High-resolution vertical wind and turbulence measurements with quadcopter uncrewed aerial systems: wind tunnel calibration and field validation

TL;DR

This study calibrates and validates high-resolution vertical wind and turbulence measurements using quadcopter drones in wind tunnels and field conditions, enabling accurate turbulence analysis in complex atmospheric flows.

Contribution

The paper introduces a wind tunnel calibration method for drone-based vertical wind measurements and validates it in field conditions, improving turbulence data accuracy.

Findings

High accuracy in turbulence statistics measurements.

RMSE of 0.12 m²/s² for vertical wind variance.

Agreement of variance spectra with reference sensors.

Abstract

The SWUF-3D drone fleet is used in the atmospheric boundary layer (ABL) for in situ turbulence measurements of complex flows, such as in mountainous terrain or wind turbine wakes. Previous calibrations for measuring vertical wind speed $w$ using the drones' avionics data were performed on field data, limiting applicability to low winds ($\leq 8~\mathrm{m}\,\mathrm{s}^{-1}$) and being prone to high uncertainties. To overcome these limitations, we calibrate $w$ measurement in a wind tunnel and validate it in field measurements. Calibration is performed in a wind tunnel with an active grid used to deflect horizontal flow into the vertical. This wind is measured with a multi hole probe, while wind forces acting on the drone are determined from the avionics data, allowing an empirical relationship between these quantities. For validation, we conduct comparative fleet measurements with up to 10 drones simultaneously around an array of meteorological masts equipped with sonic anemometers. The results show high accuracy for turbulence statistics: the variance determination for $w$ has a root mean square error (RMSE) of 0.12~$\mathrm{m^2\,s^{-2}}$ and a normalized RMSE (nRMSE) of 17.0~\%, for the horizontal wind components the RMSEs are $\sim$0.3~$\mathrm{m^2\,s^{-2}}$ and nRMSEs $\sim$25~\%. The RMSEs for the covariances of the components are $<\,$0.3~$\mathrm{m^2\,s^{-2}}$. The variance spectra of $w$ measured with drones and reference sensors agree in all frequency ranges, the RMSE for covariances between different measurement points is $\sim$0.1~$\mathrm{m^2\,s^{-2}}$. Accurate $w$ retrieval at all wind speeds sustainable by the drone enables studies of strongly three-dimensional flows, supports eddy-covariance flux estimation, enables resolving diurnal turbulence evolution in the ABL, and improves spatial turbulence characterization.