KiDS-Legacy calibration: unifying shear and redshift calibration with the SKiLLS multi-band image simulations

TL;DR

KiDS-Legacy aims to achieve percent-level weak-lensing calibration by unifying shear and redshift calibrations in an end-to-end, multi-band image-simulation framework. SKiLLS combines SURFS-Shark cosmological simulations with COSMOS-derived morphologies and realistic KiDS+VIKING imaging to produce nine-band mock catalogs and images, enabling joint shear and photometric redshift estimates. The study demonstrates robust shear calibration against blending, PSF modelling, and observational variations, while highlighting residual higher-order effects that will matter for future surveys and motivating Metacalibration approaches. Overall, SKiLLS provides a powerful, transferable pipeline for precise tomographic weak lensing analyses in KiDS and future Stage IV experiments.

Abstract

We present SKiLLS, a suite of multi-band image simulations for the weak lensing analysis of the complete Kilo-Degree Survey (KiDS), dubbed KiDS-Legacy analysis. The resulting catalogues enable joint shear and redshift calibration, enhancing the realism and hence accuracy over previous efforts. To create a large volume of simulated galaxies with faithful properties and to a sufficient depth, we integrated cosmological simulations with high-quality imaging observations. We also improved the realism of simulated images by allowing the point spread function (PSF) to differ between CCD images, including stellar density variations and varying noise levels between pointings. Using realistic variable shear fields, we accounted for the impact of blended systems at different redshifts. Although the overall correction is minor, we found a clear redshift-bias correlation in the blending-only variable shear simulations, indicating the non-trivial impact of this higher-order blending effect. We also explored the impact of the PSF modelling errors and found a small yet noticeable effect on the shear bias. Finally, we conducted a series of sensitivity tests, including changing the input galaxy properties. We conclude that our fiducial shape measurement algorithm, lensfit, is robust within the requirements of lensing analyses with KiDS. As for future weak lensing surveys with tighter requirements, we suggest further investments in understanding the impact of blends at different redshifts, improving the PSF modelling algorithm and developing the shape measurement method to be less sensitive to the galaxy properties.

Figures (31)

• Figure 1: Number of galaxies per square degree per $0.1$ magnitude in the input apparent magnitudes. The green dashed lines are from the original SURFS-Shark mock catalogue, whilst the blue solid lines denote the modified results. The red solid lines correspond to the COSMOS2015 observations with flags applied for the UltraVISTA area inside the COSMOS field after removing saturated objects and bad areas ($1.38~{\rm deg}^2$ effective area, Table 7 of Laigle2016ApJS). The analytical fitting result in the $r$-band (black dashed line) is from Conti2017MNRAS. The $g$-band photometry is not in the COSMOS2015 catalogue and, thus, not shown in the plot. We note that the COSMOS2015 catalogue is incomplete at $K_s\gtrsim 24.5$Laigle2016ApJS.
• Figure 2: Flowchart summarising the algorithm to construct the SKiLLS input mock catalogue. The SKiLLS galaxies inherit the synthetic multi-band photometry and $N$-body 3D positions from the SURFS-Shark simulations, whilst the morphology is learned from the observations in the COSMOS field using an algorithm based on the vine-copula modelling (see Sect. \ref{['Sec:inputGalShape']} for details).
• Figure 3: Comparison of the overall magnitude-morphology relations in several redshift bins. The red solid and blue dashed lines denote the training and target samples, respectively. The left panel shows the mean half-light radius as a function of $r$-band magnitude, whilst the right panel presents the mean ellipticity as a function of $r$-band magnitude. The statistical uncertainties shown are calculated from 500.0 bootstraps. The left panel also shows the histograms of the normalised magnitude distributions, demonstrating that the extra high-redshift bright galaxies in the simulation contribute little to the overall population.
• Figure 4: Two-dimensional kernel density plots of morphological parameters in several magnitude bins. The red solid and blue dashed lines denote the training and target samples, respectively. The left panel shows the correlation between the size and ellipticity, whilst the right panel presents the correlation between the size and Sérsic index. The plotted contour levels are $20\%$, $40\%$, $60\%$, $80\%$.
• Figure 5: Input magnitude distributions in the $r$-band for the six stellar catalogues used by SKiLLS. Labels indicate the pointing centres (RA, Dec), except for 'COllege', which denotes the stellar catalogue used by Kannawadi2019AA.