SEARCH · Geology Reports

Results for “Frontiers for Young Minds”

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

SourceResearch connectionFrom yearThrough year

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

1,056 records · Page 7Linked to original sources

USGS2026

Computational electromagnetic geophysics for groundwater system studies: A review on established practices and recent advances

Identifying effective solutions for locating groundwater resources and ensuring the quality of drinking water is increasingly urgent, given the challenges posed by climate change and population growth. This review investigates electromagnetic geophysical imaging techniques, in both time- and frequency-domain, that can provide valuable insights for groundwater assessment. We explore computational electromagnetic methods used to evaluate electromagnetic data and several recent hydrogeophysical case studies. As open-source frameworks for modeling electromagnetic geophysical problems become available, a broader range of researchers can interpret their data with computationally advanced software. We provide an overview of documented open-source codes for evaluating electromagnetic data and analyze various hydrological targets in relation to their electromagnetic surveying technique and the computational method applied. Furthermore, we evaluate the potential of advanced computational techniques, including three-dimensional modeling, non-deterministic inversion and machine learning, to couple geophysical with numerical groundwater modeling and apply it in groundwater system studies. Despite obstacles such as complexity and resource demands, our findings indicate that the quantification and integration of predictive uncertainties from both electromagnetic and hydrological data and simulations would significantly improve the reliability of hydrogeophysical models. This can lead to a deeper understanding of groundwater systems and improved management practices.

Journal of Hydrology↗

USGS2025

Systematic approach to prioritize wells for effective groundwater monitoring and management in the Arkansas Headwaters Basin, Colorado, USA

Study region The Arkansas Headwaters Basin, an intermountain basin in the Southern Rocky Mountains of North America. Study focus Our specific focus is choosing a set of wells to support a possible future regional groundwater-surface water model that would support water management. We present a three-step process using multiple criteria to score, predict, and choose prioritized wells that capture the full distribution of data including extremes. The three-step process provides accessible visualizations, fiscally efficient well prioritization, and screening useful for subsequent groundwater modeling. The novelty of the proposed methodology is the systematic approach integrating a scoring and a predictive approach to support a selection path. The systematic approach may be broadly adapted for other basins. New hydrological insights for the region Understanding regional hydrology hinges on efficient collection of hydrologic data that captures the relevant dynamics including extremes. The present study, a case study for a particular basin in the Southern Rocky Mountains, is the first use of a scripted (R software) strategy to select an economical and representative set of monitoring wells. Our findings suggest caution when using proximity as a proxy for correlation, because proximal wells in the same geologic formation and similar depths are not always correlated. In the Arkansas Headwaters Basin, subsurface geology may be less influential on groundwater elevations than broader hydrologic influences, such as regional drought.

Colorado↗

USGS2026

Making many out of one: Synthetic geologic deformation model distributions for use in USGS NSHM25‐PRVI Puerto Rico-U.S. Virgin Island update

A key use‐case of geologic slip rates is within deformation models used in probabilistic seismic hazard analyses. Field‐derived geologic slip rates have formed the cornerstone of deformation models in such applications for decades. Recent advancements in seismic hazard analyses have expanded the use of faults for which geologic slip rates are not well constrained using categorical slip rate estimates. Because of these advancements, application of a geologic deformation model for use in 2025 U.S. Geological Survey National Seismic Hazard Model Puerto Rico‐U.S. Virgin Islands (NSHM25‐PRVI) proved challenging due to: (1) a lack of field‐based geologic slip rates, and (2) a lack of epistemic uncertainty distributions within a broad range of estimated slip rates. Preliminary versions of the NSHM25‐PRVI model sampled these slip rate bins in a coincident manner along preferred and extreme value branches, which yielded untenable correlations in mean hazard results. To minimize the influence of correlated uncertainties amid these challenges, we develop a synthetic epistemic uncertainty distribution for deformation rate on each crustal fault. Each fault has a weighting schema across four possible distribution shapes: uniform, normal, triangular favoring local minima, and triangular favoring local maxima. The synthetic distributions are then sampled several times for each logic tree branch. The results provide a more realistic distribution of rates across the study region as compared with using correlated extrema sampling. This exploration of our method in a small region like PRVI can pave the way for larger‐scale, more complicated applications (e.g., western United States).

Puerto Rico, US Virgin Islands↗

USGS2025

Using gridded seismicity to forecast the long-term spatial distribution of earthquakes for the 2025 Puerto Rico and U.S. Virgin Islands National Seismic Hazard Model

Gridded (or background) seismicity models are a critical component of probabilistic seismic hazard assessments, accounting for off‐fault and smaller‐magnitude earthquakes. They are typically developed by declustering and spatially smoothing an earthquake catalog to estimate a long‐term seismicity rate that can be used to forecast future earthquakes. Here, we present new gridded seismicity models for use in the 2025 National Seismic Hazard Model (NSHM) for Puerto Rico and the U.S. Virgin Islands (PRVI). The previous PRVI NSHM was released in 2003, and our new models incorporate updates to both data and methodology. We utilize an updated earthquake catalog based on improved Puerto Rico Seismic Network data with newly characterized completeness epochs. The catalog is divided into crustal, subduction interface, and intraslab seismicity using new methods and Slab2 subduction zone geometries. To forecast the long‐term spatial distribution of earthquakes, we use an updated methodology developed for the 2023 U.S. 50‐state NSHM, considering three declustering methods and two spatial smoothing methods based on 2D Gaussian kernels. To adapt it for the complex seismotectonics of the region, we also adopt probabilistic methods to account for events with unknown depths and uncertainties in tectonic classification, and develop a new method for spatial scaling to counteract the effects of spatial variability in network coverage while maintaining the use of smaller events. Finally, we test the performance of these spatial models in forecasting the location of M w ≥ 5earthquakes in the region. Our updated methodology improves the representation of epistemic uncertainty relative to the 2003 model, and our results demonstrate the effectiveness of the new measures we have introduced to address heterogeneities in network detection and systematically evaluate forecast performance.

Puerto Rico, U.S. Virgin Islands↗

USGS2025

Three-dimensional seismic velocity model for the Cascadia Subduction Zone with shallow soils and topography, version 1.7

The U.S. Geological Survey’s seismic velocity model for the Cascadia Subduction Zone provides P- and S-wave velocity ( V P and V S , respectively) information from 40.2° to 50.0° N. latitude and −129.0° to −121.0° W. longitude, and is used to support a variety of research topics, including three-dimensional (3D) earthquake simulations and seismic hazard assessment in the Pacific Northwest. This report describes an update to the previous version (v) 1.6 of the 3D seismic velocity model for the Cascadia Subduction Zone. This new model (herein referred to as v1.7) contains more detailed near-surface structure for improved earthquake ground motion modeling. Updated features include the addition of a new shallow soil velocity model in the top few hundred meters and the option of adding user-specified topography. Although v1.6 of the Cascadia seismic velocity model has a minimum V S of 600 meters per second (m/s), the new model (v1.7) has a minimum V S of approximately 40 m/s. Overall, this update will allow for more accurate ground motion estimates from 3D simulations of scenario earthquakes in the Cascadia Subduction Zone region.

British Columbia, California, Oregon, Washington↗

USGS2023

Evaluation of models for estimating hydraulic conductivity in glacial aquifers from NMR logging

Nuclear magnetic resonance (NMR) logging is a promising method for estimating hydraulic conductivity ( K ). During the past ∼60 years, NMR logging has been used for petroleum applications, and different models have been developed for deriving estimates of permeability. These models involve calibration parameters whose values were determined through decades of research on sandstones and carbonates. We assessed the use of five models to derive estimates of K in glacial aquifers from NMR logging data acquired in two wells at each of two field sites in central Wisconsin, USA. Measurements of K , obtained with a direct push permeameter (DPP), K DPP , were used to obtain the calibration parameters in the Schlumberger-Doll Research, Seevers, Timur-Coates, Kozeny-Godefroy, and sum-of-echoes (SOE) models so as to predict K from the NMR data; and were also used to assess the ability of the models to predict K DPP . We obtained four well-scale calibration parameter values for each model using the NMR and DPP measurements in each well; and one study-scale parameter value for each model by using all data. The SOE model achieved an agreement with K DPP that matched or exceeded that of the other models. The Timur-Coates estimates of K were found to be substantially different from K DPP . Although the well-scale parameter values for the Schlumberger-Doll, Seevers, and SOE models were found to vary by less than a factor of 2, more research is needed to confirm their general applicability so that site-specific calibration is not required to obtain accurate estimates of K from NMR logging data.

Wisconsin↗

USGS2024

Flood-inundation maps for the Cuyahoga River in and near Independence, Ohio, 2024

Digital flood-inundation maps for a 9.9-mile reach of the Cuyahoga River in and near Independence, Ohio, were created by the U.S. Geological Survey (USGS) in cooperation with the Northeast Ohio Regional Sewer District Board of Trustees. 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 (2024) stage-streamflow relation (rating curve 43.0) for the USGS streamgage 04208000, Cuyahoga River at Independence, Ohio. The resulting hydraulic model was then used to compute 13 water-surface profiles for water levels (flood stages) ranging from 14.00 to 26.00 feet. The flood stages range from “action stage” to above “major flood stage” as reported by the National Weather Service. 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 protect life and property, to coordinate flood response activities such as evacuations and road closures, and to aid postflood recovery efforts.

Ohio↗

USGS2026

Earthquake ground-motion model adjustments for the San Francisco Bay area

We develop adjustments to ergodic ground‐motion models (GMMs) to improve their performance in the San Francisco Bay Area (SFBA). GMMs are widely used in hazard assessments to estimate characteristics of ground shaking based on known properties of the source, path, and site. Such models are often developed using datasets containing records from various regions, resulting in models that represent median ground‐motion behavior, which may not adequately represent ground motions within subregions. This is true for the SFBA, where ground motions attenuate more rapidly with distance than in many other parts of California that dominate GMM databases. To support improved seismic hazard estimates in the SFBA, we calculate regional constants and anelastic attenuation coefficient adjustments relative to two commonly used ergodic GMMs: BSSA14 ( Boore et al. , 2014 ) and ASK14 ( Abrahamson et al. , 2014 ). These adjustments are obtained for a suite of ground‐motion intensity measures (peak ground acceleration, peak ground velocity, and 5%‐damped pseudospectral acceleration at oscillator periods ranging from 0.075 to 10 s) using mixed‐effects regression. Use of the regionally adjusted models reduces the overall bias by up to 0.5 natural log units for BSSA14 and up to 0.6 natural log units for ASK14. We demonstrate one application of our attenuation adjustments and their implications in an earthquake early warning case study of the 2014 M 6.0 South Napa earthquake. The predicted extent of shaking using the adjusted models better matches observed shaking at large source‐to‐site distances, especially for lower shaking intensities, thus potentially reducing overalerting. We encourage the use of our model adjustments when ergodic models are considered for seismic hazard studies in the SFBA.

California↗

USGS2026

Localization of spatiotemporally heterogeneous subsurface flows using autoencoder-based deep learning framework for time-lapse self-potential tomography

Self-potential (SP) monitoring has emerged as a valuable method for characterizing subsurface hydrogeological features and processes due to its sensitivity to fluid-induced electrokinetic effects. Despite advancements in SP inversion, challenges remain in imaging groundwater dynamics from SP activities due to complex hydrological settings and transient noise. In this study, a deep learning autoencoder (AE)-based framework is proposed for the spatiotemporal localization of subsurface fluid movement from time-lapse SP tomography. Temporal segments of time-lapse numerical inversions were first derived from long-term SP monitoring conducted from a floodplain site in Oak Ridge, Tennessee, known for active hyporheic exchange. Subsequently, AE models based on vision transformer (ViT), convolutional long short-term memory (ConvLSTM), convolutional neural network, and temporal convolutional network were individually trained and compared on the SP tomography segments for reconstruction performance. Finally, the reconstruction error over time serves as an anomaly score to identify moments of active SP variation, whereas spatial distributions of errors within these moments are analyzed to image and localize regions associated with anomalous subsurface fluid movement. The results demonstrate that ConvLSTM- and ViT-AE are most capable for the localization task with contrasting error distributions and consistent delineation of anomalies. Applying the method to both SP arrays parallel and perpendicular to the stream produced consistent anomaly zones near a fault or karst feature, validating the robustness and generalization of the approach. These results demonstrate the potential of the proposed framework as a scalable and interpretable tool for spatiotemporal analysis of subsurface flow dynamics in complex hydrogeological systems.

Tennessee↗

USGS2026

Linking distribution and return-on-investment models to optimize woody management for prairie grouse in Nebraska

Grasslands in Nebraska, USA, face threats from agricultural conversion, urban development, and woody encroachment, all of which negatively affect prairie grouse ( Tympanuchus spp.) populations. To optimize conservation planning, Nebraska wildlife agencies developed probabilistic area-based surveys for greater prairie-chicken ( T. cupido ) and sharp-tailed grouse ( T. phasianellus ) to sample landscapes across a range of environmental conditions. This design improves historical surveys and enables the development of distribution models that quantitatively define habitat associations and support scenario-based conservation planning. Using survey data collected during 2020–2022, we modeled prairie grouse occurrence and abundance as functions of land cover, topography, and climate using Bayesian logistic and zero-inflated negative binomial models with regularized horseshoe priors. We then conducted a maximum potential return-on-investment analysis of woody cover treatments, assuming sustained treatment success, relative to projected impacts of woody encroachment on prairie grouse populations by 2050. Among modeled associations were a positive association with grasslands having low woody cover and a negative association with grasslands having high woody cover. Across the 3-year period, median estimated annual populations were 142,380 for greater prairie-chicken (range of 95% CIs across years = 68,821–277,615) and 64,154 for sharp-tailed grouse (range of 95% CIs across years = 27,550–146,559). Under projected woody encroachment, mean predicted population declines were 10% for greater prairie-chicken (range of 95% CIs = 7–14%) and 6% for sharp-tailed grouse (range of 95% CIs = 5–7%). Areas with high prairie grouse density and low treatment costs in 2021, and high projected woody encroachment and population loss by 2050, offered the greatest return on investment for woody management. Return on investment was greatest in the northwestern Shortgrass Prairie ecoregion (northwestern Nebraska) for sharp-tailed grouse and the eastern Sandhills ecoregion (central Nebraska) for both species. These models underscore the value of evidence-based, quantitative approaches for prioritizing conservation actions on working lands. Scenario-based modeling could be extended to guide other treatments, such as optimizing restoration (e.g., Conservation Reserve Program) or incentivizing grassland persistence in areas with predicted climate resilience.

Nebraska↗

USGS2026

Scenarios to assess the future water availability in the Mississippi River Valley Alluvial Aquifer for the Cache River and Grand Prairie Regions of Arkansas

The U.S. Geological Survey, as part of the Arkansas Groundwater Initiative, developed forecast scenarios using previously calibrated MODFLOW 6 groundwater models that focused on the Cache and Grand Prairie Critical Groundwater Areas to assess the impact of future climate and water management strategies on the Mississippi River Valley alluvial aquifer. A Soil Water Balance model was used to forecast recharge and irrigation water use. The forecast scenario period was from January 1, 2019, through December 31, 2055, with monthly stress periods. Twenty scenarios were simulated and included seven alternate climate forecasts, five 13 general groundwater pumping reduction scenarios (round 1), and groundwater pumping reduction scenarios by crop type and for the Bayou Meto Water Management Project and Grand Prairie Area Demonstration Project (round 2). Declines in saturated thickness within the Cache Critical Groundwater Area were larger for 18 of the 20 scenarios as compared to outside of the Critical Groundwater Area. The largest average increase in saturated thickness inside the Critical Groundwater Area was 6.4 m which occurred for the round 1, 50 percent reduction scenario. Automatic reductions in groundwater pumping by MODFLOW 6 in the Cache simulation ranged from 0.02 to 13.1 percent of total groundwater pumping. For the Grand Prairie model domain, the average change in saturated thickness of the Mississippi River Valley alluvial aquifer inside the Critical Groundwater Area for the forecast period ranged between -6.6 to 1.7 m. The average saturated thickness of the Mississippi River Valley alluvial aquifer inside the Grand Prairie Critical Groundwater Area declined for 16 of the 20 scenarios. The average reduction in requested groundwater pumping for all scenarios inside the Grand Prairie Critical Groundwater Area was 25.1 percent, and the largest reduction was 46.5 percent.

ESS Open Archive↗

USGS2025

A comprehensive geologic framework of the National Crustal Model for seismic hazard studies in the conterminous United States

A three-dimensional (3D) geologic framework has been developed for the conterminous United States (U.S.) as part of the U.S. Geological Survey National Crustal Model to enhance seismic hazard modeling. The geologic framework is created from geologic maps and multiple subsurface geologic unit boundaries including the base of the Miocene, Cenozoic, Phanerozoic, and the Mohorovičić discontinuity. Modifications are made to surficial geologic maps to remove discontinuities across state and country borders. The subsurface distribution of rock type and age is extrapolated from the surface, seeded with subsurface geologic information, and constrained by a map of basement geology. The framework provides the basis for estimates of subsurface seismic velocity and density that is needed to improve estimates of earthquake ground shaking and seismic hazard. The present framework greatly expands and updates a previously published 3D geologic framework of the western part of the U.S. that was itself a first-of-its-kind digital 3D portrayal of the nation.

conterminous United States↗

USGS2025

Integrating marine historical ecology into management of Alaska’s Pacific cod fishery for climate readiness

The Pacific cod ( Gadus macrocephalus ) fishery was closed in 2020 after a rapid decline in biomass caused by the marine heat waves of 2014–2019. Pacific cod are exceptionally thermally sensitive and management of this fishery is now challenged by increasingly unpredictable climate conditions. Fisheries monitoring is critical for climate readiness, but short-term monitoring data may be inadequate for recognizing and anticipating change under rapid climate changes. We propose an interdisciplinary, marine historical ecology framework that looks to long-term records (local and traditional knowledge, history, archaeology, and paleoclimatology) to capture a long range of ecological variability and provide historical context for management. In order to connect to contemporary fisheries management, this framework must be built on a common vocabulary and an understanding of the key metrics used in fisheries stock assessments. Here, we propose metrics derived from Pacific cod stock assessment and synthesize information relevant to understanding the effects of past warming periods on cod populations across the Gulf of Alaska and Bering Sea. This case study provides a framework for thinking about how to use these historical records in the context of fisheries management under rapidly changing climate conditions.

Alaska↗

USGS2026

Groundwater tracing used to delineate recharge areas and map karst groundwater pathways for subterranean streams at Oregon Caves National Monument and Preserve

Oregon Caves National Monument and Preserve in southwestern Oregon is a 4,554-acre area managed by the National Park Service that is home to several cave systems, including Oregon Caves, which is the longest cave in Oregon, with 3.03 miles of mapped passages. Because of the interconnected nature of karst hydrologic systems, it is critical to understand the areas that can influence water quality and quantity in karst environments. Toward this goal, dye tracing was conducted by the U.S. Geological Survey from 2021 to 2024 to better understand the pathways that karst groundwater follows at Oregon Caves National Monument and Preserve and to delineate recharge areas for two caves, Oregon Caves and Cave Next Door. During the project, eight dye injections were conducted, delineating a 0.51-square-mile recharge area for Oregon Caves and a 0.69-square-mile recharge area for Cave Next Door. Additionally, the study helped to identify three resurgences associated with Oregon Caves that were previously unknown and showed that the recharge areas for the two caves were distinct from one another. The dye traces also illuminated some unique recharge characteristics of the karst at Oregon Caves, including a high variance in karst groundwater velocities, retention within the karst aquifers, and a significant diffuse-flow component.

Oregon↗

USGS2025

The 3D National Topography Model Call for Action—Part 2: The Next Generation 3D Elevation Program

The three-dimensional (3D) National Topography Model initiative to integrate elevation and hydrography data includes the next generation of hydrography data from the 3D Hydrography Program and the next generation of elevation data from the 3D Elevation Program (3DEP). The first-ever collection of light detection and ranging (lidar) data for the nation (IfSAR for Alaska) provides a critical baseline reference, and the addition of multiple repeat elevation mapping projects as part of the next generation of 3DEP would substantially expand analysis capabilities. As the U.S. Geological Survey (USGS) is closing in on our goal of complete coverage with 3DEP data available or in progress for 98.3 percent of the Nation at the end of fiscal year 2024, the USGS is already transitioning to the next generation of 3DEP. Based on the 3D Nation Study results and input from a broad range of stakeholders, the USGS National Geospatial Program has finalized a new design for 3DEP that provides increased lidar quality levels and refresh rates. The new program is designed with more flexibility to meet changing user needs and take advantage of improvements in mapping technologies. The program will aim to expand the level of interagency coordination for topobathymetric lidar acquisition for inland rivers. The next generation of 3DEP will also aim to emphasize research, including advancing program design, products, and services and engaging and leveraging the evolving 3D industry. Research goals also include becoming more flexible in meeting user needs and taking advantage of evolving remote-sensing technologies. The program also plans to move from focusing on producing standard products to producing a concept of a 3D Nation Ecosystem with a variety of inputs, products, and services.

Circular↗

USGS2021

Critical review of mercury methylation and methylmercury demethylation rate constants in aquatic sediments for biogeochemical modeling

Mercury is a toxin that causes neurological impairments in adults, is particularly harmful for fetuses and children, and is deadly in severe cases, making it a worldwide health concern. Methylmercury (MeHg) is the environmentally relevant form of mercury (Hg) because it biomagnifies along the food chain. Methylmercury is mainly produced in aquatic sediments via methylation of inorganic Hg (Hg(II)) and transformed back via demethylation. Because transformation rates determine MeHg concentrations, quantification of methylation and demethylation rates is needed to inform management of MeHg. Published rate constants for Hg(II) methylation ( 𝑘 𝑚 ) and MeHg demethylation ( 𝑘 𝑑 ) vary greatly, stemming partly from differences in experimental methods. We conducted a comprehensive review of rate laws, evaluated published rate constants, and performed biogeochemical simulations to assess variability in reported 𝑘 𝑚 and 𝑘 𝑑 . Based on selected studies employing the same pseudo-first-order rate law and similar experimental methods, we found that 𝑘 𝑚 = 0.04 ± 0.03 d −1 is a reasonable range for wetland sediments. Over a number of environments, maximum 𝑘 𝑑 was smaller at sites without Hg source ( 𝑘 𝑑 = 0.5 d −1 ) than at sites with identified Hg source ( 𝑘 𝑑 = 1.8 d −1 ). Larger variability and higher uncertainty in 𝑘 𝑑 compared to 𝑘 𝑚 highlight the need for more research on MeHg demethylation rates. This critical review: (a) aids the design of future experimental studies of 𝑘 𝑚 and 𝑘 𝑑 ; (b) provides guidance for comparing rate constants from different studies; (c) presents a biogeochemical reaction model to assess rate constants; and (d) informs selection of 𝑘 𝑚 and 𝑘 𝑑 values from the literature for use in model simulations.

Critical Reviews in Environmental Science and Tech↗

USGS2025

International data gaps at the Center for Engineering Strong Motion Data

The Center for Engineering Strong Motion Data (CESMD) is utilized by seismologists, engineers, and disaster management professionals in the US and has historically achieved and distributed waveforms from across the globe for significant earthquakes. The increased access to the waveforms via Web API (Application Programming Interface) offers a unique opportunity to provide the community complete datasets, sampling a variety of tectonic environments and geologic conditions, increasing the number of available ground motion records for use in ground motion models (GMMs) and improving the accuracy of earthquake engineering evaluations. The objective of this study is to programmatically identify gaps in global event data from the past decade and backfill missing data gaps at CESMD. We first compare the CESMD catalog with the Advanced National Seismic System (ANSS) Comprehensive Earthquake Catalog identifying regions and time periods where strong-motion data is limited or inadequate. To backfill datasets at CESMD for significant events, we pinpoint regions and time intervals that lack information, creating a list of events for which we’d like to obtain data. An important facet of this work is identifying the source of data and metadata across earthquake repositories around the world and integrating these data repositories into our current strong-motion data processing workflow. In parallel with these newly processed datasets, we are developing a script to produce data origination citations to include provenance and attribution information to associate with respective datasets at CESMD. We showcase our methodology for identifying and filling data gaps at CESMD using three case studies (the 2018 Anchorage Alaska earthquake sequence, seismicity associated with the 2018 Hawaiian Kilauea volcano eruption, and several earthquakes in Turkey) and then outline our strategy to apply our data gap backfilling methods on an international scale.

Conference Paper↗

USGS2025

Preventing overfitting when using tree-based methods for mapping hydrothermal favorability

Ensemble tree-based algorithms are robust tools for estimating sparsely distributed resources with non-linear dependencies (e.g., hydrothermal systems). These algorithms naturally accommodate the threshold conditions necessary to enable and support hydrothermal systems (e.g., having sufficient heat and permeability) and are simpler than many other non-linear machine learning strategies (e.g., artificial neural networks), which is an advantage when working with few labeled examples from which to learn. In previous work, we used eXtreme Gradient Boosting (XGBoost) to produce regional prediction and uncertainty maps of hydrothermal favorability; however, recent studies suggest that, even when properly applied, XGBoost has some risk of overfitting when there are few labeled examples from which to learn. To evaluate overfitting when constructing hydrothermal favorability maps with tree-based methods, we compare XGBoost with Extremely Randomized Trees (ExtraTrees), another ensemble tree-based algorithm that has the potential to underfit when using few labeled examples. We hold all other modeling parameters constant, resulting in two contrasting favorability maps of conventional geothermal resources for the Great Basin. Our results indicate that ExtraTrees demonstrably reduces overfitting compared with XGBoost. After considering overall performance, we conclude that ExtraTrees provides a more suitable modeling approach than XGBoost for the purposes of conventional hydrothermal resource assessments.

Conference Paper↗