Spatio-temporal observations of temperature and wind velocity using drone-based acoustic atmospheric tomography

This paper reports on a sequence of trials in which the acoustic signature of a small remotely piloted vehicle (drone) has been used to obtain spatio-temporal estimates of atmospheric temperature and wind vectors. Sound fields are recorded onboard the aircraft and by microphones on the ground. Observations are compared and the resulting propagation delays computed for each intersecting ray path transecting the intervening atmosphere. A linear model of sound speed corresponds to virtual temperature and wind velocity, plus tomographic inversion combined with regularisation, then allows vertical cross-sections and volumes of temperature and wind profile to be computed. These two- and three-dimensional profiles are represented as a lattice of elliptical radial basis functions, which enables the medium to be visualised at high levels of resolution. The technique has been used to provide spatio-temporal visualisation of atmospheric dynamics up to altitudes of 1200 m over baselines of 600 m. Independent measurements taken by co-located instruments such as a Doppler SODAR, ZephIR 300 LIDAR and temperature sensors carried onboard drones flying within the remotely sensed atmosphere show real world performance suggests accuracies of around 0.3 °C, 0.5 m/s and 0.2 m/s for temperature, horizontal and vertical wind speeds respectively may be anticipated. The real world performance also compares very favourably to error envelopes anticipated from propagation models based on large eddy simulation.