In the companion "Theory" article, we presented a new framing of the seismic location problem in terms of differential geometry (Harris et al., 2025). From that viewpoint, we developed a "project and correct" approach for estimating the relative locations of earthquakes. Here, we use project and correct to estimate high-precision relative locations of events from an earthquake sequence beneath the town of Pahala, Hawaii, using high-precision correlation-derived picks. The sequence was active from 2020 through 2022 and produced many highly correlated signals at Hawaii Volcano Observatory (HVO) stations on the island of Hawaii. The data we inverted consisted of 2882 events with observations at 5 HVO stations. For comparison with the travel-time image, we also produced conventional hypocenter solutions using both the Bayesloc program (Myers et al., 2007, 2009) and a purpose-built double-difference code. There were obvious structural elements in the resulting image, the resolution of which we used to test the performance of the project and the correct algorithm. For the projection step, we first produced a 3D local basis using an singular value decomposition (SVD) of the 2882 groups of times. Projection of the travel-time vectors into this basis resulted in an image with structures similar to those produced by our conventional locators, but with distortion as predicted by theory. Removing the distortion requires an inverse operator generated from the metric tensor at the geometric centroid of the events. We compared two approaches to obtaining such an inverse operator. The first uses an estimate of the geographic centroid of the event cloud from the centroid of the travel-time data. The second approach uses the centroid of the conventionally produced locations. The first approach produces a corrected image very similar to the conventional results, but with a rotation. The corrected image produced using the conventionally derived centroid is a near-exact match to the conventional locations.

Accurate event locations are important for many endeavors in seismology, and understanding the factors that contribute to uncertainties in those locations is complex. In this article, we present a case study that takes an in-depth look at the accuracy and precision possible for locating nine shallow earthquakes in the Rock Valley fault zone in southern Nevada. These events are targeted by the Rock Valley Direct Comparison phase of the Source Physics Experiment, as candidates for the colocation of a chemical explosion with an earthquake hypocenter to directly compare earthquake and explosion sources. For this comparison, it is necessary to determine earthquake hypocenters as accurately as possible so that different source types have nearly identical locations. Our investigations include uncertainty analysis from different sets of phase arrivals, stations, velocity models, and location algorithms. For a common set of phase arrivals and stations, we find that epicentral locations from different combinations of velocity models and algorithms are within 600 m of one another in most cases. Event depths exhibit greater uncertainties, but focusing on the S-P times at the nearest station allows for estimates within approximately 500 m.

This paper combines the power of deep-learning with the generalizability of physics-based features, to present an advanced method for seismic discrimination between earthquakes and explosions. The proposed method contains two branches: a deep learning branch operating directly on seismic waveforms or spectrograms, and a second branch operating on physics-based parametric features. These features are high-frequency P/S amplitude ratios and the difference between local magnitude (ML) and coda duration magnitude (MC). The combination achieves better generalization performance when applied to new regions than models that are developed solely with deep learning. We also examined which parts of the waveform data dominate deep learning decisions (i.e., via Grad-CAM). Such visualization provides a window into the black-box nature of the machine-learning models and offers new insight into how the deep learning derived models use data to make the decisions.

High-frequency ( similar to 2 Hz) seismic P/S amplitude ratios are well-established as a discriminant to distinguish between natural earthquakes and underground explosions at regional distances ( similar to 200-1500 km). As research shifts toward identifying lower-yield events, work has begun to investigate the potential of this discriminant for use at local distances (<200 km), in which initial results raise questions about its effectiveness. Here, we utilize data from several chemical explosion experiment series at the Nevada National Security Site in southern Nevada in the United States to study explosion Pg/Lg ratios across the range of local to regional distances. The experiments are conducted over differing emplacement conditions, with contrasting geologies and a variety of yields and depths of burial, including surface explosions. We first establish the similarities of Pg/Lg ratios from chemical explosions to those from historic nuclear tests and conclude that, as previous data have suggested, chemical explosion ratios are good proxies for nuclear tests. We then examine Pg/Lg ratios from the new experiment series as functions of distance, yield, depth of burial, and scaled depth of burial (SDOB). At far-local and regional distances, we observe consistently higher ratios from hard-rock explosions compared to ones in a weaker dry alluvium medium, consistent with prior regional distance results. No other trends with yield, depth of burial, or SDOB are strongly evident. Scatter in the observed ratios is very high, particularly at the shortest event-to-station distances, suggesting that small-scale path effects play a significant role. On average, the local distance explosion Pg/Lg ratios show remarkable consistency across all the variations in emplacement. Explosion source models will need to reproduce these results.

ABSTRACT Seismologists distinguish underground nuclear explosions from more commonly occurring earthquakes using moment tensor inversion, high-frequency P/S amplitude ratios, mb:Ms comparisons, and P-pP differential travel times. These methods are generally successful for large seismic events (M>3–4) well recorded at regional-to-teleseismic distances (>150 km); however, it is unclear whether they can be modified to work for small events (M<3) well recorded only at local distances (<150 km). Here, we evaluate a recently proposed, local-distance seismic source discriminant—the difference between local magnitude (ML) and coda duration magnitude (MC)—using seismograms of earthquakes and buried, single-fired chemical explosions recorded in three regions of the western United States. The quantity ML–MC was previously found to be sensitive to source depth, effectively discriminating mine blasts, induced earthquakes, and very shallow tectonic earthquakes from deeper crustal earthquakes. In this study, we report the first evaluation of ML–MC as a depth discriminant using data from buried, single-fired explosions that, unlike the seismic sources studied earlier, are good analogs for underground nuclear explosions. We find that even when using generic, uncalibrated methods of assigning magnitudes, ML–MC separates single-fired explosions and earthquakes. The area under the receiver operating characteristic curve is 0.92 for 19 explosions and 14 earthquakes in Washington, 0.90 for 22 explosions and 90 earthquakes in Wyoming, and 0.99 for three explosions and 149 earthquakes in Nevada. ML:MC comparisons have the potential to enhance discrimination based on high-frequency P/S amplitudes ratios—which perform less well at local than regional distances—because the two metrics have complementary sensitivities.

A series of four chemical explosions were detonated in a deep borehole within the Yucca Flat Dry Alluvium Geology (DAG) at the Nevada National Security Site between 2018 and 2019. The two larger chemical explosions of 50 tons (DAG‐2) and 10 tons (DAG‐4) TNT equivalent yield triggered energetic aftershock sequences numbering 1392 and 347 microearthquakes, respectively, within the first 10 days. No significant aftershock activity was observed for the two smaller 1‐ton explosions (DAG‐1 and DAG‐3). We used a back‐projection method based on travel‐time migration and stacking of signal‐to‐noise ratio traces to detect, associate and locate aftershocks from a subset of 22‐geophones within a larger 2 × 2 km seismic array surrounding the borehole. The aftershocks located within 300 m of the borehole and the depths were above the working points of 300 and 50 m depths of DAG‐2 and DAG‐4, respectively, ruling out triggering slip on geologic faults or disturbances beneath neighboring collapse craters. DAG‐2 and DAG‐4 aftershocks decayed at similar rates, with power‐law exponents of p = 1.48 and p = 1.49, respectively. These decay rates are comparable to aftershocks sequences triggered by earthquakes and historical nuclear explosions at Yucca Flat. A smooth power‐law aftershock decay within the first 10 days suggests a triggering mechanism from explosion generated stress relaxation due to the diffusion of high gas pressures in the cavity and radial fractures. A more random and episodic aftershock rate would be expected due to cavity collapse or falling rubble in chimney formation.

Seismic arrays provide useful tools for regional seismic monitoring. Two small-aperture, regionally-oriented arrays, QWAR and HQAR, were deployed in Saudi Arabia and Oman in 2012 and 2016. We utilize a time-domain beampacking method, similar to frequency-wavenumber analysis, to examine the performance of the arrays in terms of slowness and azimuthal bias and event detection capabilities. Additionally, we investigate persistent ambient noise sources recorded by the arrays. We find that the arrays provide slowness vectors with biases comparable with similar-sized arrays of the International Monitoring System in other locations around the world. At QWAR, regional events of magnitude 3.0 and above are detected a majority of the time, and as magnitudes increase to 4.0 and above, the detection rate is greater than 82%. Strong noise generation, primarily with slownesses characteristic of Lg waves, is found in the directions of the northern Arabian/Persian Gulf and southern Red Sea and may be a factor in event detection capabilities.

The ability to discriminate between underground explosions and earth-quakes is an important operation for global security and test-ban-treaty monitoring, and seismology plays a key role in this endeavor. The use of seismic P- to S-wave amplitude ratios has been demonstrated to be an effective discriminant at regional distances for a variety of emplacement conditions and test sites around the world. Increasing interest in detecting and discriminating smaller events, both for purposes of nuclear monitoring and seismic hazard estimation, requires the use of data with shorter event-to-station paths. At these local distances (<= 200 km), the use of P/S ratios for event discrimination is poorly studied compared with regional distances (200-1500 km). The path averaging that allows for relatively simple corrections to P/S ratios to remove bias from along-path structural differences at regional distances may be insufficient to remove enough scatter from the ratios to allow for accurate event identification at local distances. We use data from a series of chemical explosions detonated as part of the Source Physics Experiment at the Nevada National Security Site in southern Nevada and surrounding seismicity to investigate the discrimination performance of P/S ratios at local distances. We find that discrimination at individual stations is highly variable, regardless of frequency. Network averaging of ratios improves discrimination performance above 6 Hz, as a result of the reduction of scatter among the earthquake ratios. However, azimuthal and distance-limited averaging suggests that explosion ratios are affected by a complicated source and path relationship that may lead to substantial variability in local P/S ratios at different locations.

We develop high-resolution, laterally varying attenuation models for the regional crustal phases of Pg and Lg in the area surrounding the Basin and Range Province in the western United States. The models are part of the characterization effort for the Source Physics Experiment (SPE), a series of chemical explosions at the Nevada National Security Site designed to improve our understanding of explosion source phenomenology. To aid in SPE modeling efforts, we focus on improving our ability to accurately predict amplitudes in a set of narrow frequency bands ranging from 0.5 to 16.0 Hz. To explore constraints at higher frequencies where data become more sparse, we test the robustness of the empirically observed power-law relationship between quality factor Q and frequency (Q = Q(0)f(gamma)). Our methodology uses a staged approach to consider attenuation, physics-based source terms, site terms, and geometrical spreading contributions to amplitude measurements. Tomographic inversion results indicate that the frequency dependence is a reasonable assumption as attenuation varies laterally for this region through all frequency bands considered. Our 2D Pg and Lg attenuation models correlate with underlying physiographic provinces, with the highest Q located in the Sierra Nevada Mountains and the Colorado plateau. Compared to a best-fitting 1D model for the region, the 2D model provides an 81% variance reduction overall for Lg residuals and a 75% reduction for Pg. These detailed attenuation maps at high frequencies will facilitate further study of local and regional distance P/S amplitude discriminants that are typically used to distinguish between earthquakes and underground explosions.

The Source Physics Experiment (SPE) is a series of chemical explosions at the Nevada National Security Site (NNSS, formerly the Nevada Test Site) designed to improve our understanding of explosion physics. A future SPE will place an explosion at the hypocenter of a small, shallow earthquake, providing a direct earthquake-to-explosion experiment. Candidate earthquakes for this novel experiment come from a sequence of over 200 unusually shallow events that occurred in Rock Valley, Nevada, in the southeastern portion of the NNSS during 1993. We apply the Bayesloc multiple-event location algorithm (Myers et al., 2007, 2009) to determine the best possible locations and depths for these events. Past nuclear tests in the nearby Yucca Flat on the NNSS are relocated with the same method to provide insight into the accuracy and uncertainties associated with the Bayesloc location results for the Rock Valley earthquakes. This test suggests that we can accurately pinpoint the location of the Rock Valley events within approximately 1 km of their true locations using direct arrival times only. The incorporation of differential arrival times and a potential ground-truth event can significantly decrease the already small uncertainties associated with the epicenter locations. Depth determinations have uncertainties of a few kilometers. Depth uncertainty may be reduced by developing an accurate 3D model of P-wave and S-wave velocity for Rock Valley.