Using a combination of data from Match, POAM II, POAM III and MLS we show that the chemical loss rate of Arctic O3 during January of four cold winters (1992, 1995, 1996, and 2000) is consistently faster than can be accounted for by assuming complete activation of reactive chlorine and standard reaction kinetics. However, O3 loss rates measured during late February and early March 1996 are shown to be consistent with observations of ClO. The faster than expected O3 loss rates during January are shown to occur when air parcels are illuminated at high solar zenith angles (SZAs between ~85 and 94°), and to result in cumulative O3 loss of ~0.5 ppmv. The cause of the rapid January O3 loss is unclear, but may be related to a photolytic process at high SZA that is poorly represented by current photochemical models.