Transducer Shadowing Explains Observed Underestimates in Vertical Wind Velocity from Non-orthogonal Sonic Anemometers

Abstract:

Sonic anemometry is fundamental to all eddy-covariance studies of surface energy and ecosystem carbon and water balance. While recent studies have shown that some anemometers underestimate vertical wind, we hypothesize that this is caused by the lack of transducer shadowing correction in non-orthogonal models. We tested this in an experiment comparing three sonic anemometer designs: orthogonal (O), non-orthogonal (NO), and quasi-orthogonal (QO); using four models: K-probe (O) and A-probe (NO) (Applied Technologies, Inc.) and CSAT3 (NO) and CSAT3V (QO) (Campbell Scientific, Inc.). For each of a 12-week experiment at the GLEES AmeriFlux site, five instruments from a pool of twelve (three of each model) were randomly selected and located around a control (CSAT3); mid-week all but the control were re-mounted horizontally. We used Bayesian analysis to test differences between models in half-hour standard deviations (σ_u, σ_v, σ_w, and σ_T), turbulent kinetic energy (TKE), and the ratio between vertical/horizontal TKE (VHTKE). The K-probe experiences horizontal transducer shadowing which is effectively corrected using an established wind-tunnel derived algorithm. We constructed shadow correction algorithms for the NO/QO anemometers by applying the K-probe function to each non-orthogonal transducer pair (SC1) as well as a stronger correction of twice the magnitude (SC2). While the partitioning of VHTKE was higher in O than NO/QO anemometers, the application of SC1 explained 45-60% of this discrepancy while SC2 overcorrected it. During the horizontal manipulation changes in the NO/QO were moderate in σ_u (4-8% decrease), very strong in σ_v (9-11% decrease), and minimal in σ_w (-3 to 4% change) while only σ_u measurements changed (3% decrease) with the K-probe. These changes were predicted by both shadow correction algorithms, with SC2 better explaining the data. This confirms our hypothesis while eliminating others that attribute the underestimate to a systematic bias in the w-axis. All flux sites that employ these non-orthogonal sonic anemometers will underestimate vertical fluxes of mass and energy, but since 1) the correction is a function both of sensor geometry and local wind and 2) the true shadow correction function is unknown at this time, the actual underestimate is uncertain and will vary between sites.