Search USGS⌕ Search

SEARCH · Search USGS

Results for “Journal of Computational Physics”

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.

1,661 records · Page 3Linked to original sources
USGS2025

Implications of physics-based M9 ground motions on liquefaction-induced damage in the Cascadia Subduction Zone: Looking forward and backward

Given the likelihood of future M9 Cascadia Subduction Zone (CSZ) earthquakes, various estimates of the resulting, regional ground motions have been made, including a suite of 30 physics-based simulations that reflect key modeling uncertainties. However, because the last CSZ interface rupture occurred in 1700 CE, the shaking expected in such an event is especially uncertain, as are the impacts to the built and living environments. Like other coseismic impacts, soil liquefaction poses a significant threat and must be considered by any scenario study used to inform planning and response, or to focus mitigation resources. Liquefaction is also notable for its potential to “ground truth” ground-motion estimates, given that its presence or absence in the geologic record can provide constraint on the intensities of shaking in past events. It is thus an important phenomenon looking both forward and backward. Accordingly, using recent physics-based simulations, this study (1) predicts liquefaction in M9 CSZ ruptures at 400 locations in Oregon, Washington, and British Columbia (BC) using an array of cone-penetration-test based models and (2) uses paleoliquefaction evidence at ten sites spanning from Southern Oregon to Vancouver, BC to constrain possible ground-motion intensities experienced in the 1700 CE earthquake. The forward predictions indicate that liquefaction in M9 events could be pervasive in the region and affect numerous population hubs, with the potential for damage across hundreds of square kilometers. The backward analyses suggest that 1700 CE ground-motion intensities may have been less than expected from M9 simulations in some northern portions of the CSZ (e.g. Seattle), given the paucity of 1700 CE liquefaction evidence in these areas. Ultimately, further discovery and analysis of CSZ paleoliquefaction, or lack thereof, will confirm or modify this possibility and the conclusions drawn herein.

Oregon, Washington↗
USGS2024

Global variability of the composition and temperature at the 410-km discontinuity from receiver function analysis of dense arrays

Seismic boundaries caused by phase transitions between olivine polymorphs in Earth's mantle provide thermal and compositional markers that inform mantle dynamics. Seismic studies of the mantle transition zone often use either global averaging with sparse arrays or regional sampling from a single dense array. The intermediate approach of this study utilizes many densely spaced seismic arrays distributed around the globe. We systematically compute teleseismic P-to-S receiver functions for each seismic array and invert for the 1-D seismic velocity structure of the mantle transition zone beneath each array to facilitate a comparison between densely sampled regions. We stack 3,600 receiver functions on average at 67 arrays in total. The stack is used in a probabilistic inversion to estimate the mantle transition zone interface depths and velocities beneath each array. We focus on the 410-km discontinuity (410) because it is a prominent seismic interface that is clearly linked to a single mineral phase transition between olivine and wadsleyite. The depths and velocity contrasts of the 410 are mapped to temperatures and compositions using mineral physics constraints. The depth of the 410 ranges from ∼405–440 km, which is consistent with a ∼360 K temperature range in a dry mantle and a ∼260 K temperature range in a wet mantle (2 wt. % water). The Vs contrast across the 410 ranges from ∼2.5–8 %, which is consistent with ∼20–70 vol. % olivine composition in a dry mantle and ∼25–80 vol. % in a wet mantle. The bulk composition of the upper mantle near the 410-km discontinuity is typically considered to be well-mixed because there is no thermodynamic impediment to convection at the olivine to wadsleyite phase transition. However, the wide range of inferred olivine content from our study suggests that there are large lateral variations in the bulk composition of the upper mantle near the 410-km discontinuity.

Earth and Planetary Science Letters↗
USGS2024

Stream nitrate dynamics driven primarily by discharge and watershed physical and soil characteristics at intensively monitored sites: Insights from deep learning

We developed a suite of models using deep learning to make hindcast predictions of the 7‐day average backward‐looking nitrate concentration at 46 predominantly agricultural sites across the midwestern and eastern United States. The models used daily observations of discharge and meteorological variables and watershed attributes describing anthropogenic modification to hydrology, nitrogen application, climate, groundwater, land use, watershed physiographic attributes, and soils. Across all sites, discharge and watershed soil and physiographic attributes showed a strong influence on model performance. Analysis of drivers across sites revealed considerable regional differences related to controlling processes such as groundwater contributions. We tested several ways to pool data across sites to develop accurate models and make the most effective use of available data. Single‐site models, in which models are trained and tested at a single location, showed generally strong predictive performance (median Kling‐Gupta Efficiency = 0.66), and accuracy at poorly performing sites could be improved by grouping sites with similar characteristics. Developing a single model for all sites reduced performance at several locations with distinct characteristics, suggesting that there is a threshold of dissimilarity beyond which more data does not improve the model. While many deep learning studies have shown that national or even global models can outperform local models, it is not clear that this is true for water quality constituents. This study demonstrates how data can be combined effectively, using deep learning to develop accurate and interpretable models of instream nitrate at sites where varying processes are responsible for changes in nitrate concentration.

Water Resources Research↗
USGS2026

A roadmap for identifying and interpreting physical processes and national water model prediction bias associated with baseflow index regimes across the contiguous United States

Understanding how groundwater–surface water interactions shape streamflow variability is critical for diagnosing low flow behavior and prediction bias in continental scale hydrologic models. We present a process informed framework that links observed baseflow (BF) dynamics, watershed attributes, and National Water Model (NWM) performance across the contiguous United States. Using daily observed streamflow from 797 reference quality streamgages, we developed monthly baseflow index (BFI) signatures using a streamgage specific, calibrated digital filter. Hierarchical clustering of these signatures identified seven distinct BFI regimes capturing regional and seasonal variability. We evaluated NWM v3.0 retrospective streamflow performance within each regime using multiple hydrograph and flow duration curve-based metrics. Model skill varied systematically across regimes: mixed flow systems were simulated most accurately, while predominantly BF dominated and quickflow dominated regimes exhibited substantially poorer performance. Across nearly all regimes, the NWM underestimated observed BFI magnitude and frequently failed to reproduce seasonal BF patterns, indicating systematic biases in simulated low flow contributions. To relate these regimes to potential process controls, we trained a Random Forest classifier using static watershed attributes and applied Shapley Additive Explanations to identify features most strongly associated with each regime. Results highlight regionally varying influences, including the dominant role of snow fraction and seasonal runoff timing in snow dominated basins and the importance of evapotranspiration and aridity in quickflow dominated systems. Collectively, these findings demonstrate how hydrologic signatures combined with interpretable machine learning can diagnose regime specific model biases and generate process-based hypotheses about limitations in large scale hydrologic prediction systems.

contiguous United States↗
USGS1976

Effect of deicing chemicals on the hydrologic environment in Massachusetts; evaluation of surface-water data collection

The objective of this study is to develop predictive relationships that can be used to describe the impact of highway deicing salts on the hydrologic environment. A method, presented in an earlier report, for estimating yearly mean chloride concentrations from estimated or actual runoff and salt-application data was applied to the period 1971 through1974 and the results compared with yearly mean chloride concentrations computed from records of specific conductance. Estimates were found to be from minus 63 percent to plus 56 percent in error when the error was expressed as the difference between estimated and computed values as a percentage of the computed values. Chloride concentrations used in this study are computed from specific conductance/chloride concentration relationships and records of specific conductance. Preliminary results from graphic analyses and least-squares regression analysis show that relationships between measured values of specific conductance and chloride concentration have correlation coefficients ranging from 0.30 to 0.97. Analysis of streamflow for all major dissolved constituents is recommended with the purpose of attempting to describe the variations in the specific conductance/chloride concentration relationships.

Massachusetts↗
USGS2025

Extracting data from maps: Lessons learned from the artificial intelligence for critical mineral assessment competition

The U.S. Geological Survey (USGS), Defense Advanced Projects Research Agency (DARPA), NASA Jet Propulsion Laboratory (JPL), and MITRE ran a 12-week machine learning competition aimed at accelerating development of AI tools for critical mineral assessments. The Artificial Intelligence for Critical Mineral Assessment Competition solicited innovative solutions for two challenges: 1) automated georeferencing of historical maps, and 2) automated feature extraction from historical maps. Competitors used a new dataset of historical map images to train, validate, and evaluate their models. Automated georeferencing pipelines attained a median root-mean square error of 1.1 km. Prompt-based extraction (i.e., with user input) of polygons, polylines, and points from geologic maps yielded median F1-scores of 0.77, 0.56, 0.35, respectively. Geologic maps pose numerous challenges for AI workflows because they vary significantly. However, despite its short duration, the competition yielded promising results that have since spurred further innovation in this area and led to the development of new AI tools to semi-automate key, time-consuming parts of the assessment workflow.

Applied Computing and Geosciences↗
USGS2023

Don’t Let Negatives Hold You Back: Accounting for Underlying Physics and Natural Distributions of Hydrothermal Systems When Selecting Negative Training Sites Leads to Better Machine Learning Predictions

Selecting negative training sites is an important challenge to resolve when utilizing machine learning (ML) for predicting hydrothermal resource favorability because ideal models would discriminate between hydrothermal systems (positives) and all types of locations without hydrothermal systems (negatives). The Nevada Machine Learning project (NVML) fit an artificial neural network to identify areas favorable for hydrothermal systems by selecting 62 negative sites where the research team had confidence that no hydrothermal resource exists. Herein, we compare the implications of the expert selection of negatives (i.e., the NVML strategy) with a random sample strategy, where it is assumed that areas outside the favorable structural ellipses defined by NVML are negative. Because hydrothermal systems are sparse, it is highly probable that, in the absence of a favorable geological structure, hydrothermal favorability is low. We compare three training strategies: 1) the positive and negative labeled examples from NVML; 2) the positive examples from NVML with randomly selected negatives in equal frequency as NVML; and 3) the positive examples from NVML with randomly selected negatives reflecting the expected natural distribution of hydrothermal systems relative to the total area. We apply these training strategies to the NVML feature data (input data) using two ML algorithms (XGBoost and logistic regression) to create six favorability maps for hydrothermal resources. When accounting for the expected natural distribution of hydrothermal systems, we find that XGBoost performs better than the NVML neural network and its negatives. Model validation was less reliable using F1 scores, a common performance metric, than comparing probability estimates at known positives, likely because of the extreme natural class imbalance and the lack of negatively labeled sites. This work demonstrates that expert selection of negatives for training in NVML likely imparted modeling bias. Accounting for the sparsity of hydrothermal systems and all the types of locations without hydrothermal systems allows us to create better models for predicting hydrothermal resource favorability.

Geothermal Resources Council Transactions↗
USGS2026

SlideDetect: Spatio-temporal landslide detection using a three-dimensional convolutional neural network

Landslides pose a serious and ongoing threat to both human lives and infrastructure worldwide; therefore, it is of interest to predict where and when landslides are likely to occur. Advances in machine learning techniques have spurred numerous studies aimed at estimating relative landslide propensity, but are limited to spatial (as opposed to temporal) prediction due to the sparsity of landslide timing data. We address this data gap by training SlideDetect, a 3-dimensional convolutional neural network (3D CNN), to identify landslides based on their spatial and temporal occurrence within multitemporal image stacks. We use an inventory of landsides triggered by the 2018 Hokkaido earthquake and two years of monthly composite optical imagery spanning this event. The model can identify not only landslide location but also landslide date with an area under the precision-recall curve (PR-AUC) of 0.84. We further present a new standard for presenting PR curve results that explicitly compares model performance at different confidence thresholds, allowing for clearer model evaluation and comparison. Our new approach to constraining landslide timing paired with this more consistent and objective method for evaluating model performance shows considerable promise, and with further application and testing, SlideDetect could enhance the data availability and tools needed to advance landslide hazard and risk assessments.

JGR Machine Learning and Computation↗
USGS2025

Measurement of in situ-produced cosmogenic nuclides in fine-grained quartz from shale

In situ -produced 10 Be in quartz is widely used to constrain exposure ages and denudation rates, traditionally measured in sand-sized grains. Here we report a new method for isolating fine-grained quartz from shale and demonstrate its reliability for grain sizes down to single microns. Sequential dissolution tests and analyses of grain size separates show that meteoric 10 Be is eliminated and that recoil losses during cosmogenic nuclide production are balanced by implantation from other mineral grains. High-temperature combustion leads to diffusion of meteoric 10 Be into fine-grained quartz and must be used with caution.

Nuclear Instruments and Methods in Physics Researc↗
USGS2026

Velocity-independent dry friction on mica: A realization of ideal Amontons-Coulomb friction

The Amontons-Coulomb friction law assumes that the frictional force between materials is independent of sliding velocity. However, as Coulomb noted, this is a rough approximation, and a second-order dependence of friction on the logarithm of sliding velocity is incorporated in a commonly used ‘rate- and state-dependent’ friction representation. Here we conduct shear experiments on mica, a layer-structured mineral, at temperatures ranging from 25 to 200ºC and under normal stress of 100 MPa. The friction coefficient clearly depends on the logarithmic sliding velocity at 25ºC, but rate sensitivity decreases with increasing temperature until at 200ºC, the friction coefficient is independent of sliding velocity. Our findings could initiate the development of velocity-independent frictional materials, realizing the ideal Amontons-Coulomb friction.

Physical Review Letters↗
USGS2026

Computation of regional groundwater budgets for the Virginia Coastal Plain aquifer system

Computation of detailed groundwater flow budgets for subdivisions of Virginia’s Coastal Plain aquifer system has enabled quantification and more thorough understanding of groundwater flow within this important water resource. A zone budget analysis conducted on previously published groundwater models of the Virginia Coastal Plain and Virginia Eastern Shore shows that groundwater conditions vary substantially throughout the Coastal Plain aquifer system due to local variations in hydrogeology and historical and ongoing variations in groundwater use and management. Decades of substantial groundwater withdrawal from the Coastal Plain aquifer system have fundamentally altered groundwater flow from pre-development conditions. Rates of sustainable withdrawal are limited because the downward groundwater flow rate into confined aquifers supplying groundwater is a relatively small portion of the total groundwater water budget for the aquifer system. Analyses of groundwater budgets from the Virginia Coastal Plain model show that groundwater flow is generally outward from the surficial aquifer to rivers and coastal water bodies and downward through a series of underlying aquifers and confining units to the Potomac aquifer, which is the deepest aquifer and the source of most groundwater withdrawals. Downward flow into the Potomac aquifer currently is estimated to be only 7 percent of total net precipitation-derived net recharge at the land surface but makes up about 66 percent of inflow to the aquifer in Virginia, with much of the remaining inflow occurring laterally from areas outside of defined groundwater budget regions in Virginia. For several decades prior to 2010, high rates of withdrawal from the Potomac aquifer resulted in substantial decline in groundwater storage in the aquifer and in most overlying aquifers and confining units. From 2010 to 2025, rates of withdrawal substantially lower than the historical maximum have resulted in small net increases in groundwater storage in the confined aquifer system for most regions of the Virginia Coastal Plain. Nevertheless, for the same period, groundwater storage for the entire model domain continues to incrementally decline, indicating that storage recovery in Virginia is offset by a continued decrease in storage in areas beneath the Chesapeake Bay or in adjacent areas of Maryland and North Carolina. Withdrawals from the Potomac aquifer have induced substantial downward flow which is a large part of groundwater budgets for confined aquifers such as the Potomac. Downward groundwater flow continues under current conditions, but because vertical flow rates are a function of the difference between water pressure in the upper surficial systems and lower confined units, those rates are lower than those in earlier decades as the confined water levels partially recover from larger groundwater withdrawals in the past. Geographically, groundwater flow is generally inward from perimeter regions of the Virginia Coastal Plain toward central regions with the largest withdrawal rates. Estimated groundwater inflow from coastal regions could be contributing to saltwater intrusion, though that was not measured directly in this study. Analyses of groundwater budgets from the Virginia Eastern Shore peninsula, a geographic region of the Virginia Coastal Plain, show that groundwater flow for that isolated aquifer system is generally outward from the surficial aquifer to coastal water bodies and downward into the confined Yorktown-Eastover aquifer system, which is the source of most withdrawals. Downward groundwater flow into the confined Yorktown-Eastover aquifer system is estimated to be less than 2 percent of total recharge and less than 9 percent of net recharge at the water table but makes up over 93 percent of all inflow to the confined aquifer system. Decades of substantial but relatively consistent groundwater withdrawals have induced greater downward flow rates into the confined aquifer system but also have resulted in loss of groundwater from storage. Currently, estimated storage loss accounts for slightly under 7 percent of withdrawals from the confined aquifer system. The current withdrawal rate from the confined Yorktown-Eastover system is near the highest reported rate for the Eastern Shore, which means that the storage depletion is expected to continue, even though groundwater levels appear to be relatively stable. Estimated groundwater flow rates upward from the confining unit underlying the Yorktown-Eastover system and small rates of inflow from coastal water bodies underscore ongoing concerns about up-coning and lateral intrusion of salty groundwater.

EarthArXiv↗
USGS2026

3D semantic mapping of surface geological features

Semantic mapping in 3D is fundamental to a wide range of geoscientific studies and applications, including geomorphology, hazard assessment, and environmental monitoring. However, automatically segmenting geological features from large-scale photogrammetric datasets remains a significant challenge. We present a methodology to address this gap. Using overlapping images collected over environments of interest, Structure-from-Motion (SfM) produces georeferenced point clouds and estimates camera poses. Existing large vision models, such as Segment Anything Model, segment objects in the images, generating pixel-segmentation associations. To produce pixel-point associations, we project the points back onto the camera image planes. As objects are independently segmented across multiple images with different perspectives, we develop a segmentation mosaicking algorithm to build probabilistic point-segmentation associations that combines the pixel-segmentation associations and pixel-point associations. Our methodology is validated using both synthetic data generated by Kubric and real-world UAV-SfM data. The implementation is designed to be compatible with existing SfM software, including Agisoft and OpenDroneMap, for photogrammetry mapping in geoscience studies. As a case study, we apply our method to the semantic mapping of precariously balanced rocks (PBRs), which provide upper-bound constraints on historical ground motion shaking intensity. To support object-level identification of PBRs, we additionally integrated Grounding DINO, enabling text-prompted segmentation of features of interest within UAV imagery. This case study demonstrates the effectiveness of our method in generating a 3D semantic map of PBRs, enabling spatial distribution of PBR fragility for earthquake hazard analysis.

Computers & Geosciences↗
USGS2026

Computation of regional groundwater budgets for the Virginia Coastal Plain aquifer system

Computation of detailed groundwater flow budgets for subdivisions of the Virginia Coastal Plain aquifer system has enabled quantification and more thorough understanding of groundwater flow within this important water resource. A zone budget analysis based on previously published groundwater models of the Virginia Coastal Plain and Virginia Eastern Shore indicates that groundwater conditions vary substantially throughout the Coastal Plain aquifer system because of local variations in hydrogeology and historical and ongoing variations in groundwater use and management. Decades of substantial groundwater withdrawal from the Coastal Plain aquifer system have altered groundwater flow from predevelopment conditions. Rates of sustainable withdrawal are limited because the downward groundwater flow rate into confined aquifers is a relatively small part of the total groundwater budget for the aquifer system compared to the rate of recharge at the land surface. Analyses of groundwater budgets from the Virginia Coastal Plain model indicate that groundwater flow is generally outward from the surficial aquifer to rivers and coastal waterbodies and downward through a series of underlying aquifers and confining units to the Potomac aquifer, which is the deepest aquifer and the source of most groundwater withdrawals. Downward flow into the Potomac aquifer is estimated to be only 7 percent of total net precipitation-derived net recharge at the land surface but makes up about 66 percent of inflow to the aquifer in Virginia, with much of the remaining inflow occurring laterally from outside of defined groundwater budget regions in Virginia. For several decades prior to 2010, high rates of withdrawal from the Potomac aquifer resulted in substantial decline in groundwater storage in the aquifer and in most overlying aquifers and confining units. From 2010 to 2023, rates of withdrawal substantially lower than the historical maximum resulted in small net increases in groundwater storage in the confined aquifer system for most regions of the Virginia Coastal Plain. Nevertheless, for the same period, groundwater storage for the entire model domain continues to incrementally decline, indicating that storage recovery in Virginia is offset by a continued decrease in storage in areas beneath the Chesapeake Bay or adjacent areas of Maryland and North Carolina. Withdrawals from the Potomac aquifer have induced substantial downward flow which is a large part of groundwater budgets for confined aquifers such as the Potomac. For the most recent simulated conditions (2023) downward groundwater flow continues, but because vertical flow rates are a function of the difference between water pressure in the upper surficial systems and lower confined units, rates of downward flow are lower than those in earlier decades as the confined water levels partially recover from larger groundwater withdrawals in the past. Geographically, groundwater flow is generally inward from perimeter regions of the Virginia Coastal Plain toward central regions with the largest withdrawal rates. Groundwater inflow from coastal regions could be contributing to saltwater intrusion, even though that was not measured in this study. Analyses of groundwater budgets from the Virginia Eastern Shore peninsula, a geographic region of the Virginia Coastal Plain, indicate that groundwater flow for that isolated aquifer system is generally outward from the surficial aquifer to coastal water bodies and downward into the confined Yorktown-Eastover aquifer system, which is the source of most withdrawals. Downward groundwater flow into the confined Yorktown-Eastover aquifer system is estimated to be less than 2 percent of total recharge and less than 9 percent of net recharge at the water table but makes up more than 93 percent of all inflow to the confined aquifer system. Decades of substantial but relatively consistent groundwater withdrawals have induced greater downward flow rates into the confined aquifer system but also have resulted in loss of groundwater from storage. For the most recent simulated period (2023), estimated storage loss accounts for slightly under 7 percent of withdrawals from the confined aquifer system. The reported withdrawal rate for this period from the confined Yorktown-Eastover system is near the highest reported rate for the Virginia Eastern Shore, which means that the storage depletion is expected to continue, even though groundwater levels appear to be relatively stable. Estimated groundwater flow rates upward from the confining unit underlying the Yorktown-Eastover system and low rates of inflow from coastal water bodies underscore ongoing concerns about up-coning and lateral intrusion of salty groundwater.

Virginia↗
USGS2025

Statistical approaches for modeling correlated grade and tonnage distributions and applications for mineral resource assessments

Correlations between grade and tonnage exist in mineral resource data compiled from published reports, but they are not always addressed during quantitative assessment of undiscovered mineral resources. Failure to account for correlated grade and tonnage distributions can result in geologically unrealistic assessment results. Current software tools simulate univariate ore tonnage and multivariate resource grades of undiscovered deposits independently. As a result, analysts are forced to rely on ad-hoc solutions to minimize the correlation issues by: 1) creating subsets of data with restricted criteria; 2) truncating grade and tonnage distributions; and 3) testing model robustness using exploratory data analysis. While these methods represent pragmatic solutions, the statistical solutions presented here provide additional options to address real correlations in grade and tonnage data used for mineral resource assessments. We present a modified version of the MapMark4 package in R that introduces two alternatives for modeling grade and tonnage distributions, consisting of a multivariate solution that accounts for correlations between ore tonnage and metal grades and an empirical solution that utilizes simple random sampling with replacement to reproduce coupled grades and tonnages from the input data. We present simulations for contained ore and metal for three case studies representing tungsten skarn, komatiite-hosted nickel, and sediment-hosted carbonate amagmatic zinc-lead (Mississippi Valley-type) deposits. Employing the methods presented here yields quantitative mineral resource assessment results that more closely reflect the empirical distributions of grades and tonnages observed in nature and expands the applicability of these tools for ongoing critical mineral resource assessments.

Applied Computing and Geosciences↗