Abstract Machine learning models for microseismicity detection are often limited by the scarcity of large and high‐quality labeled data sets in many regions. To address this need, we introduce the Oklahoma Labeled AI Dataset (OKLAD), a manually curated data set compiled by the Oklahoma Geological Survey (OGS). OKLAD is designed to support studies of induced seismicity and serves as a benchmark for evaluating deep‐learning detection models in local and regional monitoring contexts. Using OKLAD, we fine‐tuned several established phase‐picking models and observed substantial improvement in local and regional detection. The best performing model achieved recalls of 91.1% for first arrival P‐ detection and 89.8% for first arrival S‐wave detection. Validating this model on continuous data in Oklahoma, we recovered 96.8% of the OGS‐cataloged events and identified 146.8% more events after associative comparison with the events reported by routine network operations. Comparable improvements were also observed when applying the best performing models to other induced seismicity settings, such as west Texas. These results establish OKLAD as a benchmark data set for induced seismicity and demonstrate the effectiveness of transfer learning for improving regional seismic monitoring. This approach provides a replicable framework for enhancing microseismicity detection in challenging environments and can be extended to other regions where generalized deep‐learning pickers may underperform.

Intraplate intermediate-depth earthquakes (IDEs) occur at depths >70 kilometers, which is well below the brittle‒ductile transition in Earth's crust, and their origins are difficult to reconcile with traditional plate tectonics. We applied a deep-learning technique to regional seismic data and identified intraplate IDEs beneath Antarctica. Their occurrence can be explained by concentrated bending stresses resulting from variable, thermally driven uplift along the edge of East Antarctica, where there is an abrupt change in lithospheric strength. Our results support geodynamic model predictions that intraplate seismicity strongly depends on lithosphere‒asthenosphere interactions. The Antarctic ice sheet can also affect regional stress conditions, highlighting the complex interplay between Earth systems. Advanced seismic detection capabilities, such as those used here, could reveal that intraplate IDEs are more common globally than currently recognized.

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.

Regions of intra-sedimentary fluid injections, such as wastewater disposal, CO2 storage, and enhanced geothermal systems, commonly record near- and far-field basement-hosted earthquakes, highlighting complex fluid migration pathways and triggering mechanisms. Here, we present results from a rapid-response magnetotelluric geophysical survey of the 2016 M(w)5.8 Pawnee, Oklahoma earthquake epicenter region, the largest recorded injection-induced event, that aimed to illuminate the fluid migration pathways from the sedimentary Arbuckle carbonate injection zones into the intrabasement hypocentral zones. Intriguingly, our resistivity model reveals a broadly resistive seismogenic basement, relatively conductive shallow basement and Arbuckle intervals, overlain by discontinuous but highly conductive post-Arbuckle units. Porosity estimates indicate extremely low (<3%) basement porosity at >2 km depths, contrasting with intermediate porosity in the fractured top-basement and Arbuckle unit, and highly porous (10%-30%) post-Arbuckle sequences. Arbuckle-to-basement fracture distribution in core samples shows a prominent decrease in fracture intensities across the basement-sedimentary interface, representing a fluid flow barrier, and greater intensities in the Upper-Arbuckle and deeper basement levels. These observations present new evidence suggesting large-scale confinement of injected saline fluids to flow units within the Arbuckle, and potentially significant lateral dissipation of the fluids than previously thought. The results support a remote triggering model for the 2016 event and provide useful insights for understanding seismic triggering in similar fluid-injection settings.

Hydraulic fracturing activities and the subsequent wastewater injection into the subsurface have increased the number of felt seismicity in Oklahoma. Some of the smaller magnitude (M < 3.5) earthquakes have been associated with reported damage to the built environment. We evaluate the subsurface structure of the near-surface sediments in Union City, Oklahoma to identify and delineate subsurface structures that amplify ground motion in the area. The study area is adjacent to the Canadian River, a major braided meandering river that traverses Oklahoma generally from west to east, that has significant alluvial depositions that are visible near roadways and from aerial imagery. We analyzed ambient noise data from 60 continuously-recording seismic nodes as well as approximately two kilometers of electrical resistivity tomography (ERT) and induced polarization (IP) data. We performed horizontal-vertical spectral ratio (HVSR) to deduce the dominant resonant frequencies and paired the resonant frequency information with the higher resolution ERT to resolve the sedimentary structures that produce the observed resonance. Our analysis indicates that the higher resonance frequencies are found further away from the riverbanks, and the inverted depth matches the depth and thickness of the mapped terrace deposits in the area. In the southern area, closer to the riverbank, the dominant resonant frequencies are controlled by thick clay layers that are deeper than the alluvial deposits from the Canadian river. The lower frequency resonance corresponds to the depths where the transition between Permian and Pennsylvanian deposits produce high impedance contrast at the strata boundaries resulting in amplified ground motion.

Between June 2019 and June 2022, 1192 earthquakes of magnitude ML 0.2–3.7 were recorded by the Oklahoma state seismic network near the town of Quinton in eastern Oklahoma. In our analysis of the area, we find that the earthquake sequence was plausibly enhanced due to adjacent hydraulic fracturing (HF) activity that broadened the influence of wastewater injection at a single disposal well. The initial cataloged seismicity indicated earthquake epicenters aligning as a northeast-striking fault south of northeast-striking mapped faults. However, careful relocation suggests the seismicity epicenters overlay a series of east–west-striking faults that are subparallel to the maximum-horizontal stress direction. We reanalyze the seismicity to establish the interaction between HF and wastewater disposal (WD) and their role in inducing seismicity in the area. We relocated earthquakes using a double-difference scheme and a local 1D velocity model to reveal three fault segments oriented in a northwest direction. The spatiotemporal occurrence of seismicity suggests increased pore pressure from the point of injection as the primary driver of seismicity. HF is found to “prime” the faults for the WD to induce and propagate seismicity. Furthermore, the preceding HF may have increased the vertical hydraulic connectivity, allowing a direct vertical propagation of fault permeability. The frequency-index method, which utilizes earthquake waveform frequency content, indicates fluid inclusion into the seismogenic faults during the fault reactivation by wastewater injection.

Determination of the in-situ stress orientations in the subsurface is key to understanding crustal behavior. For example, in Oklahoma and Kansas a surge in seismic activity occurred between 2010 and 2019 with the vast majority of hypocenters located in the crystalline basement. This prompted significant interest in characterizing the stress state in this region through indirect geophysical methods, such as shear-wave anisotropy, which is a technique used to identify the principal stress directions through identification of seismic anisotropy. The interpretation of apparent anisotropy from regional-scale seismic measurements can be somewhat limited due to assumptions regarding the physical mechanism for the observed S-wave velocity polarizations and the difficulty in separating the intrinsic anisotropy from other factors. In this work we have investigated the intrinsic velocity anisotropy of crystalline basement rocks from Oklahoma and Kansas using direct laboratory velocity measurement techniques. Two sets of tests were conducted to measure the horizontal and vertical velocities of each rock sample. Tests were conducted under hydrostatic conditions so that the intrinsic rock properties would be the dominant factor in the observed velocity anisotropy. Stereologic techniques were used to quantify the microstructural variation and relate it to both the laboratory and field observations. The results indicate that there is a non-trivial degree of velocity anisotropy in both the horizontal and vertical directions, varying for rocks from different locations. Microstructural observations of fractures show that horizontal fractures orientations dominate the samples, coinciding with the strike-slip regime of the region. However, velocity polarization and fracture orientations do not always align well. The results indicate a clear intrinsic anisotropy in the basement rocks of Oklahoma and Kansas and our work highlights the need for a suite of other measurements (i.e. borehole breakouts, stress inversion, or others) to aid in determining the stress orientations, aside from relying solely upon shear-wave polarization to determine subsurface relative stress orientations.

SUMMARY Microseismic monitoring is an important technique to obtain detailed knowledge of in-situ fracture size and orientation during stimulation to maximize fluid flow throughout the rock volume and optimize production. Furthermore, considering that the frequency of earthquake magnitudes empirically follows a power law (i.e. Gutenberg–Richter), the accuracy of microseismic event magnitude distributions is potentially crucial for seismic risk management. In this study, we analyse microseismicity observed during four hydraulic fracture treatments of the legacy Cotton Valley experiment in 1997 at the Carthage gas field of East Texas, where fractures were activated at the base of the sand-shale Upper Cotton Valley formation. We perform waveform cross-correlation to detect similar event clusters, measure relative amplitude from aligned waveform pairs with a principal component analysis, then measure precise relative magnitudes. The new magnitudes significantly reduce the deviations between magnitude differences and relative amplitudes of event pairs. This subsequently reduces the magnitude differences between clusters located at different depths. Reduction in magnitude differences between clusters suggests that some attenuation-related biases could be effectively mitigated with relative magnitude measurements. The maximum likelihood method is applied to understand the magnitude frequency distributions and quantify the seismogenic index of the clusters. Statistical analyses with new magnitudes suggest that fractures that are more favourably oriented for shear failure have lower b-value and higher seismogenic index, suggesting higher potential for relatively larger earthquakes, rather than fractures subparallel to maximum horizontal principal stress orientation.

Disposal of industrial wastewater and activities such as CO2 ${\mathrm{C}\mathrm{O}}_{2}$ underground sequestration depend upon pressure conditions within deep geologic reservoirs. Sometimes, injection and storage are associated with induced seismicity, suggested to result from reservoir compartmentalization, leakage into faults, or other mechanisms in the subsurface. To understand subsurface pressure conditions within a major regional disposal reservoir, the Arbuckle Group of Oklahoma, we monitored the water levels in 15 inactive injection wells. The wells were monitored at 30-s intervals, with eight wells monitored since September 2016, and an additional seven from July 2017. All the wells were monitored until early March 2020. Since 2016, hydraulic head decreased in 13 of the 15 wells, proportional to near-borehole fluid pressure even considering decreasing regional injection volumes during this period. The well pressures respond to three types of perturbations: (i) gravitational fluctuations (a.k.a. solid-earth tides) (ii) distal and proximal earthquakes, and (iii) injections into nearby wells. Parameterization of tidal responses illustrates that the near wellbore environments have negative fluid flux (i.e., are leaking). Earthquakes cause differing pressure responses from well to well, with some highly sensitive to proximal events, some to distal events, and some apparently insensitive. Injections have variable impacts in some cases masking tidal and earthquake pressure signals. Collectively, there appears to be a threshold injection rate above which well pressure becomes less sensitive to the volume of injections within 15 km. Multi-scale geological structure and temporal permeability changes are likely controlling the pressure field, indicating leakage of fluids across the system.

Controls on landslides vary as a function of landscape and regional activity. For example, low-relief, woodland regions have slope gradients, soil types, and substrate lithologies that contrast with steeper mountainous regions prone to rock fall and debris flows. Similarly, regional variations in precipitation, earthquakes, and other impacts on landslide surfaces create regional variations in landslide properties. While the controls on landslide characteristics have been extensively studied for high-relief coastal and tectonically active regions, controls on lowrelief landslides have received comparatively less attention. We focus here on a part of the Ozark and Ouachita Mountains in the US southern mid-continent to explore such characteristics of landslides and potential controls in low-relief regions. The area exhibits frequent landslides in soil-covered low-relief forested hillslopes. We evaluated the frequency-size scaling of landslides occurred during periods of different earthquake frequency and precipitation amount (pre- and post-2005). We also produced maps of landslide susceptibility based on random forest machine learning applied to remotely sensed data. We found that landslides are clustered mostly in upland hillslopes, and that small landslides dominate the area, quantified by a landslide frequency-size distribution fitting a double Pareto curve. Additionally, the overall landslide frequency, and potentially the porportion of smaller landslides relative to the larger ones, significantly increased after 2005, the period during which the area also experienced increased induced seismicity and extreme storm events. Approximately 94 % of historical landslides were within random-forest-classified high-landslide probability (probability > 0.5) zones, coinciding with moderate to steep (18(degrees) +/- 9(degrees)) and convergent upland slopes underlain by shale and sandstone. Anomalously high frequency landslides appear to result from triggering by extreme weather, human-induced earthquake activity, and human-induced hillslope modification.

Seismic imaging in 3-D holds great potential for improving our understanding of ice sheet structure and dynamics. Conducting 3-D imaging in remote areas is simplified by using lightweight and logistically straightforward sources. We report results from controlled seismic source tests carried out near the West Antarctic Ice Sheet Divide investigating the characteristics of two types of surface seismic sources, Poulter shots and detonating cord, for use in both 2-D and 3-D seismic surveys on glaciers. Both source types produced strong basal P-wave and S-wave reflections and multiples recorded in three components. The Poulter shots had a higher amplitude for low frequencies (<10 Hz) and comparable amplitude at high frequencies (>50 Hz) relative to the detonating cord. Amplitudes, frequencies, speed of source set-up, and cost all suggested Poulter shots to be the preferred surface source compared to detonating cord for future 2-D and 3-D seismic surveys on glaciers.

As seismic data collection continues to grow, advanced automated processing techniques for robust phase identification and event detection are becoming increasingly important. However, the performance, benefits, and limitations of different automated detection approaches have not been fully evaluated. Our study examines how the performance of conventional techniques, including the Short-Term Average/Long-Term Average (STA/LTA) method and cross-correlation approaches, compares to that of various deep learning models. We also evaluate the added benefits that transfer learning may provide to machine learning applications. Each detection approach has been applied to three years of seismic data recorded by stations in East Antarctica. Our results emphasize that the most appropriate detection approach depends on the data attributes and the study objectives. STA/LTA is well-suited for applications that require rapid results even if there is a greater likelihood for false positive detections, and correlation-based techniques work well for identifying events with a high degree of waveform similarity. Deep learning models offer the most adaptability if dealing with a range of seismic sources and noise, and their performance can be enhanced with transfer learning, if the detection parameters are fine-tuned to ensure the accuracy and reliability of the generated catalog. Our results in East Antarctic provide new insight into polar seismicity, highlighting both cryospheric and tectonic events, and demonstrate how automated event detection approaches can be optimized to investigate seismic activity in challenging environments.

Landform maps are important tools in assessment of soil- and eco-hydrogeomorphic processes and hazards, hydrological modeling, and natural resources and land management. Traditional techniques of mapping landforms based on field surveys or from aerial photographs can be time and labor intensive, highlighting the importance of remote sensing products based automatic or semi-automatic approaches. In addition, the time-intensive manual labeling can also be subjective rather than an objective identification of the landform. Here we implemented such an objective approach applying a random forest machine learning algorithm to a set of observed landform data and 1m horizontal resolution bare-earth digital elevation model (DEM) developed from airborne light detection and ranging (LiDAR) data to rapidly map various landforms of a hilly landscape. The landform classification includes upland plateaus, ridges, convex slopes, planar slopes, concave slopes, stream channels, and valley bottoms, across a 400 km(2) hilly landscape of the Ozark Mountains in northeastern Oklahoma. We used 4200 landform observations (600 per landform) and eight topographic indices derived from 2 m, 5 m and 10 m resolution LiDAR DEM in random forest algorithm to develop 2 m, 5 m and 10 m resolution landform models. We test the effectiveness of DEM resolution in mapping landforms via comparison of observed landforms with modeled landforms. Results showed that the approach mapped similar to 84% of observed landforms when covariates were at 2 m resolution to similar to 89% when they were at 10 m resolution. However, predicted maps showed that the 2 m resolution covariates performed better at capturing accurate landform boundaries and details of small-sized landforms such as stream channels and ridges. The approach presented here significantly reduces the time required for mapping landforms compared to traditional techniques using aerial imagery and field observations and allows incorporation of a wide variety of covariates. The landform map developed using this approach has several potential applications. It could be utilized in hydrological modeling, natural resource management, and characterizing soil-geomorphic processes and hazards in a hilly landscape.

ABSTRACT In the period between 1961 and 2008, Oklahoma, USA, averaged about two M ≥3.0 earthquakes per year, with no damage to any built infrastructure. A substantial increase in seismic activity was first observed in 2009, when there were 20 M ≥3.0 earthquakes, and activity peaked in 2015, when over 900 M ≥3.0 earthquakes occurred. Because of the unprecedented increase in seismic activity, the governor’s office of Oklahoma formed a Coordinating Council of researchers, regulators, industry, and other stakeholders in 2015. The Coordinating Council was led by the Secretary of Energy and Environment and charged with understanding and attempting to mitigate (that is, reduce, if not eliminate) induced seismicity and potential impacts. Major outcomes of the coordinated efforts included delineation of an area of interest (AOI) for seismicity in Oklahoma, modifications to underground injection control (UIC) well completion depths and injection rates into UIC wells in the AOI, development of the Oklahoma Well and Seismic Monitoring (OWSM) application used for regulatory oversight and action, modified well completion protocols, a more robust seismic network, and numerous scientific investigations and publications. Because of concerted efforts between regulators and industry, disposal into the Arbuckle Group, the primary zone for wastewater disposal, in the AOI was reduced by more than 50% though oil production continued to increase. Seismic activity decreased over a 6 yr period with 619, 302, 195, 65, 39, and 29 M ≥3.0 earthquakes occurring in 2016, 2017, 2018, 2019, 2020, and 2021, respectively. At the time of latest updates to this chapter (16 October 2022), there have been 12 M ≥3.0 earthquakes and one M ≥4.0 earthquake, so the projected total of M ≥3.0 earthquakes in 2022 is 17. Using these metrics, the coordinated efforts of Oklahoma stakeholders appear to have successfully reduced seismicity with respect to frequency and number in the range of minor but often felt (M 3.0–3.9), light (M 4.0–4.9), and moderate (M 5.0–5.9) earthquakes. So, the Oklahoma case provides examples of how stakeholder action diminished seismic hazards and how similar actions could be used to reduce induced seismicity in other areas where injections occur.

Underground wastewater injection into deep reservoirs confronts complex hydrogeologic conditions potentially leading to induced seismicity. We focus on the geologic units of the Arbuckle Group carbonates in Oklahoma that are frequently used for disposing of produced water associated with oil and gas production. Subsurface pressures and fluid flow figure prominently in most explanations for induced seismicity and are important in evaluations of future storage potential of the Arbuckle. To understand subsurface pressure conditions within the Arbuckle we monitored the water levels in 15 inactive wells. The wells were monitored at 30-second intervals, with eight wells monitored since September 2016, and an additional seven from July 2017. All of the wells were monitored until early March 2020. Since 2016, 13 of the 15 wells showed a net decrease in well level (a.k.a. hydraulic head), proportional to near-borehole fluid pressure. The pressure patterns observed in each well vary from one another, and some wells display a gradual decrease in pressure over time, some have a rapid decrease, and others show irregular changes. The well pressures respond to Earth tides and injections into nearby wells as well as response to distal and proximal seismic waves, though responses vary considerably between wells. Moreover, there appears to be a threshold injection rate above which pressure changes level off. The data illustrate that Arbuckle hydrogeology is a multi-scale, temporally dynamic system, with regional heterogeneity of porosity and permeability. These dynamics exert important controls on seismic hazard and storage capacity estimates.

During routine operations monitoring Oklahoma earthquakes, we found that certain earthquakes occurred closely both in space and time and had overlapping phases at the recording stations. Through further scrutiny and analysis, we determined that rather than being distinctly different earthquakes, some of the earthquakes exhibited multiphase arrivals and longer than expected coda due to unique ray paths that encounter impedance contrasts such as at the sedimentary rock-basement. Of course, some of these events truly were distinct events, which we term overlapping earthquakes, for which perceived coda duration overlaps and obscures the phase arrivals of the second event due to the source proximity in both time and space. We detail our classification scheme to separate the local earthquakes in Oklahoma as single, overlapping earthquakes, or those associated with multiphase arrivals. We forward model seismic wave propagation in a 2D crustal model and develop a methodology that utilizes waveform correlation to distinguish phases from overlapping earthquakes to those from crustal reverberations. Duration analysis shows a more elongated duration, qualitatively similar to the duration produced by overlapping earthquakes, at the sites where multiphase arrivals are observed.

ABSTRACT We report on the source of seismoacoustic pulses that were observed across the state of Oklahoma (OK) during summer of 2019, and the subject of national media coverage and speculation. Seismic network data collected across four U.S. states and interviews with witnesses to the pulse’s effect on residential structures demonstrate that they were triggered by routine ammunition disposal operations conducted by McAlester Army Ammunition Plant (McAAP). During these operations, conventional explosives destroy obsolete munitions stored in pits through a controlled sequence of electronically timed shots that occur over tens of minutes. Despite noise-abatement efforts that reduce coupling of acoustic energy with air, some lower frequency, subaudible (infrasonic) sound radiates from these shots as discrete pulses. We use nine months of blast log documents, seismic network records, analyst picks, and physical modeling to demonstrate that seismic stations as far as 640 km from McAAP sample these pulses, which record seasonal patterns in stratospheric and tropospheric winds, as well as the dynamic formation of waveguides and shadow zones. Digital short-term average to long-term average detectors that we augment with dynamic thresholds and time-binning operations identify these pulses with a fair probability, when compared with visual observations. Our analyses thereby provide estimates of observation rates for both partial and full sequences of these pulses, as well as single shots. We suggest that disposal operations can exploit existing, composite seismic networks to predict where residents are likely to witness blasting. Crucially, our data also show that dense seismic networks can record multiscale atmospheric processes in the absence of infrasound arrays.

To better understand relationships among crustal anisotropy, fracture orientations, and the stress field in Oklahoma and southern Kansas, we conduct shear-wave splitting analysis on the last 9 yr of data (2010-2019) of local earthquake observations. Rather than a predominant fast direction (o), we find that most stations have a primary fast direction of polarization (opri) and a secondary fast direction of polarization (osec). At most stations, either the primary fast direction of polarization (opri) or the secondary fast direction of polarization (osec) is consistent with the closest estimated maximum horizontal stress (sigma H max) orientation in the vicinity of the observation. The general agreement between fast directions of polarization (o) and the maximum horizontal stress orientations (sigma H max) at the regional level implies that the fast polarization directions (o) are extremely sensitive to the regional stress field. However, in some regions, such as the Fairview area in western Oklahoma, we observe discrepancies between fast polarization directions (o) and maximum horizontal stress orientations (sigma H max), in which the fast directions are more consistent with local fault structures. Overall, the primary fast direction of polarization (opri) is mostly controlled and influenced by the stress field, and the secondary fast direction of polarization (osec) likely has some geologic structural control because the secondary direction is qualitatively parallel to some mapped north-striking fault zones. No significant changes in fast directions over time were detected with this technique over the 5 yr (2013-2018) of measurements, suggesting that pore pressure may not cause a significant enough or detectable change above the magnitude of the background stress field.