The release of carbon dioxide from Arctic permafrost soils is both a consequence and a driver of the amplified increase of near-surface temperatures observed in the Arctic. The warming is most intense during the winter months where the CO2 exchange is still poorly explored due to the extreme conditions of the Arctic winter. This study aims to contribute to a better understanding of the physical processes driving the CO2 fluxes during the winter months based on a one year time series of eddy covariance flux estimates from the high Arctic Bayleva site. Its focus is on time periods of hours to days, where the magnitude of the CO2 fluxes significantly exceeds its typical range of variation. These time periods are generally associated with an exchange of the large scale air masses influencing the study site and particularly occur simultaneously with high wind speeds. While fluxes indicating an unexpected uptake of CO2 by the soil during the winter months are most likely related to sensor limitations, no clear measurement issues could be identified for time periods of apparent strong CO2 release to the atmosphere. The advection of CO2 rich air to the study site from remote regions is likely to be the most important cause of these observations, while the abrupt release of CO2 stored below the measurement height or inside the snow pack might have an additional, although smaller, impact. In order to separate the individual contributions and to exclude any methodological issues, further research on the topic is urgently needed. Since the CO2 flux events in total account for a significant part of 30 % of the net annual carbon uptake, their handling has pronounced implications for the quantification of the current Arctic carbon cycle. However, since the events of apparent intense CO2 exchange are not yet sufficiently understood, their handling probably differs significantly among studies. Especially in the context of long term studies the events might easily even go unnoticed. When using already available flux data sets as an input for large scale modelling efforts, the presence of winter time CO2 flux events at comparable study sites all over the Arctic region might thus introduce an unexpected uncertainty to future climate projections.