USGS Science⌕ Search

SEARCH · USGS Science

Results for “Environmental Modeling and Software”

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.

1,662 records · Page 26Linked to original sources
USGS2026

Detecting earthquakes in noisy real-time GNSS data with deep learning for improved PGD magnitude estimation

To disseminate accurate and useful warnings, earthquake early warning (EEW) systems must quickly determine the size and location of an earthquake to estimate expected shaking. Traditional seismic‐based algorithms tend to underestimate the true magnitudes of large earthquakes, a phenomenon known as magnitude saturation. This limitation motivated the recent inclusion of Global Navigation Satellite Systems (GNSS) data into the U.S. Geological Survey’s ShakeAlert EEW system with the Geodetic First Approximation of Size and Time (GFAST) algorithm because GNSS data do not saturate with large ground motions. However, the noise levels of GNSS data are very high compared with traditional seismic data, which obscures P ‐wave arrivals and can result in less accurate magnitude estimations if displacement amplitudes are low, such as for lower magnitude earthquakes or large source–station distances. In this study, we develop a deep‐learning model that detects earthquakes in GNSS data and use the Ridgecrest, California, earthquake sequence as a case study to demonstrate how the model could act as a filter to reduce the amount of low‐quality data that enters an algorithm like GFAST. To preserve our limited real earthquake data for model inference, we generated a training dataset composed of >700,000 synthetic displacement waveforms. We combined the synthetic waveforms with real‐time GNSS noise to produce realistically noisy training waveforms and then tested our model on additional synthetic data and performed inference using the real data that were held back. We discuss the performance of our trained model on both the unseen synthetic data and real inference data. Our model can be used to selectively filter only high‐quality data where an earthquake signal is observed for input into an algorithm like GFAST (outperforming a simple signal‐to‐noise ratio–based filter) to reduce the error in GFAST’s real‐time earthquake magnitude estimations.

California↗
USGS2025

An empirical Green’s function approach for isolating directivity effects in earthquake ground-motion amplitudes

In this study, we apply an empirical Green’s function (eGf) method within a ground‐motion modeling framework to mitigate trade‐offs between source, path, and site effects. Many physical processes contribute to spatial variations in observed ground motions, including earthquake radiation pattern, directivity, variable path attenuation, and site effects. Current nonergodic ground‐motion models use spatially varying coefficients for path and site effects, but they do not address trade‐offs with complex earthquake source effects. To quantify the influence of directivity on ground‐motion amplitudes, we use records from multiple smaller earthquakes with epicenters near that of a larger event. We use these small magnitude events as eGfs and estimate repeatable path and site effects at individual stations, assuming that the average adjustments are not controlled by directivity. We adjust residuals from the larger earthquake using the eGf terms, isolating effects related to the rupture. This method clearly enhances the observed broadband directivity observed in the 2022 M 5.1 and 2007 M 5.4 Alum Rock earthquake ground motions, reinforcing the conclusion that their ruptures were unilateral. For the 2004 M 6.0 Parkfield earthquake, we find a bilateral rupture model better fits the data because variations in rupture velocity, slip rate, and slip distribution seem to have a stronger effect on the ground motions than rupture direction alone. Applying eGf adjustments reduces the standard deviation of the rupture models over the three earthquakes by 32% on average and by up to 57% for the 2022 Alum Rock earthquake, confirming we have effectively removed repeatable effects related to the wave propagation path and site response. We propose a novel measure of the frequency‐dependent directivity amplification strength as the reduction in ground‐motion residual variability gained by fitting a directivity model; for the three earthquakes considered, this parameter varies between 25% and 75%, indicating that directivity can strongly influence ground motions and should be considered in ground‐motion modeling.

California↗
USGS2025

Flood-inundation maps for Río de la Plata in and near Comerío, Puerto Rico, 2025

Digital flood-inundation maps for a 3.1-mile reach of Río de la Plata in and near Comerío, Puerto Rico, were created by the U.S. Geological Survey (USGS). Water-surface profiles were computed for the stream reach by using a one-dimensional steady-state step-backwater model. The model was calibrated to the current (2025) stage-streamflow relation (rating curve 11.0) for the USGS streamgage 50043800, Río de la Plata at Comerío, Puerto Rico. The resulting hydraulic model was then used to compute 16 water-surface profiles for water levels (flood stages) ranging from 10.00 to 40.00 feet at the streamgage and ranging from “action stage” to above “major flood stage” as reported by the National Weather Service. The 40.00-foot stage was selected because it exceeds the peak stage of 34.86 ft recorded during Hurricane Maria at the USGS streamgage 50043800, Río de Plata at Comerío, Puerto Rico. The simulated water-surface profiles were then used in combination with a digital elevation model derived from light detection and ranging data to map the inundated areas associated with each flood profile. The flood-inundation maps and the supporting hydraulic model produced by this study can be used by emergency managers and local officials to assess flood mitigation strategies and to define flood hazard areas to help protect life and property, to coordinate flood response activities such as evacuations and road closures, and to aid post-flood recovery efforts.

Comerio↗
USGS2024

Separating signals in elevation data improves supervised machine learning predictions for hydrothermal favorability

A recent study identified topography (land surface elevation above sea level) as an important input dataset (feature) for predicting the location of hydrothermal systems in the Great Basin in Nevada. Yet, topography is generally a result of more than one geological process and may consequently contain multiple distinct signals. For example, the geologic evolution of the Great Basin has produced both crustal thickening (i.e., regional-scale trends in elevation) and thinning via Basin and Range extensional faulting (i.e., valley-scale topographic relief). We postulate that these geologic processes may affect the occurrence of hydrothermal systems differently. Therefore, we separate the regional trend from the valley-scale signal in the Great Basin, and then use them separately to evaluate the importance of each as predictors for hydrothermal favorability. Our prior work applying supervised machine learning (ML) using the data from the Nevada Machine Learning Project demonstrated that employing a training strategy that randomly selects negative training sites produces better performing models for predicting hydrothermal favorability than a training strategy that uses expert-selected negatives. The models created using both training strategies exhibited a west-east geographic trend in the predictions for the favorability of hydrothermal resources. These models generally predicted higher favorability in western Nevada and lower favorability in eastern Nevada. This west-east trend in predicted favorability correlates with elevation across the Great Basin, which trends higher from west to east. By separating the original elevation feature into distinct features for elevation trend (i.e., regional-scale topography) and detrended elevation (i.e., valley-scale or local relative topography), we find that models using the separated topographic signals consistently outperform competing models that use the original elevation feature. Although western Nevada still exhibits higher favorability than eastern Nevada, using separated signals for regional elevation and local structure reduces the west-east prediction trend in the region and emphasizes structures associated with hydrothermal upflow. This work emphasizes how carefully engineering features to represent geological conditions relevant to hydrothermal systems allows ML algorithms to detect important patterns for predicting hydrothermal resource favorability and leads to better model performance.

Conference Paper↗
USGS2024

Cold blood in warming waters: Effects of air temperature, precipitation, and groundwater on Gulf Sturgeon thermal habitats in a changing climate

Objective In a changing climate, the effects of air temperature, precipitation, and groundwater on water temperature and thermal habitat suitability for Gulf Sturgeon Acipenser desotoi , listed as threatened under the U.S. Endangered Species Act, are not well understood. Hence, we incorporated these factors into thermal habitat models to forecast how Gulf Sturgeon may be affected by wide‐ranging climate change scenarios in 2024–2074. Methods Using data from the Choctawhatchee River, Florida, we developed precipitation‐ and groundwater‐corrected air–water temperature models, compared their accuracy with that of conventional air–water temperature models used in fisheries management, and projected future Gulf Sturgeon thermal habitat suitability for normal physiological functioning and fieldwork (i.e., population sampling and telemetry surgeries) in summer (May–August) under 16 climate change scenarios. Result Precipitation‐ and groundwater‐corrected models were more accurate than conventional air–water temperature models (mean improvement in adjusted R 2 = +0.45; range = +0.09 to +0.75). Water temperature was projected to warm at widely variable rates across climate change scenarios encompassing different air temperature, precipitation, and groundwater regimes. Importantly, Gulf Sturgeon summer aggregation areas were cooler and influenced more by precipitation and groundwater and less by air temperature than were non‐aggregation areas. If precipitation and groundwater—as drivers of cooling—become warm in a changing climate, summer aggregation areas were projected to exhibit thermal habitat degradation equivalent to or greater than that of non‐aggregation areas. Conclusion Our results add hydrological context to the premise that aggregation areas provide cool water and energetic savings for Gulf Sturgeon during summer, underscoring the importance of protecting these habitats through groundwater conservation, water quality monitoring, and riparian/watershed habitat management. Our findings indicate that identifying thermally appropriate times for fieldwork activities will be increasingly important and time‐restricted as climate change intensifies. However, our research provides managers with a portfolio of water temperature models and an accurate, cost‐effective, management‐relevant approach to forecasting thermal habitat conditions for Gulf Sturgeon and other species in a changing climate.

Alabama, Florida↗
USGS2025

Multi-lithofacies alluvial characterization via airborne electromagnetic-borehole fusion using ordinary interval kriging and geologic constraints

Airborne electromagnetic (AEM) survey provides extensive spatial coverage and detailed resolution at the near surface and can be used to develop hydrogeological models. However, utilization of AEM data is not straightforward because AEM resistivity is an indirect measurement for inferring sediment types. This study develops an ordinary interval kriging (OIK) algorithm and a resistivity-to-multi-lithofacies (R2ML) data fusion workflow for multi-lithofacies alluvial characterization. OIK utilizes irregular interval data to construct three-dimensional (3D) resistivity fields from one-dimensional inverted AEM resistivity models. The R2ML workflow maps the resistivity field generated from OIK into a multi-facies lithological model, incorporating geologic constraints derived from well logs and geological observations. The numerical and real-world cases demonstrate that OIK is computationally efficient, accounts for 3D anisotropy, and minimizes the smoothing effect, thereby preserving resistivity contrasts and reducing interpolation uncertainty. The methodology is applied to lithologic characterization of the Mississippi River Valley alluvial aquifer (MRVA) in the Shellmound area, Mississippi, U.S. A frequency-domain AEM survey was conducted to support groundwater studies for the managed aquifer recharge (MAR) to the MRVA. The resulting lithological model, including four types of lithofacies—clay, very fine sand, fine-medium sands, and graveliferous sands, illustrates the geomorphological processes of the MRVA and implies potential MAR. The alignment between the lithological model and existing geological and hydrogeological investigations demonstrates that OIK and R2ML workflow effectively capture the subsurface architecture of the MRVA. The methods have broad applicability for characterizing alluvial aquifers through AEM-borehole data fusion, supporting sustainable groundwater management.

Mississippi↗
USGS2025

A 10-year continuous daily simulation of chloride flux from a suburban watershed in Fairfax County, Virginia, USA

Increasing levels of chloride in surface water are associated with detrimental effects on water quality, aquatic ecosystems, infrastructure, and human health. Numerous mass-balance studies have inferred watershed transport processes by interpreting chloride inputs and outputs, but few represent internal dynamics explicitly. We constructed a coupled water/chloride mass balance model to gain insights into storage, residence time, and transport processes in a 10-km 2 urban watershed. The model, which operates over a 10-year period at a daily time scale, represents storage in a dynamic soil-moisture reservoir, quick-flow runoff from storm events, and slow-flow runoff that sustains streamflow in dry weather. The calibrated model accurately represented (a)the observed transition from a streamflow enrichment regime in cold months to a dilution regime in warmer months, (b) the observed tendency for late-summer concentrations to be higher after winters with heavy snowfall, and (c) a period-of-record downward trend in chloride concentration likely associated with a downward trend in annual snowfall. Estimated chloride inputs averaged 195 metric tons per year, while the average output was 270 metric tons per year. In contrast, estimated storage was only 107 metric tons. The estimated mean residence time in groundwater was 1.27 years. This short residence time indicates that efforts to reduce inputs will manifest as decreased concentrations in streamflow on a management-relevant time scale of several years. The coupled mass balance model yielded insights into internal watershed dynamics that would not be possible from simple input/output analysis; such models can be useful tools for gaining insight into small watershed hydrology and pollutant transport.

Virginia↗
USGS2025

Divide and conquer: Separating the two probabilities in seismic phase picking

There are two fundamental probabilities in the seismic phase picking process – the probability of the existence of a seismic phase (detection probability) and the probability of correctly identifying the phase arrival time (timing probability). The nearly ubiquitous approach in developing deep learning phase picking models is to use a kernel, such as a truncated Gaussian, to mask the labeled phase arrival time, and train a segmentation model. Once a model is trained, the times of the peaks in the output are taken as phase arrival times (picks) and the height of the peaks are taken as “probability” of the picks. Here, we show that this “probability” represents neither the detection nor the timing probabilty because this approach forces the output to follow the shape of the kernel. We introduce an approach using two models to estimate these two distinct probabilities. We use a binary classifier with a calibrated confidence to address the detection probability and a multi-class classifier to obtain a probability mass function to address the timing probability. This new approach makes the deep learning-based phase picking process more interpretable and gives us options to logically control seismic monitoring workflows.

Geophysical Journal International↗
USGS2025

Evaluating Laramide orogenesis via flexural basin response in the San Juan basin, New Mexico and Colorado

A challenge in interpreting the location, timing, and magnitude of ancient orogenic events is that ongoing uplift and erosion in the hinterlands often destroys much of the primary record of these events. However, basin-thickness patterns in the sedimentary record can provide complimentary evidence of uplift via flexural effects. Here, we deploy well-log correlation, isochores, basin modeling, flexural modeling, and subcrop mapping to evaluate the Late Cretaceous to Paleogene basin response to Laramide tectonism in the San Juan basin. A wedge of upper Campanian to Maastrichtian sedimentary rock thickens from 200 to 800 meters from southeast to northwest in the basin. This pattern can be successfully simulated via flexural modeling if we infer early Laramide uplift along the northwest basin flank that produced a 0.8 km high topographic load. The Laramide unconformity bounds the top of this Upper Cretaceous sedimentary wedge and truncates progressively older strata to the east, further supporting a westward tilt of the basin. The onset of Campanian Laramide flexure may have also contributed to the profound transgression from the upper Menefee Formation to the Lewis Shale. The Paleocene isochore map displays an approximately symmetrical pattern, with thickening towards the center of the basin. This suggests the possibility of competing flexural loads. The base Eocene structure indicates an asymmetric deep on the northeast flank of the basin, providing flexural evidence of contemporaneous uplift/loading of the Nacimiento uplift and Archuleta arch; this has been modeled as ~2.1 km load height. Both Cretaceous and Paleocene sedimentary wedges are narrow, suggesting low flexural rigidity; modeled effective elastic thicknesses (EET) are 20-30 km, comparable to estimates of modern EET for the region.

Colorado, New Mexico↗
USGS2025

Fluid inclusion constraints on the geometry of the magmatic plumbing system beneath Mauna Loa – Part 2: Xenoliths

Mauna Loa volcano erupts crystal-poor material at its summit and more crystal-rich material on its rift zones. Some of the more olivine-rich lava flows contain xenoliths with diverse mineralogy, including cumulate harzburgites with high-Mg# orthopyroxenes and high-Fo olivines (both > 84). Previous experimental work and thermodynamic modelling has proposed that high-Mg# orthopyroxenes only crystallize from Mauna Loa melts at high pressures (> 6 kbar, > 20 km), leading to suggestions that there is a region of sub-Moho magma storage at Mauna Loa in addition to the geophysically imaged magma reservoir at 2–5 km depth below the summit. We use melt and fluid inclusion barometry combined with thermodynamic models to further investigate this suggestion. Fluid inclusion data from harzburgites and dunitic xenoliths yield storage depths remarkably similar to those found in non-xenolithic crystals from lavas and tephras, with a clear peak at ~ 2–3 km (below the summit). Depths from melt inclusions in these xenoliths overlap with fluid inclusion pressures, ruling out the possibility of fluid inclusion re-equilibration during a period of stalling in a shallower reservoir. We examine five different thermodynamic models and find that the minimum pressure of olivine-orthopyroxene co-saturation varies by ~ 4 kbar (~ 12 km). These models also fail to predict that orthopyroxene is stable in ~ 15–80% of compositionally relevant experimental charges which grew orthopyroxene. Overall, this shows that phase stability modelling is an unreliable method of determining magma storage depth at Mauna Loa. We suggest that model discrepancies reflect a lack of experimental constraints on orthopyroxene stability at > 1200 ℃ and 0.01–5 kbar. Based on the presence of large oikocrystic orthopyroxenes completely enclosing rounded olivine chadacrysts, we suggest that these harzburgitic xenoliths formed through the reaction of intruding melts with olivine mush piles within the Mauna Loa edifice at ~ 3 km depth below the summit, with no need for a deeper storage reservoir. The predominance of pre-eruptive shallow storage means that there is more chance of detecting reservoir destabilization with geophysical monitoring techniques compared to a scenario where melts are supplied from sub-Moho reservoirs.

Hawaii↗
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↗
USGS2024

Development of USGS NSHMs: Do small changes in hazard imply small changes in risk?

One of the flagship products from the U.S. Geological Survey (USGS) is the National Seismic Hazard Model (NSHM). Since 1976, the NSHM has been periodically updated to reflect newly published earthquake science and provide probabilistic estimates of seismic hazard for the United States. During each update cycle, alternative models are deliberated, analyzed, and documented through logic trees and their corresponding logic tree branch weights. For example, the decision to modify a logic tree branch weight may be influenced by sensitivity analyses of the logic tree branches in their effects on the mean hazard. However, do small changes in traditional measures of hazard imply small changes in risk? In this study, we make use of two update cycles of the USGS NSHMs and a National Bridge Inventory (NBI) from the Federal Highway Administration (FHWA) to explore the preceding question. Specifically, we first identify geographic locations in the conterminous United States in which the change in hazard from one cycle to another is relatively small. Next, we model the seismic risk to highway bridges for these locations and for each update cycle, while simultaneously distinguishing low hazard environments from high hazard environments. These data enable quantitative analysis of how much changes in risk can be expected given small changes in hazard, investigating the importance of risk models in decision-making during development of the USGS NSHMs.

Conference Paper↗
USGS2025

Simulated effects of future water availability and protected species habitat in a perennial wetland, Santa Barbara County, California

This study evaluates the potential water availability in Barka Slough and the effects of changing hydrological conditions on the aquatic habitat of five protected species. Barka Slough is a historically perennial wetland at the downstream western end of the San Antonio Creek Valley watershed (SACVW). A previously published hydrologic model of the SACVW for 1948–2018 was extended to include 2019–2021 and then modified to simulate the future years of 2022–2051. Two models simulating the future years of 2022–2051 were constructed, each with different climate inputs: (1) a repeated historical climate and (2) a 2070-centered Drier Extreme Warming climate (2070 DEW). The model with the 2070 DEW climate had warmer temperatures and an increase in average annual precipitation driven by larger, albeit more infrequent, precipitation events than the model with the historical climate. Simulated groundwater pumpage resulted in cumulative groundwater storage depletion and groundwater-level decline in Barka Slough in both future models. The simulations indicate that Barka Slough may transition from a perennial to an ephemeral wetland. Streamflow, stream disconnection, and depth to groundwater are key habitat metrics for federally listed species in Barka Slough. Future seasonal conditions for each metric are more likely to affect federally listed species’ habitats under 2070 DEW climatic conditions. Future seasonal streamflow volume may negatively impact unarmored threespine stickleback ( Gasterosteus aculeatus williamsoni ) and tidewater goby ( Eucyclogobis newberryi) habitats. Future seasonal stream disconnection may negatively impact the unarmored threespine stickleback habitat. Future groundwater-level decline may negatively impact Gambel’s watercress ( Nasturtium gambelii ) and La Graciosa thistle ( Cirsium scariosum var. loncholepis ) habitats and could influence the ability to use Barka Slough as a restoration or reintroduction site for these species. Results from this study can be used to inform water management decisions to sustain future groundwater availability in the SACVW.

California↗
USGS2026

Landslide volume estimation from seismic waveform features using a global catalog of seismogenic events

Seismic waves generated by large, rapid landslides encode information about the source and can be analyzed rapidly following an event. Even remote landslides can pose hazards to downstream communities, so rapid detection and characterization using existing seismic monitoring networks could be beneficial. In this study, we expand on past regionally limited work by presenting a globally applicable method for estimating landslide volume from seismic features that could be integrated into future landslide seismic monitoring frameworks. We train the model using multivariable linear regression and five seismic features derived from recordings of 129 landslide events of a range of styles and locations with independently estimated volumes. We present two preferred models, one that combines long‐period (LP) and high‐frequency (HF) features and one for use on smaller landslides without observed long‐period signals. We find that our best‐performing model, applicable to landslides larger than 100,000 m 3 with signals containing observable long‐period energy, requires only two features: LP (20–100 s) absolute maximum amplitude and HF (1–5 Hz) rise time (time between the signal start and the maximum envelope amplitude) and has an R 2 score of 0.79. This model predicted volumes within one order of magnitude for 55 out of 58 events. We find that this combination mitigates the trade‐off between mass and acceleration and other variations of landslide style that limit methods based on amplitude alone.

Seismological Research Letters↗
USGS1990

Preconditioned Conjugate-Gradient 2 (PCG2), a computer program for solving ground-water flow equations

This report documents PCG2: a numerical code to be used with the U.S. Geological Survey modular three-dimensional, finite-difference, ground-water flow model. PCG2 uses the preconditioned conjugate-gradient method to solve the equations produced by the model for hydraulic head. Linear or nonlinear flow conditions may be simulated. PCG2 includes two reconditioning options: modified incomplete Cholesky preconditioning, which is efficient on scalar computers; and polynomial preconditioning, which requires less computer storage and, with modifications that depend on the computer used, is most efficient on vector computers. Convergence of the solver is determined using both head-change and residual criteria. Nonlinear problems are solved using Picard iterations. This documentation provides a description of the preconditioned conjugate gradient method and the two preconditioners, detailed instructions for linking PCG2 to the modular model, sample data inputs, a brief description of PCG2, and a FORTRAN listing.

Water-Resources Investigations Report↗
USGS2025

Accurate simulation of flow through dipping aquifers with MODFLOW 6 using enhanced cell connectivity

In simulations of groundwater flow through dipping aquifers, layers of model cells are often “deformed” to follow the top and bottom elevations of the aquifers. When this approach is used in MODFLOW, adjacent cells within the same model layer are vertically offset from one another, and the standard conductance-based (two-point) formulation for flow between cells does not rigorously account for these offsets. The XT3D multi-point flow formulation in MODFLOW 6 is designed to account for geometric irregularities in the grid, including vertical offsets, and to provide accurate results for both isotropic and anisotropic groundwater flow. A recent study evaluated the performance of the standard formulation and XT3D using a simple, synthetic benchmark model of a steeply dipping aquifer. Although XT3D generally improved the accuracy of flow simulations relative to the standard formulation as expected, neither formulation produced accurate flows in cases that involved large vertical offsets. In this paper, we explain that the inability of XT3D to produce accurate flows in the steeply dipping aquifer benchmark was not due to an inherent limitation of the flow formulation, but rather to the limited cell connectivity inherent in the most commonly used discretization packages in MODFLOW 6. Furthermore, we demonstrate that XT3D is able to produce the expected accuracy when adequate cell connectivity is introduced using MODFLOW's unstructured grid type and the aquifer is discretized vertically using at least two model layers.

Groundwater↗
USGS2024

Thermo-hydrologic processes governing supra-permafrost talik dynamics in discontinuous permafrost near Umiujaq (Québec, Canada)

Widespread supra-permafrost talik formation is currently recognized as a critical mechanism that could accelerate permafrost thaw in the Arctic (e.g., Connon et al. 2018; Farquharson et al. 2022). However, the trajectory of permafrost dynamics following talik formation may prove difficult to predict. Physically-based cryohydrogeologic models provide a powerful tool for understanding processes and factors controlling talik dynamics and, ultimately, how permafrost will respond to climate change. Such models are typically used to represent multiple non-linear processes relevant for groundwater systems in cold regions, such as coupled heat and groundwater movement, including freeze-thaw dynamics and the effects on the surface energy balance and the subsurface thermal and hydraulic properties (Lamontagne-Hallé et al. 2020). Though cryohydrogeologic modeling advances have been made in simulating talik dynamics, few applications have been tested against robust long-term hydrometeorological and subsurface observations.

Quebec↗
USGS2025

Estimated ultimate recovery (EUR) Prediction for Eagle Ford Shale using integrated datasets and artificial neural networks

The estimated ultimate recovery (EUR) is an important parameter for forecasting oil and gas production and informing decisions regarding field development strategies. In this study, we combined site-specific geologic, completion, and operational parameters with the predictive capabilities of machine learning (ML) models to predict EURs of the wells for the Eagle Ford Marl Continuous Oil Assessment Unit. We developed an extensive dataset of wells that have produced from the lower and upper Eagle Ford Shale intervals and reduced the model complexity using principal component analysis. We tested the ML models and estimated the sensitivities of ML-predicted EURs to changes in the values of different input variables. The results of applying the optimized ML model to the Eagle Ford suggest that the approach developed in this study could be promising. The ML estimates of the EURs fit the DCA-based values with an R 2 ~ 0.9 and a mean absolute error of ~36 × 10 3 bbl. In the lower Eagle Ford Shale, the EUR estimates were found to be most sensitive to changes in porosity, net thickness of the interval, clay volume, and the API gravity of the oil; and that in the upper Eagle Ford Shale they were most sensitive to changes in the total organic carbon and water saturation, which suggests that it could be important to consider these parameters in assessing these intervals or close analogs.

Louisiana, Mississippi, Texas↗