Because most forests are multi-specific and vertically heterogeneous, understanding the functioning of such complex ecosystems remains a challenge for quantifying and modeling the fluxes of CO2 between the atmosphere and the vegetation at regional scale. In particular, there is lack of knowledge concerning flux partitioning between overstorey and understorey of forests across different climates and forest types. The main objective of this study is to compare the micrometeorology and the fluxes of CO2, water and sensible heat measured by the eddy-covariance method above and below the main canopy of different forests. The eight study sites are part of the Fluxnet network, and range through various climates and structural types. Across these different biomes, the leaf area index (LAI) of the overstorey is the main factor influencing flux footprint and turbulent mixing below the canopy, and flux partitioning between overstorey and understorey. Wind direction and rates of turbulent mixing generally differ between the two levels for close canopies and low wind speed in the boundary layer. Forcing of understorey turbulent mixing by overstorey conditions is stronger in case of low overstorey LAI, while trunk space flow tends to decouple from that in the boundary layer in closed canopies. Vertical light extinction is positively correlated with overstorey LAI. During the day, efflux of CO2 from the understorey is positively related to soil temperature. Uptake of CO2 by the understorey only occurs if a significant amount of light penetrates through the overstorey and reaches a significant amount of understorey leaf area. During the night, the probability that CO2 emitted from the soil reach the top of the canopy is negatively correlated with overstorey LAI. These results will help developing sound ecosystem modeling schemes for quantifying carbon budgets at regional scales, taking into account the wide variety of mixed and vertically complex forest canopies.