Geology Reports⌕ Search

SEARCH · Geology Reports

Results for “Map”

Search indexed USGS publications on groundwater, aquifers, geologic maps, mineral resources and earthquakes. Explore source records by subject and place.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

452 records · Page 11Linked to original sources
USGS2024

Testing megathrust rupture models using tsunami deposits

The 26 January 1700 CE Cascadia subduction zone earthquake ruptured much of the plate boundary and generated a tsunami that deposited sand in coastal marshes from northern California to Vancouver Island. Although the depositional record of tsunami inundation is extensive in some of these marshes, few sites have been investigated in enough detail to map the inland extent of sand deposition and depict variability in tsunami deposit thickness and grain size. We collected 129 cores in marshes of the Salmon River estuary in Oregon and reanalyzed 114 core logs from a 1987–88 study that mapped the inland extent of circa 1700 CE sandy tsunami deposits. The ca. 1700 CE tsunami deposit in the Salmon River estuary is easily recognized in cores ≤1 m deep in which a buried marsh peat is overlain by a well sorted sand bed with a sharp lower contact that thins and fines inland. We use tsunami deposit data and models of sandy tsunami sediment transport (using Delft3D-FLOW) to test 15 rupture models that could represent a ca. 1700 CE earthquake. At least 12–16 m of slip offshore of the Salmon River, which results in 0.8–1.0 m of coastal coseismic subsidence, is required to match the ca. 1700 CE sand deposit's inland extent, which is consistent with models of heterogeneous megathrust slip in ca. 1700 CE. Our methods of detailed tsunami deposit mapping, combined with sediment transport modeling, can be used to test models of megathrust ruptures and their tsunamis to potentially improve earthquake and tsunami hazard assessments.

JGR Earth Surface↗
USGS2026

USGS Geochron Database

Introduction Geochronology helps us understand Earth’s history by determining when important events, like volcanic eruptions, the rise of mountains, the formation of mineral resources, and changes in the landscape, happened. Geochronological data directly support geologic mapping and can inform decisions about geologic hazard mitigation, natural resource management, and infrastructure resilience. The U.S. Geological Survey (USGS) Geochron database provides access to more than 300,000 published, publicly available age measurements from more than 40,000 geological samples. This database is the result of a collaborative effort with State geological surveys and geoscientists from across the globe. The USGS Geochron database is the most comprehensive collection of geochronological data available for the United States. Users can view data through an interactive map explorer, download datasets, and integrate data into geospatial software or other analysis tools.

Fact Sheet↗
USGS2024

Improving crop-specific groundwater use estimation in the Mississippi Alluvial Plain: Implications for integrated remote sensing and machine learning approaches in data-scarce regions

Study region The Mississippi Alluvial Plain (MAP) in the United States (US). Study focus Understanding local-scale groundwater use, a critical component of the water budget, is necessary for implementing sustainable water management practices. The MAP is one of the most productive agricultural regions in the US and extracts more than 11 km 3 /year for irrigation activities. Consequently, groundwater-level declines in the MAP region pose a substantial challenge to water sustainability, and hence, we need reliable groundwater pumping monitoring solutions to manage this resource appropriately. New hydrological insights for the region We incorporate remote sensing datasets and machine learning to improve an existing lookup table-based model of groundwater use previously developed by the U.S. Geological Survey (USGS). Here, we employ Distributed Random Forests, an ensemble machine learning algorithm to predict annual and monthly groundwater use (2014–2020) throughout this region at 1-km resolution, using pumping data from existing flowmeters in the Mississippi Delta. Our model compares favorably with the existing USGS model, with higher R 2 (0.51 compared to 0.42 in the previous model), and lower root mean square error (RMSE) and mean absolute error (MAE)— 0.14 m and 0.09 m, respectively in our model, compared to 0.15 m and 0.1 m in the previous model. Therefore, this work advances our ability to predict groundwater use in regions with scarce or limited in-situ groundwater withdrawal data availability.

Journal of Hydrology Regional Studies↗
USGS2026

Surface variable‐based machine learning for scalable arsenic prediction in undersampled areas

In the United States, private wells are not federally regulated, and many households do not test for Arsenic (As). Chronic exposure is linked with multiple health outcomes, and risk can change sharply over short distances and with well depth. Coarse maps or sparse sampling often miss exceedances. Most existing models operate at ∼1 km resolution and use groundwater chemistry or detailed geologic logs, which limits their use in undersampled areas where improved guidance is most needed. We overcome these limitations by developing a machine learning model for Minnesota, USA, that predicts As exposure risk using only surficial variables from remote sensing and global data sets. Variables related to surface water hydrology and geomorphology are selected based on mechanistic links that control redox conditions and As mobilization. Local training was essential, and surficial geology variables that are more sensitive to local conditions were needed to maximize model accuracy. The resulting complete model was sufficiently sensitive to generate accurate and detailed risk maps and depth profiles of As concentrations above the 10 μg/L maximum contaminant level. Accuracy depended on local training data density. We identified a training data density of 0.07 wells/km 2 as a practical target for stable county-level performance. Maps of exceedance probabilities highlight priority areas for testing that are particularly important in rural communities that have received less sampling. These results support public health action by guiding where to install wells and where to test them, how much new sampling is needed, and where treatment outreach is most urgent.

Minnesota↗
USGS2024

Earthquake-triggered ground-failure inventory associated with the M7.1 2018 Southcentral Alaska earthquake

The 30 November 2018, magnitude (Mw) 7.1 earthquake in Southcentral Alaska triggered substantial landslides, liquefaction, and ground cracking throughout the region, resulting in widespread geotechnical damage to buildings and infrastructure. Despite a challenging reconnaissance and remote-sensing environment, we constructed a detailed digital inventory of ground failure associated with the event from several sources. Sources included information derived from remotely sensed data, and data compiled from literature, social media postings, and earthquake damage information compiled by local, state, and federal agencies. Each instance of ground failure within the inventory contains information on the location and type of observed ground failure, and the methods and data used to document the occurrence. Where high-quality data, such as LIDAR or satellite imagery, were available and showed the ground-failure instance clearly, the extent is mapped as a polygon or polyline. All other locations are mapped as points. There are a total of 886 ground-failure instances documented within the inventory (400 landslides, 286 liquefaction features, and 200 features unattributed to specific processes). A semi-quantitative confidence scheme is used to describe mapping certainty associated with each ground-failure feature. This inventory represents a relatively moderate ground-failure-triggering event that occurred in a subarctic environment. This data paper describes the content within the inventory, the inventory data collection procedures, and limitations of the data. Events of this type are not often documented in detail; thus, adding the inventory data to the US Geological Survey Open Repository of Earthquake-Triggered Ground-Failure Inventories further diversifies the datasets available to the scientific community to be used to better understand and model earthquake-triggered ground failure.

Alaska↗
USGS2026

High-resolution magnetic survey using an unoccupied aerial vehicle to constrain buried lava flow geometry, volume, and eruptive history of Little Cones, Crater Flat, Nevada

Magnetic surveys are an important tool used to augment geologic mapping in distributed volcanic fields. Using magnetic anomalies, it is possible to model the geometry of shallowly buried volcanic features, such as conduits, sills, and lava flows. This subsurface mapping is important for understanding eruption dynamics and emplacement of lava flows, and it sometimes reveals buried volcanoes no longer visible at the surface. These data are critical to better interpret the numbers, styles, and magnitudes of eruptions in distributed volcanic fields and their associated volcanic hazards. New advances in unoccupied aerial vehicles (UAVs) offer an attractive middle range of resolution and aerial coverage between ground-based magnetic surveys and aeromagnetic surveys. Here, we present the results of a UAV fluxgate magnetic survey of the Little Cones, Nevada, scoria cones, which have been the target of previous ground and aeromagnetic surveys. The magnetic anomalies at Little Cones are of interest because the surrounding alluvium conceals lava flows that erupted from Little Cones, making it very difficult to understand the volume and morphology of lava flows from geologic mapping alone. Nonlinear inversion of UAV-collected magnetic data were used to model the thickness and morphology of buried Little Cones’ lava flows with higher precision than achieved previously. The sequence of events and calculated flow characteristics are then interpreted. The total volume of Little Cones, including concealed lava flows, is approximately 0.016 cubic kilometer, and the initial sheet flow erupted in less than 24 hours. The findings presented herein demonstrate that UAV-based magnetic surveys are a reliable method of data collection and an efficient alternative to other survey methods, facilitating development of a three-dimensional perspective of distributed volcanic fields.

Nevada↗
USGS2026

Abandonment of the Upper Devonian Greenland Gap Group and Scherr Formation and revision of the Upper Devonian Brallier and Foreknobs Formations in the central Valley and Ridge Province

This study revises the lithostratigraphic framework of the Upper Devonian interval traditionally assigned to the Greenland Gap Group across the central Appalachian Valley and Ridge Province. The work aims to modernize and standardize lithostratigraphic nomenclature, establish a new reference section and demonstrate how the revised stratigraphy improves edge-matching of 1:24,000 scale geologic maps and supports compilation mapping at scales of 1:100,000 and larger. The revision eliminates the names Greenland Gap Group, Scherr Formation and the Minnehaha Springs Member of the Scherr Formation; reassigns all strata previously designated as Scherr Formation by Dennison (1970) to the upper Brallier Formation; and abandons the basal Mallow Member of the Foreknobs Formation, placing its strata within the upper Brallier as originally defined by Butts (1918). The contact between the Brallier and Foreknobs formations is placed at the base of the first mappable, ridge-forming package of fine- to coarse-grained, cross-bedded, sandstone beds, often containing rounded quartz pebbles with minor interbeds of shale and siltstone. This contact may be gradational in places but, even in absence of good exposure, can usually be distinguished topographically in recently produced lidar-derived imagery as having elevated relief due to the presence of more resistant, compositionally mature coarse-grained sandstone-rich strata. Applying this criterion for mapping the contact between the Brallier and Foreknobs formations has resulted in reconciliation of mismatches of geologic contacts along several 7.5-minute quadrangle boundaries in the states of Virginia, West Virginia, Maryland and Pennsylvania. A new reference section at Baker, West Virginia showcases the contacts between the Harrell Shale, Brallier Formation, Foreknobs Formation and Hampshire Formation. A digital outcrop model of the reference section is provided for future preservation.

Maryland, Pennsylvania, Virginia, West Virginia↗
USGS2025

Using public participatory geographic information systems (PPGIS) to explore uses and values for Mojave Trails National Monument, California

Many people ascribe a variety of values to public lands and waters, but some values are more difficult to assess and quantify than others. Public participatory geographic information systems (PPGIS) are tools that have been used to help quantify and map the public’s diverse values for a landscape. This work describes the first known Office of Management and Budget–approved use of PPGIS by a Department of the Interior bureau. The U.S. Geological Survey developed an internet-based application to aid in gathering PPGIS data, called Values Mapping for Planning in Regional Ecosystems (VaMPIRE). Further, this work describes the first pilot of the VaMPIRE application in coordination with the Bureau of Land Management to collect spatial data and other survey data regarding the public’s uses of and values for locations within Mojave Trails National Monument. We emailed the link to the VaMPIRE application to an interested party email list in 2024 with 207 valid emails and received 74 responses; we also received 47 responses from members of an off-roading social media group. Of the list of 16 value options, recreation was the most popular value for the monument, followed by wilderness and inspirational. Over 1,000 points were placed throughout the monument, indicating locations people use or value, with the locations spread throughout the entire monument. Additionally, most survey respondents stated their ability to receive benefits in locations they mapped would not change in response to a hypothetical scenario related to recreational facility development. This report describes exploratory results from the first use of the VaMPIRE tool in Mojave Trails National Monument and includes reflections on how the process went and considerations for future use of VaMPIRE.

California↗
USGS2026

Generating geochemical and mineralogy distributions of soil in the conterminous United States using Bayesian hierarchical spatial models

Characterizing geochemical and mineralogical soil distributions across large spatial extents is essential for understanding mineral resources, ecosystem processes, and environmental risks. Rasters of soil geochemical distributions for the conterminous United States, however, are limited. We present a Bayesian modeling workflow and tool for generating predictive geochemical and mineralogy distribution maps for the conterminous United States using integrated nested Laplace approximation (INLA) with the stochastic partial differential equation approach. By modeling soil geostatistical data with environmental covariates (soil properties, topography, climate, and land cover), we generate predictive distributions of soil geochemistry that can be mapped or extracted for further analyses. As an example, we model the spatial distribution of trace elements in soil relevant to vertebrate health (cobalt, copper, iron, manganese, selenium, and zinc) and provide a workflow that can be used to generate and visualize predictive distributions of 39 other major and trace elements and 21 minerals of the soil survey, supporting a variety of ecological, environmental, and agricultural applications.

MethodsX↗
USGS2024

Recent advances in characterizing the crustal stress field and future applications of stress data: Perspectives from North America

The stress field controls patterns of crustal deformation, including which faults are likeliest to cause earthquakes or transmit fluids. Since the 1950s, maps of maximum horizontal stress ( S Hmax ) orientations have advanced dramatically, and the style of faulting (relative principal stress magnitudes) has recently been mapped in some regions as well. This perspectives paper summarizes developments in characterizing stress orientations and (relative) magnitudes, including new seismic and borehole methods, as well as progress in identifying the causes of stress variations. Despite these advances, adding far more spatiotemporal detail would allow geoscientists to address many of today's key challenges regarding natural hazards, energy development, and geodynamics. In particular, it is critically important to characterize stress heterogeneity at multiple scales while also recognizing the coherent variability of the stress field. The second part of the paper considers how more detailed stress datasets could prove essential to addressing some of the grand questions in geoscience, including deciphering the poorly understood feedbacks between crustal dynamics and surface processes, improving earthquake and eruption forecasts, and determining the origins and shared properties of plate boundaries.

Geological Society, London, Special Publications↗
USGS2025

Revisiting an enigma on California's north coast: The Mw6.5 Fickle Hill earthquake of 21 December 1954

Many earthquakes occur along the North Coast of California in the vicinity of the Mendocino Triple Junction (MTJ), where the Pacific, Gorda, and North American (NA) plates meet, and on the adjacent plate boundaries. The MTJ marks the nexus of the Mendocino and San Andreas faults with the Cascadia subduction zone (CSZ). Historically, most large earthquakes around the MTJ have been within the offshore Gorda plate and its subducted portion beneath the NA plate. North of the MTJ, active faults mapped in the NA plate are part of the CSZ fold‐and‐thrust belt. Although some events have been detected in the NA plate, no large historic events have been associated with mapped surface faults. The 21 December 1954 M w 6.5 earthquake in Humboldt County is one possible exception. Using published data from catalogs and articles, unpublished data from Berkeley’s archives, and S ‐ P times interpreted from two U.S. Coast and Geodetic Survey (USCGS) accelerometers, we determine a probability cloud for the earthquake’s hypocenter using NonLinLoc. The highest probability location lies beneath Fickle Hill just east of the city of Arcata, California, at 40.87° N, 124.03° W, and ∼11 km depth. Using P ‐wave polarities from Berkeley stations and the digitized waveforms from the accelerometers, we find that the focal mechanism most consistent with the data indicates thrust movement with strike, dip, and rake of 350°, 10°, and 90°, respectively, at a depth of 14 km. Given the depth uncertainties of both this event and the megathrust, this implies that the earthquake most likely took place on the subduction interface rather than on the mapped faults in the Mad River fault zone that trend 322° and dip to the northeast. The revisited intensity in the epicentral region also supports a location beneath Fickle Hill to the east of the city of Arcata, California.

California, Idaho, Nevada, Oregon, Washington↗
USGS2025

Preliminary field report of landslide hazards following Hurricane Helene

Executive Summary This report reflects our knowledge regarding the widespread landslide activity associated with Hurricane Helene observed during the U.S. Geological Survey’s (USGS) mission assignment to North Carolina in October 2024. The material in this report was originally prepared for the Federal Emergency Management Agency under mission assignment DR-4827-NC. The data and commentary in this report are reflective of a report provided to the Federal Emergency Management Agency (FEMA) on October 18, 2024, as well as information provided in briefings at the Buncombe County Emergency Operations Center. The report has been modified for public dissemination. This assessment was based on systematic visual examination and mapping of landslide locations from aerial and satellite imagery, visual and photographic observations from low-level helicopter overflights and conversations with local landslide experts from the North Carolina Geological Survey and Appalachian Landslide Consultants PLLC, and more than 50 years of combined landslide hazard professional experience of the mission-assigned field team. No systematic field investigations were done by the USGS. While responding to the event, the USGS did not identify any landslides that posed an immediate major threat to recovery personnel in parts of nine counties in North Carolina (Avery, Buncombe, Henderson, McDowell, Mitchell, Polk, Rutherford, Watauga, and Yancey); however, threats from renewed landslide activity may remain heightened in localized areas for months or even years. Known areas of the most abundant landslide occurrence include Bat Cave, Lake Lure, Chimney Rock, Swannanoa, Black Mountain, Fairview, steep areas in Asheville, and the Blue Ridge Parkway. The USGS shared detailed locations of known landslides with the Emergency Operations Centers. The thousands of landslide scars on hillsides and landslide deposits on flatter ground may present some threat to recovery activities. Soil and rocks will continue to erode from newly exposed landslide scars and may pose a threat to people and infrastructure who are immediately nearby. In general, the steeper and taller the landslide scar, the greater the potential threat. This threat is heightened during periods of rainfall and increases with the duration and intensity of rainstorms. Very heavy rainfall, or repeated rainfall events during short periods, could also initiate new landslides on steep slopes. Excavation of landslide deposits, particularly excavation of those deposits directly adjacent to steep slopes, may also pose a threat to nearby people and equipment. An interagency collaborative mapping effort led by the USGS that informed this assessment identified 1,155 landslide locations by the October 2024 briefings, but that number increased to 2,217 in a final reviewed version of the locations published in January 2025. Locations were mapped from satellite imagery, fixed-wing and helicopter surveys, media and social media, and field reports in the 3 weeks following the passage of the remnants of Hurricane Helene. USGS products outlined in this report are publicly available and include geotagged photographs from aerial reconnaissance, hazard models, an interactive view of mapped landslide locations, and landslide safety and education resources.

North Carolina, South Carolina, Tennessee, Virgini↗
USGS2025

Land cover change within wetland complexes at Dixie Meadows, Churchill County, Nevada: 2015 – 2023

Dixie Meadows, Nevada, is a system of geothermal springs and seeps that feed a complex of marshes and wetland meadows that are located within lands managed by the Bureau of Land Management (BLM) and the Department of Defense (DOD). A previous U.S. Geological Survey report documented variability in satellite imagery-based land cover classifications for seven wetland complexes at near monthly time intervals between October 2015 and January 2022. This report presents additional data, extending analysis to November 2023. Land cover classifications between October 2015 and November 2023 demonstrated an association between vegetation cover characteristics and surface moisture, with Class 1 having dry, bare soil or sparse upland vegetation, Class 2 having moist, bare soil or sparse to small vegetation, Class 3 having dense green vegetation with potentially saturated soil conditions, Class 4 having a mix of shallow surface water, saturated soil, and dense green vegetation, and Class 5 having open surface water. Most of the wetland complexes occur close to spring outflows primarily within land managed by the DOD, though portions are also within BLM lands. The intervening and surrounding landscape outside of the wetland complexes assessed in this study are managed by the BLM. As a result, Class 1 land covers had the largest areal coverage for BLM managed lands. Classes 2 and 3 land covers were primarily mapped inside the wetland complexes and thus had the largest area coverage within DOD managed lands. Class 4 was almost exclusively mapped within the wetland complexes and thus was largely contained within DOD managed lands. Class 5 (open water) was exclusively mapped in and adjacent to a single wetland complex with catchment ponds on land managed by the BLM. The distribution of these land cover classes over the study period was seasonally and annually variable. Land cover areas of Classes 1 and 2 were larger during the spring months. Conversely, land cover areas of Classes 3 and 4 tended to be greatest during the summer or fall. These patterns might be influenced by differences in seasonal water sources and phenology.

Nevada↗
USGS2024

Evaluation of the lithium resource in the Smackover Formation brines of southern Arkansas using machine learning

Global demand for lithium, the primary component of lithium-ion batteries, greatly exceeds known supplies, and this imbalance is expected to increase as the world transitions away from fossil fuel energy sources. High concentrations of lithium in brines have been observed in the Smackover Formation in southern Arkansas (>400 milligrams per liter). We used published and newly collected brine lithium concentration data to train a random forest machine-learning model using geologic, geochemical, and temperature explanatory variables and create a map of predicted lithium concentrations in Smackover Formation brines across southern Arkansas. Using these predicted lithium maps with reservoir parameters and geologic information, we calculated that there are 5.1 to 19 million tons of lithium in Smackover Formation brines in southern Arkansas, which represents 35 to 136% of the current US lithium resource estimate. Based on these calculations, in 2022, 5000 tons of dissolved lithium were brought to the surface within brines as waste streams of the oil, gas, and bromine industries.

Arkansas↗
USGS2025

Exploration for blind geothermal systems in the eastern Great Basin of Utah: An update on the “Lund North” INGENIOUS detailed study site

Existing geothermal production in Utah is commonly collocated with surficial expressions of geothermal heat including active hot springs and hot spring deposits. However, geothermal potential across the Great Basin region is thought to be much higher for hidden or blind geothermal systems. Accordingly, exploration techniques that can locate geothermal resources that lack surface thermal features could support future development of these systems. The goal of the INGENIOUS project is to reduce exploration risk and discover new, economically viable hidden geothermal systems in the Great Basin region. This paper summarizes the efforts and preliminary results of blind geothermal resource prospecting in Utah as part of the larger INGENIOUS project. The Lund North site in the Basin and Range province of southwest Utah was designated as the fourth detailed study site for the project. The site was initially identified from the preliminary Play Fairway Analysis (PFA) geothermal favorability map, which is based on various data elements, and the identification of favorable structural settings across Utah. This site contains existing legacy data including a thermal gradient borehole with a heat-flow value exceeding 220 mW/m 2 . New data collected at this site include terrestrial gravity, magnetotellurics, transient electromagnetics, fluid geochemistry, geologic mapping, UAV-based lidar, paleomagnetism, aeromagnetic surveys, and a shallow temperature survey. Preliminary data and model interpretations corroborate the presence of a large, east-dipping, northeast-southwest-trending normal fault on the edge of a typical Basin and Range graben. Quaternary fault mapping refined by high-resolution lidar surveys suggests a complex step over fault geometry. Utilizing multi-disciplinary datasets for further site assessment at Lund North will facilitate the development of a local-scale PFA and geothermal conceptual model. This new data will be used to site successive thermal gradient drilling to verify heat flow with temperatures at depth at the Lund North site and further investigate the existence of a local geothermal system.

Utah↗
USGS2023

Stratigraphic notes—Volume 1, 2022

This is the first volume in the U.S. Geological Survey (USGS) series of reports on stratigraphy entitled “Stratigraphic Notes,” which consists of short papers that highlight stratigraphic studies, changes in stratigraphic nomenclature, and explanations of stratigraphic names and concepts used on published geologic maps. “Stratigraphic Notes” is a long-term (multiyear), multivolume publication containing articles that address updates or revisions to stratigraphic nomenclature (and whose content ultimately will be incorporated by National Geologic Map Database personnel into Geolex, https://ngmdb.usgs.gov/Geolex/ ). We welcome papers for the “Stratigraphic Notes” series from geoscientists of the USGS, of State Geological Surveys, and from academicians. Papers can be submitted for publication in “Stratigraphic Notes” by contacting the USGS Geologic Names Committee ( gnc@usgs.gov ). As new “Stratigraphic Notes” volumes are published, links to the volumes will be posted at https://doi.org/10.3133/pp1879 . This first volume ("Stratigraphic notes—Volume 1, 2022") includes articles that provide guidance for those who wish to submit papers to “Stratigraphic Notes,” as well as information on how to make your manuscripts compliant for geologic names reviews and how to organize your paper’s content to facilitate inclusion of new or revised names in Geolex. This volume also includes some specific guidance on conducting geologic names reviews of geologic and hydrogeologic reports.

Professional Paper↗
USGS2024

Perspectives on transportable array Alaska background noise levels

Background seismic noise fundamentally sets a lower bound on our ability to record signals arising from earthquakes. The background noise spectrum at a station is a combination of cultural noise, ocean-generated microseism noise, intrinsic instrument self-noise, and the sensitivity of the instrument to nonseismic noise sources. The USArray-Transportable Array Alaska deployed 195 stations across Alaska and parts of Canada (Yukon, British Columbia, and Northwest Territories). These stations were all installed using similar techniques and made use of instruments with similar self-noise levels. As such, this network provides an opportunity to look at how geographic location influences seismic background. Using these broadband stations, we report background noise levels from 0.2 to 75 s period in six discrete bands. By constructing “noise maps,” we depict both spatial and temporal changes in the background noise field. Using these maps, combined with targeted analysis, we infer sources and contributing factors to noise levels in these different period bands. These include cultural noise, the formation of sea ice, seasonal changes in permafrost and wave activity in the Gulf of Alaska, and magnetic field variability. We use this study as an opportunity to review several previous studies examining seismic noise in Arctic regions.

Book chapter↗
USGS2025

Total uncertainty quantification in inverse solutions with deep learning surrogate models

We propose an approximate Bayesian method for quantifying the total uncertainty in inverse partial differential equation (PDE) solutions obtained with machine learning surrogate models, including operator learning models. The proposed method accounts for uncertainty in the observations, PDE, and surrogate models. First, we use the surrogate model to formulate a minimization problem in the reduced space for the maximum a posteriori (MAP) inverse solution. Then, we randomize the MAP objective function and obtain samples of the posterior distribution by minimizing different realizations of the objective function. We test the proposed framework by comparing it with the iterative ensemble smoother and deep ensembling methods for a nonlinear diffusion equation with an unknown space-dependent diffusion coefficient. Among other applications, this equation describes the flow of groundwater in an unconfined aquifer. Depending on the training dataset and ensemble sizes, the proposed method provides similar or more descriptive posteriors of the parameters and states than the iterative ensemble smoother method. Deep ensembling underestimates uncertainty and provides less-informative posteriors than the other two methods. Our results show that, despite inherent uncertainty, surrogate models can be used for parameter and state estimation as an alternative to the inverse methods relying on (more accurate) numerical PDE solvers.

Journal of Computational Physics↗