Omar Marcillo

On infrasound generated by wind farms and its propagation in low‐altitude tropospheric waveguides

Infrasound from a 60-turbine wind farm was found to propagate to distances up to 90 km under nighttime atmospheric conditions. Four infrasound sensor arrays were deployed in central New Mexico in February 2014; three of these arrays captured infrasound from a large wind farm. The arrays were in a linear configuration oriented southeast with 13, 54, 90, and 126 km radial distance and azimuths of 166°, 119°, 113°, and 111° from the 60 1.6-MW turbine Red Mesa Wind Farm (RMWF), Laguna Pueblo, New Mexico, USA. Peaks at a fundamental frequency slightly below 0.9 Hz and its harmonics characterize the spectrum of the detected infrasound. The generation of this signal is linked to the interaction of the blades, flow gradients, and the supporting tower. The production of wind-farm sound, its propagation, and detection at long distances can be related to the characteristics of the atmospheric boundary layer. First, under stable conditions, mostly occurring at night, winds are highly stratified, which enhances the production of thickness sound (TS) and the modulation of other higher-frequency wind turbine sounds. Second, nocturnal atmospheric conditions can create low-altitude waveguides (with altitudes on the order of 100s of meters) allowing long distance propagation. Third, night and early morning hours are characterized by reduced background atmospheric noise that enhances signal detectability. This work describes the characteristics of the infrasound from a quasi-continuous source with the potential for long-range propagation that could be used to monitor the lower part of the atmospheric boundary layer.

Analysis and modeling of infrasound from a four-stage rocket launch

Infrasound from a four-stage sounding rocket was recorded by several arrays within 100 km of the launch pad. Propagation modeling methods have been applied to the known trajectory to predict infrasonic signals at the ground in order to identify what information might be obtained from such observations. There is good agreement between modeled and observed back azimuths, and predicted arrival times for motor ignition signals match those observed. The signal due to the high-altitude stage ignition is found to be low amplitude, despite predictions of weak attenuation. This lack of signal is possibly due to inefficient aeroacoustic coupling in the rarefied upper atmosphere.

Extracting changes in air temperature using acoustic coda phase delays

Blast waves produced by 60 high-explosive detonations were recorded at short distances (few hundreds of meters); the corresponding waveforms show charge-configuration independent coda-like features (i.e., similar shapes, amplitudes, and phases) lasting several seconds. These features are modeled as reflected and/or scattered waves by acoustic reflectors/scatters surrounding the explosions. Using explosion pairs, relative coda phase delays are extracted and modeled as changes in sound speed due to changes in air temperature. Measurements from nearby weather towers are used for validation.

Infrasound Signal Detection: Re-examining the Component Parts that Makeup Detection Algorithms

Detecting a Signal Of Interest (SOI) is the first step in many applications of infrasound monitoring. This intuitively simple task is defined as separating out signals from background noise on the basis of the characteristics of observed data; it is, however, deceptively complex. The problem of detecting signals requires multiple processes that are divisible at their highest level into several fundamental tasks. These tasks include (1) defining models for SOIs and noise that properly fit the observations, (2) finding SOIs amongst noise, and (3) estimating parameters of the SOI (e.g., Direction Of Arrival (DOA), Signal-to-Noise Ratio (SNR) and confidence intervals) that can be used for signal characterization. Each of these components involves multiple subcomponents. Here, we explore these three components by examining current infrasound detection algorithms and the assumptions that are made for their operation and exploring and discussing alternative approaches to advance the performance and efficiency of detection operations. This chapter does not address new statistical methods but does offer some insights into the detection problem that may motivate further research.

Estimation of explosive yield using regional distance infrasound

Infrasound signals from large above-ground explosions are known to propagate to significant distances while remaining at observable levels due to a combination of the large source energy for such events and decreased absorption for acoustic energy at lower frequencies. Previously, propagation-based, stochastic path geometry and travel time models have been used to improve estimates of location and time for infrasonic sources with promising results. A similar approach has been applied to develop stochastic transmission loss models for infrasonic signals. The resulting models have been utilized in a Bayesian framework to compute a near-source estimate of the acoustic signal, which can then be used to compute a probability distribution for possible explosive yields of the source. Details of the propagation models and yield estimation framework will be presented along with results for application to a number of explosive events.

Improvements to infrasonic detection capabilities through application of a generalized least squares beamformer

A common challenge in seismoacoustic research and analysis is detection of low signal-to-noise ratio (SNR) signals in a variable and often coherent background. Previously, the application of an adaptive F-detector to infrasonic data has successfully reduced false detection rates attributed to coherent noise across array elements; however, the detector is applied post-processing after analysis using a standard (Bartlett) beamformer, which raises the detection threshold and can lead to missed detections. Application of a generalized least squares (GLS) beamformer can enhance processing of transient infrasonic signals, particularly in the presence of correlated noise, by adaptively accounting for the background noise characteristics. The GLS beamformer enhancement leads to improved capabilities for accurately detecting low amplitude infrasonic events of interest that would not normally produce a sufficiently high F-statistic to be declared a detection. A statistical analysis of GLS beamformer performance in ...

Improving infrasonic location estimates for underground nuclear explosions Improving infrasonic location estimates for underground nuclear explosions

Infrasound data from underground nuclear explosions conducted by North Korea in 2006, 2009, 2013 and 2016 were recorded on six seismo-acoustic arrays co-operated by Southern Methodist University (SMU) and the Korean Institute of Geosciences and Mineral Resources (KIGAM). No infrasound signals were observed during the 2006 test, while signals from the others have been used to determine event locations and yield estimations. Prior location studies have demonstrated that wind corrections for back azimuth deviation improve location estimates. Additionally, recent improvements to the Bayesian Infrasonic Source Localization (BISL) methodology have shown to reduce 90% confidence contours for location by 40% through the utilization of propagation-based likelihood priors for celerity and backazimuth deviation from seven years of archival atmospheric specifications. Relocations of the 2009, 2013 and 2016 nuclear explosions will be presented to demonstrate the application of BISL to underground nuclear explosions.

Analysis of explosive sources using a distributed network of infrasound arrays

Analysis of signatures observed on a distributed network requires a model describing the propagation of the energy between network nodes as well as a statistical framework to quantify confidence in analysis conclusions. For the case of infrasound analysis, the propagation model must account for the inhomogeneous, dynamic nature of the atmosphere and the resulting temporal and spatial variations of propagation effects as well as the anisotropic nature of infrasonic propagation due to the dependence on the directionality of the winds. Association and localization methods have been developed utilizing a Bayesian framework to identify explosive sources from distant, infrasonic observations with possible applications to other infrasonic sources. The confidence of the analysis conclusions is dependent on the accuracy and realism of the likelihood definitions utilized in the method, which are therefore an area of ongoing research. An overview of the association and localization methods will be presented along wi...

Measuring the coherency of microbaroms at long distances for the inversion of stratospheric winds

We demonstrate the design of an infrasound network and the associated analysis for measuring the coherency of microbaroms at large distances (10s of km) and inverting for stratospheric winds. We have developed a mathematical framework for the inversion of local stratospheric winds using microbaroms, and found constraints on the optimum sensor network topology. Based on these results, we deployed a prototype sensor network over the winter months (January to March, 2013) that comprised three single-sensor stations, one 30-m and two 1-km arrays with separations between 5 and 70 km. The initial analysis shows periods of very high coherency lasting several hours with tropospheric and low stratospheric celerities. Coherency decreases rapidly with distance and azimuth compared to the direction of propagation of microbaroms. We are exploring topography as the cause of low signal coherency at long distances. Following this pilot study, we are designing a denser sensor network further optimized to capture microbaro...

Processing international monitoring system infrasound data to detect and locate global events using probabilistic algorithms

Automating the detection and location of events using the International Monitoring (IMS) System infrasound network is a significant challenge. Any algorithm must reliably detect nuclear tests in the atmosphere with a minimum number of false alarms. Here, we report on the application of probabilistic techniques for detection, association, and location of infrasound events to data from the IMS network. We compare our results with the SEL3 automatic event detections obtained by the CTBTO.