Scholarlink

The interest of geostatistical methods and modelisation in the environmental characterization of contaminated sites is widely recognized and documented. Recent democratization of field spectroscopic measurement devices for quantification of contaminants in soils is both a technical and methodological breakthrough for the industry. Field measurements provide 3 to 5 times more data compared to laboratory analysis. Properly interpreted they can be used to adapt sampling designs and are key to perform onsite and in one intervention the delimitation of pollutants. The downside of those modern devices based on spectroscopy is that the quantification of contaminants requires chemometrics analysis reinforced with AI. Those predictions although very efficient, raise the need for a quick assessment and QAQC of the predictions. Since natural heterogeneity of soils is a major source of interference, AI models for the quantification of contaminants are often site-specific and created after an initial sampling phase. We consider using this prior data to optimize the delimitation phase and reinforce the impact of each drillhole. The present work is a first review of: At term, an industrial-ready integrated approach of chemometric, geospatial analysis and AI would be relevant for a better use of spectroscopic onsite measurements of contaminants.

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

Real-Time Geostatistics for an Uncertainty Driven Environmental Soils Characterization—Field Algorithms Reviewed and Discussed

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Antoine Heude,
  • Justine Elias,
  • Victorine Herbin

摘要

The interest of geostatistical methods and modelisation in the environmental characterization of contaminated sites is widely recognized and documented. Recent democratization of field spectroscopic measurement devices for quantification of contaminants in soils is both a technical and methodological breakthrough for the industry. Field measurements provide 3 to 5 times more data compared to laboratory analysis. Properly interpreted they can be used to adapt sampling designs and are key to perform onsite and in one intervention the delimitation of pollutants. The downside of those modern devices based on spectroscopy is that the quantification of contaminants requires chemometrics analysis reinforced with AI. Those predictions although very efficient, raise the need for a quick assessment and QAQC of the predictions. Since natural heterogeneity of soils is a major source of interference, AI models for the quantification of contaminants are often site-specific and created after an initial sampling phase. We consider using this prior data to optimize the delimitation phase and reinforce the impact of each drillhole. The present work is a first review of: At term, an industrial-ready integrated approach of chemometric, geospatial analysis and AI would be relevant for a better use of spectroscopic onsite measurements of contaminants.