Scholarlink

<p>This study specifically aims to investigate soil contamination from oil spills in the Sarkhun region, located in western Iran. Following the Maroon oil pipeline spill in 2018—which resulted in significant soil, water, and air contamination—this study used acombination of gravimetric analysis to determine oil concentrations (0.1 to 5.0&#xa0;mg/g). The analysis relies on ground-based data from 47 soil samples. This study also employs remote sensing indices including NDWI (Normalized Difference Water Index), NDVI (Normalized Difference Vegetation Index), FI (Fuel Index), and GRSWIR (Green-Red Shortwave Infrared Index) from the Landsat 8 (2018) and Sentinel-2 (2020) platforms. For Landsat 8, the NIR (Near-Infrared) band ranges from 0.85 to 0.88&#xa0;μm and the SWIR (Shortwave Infrared) band ranges from 2.11 to 2.29&#xa0;μm; for Sentinel-2, bands ranging from 0.443 to 0.665&#xa0;μm and a 2.190&#xa0;μm band are applied to determine the relationship between the mentioned contaminations and electromagnetic waves. Also, the K-means and SOM (Self Organizing Map) algorithms are also used to determine contaminated areas. Additionally, the Topographic Position Index (TPI) is used to analyze how geomorphological conditions affect the distribution of contamination. Key findings show that NDWI has high sensitivity to oil contamination, with a correlation coefficient of 0.87. Contamination levels were higher in 2018 (<i>p</i> &lt; 0.05), and hydrocarbon accumulation is greater in valley landforms. The results also show that the K-means method has better performance in identifying contaminated sites, with a clustering accuracy of 0.82, compared to SOM (which has an accuracy of 0.83). This study employs an integrated approach combining remote sensing, machine learning, and TPI analysis to provide a robust framework for monitoring oil contamination. This integrated approach distinguishes it from other contamination studies. The results of this study can be used to manage pollution effectively and reduce environmental impacts in vulnerable areas.</p> Graphical abstract <p>This study comprehensively examined oil contamination in the Serkhun region of western Iran, an area of significant environmental importance. Researchers employed a dual methodology, combining Gravimetric Analysis to pinpoint contaminated zones with advanced remote sensing techniques. They utilized key indicators like NDWI, NDVI, FI, and GRSWIR to map oil-affected areas in both 2018 and 2020. For data categorization, K-means and Self-Organizing Maps (SOM) image classifiers were applied. The findings revealed a consistent decrease in NDWI values in 2018, strongly indicating elevated soil pollution, with specific points showing the lowest values confirming concentrated pollutants. Furthermore, lower NDVI and FI values in 2018 corroborated this contamination, a finding supported by GRSWIR values. A multi-indicator analysis consistently emphasized substantial soil pollution in Serkhun in 2018. The investigation, conducted across 47 points, underscored the strong correlation between NDWI and oil pollution, establishing NDWI as a valuable tool for precise assessments. Notably, the K-means method demonstrated superior accuracy (0.82) in identifying oil-contaminated areas compared to the SOM method. Finally, the study linked landform classification to pollution, showing a higher degree of contamination in 1st class landforms, such as canyons and deeply incised streams.</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

Detection of Soil Oil Contamination Using Remote Sensing and Machine Learning Techniques

  • 期刊论文

  • 发表日期：

    2025年10月15日

查看全文
  • Marzieh Mokarram,
  • Saeed Negahban,
  • Tam Minh Pham

摘要

This study specifically aims to investigate soil contamination from oil spills in the Sarkhun region, located in western Iran. Following the Maroon oil pipeline spill in 2018—which resulted in significant soil, water, and air contamination—this study used acombination of gravimetric analysis to determine oil concentrations (0.1 to 5.0 mg/g). The analysis relies on ground-based data from 47 soil samples. This study also employs remote sensing indices including NDWI (Normalized Difference Water Index), NDVI (Normalized Difference Vegetation Index), FI (Fuel Index), and GRSWIR (Green-Red Shortwave Infrared Index) from the Landsat 8 (2018) and Sentinel-2 (2020) platforms. For Landsat 8, the NIR (Near-Infrared) band ranges from 0.85 to 0.88 μm and the SWIR (Shortwave Infrared) band ranges from 2.11 to 2.29 μm; for Sentinel-2, bands ranging from 0.443 to 0.665 μm and a 2.190 μm band are applied to determine the relationship between the mentioned contaminations and electromagnetic waves. Also, the K-means and SOM (Self Organizing Map) algorithms are also used to determine contaminated areas. Additionally, the Topographic Position Index (TPI) is used to analyze how geomorphological conditions affect the distribution of contamination. Key findings show that NDWI has high sensitivity to oil contamination, with a correlation coefficient of 0.87. Contamination levels were higher in 2018 (p < 0.05), and hydrocarbon accumulation is greater in valley landforms. The results also show that the K-means method has better performance in identifying contaminated sites, with a clustering accuracy of 0.82, compared to SOM (which has an accuracy of 0.83). This study employs an integrated approach combining remote sensing, machine learning, and TPI analysis to provide a robust framework for monitoring oil contamination. This integrated approach distinguishes it from other contamination studies. The results of this study can be used to manage pollution effectively and reduce environmental impacts in vulnerable areas.

Graphical abstract

This study comprehensively examined oil contamination in the Serkhun region of western Iran, an area of significant environmental importance. Researchers employed a dual methodology, combining Gravimetric Analysis to pinpoint contaminated zones with advanced remote sensing techniques. They utilized key indicators like NDWI, NDVI, FI, and GRSWIR to map oil-affected areas in both 2018 and 2020. For data categorization, K-means and Self-Organizing Maps (SOM) image classifiers were applied. The findings revealed a consistent decrease in NDWI values in 2018, strongly indicating elevated soil pollution, with specific points showing the lowest values confirming concentrated pollutants. Furthermore, lower NDVI and FI values in 2018 corroborated this contamination, a finding supported by GRSWIR values. A multi-indicator analysis consistently emphasized substantial soil pollution in Serkhun in 2018. The investigation, conducted across 47 points, underscored the strong correlation between NDWI and oil pollution, establishing NDWI as a valuable tool for precise assessments. Notably, the K-means method demonstrated superior accuracy (0.82) in identifying oil-contaminated areas compared to the SOM method. Finally, the study linked landform classification to pollution, showing a higher degree of contamination in 1st class landforms, such as canyons and deeply incised streams.