Abstract Hispaniola lies at the complex boundary between the North American (NOAM) and Caribbean plates, where subduction, collision, and strike-slip tectonic processes generate frequent and damaging earthquakes. However, existing network operations have limited real-time access to stations and, as a result, catalogs are not sufficiently comprehensive to resolve fault structures and subducting slab geometries, constraining seismic hazard assessment. In this study, we applied three widely used deep-learning phase pickers (EQTransformer, generalized phase detection, and PhaseNet) to continuous seismic data from 2014 to 2023. After phase association using PyOcto and event location with Hypoinverse, depth resolution improved significantly, revealing a southeastward-deepening seismicity trend consistent with the subduction processes in Hispaniola. Merging the individual catalogs produced an aggregated machine-learning (ML) catalog containing over 47,000 earthquakes within the Hispaniola region, representing roughly three times more than those reported by local seismic networks over the same period. This ML catalog shows strong agreement with local network catalogs in phase arrival times and event locations, while also capturing low-magnitude seismicity concentrated along major fault zones, including the Enriquillo–Plantain Garden and Septentrional–Oriente fault zone, as well as previously undercharacterized clusters near San José de Ocoa and the margins of Enriquillo Lake. After relocation with hypoDD, the aggregated catalog also enabled imaging of the subducting NOAM slab, which may extend westward to the Haiti–Dominican Republic border. Near ∼70° W, seismicity becomes diffuse and does not clearly define a coherent slab geometry. East of this longitude, seismicity delineates a collision zone where the north-dipping Caribbean slab meets the south-dipping NOAM slab, supporting earlier double-slab interaction models. These results demonstrate that ML-based workflows can generate high-quality seismic catalogs that resolve slab geometries and tectonic processes, thereby improving seismic hazard assessment and supporting real-time monitoring in complex plate boundary regions like Hispaniola.

We use array analysis of auto- and cross-correlations of ambient noise recorded by closely-spaced station pairs in a dense array near Sweetwater, Texas to image the P-wave reflectivity of Earth's upper mantle structure. Imaging was performed with 18 linear arrays, each comprised of 80-100 Fairfield Nodal vertical-component instruments that recorded seismic noise continuously for 8-11 days in March 2014. A coherent signal at similar to 26 s two-way time (TWT) appears consistently across all profiles, including in separate analyses of daytime and nighttime records. This event persists even in gathers formed from single-day recordings without any reported teleseismic events of magnitude >= 5.5, suggesting that teleseismic events are not the sole source of illumination. Analysis of the H/V ratio from a co-located three-component broadband station shows persistently low values over the deployment period, consistent with a wavefield that is dominated by body waves rather than surface waves. These conditions were likely critical to the successful extraction of the reflection response. We interpret the event at similar to 26 s TWT, following a forward modeling exercise, as a PpP-type reflection response from the Lithosphere-Asthenosphere Boundary (LAB) at similar to 90 km depth. We support this interpretation using teleseismic P-wave coda autocorrelograms computed for nearby broadband stations. The presence of this signal in a relatively high frequency band 0.4-4.0 Hz with a positive polarity in autocorrelograms indicates a steep, negative velocity gradient at the LAB.

The Bengal Basin is a sedimentary basin in the northeast region of the Indian subcontinent. It lies between the Indian Shield and the Indo-Burma Ranges, where the India plate is obliquely subducting under the Burma microplate. Multiple interpretations of the nature of the crust here have been proposed. Using a compilation of data from 40 regional broadband stations, we determine the crustal structure by waveform modeling receiver functions and autocorrelograms. We obtain useful velocity models for 30 stations with 2-3 sedimentary units overlying the crystalline crust. The sedimentary section is up to 16.4 km thick with depths increasing from northwest to southeast. The first two sedimentary units have mean thicknesses of similar to 3.2 and similar to 6.5 km and Vp values of similar to 2.8 and similar to 4.9 km/s, respectively. Below these units, large negative Ps conversions are present, which we interpret as two low-velocity zones in the deepest portion of the Bengal Basin, with average Vp and Vp/Vs values of 4.2 km/s and 1.90. The low seismic velocities could be a result of fluids trapped in the deepest sedimentary unit. Below the sedimentary section the thickness of the crystalline crust varies from 12.9 to 34 km, thinning from northwest to southeast in the opposite general trend of basin depth, with an average Vp of 6.7 km/s. The crystalline crust is thinner and faster than typical continental crust and thicker and slower than typical oceanic crust. We suggest the region has extended continental crust that was altered during the Cretaceous rifting that created the Bengal Basin.

Here we model crustal structure across a profile in the south-central United States extending from the Gulf of Mexico to northern Arkansas by jointly modeling P receiver functions and autocorrelograms. Interfaces visible in the P receiver functions vary across the profile. Beneath the Gulf Coastal Plain we find an intra-sedimentary contact, the basin bottom, and the Moho. Beneath the Ouachita Mountains two impedance contrasts are visible, as well as the Moho. In the Arkoma Basin, the basin bottom and Moho are visible. At the northern end of the profile only the Moho is visible. We observe basin depths of 0-13 km, and a Moho depth ranging from similar to 26 to 47 km, with the Moho deepening to the north and shallowing to the south. We also observe faster crustal P velocities south of the Ouachita Mountains and slower values to the north, potentially reflecting changes in bulk crustal composition.

The Caribbean Island of Hispaniola resulted from plate convergence that brought together distinct terranes and may include one or more active microplates. These processes may have brought seismically anisotropic materials to the island or created anisotropic structures through deformation. Our goal is to identify and model interfaces that bound seismically anisotropic media and/or dipping interfaces between isotropic layers beneath Hispaniola and assess the implications of our results for the region's tectonics. We perform harmonic decomposition of receiver function gathers for 17 broadband seismic stations deployed on the island. Our results reveal (a) a layer bounded by interfaces that produce opposite polarity on the transverse receiver functions at 25 and 41 km depth in the northern and eastern Dominican Republic that corresponds to the subducting North American oceanic lithosphere, (b) a layer at ∼8–17 km depth (also bounded by interfaces that produce opposite polarity) beneath the Cordillera Central in the central and eastern parts of the island that may indicate the crystalline, well‐differentiated lower crust of this long‐lived terrane, and (c) an interface that bounds anisotropic material at 22 km depth beneath southwestern Hispaniola. This last interface may correspond to the underthrusting Caribbean Large Igneous Province (CLIP), in which case it would mark the CLIP's upper boundary and the lower boundary is too gradual a transition to produce conversions.

SUMMARY We use continuous data from more than 200 vertical-component broad-band seismic stations for the years 2012 and 2013 to model the crustal and lithospheric structure of the southeastern United States (SEUS). Seismic interferometry via cross-coherence is used to retrieve the surface wave Empirical Green's function (EGF) between station pairs. We mitigate the problem of non-stationary sources contributing to the extracted EGF by applying double beamforming to pairs of subarrays of stations. The recovered Rayleigh waves are used to compute group velocity dispersion maps which are then inverted to find shear wave velocity profiles using a Markov Chain Monte Carlo technique. EGFs are computed for more than 80 000 station pairs after filtering to the period band 7–60 s. Results reveal tectonic features in the SEUS that support recent claims that the northernmost extent of the Suwannee suture is located near the boundary of the Carolinia and Charleston terranes and that the Charleston and Suwannee terranes are not seismologically distinct within the resolution limits of the data set.

We present a novel strategy for joint modelling to estimate shear wave velocities at ten sites in Dhanbad City, India. The joint modelling technique utilizes the complementary sensitivities of disparate datasets, such as horizontal-to-vertical spectral ratios (HVSR) and Rayleigh wave phase velocity dispersion curves, obtained from passive and active seismic surveys, respectively. The model search space is explored rigorously via the very fast simulated annealing (VFSA) global optimization technique and uncertainties are quantified with estimates of marginal posterior probability distributions (PPDs) and parameter correlation matrices (PCMs). These tools help us identify the portions of acceptable models that are well, or poorly, constrained. A confirmatory test is conducted to demonstrate the effectiveness of the modelling scheme by synthetic data as well as field data recorded in Dhanbad City. The depth to engineering bedrock is found to reach its maximum (~ 42 m) and minimum (~ 9.5 m) in the southwestern and northeastern parts of the city, respectively, while dominant frequencies range between 2 and 14.6 Hz. The largest Vs30 value found is 562 m/s at Dhanbad’s southern border and 351 m/s in south-central neighbourhoods. Shear wave velocity profiles are consistent with borehole lithology data. The seismic characterization of Dhanbad City presented here will aid in seismic hazard mitigation efforts and facilitate infrastructure that can sustain the dynamic load imposed by earthquakes.

We develop an original algorithm for velocity estimation that incorporates the constraints of three seismic functionals: surface wave dispersion, P-receiver function, and S-waveform. Here, we use a parallelized reflectivity algorithm to generate synthetic seismograms and match the observed functionals by a global optimization scheme called very fast simulated annealing (VFSA). This method also allows us to assess the uniqueness and parameter independence of the resultant models. Synthetic tests using surface wave (SW) dispersion and receiver functions (RF), and then SW, RF, and waveforms windowed around the S arrival, establish that inclusion of the third functional results in the best recovery of the model. We employ the algorithm to model the Kachchh basin in Gujarat, India, because the area is of active interest for monitoring purposes, and prior results of RF and SW modeling are available for comparison. The waveform functionals used here are generated from broadband seismograms of teleseismic, deep (396–609 km), and moderate- to large-magnitude (6–6.8) earthquake events recorded at semipermanent seismograph stations in the Kachchh basin. Joint inversion of SW, RF, and S-waveform improves the velocity structure by revealing layers that were not identified in previous modeling that used SW and RF alone. Our model clearly shows low-velocity zones (LVZs), crustal and lithospheric thinning, and the lithosphere–asthenosphere boundary. We estimate the crustal thickness to be 41 km at all but one station, and the lithosphere to be in the range of 70–80 km. P- and S-velocities from the uppermost crust to the Moho vary from 4.7 to 7.0 km/s, and 2.7–4.1 km/s, respectively.

We model basin and Moho structure in the Permian Basin region of west Texas and southeastern New Mexico using a method for waveform matching via global optimization of P‐to‐S receiver functions, vertical autocorrelograms, and horizontal autocorrelograms. The algorithm is driven by Particle Swarm Optimization, whose search history can be used to assess the strength of data constraints on model parameters. A common drawback in receiver function modeling is the need to assume a value for Vp before Vs and layer thickness can be estimated. But constraints on Vp can be provided by vertical autocorrelograms of teleseismic arrivals, which detect reverberating P waves, because the phase delay times depend only on Vp and interface depth. P‐to‐S receiver functions and vertical and radial autocorrelograms are computed for M > 5.5 events at 30°–100° epicentral distance, then edited and binned by ray parameter. Synthetic seismograms are computed for layered models using a 1D reflectivity method. The free parameters in the algorithm include basin depth, basin Vp, basin Vp/Vs, and thickness and Vp/Vs of the crystalline crust. Where vertical autocorrelograms prove insufficient for determining basin Vp, Vp is assumed from nearby stations that were modeled successfully. We find an average basin Vp of 4.57 km/s and basin depths to ∼8 km. Moho depths are generally 40–48 km. A basin depth of at least 3.5 km is typically needed to produce sufficient phase separation to allow an average Vp to be determined with confidence. Modeling autocorrelograms jointly with receiver functions therefore improves constraints on deep basin structure.

Recent technological advances have reduced the complexity and cost of developing sensor networks for remote environmental monitoring. However, the challenges of acquiring, transmitting, storing, and processing remote environmental data remain significant. The transmission of large volumes of sensor data to a centralized location (i.e., the cloud) burdens network resources, introduces latency and jitter, and can ultimately impact user experience. Edge computing has emerged as a paradigm in which substantial storage and computing resources are located at the “edge” of the network. In this paper, we present an edge storage and computing framework leveraging commercially available components organized in a tiered architecture and arranged in a hub-and-spoke topology. The framework includes a popular distributed database to support the acquisition, transmission, storage, and processing of Internet-of-Things-based sensor network data in a field setting. We present details regarding the architecture, distributed database, embedded systems, and topology used to implement an edge-based solution. Lastly, a real-world case study (i.e., seismic) is presented that leverages the edge storage and computing framework to acquire, transmit, store, and process millions of samples of data per hour.

Seismic field logistics, particularly recording geometry and duration, can be made significantly more efficient in ambient-noise interferometry if strategies for ensuring timely convergence of Empirical Green's Functions (EGFs) can be devised. Quantitative measures of each cross-correlation panel's constructive contribution to the final result are needed to ensure the best result with the shortest recording duration. We investigate the utility of a variety of seismic data features to improve the convergence of EGFs in ambient-noise interferometry with a dataset acquired at the San Emidio Geothermal Field, Nevada. We identify two features, "rms energy" and "kurtosis", that succeed in distinguishing cross-correlation panels that contribute to EGF convergence from those that degrade the stacked result. Using either of these features allows us to selectively stack data and enhance signal quality at greater offsets and recover weaker arrivals than is possible by stacking all available data. The data used in this study were acquired in May 2019 by a 144-element array of 4.5 Hz geophones recorded by Reftek 130 digitizers in a configuration that was augmented by embedded processors to perform real-time, in-field processing. Such capabilities, coupled with selective stacking strategies, may help minimize the time, cost, and effort required to determine the best possible EGFs.

Seismic interferometry has been shown to extract body wave arrivals from ambient noise seismic data. However, surface waves dominate ambient noise data, so cross-correlating and stacking all available data may not succeed in extracting body wave arrivals. A better strategy is to find portions of the data in which body wave energy dominates and to process only those portions. One challenge is that passive seismic recordings comprise huge volumes of data, so identifying portions with strong body-wave energy could be difficult or time-consuming. We use spatio-temporal features, calculated with data recorded by all receivers together, to perform unsupervised clustering. Using data recorded by a dense seismic array in Sweetwater, TX we were able to identify five clusters, representing a subsets of the complete dataset that contain similar features, and extract a 7 km/s body wave arrival from one cluster. This arrival did not emerge when we performed the same cross-correlation and stacking regimen on the entire dataset.

SUMMARY The northeastern Indian region is characterized by complex lithospheric structure that developed due to collision between the Indian and Eurasian tectonic plates, in the north, and to subduction beneath the Burmese arc, in the east. We report results from joint modelling of Ps and Sp receiver functions and Rayleigh wave group velocity dispersion curves in which a broad search for acceptable models is performed via simulated annealing. We identify three tectonic domains, the Shillong plateau, Brahmaputra valley and Indo-Burma convergence zone (IBCZ), sampled by teleseismic earthquake data recorded by nine broad-band seismic stations. Our results reveal that the region's thinnest crust lies beneath the Shillong plateau, where it increases slightly from the plateau's eastern edge to its centre and reaches a maximum at the western edge of the plateau. Crustal Vp/Vs ratios range between 1.69 and 1.75 for the Shillong plateau, which is consistent with a felsic composition. Deeper Moho depths beneath the Brahmaputra valley, adjacent to the northern front of the Shillong plateau, may be due to the flexure of Indian lithosphere subducting beneath Asia. Low velocity zones are indicated at ∼5–10 km depth beneath the Brahmaputra valley, which may have been developed by NE–SW trending compressional stresses from the collision at the Himalayan arc and subduction at the Burmese arc. The crust is thickest in Kohima, beneath the Naga thrust in the IBCZ, where a high velocity zone is observed for both Vp and Vs at a depth of 25–40 km. This anomaly may be associated with a high velocity slab, trending N–NE to S–SW, that comprises the subducting Indian lithosphere in the IBCZ.

The southern U.S. continental margin records a rich tectonic history spanning ~1.35 Ga, including two complete Wilson cycles. Due to a thick sediment cover, the paucity of local seismicity, and sparse seismic instrumentation, the tectonomagmatic evolution of this passive margin remains uncertain. The recent USArray seismic network offers an opportunity to address unresolved issues by imaging the crust and upper mantle. Direct teleseismic P and S traveltimes recorded at ~1,600 stations are mapped into 3‐D velocity perturbations. P and S tomography models, although generated independently, are broadly consistent, with a strong correlation between anomaly patterns in the sublithospheric mantle and known crustal features. Fast mantle velocities correspond to the Laurentia Craton and Llano and Grenville terranes, while intracratonic rift systems and Paleozoic arc terranes are marked by slow anomalies in the upper mantle.