Abstract Ultrasonic anemometers are widely used to investigate turbulence in the surface layer. Some of their advantages are high-frequency sampling, the ability to work for long periods without a resident operator (even in adverse meteorological conditions), and their calibration related only to design parameters. In this paper an analysis of the random uncertainty associated with ultrasonic anemometer measurements is reported. The analysis is based on a statistical procedure that compares the simultaneous data taken with two identical anemometers operating in nominally identical conditions. Postprocessing of data has been carried out in different reference systems in order to evaluate how the random uncertainties change according to the postprocessing procedure used. Results show that uncertainty on wind velocity decreases with averaging time, and it can be as low as 1 cm s−1 for a typical averaging time of 30 min. The random uncertainty on average vertical wind velocity 〈w〉 could also be as low as 1 cm s−1, and it is very sensitive to the effects of vertical misalignment. The analysis is based on six different measurement sets in which the anemometers have been deployed on a single mast or in two separate masts with and without additional detectors placed near the anemometers themselves. Results indicate that the uncertainty of all the measured parameters increases when the anemometer is used in configurations in which they are placed on separate masts. Several parameters also show an additional increase of uncertainty if other detectors are placed nearby. The relative random uncertainty on momentum and sensible heat fluxes, for typical averaging time, can be as low as 6%–7% and it could increase to a factor of 2–3 when they are placed on separate masts. Only small effects due to the influence of flow distortions caused by the presence of additional sensors have been found on fluxes and are mainly related to sensible heat flux.