Detection and Tracking of Tornadic Storms in the Southeast with Arrays of Infrasound Sensors

Working in co-ordination with the NOAA Vortex Southeast (Vortex SE) research program, several infrasound sensor arrays were deployed at fixed sites across Alabama, Georgia, and Tennessee during Spring of 2017 and Spring of 2018 to investigate the emission and characterization of infrasonic acoustic energy from tornadoes and related phenomena. Each array consisted of several calibrated, broadband infrasound sensors. The arrays were configured in a pattern such that accurate bearings to acoustic sources could be obtained over a broad range of frequencies (nominally from 1 Hz to 100 Hz). Data were collected synchronously at a rate of 1000 samples per second.

Subsequent processing of the data from the arrays revealed acoustic emissions from the tornadic storms ranging in frequencies below 1 Hz to frequencies greater than 10 Hz. Accurate bearings, as verified by correlation with radar data, to the storms have been calculated from distances greater than 60 km. Analysis has revealed that continuous emissions occurred prior to the estimated touchdown times while the storms were on the ground and for short periods after the tornadoes lifted; however, the strongest emissions appeared to occur while the storms were on the ground.

Additional results from the analysis of the infrasound data will be presented that demonstrates that automatic, real-time, signal processing methods exist to distinguish sound from thunder (transient), from sound created by tornadic storms (quasi-stationary) and to calculate bearings to more than one storm simultaneously (source separation) using the same array. The analysis also reveals that, not unexpectedly, the distance from which the emissions can be detected is dependent on regional-scale atmospheric conditions.

For some time, it has been recognized that tornadoes emit acoustic radiation at audio and infrasonic frequencies, although all the physical mechanisms that radiate acoustic energy have yet to be fully understood. In the middle 1970s, work by T. M. Georges established that convective storms and tornadic storms radiate infrasound detectable at ranges on the order of hundreds of kilometers (Georges, 1973; Georges and Greene, 1975), while analysis of opportunistic recordings of nearby tornadoes supplied evidence that the acoustic energy at audio frequencies exhibits a broad peak and then decreases as frequency increases (Arnold et al., 1976) similar to that of a classical turbulence energy cascade. A review of the history and development of severe storm infrasound programs and associated radiation models from the 1970s through 2004 was provided by Bedard (2005) that helped reinforce that infrasonic tornado detection is feasible with existing technology. He examined infrasound in the 0.5–10 Hz band and found the best correlation between sensors (a proxy for signal-to-noise ratio and the basis of storm detection) was between 0.5 and 2.5 Hz. Analysis of the storms detected with infrasound during the summer of 1985 found that 89% of these produced hail, had tops greater than 13.7 km, or produced a hook radar echo. Another relevant finding was that vortex motions aloft were apparently detectable by infrasound 20–30 min prior to the formation of a tornado.

Contemporary research has focused on identifying physically plausible radiation modes in the infrasound band and has suggested several mechanisms in addition to the canonically assumed turbulence radiation (Tatom et al., 1995). For example, three dimensional Rossby waves can radiate from the vortex if the maximum wind speed of the vortex exceeds a modest threshold (Schecter et al., 2008), and adiabatic processes involving hail and moisture evaporation/ condensation are likely sources within the supercell (Akhalkatsi and Gogoberidze, 2009, 2011; Schecter, 2011b; Schecter and Nicholls, 2010). However, periodic vortex expansion and contraction is not likely to be a source of infrasound (Schecter, 2011a). No attempt to attribute acoustic energy to specific tornadic mechanisms is made in this investigation, but the ability to reliably detect tornadic activity in the infrasonic and audio bands at distances of practical value is demonstrated, and an argument for the potential use of relatively small, inexpensive arrays (e.g., those with an array diameter on the order of a few tens of meters) of low-frequency audio acoustic sensors to detect, locate, and possibly help characterize tornadoes in real time is demonstrated. Work very similar to that presented here has been reported by Frazier, et al (2014) and Rinehart (2012).