Search USGS⌕ Search

Geology topics

Elizabeth S. Cochran

Publications and source records attributed to Elizabeth S. Cochran.

At least 37 records · Page 2Linked to original sources
USGS2024

Modeling protective action decision-making in earthquakes by using explainable machine learning and video data

Earthquakes pose substantial threats to communities worldwide. Understanding how people respond to the fast-changing environment during earthquakes is crucial for reducing risks and saving lives. This study aims to study people’s protective action decision-making in earthquakes by leveraging explainable machine learning and video data. Specifically, this study first collected real-world CCTV footage and video postings from social media platforms, and then identified and annotated changes in the environment and people’s behavioral responses during the M7.1 2018 Anchorage earthquake. By using the fully annotated video data, we applied XGBoost, a widely-used machine learning method, to model and forecast people’s protective actions (e.g., drop and cover , hold on , and evacuate ) during the earthquake. Then, explainable machine learning techniques were used to reveal the complex, nonlinear relationships between different factors and people’s choices of protective actions. Modeling results confirm that social and environmental cues played critical roles in affecting the probability of different protective actions. Certain factors, such as the earthquake shaking intensity and number of people shown in the environment, displayed evident nonlinear relationships with the probability of choosing to evacuate . These findings can help emergency managers and policymakers design more effective protective action recommendations during earthquakes.

Scientific Reports↗
USGS2024

Incorporating intensity distance attenuation into PLUM ground-motion-based earthquake early warning in the United States: The APPLES configuration

We develop Attenuated ProPagation of Local Earthquake Shaking (APPLES), a new configuration for the United States West Coast version of the Propagation of Local Undamped Motion (PLUM) earthquake early warning (EEW) algorithm that incorporates attenuation into its ground-motion prediction procedures. Under APPLES, instead of using a fixed radius to forward-predict observed peak ground shaking to the area surrounding a seismic station, the forward-predicted intensity at a location depends on the distance from the station using an intensity prediction relationship. We conduct conceptual tests of maximum intensity distribution predictions in APPLES and PLUM using a catalog of ShakeMaps to confirm that the attenuation relationship in APPLES is appropriately modeling shaking distributions for West Coast earthquakes. Then, we run APPLES and PLUM in simulated real-time tests to determine warning time performance. Finally, we compare real-time alert behavior during the 2022 M 6.4 Ferndale, California, earthquake and other recent events. We find that APPLES presents two potential improvements to PLUM by reducing over-alerting during smaller magnitude earthquakes and by increasing warning times in some locations during larger earthquakes. APPLES can produce missed and late alerts in locations that experience shaking intensities close to the level used to issue alerts, so preferred alerting strategies with APPLES would use alert thresholds that are lower than the intensities targeted for EEW alerts. We find alerts using APPLES are also similar to those for the source-based approaches currently used in the ShakeAlert EEW system, which will make APPLES easier to integrate into the system.

California, Oregon, Washington↗
USGS2023

Quantifying site effects and their influence on earthquake source parameter estimations using a dense array in Oklahoma

We investigate the effects of site response on source parameter estimates using earthquakes recorded by the LArge-n Seismic Survey in Oklahoma (LASSO). While it is well known that near-surface unconsolidated sediments can cause an apparent breakdown of earthquake self-similarity, the influence of laterally varying site conditions remains unclear. We analyze site conditions across the 1825-station array on a river plain within an area of 40 km by 23 km using vertical ground motions from 14 regional earthquakes. While the source radiation pattern controls P-wave ground motions below 8 Hz, the surface geology correlates with P-wave ground motions above 8 Hz and S-wave ground motions at 2–21 Hz. Stations installed in alluvial sediments have vertical ground motions that can exceed three times the array median. We use the variation of ground motion of regional earthquakes across the array as a proxy for site effects. The corner frequencies and stress drops of local earthquakes ( M L = 0.01–3) estimated using a standard single-spectra approach show negative correlations with the site-effect proxy, while the seismic moments show positive correlations. In contrast, the spectral-ratio approach effectively shows no correlation. The overall bias is small as expected for this relatively homogeneous structure; accurate estimation of site-related biases requires at least 30 stations. Correcting for site-related biases reduces the standard deviations of the source parameters by less than 13% of the total variations. Remaining variations are partially associated with source directivity and model misfits— as small earthquakes can have complex ruptures.

Oklahoma↗
USGS2023

A detailed view of the 2020-2023 southwestern Puerto Rico seismic sequence with deep learning

The 2020–2023 southwestern Puerto Rico seismic sequence, still ongoing in 2023, is remarkable for its multiple‐fault rupture complexity and elevated aftershock productivity. We applied an automatic workflow to continuous data from 43 seismic stations in Puerto Rico to build an enhanced earthquake catalog with ∼180,000 events for the 3+ yr sequence from 28 December 2019 to 1 January 2023. This workflow contained the EQTransformer (EQT) deep learning model for event detection and phase picking, the EikoNet‐Hypocenter Inversion with Stein Variational Inference probabilistic earthquake location approach with a neural network trained to solve the eikonal wave equation, and relocation with event‐pair waveform cross correlation. EQT increased the number of catalog events in the sequence by about seven times, though its performance was not quite as good as thorough analyst review. The enhanced catalog revealed new structural details of the sequence space–time evolution, including sudden changes in activity, on a complex system of many small normal and strike‐slip faults. This sequence started on 28 December 2019 with an M 4.7 strike‐slip earthquake followed by 10 days of shallow strike‐slip foreshocks, including several M 5+ earthquakes, in a compact region. The oblique normal fault M w "> M w 6.4 mainshock then happened on 7 January 2020. Early aftershocks in January 2020, with several M 5+ earthquakes, quickly expanded into two intersecting fault zones with diffuse seismicity: one extending ∼35 km on a northward‐dipping normal fault and the other ∼60‐km‐long and oriented west‐northwest–east‐southeast on strike‐slip faults. Months to years later, aftershocks moved westward, deeper, and to outer reaches of the active fault zones, with abrupt rapid seismicity migration following larger M 4.7+ aftershocks in May, July, and December 2020. The observed seismicity evolution indicates cascading failure from stress transfer on multiple critically stressed faults. High aftershock productivity results from the complex multiple‐fault network hosting the sequence, which is characteristic of an immature fault system in the diffuse deformation zone around Puerto Rico, at the complicated North American–Caribbean plate boundary region.

Puerto Rico↗
USGS2023

Fault roughness at seismogenic depths and links to earthquake behavior

Fault geometry affects the initiation, propagation, and cessation of earthquake rupture, as well as, potentially, the statistical behavior of earthquake sequences. We analyze 18,250 (−0.27 < M < 4.4) earthquakes of the 2016–2019 Cahuilla, California, swarm and, for the first time, use these high‐resolution earthquake locations to map, in detail, the roughness across an active fault surface at depth. We find that the strike‐slip fault is 50% rougher in the slip‐perpendicular direction than parallel to slip. 3D mapping of fault roughness at seismogenic depths suggests that roughness varies by a factor of 8 for length scales of 1 km. We observe that the largest earthquake ( M 4.4) occurred where there is significant fault complexity and the highest measured roughness. We also find that b ‐values are weakly positively correlated with fault roughness. Following the largest earthquake, we observe a distinct population of earthquakes with comparatively low b ‐values occurring in an area of high roughness within the rupture area of the M 4.4 earthquake. Finally, we measure roughness at multiple scales and find that the fault is self‐affine with a Hurst exponent of 0.52, consistent with a Brownian surface.

California↗
USGS2022

The 2020 Westmorland, California earthquake swarm as aftershocks of a slow slip event sustained by fluid flow

Swarms are bursts of earthquakes without an obvious mainshock. Some have been observed to be associated with transient aseismic fault slip, while others are thought to be related to fluids. However, the association is rarely quantitative due to insufficient data quality. We use high-quality GPS/GNSS, InSAR, and relocated seismicity to study a swarm of >2,000 earthquakes which occurred between 30 September and 6 October 2020, near Westmorland, California. Using 5 min sampled Global Positioning System (GPS) supplemented with InSAR, we document a spontaneous shallow M w 5.2 slow slip event that preceded the swarm by 2–15 hr. The earthquakes in the early phase were predominantly non-interacting and driven primarily by the slow slip event resulting in a nonlinear expansion. A stress-driven model based on the rate-and-state friction successfully explains the overall spatial and temporal evolution of earthquakes, including the time lag between the onset of the slow slip event and the swarm. Later, a distinct back front and a square root of time expansion of clustered seismicity on en-echelon fault structures suggest that fluids helped sustain the swarm. Static stress triggering analysis using Coulomb stress and statistics of interevent times suggest that 45%–65% of seismicity was driven by the slow slip event, 10%–35% by inter-earthquake interactions, and 10%–30% by fluids. Our model also provides constraints on the friction parameter and the pore pressure and suggests that this swarm behaved like an aftershock sequence but with the mainshock replaced by the slow slip event.

California↗
USGS2022

Stress heterogeneity as a driver of aseismic slip during the 2011 Prague, Oklahoma aftershock sequence

The interaction of aseismic and seismic slip before and after an earthquake is fundamental for both earthquake nucleation and postseismic stress relaxation. However, it can be difficult to determine where and when aseismic slip occurs within the seismogenic zone because geodetic techniques are limited to detecting moderate to large slip amplitudes or long duration small slip amplitudes. Here, we use repeating earthquakes (earthquakes that re-rupture the same fault patch) as a proxy for aseismic slip during the 2011 Prague, Oklahoma earthquake sequence. We find that aseismic slip in the Prague earthquake sequence occurs both within the granitic basement and the overlying sedimentary rocks. The repeating earthquakes show that patches of aseismic slip are mostly located at fault intersections. These fault intersections hosted possible mainshock slip, abundant aftershocks, and afterslip. We estimate that ∼40% of the aftershocks are driven by afterslip. We interpret that aseismic slip occurs at fault intersections where stress heterogeneity creates patches of lower stress that are stable within a nonsteady state, rate-state framework.

Oklahoma↗
USGS2022

Real-time earthquake detection and alerting behavior of PLUM ground-motion-based early warning in the United States

We examine the real‐time earthquake detection and alerting behavior of the Propagation of Local Undamped Motion (PLUM) earthquake early warning (EEW) algorithm and compare PLUM’s performance with the real‐time performance of the current source‐characterization‐based ShakeAlert System. In the United States (U.S.), PLUM uses a two‐station approach to detect earthquakes. Once a detection is confirmed, observed modified Mercalli intensity (MMI) distributions are forecast onto a regular grid, in which the preferred alert regions are grid cells with MMI 4.0+ forecasts. Although locations of dense station coverage allow PLUM to detect small ( M < 4.5) earthquakes typically not considered for EEW in the U.S., a PLUM detection on a small earthquake does not always generate an alert. This is because PLUM alerts are determined by current shaking distributions. If the MMI 4.0+ shaking subsides prior to detection confirmation by shaking at a second neighboring station, the prior MMI 4.0+ information will not be in the alert forecasts. Of the 432 M 3.0+ U.S. West Coast earthquakes in 2021, 33 produced ground motions large enough to be detected by PLUM. Twenty‐four generated MMI 4.0+ PLUM alerts, whereas ShakeAlert issued public EEW alerts for 13 of these earthquakes. We compare PLUM and ShakeAlert alert regions with ShakeMap and “Did You Feel It?” intensity distributions. Because PLUM alert regions surround stations observed to have strong ground motions (regardless of earthquake magnitude), PLUM alerts reliably include locations that experience significant shaking. This is not necessarily the case for ShakeAlert alert regions when there are large errors in magnitude or epicenter estimates. For two of the largest earthquakes in our real‐time dataset, the M 6.0 Antelope Valley and M 5.1 Petrolia earthquakes, the inclusion of PLUM would have improved real‐time ShakeAlert performance. Our results indicate that incorporation of PLUM into ShakeAlert will improve the robustness of the EEW system.

California, Oregon, Washington↗
USGS2022

What to expect when you are expecting earthquake early warning

We present a strategy for earthquake early warning (EEW) alerting that focuses on providing users with a target level of performance for their shaking level of interest (for example, ensuring that users receive warnings for at least 95 per cent of the occurrences of that shaking level). We explore the factors that can affect the accuracy of EEW shaking forecasts including site conditions (which can make shaking more or less varied), the share of earthquake hazard that comes from outlier events, and even the metric of shaking being forecast (with modified Mercalli intensity being the most uncertain quantity to forecast). To provide different locations with the same level of EEW performance, the alert issuance criteria will need to change based on these factors. Finally, we present a hierarchy of information approach with which to communicate EEW system design and performance to the public and other stakeholders, including how frequently alerts will be issued and how accurate those alerts will be.

Geophysical Journal International↗
USGS2022

Very low frequency earthquakes in between the seismogenic and tremor zones in Cascadia?

Megathrust earthquakes and their associated tsunamis cause some of the worst natural disasters. In addition to earthquakes, a wide range of slip behaviors are present at subduction zones, including slow earthquakes that span multiple orders of spatial and temporal scales. Understanding these events may shed light on the stress or strength conditions of the megathrust fault. Out of all types of slow earthquakes, very low frequency earthquakes (VLFEs) are most enigmatic because they are difficult to detect reliably, and the physical nature of VLFEs are poorly understood. Here we show three VLFEs in Cascadia that were dynamically triggered by a 2009 Mw 6.9 Canal de Ballenas earthquake in the Gulf of California. The VLFEs likely locate in between the seismogenic zone and the Cascadia episodic tremor and slip (ETS) zone, including one event with a moment magnitude of 5.7. This is the largest VLFE reported to date, causing clear geodetic signals. Our results show that the Cascadia megathrust fault might slip rapidly at some spots in this gap zone, and such a permissible slip behavior has direct seismic hazard implications for coastal communities and perhaps further inland. Further, the observed seismic sources may represent a new class of slip events, whose characteristics do not fit current understandings of slow or regular earthquakes.

Oregon, Washington↗
USGS2022

Quantifying the sensitivity of microearthquake slip inversions to station distribution using a dense nodal array

To investigate the sensitivity of slip inversions to station distribution and choice of empirical Green’s function (EGF), we examine three microearthquakes that occurred within the high‐density LArge‐n Seismic Survey in Oklahoma (LASSO) nodal seismic array. The LASSO array’s dense distribution of 1825 geophones provides an exceptional level of spatial and azimuthal coverage, allowing for more accurate inversions of slip than are possible with typical station distributions. The highly accurate slip inversions, in turn, allow for the exploration of the sensitivity of slip inversions to station distribution and parameter choices. We examine the effects of these choices using three well‐recorded strike‐slip microearthquakes ( M L 1.7, 2.3, and 2.7) using an EGF method. From this analysis and the systematic testing of varied network arrangements, we find that station distributions that have uniform coverage of azimuth and distance can retrieve the overall pattern of slip, but the estimated amplitude of slip can vary by 30% for high‐slip regions due to small variations in station location. In addition, we find that the distance range that accurately resolves the overall pattern of slip is the one that contains the takeoff angles of 45°–65°. Concerning azimuthal coverage, a network with >270° performs similarly to having complete coverage. The choice of EGF can shift the location of resolved areas of slip and their amplitude, depending on its similarity in location and radiation pattern to the target earthquake.

Bulletin of the Seismological Society of America↗
USGS2022

Fast rupture of the 2009 Mw 6.9 Canal de Ballenas earthquake in the Gulf of California dynamically triggers seismicity in California

In the Gulf of California, Mexico, the relative motion across the North America-Pacific boundary is accommodated by a series of marine transform faults and spreading centers. About 40 M>6 earthquakes have occurred in the region since 1960. On 3 August 2009, an Mw 6.9 earthquake occurred near Canal de Ballenas in the region. The earthquake was a strike-slip event with a shallow hypocenter that is likely close to the seafloor. In contrast to an adjacent M7 earthquake, this earthquake triggered a ground-motion-based earthquake early warning algorithm being tested in southern California (∼600 km away). This observation suggests that the abnormally large ground motions and dynamic strains observed for this earthquake relate to its rupture properties. To investigate this possibility, we image the rupture process and resolve the slip distribution of the event using a P-wave back-projection approach and a teleseismic, finite-fault inversion method. Results from these two independent analyses indicate a relatively simple, unilateral rupture propagation directed along-strike in the northward direction. However, the average rupture speed is estimated around 4 km/s, suggesting a possible supershear rupture. The supershear speed is also supported by a Rayleigh wave Mach cone analysis, although uncertainties in local velocity structure preclude a definitive conclusion. The Canal de Ballenas earthquake dynamically triggered seismicity at multiple sites in California, with triggering response characteristics varying from location-to-location. For instance, some of the triggered earthquakes in California occurred up to 24 hours later, suggesting that nonlinear triggering mechanisms likely have modulated their occurrence.

Baja California↗
USGS2022

Alert optimization of the PLUM earthquake early warning algorithm for the western United States

We determine an optimal alerting configuration for the propagation of local undamped motion (PLUM) earthquake early warning (EEW) algorithm for use by the U.S. ShakeAlert system covering California, Oregon, and Washington. All EEW systems should balance the primary goal of providing timely alerts for impactful or potentially damaging shaking while limiting alerts for shaking that is too low to be of concern (precautionary alerts). The PLUM EEW algorithm forward predicts observed ground motions to nearby sites within a defined radius without accounting for attenuation, avoiding the earthquake source parameter estimation step of most EEW algorithms. PLUM was originally developed in Japan where the alert regions and ground motions for which alerts are issued differ from those implemented by ShakeAlert. We compare predicted ground motions from PLUM to ShakeMap‐reported ground motions for a set of 22 U.S. West Coast earthquakes of magnitude 4.4–7.2 and evaluate available warning times. We examine a range of prediction radii (20–100 km), thresholds used to issue an alert (alert threshold), and levels of impactful or potentially damaging shaking (target threshold). We find optimal performance when the alert threshold is close to the target threshold, although higher target ground motions benefit from somewhat lower alert thresholds to ensure timely alerts. We also find that performance, measured as the cost reduction that a user can achieve, depends on the user’s tolerance for precautionary alerts. Users with a low target threshold and high tolerance for precautionary alerts achieve optimal performance when larger prediction radii (60–100 km) are used. In contrast, users with high target thresholds and low tolerance for precautionary alerts achieve better performance for smaller prediction radii (30–60 km). Therefore, setting the PLUM prediction radius to 60 km balances the needs of many users and provides warning times of up to ∼20 s.

California, Oregon, Washington↗
USGS2022

A unified perspective of seismicity and fault coupling along the San Andreas Fault

The San Andreas Fault (SAF) showcases the breadth of possible earthquake sizes and occurrence behavior; in particular, the central SAF is a microcosm of such diversity. This section also exhibits the spectrum of fault coupling from locked to creeping. Here, we show that the observations of aseismic slip, temporal clustering of seismicity, and spatial variations in earthquake size distributions are tightly connected. Specifically, the creep rate along the central SAF is shown to be directly proportional to the fraction of nonclustered earthquakes for the period 1984–2020. This relationship provides a unified perspective of earthquake phenomenology along the SAF, where lower coupling manifests in weaker temporal clustering, with repeating earthquakes as an end-member. This new paradigm provides additional justification for characterizing the northwest ∼75 kilometers of the creeping segment as a transition zone, with potential implications for seismic hazard.

California↗
USGS2021

The PLUM earthquake early warning algorithm: A retrospective case study of West Coast, USA, data

The PLUM (Propagation of Local Undamped Motion) earthquake early warning (EEW) algorithm differs from typical source-based EEW algorithms as it predicts shaking directly from observed shaking without first deriving earthquake source information (e.g., magnitude and epicenter). Here, we determine optimal PLUM event detection thresholds for U.S. West Coast earthquakes using two data sets: 558 M3.5+ earthquakes (California, Oregon, Washington; 2012–2017) and the ShakeAlert test suite of historic and problematic signals (1999–2015). PLUM computes Modified Mercalli Intensity ( I MMI ) using velocity and acceleration data, leveraging co-located sensors to avoid problematic signals. An event detection is issued when the observed I MMI exceeds a given threshold(s). We find a two-station detection method using I MMI trigger thresholds of 4.0 and 3.0 for the first and second stations, respectively, is optimal for detecting M4.5+ earthquakes. PLUM detected 79 events in the 2012–2017 data set, reporting (not including telemetry or alert dissemination) detection times on par, and sometimes faster than current EEW methods (mean 8 s; median 6 s). As expected, detection times were slower for the older 1999–2015 earthquakes ( N = 21; mean 11 s; median 6 s) when station coverage was sparser. Of the 31 PLUM detected M5+ events (10 2012–2017; 21 1999–2015), theoretically 20 (∼65%) could provide timely warnings. PLUM issued no false detections and avoided issuing detections for all calibration/anomalous signals, regional and teleseismic events. We conclude PLUM can successfully identify I MMI 4+ shaking from local earthquakes and could complement and enhance EEW in the U.S.

British Columbia, California, Oregon, Washington↗
USGS2021

Robust earthquake early warning at a fraction of the cost: ASTUTI Costa Rica

We show that a fixed smartphone network can provide robust Earthquake Early Warning for at least two orders of magnitude less cost than scientific-grade networks. Our software and cloud-based data architecture that we have constructed for the Alerta Sismica Temprana Utilizando Teléfonos Inteligentes (ASTUTI; Earthquake Early Warning Utilizing Smartphones) network in Costa Rica is easily scaled and exported. Implementation comprises provisioning and installing modern smartphones in judicious locations. Stand-up time for regionally operational networks can be on the order of days. We evaluated a non-parametric ground-motion detection and alerting strategy that would alert the entire Costa Rican population of any event with a ground motion detection threshold of 0.55–0.65 %g at four neighboring stations. During a 6-month evaluation period ASTUTI detected and alerted on five of 13 earthquakes with M w 4.8–5.3 that caused felt Modified Mercalli Intensity shaking levels of 4.3–6. The system did not produce any false alerts and the undetected events did not produce wide-spread or significant felt shaking. System latencies were less than or similar to scientific-grade latencies. Alerts for all five detected events would have reached the capital city, San Jose, before strong S -wave shaking. This would have afforded time for Drop Cover Hold On actions by most residents. Two of the five alerts were triggered by P -waves suggesting that smartphone-based networks could approach the fastest theoretical EEW performance, especially with future expected improvements in smartphone sensors and processing algorithms.

AGU Advances↗
USGS2021

Wastewater disposal has not significantly altered the regional stress state in southern Kansas

Wastewater disposal is primarily responsible for the increased seismicity rate since ~2013 in southern Kansas. Previous work that used shear wave splitting (SWS) in southern Kansas interpreted a ~90º temporal rotation in the fast polarization direction and attributed it to increased pore pressures resulting from fluid injection. However, this interpreted rotation coincided with a change in the stations used to make the SWS measurements. We investigate the temporal variability of fast azimuths in southern Kansas by making SWS measurements on earthquake families with similar source-receiver paths recorded on a stable local seismic network. We select high-quality SWS measurements by investigating the stability of results across 65 different frequency bands between 0.5-15 Hz. We find that the fast polarization direction in southern Kansas is relatively constant with an average ENE (~N79ºE) orientation between 2014-2017. Our fast polarization measurements are primarily a reflection of the maximum principal horizontal stress direction (SHmax). We observe a slight spatial change in SHmax to the NE (~N55°E) near the Nemaha Ridge in Oklahoma. However, we do not observe any significant temporal rotation of SHmax or variation in delay time (i.e., crack density) in southern Kansas, contrary to the earlier study. The previously interpreted ~90º rotation may either be a reflection of a very local stress change or a misinterpretation of SWS results potentially due to the use of inconsistent source-receiver paths. Our SWS measurements cover the period of peak wastewater disposal and seismicity rates and suggest an absence of significant temporal rotations in the local anisotropy and stress orientations associated with wastewater disposal.

Kansas↗