Underestimation of nocturnal CO2 respiration under calm conditions remains an unsolved problem at many forest flux stations. In this paper, the hypothesis is tested that horizontal mean transport, not previously measured, may account for the missing CO2. A systematic methodology was developed that comprises characterizing the subcanopy motions, determining the appropriate size of the subcanopy network required to make the measurements, developing a method of integrating the measurements in the vertical, and determining the required averaging time. Measurements were performed at the Harvard Forest (Petersham, MA), over 4 years. The appropriate size of the network of wind and CO2 sensors was shown to be on the order of 100 m, ensuring that sensors were generally observing coherent processes on this scale or larger and thus displayed some correlation. Horizontal transport of CO2 at Harvard Forest was found to be restricted to the bottom ∼10 m of the forest. The fraction of the negative buoyancy force in the sum of dynamic driving forces described nights with missing flux problems (“deficit nights”) significantly better than the commonly used friction velocity criterion. Including the measured horizontal transport terms did not on average fully account for the observed difference in NEE of 1.2±0.3 μmol m−2 s−1 between deficit and non-deficit nights, but decreased the difference to 0.7±0.5 μmol m−2 s−1. Horizontal transport did account for the difference, to within measurement error, during summer months, but not during spring or fall.