In most studies of the net ecosystem-atmosphere exchange of CO2 (NEE) using tower-based eddy covariance (EC) systems is has been assumed that advection is negligible. In this study we use a scalar conservation budget method to estimate the contribution of advection to NEE measurements from a very tall tower in northern Wisconsin. We examine data for June--August 1997. Measured NEE0, calculated as the sum of the EC flux plus the rate of change of storage below the EC measurement level, is expected to be constant with measurement height, and we take the differences between levels as a measure of advection. We find that the average difference in total advection ΔFCadtot between 30 and 122 m is as large as 6 μmol m-2 s-1 during the morning transition from stable to convective conditions and the average difference ΔFCadtot between 122 and 396 m is a large as 4 μmol m-2 s-1 during daytime. For the month of July, advection between 30 and 122 m is 27% of the diurnally integrated NEE0 at 122 m, and advection between 122 and 396 m accounts for 5% of the NEE0 observed at 396 m. The observed differences of advection often have significant correlation with the vertical integral of wind speed within the same layer. This indicates that the horizontal advection contribution to NEE could be significant. Direct observations of the vertical gradient in CO2 show that ΔFCadtot cannot be explained by vertical advection alone. It is hypothesized that differing flux footprints and pooling of CO2 in the heterogeneous landscape causes the advection contribution. The magnitudes of the total advection component FCadtot of NEE at the 30 m level are roughly estimated by a linear extrapolation. A peak in FCadtot at 30 m of ~3 μmol m-2 s-1 during the morning transition is predicted for all three months. The July integrated FCadtot is estimated to be 10% of the diurnally integrated NEE0 at 30 m. ¿ 2000 American Geophysical Union