Search

We develop a deep learning-based infrasonic detection and categorization methodology that uses convolutional neural networks with self-attention layers to identify stationary and non-stationary signals in infrasound array processing results. Using features extracted from the coherence and direction-of-arrival information from beamforming at different infrasound arrays, our model more reliably detects signals compared with raw waveform data. Using three infrasound stations maintained as part of the International Monitoring System, we construct an analyst-reviewed data set for model training and evaluation. We construct models using a 4-category framework, a generalized noise vs non-noise detection scheme, and a signal-of-interest (SOI) categorization framework that merges short duration stationary and non-stationary categories into a single SOI category. We evaluate these models using a combination of k-fold cross-validation, comparison with an existing “state-of-the-art” detector, and a transportability analysis. Although results are mixed in distinguishing stationary and non-stationary short duration signals, f-scores for the noise vs non-noise and SOI analyses are consistently above 0.96, implying that deep learning-based infrasonic categorization is a highly accurate means of identifying signals-of-interest in infrasonic data records.

The 15 January 2022 climactic eruption of Hunga volcano, Tonga, produced an explosion in the atmosphere of a size that has not been documented in the modern geophysical record. The event generated a broad range of atmospheric waves observed globally by various ground-based and spaceborne instrumentation networks. Most prominent was the surface-guided Lamb wave (≲0.01 hertz), which we observed propagating for four (plus three antipodal) passages around Earth over 6 days. As measured by the Lamb wave amplitudes, the climactic Hunga explosion was comparable in size to that of the 1883 Krakatau eruption. The Hunga eruption produced remarkable globally detected infrasound (0.01 to 20 hertz), long-range (~10,000 kilometers) audible sound, and ionospheric perturbations. Seismometers worldwide recorded pure seismic and air-to-ground coupled waves. Air-to-sea coupling likely contributed to fast-arriving tsunamis. Here, we highlight exceptional observations of the atmospheric waves.

SUMMARY More realistic models for infrasound signal propagation across a region can be used to improve the precision and accuracy of spatial and temporal source localization estimates. Motivated by incomplete infrasound event bulletins in the Western US, the location capabilities of a regional infrasonic network of stations located between 84–458 km from the Utah Test and Training Range, Utah, USA, is assessed using a series of near-surface explosive events with complementary ground truth (GT) information. Signal arrival times and backazimuth estimates are determined with an automatic F-statistic based signal detector and manually refined by an analyst. This study represents the first application of three distinct celerity-range and backazimuth models to an extensive suite of realistic signal detections for event location purposes. A singular celerity and backazimuth deviation model was previously constructed using ray tracing analysis based on an extensive archive of historical atmospheric specifications and is applied within this study to test location capabilities. Similarly, a set of multivariate, season and location specific models for celerity and backazimuth are compared to an empirical model that depends on the observations across the infrasound network and the GT events, which accounts for atmospheric propagation variations from source to receiver. Discrepancies between observed and predicted signal celerities result in locations with poor accuracy. Application of the empirical model improves both spatial localization precision and accuracy; all but one location estimates retain the true GT location within the 90 per cent confidence bounds. Average mislocation of the events is 15.49 km and average 90 per cent error ellipse areas are 4141 km2. The empirical model additionally reduces origin time residuals; origin time residuals from the other location models are in excess of 160 s while residuals produced with the empirical model are within 30 s of the true origin time. We demonstrate that event location accuracy is driven by a combination of signal propagation model and the azimuthal gap of detecting stations. A direct relationship between mislocation, error ellipse area and increased station azimuthal gaps indicate that for sparse networks, detection backazimuths may drive location biases over traveltime estimates.

Infrasonic signals are frequently observed at large propagation distances of 100s to 1000s of kilometers from the source and, in extreme cases, signals have been observed to circle the globe. Simulating acoustic propagation at these large scales requires a framework that not only accounts for 3D propagation effects including along- and cross-path winds, but also for the non-Cartesian geometry of the atmospheric layer surrounding the globe. Furthermore, interaction of propagating infrasonic energy with terrain features is known to have a notable impact on signals, particularly when the wavelength of the signals is comparable to the terrain scale. Finally, spatial variations in the atmospheric structure beyond simple stratification can have significant impacts on propagation paths. Numerical tools for acoustic ray tracing that account for these and other complications will be detailed and demonstrated using several recent infrasonic events of interest.

Infrasound observations and complimentary numerical simulations have shown that infrasound propagation is strongly influenced by topography within approximately 10 km from the source. Recent computational efforts using ray theory have shown that topographic influence extends over hundreds of km and is especially strong when considering propagation through the troposphere. Wind and temperature gradients also have a strong influence on propagation at these distances, which suggests that both topography and 3-D atmospheric structure need to be accounted for in long range waveform modeling. Here we show preliminary results from numerical simulations of linear acoustic propagation through a moving, inhomogeneous atmosphere using an in-development 3-D finite difference time-domain (FDTD) propagation code. We compare our synthetic waveforms in two and three dimensions with existing community infrasound propagation codes and discuss future developments, including open source licensing. Lastly, we present preliminary results from applying this code to the Humming Roadrunner experiments and similar data sets.

Numerous infrasound observations and complementary numerical simulations have shown that infrasound propagation is strongly influenced by terrain within approximately 10 km from the source. Recent computational efforts using ray theory have shown that terrain influence extends over hundreds of km and is especially strong for waves ducted in the troposphere. Wind and temperature gradients also have a strong influence on propagation at these distances, which suggests that both terrain and atmospheric structure need to be accounted for in waveform modeling at a wide range of distances. Here we show preliminary results from numerical simulations of linear acoustic propagationthrough a moving, inhomogeneous atmosphere using a finite-difference time domain propagation code. We compare our synthetic waveforms in two and three dimensions with existing community infrasound propagation codes and discuss future developments, including open source licensing. Finally, we present preliminary comparisons between modeling and observations from a large-scale explosion at the Utah Testing and Training Range. This scenario highlights the importance of using full-wave modeling compared to ray methods to explain the observed waveforms. [Work funded by the Defense Threat Reduction Agency. Cleared for release.]

Supersonic sources such as bolides, aircraft, and rocket launches are known to produce a wide range of acoustic and infrasonic signals. The lower frequency infrasonic component of these signals can propagate large distances through the atmosphere due to the decreased thermo-viscous losses at low frequency and presence of temperature and wind induced infrasonic waveguides so that signals are often detected at regional distances of hundreds of kilometers. The dominant radiator of energy from such sources is the Mach cone signal that emanates from the source with very specific geometry. Ray tracing analysis using a Mach cone source model to predict infrasonic signals at regional distances will be presented and several parametric studies and comparisons with observed signals from several representative supersonic sources will be presented.

Infrasound propagation depends critically on the temperature and wind velocity profiles at the time of signal propagation. These vary wildly, showing at best qualitative systematic behavior. It follows that any predictions of signal detection capability are necessarily statistical in nature. One approach is to collect a large number of historical temperature and wind velocity profiles, for a specific time and location, and use this as a sample space from which to generate a statistical model for the atmosphere as a propagation medium. To generate a statistical model for the expected transmission loss one must run a propagation model through a sampling of atmospheric profiles sufficient to reproduce the statistical behavior. The sampling is done using an Empirical Orthogonal Function (EOF) decomposition. The advantages of the use of an EOF decomposition will be discussed and examples from transition zone (ranges less than 100 km) and regional (ranges of several hundreds of kilometers) will be presented. Given a model for the turbulent pressure fluctuation levels near a receiving array, signal detection probability functions can be estimated.