Scholarlink

<p>This research develops Machine Learning models using Random Forest (RF), Support Vector Machine (SVM), and satellite-based features extracted from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) data to map active petroleum seepages in the Aghajari oil field, Zagros Fold-Thrust Belt (ZFTB), SW Iran. The sulfate-rich Gachsaran Formation (primary evaporite cap rock) has undergone significant alteration as a result of hydrocarbon migration. However, little is known about how petroleum seepages influence evaporite rocks and how to use satellite data to detect them. The spectral and chemical properties of altered and unaltered rock units of the Gachsaran Formation, were analyzed using Reflectance spectroscopy, optical microscopy, and X-ray diffraction (XRD) analysis. The Boruta algorithm, Pearson’s correlation coefficient, and variable importance plots were utilized to identify key satellite-based features, evaluate their linear correlations, and ascertain their significance. The ML models were trained by randomly splitting field samples into two groups (70% for training and 30% for model testing), and their accuracy were assessed using the confusion matrix, performance evaluation metrics and baseline comparison. Spectroscopic analysis revealed meaningful changes in the spectral properties of evaporitic rocks between altered and unaltered units. The results of petrographic studies disclosed the presence of solid residue of organic compounds in the altered samples. XRD analysis indicated increase of anhydrite and bassanite in altered rocks. The RF algorithm outperformed SVM in detecting petroleum seepages based on the confusion matrix and performance evaluation metrics. According to the results, PCA6, BR2/1, BR8/5, BR8/4, BR8/6, ICA3, and PCA7 were the main features, and PCA was the most important data set. The ML models can be utilized to detect petroleum seepages and observe environmental effects.</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

Evaluating machine learning algorithms and satellite-based features to spatially detect active petroleum seepages in the Aghajari oil field, SW Iran

  • 期刊论文

  • 发表日期：

    2026年1月7日

查看全文
  • Mohammad H. Tayebi,
  • Diego Fernando Ducart,
  • Carlos Roberto de Souza Filho,
  • Saeid Asadzadeh,
  • Majid H. Tangestani,
  • Soheyla Esmaeili

摘要

This research develops Machine Learning models using Random Forest (RF), Support Vector Machine (SVM), and satellite-based features extracted from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) data to map active petroleum seepages in the Aghajari oil field, Zagros Fold-Thrust Belt (ZFTB), SW Iran. The sulfate-rich Gachsaran Formation (primary evaporite cap rock) has undergone significant alteration as a result of hydrocarbon migration. However, little is known about how petroleum seepages influence evaporite rocks and how to use satellite data to detect them. The spectral and chemical properties of altered and unaltered rock units of the Gachsaran Formation, were analyzed using Reflectance spectroscopy, optical microscopy, and X-ray diffraction (XRD) analysis. The Boruta algorithm, Pearson’s correlation coefficient, and variable importance plots were utilized to identify key satellite-based features, evaluate their linear correlations, and ascertain their significance. The ML models were trained by randomly splitting field samples into two groups (70% for training and 30% for model testing), and their accuracy were assessed using the confusion matrix, performance evaluation metrics and baseline comparison. Spectroscopic analysis revealed meaningful changes in the spectral properties of evaporitic rocks between altered and unaltered units. The results of petrographic studies disclosed the presence of solid residue of organic compounds in the altered samples. XRD analysis indicated increase of anhydrite and bassanite in altered rocks. The RF algorithm outperformed SVM in detecting petroleum seepages based on the confusion matrix and performance evaluation metrics. According to the results, PCA6, BR2/1, BR8/5, BR8/4, BR8/6, ICA3, and PCA7 were the main features, and PCA was the most important data set. The ML models can be utilized to detect petroleum seepages and observe environmental effects.