Sonic anemometers are the principal instruments in micrometeorological studies of turbulence and ecosystem fluxes. Common designs underestimate vertical wind measurements because they lack a correction for transducer shadowing, with no consensus on a suitable correction. We present a subset of data collected during field experiments in 2011 and 2013 featuring two or four CSAT3 sonic anemometers. We introduce a Bayesian analysis to resolve the three-dimensional correction by optimizing differences between anemometers mounted both vertically and horizontally. A grid of 512 points (~ ±5° resolution in wind location) is defined on a sphere around the sonic anemometer, from which the shadow correction for each transducer-pair is derived from a set of 138 unique state variables describing the quadrants and borders. Using the Markov chain Monte Carlo (MCMC) method, the Bayesian model proposes new values for each state variable, recalculates the fast-response dataset, summarizes the five-minute wind statistics, and accepts the proposed new values based on the probability that they make measurements from vertical and horizontal anemometers more equivalent. MCMC chains were constructed for three different prior distributions describing the state variables: no shadow correction, the Kaimal correction for transducer shadowing, and double the Kaimal correction, all initialized with 10% uncertainty.
This archive includes all data and source code required to create the 3-D posterior correction and to analyze its effect during the 2011 and 2013 field experiments plus a 2016 field validation experiment. There are 20 Hz wind velocity and temperature data from the 2011 and 2013 field experiments contained within 74,664 files (one file each per five-minute period, replicate anemometer, and separated between wind and temperature) and 20 Hz wind velocity and temperature data from the 2016 experiment contained within 1883 files (one file each per each five-minute period, each file contains 3 replicate anemometers). R code is provided to estimate the posterior correction from the 2011 and 2013 wind data using Bayesian analysis assuming three prior corrections: a flat prior, the Kaimal correction, and double the Kaimal correction. Matlab code is provided to normalize, thin, and scale the posterior correction into its final form. Matlab code is provided to apply the posterior correction and to analyze is effect for both sets of data. Auxiliary files to execute the code are also provided.