DATA DESCRIPTION

Raw Data Reduction

ANU/WiFeS

The observed single frame datacubes were generated and pre-processed using the PYTHON package pyWiFeS (Childress et al. 2014). This process includes cosmic ray removal, wavelength calibration, telluric correction, atmospheric extinction correction, and flux calibration using standard stars. The reduced individual exposure frames were further post-processed to produce the final mosaic data cubes, as described below.

  1. The barycentric correction was done for the wavelength solution of each exposure frame through astropy.AP.We then visually checked the relative wavelength calibration among frames through strong skylines and galaxy emission lines.
  2. We checked whether the WiFeS pipeline produced the correct variance array by checking the sky exposures, where the pipeline-produced errors should be equal to the standard deviations of their fluxes over the FOV at each wavelength. The ratio of the two was found to be around unity with a scatter of about 20%.
  3. For each target frame, the sky frame observed right before or after the target was used for sky subtraction. At each wavelength, the mean sky was estimated from the sky frame through the PYTHON code mmm.mmm and then was subtracted from the target frame. The error of the mean sky was then added quadratically to the error array of the target. We further masked the wavelength regions of strong skylines by identifying wavelength grids that were 10-sigma above the sky continuum in the sky datacubes. In practice, the mask was done by multiplying the variance at these wavelengths with -1.
  4. For a given grism, the sky-subtracted target frames obtained at different epochs were first reprojected to a common wavelength grid in a flux-conserving manner. The mean of these frames was then adopted as the final combined spectrum for this grism.
  5. To combine different grisms, a wavelength grid starts at 3500 AA and ends at 9000 AA with a constant interval in logarithm of 1/7000/2/ln(10) was created. All grisms were then projected onto this wavelength grid with flux conservation. U7000 and I7000 were normaized to B7000 and R7000, respectively. These two were then further normalized to each other. Note that we will further calibrate the flux based on the broad-band images so that the exact grism for the flux normalization does not matter.
  6. For the final combined datacube, an initial world coordinate system (WCS) was added using the input position and PA from the observation. The datacube was then rotated to have north up and east to the left.
  7. A synthetic broad-band image in the r band was created from the datacube and then compared to the observed broad-band image from SkyMapper (Onken et al. 2024) to further update the WCS. The final accuracy of WCS is less than 1 arcsecond, limited by the FOV or the lack of point sources in the field.
  8. For the final flux calibration, we produced synthetic v, g, r, and i images from the datacube and measured their synthetic photometry within large apertures. We then compared this synthetic photometry with measurements from SkyMapper broad-band images taken in the same bands and with the same apertures. The filter curves for the four broad bands adopted cover the spectral range of our WiFeS spectra. We fitted a double power-law flux calibration curve to the relative flux ratios in bands detected in the synthetic and real broadband images.
  9. Finally, the datacubes were corrected for the MilkyWay (MW) foreground extinction by applying the extinction curve from Cardelli et al. (1989) with Rv = 3.1. The E(B-V) were from Dale et al. (2009), which they adopted from Schlegel et al. (1998).

VLT/MUSE

The MUSE pipeline (Weilbacher et al. 2020) automatically reduces the raw data under the workflow engine of the esoreflex environment (Freudling et al. 2013). Briefly, the pipeline corrects for bias, dark current, flat-field, and sky-flat. It further calibrates wavelengths by arc lamps, flux by standard stars, geometric and astrometric coordinates, as well as atmospheric and telluric extinction. For each frame, a sky background is observed for an off-target blank sky. Residual sky removal for individual frames: Our MUSE observations were carried out under all weather conditions. To improve the flux calibration, we carried out further post-processing on the individual frames. Following steps are adopted.

  1. Firstly, to account for possible variations in sky brightness between targets and off-target sky pointings, we used broad-band images to guide additional sky subtraction in the target frames. The broad-band images were collected from the data archives of DESI (Dey et al. 2019), Pan-STARRS (Chambers et al. 2016), or SkyMapper (Onken et al. 2024). Given that DESI images have deep exposure, good S/N, and cover most of our targets, flux calibration was primarily based on DESI, then Pan-STARRS and SkyMapper. Based on the white images of each exposure produced by the MUSE pipeline, the WCS of all frames were matched to the r-band images by aligning (like-) point sources, then they were reprojected to a common frame by PYTHON package of reproject.reproject interp.
  2. Based on IRAC 3.6 μm images, broad r-band images, and MUSE white images, we selected several clean backgrounds and subtracted their median spectra from the entire datacube. This process helps reduce residual sky after the off-target sky subtraction performed by the pipeline. The standard deviations of the subtracted median sky spectra were added to the errors of datacubes. If no clean background was available within the FOV, this step was skipped. For some galaxies where diffuse optical light fills the entire FOV, we still performed background subtraction, following the principle of choosing the lesser of two evils.
  3. Calibration of the spectral shape of individual frames: For each target frame, we selected an aperture covering the main body of the target within the MUSE FOV and used the median spectra within this aperture as the integrated spectrum. We found that each frame still showed discrepancies in absolute flux and, particularly towards the blue end, variations in spectral shape. To correct the former, we convolved the integrated spectrum of each frame with the filter curves from the DESI, Pan-STARRS, or SkyMapper surveys to obtain synthetic photometry. This synthetic photometry was then compared to the observed photometry from broad-band images to correct the absolute flux. To calibrate the spectral shape, we first discarded frames with either negative fluxes over a significant portion of the wavelength range or highly complex spectral shapes. For the remaining frames, we selected the frame with a smooth median spectrum and a synthetic broadband color similar to the observed one as the reference.Each spectrum was smoothed by calculating the median over wavelength bins of 150 AA. The ratio of each spectrum to the reference was fitted with a 3rd-order polynomial function. The fitting wavelength range stopped at 8800 AA 481 to avoid strong skylines but still covered the Caii triplet. We visually inspected each fit, and when the fitted polynomial showed distorted tails, a simple linear function was used instead. Each pixel of every frame was then scaled by the corresponding fitted ratios, and these ratios were also applied to the errors to maintain the S/N. Finally, we combined all frames to produce the final datacube, weighting by exposure time.
  4. Final Flux Calibration: We produced synthetic photometry from the final datacube and compared it to the broad-band images within a large aperture that covered the main body of the galaxy. The ratio between the two was fitted with a 3rd-order polynomial function or, if the polynomial resulted in distorted tails, a linear function. This fitted ratio was then applied to all pixels to achieve the final flux calibration.
  5. The datacubes were 2 x 2 pixel rebinned, resulting in a final spatial sampling of 0.4" x 0.4". This process promotes the S/N of each single spectrum. Besides, the wavelengths were also resampled to a constant interval in logarithm of 1/2500/2/ln(10).
  6. The WCS was corrected by comparing the position of the brightest point source in the MUSE white image with that in the Pan-STARRS, DESI, or SkyMapper images, again.
  7. Finally, the MW foreground extinction was corrected with Cardelli et al. (1989) extinction curve and adopted E(B-V) from Dale et al. (2009).

Spectra Analysis & High-level Datacubes

We generated high-level data products for DGIS using the Galaxy IFU Spectroscopy Tool (GIST, Bittner et al. (2019)) 1. The PYTHON-based GIST pipeline reads IFS datacube, performs spaxel binning using Voronoi, fits spectra with pPXF, fits emission/absorption lines with GandALF (Gas and Absorption Line Fitting software, Sarzi et al. (2006)), and extracts stellar population properties and non-parametric SFH from pPXF. The pipeline provides spaxel- and/or bin-level data products, including stellar and gas kinematics, emission and absorption line fluxes, as well as stellar age, stellar metallicity, and SFH. Furthermore, GIST provides an interactive visualization tool Mapviewer, which facilitates the quick generation of various parameter maps and allows users to check pixel-specific spectra and their spectral fits. For spectral fitting, we employed the full HR-PYPOPSTAR stellar template with Chabrier (2003) IMF. Emission lines and skylines were masked during the continuum fitting process. Pixels with a continuum S/N lower than 1.5 were excluded, and an S/N of 20 was required after Voronoi binning. The pipeline was executed at both the SPAXEL and BIN levels.