Search USGS⌕ Search

SEARCH · Search USGS

Results for “SELECT Journal”

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.

At least 1,135 records · Page 63Linked to original sources
USGS1992

Selenium and sulfur relationships in alfalfa and soil under field conditions, San Joaquin Valley, California

Relationships between total Se and S or soluble SeO 4 and SO 4 in soils and tissue concentrations in alfalfa ( Medicago sativa L.), under field conditions in the San Joaquin Valley of California, suggest that the rate of accumulation of Se in alfalfa may be reduced in areas where high Se and S concentrations in soils were measured. These data suggest that the balance between carbonate and sulfate minerals in soil may have a greater influence on uptake of Se by alfalfa than does the balance of SeO 4 and SO 4 in soil solution. Soil and alfalfa were sampled from areas representing a wide range in soil Se and S concentrations. Specific sampling locations were selected based on a previous study of Se, S, and other elements where 721 soil samples were collected to map landscape variability and distribution of elements. Six multiple-linear regression equations were developed between total and/or soluble soil chemical constituents and tissue concentrations of Se in alfalfa. We chose a regression model that accounted for 72% of the variability in alfalfa Se concentrations based on an association of elements in soil (total C, S, Se, and Sr) determined by factor analysis. To prepare a map showing the spatial distribution of estimated alfalfa Se concentrations, the model was applied to the data from the previously collected 721 soil samples. Estimated alfalfa Se concentrations in most of the study area were within a range that is predicted to produce alfalfa with neither Se deficiency nor toxicity when consumed by livestock. A few small areas are predicted to produce alfalfa that potentially would not meet minimum dietary needs of livestock.

Journal of Environmental Quality↗
USGS2005

Modeling tritium transport through a deep unsaturated zone in an arid environment

Understanding transport of tritium ( 3 H) in unsaturated zones is critical to evaluating options for waste isolation. Tritium typically is a large component of low-level radioactive waste (LLRW). Studies at the U.S. Geological Survey's Amargosa Desert Research Site (ADRS) in Nevada investigate 3 H transport from a closed LLRW facility. Two boreholes are 100 and 160 m from the nearest waste trench and extend to the water table at 110 m. Soil-water vapor samples from the deep boreholes show elevated levels of 3 H at all depths. The objectives of this study were to (i) test source thermal and gas-advection mechanisms driving 3 H transport and (ii) evaluate model sensitivity to these mechanisms and to selected physical and hydraulic properties including porosity, tortuosity, and anisotropy. A two-dimensional numerical model incorporated a non-isothermal, heterogeneous domain of the unsaturated zone and instantaneous isotopic equilibrium. The TOUGH2 code was used; however, it required modification to account for temperature dependence of both the Henry's law equilibrium constant and isotopic fractionation with respect to tritiated water. Increases in source temperature, pressure, and porosity enhanced 3 H migration, but failed to match measured 3 H distributions. All anisotropic simulations with a source pressure component resembled, in shape, the upper portion of the 3 H distribution of the nearest borehole. Isotopic equilibrium limited migration of 3 H, while effects of radioactive decay were negligible. A 500 Pa pressure increase above ambient pressure in conjunction with a high degree of anisotropy (1:100) was necessary for simulated 3 H transport to reach the nearest borehole.

Vadose Zone Journal↗
USGS1979

A preliminary annotated bibliography on siliceous algal cysts and scales

As part of an investigation of the stratigraphic and environmental distributions of siliceous algal resting cysts, we have assembled this bibliography. The emphasis is on papers that illustrate these cysts, and our annotation concentrates on the nature of the illustrations. In some cases, we have also listed taxonomic names of the species illustrated in the paper, as well as those of any new species described, but this part of the annotation is not yet complete. The taxonomic names given here are always those given by the original author. The references have been selected by starting with a few papers and extracting relevant titles from their reference lists. We have included not only relevant works, but also a few that sounded relevant but turned out not to be. This bibliography is, we believe, fairly complete for references that describe fossil occurrences of cysts, but much less thorough for the taxonomic literature. The list of references is still evolving, and we welcome additions or corrections. We have checked and verified many of the references listed here; those that we have not yet seen are indicated by an asterisk preceding the authorship. We have also indicated the source of each reference which we have not seen. Most such references were found in the reference lists of other papers listed here. Other references have been taken from the VanLandingham bibliography, which is an unpublished manuscript on file at the Diatom Library of the California Academy of Sciences in San Francisco. We wish to thank Mrs. Margaret Hanna, curator of the Diatom Library at the California Academy of Sciences (San Francisco) and Mrs. Jacquelyn Freeberg, of the USGS Menlo Park Library, for help in locating many of the references, which are often tantalizingly obscure. Some of the references are incomplete; they have been included anyway in order to preserve as much information as possible. For the same reason, the bibliographic style of the references is not consistent throughout; we prefer to include such information as the authors' full names and issue numbers of journals when we have it available, rather than discard good information about one paper just because we lack similar information about another.

Open-File Report↗
USGS2016

Using GPS telemetry to determine roadways most susceptible to deer-vehicle collisions

More than 1 million wildlife-vehicle collisions occur annually in the United States. The majority of these accidents involve white-tailed deer (Odocoileus virginianus) and result in >US $4.6 billion in damage and >200 human fatalities. Prior research has used collision locations to assess sitespecific as well as landscape features that contribute to risk of deer-vehicle collisions. As an alternative approach, we calculated road-crossing locations from 25 GPS-instrumented white-tailed deer near Madison, Georgia (n=154,131 hourly locations). We identified crossing locations by creating movement paths between subsequent GPS points and then intersecting the paths with road locations. Using AIC model selection, we determined whether 10 local and landscape variables were successful at identifying areas where higher frequencies of deer crossings were likely to occur. Our findings indicate that traffic volume, distance to riparian areas, and the amount of forested area influenced the frequency of road crossings. Roadways that were predominately located in wooded landscapes and 200–300 m from riparian areas were crossed frequently. Additionally, we found that areas of low traffic volume (e.g., county roads) had the highest frequencies of deer crossings. Analyses utilizing only records of deer-vehicle collision locations cannot separate the relative contribution of deer crossing rates and traffic volume. Increased frequency of road crossings by deer in low-traffic, forested areas may lead to a greater risk of deer-vehicle collision than suggested by evaluations of deer-vehicle collision frequency alone.

Georgia↗
USGS2018

Relating river discharge and water temperature to the recruitment of age‐0 White Sturgeon (Acipenser transmontanus Richardson, 1836) in the Columbia River using over‐dispersed catch data

The goals were to (i) determine if river discharge and water temperature during various early life history stages were predictors of age‐0 White Sturgeon, Acipenser transmontanus , recruitment, and (ii) provide an example of how over‐dispersed catch data, including data with many zero observations, can be used to better understand the effects of regulated rivers on the productivity of depressed sturgeon populations. An information theoretic approach was used to develop and select negative binomial and zero‐inflated negative binomial models that model the relation of age‐0 White Sturgeon survey data from three contiguous Columbia River reservoirs to river discharge and water temperature during spawning, egg incubation, larval, and post‐larval phases. Age‐0 White Sturgeon were collected with small mesh gill nets in The Dalles and John Day reservoirs from 1997 to 2014 and a bottom trawl in Bonneville Reservoir from 1989 to 2006. Results suggest that seasonal river discharge was positively correlated with age‐0 recruitment; notably that discharge, 16 June–31 July was positively correlated to age‐0 recruitment in all three reservoirs. The best approximating models for two of the three reservoirs also suggest that seasonal water temperature may be a determinant of age‐0 recruitment. Our research demonstrates how over‐dispersed catch data can be used to better understand the effects of environmental conditions on sturgeon populations caused by the construction and operation of dams.

Columbia River↗
USGS2016

Spawning site fidelity of wild and hatchery lake trout ( Salvelinus namaycush ) in northern Lake Huron

Fidelity to high-quality spawning sites helps ensure that adults repeatedly spawn at sites that maximize reproductive success. Fidelity is also an important behavioural characteristic to consider when hatchery-reared individuals are stocked for species restoration, because artificial rearing environments may interfere with cues that guide appropriate spawning site selection. Acoustic telemetry was used in conjunction with Cormack–Jolly–Seber capture–recapture models to compare degree of spawning site fidelity of wild and hatchery-reared lake trout ( Salvelinus namaycush ) in northern Lake Huron. Annual survival was estimated to be between 77% and 81% and did not differ among wild and hatchery males and females. Site fidelity estimates were high in both wild and hatchery-reared lake trout (ranging from 0.78 to 0.94, depending on group and time filter), but were slightly lower in hatchery-reared fish than in wild fish. The ecological implication of the small difference in site fidelity between wild and hatchery-reared lake trout is unclear, but similarities in estimates suggest that many hatchery-reared fish use similar spawning sites to wild fish and that most return to those sites annually for spawning.

Canadian Journal of Fisheries and Aquatic Sciences↗
USGS1970

Soil-food chain-pesticide wildlife relationships in aldrin-treated fields

Soil-food-chain-pesticide wildlife relationships were investigated to learn the concentration of pesticide residues present in soils, macro-invertebrates, vertebrates, and seeds as a result of annual applications of aldrin at recommended rates for pest control. Two central Missouri cornfields treated with aldrin at 1 lb/acre, for 16 and 15 of the past 17 years, were selected for study during 1965-67. Primary samples collected for residue analyses included soils, earthworms (Lumbricidae), crickets (Gryllidae), and two kinds of ground beetles (Carabidae) obtained during early April, June, August, and October. Vertebrates and plant seeds collected during 1967 included white-footed mice ( Peromyscus maniculatus ), toads ( Bufo americanus ), snakes ( Thamnophis sirtalis and Pituophis sayi ), corn ( Zea Mays ), foxtail ( Setaria Faberii ), and annual sunflower ( Helianthus annuus ). Pesticide residues consisted primarily of dieldrin, the degradation product of aldrin. Combined aldrin and dieldrin residues, as two field all-season averages, wet weight basis, were: soils, 0.31 ppm; earthworms, 1.49 ppm; crickets, 0.23 ppm; Harpalus ground beetles, 1.10 ppm; Poecilus ground beetles, 9.67 ppm; white-footed mice, 0.98 ppm; toads, 3.53 ppm; garter snakes, 12.35 ppm; and corn, foxtail, and sunflower seeds less than 0.02 ppm each. Unusually high average residues (37.48 ppm) in Poecilus beetles during June, 1967, were attributed to abnormally high soil moisture and predacious feeding habits of these insects.

Missouri↗
USGS2012

Parasites of fishes in the Colorado River and selected tributaries in Grand Canyon, Arizona.

As part of the endangered humpback chub (HBC; Gila cypha) Adaptive Management Program, a parasite survey was conducted from 28 June to 17 July 2006 in 8 tributaries and 7 adjacent sections of the main stem of the Colorado River, U.S.A. In total, 717 fish were caught, including 24 HBC. Field necropsies yielded 19 parasite species, 5 of which (Achtheres sp., Kathlaniidae gen. sp., Caryophyllaidae gen. sp., Myxidium sp., and Octomacrum sp.) are new records for Grand Canyon, Arizona, U.S.A. Spearman's correlation coefficient analyses showed no correlations between parasite burden and fork length for various combinations of fish and parasite species. Regression analyses suggest that no parasite species had a strong effect on fish length. The most diverse parasite community (n=14) was at river kilometer (Rkm) 230, near the confluence of Kanab Creek. The most diverse parasite infracommunity (n=12) was found in the non-native channel catfish (CCF; Ictaluris punctatus). Overall parasite prevalence was highest in CCF (85%) followed by that in HBC (58%). The parasite fauna of humpback chub was mainly composed of Bothriocephalus acheilognathi and Ornithodiplostomum sp. metacercariae.

Arizona↗
USGS2025

Predicting pup-rearing habitat for Mexican wolves

Population monitoring is essential to document recovery efforts for threatened and endangered species. Mexican wolves ( Canis lupus baileyi ) are an endangered subspecies of gray wolves that historically occupied large portions of the American Southwest and Mexico. Recently, the Mexican wolf population in the United States has been growing rapidly and traditional approaches for population monitoring (e.g., capture and radio collaring) are becoming difficult and expensive as wolves expand into new areas. We developed predictive models of pup-rearing habitat (i.e., den and rendezvous sites) that could help guide future population monitoring efforts. We located 255 den sites and 129 rendezvous sites in Arizona and New Mexico, USA (1998–2023) using tracking collars and site visits. We sampled habitat conditions in wolf-occupied regions of Arizona and New Mexico and fit logistic regressions to these data following a use–available study design to estimate resource selection functions (RSF) for den and rendezvous sites. We hypothesized wolves would select areas that offered greater physical protection, lower human-disturbance, and access to reliable water sources for pup-rearing but that the relative importance of these features would differ between the denning and rendezvous site seasons. Mexican wolves selected den sites at higher elevations in steeper and rougher terrain that were closer to permanent waterbodies but farther from rural roads. Selection of rendezvous sites was also associated with higher elevations and proximity to waterbodies but varied with availability of green leaf biomass on the landscape. While still highly predictive, our rendezvous site model was less predictive than our den model (Spearman's correlation averaged 0.81 [SE = 0.05] vs. 0.90 [SE = 0.03], respectively), possibly because water and green leaf biomass are more spatially diffuse and variable because of monsoonal rains during the rendezvous site season. Our results suggest that terrain features associated with physical protection and access to reliable water were most important in characterizing suitable pup-rearing habitat for Mexican wolves. By predicting suitable den and rendezvous site habitat across portions of the Mexican Wolf Experimental Population Area, our models can help guide future population monitoring by reducing the total search area when surveying for wolves and increase the probability of detecting all members of a pack.

Arizona, New Mexico↗
USGS2007

An automated approach to detecting signals in electroantennogram data

Coupled gas chromatography/electroantennographic detection (GC-EAD) is a widely used method for identifying insect olfactory stimulants present in mixtures of volatiles, and it can greatly accelerate the identification of insect semiochemicals. In GC-EAD, voltage changes across an insect's antenna are measured while the antenna is exposed to compounds eluting from a gas chromatograph. The antenna thus serves as a selective GC detector whose output can be compared to that of a "general" GC detector, commonly a flame ionization detector. Appropriate interpretation of GC-EAD results requires that olfaction-related voltage changes in the antenna be distinguishable from background noise that arises inevitably from antennal preparations and the GC-EAD-associated hardware. In this paper, we describe and compare mathematical algorithms for discriminating olfaction-generated signals in an EAD trace from background noise. The algorithms amplify signals by recognizing their characteristic shape and wavelength while suppressing unstructured noise. We have found these algorithms to be both powerful and highly discriminatory even when applied to noisy traces where the signals would be difficult to discriminate by eye. This new methodology removes operator bias as a factor in signal identification, can improve realized sensitivity of the EAD system, and reduces the number of runs required to confirm the identity of an olfactory stimulant. ?? 2007 Springer Science+Business Media, LLC.

Journal of Chemical Ecology↗
USGS1994

Use of sentinel mallards for epizootiologic studies of avian botulism

Captive-reared mallards (Anas platyrhynchos) were used as sentinels to study the epizootiology of avian botulism at the Sacramento National Wildlife Refuge, Willows, California (USA) from 1986 to 1989. Sentinel mallards were wing-clipped, and 40 to 50 birds were confined in 1.6-ha enclosures in 11 selected wetlands (pools). Enclosures were searched intensively three to four times weekly from July through October. Sick and dead wild and sentinel birds were collected, necropsied, and tested for type C botulism toxin. Botulism epizootics occurred in sentinel mallards in 1986, 1987, and 1989, but only a few isolated cases of botulism were detected in 1988. In most epizootics, botulism also was detected simultaneously in wild birds using the same pool outside the enclosure. Epizootics in sentinels were initiated and perpetuated in the absence of vertebrate carcasses. A sex-specific trend in the probability of intoxication was detected, with males contracting botulism at a higher rate than females. Daily mortality rates of sentinels during botulism epizootics ranged from 0.0006 to 0.0600, with a mean of 0.0190. These rates would result in the daily loss of 0.6 to 60 birds per thousand at risk. The use of sentinel birds provided an effective means of gathering site-specific epizootiologic data.

California↗
USGS1989

Analysis of the shallow groundwater flow system near Connetqout Brook, Long Island, New York

Streamflow on Long Island is derived principally from shallow groundwater that flows above the deeper regional flow system. The movement of shallow groundwater was studied during 1975-1982 at Connetquot Brook - an undisturbed stream in Connetquot River State Park - in south-central Long Island. The investigation encompassed: (1) field studies of streamflow, groundwater levels, and age of water as indicated by tritium concentrations, and (2) numerical simulation of the shallow flow system to evaluate the hydraulic factors that influence groundwater flow near and beneath the stream. Analysis of water-level data indicates that groundwater flow is essentially horizontal throughout the drainage basin except near and beneath the stream, where it moves upward diagonally and discharges into the streambank at three sites were 1-2 ft higher than stream stage in the and in wells driven into the streambank at three sites were 1-2 ft higher than stream stage in the first 5 ft of penetration. Increases in head, which were detected to depths of 30 ft beneath the streambed, indicate upward movement of water above that depth. Water samples from selected wells were analyzed for tritium concentration to determine the relative age of water to locate the bottom boundary of the shallow flow system. Tritium concentrations indicate that the lower boundary is from 45 to 100 ft below the water table. A two-dimensional cross-sectional flow model of the shallow flow system indicated that: (1) stream width and streambed hydraulic conductivity influence heads mostly within about 50 ft of the stream; (2) the thickness of the shallow flow system influences heads more distant from the stream but has a negligible effect near the stream; and (3) the quantity of water entering the system as recharge from precipitation influences the heads throughout the area. Field measurements of hydraulic head indicate the shallow flow system to extend to about 30 ft below the stream channel. Results of the sensitivity analysis indicate that the thickness of the shallow system has a negligible effect on head distribution beneath the stream.

New York↗
USGS2023

Using DC resistivity ring array surveys to resolve conductive structures around tunnels or mine-workings

In underground environments, conventional direct current (DC) resistivity surveys with a single linear array of electrodes produce fundamentally non-unique inversions. These non-uniqueness and model resolution issues stem from limitations placed on the location of transmitters (TXs) and receivers (RXs) by the geometry of existing tunnels and boreholes. Poor excitation and/or sampling of the region of interest (ROI) can create artifacts and reduce the resolution of the recovered model. To address these problems we propose the use of an ensemble of ring arrays, which are created by placing one or more electrodes in each face (sidewalls, floor, and ceiling) of the tunnel to form a ring of electrodes at each along-tunnel location. Using a series of increasingly complex synthetic models, we assess the benefits of ring arrays and show that they can be used to better constrain the location and shape of anomalous bodies around the tunnel. Although ring arrays significantly improve the resolution of the recovered model, the size of the comprehensive ring array survey increases rapidly with the number of electrodes used. To balance model resolution and survey size, we developed a physics-based survey design methodology. In this methodology, TXs are selected based upon secondary charge accumulations on a test block that is moved throughout the ROI. Although this survey design methodology does not produce a strictly optimal survey, it balances model resolution and survey size in a practical and computationally efficient manner. Since the ring array more accurately estimates the around-tunnel location of targets and ensures that targets on all sides of the tunnel are detected, it is ideally suited to tunnel-based environments. Our results show that only about 6 % "> 6% of the possible TXs and 0.5 % "> 0.5% of the RXs in the comprehensive ring array survey are needed to retain the improvements in resolution. Therefore, economical ring array surveys can be designed for both reconnaissance and target characterization. Following the inversion of the reconnaissance dataset, additional rings can be added to reduce the inter-ring spacing or off-tunnel boreholes can be added to the region around identified anomalies to increase resolution as required.

Journal of Applied Geophysics↗
USGS1995

Lake trout spawning habitat in the Great Lakes - a review of current knowledge

We review existing information on lake trout spawning habitat, which might indicate whether habitat is now a limiting factor in lake trout reproductive success. Lake trout spawning habitat quality is defined by the presence or absence of olfactory cues for homing, reef location with respect to the shoreline, water depth, proximity to nursery areas, reef size, contour, substrate size and shape, depth of interstitial spaces, water temperature at spawning time, water quality in interstitial spaces, and the presence of egg and fry predators. Data on factors which attracted native spawners to spawning reefs are lacking, due to the absence of historic data on egg deposition. No direct evidence of egg deposition has been collected from sites deeper than 18 m. Interstitial space and, therefore, substrate size and shape, appear to be critical for both site selection by adults and protection of eggs and fry. Water quality is clearly important for egg incubation, but the critical parameters which define water quality have not yet been well determined in the field. Exposure to wave energy, dictated in part by reef location, may maintain high water quality but may also damage or dislodge eggs. The importance of olfactory cues, water temperature, and proximity to nursery habitat to spawning trout is unclear. Limited data suggest that egg and fry predators, particularly exotic species, may critically affect fry production and survival. Although availability of physical spawning habitat is probably not limiting lake trout reproduction, changes in water quality and species composition may negatively affect early life stages. This review of habitat factors that affect early life stages of lake trout suggests several priorities for research and management.

Journal of Great Lakes Research↗
USGS2014

Reach-scale comparison of habitat and mollusk assemblages for select sites in the Clinch River with regional context

Several hypotheses, including habitat degradation and variation in fluvial geomorphology, have been posed to explain extreme spatial and temporal variation in Clinch River mollusk assemblages. We examined associations between mollusk assemblage metrics (richness, abundance, recruitment) and physical habitat (geomorphology, streambed composition, fish habitat, and riparian condition) at 10 sites selected to represent the range of current assemblage condition in the Clinch River. We compared similar geomorphological units among reaches, employing semi-quantitative and quantitative protocols to characterize mollusk assemblages and a mix of visual assessments and empirical measurements to characterize physical habitat. We found little to no evidence that current assemblage condition was associated with 54 analyzed habitat metrics. When compared to other sites in the Upper Tennessee River Basin (UTRB) that once supported or currently support mollusk assemblages, Clinch River sites were more similar to each other, representing a narrower range of conditions than observed across the larger geographic extent of the UTRB. A post-hoc analysis suggested stream size and average boundary shear stress at bankfull stage may have historically limited species richness in the UTRB (p < 0.001). Associations between mollusk assemblages and physical habitat in the UTRB and Clinch River currently appear obscured by other factors limiting richness, abundance, and recruitment.

Tennessee, Virginia↗
USGS2004

Simulating the dynamics of linear forests in Great Plains agroecosystems under changing climates

Most forest growth models are not suitable for the highly fragmented, linear (or linearly shaped) forests in the Great Plains agroecosystems (e.g., windbreaks, riparian forest buffers), where such forests are a minor but ecologically important component of the land mosaics. This study used SEEDSCAPE, a recently modified gap model designed for cultivated land mosaics in the Great Plains, to simulate the effects of climate change on the dynamics of such linear forests. We simulated the dynamics of windbreaks with different initial planting species richness and widths (light changes as the selected resulting factor) using current climate data and nested regional circulation models (RegCMs). Results indicated that (i) it took 70-80 simulation years for the linear forests to reach a steady state under both normal (present-day) and warming climates; (ii) warming climates would reduce total aboveground tree biomass and the spatial variation in biomass, but increase dominance in the linear forests, especially in the upland forests; (iii) linear forests with higher planting species richness and smaller width produced higher aboveground tree biomass per unit area; and (iv) the same species performed very differently with different climate scenarios, initial planting diversity, and forest widths. Although the model still needs further improvements (e.g., the effects of understory species should be included), the model can serve as a useful tool in modeling the succession of linear forests in human-dominated land mosaics under changing climates and may also have significant practical implications in other systems. ?? 2004 NRC Canada.

Canadian Journal of Forest Research↗
USGS2026

Morphometric properties of the CP-21 landing site on the Moon at Mons Gruithuisen Gamma

Characterizing terrain surface properties is an essential step in assessing the feasibility of landing successfully at a location on a planetary surface. Slopes and terrain ruggedness index (TRI) values derived from high-resolution (2 m pixel −1 ) digital terrain models provided important constraints in selecting the landing site for the upcoming Payloads and Research Investigations on the Surface of the Moon program as part of the Commercial Lunar Payload Services task order CP-21 mission. The selected landing site needed to balance safety requirements with the ability to achieve the science and exploration goals of the Lunar Vulkan Imaging and Spectroscopy Explorer payload. In this study, we compare several morphometric parameters in the context of the CP-21 landing site on Mons Gruithuisen Gamma, or the Gamma dome, and quantify the information they convey about lunar surface properties to assess their utility for future landing site evaluation. TRI was found to be a useful metric for assessing landing site safety. Metrics that better decouple slope and surface roughness, the vector ruggedness measure and the standard deviation of slope, provided additional information about surface characteristics and textures such as the degree to which roughness is isotropic.

Planetary Science Journal↗
USGS2022

Machine learning and data augmentation approach for identification of rare earth element potential in Indiana Coals, USA

Rare earth elements and yttrium (REYs) are critical elements and valuable commodities due to their limited availability and high demand in a wide range of applications and especially in high-technology products. The increased demand and geopolitical pressures motivate the search for alternative sources of REYs, and coal, coal waste, and coal ash are considered as new sources for these critical elements. This research evaluates the REY potential of coals from Indiana (USA). However, although coal data revealed REY potential, it suffered from sparse samples with complete REY measurements. Therefore, we explore the applicability of machine learning (ML) models and data augmentation techniques to demonstrate their applicability to evaluate REY potential in Indiana, and other areas in coal basins, using selected coal parameters (Al2O3, Fe2O3, C, Ash, S, P, Mo, Zn, and As contents) as covariates (indicators). Due to the relatively small sample size with complete REY data in the Indiana Coal Database, two data augmentation techniques (Random Over-Sampling Examples and Synthetic Minority Over-Sampling Technique) were used. Four machine learning algorithms (linear discriminate analysis, support vector machine, random forest, and artificial neural networks) were applied for modeling REY potential as a classification problem. The results show that application of Synthetic Minority Over-Sampling Technique prior to development of the support vector machine (SVM) models generated the best REY classification with an accuracy of 95%. The encouraging results based on Indiana coal data may suggest that a similar approach can be used for other coal basins for screening the locations with REY potential. Those locations then can be targeted for more detailed geochemical surveys to identify most promising areas and evaluate overall REY resources.

Indiana↗