The performance of non-orthogonal sonic anemometers has been the subject of much discussion in recent years. It followed findings that they underestimated vertical winds and vertical heat fluxes by 10-15%. For agricultural and forestry scientists who depended on them for their large-scale energy balance studies this was a bad surprise. They had collected turbulence data from hundreds of monitoring stations and found imbalances of that order in stations using non-orthogonal probes. The stations (fewer in number) that used orthogonal sonic anemometers seemed to do well.
Studies of past field data and the results of their own intercomparison experiments led the scientists to conclude that the underestimations were a consequence of the non-orthogonal design and that the best vertical winds were those measured along a vertical acoustic path. The exact cause of the underestimation was never determined.
In a recent Applied Technologies Application Note (Kaimal and Zimmerman, 2014), we started to look for clues to this underestimation in the coordinate transformation equations. We wondered if there is something in the equation for the vertical component that made it vulnerable to interference from the probe’s support structure. The equations we examined converted wind measurements along the three tilted paths to components along fixed orthogonal coordinates as defined in the ATI - K-Probe: U along the probe’s support boom, V pointing sideways and W vertical. W turned out to be simply the sum of the winds along the three tilted axes (times 0.385). Any slowing down of winds along the non-orthogonal paths would directly affect W. Support structures needed to hold the transducers in place could cause that if the updrafts and downdrafts encountered are steep enough. If such events are frequent, how badly would they distort the W signals? To answer that, we needed to look at actual signals from orthogonal and non-orthogonal sonic anemometers over a range of atmospheric conditions.
The data collected so far show a very consistent pattern. Under moderately unstable to slightly stable conditions the two probes track U, V, and T very closely but W is clearly being underestimated by the A-Probe.