Bromine oxide (BrO) was measured in situ during the Arctic Tropospheric Ozone Chemistry (ARCTOC) '96 campaign in Ny ¿lesund, Spitsbergen (April--May 1996), and during the Alert 2000 Polar Sunrise Experiment in Alert, Nunavut, Canada (April--May 2000). Measurements were made in near-surface air during low-ozone events in early May at both sites. The average of the in situ concentrations of BrO at Ny ¿lesund is consistent with the path average of near-simultaneous long-path differential optical absorption spectroscopy measurements, but there is considerable scatter in a point-by-point comparison. The differences between these observations are consistent with a strong surface influence on reactive bromine. We see similar variability of the in situ measurements at Alert, which reflects the real variability of both surface sources and sinks. The fluctuations of BrO abundances are used to assess the expected loss rates and associated lifetimes of ozone during depletion events. We show that ozone loss will be underpredicted by any temporal or spatial average of BrO.