<p><strong>Abstract.</strong> Mean vertical velocity measurements obtained from Radars at polar latitudes using Polar Mesosphere Summer Echoes (PMSE) as an inert tracer have been considered as non-representative of the mean vertical winds over the last couple of decades. PMSEs observed with the Middle Atmosphere Alomar Radar System (MAARSY) over And&#248;ya, Norway (69.30&#176;&#8201;N, 16.04&#176;&#8201;E) during summers of 2016 and 2017 are used to derive mean vertical winds in the upper mesosphere. The 3D vector wind components (zonal, meridional and vertical) are based on a Doppler beam swinging experiment using 5-beam directions (one vertical and four obliques). The 3D wind components are computed using a recently developed wind retrieval technique. The method includes full non-linear error-propagation, spatial and temporal regularization as well as beam pointing corrections and angular pointing uncertainties. Measurement uncertainties are used as weights to obtain seasonal weighted averages and characterize seasonal mean vertical velocity. Weighted average values of vertical velocities reveal a weak upward behaviour at altitudes 84&#8211;87&#8201;km after eliminating the influence of ice falling speed. At the same time, a sharp decrease/increase in the mean vertical velocities at the lower/upper edges of the summer mean altitude profile prevails, which are attributed to the sampling issues of PMSE due to disappearing of the target corresponding to the certain regions of motions and temperatures. Thus the mean vertical velocities can be biased with decrease up-/down-ward velocity measurements at lower/upper edges, while at the main central region the obtained mean vertical velocities are consistent with expected values of mean vertical winds after considering ice particle sedimentation.</p>