Apparent winter CO2 uptake by a boreal forest due to decoupling

•EC measurements above and below an open forest canopy reveal frequent decoupling.•Above-canopy filtering approaches identify decoupled periods insufficiently.•Local topography may have a profound influence on above-canopy EC measurements.•Decoupling and sub-canopy horizontal flow may bias forest EC measurements globally.

Net uptake of carbon dioxide (CO2) was observed during the winter when using the eddy covariance (EC) technique above a ∼90-year-old Scots pine (Pinus sylvestris L.) stand in northern Sweden. This uptake occurred despite photosynthetic dormancy. This discrepancy led us to investigate the potential impact of decoupling of below- and above-canopy air mass flow and accompanying below-canopy horizontal advection on these measurements. We used the correlation of above- and below-canopy standard deviation of vertical wind speed (σw), derived from EC measurements above and below the canopy, as the main mixing criterion. We identified 0.33 m s−1 and 0.06 m s−1 as site-specific σw thresholds for above and below canopy, respectively, to reach the fully coupled state. Decoupling was observed in 45% of all cases during the measurement period (5.11.2014-25.2.2015). After filtering out decoupled periods the above-canopy mean winter NEE shifted from −0.52 μmol m−2 s−1 to a more reasonable positive value of 0.31 μmol m−2 s−1. None of the above-canopy data filtering criteria we tested (i.e., friction velocity threshold; horizontal wind speed threshold; single-level σw threshold) ensured sufficient mixing. All missed critical periods that were detected only by the two-level filtering approach. Tower-surrounding topography induced a predominant below-canopy wind direction and consequent wind shear between above- and below-canopy air masses. These processes may foster decoupling and below-canopy removal of CO2 rich air. To determine how broadly such a topographical influence might apply, we compared the topography surrounding our tower to that surrounding other forest flux sites worldwide. Medians of maximum elevation differences within 300 m and 1000 m around 110 FLUXNET forest EC towers were 24 m and 66 m, respectively, compared to 24 m and 114 m, respectively, at our site. Consequently, below-canopy flow may influence above-canopy NEE detections at many forested EC sites. Based on our findings we suggest below-canopy measurements as standard procedure at sites evaluating forest CO2 budgets.