The ozone-tracer correlation method is used to deduce the stratospheric ozone loss in the Arctic winter 1996--1997. Improvements of the technique are applied, such as a new calculation of the vortex edge <Nash et al., 1996> and an improved early vortex reference function. Winter 1996--1997 is characterized by a late formation and an unusually long lifetime of the polar vortex. Remnants of vortex air were found until May. Chemical ozone losses deduced from two satellite data sets, namely Improved Limb Atmospheric Spectrometer (ILAS) and Halogen Occultation Experiment (HALOE), are discussed. The ILAS observations allow a detailed analysis of the temporal evolution of the ozone-tracer correlation inside the polar vortex and, in particular, of the development of the early vortex. For November and December 1996, it is shown that horizontal mixing still influences the ozone-tracer relation. Significant PSC related chemical ozone loss occurred beginning at mid-February, and the averaged column ozone loss is increasing toward the middle of May. From April onwards, ozone profiles in the vortex became more uniform. The decrease of ozone in the vortex remnants in April and May occurred due to chemistry. HALOE observations are available for March to May 1997. In the period 4--16 March 1997, the calculated ozone loss deduced from HALOE and ILAS is in good agreement. The average of the result from the two instruments is 15 ¿ 7 Dobson units (DU) inside the vortex core, in the altitude range of 450--550 K. At the end of March, a discrepancy between HALOE and ILAS ozone loss arises due to a significant difference (0.6 ppmv) between the two data sets in the relatively low ozone minimum measured at 475 K. Nonetheless, both data sets consistently show an inhomogeneity in ozone loss inside the vortex core at the end of March. The vortex is separated in two parts, one with a large ozone loss (HALOE 40--45 DU, ILAS 30--35 DU) and one with a moderate ozone loss (HALOE 15--30 DU, ILAS 5--25 DU) for 450--550 K. The ozone loss from HALOE in 380--550 K at that time was calculated to be 90--110 DU for the large ozone loss and 20--80 DU for the moderate ozone loss. The vortex average of column ozone loss from HALOE inside the vortex core at the end of March is 61 ¿ 20 DU, which is an increase of about 20% compared to the earlier study by M¿ller et al. <1997b> brought about by the improvement of the technique.