Continental mantle earthquakes are uncommon but hold important clues for understanding lithospheric rheology. Few of these earthquakes (<10) have been documented in western North America, though it is likely more exist owing to difficulties in resolving focal depth for small earthquakes. Mapping their distribution in western North America is key to better characterizing cratonic evolution in this area. Here, we evaluate 25 nominally lower crustal and upper mantle earthquakes in the Milk River region of southwestern Alberta. We absolutely relocate each earthquake and compare depths to published estimates of Moho depth and find 17 earthquakes locate similar to 8-16 km below the Moho. High-precision relative relocations and first-motions define a steeply west-dipping, Moho-crossing normal fault. We hypothesize these earthquakes are related to regional mantle flow patterns interacting with lithospheric edges and are part of a larger trend of upper mantle seismicity along the western edge of North American cratonic lithosphere.

The 10 September 2025 Mw 4.1 earthquake in northeastern Utah, United States, had a focal depth 68 km beneath sea level, which is 20-25 km greater than estimates of local crustal thickness, making it a rare example of a continental mantle earthquake (CME). The focal depth is well resolved from arrival-time inversion (nearest station 13 km away) and moment tensor inversion of regional waveforms. Similar to other CMEs in the Intermountain West, there were no obvious aftershocks or foreshocks, and the waveforms were enriched in high-frequency energy. Spectral modeling gives a stress drop of 80 MPa and a radiation efficiency of 0.08, albeit with large uncertainties. The high stress drop and low radiation efficiency are consistent with a dissipative source process such as thermal runaway. Also similar to previous Intermountain West CMEs, the event occurred along the boundary of the Archean Wyoming craton, where pressure-temperature conditions favor ductile deformation. We hypothesize that edge-driven or regional-scale mantle convection produces increased strain rates near the craton boundary that make either conventional brittle failure or thermal runaway feasible at relatively high pressure-temperature conditions. High conductivity inferred around the edge of the craton may suggest that fluids also contribute to CME occurrence.

Abstract Computing magnitudes is a key part of characterizing earthquakes. They are especially important for developing the magnitude–frequency distributions that are used to forecast shaking from future large earthquakes. Although dozens of magnitude scales have been proposed, moment magnitude (Mw) is preferred because it does not saturate for large events and it is related to physical source properties such as fault area. However, it is challenging to resolve Mw for small earthquakes (Mw<3.5), which dominate regional earthquake catalogs, and seismic networks typically report regional magnitudes (e.g., mbLg, ML, Mc, and Md). Therefore, it is common to develop simple relations that convert regional magnitudes into Mw before calculating magnitude–frequency distributions. Conversion relations can add bias and variance to seismic hazard analysis, and ideally Mw would be calculated directly for all earthquakes. To address this issue, we applied a spectral modeling framework to estimate Mw from Sg waves recorded at distances of 10–300 km for 25,712 natural tectonic earthquakes that occurred in or near Utah between 1 January 2001 and 30 June 2025. The procedure is simple, fast, and robust, and can be run automatically in near-real time, thus potentially contributing to earthquake response products. Refinements to time windows or waveform selection can be implemented during analyst review, similar to the existing workflow for ML or Mc. We determined high-quality moment magnitudes for 63% of the target earthquakes, increasing the number of Utah events with an Mw by a factor of ∼170. The resulting Mw catalog contains 16,173 values ranging from 0.54 to 5.54 with a median of 2.06. A declustered version has a magnitude of completeness of 2.2 and a b value of 1.14 ± 0.02. Its large size will enable improved resolution of spatial and temporal variations in b value, which may allow inference of crustal stress variations in Utah.

In terms of annualized earthquake loss and casualties, Utah has the fourth-highest seismic risk among U.S. states, after California, Washington, and Oregon. Seismic hazard is highest in the urban corridor along the Wasatch fault zone, which has hosted >= 24 surface-rupturing earthquakes (M 6.75+) in the last 6500 yr. Motivated by the 2021 rollout of ShakeAlert along the U.S. West Coast-and the increasing popularity of earthquake early warning (EEW) systems worldwide-the Utah legislature commissioned a study on the feasibility of implementing EEW in the state. Here, we expand on the seismology component of that report, focusing on the required instrumentation upgrades to the existing network, the idealized warning times expected for six representative earthquake scenarios, and the frequency with which alerts would be issued. We investigate a regional EEW system, relying on the requirements of the existing West Coast ShakeAlert system. If the existing Utah seismograph stations were upgraded with low-latency recorders and improved telemetry, we find that a prototype EEW system could currently be operated along the Wasatch fault zone. Installation of 64 additional stations is required for EEW to be fully operational throughout the Wasatch Front, and another 117 stations for EEW to fully operate throughout the Utah Intermountain Seismic Belt. Areas expected to have damaging ground shaking (modified Mercalli intensity [MMI] >= 5.5) in the six earthquake scenarios would receive 0-10 s of warning to prepare for shaking, whereas areas expected to have noticeable ground shaking but little-to-no damage (MMI 1-5.5) would receive 10-55 s of warning. A fully operational Wasatch Front EEW system set to alert on M 4.5+ earthquakes, like the West Coast ShakeAlert system, would be expected to activate once every four years on average, although the alerts would strongly cluster in time, reflecting the natural patterns of earthquake activity.

The Wasatch Front Community Velocity Model (WFCVM) is the most complete and detailed Earth model for the Wasatch Front region in north-central Utah (USA). Until recently, it had not been well evaluated with strong ground motion observations due to a lack of local earthquakes with magnitude M5+. The 2020 March 18 M-w 5.7 Magna, Utah, earthquake generated excellent strong ground motion data at dozens of stations along the Wasatch Front, with peak ground accelerations up to 0.54 g. Here, we use the forward finite-difference code SW4 to simulate waveforms of the 2020 Magna mainshock in the WFCVM up to 3 Hz and compare its predictions to observations from 35 nearby stations at epicentral distances of 4-46 km. We use a finite fault source model with a semistochastic slip distribution and overlay stochastic velocity perturbations (S) and surface topography (T) on the WFCVM, which we refer to as the 3D+S+T model. Observed-predicted amplitude ratios and Goodness-of-Fit (GOF) scores for peak ground acceleration, velocity, displacement, Arias intensity and duration, cumulative energy and duration are calculated. Our 3D+S+T model performed fairly, matching the general character of the observations with an average GOF score of 5.20 (out of a maximum of 10), slightly better than the unaltered WFCVM score (GOF = 4.97). Stochastic velocity perturbations mostly affect peak ground motions at the closest sites (<20 km), while surface topography improves durations for basin sites and generates more realistic signals at higher frequencies. Neither addition resolves underprediction of basin amplification in the eastern Salt Lake Basin and overprediction of ground motion at basin-edge sites, which likely reflect inaccurate representations of basin structure in the WFCVM. Based on these results, we recommend including stochastic velocity perturbations and topography in future simulations but conclude that updating deterministic models of basin structure will lead to the biggest improvement in forecasting ground motion for future large (M6.75+) earthquakes in the Wasatch Front region.

The inner core has been inferred to change its rotation rate or shape over years to decades since the discovery of temporal variability in seismic waves from repeating earthquakes that travelled through the inner core. Recent work confirmed that the inner core rotated faster and then slower than the rest of Earth in the last few decades; this work analysed inner-core-traversing (PKIKP) seismic waves recorded by the Eielson (ILAR) and Yellowknife (YKA) arrays in northern North America from 121 repeating earthquake pairs between 1991 and 2023 in the South Sandwich Islands. Here we extend this set of repeating earthquakes and compare pairs at times when the inner core re-occupied the same position, revealing non-rotational changes at YKA but not ILAR between 2004 and 2008. We propose that these changes originate in the shallow inner core, and so affect the inner-core-grazing YKA ray paths more than the deeper-bottoming ray paths to ILAR. We thus resolve the long-standing debate on whether temporal variability in PKIKP waves results from rotation or more local action near the inner-core boundary: it is tentatively both. The changes near the inner-core boundary most likely result from viscous deformation driven by coupling between boundary topography and mantle density anomalies or traction on the inner core from outer-core convection.

Discriminating low-yield underground nuclear explosions from small earthquakes is a key task in monitoring nuclear test ban treaties. P/S amplitude ratios have been an effective discriminant for moderate-sized events recorded at regional distances, but it is unclear if they are as effective in discriminating small seismic events recorded at local distances (<150 km). The difference between local magnitude (ML) and coda duration magnitude (Mc) has been proposed as a new discriminant that may complement P/S amplitude ratios at local distances. Here, we calculate high-frequency (up to ∼4 Hz) synthetic seismograms at epicentral distances of 0–30 km in realistic models of the Salt Lake basin (Utah, United States) to better understand how variations in source type and depth affect ML−Mc values. The Earth models incorporate simplified 1D and deterministic 3D structures, small-wavelength stochastic velocity perturbations, and surface topography. Coda waves are enhanced for the more complicated models compared to the base 1D model, but still underpredict observed durations by about a factor of two, which results in overprediction of amplitude to duration ratios (i.e., ML−Mc values) for a near-surface explosion and a 7 km deep earthquake. For both source types, the predicted ML and Mc values decrease as source depth increases, and ML−Mc shows only minor variation with depth; however, ML−Mc is on average ∼0.5 units smaller for explosions than earthquakes. This finding may imply that ML−Mc has sensitivity to source type, in addition to being a depth discriminant, but more modeling is needed given the limitations of the current study. Future modeling should incorporate higher-frequency (≳5 Hz) simulations over a larger distance range (0–150 km), where ML and Mc are commonly measured, while honoring low shear velocities (<300 m/s) near the surface and sampling a wider range of earthquake and explosion source mechanisms.

Earthquakes in continental regions overwhelmingly occur in the crust where low pressure and temperature promote brittle failure in response to tectonic stress. In rare cases, primarily in the thickened lithosphere near the Himalayas and Tibet, continental earthquakes occur in the uppermost mantle, perhaps implying an abnormally deep brittle‐ductile transition zone created by relatively low temperatures (≲600°C) and the increased strength of olivine‐rich mantle rocks. Here we present evidence for nine mantle earthquakes—only four of which were previously recognized—along the edge of the Wyoming Craton in the western U.S. Eight of the nine earthquakes occurred >15 km beneath the Moho where temperatures are likely above 700°C. We infer a mixture of brittle and ductile (thermal runaway) source processes facilitated by elevated strain rates from regional or edge‐driven mantle convection, which is thought to be a primary force behind crustal seismicity in the Intermountain West.

Accurately computing the magnitude of small earthquakes during periods of high seismicity rates can be a challenging problem. Here, we introduce a machine learning method which uses features derived from the waveforms of individual phase arrivals and event source parameters to predict the local magnitude (ML) of earthquakes recorded by the Yellowstone seismic network. Our approach works for events with small temporal separation, does not require three-component broadband stations, and produces magnitudes that are consistent with the existing Yellowstone earthquake catalog. We train one support vector machine (SVM) per station-phase pair, resulting in 34 models using P features and 18 models using S features. Producing a model for each station is a straightforward approach to ensure the SVMs learn individual station corrections. In addition, we can easily interrogate, update, and add individual station models going forward. For each station-phase pair, we introduce a recursive feature elimination (RFE) algorithm that simplifies and slightly improves the overall predictive performance of the models. For P and S features, our RFE algorithm reduces 45 potential features to a set of seven common features that work well for nearly all models. In addition, our RFE algorithm limits selection bias and can be used for different monitoring regions, regression algorithms, magnitude types, and features. Using a simple network average of the magnitude predictions of each station model, we achieve excellent ML prediction (R2 similar to 0:95) for two distinct testing sets in the Yellowstone region. Overall, our approach produces accurate ML estimates in the magnitude range of similar to 0-3.5 and increases the number of stations available to compute ML in the Yellowstone region from 14 to 34, lowering the variance of ML estimates. With our approach, we can efficiently compute ML for swarm events and significantly lower the ML magnitude of completeness in the region.

Determining the depths of small crustal earthquakes is challenging in many regions of the world, because most seismic networks are too sparse to resolve trade-offs between depth and origin time with conventional arrival-time methods. Precise and accurate depth estimation is important, because it can help seismologists discriminate between earthquakes and explosions, which is relevant to monitoring nuclear test ban treaties and producing earthquake catalogs that are uncontaminated by mining blasts. Here, we examine the depth sensitivity of several physics-based waveform features for similar to 8000 earthquakes in southern California that have well-resolved depths from arrival-time inversion. We focus on small earthquakes (2< M-L< 4) recorded at local distances (<150 km), for which depth estimation is especially challenging. We find that differential magnitudes (M-w/M-L-M-c) are positively correlated with focal depth, implying that coda wave excitation decreases with focal depth. We analyze a simple proxy for relative frequency content, Phi equivalent to log(10 )(M-0) + 3log(10 )(f(c)), and find that source spectra are preferentially enriched in high frequencies, or "blue-shifted," as focal depth increases. We also find that two spectral amplitude ratios Rg 0.5-2 Hz/Sg 0.5-8 Hz and Pg/Sg at 3-8 Hz decrease as focal depth increases. Using multilinear regression with these features as predictor variables, we develop models that can explain 11%-59% of the variance in depths within 10 subregions and 25% of the depth variance across southern California as a whole. We suggest that incorporating these features into a machine learning workflow could help resolve focal depths in regions that are poorly instrumented and lack large databases of well-located events. Some of the waveform features we evaluate in this study have previously been used as source discriminants, and our results imply that their effectiveness in discrimination is partially because explosions generally occur at shallower depths than earthquakes.

The Utah Legislature, in the 2022 General Session, appropriated funding to study the feasibility of implementing an Earthquake Early Warning (EEW) system in Utah. Funding was provided to the three primary agencies in the Utah Earthquake Program—the Utah Division of Emergency Management (UDEM), the Utah Geological Survey (UGS), and the University of Utah Seismograph Stations (UUSS). The study consisted of four main activities: (1) reviewing the history and development of EEW systems within the U.S. and around the world; (2) assessing the potential performance of an EEW system in Utah; (3) determining what enhancements to the existing Utah seismic network would be needed to implement an EEW system; and (4) conducting an online survey of Utah stakeholders to assess their knowledge of and potential interest in an EEW system.

The solid inner core, suspended within the liquid outer core and anchored by gravity, has been inferred to rotate relative to the surface of Earth or change over years to decades based on changes in seismograms from repeating earthquakes and explosions 1,2 . It has a rich inner structure 3-6 and influences the pattern of outer core convection and therefore Earth's magnetic field. Here we compile 143 distinct pairs of repeating earthquakes, many within 16 multiplets, built from 121 earthquakes between 1991 and 2023 in the South Sandwich Islands. We analyse their inner-core-penetrating PKIKP waves recorded on the medium-aperture arrays in northern North America. We document that many multiplets exhibit waveforms that change and then revert at later times to match earlier events. The matching waveforms reveal times at which the inner core re-occupies the same position, relative to the mantle, as it did at some time in the past. The pattern of matches, together with previous studies, demonstrates that the inner core gradually super-rotated from 2003 to 2008, and then from 2008 to 2023 sub-rotated two to three times more slowly back through the same path. These matches enable precise and unambiguous tracking of inner core progression and regression. The resolved different rates of forward and backward motion suggest that new models will be necessary for the dynamics between the inner core, outer core and mantle. Matching seismic waveforms show that the inner core of Earth gradually super-rotated from 2003 to 2008, and then more slowly sub-rotated from 2008 to 2023 back through the same path.

Summary Classifying the source type of small seismic events is a key task in seismology. A common goal is distinguishing tectonic earthquakes from explosions and human induced seismicity. To this end, we applied a spectral modeling workflow to Pg and Sg waves from ∼10,000 seismic events that occurred in or near Utah and were recorded by broadband seismometers in the western U.S. at distances of 10–300 km. The events were a mixture of tectonic earthquakes (EQ), industrial explosions (EX), and mining-induced seismicity (MIS, primarily collapses) and were mostly small (median magnitude of 1.34 MC). Our spectral modeling was successful for 54% of the events, resulting in a new catalog of M0 and fc values. We evaluated 13 physics-based features—including differential magnitudes, Pg/Sg spectral amplitude ratios, long-period/short-period spectral amplitude ratios, and spectral misfit—as source classifiers. We found that Φ ≡ log10(M0) + 3log10(fc) was the most effective individual feature for distinguishing EQ from EX and MIS sources because EQ spectra are relatively enriched in high frequencies. We selected five less correlated features that spanned the feature space and used a naïve Bayes approach to create a three-way classification model. The model had 97.5% accuracy when applied to an independent test dataset. Model performance deteriorated when more than six features were combined. We conclude that models developed with a few physics-based waveform features can classify small seismic events with performance comparable to high-dimensional deep-learning models. Simple models that rely on physics-based features require less training data and make more interpretable decisions than deep-learning models, though they may require higher signal-to-noise ratios.

ABSTRACT Traditional seismic phase pickers perform poorly during periods of elevated seismicity due to inherent weakness when detecting overlapping earthquake waveforms. This weakness results in incomplete seismic catalogs, particularly deficient in earthquakes that are close in space and time. Supervised deep-learning (DL) pickers allow for improved detection performance and better handle the overlapping waveforms. Here, we present a DL phase-picking procedure specifically trained on Yellowstone seismicity and designed to fit within the University of Utah Seismograph Stations (UUSS) real-time system. We modify and combine existing DL models to label the seismic phases in continuous data and produce better phase arrival times. We use transfer learning to achieve consistency with UUSS analysts while maintaining robust models. To improve the performance during periods of enhanced seismicity, we develop a data augmentation strategy to synthesize waveforms with two nearly coincident P arrivals. We also incorporate a model uncertainty quantification method, Multiple Stochastic Weight Averaging-Gaussian (MultiSWAG), for arrival-time estimates and compare it to dropout—a more standard approach. We use an efficient, model-agnostic method of empirically calibrating the uncertainties to produce meaningful 90% credible intervals. The credible intervals are used downstream in association, location, and quality assessment. For an in-depth evaluation of our automated method, we apply it to continuous data recorded from 25 March to 3 April 2014, on 20 three-component stations and 14 vertical-component stations. This 10-day period contains an Mw 4.8 event, the largest earthquake in the Yellowstone region since 1980. A seismic analyst manually examined more than 1000 located events, including ∼855 previously unidentified, and concluded that only two were incorrect. Finally, we present an analyst-created, high-resolution arrival-time data set, including 651 new arrival times, for one hour of data from station WY.YNR for robust evaluation of missed detections before association. Our method identified 60% of the analyst P picks and 81% of the S picks.

Earth's inner core acquires texture as it solidifies within the fluid outer core. The size, shape and orientation of the mostly iron grains making up the texture record the growth of the inner core and may evolve over geologic time in response to geodynamical forces and torques1. Seismic waves from earthquakes can be used to image the texture, or fabric, of the inner core and gain insight into the history and evolution of Earth's core2-6. Here, we observe and model seismic energy backscattered from the fine-scale (less than 10 km) heterogeneities7 that constitute inner core fabric at larger scales. We use a novel dataset created from a global array of small-aperture seismic arrays-designed to detect tiny signals from underground nuclear explosions-to create a three-dimensional model of inner core fine-scale heterogeneity. Our model shows that inner core scattering is ubiquitous, existing across all sampled longitudes and latitudes, and that it substantially increases in strength 500-800 km beneath the inner core boundary. The enhanced scattering in the deeper inner core is compatible with an era of rapid growth following delayed nucleation.

We conducted a multidisciplinary geophysical experiment to study the relationship between wind, waves, bathymetry, and microseisms in Yellowstone Lake—a small, isolated body of water in which wind‐driven gravity waves excite low‐amplitude seismic energy at periods ∼1.0 s with a regular daily pattern. During 25 June–4 August 2018, we deployed 40 short‐period seismometers, four meteorological stations, and two wave gauges in and around Yellowstone Lake. These instruments were complemented by seven permanent seismometers on the lake perimeter, and a contemporaneous deployment of 10 lake bottom seismometers (LBS) on the floor of the north‐central portion of the lake. The daily wind cycle, with southwesterly winds rising from ∼0 m/s in the early morning to maxima of 10–15 m/s in the late afternoon, generated lake waves with significant heights of ∼0.3–0.4 m and dominant periods of 1.5–2.5 s, which in turn generated microseisms with dominant periods of ∼0.8–1.2 s, in excellent agreement with the double‐frequency generation mechanism. Microseism energy was strongest on the lake floor stations, followed by the island stations and those along the northeastern coast of the lake, suggesting that these sites are closest to the source region where interference between incident and shoreline‐reflected waves is responsible for microseism generation. Smaller amplitude single‐frequency microseism energy was observed primarily along the northeastern lake shoreline. Both types of microseism energy are sensitive to shoreline morphology and lake wave dynamics, such that long‐term monitoring of lake microseisms may further our understanding of how these processes change over time.

ABSTRACT The Yellowstone volcanic region is one of the most seismically active areas in the western United States. Assigning magnitudes (M) to Yellowstone earthquakes is a critical component of monitoring this geologically dynamic zone. The University of Utah Seismograph Stations (UUSS) has assigned M to 46,767 earthquakes in Yellowstone that occurred between 1 January 1984 and 31 December 2020. Here, we recalibrate the local magnitude (ML) distance and station corrections for the Yellowstone volcanic region. This revision takes advantage of the large catalog of earthquakes and an increase in broadband stations installed by the UUSS since the last ML update in 2007. Using a nonparametric method, we invert 7728 high-quality, analyst-reviewed amplitude measurements from 1383 spatially distributed earthquakes for 39 distance corrections and 20 station corrections. The inversion is constrained with four moment magnitude (Mw) values determined from time-domain inversion of regional-distance broadband waveforms by the UUSS. Overall, the new distance corrections indicate relatively high attenuation of amplitudes with distance. The distance corrections decrease with hypocentral distance from 3 km to a local minimum at 80 km, rise to a broad peak at 110 km, and then decrease again out to 180 km. The broad peak may result from superposition of direct arrivals with near-critical Moho reflections. Our ML inversion doubles the number of stations with ML corrections in and near the Yellowstone volcanic region. We estimate that the additional station corrections will nearly triple the number of Yellowstone earthquakes that can be assigned an ML. The new ML distance and station corrections will also reduce uncertainties in the mean MLs for Yellowstone earthquakes. The new MLs are ∼0.07 (±0.18) magnitude units smaller than the previous MLs and have better agreement with 12 Mws (3.15–4.49) determined by the UUSS and Saint Louis University.

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.

Modal analysis of freestanding rock formations is crucial for evaluating their vibrational response to external stimuli, aiding accurate assessment of associated geohazards. Whereas conventional seismometers can be used to measure the translational components of normal modes, recent advances in rotational seismometer technology now allow direct measurement of the rotational components. We deployed a portable, three-component rotational seismometer for a short-duration experiment on a 36 m high sandstone tower located near Moab, Utah, in addition to conducting modal analysis using conventional seismic data and numerical modeling. Spectral analysis of rotation rate data resolved the first three natural frequencies of the tower (2.1, 3.1, and 5.9 Hz), and polarization analysis revealed the orientations of the rotation axes. Modal rotations were the strongest for the first two eigenmodes, which are mutually perpendicular, full-height bending modes with horizontal axes of rotation. The third mode is torsional with rotation about a subvertical axis. Measured natural frequencies and the orientations of displacements and rotation axes match our numerical models closely for these first three modes. In situ measurements of modal rotations are valuable at remote field sites with limited access, and contribute to an improved understanding of modal deformation, material properties, and landform response to vibration stimuli.