2.1 WiFeS data reduction

The WiFeS data were reduced using the WIFES pipeline (updated on 2011 November 21), which is based on the Gemini IRAF² package (version 1.10; IRAF version 2.14.1) developed by the Gemini Observatory for the integral-field spectroscopy.

Each CCD pixel in the WiFeS camera has a slightly different sensitivity, giving pixel-to-pixel variations in the spectral direction. This effect is corrected using the dome flat-field frames taken with a quartz iodine (QI) lamp. Each slitlet is corrected for slit transmission variations using the twilight sky frame taken at the beginning of the night. The wavelength calibration was performed using Ne-Ar arc exposures taken at the beginning of the night and throughout the night. For each slitlet the corresponding arc spectrum is extracted, and then wavelength solutions for each slitlet are obtained from the extracted arc lamp spectra using low-order polynomials. The spatial calibration was accomplished by using so called `wire' frames obtained by diffuse illumination of the coronagraphic aperture with a QI lamp. This procedure locates only the centre of each slitlet, since small spatial distortions by the spectrograph are corrected by the WiFeS cameras. Each wavelength slice was also corrected for the differential atmospheric refraction by relocating each slice in $x$ and $y$ to its correct spatial position.

In the N&S mode, the sky spectra are accumulated in the unused 80 pixel spaces between the adjacent object slices. The sky subtraction is conducted by subtracting the image shifted by 80 pixels from the original image. The cosmic rays and bad pixels were removed from the raw data set prior to sky subtraction using the IRAF task LACOS_IM of the cosmic ray identification procedure of van Dokkum (2001), which is based on a Laplacian edge detection algorithm. However, a few bad pixels and cosmic rays still remained in raw data, and these were manually removed by the IRAF/STSDAS task IMEDIT.

We calibrated the science data to absolute flux units using observations of spectrophotometric standard stars observed in classical mode (no N&S), so sky regions within the object data cube were used for sky subtraction. An integrated flux standard spectrum is created by summing all spectra in a given aperture. After manually removing absorption features, an absolute calibration curve is fitted to the integrated spectrum using third-order polynomials. The flux calibration curve was then applied to the object data to convert to an absolute flux scale. The $[$O I $]\,\lambda$5577Å night sky line was compared in the sky spectra of the red and blue arms to determine a difference in the flux levels, which was used to scale the blue spectrum of the science data. Our analysis using different spectrophotometric standard stars (LTT9491 and HD26169) revealed that the spectra at the extreme blue have an uncertainty of about 30% and are particularly unreliable for faint objects due to the CCD's poor sensitivity in this area.