The standard seismic explosion-monitoring paradigm is based on a sparse, spatially aliased network of stations to monitor either the whole Earth or a region of interest. Under this paradigm, state-of-the-art event-detection methods are based on seismic phase picks, which are associated at multiple stations and located using 3D Earth models. Here, we revisit a concept for event-detection that does not require phase picks or 3D models and fuses detection and association into a single algorithm. Our pickless event detector exploits existing catalog and waveform data to build an empirical stack of the full regional seismic wavefield, which is subsequently used to detect and locate events at a network level using correlation techniques. We apply our detector to seismic data from Utah and evaluate our results by comparing them with the earthquake catalog published by the University of Utah Seismograph Stations. The results demonstrate that our pickless detector is a viable alternative technique for detecting events that likely requires less analyst overhead than do the existing methods.
Park, Junghyun; Stump, Brian W.; Hayward, Chris; Arrowsmith, Stephen J.; Che, Il Y.; Drob, Douglas P.
This work quantifies the physical characteristics of infrasound signal and noise, assesses their temporal variations, and determines the degree to which these effects can be predicted by time-varying atmospheric models to estimate array and network performance. An automated detector that accounts for both correlated and uncorrelated noise is applied to infrasound data from three seismo-acoustic arrays in South Korea (BRDAR, CHNAR, and KSGAR), cooperatively operated by Korea Institute of Geoscience and Mineral Resources (KIGAM) and Southern Methodist University (SMU). Arrays located on an island and near the coast have higher noise power, consistent with both higher wind speeds and seasonably variable ocean wave contributions. On the basis of the adaptive F-detector quantification of time variable environmental effects, the time-dependent scaling variable is shown to be dependent on both weather conditions and local site effects. Significant seasonal variations in infrasound detections including daily time of occurrence, detection numbers, and phase velocity/azimuth estimates are documented. These time-dependent effects are strongly correlated with atmospheric winds and temperatures and are predicted by available atmospheric specifications. This suggests that commonly available atmospheric specifications can be used to predict both station and network detection performance, and an appropriate forward model improves location capabilities as a function of time.
Sandia journal manuscript; Not yet accepted for publication
Park, Junghyun P.; Hayward, Chris H.; Zeiler, Cleat P.; Arrowsmith, Stephen J.; Stump, Brian S.
Automated detections calculated by the progressive multi-channel correlation (PMCC) method (Cansi, 1995) and the adaptive F detector (AFD) (Arrowsmith et al., 2009) are compared to the signals identified by five independent analysts. Each detector was applied to a four-hour time sequence recorded by the Korean infrasound array CHNAR. This array was used because it is composed of both small (<100 m) and large (~1000 m) aperture element spacing. The four hour time sequence contained a number of easily identified signals under noise conditions that have average RMS amplitudes varied from 1.2 to 4.5 mPa (1 to 5 Hz), estimated with running five-minute window. The effectiveness of the detectors was estimated for the small aperture, large aperture, small aperture combined with the large aperture, and full array. The full and combined arrays performed the best for AFD under all noise conditions while the large aperture array had the poorest performance for both detectors. PMCC produced similar results as AFD under the lower noise conditions, but did not produce as dramatic an increase in detections using the full and combined arrays. Both automated detectors and the analysts produced a decrease in detections under the higher noise conditions. Comparing the detection probabilities with Estimated Receiver Operating Characteristic (EROC) curves we found that the smaller value of consistency for PMCC and the larger p-value for AFD had the highest detection probability. These parameters produced greater changes in detection probability than estimates of the false alarm rate. The detection probability was impacted the most by noise level, with low noise (average RMS amplitude of 1.7 mPa) having an average detection probability of ~40% and high noise (average RMS amplitude of 2.9 mPa) average detection probability of ~23%.