TERAPIX processing steps from the download of Elixir pre-processed CFHT images to the final stacked images and catalogues is illustrated in Fig. 3. The T0007 CFHTLS pre-processed images described in the previous section were transferred from CFHT and validated against the T0007 image lists. All Queued Service Observing validation flags archived at CFHT were also downloaded11 .
For the T0007 release we use Elixir “B5/SNLS” pre-reduced images. The images were first ingested with the Terapix processing pipeline Youpi, producing weight-maps and input catalogues. We use quality grades from previous CFHTLS releases for images which were already in the database; new images from the “L99” photometric calibration program and the VIPERS Director’s Discretionary Time observations (described later) were graded using the Youpi grading interface, described below.
Images were then divided into each of the four Wide and four Deep fields and processed with Scamp (?) in order to derive the astrometric as well as the initial photometric calibrations. As explained below, as a consequence of the new calibration scheme based on L99 images, the astrometric and photometric calibrations were performed separately.
Once aligned astrometrically, images are co-added by SWarp (?) using the Scamp initial photometric rescaling. For the Deep field, images lists are derived using seeing and photometric rescaling constraints. Two series of Deep stacks are produced using two images combination schemes: median and sigma-clipped, denoted throughout the current document as “MEDIAN” and “SIGWEI” stacks.
Because of the inherent limits in using a photometric calibration derived solely from the pre-existing CFHT survey photometric calibration (lists of observations previously flagged as photometric by the CFHT and photometric header calibration written by Elixir) in T0007 we use a different approach to calibrating the survey. Both Deep and Wide stacks are tied to the SNLS photometric system using photometric standards which lie within the CFHTLS Deep field (?). For the Deep fields, which contain the Deep field calibrators, we can rescale images directly to the SNLS system, whereas for the CFHTLS Wide fields we use L99 short photometric exposures which are bracketed by observation of the Deep fields containing the calibrators. These L99 observations are then used to rescale the Wide tiles. The original SNLS Vega photometric system is finally converted to the natural AB system which has been used in previous CFHTLS releases (4.4).
Several types of final catalogues are produced: individual catalogues for each image, merged catalogues with all filters for each tile, Wide-patch merged catalogue (all tiles in all filters for each Wide patch). The objects are flagged according to the potential saturation in each filter and location in masked regions, object type (star or galaxy).