Abstract. Since turbulence measurements from Doppler lidars are being increasingly used within wind energy and boundary-layer meteorology, it is important to assess and improve the accuracy of these observations. While turbulent quantities are measured by Doppler lidars in several different ways, the simplest and most frequently used statistic is vertical velocity variance (w 2 ) from zenith stares. However, the competing effects of signal noise and resolution volume limitations, which respectively increase and decrease w 2 , reduce the accuracy of these measurements. Herein, an established method that utilises the autocovariance of the signal to remove noise is evaluated and its skill in correcting for volume-averaging effects in the calculation of w 2 is also assessed. Additionally, this autocovariance technique is further refined by defining the amount of lag time to use for the most accurate estimates of w 2 . Through comparison of observations from two Doppler lidars and sonic anemometers on a 300 m tower, the autocovariance technique is shown to generally improve estimates of w 2 . After the autocovariance technique is applied, values of w 2 from the Doppler lidars are generally in close agreement (R 2 ≈ 0.95 − 0.98) with those calculated from sonic anemometer measurements.