Scholarlink

<p>Accurate identification of soil parent materials is essential for soil classification, mapping, and land management. This study evaluated the performance of three machine learning algorithms, Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Light Gradient Boosting Machine (LightGBM), in predicting soil parent material classes using data obtained from Visible and Near-Infrared Reflectance Spectroscopy (VNIRS), X-ray Fluorescence (XRF), and Inductively Coupled Plasma (ICP) analyses. A total of 59 surface soil samples were collected from four dominant parent materials (Mudflow, Limestone, Marl, and Basalt) in Şanlıurfa Province, Türkiye. Samples were analyzed for spectral and elemental properties, and predictive models were trained and optimized using Grid Search Cross-Validation. Model performance was evaluated using standard classification metrics including accuracy, precision, recall, and F1-score. The results showed that ICP data produced the most accurate predictions across all models, with LightGBM achieving the highest overall performance (accuracy = 0.81, F1-score = 0.81). XRF-based models also yielded strong performance, particularly when combined with RF. In contrast, VNIRS-based models demonstrated relatively low prediction accuracy, with substantial misclassification observed between certain parent material classes. Among the algorithms, RF and LightGBM consistently outperformed XGBoost across two of the three datasets. Overall, the findings highlight the effectiveness of ensemble machine learning methods in predicting soil parent materials and underscore the importance of data source selection. ICP and XRF data, when paired with robust algorithms like RF and LightGBM, offer promising tools for improving soil classification and digital soil mapping.</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

Predicting soil parent materials using machine learning models and spectral–elemental data (VNIRS, XRF, ICP)

  • 期刊论文

  • 发表日期：

    2026年7月1日

查看全文
  • Yüsra İnci,
  • Miraç Kılıç,
  • Ali Volkan Bilgili,
  • Süreyya Betül Rufaioğlu,
  • Recep Gündoğan

摘要

Accurate identification of soil parent materials is essential for soil classification, mapping, and land management. This study evaluated the performance of three machine learning algorithms, Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Light Gradient Boosting Machine (LightGBM), in predicting soil parent material classes using data obtained from Visible and Near-Infrared Reflectance Spectroscopy (VNIRS), X-ray Fluorescence (XRF), and Inductively Coupled Plasma (ICP) analyses. A total of 59 surface soil samples were collected from four dominant parent materials (Mudflow, Limestone, Marl, and Basalt) in Şanlıurfa Province, Türkiye. Samples were analyzed for spectral and elemental properties, and predictive models were trained and optimized using Grid Search Cross-Validation. Model performance was evaluated using standard classification metrics including accuracy, precision, recall, and F1-score. The results showed that ICP data produced the most accurate predictions across all models, with LightGBM achieving the highest overall performance (accuracy = 0.81, F1-score = 0.81). XRF-based models also yielded strong performance, particularly when combined with RF. In contrast, VNIRS-based models demonstrated relatively low prediction accuracy, with substantial misclassification observed between certain parent material classes. Among the algorithms, RF and LightGBM consistently outperformed XGBoost across two of the three datasets. Overall, the findings highlight the effectiveness of ensemble machine learning methods in predicting soil parent materials and underscore the importance of data source selection. ICP and XRF data, when paired with robust algorithms like RF and LightGBM, offer promising tools for improving soil classification and digital soil mapping.