We describe, and use, a limited area, 3-dimensional transport model. The model domain is located over the Arctic, but includes the majority of the anthropogenic CO2 emissions in western and eastern Europe, which together make up about 1/3 of the global CO2 emissions. The model is run for several winter periods, using anthropogenic CO2 emissions only, and the results are compared with independent CO2 measurements taken at a monitoring station on Spitsbergen in the high Arctic. We show that the initial concentrations and boundary values of the domain are not crucial for the results, and conclude that most of the measured variability above the winter baseline in CO2 at the Arctic monitoring station emanates from recent CO2 sources within the model domain. From the observed small spatial variability in the monthly mean atmospheric CO2 mixing ratio in the north Atlantic region, we assume that there is only little net exchange between the atmosphere and ocean during the studied periods. Based on the co-variation between CO2 and particulate mass,we hypothesise that most of the measured CO2 variability is due to anthropogenic fossil fuel emissions, although we can not rule out a biogenic CO2 component. Using the transport model, we compare different estimates of fossil-fuel consumption in the mid-latitudes. We find that the industrial centres and the surrounding gas-fields in the lower-Ob region (60 degrees-72 degrees N, 65 degrees-80 degrees E) occasionally have a much larger impact on the CO2 measurements at Spitsbergen than follows from a recent CO2 emission inventory. This implies that there may be an overlooked CO2 source in this region, possibly flaring of gas.