Scholarlink

<p>Drill targeting is the most consequential step in mineral exploration, yet false‑positive geophysical anomalies have become a bottleneck in exploration projects. We develop a new data science method that ranks potential target areas based on their geophysical similarity to known economic deposits. Comprehensive multivariate geophysical (magnetic, gravity and radiometric) and drilling datasets from Brazil’s Curaçá Valley Belt are utilized for this purpose. Through multivariate geophysical data analysis of 15 drilled prospects, we reveal that the sensitivity of geophysical variables to mineralization can differ significantly between deposits. We use binary labels of intrusion and non-intrusion classes for this purpose. We also show that different deposits can have geophysical signatures that make them dissimilar in high-dimensional feature space. These observations clarify the reasons behind high failure rates and highlight that merging all deposit samples into a single dataset and training a generalized machine learning model is unreliable across a broad exploration area. Our approach mitigates these issues by considering an economic deposit as the reference, using the reference deposit to identify the most sensitive geophysical features based on the geometry of mineralization, defining the high-dimensional spatial boundary of the known mineralized body, and ranking all potential areas by their distance to the reference deposit. Mineralization patterns in the top-ranked areas are then mapped using a logistic regression classifier trained on the reference deposit. When benchmarked against known mineralization, this workflow helped avoid misclassification, and identified targets validated through drilling data, for the three reference deposits used.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Exploration Drill Targeting based on Similarity with Known Economical Mineral Deposits

  • 期刊论文

  • 发表日期：

    2026年4月7日

查看全文
  • Adel Asadi,
  • Jef Karel Caers,
  • Pablo Mejia-Herrera,
  • David Zhen Yin

摘要

Drill targeting is the most consequential step in mineral exploration, yet false‑positive geophysical anomalies have become a bottleneck in exploration projects. We develop a new data science method that ranks potential target areas based on their geophysical similarity to known economic deposits. Comprehensive multivariate geophysical (magnetic, gravity and radiometric) and drilling datasets from Brazil’s Curaçá Valley Belt are utilized for this purpose. Through multivariate geophysical data analysis of 15 drilled prospects, we reveal that the sensitivity of geophysical variables to mineralization can differ significantly between deposits. We use binary labels of intrusion and non-intrusion classes for this purpose. We also show that different deposits can have geophysical signatures that make them dissimilar in high-dimensional feature space. These observations clarify the reasons behind high failure rates and highlight that merging all deposit samples into a single dataset and training a generalized machine learning model is unreliable across a broad exploration area. Our approach mitigates these issues by considering an economic deposit as the reference, using the reference deposit to identify the most sensitive geophysical features based on the geometry of mineralization, defining the high-dimensional spatial boundary of the known mineralized body, and ranking all potential areas by their distance to the reference deposit. Mineralization patterns in the top-ranked areas are then mapped using a logistic regression classifier trained on the reference deposit. When benchmarked against known mineralization, this workflow helped avoid misclassification, and identified targets validated through drilling data, for the three reference deposits used.