Integrated geophysical and unsupervised machine learning approach for 3D geological modeling of a limestone-gypsum deposit in Algeria: insights from borehole and resistivity data

Understanding the underground distribution of geological facies is essential for modern mining operations. This study integrates 11 borehole data points (ranging from 7 to 15 m in depth) and 34 Vertical Electrical Sounding (VES) measurements to model the geological structure of the Gouamiz limestone...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Carbonates and evaporites Ročník 40; číslo 3; s. 91
Hlavní autoři: Cheikhaoui, Youcef, Cheniti, Hamza, Benmelik, Ismail, Aouissi, Hani Amir
Médium: Journal Article
Jazyk:angličtina
Vydáno: Berlin/Heidelberg Springer Berlin Heidelberg 01.09.2025
Springer Nature B.V
Témata:
Accuracy
Algorithms
Boreholes
Data points
Earth and Environmental Science
Earth Sciences
Electrical resistivity
Evaporites
Geological data
Geological structures
Geology
Geophysical methods
Gypsum
Learning algorithms
Limestone
Lithology
Machine learning
Mineral Resources
Mineralogy
Original Article
Resource management
Sedimentary facies
Seismic surveys
Supervised learning
Unsupervised learning
Algeria
Limestone
Gypsum
3D geological modeling
Predictive modeling
Resistivity
Machine learning
ISSN:0891-2556, 1878-5212
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:Understanding the underground distribution of geological facies is essential for modern mining operations. This study integrates 11 borehole data points (ranging from 7 to 15 m in depth) and 34 Vertical Electrical Sounding (VES) measurements to model the geological structure of the Gouamiz limestone/gypsum evaporite deposit in Algeria. Borehole data provided direct lithological information, while VES data contributed insights into subsurface resistivity, aiding in the identification of facies in areas lacking borehole coverage. However, VES is subject to limitations, such as reduced resolution at greater depths and potential ambiguities in layer boundaries. Advanced geophysical inversion techniques and unsupervised machine learning algorithms were applied to enhance lithological predictions, achieving approximately 70% overall model accuracy. The resulting 3D geological model delineated substantial reserves: an estimated 2.3 million m 3 of limestone and 0.981 million m 3 of gypsum, representing significant economic potential with prospective market applications in construction, cement manufacturing, and agriculture. The study demonstrates the effectiveness of integrating geophysical and geological data for accurate subsurface characterization and proposes a transferable methodological framework for similar evaporite deposits globally. Nevertheless, limitations such as spatial bias in the calibration dataset and the geological complexity of the deposit were identified. Future research should incorporate additional borehole data, employ supervised learning algorithms, and consider higher-resolution geophysical techniques such as seismic reflection to improve accuracy and robustness. These insights contribute to more informed resource management and support sustainable mining strategies.
Bibliografie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0891-2556
1878-5212
DOI:10.1007/s13146-025-01122-z