Machine learning strategies for classification and prediction of alteration facies: Examples from the Rosemont Cu-Mo-Ag skarn deposit, SE Tucson Arizona

Ordóñez-Calderón, Juan C.; Gelcich, Sergio

doi:10.1016/j.gexplo.2018.07.020

Cited by 25 publications

(12 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The flat-lying quartz-wolframite ± cassiterite ± sulfides ± carbonates veins have been exposed by underground mining over an area of 2500-m length, 400 to 2000-m width, and at least 500-m depth (Wheeler 2016). The paragenetic sequence 1,000,000 (LNEG 2010) and Carta Geológica de Portugal à escala 1:500,000 (Oliveira et al 1992); coordinate system: ETRS89/Portugal TM06 (EPSG: 3763)); (b) geological map of the Panasqueira mine area (adapted from Carta Geológica de Portugal à escala 1:500,000 (Oliveira et al 1992)); and (c) schematic cross section of the Panasqueira vein system, greisen and granite (modified after Thadeu (1979)] of the ore veins has been divided into six stages of mineralization: quartz-tourmaline alteration, muscovite selvage formation, main oxide-silicate stage, main sulfide stage, pyrrhotite alteration stage, and late carbonate stage. Tourmaline, white mica, and quartz coexist as gangue minerals from the early stages to the main sulfide stage (Kelly and Rye 1979;Polya 1989;Foxford et al 1991Foxford et al , 2000Polya et al 2000).…”

Section: The Panasqueira W-sn-cu Depositmentioning

confidence: 99%

The imprint of hydrothermal fluids on trace-element contents in white mica and tourmaline from the Panasqueira W–Sn–Cu deposit, Portugal

et al. 2020

View full text Add to dashboard Cite

White mica and tourmaline are the dominant hydrothermal alteration minerals at the world-class Panasqueira W–Sn–Cu deposit in Portugal. Thus, understanding the controls on their chemical composition helps to constrain ore formation processes at this deposit and determine their usefulness as pathfinder minerals for mineralization in general. We combine whole-rock geochemistry of altered and unaltered metasedimentary host rocks with in situ LA-ICP-MS measurements of tourmaline and white mica from the alteration halo. Principal component analysis (PCA) is used to better identify geochemical patterns and trends of hydrothermal alteration in the datasets. The hydrothermally altered metasediments are enriched in As, Sn, Cs, Li, W, F, Cu, Rb, Zn, Tl, and Pb relative to unaltered samples. In situ mineral analyses show that most of these elements preferentially partition into white mica over tourmaline (Li, Rb, Cs, Tl, W, and Sn), whereas Zn is enriched in tourmaline. White mica has distinct compositions in different settings within the deposit (greisen, vein selvages, wall rock alteration zone, late fault zone), indicating a compositional evolution with time. In contrast, tourmaline from different settings overlaps in composition, which is ascribed to a stronger dependence on host rock composition and also to the effects of chemical zoning and microinclusions affecting the LA-ICP-MS analyses. Hence, in this deposit, white mica is the better recorder of the fluid composition. The calculated trace-element contents of the Panasqueira mineralizing fluid based on the mica data and estimates of mica-fluid partition coefficients are in good agreement with previous fluid-inclusion analyses. A compilation of mica and tourmaline trace-element compositions from Panasqueira and other W–Sn deposits shows that white mica has good potential as a pathfinder mineral, with characteristically high Li, Cs, Rb, Sn, and W contents. The trace-element contents of hydrothermal tourmaline are more variable. Nevertheless, the compiled data suggest that high Sn and Li contents are distinctive for tourmaline from W–Sn deposits.

show abstract

Section: The Panasqueira W-sn-cu Depositmentioning

confidence: 99%

The imprint of hydrothermal fluids on trace-element contents in white mica and tourmaline from the Panasqueira W–Sn–Cu deposit, Portugal

et al. 2020

View full text Add to dashboard Cite

show abstract

“…The aR pckage 'rgr' is used to compute the variation matrix [50] and the hierarchical cluster dendrogram is generated with the R package 'dendextend' [51,52]. To reduce dimensionality and redundancy between correlated variables, composite variables are defined by the groupings of elements on the indicated in the cluster dendrogram [39,53]. The new matrix of composite geochemical variables is centered geometric mean for consistency with the algebraic geometry of compositions [46].…”

Section: Compositional Data Analysismentioning

confidence: 99%

“…In addition, balances (ilr coordinates) and clr coefficients of the composite variables were determined using the R package 'robCompositions' [56,57]. The ilr variables are determined using the sequential binary partition [46] suggested by the variation matrix cluster dendrogram [39,53]. In this study, the term ilr variables is interchangeable with balances and il coordinates.…”

Section: Compositional Data Analysismentioning

confidence: 99%

“…In this study, the term ilr variables is interchangeable with balances and il coordinates. Use of ilr variables as a substitute for raw compositional data in supervised learning may be appropriate because both can show comparable results [53]; ilr variables also offer the advantage of capturing relative information between elements and element groups.…”

Section: Compositional Data Analysismentioning

confidence: 99%

“…The resulting cluster boundaries establish a range of BWI in each class. Selection of K for the number of classes requires the consideration of class imbalance [53], which refers to how many observations belong to one class relative to another. Highly imbalanced classes can be detrimental to supervised learning because the model is trained on different class proportions and consequently overfit to better represented classes.…”

Section: Supervised Learning With Random Forestsmentioning

confidence: 99%

See 2 more Smart Citations

Application of Data Analytics Techniques to Establish Geometallurgical Relationships to Bond Work Index at the Paracutu Mine, Minas Gerais, Brazil

2019

Self Cite

View full text Add to dashboard Cite

Analysis of geometallurgical data is essential to building geometallurgical models that capture physical variability in the orebody and can be used for the optimization of mine planning and the prediction of milling circuit performance. However, multivariate complexity and compositional data constraints can make this analysis challenging. This study applies unsupervised and supervised learning to establish relationships between the Bond ball mill work index (BWI) and geomechanical, geophysical and geochemical variables for the Paracatu gold orebody. The regolith and fresh rock geometallurgical domains are established from two cluster sets resulting from K-means clustering of the first three principal component (PC) scores of isometric log-ratio (ilr) coordinates of geochemical data and standardized BWI, geomechanical and geophysical data. The first PC is attributed to weathering and reveals a strong relationship between BWI and rock strength and fracture intensity in the regolith. Random forest (RF) classification of BWI in the fresh rock identifies the greater importance of geochemical ilr balances relative to geomechanical and geophysical variables.

show abstract

Using Three-dimensional Modeling and Random Forests to Predict Deep Ore Potentials: A Case Study on Xiongcun Porphyry Copper–Gold Deposit in Tibet, China

Lou,

Lang,

Kang

et al. 2024

Math Geosci

View full text Add to dashboard Cite

Machine learning strategies for classification and prediction of alteration facies: Examples from the Rosemont Cu-Mo-Ag skarn deposit, SE Tucson Arizona

Cited by 25 publications

References 37 publications

The imprint of hydrothermal fluids on trace-element contents in white mica and tourmaline from the Panasqueira W–Sn–Cu deposit, Portugal

The imprint of hydrothermal fluids on trace-element contents in white mica and tourmaline from the Panasqueira W–Sn–Cu deposit, Portugal

Application of Data Analytics Techniques to Establish Geometallurgical Relationships to Bond Work Index at the Paracutu Mine, Minas Gerais, Brazil

Using Three-dimensional Modeling and Random Forests to Predict Deep Ore Potentials: A Case Study on Xiongcun Porphyry Copper–Gold Deposit in Tibet, China

Contact Info

Product

Resources

About