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.

ABSTRACT Cross-correlation techniques have played a long-standing and pivotal role in seismic event monitoring. However, the performance of correlation-based detectors is challenged by nuisance seismicity, or nontarget signals. Such detections are a problem when the mission is to automatically map events to the correct source region. Using aftershocks of the 2014 Mw 6.0 South Napa, California, earthquake, we demonstrate the effectiveness of utilizing a dynamic correlation processor framework in a generalized likelihood ratio test (GLRT) detector configuration to minimize nontarget detections. A GLRT maximizes a detection statistic with respect to one or more unknown parameters. In this case, the detection statistic is a template signal match against the waveform in a window sliding over a data stream, and the unknown parameter is an index variable indicating group membership of the template event or events. Detected events are assigned to the event group that yields the largest detection statistic. Our results show that a GLRT detector will outperform a suite of independently operating correlation and subspace detectors in terms of having a lower nontarget detection rate at a given missed detection rate. We also show that a GLRT detector composed of a few high-rank subspace detectors has a slightly higher nontarget detection rate, but a significantly lower missed detection rate, than a GLRT detector composed of many low-rank subspace detectors. The high-rank GLRT configuration produced impressive results even with marginal data (single channel, single station, and very low time bandwidth product), which bodes well for the utility of building efficient aftershock classification systems and global monitoring systems at larger scales. However, future work is required to assess performance at the regional scale and to assess the performance of the system at detecting target events not used in the detector template creation.

Under special circumstances, waveform observations of seismic events related by a common, spatially-distributed source process exhibit a geometric architecture that is a distorted image of event distribution in the source region. We describe a prescription for visualizing this signal space image and use a machine-learning algorithm, Isomap, and an algorithm due to Menke to invert collections of waveforms directly for relative event location. We illustrate concepts and methods with well-characterized induced seismicity at a coal mine in the U.S. state of Utah observed by two local seismic instruments, and with synthetics. We anticipate application of these methods to repetitive volcanic seismicity, icequakes and induced seismicity. This is Lawrence Livermore National Laboratory contribution LLNL-ABS-818445.

LLNL completed the FY20 workplan entailing application of the Dynamic Correlation Processor (DCP) to International Monitoring System (IMS) primary seismic data in 2019 and 2020. The effective analysis period was curtailed to January 1 through November 12, 2019 due to a data compression change that rendered later data unusable. Fixing the data compression error is a priority for our data source, and we plan to extend the period of data analysis when the fix is implemented. We found that the percentage overlap between detections generated by the International Data Centre (IDC) DFX software and detections generated by DCP varied widely by station. Overlap is 30-40% at some stations, suggesting that a large percentage of DFX detections could be screened (removed) from the detection-event association algorithm. However, detection overlap was much lower, a few percent, at many stations and the percentage overlap was low at stations known to have many mines and other local sources of seismicity. Unexpectedly low detection overlap at many stations prompted us to examine results for the ARCES station in detail. There is considerable mine activity near ARCES, but our initial FY20 analysis found only ~2% DCP/DFX detection overlap. Differences in the pre-processing (e.g. beam recipes) used by DCP and DFX appear to be the cause of low detection overlap at ARCES. DCP uses wideband filters to improve detector robustness and the IDC uses narrow-band pre-filtering to improve sensitivity. After reprocessing ARCES using the IDC beam recipe, DCP/DFX overlap increased to ~74%. This result shows that we must reprocess the IMS network using the IDC station-specific beam recipes if we are to effectively screen DFX detections and improve automatic event building performance at the IDC.

Examination of regional distance seismic data from historic nuclear test sites has led to a variety of very effective discriminants between explosions, earthquakes, and collapses. We focus on the body-wave methods. We show that ratios between P- and S-wave amplitudes (P/S ratios) above about similar to 2 Hz very effectively separate the six Democratic People's Republic of Korea (DPRK) declared nuclear tests between 2006 and 2017 from natural earthquakes in the region. Similarly, P/S ratios separate historic Nevada Test Site (NTS) nuclear explosions from western U.S. earthquakes. We show that combining P/S ratios with ratios of low-frequency to high-frequency S-wave amplitudes can effectively separate postexplosion collapse events, such as the 1982 NTS Atrisco collapse, and the apparent collapse that followed about eight and a half minutes after the 3 September 2017 DPRK explosion. Explosions often produce fewer and smaller aftershocks than comparably sized earthquakes, which has been proposed as a potential discriminant. We apply the body-wave techniques to the recent seismicity following the largest DPRK event, after first using correlation methods to build a more complete catalog of these events. Despite the empirical effectiveness of the regional body-wave discriminants, the physical basis for the generation of explosion S waves, and therefore the predictability of P/S and low/high frequency techniques, as a function of path, frequency, and event properties such as size, depth, and geology, remains incompletely understood. A goal of current research, such as the Source Physics Experiments (SPE), is to improve our physical understanding of the mechanisms of explosion S-wave generation and advance our ability to numerically model and predict them.

A dynamic correlation processor (DCP) was used to process 90 days of 3-component seismic data recorded by station BCE. Two methods of detector creation were compared. The first attempted to span the signal space with a small number of subspace detectors having minimal projections on one another, and the second used a much larger number of detectors with substantial inter-template projections. After two passes over the data the results were compared, and the second approach was found to have produced more than 3 times the number of detections. For events in common, the average correlation statistic for the second approach was 0.77 compared to a statistic of 0.65 for the first approach. Also, analysis of the detection statistics as a function of SNR showed a strong functional relationship between SNR and correlation statistic for the second approach not found in the first approach. I interpret this to mean that the intra-group variability of signals produced by the second method is due primarily to noise rather than variations in source or path. These results show that (in an environment with large numbers of repeaters) if precise classification is a goal, it is better to operate a DCP with many correlators with substantial inter-template projections and that fully span the signal space. The performance of the DCP was also measured using a small set of “ground truth” picks for BCE from the Chip Brogan catalog. The catalog spanned less than three days out of the 90 days of the experiment and is also incomplete, but the comparison identified 12 signals missed by the DCP because of low SNR. Combining detections by the DCP and picks from the Brogan catalog, identifies 76 unique signals. Of these, the DCP detected 64 (84%) and the Catalog contained between 48/76 (63%) and 48/62 (77%). Analysis of time-of-day and day-of-week statistics was not useful in identifying the signal sources. However, magnitude recurrence statistics were consistent with mining induced seismicity (a preponderance of small sources). Failure of the temporal discriminants may indicate that the events being detected are not coupled in the same way to the mining activity, e.g. secondary failures. Correlation data from explosions at the Nevada test site (NTS) were used to bound the separation of events within groups. Nearly all the NTS correlations with the same passband and time bandwidth product as the detections and having a correlation statistic >= 0.8 are from sources separated by less than 1 km. Even at correlation values of 0.6 nearly all the source separations are less than about 3 km. Because the DCP detectors were formed to span the signal space at a correlation value of 0.9 and because the intra-group variability is most likely due to noise and not source variability, it is likely that the sources associated with each detector are separated by one km or less, and almost certain that the majority are separated by no more than 3 km.

Aftershock sequences following very large earthquakes present enormous challenges to near‐real‐time generation of seismic bulletins. The increase in analyst resources needed to relocate an inflated number of events is compounded by failures of phase‐association algorithms and a significant deterioration in the quality of underlying, fully automatic event bulletins. Current processing pipelines were designed a generation ago, and, due to computational limitations of the time, are usually limited to single passes over the raw data. With current processing capability, multiple passes over the data are feasible. Processing the raw data at each station currently generates parametric data streams that are then scanned by a phase‐association algorithm to form event hypotheses. We consider the scenario in which a large earthquake has occurred and propose to define a region of likely aftershock activity in which events are detected and accurately located, using a separate specially targeted semiautomatic process. This effort may focus on so‐called pattern detectors, but here we demonstrate a more general grid‐search algorithm that may cover wider source regions without requiring waveform similarity. Given many well‐located aftershocks within our source region, we may remove all associated phases from the original detection lists prior to a new iteration of the phase‐association algorithm. We provide a proof‐of‐concept example for the 2015 Gorkha sequence, Nepal, recorded on seismic arrays of the International Monitoring System. Even with very conservative conditions for defining event hypotheses within the aftershock source region, we can automatically remove about half of the original detections that could have been generated by Nepal earthquakes and reduce the likelihood of false associations and spurious event hypotheses. Further reductions in the number of detections in the parametric data streams are likely, using correlation and subspace detectors and/or empirical matched field processing.

Seismic networks around the world use conventional triggering algorithms to detect seismic signals in order to locate local/regional seismic events. Kuwait National Seismological Network (KNSN) of Kuwait Institute of Scientific Research (KISR) is operating seven broad-band and short-period three-component stations in Kuwait. The network is equipped with Nanometrics digitizers and uses Antelope and Guralp acquisition software for processing and archiving the data. In this study, we selected 10 days of archived hourly-segmented continuous data of five stations (Figure 1) and 250 days of continuous recording at MIB. For the temporary deployment our selection criteria was based on KNSN catalog intensity for the period of time we test the method. An autonomous event detection and clustering framework is employed to test a more complete catalog of this short period of time. The goal is to illustrate the effectiveness of the technique and pursue the framework for longer period of time.

Seismotectonic studies of the 2008 Storfjorden aftershock sequence were limited to data acquired by the permanent, but sparse, regional seismic network in the Svalbard archipelago. Storfjorden’s remote location and harsh polar environment inhibited deployment of temporary seismometers that would have improved observations of sequence events. The lack of good station coverage prevented the detection and computation of hypocenter locations of many low magnitude events (mb < 2.5) in the NORSAR analyst-reviewed bulletin. As a result, the fine structure of the sequence’s space–time distribution was not captured. In this study, an autonomous event detection and clustering framework is employed to build a more complete catalog of Storfjorden events using data from the Spitsbergen (SPITS) array. The new catalog allows the spatiotemporal distribution of seismicity within the fjord to be studied in greater detail. Information regarding the location of active event clusters provides a means of inferring the tectonic structure within the fault zone. The distribution of active clusters and moment tensor solutions for the Storfjorden sequence suggests there are at least two different structures within the fjord: a NE-SW trending linear feature with oblique-normal to strike-slip faulting and E-W trending normal faults.

: The objective of this study has been to develop a data-adaptive matched field procedure to detect coherently and incoherently across networks of stations at local and regional distances. The detector extends the single-phase matched field processing approach to detection using the entire waveform. The procedure is based upon a narrowband signal representation that exposes the invariant spatial and temporal correlation structure of network signals from repeating sources. The matched field detector is referred to as an incoherent detector (spatially coherent over a receiver aperture, but temporally incoherent due to the incoherent summation over the narrow frequency bands) and is optimal for sources displaying significant variation in the source time function from event to event. The framework is designed to adapt to the statistics of source time histories for any given target and employs an exponential age-weighting approach to effectively monitor evolving sources, such as open-cast mines. For a given source of interest, an initial detector is formed. If only a single observation is available, the detector can be either a rank-1 coherent detector (a correlator) or an incoherent detector. If multiple observations are available, an event ensemble can be formed, allowing higher rank subspace detectors.

Our identification research for the past several years has focused on the problem of correctly discriminating small- magnitude explosions from a background of earthquakes, mining tremors, and other events. Small magnitudes lead to an emphasis on regional waveforms. It has been shown that at each test site where earthquake and explosions are in close proximity and recorded at the same station, clear differences in the regional body waves, such as the relative high frequency amplitudes of P and S waves, can be used to discriminate between event types. However, path and source effects can also induce such differences; therefore these must be quantified and accounted for. We have been using a specific technique called Magnitude and Distance Amplitude Correction (MDAC), (Walter and Taylor, 2002) with some success to account for some of these effects. Here we briefly present highlights from three aspects of our recent work: 1) Efficient Processing - In order to calibrate the source and path behavior of a particular region we need to be able to quickly and efficiently measure hundreds to thousands of prior events. We have developed processing software called Regional Body-wave Amplitude Processor (RBAP) that is optimized for quickly selecting data, reviewing seismograms and making amplitude measures for the four major regional phases Pn, Pg, Sn and Lg. The code includes an MDAC modeling module. The code works with a database so that changes to parameters (e.g. locations, magnitudes, MDAC values, etc.) are traced and results can easily be updated to reflect the changes. 2) Transportability - We have been re-examining the large database of the western United States (U.S.) underground nuclear tests and earthquakes assembled under a prior agreement. This western U.S. data covers a wide range of depths and material properties and has excellent ground truth information. We define the MDAC parameters based on fitting the earthquakes alone. We are comparing and contrasting how well MDAC corrected events discriminate in the western U.S. compared to MDAC-corrected events in other regions such as Eurasia. Two issues are important: first, how effectively can an earthquake-based MDAC model remove gross path and source effects. Second, how large is the variability in the MDAC- corrected explosion source spectra from region to region. This addresses the issue of whether a particular discriminant such as high frequency P/S works everywhere or only under certain circumstances. 3) Source Models - The MDAC procedure uses a new generalized Brune (1970) style earthquake model to predict expected earthquake spectra. One of its parameters is the apparent stress, a ratio of seismic energy to moment. Currently the geophysical community is split on whether apparent stress is constant or increasing with the size of the earthquake. Here we briefly present some research results examining the scaling of earthquake apparent stress. Another factor in addressing the transportability is the explosion source model. We have started to investigate how effectively the existing explosion models can be incorporated into MDAC using path parameters determined from earthquakes. We present some preliminary results here where we have tested some basic explosion source models on western U.S. explosions.

The National Nuclear Security Administration (NNSA) Ground-Based Nuclear Explosion Monitoring Research and Engineering (GNEM R&E) Program has made significant progress enhancing the process of deriving seismic calibrations and performing scientific integration with automation tools. We present an overview of our software automation efforts and framework to address the problematic issues of very large datasets and varied formats utilized during seismic calibration research. The software and scientific automation initiatives directly support the rapid collection of raw and contextual seismic data used in research, provide efficient interfaces for researchers to measure/analyze data, and provide a framework for research dataset integration. The automation also improves the researcher's ability to assemble quality controlled research products for delivery into the NNSA Knowledge Base (KB). The software and scientific automation tasks provide the robust foundation upon which synergistic and efficient development of seismic calibration research may be built. The task of constructing many seismic calibration products is labor intensive and complex, hence expensive. However, aspects of calibration product construction are susceptible to automation and future economies. We are applying software and scientific automation to problems within two distinct phases or "tiers" of the seismic calibration process. The first tier involves initial collection of waveform and parameter (bulletin) data that comprise the "raw materials" from which signal travel-time and amplitude correction surfaces are derived and is highly suited for software automation. The second tier in seismic research content development activities include development of correction surfaces and other calibrations. This second tier is less susceptible to complete automation, as these activities require the judgment of scientists skilled in the interpretation of often highly unpredictable event observations. Even partial automation of this second tier, through development of prototype tools to extract observations and make many thousands of scientific measurements, has significantly increased the efficiency of the scientists who construct and validate integrated calibration surfaces. This achieved gain in efficiency and quality control is likely to continue and even accelerate through continued application of information science and scientific automation. Data volume and calibration research requirements have increased by several orders of magnitude over the past decade. Whereas it was possible for individual researchers to download individual waveforms and make time-consuming measurements event by event in the past, with the Terabytes of data available today, a software automation framework must exist to efficiently populate and deliver quality data to the researcher. This framework must also simultaneously provide the researcher with robust measurement and analysis tools that can handle and extract groups of events effectively and isolate the researcher from the now onerous task of database management and metadata collection necessary for validation and error analysis. We have succeeded in automating many of the collection, parsing, reconciliation and extraction tasks, individually. Several software automation prototypes have been produced and have resulted in demonstrated gains in efficiency of producing scientific data products. Future software automation tasks will continue to leverage database and information management technologies in addressing additional scientific calibration research tasks.

This project has built a unique historic database of regional distance nuclear explosion, earthquake, and mine-related digital broadband seismograms for the western United States (US). The emphasis is on data from Lawrence Livermore National Laboratory (LLNL)-managed stations MNA, ELK, KNB and LAC that recorded many nuclear tests and nearby earthquakes in broadband digital form since 1980, along with a small number of earlier events that were digitized from tapes. Through the generous cooperation of Sandia National Laboratories (SNL) we have also included waveforms from their Leo Brady network (BMN, DWN, LDS, NEL,TON). In addition we include data from other open broadband stations in the western US with long operating histories and/or ties to the International Monitoring System (e.g. PFO, YKA, CMB, NEW, DUG, ANMO, TUC). These waveforms are associated with a reconciled catalog of events and station response information to facilitate analysis. The goal is to create a high- quality database that can be used in the future to analyze fundamental regional monitoring issues such as detection, location, magnitude, and discrimination.

Monitoring nuclear explosions on a global basis requires accurate event locations. As an example, a typical size used for an on-site inspection search area is 1,000 square kilometers or approximately 17 km accuracy, assuming a circular area. This level of accuracy is a significant challenge for small events that are recorded using a sparse regional network. In such cases, the travel time of seismic energy is strongly affected by crustal and upper mantle heterogeneity and large biases can result. This can lead to large systematic errors in location and, more importantly, to invalid error bounds associated with location estimates. Calibration data and methods are being developed and integrated to correct for these biases. Our research over the last few years has shown that one of the most effective approaches to generate path corrections is the hybrid technique that combines both regionalized models with three-dimensional empirical travel-time corrections. We implement a rigorous and comprehensive uncertainty framework for these hybrid approaches. Qualitative and quantitative validations are presented in the form of single-component consistency checks, sensitivity analysis, measures of robustness, and outlier testing, along with end-to-end testing of confidence measures. We focus on screening and validating both empirical and model-based calibrations as well as the hybrid form that combines these two types of calibration. We demonstrate that the hybrid approach very effectively calibrates both travel-time and slowness attributes for seismic location in the Middle East, North Africa, and Western Eurasia (ME/NA/WE). Furthermore, this approach provides highly reliable uncertainty estimates. Finally, we summarize the National Nuclear Security Administration (NNSA) validated data sets that have been provided to contractors in the last year.

Coda waves are often considered to be generated by backscattering of primary waves from randomly distributed heterogeneities in the crust; however, modeling and experimental work indicates that much of the coda may be due to scattering near the receiver. Knowing where the coda is generated is important for several reasons. Coda waves have been used to characterize site amplification, scattering and attenuation throughout the crust, and velocity changes associated with large earthquakes; however, a lack of knowledge of where coda waves are generated can lead to ambiguous interpretations. In this study we analyze 26 s of coda waves recorded at up to 78 stations using slowness stacking on two source arrays and find that at nearly all stations, regardless of distance from the source, nearly all the arrivals are within one hemisphere, most are strongly clustered near the direct arrival, and the clusters are, in general, symmetrically disposed about the array axis (fault plane). This finding suggests that the coda consists primarily of waves scattered near the stations rather than waves scattered throughout the crust. This result holds even at stations very close to the sources where the amount of coda examined is about 6 times the direct S travel time, and it suggests that much of the coseismic velocity decrease associated with the 1989 Loma Prieta earthquake [Ellsworth et al., 1992] occurred in the shallow crust near the stations. One possible mechanism for inducing such a change is a decrease in the mean stress from the mainshock slip; however, on the basis of static stress modeling, we find that this mechanism would produce a velocity increase in the region where velocity was observed to decrease. We conclude that the velocity decrease was more likely caused by crack opening or crack coalescence due to strong shaking from the mainshock.

We find that foreshocks provide clear evidence for an extended nucleation process before some earthquakes. In this study, we examine in detail the evolution of six California foreshock sequences, the 1986 Mount Lewis (M(L)=5.5), the 1986 Chalfant (M(L)=6.4), the 1986 Stone Canyon (M(L)=4.7), the 1990 Upland (M(L)=5.2), the 1992 Joshua Tree (M(W)=6.1), and the 1992 Landers (M(W)=7.3) sequence. Typically, uncertainties in hypocentral parameters are too large to establish the geometry of foreshock sequences and hence to understand their evolution. However, the similarity of location and focal mechanisms for the events in these sequences leads to similar foreshock waveforms that we cross correlate to obtain extremely accurate relative locations. We use these results to identify small-scale fault zone structures that could influence nucleation and to determine the stress evolution leading up to the mainshock. In general, these foreshock sequences are not compatible with a cascading failure nucleation model in which the foreshocks all occur on a single fault plane and trigger the mainshock by static stress transfer. Instead, the foreshocks seem to concentrate near structural discontinuities in the fault and may themselves be a product of an aseismic nucleation process. Fault zone heterogeneity may also be important in controlling the number of foreshocks, i.e., the stronger the heterogeneity, the greater the number of foreshocks. The size of the nucleation region, as measured by the extent of the foreshock sequence, appears to scale with mainshock moment in the same manner as determined independently by measurements of the seismic nucleation phase. We also find evidence for slip localization as predicted by some models of earthquake nucleation.

The June 28, 1992, Landers, California, earthquake (Mw = 7.3) was preceded for about 7 hours by a foreshock sequence consisting of at least 28 events. In this study we examine the geometry and temporal development of the foreshocks using high‐precision locations based on cross correlation of waveforms recorded at nearby stations. By aligning waveforms, rather than trying to obtain travel time picks for each event independently, we are able to improve the timing accuracy greatly and to make very accurate travel time picks even for emergent arrivals. We perform a joint relocation using the improved travel times and reduce the relative location errors to less than 100 m horizontally and less than 200 m vertically. With the improved locations the geometry of the foreshock sequence becomes clear. The Landers foreshocks occurred at a right step of about 500 m in the mainshock fault plane. The nucleation zone as defined by the foreshock sequence is southeast trending to the south and nearly north trending to the north of the right step. This geometry is confirmed by the focal mechanisms of the foreshock sequence, which are right‐lateral and follow the trend as determined by the foreshock locations on the two straight segments of the fault, and are rotated clockwise for foreshocks that occur within the step. The extent of the foreshock sequence is approximately 1 km both vertically and horizontally. Modeling of the Coulomb stress changes due to all previous foreshocks indicates that the foreshocks probably did not trigger each other. This result is particularly clear for the Mw = 4.4 immediate foreshock. Since stress transfer in the sequence appears not to have played a significant role in its development, we infer an underlying aseismic nucleation process, probably aseismic creep. Other studies have shown that earthquake nucleation may be controlled by fault zone irregularities. This appears to be true in the case of the Landers earthquake, although the size of the irregularity is so small that it is not detectable by standard location techniques.

For years, severe rockburst problems at the Lucky Friday mine in northern Idaho have been a persistent safety hazard and an impediment to production. An MP250 based microseismic monitoring system, which uses simple voltage threshold picking of first arrivals, has been used in this mine since 1973 to provide source locations and energy estimates of seismic events. Recently, interest has been expressed in developing a whole waveform microseismic monitoring system for the mine to provide more accurate source locations and information about source characteristics. For this study, we have developed a prototype whole-waveform microseismic monitoring system based on a 80386 computer equipped with a 50 kHz analog-digital convertor board. The software developed includes a data collection program, a data analysis program, and an event detection program. Whole-waveform data collected and analyzed using this system during a three-day test have been employed to investigate sources of error in the hypocenter location process and to develop an automatic phase picker appropriate for microseismic events.