A global (multi-CCD and multi-dither) astrometric calibration is calculated per filter per tile. SCAMP (Bertin 2006, ASP Conf. Series 351, 112) is used for this purpose, with a polynomial degree of 2 over the whole mosaic. The (unfiltered) 2MASS-PSC (Skrutskie et al. 2006, AJ, 131, 1163) is used as astrometric reference catalog. A more detailed description of the astrometric pipeline is in MF13.
SWARP is used to resample all exposures in a tile to the same pixel grid, with a pixel size of 0.2 arcseconds. After background subtraction the exposures are coadded using a weighted mean stacking procedure.
The accuracy of the absolute astrometry ("KiDS vs 2MASS") is uniform over a coadd, with typical 2-dimensional (2-D) RMS of 0.31 arcsec in g, r, and i, and 0.25 arcsec in u. The lower RMS in u-band is most likely due to the fact that in this band on average brighter 2MASS sources are selected as reference sources.
The accuracy of the relative astrometry ("KiDS vs KiDS"), measured by the 2-D positional residuals of sources between dithers, is also uniform across a single coadd. In the middle column of the figure in the data product quality section the accuracy of this relative astrometry of all coadds is shown. The typical 2-D RMS is ~0.03 arcsec in all filters, but with a larger scatter in u. The larger range in RMS in u-band is due to the smaller number of available reference sources.