Each stack comprises all images located inside a radius of 3′with respect to a tile center position. The shifts between each exposure in an observing sequence ensures that gaps between the CCDs are filled in the stacked image. Note that in the case of the Wide stacks, the overlapping pixels of nearest neighbor tiles are used for the field-to-field calibration but are not used during the combination process. (In the Deep D2-u* field, additional data from the COSMOS survey is used to make the final stacks). The software used for the image combination is SWARP (in version 2.17.6) developped by Emmanuel Bertin and available on Astromatic16 .
The T0007 images and weight-maps are combined using the option COMBINE_TYPE “MEDIAN” in SWarp for the Wide and Deep surveys. The weight-maps correspond to SWarp’s image type MAP_WEIGHT which correspond to maps of relative inverse variance. Images are resampled using Lanczos-3 interpolation kernel. All resulting tiles and weight-maps have 19354 × 19354 pixels of 0.186", corresponding 1 deg × 1 deg.
Additionally, for the Deep fields, a second set of stacks is produced using a sigma-clipping pixel combination algorithm, selected by setting COMBINE_TYPE “SIGWEI” in our modified version of SWarp supplied by S. Foucaud. The clipping parameter is set to ±3.0σ.
The complete list of images used in each Wide tile is presented in the Appendix A.2. For the Deep fields, similar lists can be made available on demand. For the Deep fields, images with seeing < 1.3” in u*, and < 0.95” in griyz, are selected for the stack production. From the lists, secondary lists are created containing the “85%” and “25%” best-seeing images.
All stacks have a zero-point of 30 in AB magnitudes. The magnitudes of objects in these stacks are computed as follows: m = 30 - 2.5 log(counts). The SWarp configuration file is presented in Section B.2.2.
Once the g,r and i band data are available they are combined into a “chisquare” or chi2 image which has the same position, scale and input size as the g, r and i images. This chisquared image has been shown to be the optimal way to create a detection image from a series of images taken with different filters (?). The SWarp configuration file for the production of the T0007 chi2 images is given in Section B.2.3.