Abstract: Observations of vegetation-atmosphere exchange of carbon dioxide (CO 2) by the eddy covariance (EC) technique are limited by difficult conditions such as nighttime and heterogeneous terrain. Thus, advective flux components are included into the net ecosystem exchange (NEE) budget. However, advection measurements are experimentally challenging and do not always help to solve the night flux problem of the EC technique. This study investigates alternative methods for the observation of horizontal advection, in particular horizontal concentration gradients, as well as different approaches to coordinate rotation and vertical advection. Continuous high-frequency measurements of the horizontal CO 2 concentration field are employed and compared to the often used discontinuous sequential sampling. Significant differences were found in the case of 30-min mean concentration values between the conventional discontinuous sampling approach and the complete observation of the time series by continuous sampling. Estimates of vertical advection rely on accurate estimates of vertical wind velocity (W). Therefore, different approaches to the planar fit coordinate rotation have been investigated. Sector-wise rotation was able to eliminate directional dependencies of mean W. Furthermore, the effect of the data set length used for rotation (window length) was investigated and was found to have significant impact on estimates of vertical advection, with larger window lengths yielding about 50% larger vertical advection. A sequential planar fit with controlled window length is proposed to give reproducible results. The different approaches to the measurement and calculation of horizontal and vertical advection presented are applied to data obtained during the exchange processes in mountainous region experiment at the FLUXNET site Waldstein-Weidenbrunnen (DE-Bay). Estimates of NEE including advection are compared to NEE from turbulent and storage flux alone without advection. NEE including vertical advection with sector-wise planar fit rotation and controlled window length and including horizontal advection from continuous gradient measurements, which were comprehensively bias corrected by a new approach, did compare well with the expected night flux error, with meteorological drivers of the fluxes and with soil chamber measurements. Unrealistically large and noisy values of horizontal advection from the conventional discontinuous sampling approach, which lead to unrealistic values of NEE, could be eliminated by the alternative approaches presented. We therefore suggest the further testing of those approaches at other sites in order to improve the accuracy of advection measurements and, subsequently, estimates of NEE. © 2011 Springer-Verlag.