The sensitivity of sonic anemometer-derived stress estimates to the tilt of
the anemometer is investigated. The largest stress errors are shown to occ
ur for unstable stratification (z/L < 0) and deep convective boundary layer
s. Three methods for determining the tilt angles relative to a mean streaml
ine coordinate system and for computing the tilt-corrected stresses are the
n compared. The most commonly used method, involving a double rotation of t
he anemometers' axes, is shown to result in significant run-to-run stress e
rrors due to the sampling uncertainty of the mean vertical velocity. An alt
ernative method, requiring a triple rotation of the anemometer axes, is sho
wn td result in even greater run-to-run stress errors due to the combined s
ampling errors of the mean vertical velocity and the cross-wind stress. For
measurements over the sea where the cross-stream stress is important, the
double rotation method is shown to overestimate the surface stress, due to
the uncorrected lateral tilt component. A third method, using a planar fit
technique, is shown to reduce the run-to-run stress errors due to sampling
effects, and provides an unbiased estimate of the lateral stress.