A meteorological PARASITE: High-resolution 3D wind vector from off-the-shelf UAS

Traditional in-situ and remote sensing methods leave observational gaps for the high-resolution 3D wind vector in the atmospheric boundary layer. We present PARASITE (Portable Aircraft Rucksack for Atmospheric Sensing and In-Situ Turbulence Estimation) a sensor and logging system to estimate the 3D wind vector using robust, off-the-shelf uncrewed aircraft systems (UAS) from internal avionics, independent of external flow sensors or calibration infrastructure. The approach combines a physics-based model combined with a neural network for residual error correction, both calibrated in a standalone process without requiring a reference sensor or wind tunnel. Validation campaigns took place at the German Meteorological Service (DWD) observatory in Falkenberg, Germany (Winter 2025) and Forschungszentrum Jülich, Germany (Summer 2024). The dataset includes 19 radiosonde ascents up to 2000 m above sea level and 8 flight hours adjacent to ultrasonic anemometers on a 99 m mast. Conditions ranged from 0.3 to 11 m s⁻¹ under thermally stable stratification for the sonic anemometer comparison, and convective conditions with wind speeds ranging from 0.0 to 11 m s⁻¹ for the radiosonde profiles. For 1 min averages compared to ultrasonic anemometer data, the UAS measurements show excellent correlation. Horizontal wind speed errors are low, with a root mean squared error (RMSE) of 0.30 m s⁻¹ and a mean error (ME) of 0.01 m s⁻¹. Wind direction shows an RMSE of 4^◦ and ME of 0.5^◦. Analysis of raw 10 Hz vertical wind data yields an ME of−0.04 m s⁻¹ and RMSE of 0.44 m s⁻¹. Analysis of ensemble averaged power spectra and structure functions confirms the method resolves turbulence following the Kolmogorov −5/3 law up to ∼2 Hz, comparable to reference instrumentation. Furthermore, comparisons with radiosonde profiles indicate the measurement is independent of air density.