May 22 - July 30
Determining the Optimal Sampling Rate of a Sonic Anemometer Based on the Shannon-Nyquist Sampling Theorem
Andrew Mahre and Gerry Creager
What is already known:
- Sonic anemometers use no moving parts, and therefore the temporal averaging method which has been historically used for non-sonic anemometers has no scientific basis.
- A set, optimal sampling rate would be useful for obtaining the maximum amount of information possible from the wind’s “signal”.
- Better understanding of the properties of wind would be useful for modeling of the boundary layer.
What this study adds:
- Wind data from 4 separate instruments at 10 Hz, 1 Hz, and 1/3 Hz are analyzed using spectral analysis techniques.
- A power spectrum was created for each dataset and decimated versions of each dataset using a Fourier Transform.
- Spikes in power were present in the power spectrum created from the 10 Hz dataset and from decimated versions of the 10 Hz dataset, but are possibly due to the instrument, and not the wind itself.
- No spikes in power are present at any frequency in any other dataset.
Abstract:
While sonic anemometers have been in use for nearly 50 years, there is no literature which investigates the optimal sampling rate for sonic anemometers based on the Shannon-Nyquist Sampling Theorem. In this experiment, wind is treated as a wavelet, so that sonic anemometer data with multiple sampling rates can be analyzed using spectral analysis techniques. From the power spectrum, it is then possible to determine the minimum frequency at which a sonic anemometer must sample in order to maximize the amount of information gathered from the wavelet, while minimizing the amount of data stored. Using data from the Oklahoma Mesonet and data collected on-site, no obvious peak is present in any resulting power spectra that can be definitively be considered viable. This result suggests a nearly random power distribution among frequencies, which is better-suited for averaging and integrating data collection processes.