Abell-360 merged-catalog schema

This file documents the columns of the per-tract merged catalogs produced by xlens.processor.merge.MergePipe:

File	Tract	Rows
`a360_merged_tract10463.parq`	10463	45,176
`a360_merged_tract10464.parq`	10464	25,104
`a360_merged_tract10704.parq`	10704	3,861

Storage format is Parquet (ArrowAstropy storage class); read with astropy.table.Table.read("...parq").

Each tract catalog has 74 columns. The four magnitude bands are g, r, i, z; the photometric zero point of the input fluxes is mag_zero = 31.4.

Shapelet kernel

All FPFS shapelet quantities (fpfs1_m00, fpfs1_m20, fpfs1_e1/2, {b}_flux_fpfs1, …) were measured with the fpfs1 Gaussian shapelet kernel of width

\sigma_{\rm shapelets1} = 0.54\,\text{arcsec}

set via MeasureCellCoaddsPipeConfig.fpfs.sigma_shapelets1 = 0.54 in cell_coadd_pipeline.yaml. This σ enters the Gaussian-aperture flux relation flux = m00 · 2π σ² (so the per-band fluxes inherit the same kernel width).

Coordinate convention

**The local sky frame uses the GalSim convention (+x = West, +y = North). Concretely:

g1, g2, e1, e2, … and every spin-2 quantity is expressed in this +x = West frame.
g2 flips sign under a (+x = West) → (+x = East) convention swap; every *_dg2 derivative therefore also flips sign under that swap.

1. Detection / position / selection (12 columns)

Column	Dtype	Unit	Description
`ra`	float64	deg (ICRS)	Right ascension of the detection.
`dec`	float64	deg (ICRS)	Declination of the detection.
`x1`	float64	arcsec	First-moment-derived sky-plane position (`+x = West`), measured relative to the tract origin. Divide by `pixel_scale` for tract-pixel units.
`x2`	float64	arcsec	Second axis of the same position pair (`+y = North`).
`x1_det`	float64	arcsec	First detection-axis position used by AnaCal.
`x2_det`	float64	arcsec	Second detection-axis position.
`wsel`	float64	—	AnaCal selection weight `w_sel`
`dwsel_dg1`	float64	—	First-order shear response of `wsel` to `g1`.
`dwsel_dg2`	float64	—	First-order shear response of `wsel` to `g2`.
`mask_value`	int32	bitmask	Bit-OR of bad-pixel and bright-star mask flags at the detection’s pixel.
`object_id`	int64	—	Stable per-object identifier. Unique across patches, tracts, and merged outputs.

2. Band-combined ellipticity and shear response (6 columns)

The band combination is m_combined = Σ_b w_b · m_band with Σ_b w_b = 1. MergePipe.config.band_weights selects the per-band weights (defaults to 1 / median({band}_flux_fpfs1_err)^2). e_i = m22{c,s} / (m00 + c0) where c0 = fpfs_c0 · 10^((mag_zero − 30) / 2.5). The local WCS shear + rotation correction is then applied to e1, e2 in place.

Column	Dtype	Unit	Description
`fpfs1_e1`	float64	—	Band-combined and WCS-corrected first ellipticity component.
`fpfs1_e2`	float64	—	Band-combined and WCS-corrected second ellipticity component (Flips sign if x-axis is flipped).
`fpfs1_de1_dg1`	float64	—	Diagonal shear response `∂e1 / ∂g1`.
`fpfs1_de1_dg2`	float64	—	Off-diagonal shear response `∂e1 / ∂g2`. (Flips sign if x-axis is flipped).
`fpfs1_de2_dg1`	float64	—	Off-diagonal shear response `∂e2 / ∂g1`. (Flips sign if x-axis is flipped).
`fpfs1_de2_dg2`	float64	—	Diagonal shear response `∂e2 / ∂g2`.

3. Band-combined shapelet moments and shear responses (6 columns)

Used downstream for size cuts ((m00 + m20) / m00 > 0.1) and weight response calculation.

Column	Dtype	Unit	Description
`fpfs1_m00`	float64	—	Band-combined zeroth shapelet moment. Flux-like.
`fpfs1_dm00_dg1`	float64	—	`∂m00 / ∂g1` (band-combined).
`fpfs1_dm00_dg2`	float64	—	`∂m00 / ∂g2`. (Flips sign if x-axis is flipped)
`fpfs1_m20`	float64	—	Band-combined size-like spin-0 shapelet moment.
`fpfs1_dm20_dg1`	float64	—	`∂m20 / ∂g1`.
`fpfs1_dm20_dg2`	float64	—	`∂m20 / ∂g2`. (Flips sign if x-axis is flipped)

4. Per-band fluxes (4 columns × 4 bands = 16 columns)

flux = m00 · 2π σ² (Gaussian-aperture flux derived from the shapelet m00; σ = sigma_shapelets1 set by the FPFS task). Magnitudes are derived as mag = mag_zero − 2.5 · log10(flux) with mag_zero = 31.4.

For each band b ∈ {g, r, i, z} the catalog has:

Column	Dtype	Unit	Description
`{b}_flux_fpfs1`	float64	linear flux (zero-pointed at `mag_zero`)	Per-band Gaussian-aperture flux for band `b`.
`{b}_dflux_fpfs1_dg1`	float64	flux	`∂flux / ∂g1` for band `b`.
`{b}_dflux_fpfs1_dg2`	float64	flux	`∂flux / ∂g2` for band `b`. (Flips sign if x-axis is flipped)
`{b}_flux_fpfs1_err`	float64	flux	1σ uncertainty on `{b}_flux_fpfs1` (Gaussian-aperture analytic propagation of `m00_err`).

5. Photo-z point estimates (7 keys × 5 distortions = 35 columns)

Joined onto the catalog by MergePipe._join_photoz (object_id-aligned) from the per-patch deep_cell_coadd_anacal_fzb_point produced by photoZPipe. Each row of the photo-z table holds five FlexZBoost estimates: the undistorted call (_0) plus the four ±dg distortions in g1 (_1p, _1m) and g2 (_2p, _2m), evaluated at dg = 0.01. If x-axis is flipped, the _2p and _2m suffixes are swapped, since g2 reverses sign in the LSST convention.

The same seven keys appear with each of the five suffixes s ∈ {_0, _1p, _1m, _2p, _2m}; combined that’s 7 × 5 = 35 columns.

Key (column = `key + suffix`)	Dtype	Unit	Description
`zbest{s}`	float32	redshift	Per-object best-estimate redshift (FlexZBoost best estimate).
`zmode{s}`	float32	redshift	Mode of the FlexZBoost p(z).
`z025{s}`	float32	redshift	2.5th percentile of the FlexZBoost p(z).
`z160{s}`	float32	redshift	16th percentile of the p(z).
`z500{s}`	float32	redshift	50th percentile (median) of the p(z).
`z840{s}`	float32	redshift	84th percentile of the p(z).
`z975{s}`	float32	redshift	97.5th percentile of the p(z).