Observations of ozone mixing ratios in the lower troposphere of Arctic Canada in April 1994 are summarized. Except during a brief period of anomalous transport from high altitude, air depleted in ozone was always observed above sea ice and snow throughout the northern Ellesmere Island/Lincoln Sea region, or whenever air was sampled which had been in recent contact with sea ice and snow. Ozone mixing ratios observed at a camp on the sea ice north of Alert were consistent with a 1992 study. Observations at the ice camp confirmed that ozone was depleted more frequently (74% of all observations <5 ppbv) than at coastal and inland sites near Alert (10% of observations <5 ppbv). Mixing ratios briefly attained a maximum of 36 ppbv at the ice camp but were normally completely depleted or below typical free tropospheric levels of 35--45 ppbv observed elsewhere in the region. Aircraft measurements and vertical profiles of ozone and meteorological parameters from balloon sondes confirmed that ozone depletion existed in a layer above the sea ice, from the surface up to heights of 200--400 m. Some observations showed a very abrupt transition between depleted and nondepleted conditions at the upper boundary of the layer. The occurrence of a layer of fully depleted ozone was well correlated with surface high-pressure systems. At Alert the appearance of air that was fully depleted in ozone was driven by advection from ocean areas to the north. A fortuitous set of meteorological conditions allowed the use of a simple model for testing general aspects of some proposed hypotheses for ozone destruction, during a period when it is believed the depleted layer was forming. It was found that the observed rate of ozone destruction would require levels of HO and Cl atoms much higher than would be expected for the prevailing conditions but reasonable concentrations of Br atoms. It was also found that an effective ozone deposition velocity of 0.1--0.2 cm s-1 could account for the observed depletion rate during this period. That is, the observed rate of ozone depiction during formation of a depleted layer was consistent with either a volume sink or a surface sink for ozone. ¿ 1998 American Geophysical Union