Extraction of 1-d spectra from the 2-d images will be performed by a highly automated version of CCDSPEC (Gulliver & Hill 2002). The characteristics of the CCD are automatically applied to the images, including the location of dead rows and hot pixels. In order to achieve the goal of better than 1% precision, large numbers of bias and flat field frames will be used in the reduction process. There will be a continual programme to monitor the image quality. Finally, optimally extracted spectra will be obtained, including the removal of scattered light and cosmic rays.

The Earth's atmosphere has a considerable effect on the stellar flux as measured from the surface. The principal sources of extinction, Rayleigh and aerosol scattering, ozone and telluric line absorption, are discussed, along with methods used to determine their effects on the observed spectra. Correction for telluric lines is the most problematical, due to their non-linear variation with airmass. By using a large network of constant stars to monitor atmospheric extinction it is possible to determine the extinction coefficients to generally better than 1% and to assess their temporal variability.

The spectrophotometric observations are placed on an absolute flux scale by reference to stars with known values of true flux at top of Earth's atmosphere. These standard stars have been calibrated against terrestrial sources of known properties. Unfortunately, very few stars have been calibrated directly. The ASTRA fluxes will be calibrated against the best available Vega flux distribution. The constant stars used in the extinction determinations will provide the internal calibration network of secondary flux standards.

The available absolute calibrations are accurate to typically 1–2%. Ultimately this uncertainty will limit the accuracy of the final fluxes of other stars. However, the internal precision will be significantly higher, and should moreaccurate absolute calibrations become available the fluxes can be re-calibrated to higher accuracy.