Ejection and sweep eddy motions in the atmospheric surface layer (ASL) are widely accepted as being responsible for much of land surface evaporation, sensible heat flux, and momentum flux; however, less is known about this type of eddy motion within the canopy sublayer (CSL) of forested systems. The present study analyzed the ejection-sweep properties at the canopy-atmosphere interface of a 13 m tall, uniformly aged southern loblolly pine stand and a 33 m tall, unevenly aged hardwood stand using velocity and scalar (temperature, water vapor, and carbon dioxide) fluctuation measurements at the canopy-atmosphere interface. It was found that the measured sweeps and ejections time fractions for scalars and momentum are comparable and are in good agreement with other laboratory and field experiments. This investigation demonstrates that the third-order cumullant expansion method (CEM) reproduces the measured relative flux contribution of ejections and sweeps (ΔS0) and the difference between sweep and ejection time fractions for both momentum and scalars at the canopy-atmosphere interface in contrast to findings from a previous ASL experiment. A linkage between ΔS0 and the scalar flux budget is derived and tested via the third-order CEM at the canopy-atmosphere interface for the pine and the hardwood stands. It is shown that ΔS0 can be related to the dimensionless scalar flux transport term whose gradient is central to the scalar variance budget. Also, the derived relationship is independent of canopy roughness or scalar sources and sinks. Hence this investigation establishes an analytical linkage between second-order closure models, the ejection-sweep cycle, and third-order CEM at the canopy-atmosphere interface. Dissimilarity between the ejection-sweep cycle for scalar and momentum transport is considered via conditional probability distributions at both forest stands. In contrast to a laboratory heat dispersion experiment, it is shown that while the ejection-sweep cycles for scalar and momentum transport are intimately linked, they are not identical. Therefore the results from momentum ejection-sweeps investigations cannot be extrapolated to scalar transport. Comparisons with other laboratory experiments are also discussed, especially in relation to the scalar ejection and sweep time fractions.¿ 1997 American Geophysical Union