Mitigating Shear-dependent Object Detection Biases with Metacalibration

TL;DR

Metadetection extends metacalibration to account for shear-dependent object detection biases that arise in blended galaxy images. By shearing large regions and re-running detection, it computes a response that correctly calibrates shear measurements even when detections vary with shear, demonstrated in simulations that mimic DES and LSST densities. The method reduces multiplicative biases to sub-percent level, with worst-case coherent-space shear contributing only a few tenths of a percent; PSF variation across coadds does not pose a fundamental obstacle. The work also identifies operational challenges for applying metadetection to real survey data, including handling multiple detection catalogs and masking, and points to future work to robustly integrate metadetection into end-to-end analyses.

Abstract

Metacalibration is a new technique for measuring weak gravitational lensing shear that is unbiased for isolated galaxy images. In this work we test metacalibration with overlapping, or ``blended'' galaxy images. Using standard metacalibration, we find a few percent shear measurement bias for galaxy densities relevant for current surveys, and that this bias increases with increasing galaxy number density. We show that this bias is not due to blending itself, but rather to shear-dependent object detection. If object detection is shear independent, no deblending of images is needed, in principle. We demonstrate that detection biases are accurately removed when including object detection in the metacalibration process, a technique we call metadetection. This process involves applying an artificial shear to images of small regions of sky and performing detection on the sheared images, as well as measurements that are used to calculate a shear response. We demonstrate that the method can accurately recover weak shear signals even in highly blended scenes. In the metacalibration process, the space between objects is sheared coherently, which does not perfectly match the real universe in which some, but not all, galaxy images are sheared coherently. We find that even for the worst case scenario, in which the space between objects is completely unsheared, the resulting shear bias is at most a few tenths of a percent for future surveys. We discuss additional technical challenges that must be met in order to implement metadetection for real surveys.

Figures (7)

• Figure 1: Toy example of shear-dependent object detection in the presence of a PSF. In panel (a) two objects are present, convolved by a PSF with no shear. Contours represent constant brightness. In panel (b) the objects are sheared by $\gamma = (0.0, 0.1)$after the PSF convolution. The contour levels are the same as panel (a). In this case the inner contours for the two objects overlap before and after application of the shear. This is a general property of the shear transformation in the weak regime: surface brightness is preserved under shear, and because the mapping is one-to-one, the topology is also preserved. In panel (c) the shear is applied before the PSF convolution, which mimics real sky images. In this case the inner contours do not overlap after shearing, and two objects may be detected rather than one. For case (c) an object detection algorithm that identified connected regions above a threshold as a single object would manifest a shear-dependent object detection bias.
• Figure 2: Example images of simulated galaxies used for the pair tests presented in section \ref{['sec:sims:pairs']}. From left to right in the top row, the separations are 1.0, 1.5, 2.0 arcsec. From left to right in the bottom row the separations are 3.0 and 4.0 arscec. The pixel scale is 0.263 arcsec.
• Figure 3: Example images from the DES (left) and LSST (right) simulations. Each multicolor, $gri$-band image is approximately $\sim\!2.2$ arcmin on a side. The DES images have a pixel scale 0.263 arcsec and a PSF FWHM of $\sim\!1$ arcsec. The LSST images have a pixel scale of 0.2 arcsec and a PSF FWHM of $\sim\!0.8$ arcsec.
• Figure 4: Mean multiplicative shear bias measured for pairs of simulated galaxies (see § \ref{['sec:sims:pairs']} for details) at various separations. At each separation, a large number of trials was generated with random orientations of the pair. At 4.0 arcsec separation, two objects were detected in all cases. At 1.5 arcseconds two objects were detected in half the cases. At 1.0 arcsec a single object was detected in all cases. Red triangles represent standard metacalibration with MOF deblending for modeling all detected objects. Blue circles represent metacalibration+MOF with detection included as part of the process. Green pluses represent metacalibration with detection included but without deblending, and using simple weighted moments without PSF correction as the shear estimator. Very large biases are seen for standard metacalibration+MOF as detection becomes ambiguous, for example at 1.5 arcsec separations. When detection is included in the metacalibration process the biases are greatly reduced. The bias is reduced even in the case where no deblending was performed and no PSF correction or detailed object modeling were performed. This indicates that a majority of the bias is due to shear-dependent detection, not light blending or details of the object modeling.
• Figure 5: Same as Figure \ref{['fig:toy']} but the shear was applied to the objects without shearing the space between them. This models the extreme and unphysical case where two objects are in line-of-sight projection but sheared completely independently. The contours changed less after shearing in this case as compared to the contours in Figure \ref{['fig:toy']}. This case differs from the metadetection process, in which the entire image is sheared, including the space between images.