) are small ($|m|<0.023$) and agree well with the level of bias expected for a perfect shape measurement method after accounting for selection bias (red points). The green points indicate the corresponding detection bias from SourceExtractor. For the first three bins, the multiplicative bias for lensfit (orange triangles) is similar, but it increases for more distant galaxies.