Beijing-Arizona Sky Survey (BASS)

There are significant changes in source detection and photometry between DR1 and DR2. In DR1, the source detection and corresponding photometry are performed on single-epoch images. However, we detect sources in stacked images in order to improve the detection efficiency in DR2. The DR2 photometry are made in single-epoch images based on prior positions of those sources.

Image stacking and source detection

As described in Zou et al. (2017), the full sky is equally divided into 96,775 blocks. These blocks are evenly spaced in declination. Each block has an area of 0.681×0.681 deg², equal to an image size of 5400×5400 with a pixel scale of 0.454”. There are overlaps of 0.02° in both right ascension and declination between adjacent blocks. The stacked image is generated with a simple tangent-plane projection around the block center.

We create four stacked images: three images for g, r, and z bands and one composite from these three stacks. The composite image is combined from g, r, and z-band stacked images with flux scales of 0.65, 1.0, and 1.5, respectively, which make g - r and r - z colors of F/G type stars close to zero. These flux scales are also used for generating color pictures. For a specified band and block, we collect single-epoch images that are connected to the block. These images are resampled and reprojected by Swarp, and then combined by median to form a stacked image and corresponding weight image \citep{ber02}. The sky background map for each single-exposure image is estimated by using a mesh with a grid size of 410 pixels and masking large objects from the Third Reference Catalogue of bright galaxies (RC3) and New General Catalogue (NGC). The flux of each single-epoch image is scaled to make the stack image having a zero point of 30. The single-epoch images used for stacking should satisfy the following conditions: (a) exposure time > 30 s; (b) seeing < 3.5” for BASS and < 2.5” for MzLS; © zero point RMS error < 0.2; (d) number of stars used for calculating the zero point > 50; (e) astrometric RMS error in both R.A. and decl. < 0\arcsec.5; (f) 5σ depths are at most 1.5 mag shallower than the required one; (g) sky ADU < 15000 for BASS and < 25000 for MzLS; (h) transparency > 0.5. Here the transparency is related to the level of atmospheric extinction. The figure below gives examples of a g-band stacked image, its weight map, and corresponding color image composed of three-band stacked images.

Sources are first detected in the composite image, and then detected separately in g, r, and z-band stacked images if existing. Sources detected at least twice in four combined images are regarded as true objects. We make objects unique by cross-matching sources detected in adjacent blocks with a separation of 2 pixels (0.91″).

Methods of photometry

We have developed a new Python-based package for photometry named ``AstroPhot“ (Zou et al. in preparation). It can be used for general purpose, but at present it is just applied to BASS DR2. The current version can make accurate measurements of circular aperture, elliptical aperture, and PSF magnitudes at specific positions. However, model measurements with galaxy profiles, such as deVauculeurs, exponential and composite profiles, is still in development.

Segmentation and shape measurements

Different kinds of magnitude measurements are performed on a single-epoch image for objects detected on stacked images. These objects are first projected onto the image. A watershed segmentation algorithm is then adopted to separate signal pixels belonging to each object. The signal pixels are identified with a detecting threshold at 1$\sigma$ above the sky background after the image is smoothed with a Gaussian kernel of σ = 1.5 pixels. The global sky background and its RMS map are calculated in mesh grids. Based on the segment of each object, we calculate the centroid, shape parameters, and refined center using a Gaussian-kernel window. The shape parameters describe the shape of an object as an ellipse, including semi-major (A) and semi-minor (B) axis lengths and position angle (PA). They are also called as 1σ elliptical parameters.

Circular aperture photometry

We adopt 12 apertures for circular aperture photometry with radius ranging from 3 to 40 pixels. These aperture sizes are the same as used in the South Galactic Cap $u$-band Sky Survey, which ultilized the Bok telescope and 90Prime camera to perform a u-band imaging survey. The below table shows the aperture radii in both pixel and arcsec.

Aperture No. 1 2 3 4 5 6 7 8 9 10 11 12
radius in pixel 3 4 5 6 8 10 13 16 20 25 30 40
radius in arcsec (BASS) 1.36 1.82 2.27 2.72 3.63 4.54 5.90 7.92 9.08 11.35 13.62 18.26
radius in arcsec (MzLS) 0.78 1.04 1.31 1.57 2.09 2.61 3.39 4.18 5.22 6.53 7.83 10.44

Isophotal and Kron elliptical aperture photometry

Isophotal magnitudes are derived by simply integrating pixel fluxes within segments. For better magnitude measurements of galaxies, we estimate an appropriate elliptical aperture for each object. The elliptical aperture should enclose most flux. We call it ``Kron aperture”, which was introduced in Kron (1980). The Kron aperture size is described by the Kron radius. This radius is determined in a similar way as in SExtractor\citet{ber96}. First, an characteristic radius $r_1$ is calculated as the first-order moment within an large ellipse, whose size is 6 times the 1σ ellipse: $r_1 = ΣrF( r )/ΣF( r ), where r is the elliptical distance of a pixel to the center and F( r ) is the corresponding flux in this pixel. We set the Kron radius rk to be 2.5r1. The Kron aperture magnitudes are measured by integrating the fluxes of pixels within the Kron ellipse. The major and minor axis lengths of the Kron ellipse are computed as rk√e and rk/√e, respectively, where e is the elongation. To excluding extreme small and large apertures, we set lower and upper limits of the kron radius to be 3r0 and 10r0, where r0 = √(AB). It is reported that more than 94% of the light for galaxies is located in the Kron ellipse, almost independent of their magnitudes. For both circular and Kron aperture photometry, special handling is done for objects when they are contaminated by nearby sources in order to improve the photometric accuracy. In the aperture, the pixels occupied by nearby objects are filled by mirroring the opposite pixels relative to the object center. In addition, each pixel is divided in 5×5 sub-pixels for more accurately counting the fluxes of pixels at the boundary of the aperture.

PSF photometry

PSF magnitudes are measured with the PSF model derived by PSFEx. PSFEx models the PSF as a linear combination of basis vectors. The pixel basis and automatic sampling step are selected in this software. We use a third-degree polynomial to model the position-dependent variation of the PSF. The PSF size is 45×45, which is about 12 times the seeing FWHM. Below table lists some key input parameters in PSFEx. When fitting with the PSF model, the pixels belonging to the segment of each object are used. If objects are not isolated, i.e. their segments are connected to each other, they are fitted simultaneously. We use k-means clustering to iteratively group objects. First, two groups are generated by k-means clustering algorithm. Then the members in each group are divided into two subgroups using the same algorithm. All groups are divided in this way unil the number of members in each group is not larger than 3. Finally, all groups are ranked by their brightest members, and the PSF magnitudes in each group are measured simultaneously.

Parameter Value Description
BASIS_TYPE PIXEL_AUTO Type of basis vector
PSF_SAMPLING 0.0 Sampling step in pixel units
PSF_ACCURACY 0.01 Accuracy to expect from PSF “pixel” values
PSF_SIZE 45,45 Image size of the PSF model
PSF_RECENTER Y Wether to allow recentering of PSF-candidates
CENTER_KEYS X_IMAGE,Y_IMAGE Catalogue parameters for source pre-centering
PSFVAR_DEGREE 3 Polynomial degree for position-dependent variation
SAMPLE_FWHMRANGE 2.0,20.0 Allowed FWHM range
SAMPLE_VARIABILITY 0.5 Allowed FWHM variability
SAMPLE_MAXELLIP 0.3 Maximum ellipticity

Coadding measurements

Based on the detections on stacked images, we perform circular aperture, isophotal, Kron elliptical aperture, and PSF flux measurements on single-epoch images. As described in our DR1 paper, the coordinates of objects are corrected with astrometric residuals, which mainly origin from poor charge transfer efficiencies of detectors. The mount of corrections is about a few percent of an arcsec. All magnitudes are aperture-corrected with growth curves produced by circular aperture photometry. In addition, the magnitudes are corrected with photometric residuals, which are mainly caused by the focal distortion, improper flat-fielding, and scattered light. These astrometric and photometric residuals are obtained by comparing the coordinates and magnitudes with those in the Gaia DR1 and PS1 catalogs, respectively.

The parameters measured in single-epoch images are merged to generate co-added catalogs. For each object, we take error-weighted averages of the refined centers, shape parameters, and fluxes (equivalently magnitudes). The errors of the averages are computed as weighted errors. We also calculate the parameter standard error, which is the RMS divided by square root of the number of measurements. The effective seeing and sky background, number of exposures, and average, minimum and maximum of the Julian day when observations were taken are recorded in the catalogs.

datarelease/dr2/dr2_phot/home.txt · Last modified: 2017/12/23 22:14 by Zou Hu
CC Attribution-Share Alike 4.0 International
Powered by PHP Driven by DokuWiki Recent changes RSS feed Valid CSS Valid XHTML 1.0 Valid HTML5