View in EDS

Predicting soil cadmium spatial distribution with HDXRF coupling to multi-model machine learning in the karst area.

Saved in:
Bibliographic Details
Title: Predicting soil cadmium spatial distribution with HDXRF coupling to multi-model machine learning in the karst area.
Authors: Gu, Zichen1 (AUTHOR) zichengu@126.com, Liu, Hongyan1,2 (AUTHOR) hyliu@gzu.edu.cn, Mei, Xue1 (AUTHOR) 2654961758@qq.com, Wu, Longhua3 (AUTHOR) lhwu@issas.ac.cn, Rasool, Ghulam2 (AUTHOR) rasool@gzu.edu.cn, Li, Xuexian1 (AUTHOR) xxli5@gzu.edu.cn, Chen, Yonglin2 (AUTHOR) 3235329591@qq.com, Zhang, Hai1 (AUTHOR) HaiZhangSN2000@126.com, Li, Chunyan1 (AUTHOR) 3353339550@qq.com, Song, Wei4 (AUTHOR) 3171311698@qq.com
Source: Environmental Monitoring & Assessment. Mar2026, Vol. 198 Issue 3, p1-18. 18p.
Subject Terms: *X-ray fluorescence, *Machine learning, *Inorganic soil pollutants, *Soil pollution, *Pollution, *Karst
Geographic Terms: China
Abstract: Mapping soil potentially toxic metal spatial distribution is critical for safe and sustainable use and management of the contaminating soil. However, it is challenged by high soil heterogeneity and non-linear and complex driving factors in the karst area. Our machine learning framework integrates high-definition X-ray fluorescence (HDXRF) spectra with soil properties and topographical features to predict soil Cd concentration in the karst area. We analyzed soil samples from two geogenically Cd-enriched areas and one smelting-affected area in southwestern China using six different ML models. The results demonstrated that the Elastic Net model achieved the best predictions (R2 = 0.879, RMSE (Root Mean Square Error) = 0.299 mg·kg−1) in geogenic areas, while XGBoost performed best (R2 = 0.893, RMSE = 1.33 mg·kg−1) in smelting zones, indicating distinct Cd dynamics in these environments. SHAP (SHapley Additive exPlanations) analysis revealed that the key factors driving Cd accumulation varied between the two settings. In geogenically Cd-enriched systems, the main influences were sand content and topography. In contrast, in smelting pollution areas, pH levels and interactions between silt and organic matter became the dominant drivers. The Geodetector method quantified interactions between synergistic factors, with q-values ranging from 0.60 to 0.85, illuminating clay-slope coupling in natural systems, as opposed to the silt-cation exchange capacity synergy found in contaminated zones. This framework enables efficient and cost-effective large-scale monitoring of Cd levels while providing insights into the lithology-dependent mechanisms of contamination. The soil Cd spatial distribution prediction model in this work offers critical information for precise soil management in vulnerable karst ecosystems. [ABSTRACT FROM AUTHOR]
Database: Energy & Power Source
FullText Text:
  Availability: 0
Header DbId: enr
DbLabel: Energy & Power Source
An: 192481239
AccessLevel: 6
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: Predicting soil cadmium spatial distribution with HDXRF coupling to multi-model machine learning in the karst area.
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Gu%2C+Zichen%22">Gu, Zichen</searchLink><relatesTo>1</relatesTo> (AUTHOR)<i> zichengu@126.com</i><br /><searchLink fieldCode="AR" term="%22Liu%2C+Hongyan%22">Liu, Hongyan</searchLink><relatesTo>1,2</relatesTo> (AUTHOR)<i> hyliu@gzu.edu.cn</i><br /><searchLink fieldCode="AR" term="%22Mei%2C+Xue%22">Mei, Xue</searchLink><relatesTo>1</relatesTo> (AUTHOR)<i> 2654961758@qq.com</i><br /><searchLink fieldCode="AR" term="%22Wu%2C+Longhua%22">Wu, Longhua</searchLink><relatesTo>3</relatesTo> (AUTHOR)<i> lhwu@issas.ac.cn</i><br /><searchLink fieldCode="AR" term="%22Rasool%2C+Ghulam%22">Rasool, Ghulam</searchLink><relatesTo>2</relatesTo> (AUTHOR)<i> rasool@gzu.edu.cn</i><br /><searchLink fieldCode="AR" term="%22Li%2C+Xuexian%22">Li, Xuexian</searchLink><relatesTo>1</relatesTo> (AUTHOR)<i> xxli5@gzu.edu.cn</i><br /><searchLink fieldCode="AR" term="%22Chen%2C+Yonglin%22">Chen, Yonglin</searchLink><relatesTo>2</relatesTo> (AUTHOR)<i> 3235329591@qq.com</i><br /><searchLink fieldCode="AR" term="%22Zhang%2C+Hai%22">Zhang, Hai</searchLink><relatesTo>1</relatesTo> (AUTHOR)<i> HaiZhangSN2000@126.com</i><br /><searchLink fieldCode="AR" term="%22Li%2C+Chunyan%22">Li, Chunyan</searchLink><relatesTo>1</relatesTo> (AUTHOR)<i> 3353339550@qq.com</i><br /><searchLink fieldCode="AR" term="%22Song%2C+Wei%22">Song, Wei</searchLink><relatesTo>4</relatesTo> (AUTHOR)<i> 3171311698@qq.com</i>
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="JN" term="%22Environmental+Monitoring+%26+Assessment%22">Environmental Monitoring & Assessment</searchLink>. Mar2026, Vol. 198 Issue 3, p1-18. 18p.
– Name: Subject
  Label: Subject Terms
  Group: Su
  Data: *<searchLink fieldCode="DE" term="%22X-ray+fluorescence%22">X-ray fluorescence</searchLink><br />*<searchLink fieldCode="DE" term="%22Machine+learning%22">Machine learning</searchLink><br />*<searchLink fieldCode="DE" term="%22Inorganic+soil+pollutants%22">Inorganic soil pollutants</searchLink><br />*<searchLink fieldCode="DE" term="%22Soil+pollution%22">Soil pollution</searchLink><br />*<searchLink fieldCode="DE" term="%22Pollution%22">Pollution</searchLink><br />*<searchLink fieldCode="DE" term="%22Karst%22">Karst</searchLink>
– Name: SubjectGeographic
  Label: Geographic Terms
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22China%22">China</searchLink>
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: Mapping soil potentially toxic metal spatial distribution is critical for safe and sustainable use and management of the contaminating soil. However, it is challenged by high soil heterogeneity and non-linear and complex driving factors in the karst area. Our machine learning framework integrates high-definition X-ray fluorescence (HDXRF) spectra with soil properties and topographical features to predict soil Cd concentration in the karst area. We analyzed soil samples from two geogenically Cd-enriched areas and one smelting-affected area in southwestern China using six different ML models. The results demonstrated that the Elastic Net model achieved the best predictions (R2 = 0.879, RMSE (Root Mean Square Error) = 0.299 mg·kg−1) in geogenic areas, while XGBoost performed best (R2 = 0.893, RMSE = 1.33 mg·kg−1) in smelting zones, indicating distinct Cd dynamics in these environments. SHAP (SHapley Additive exPlanations) analysis revealed that the key factors driving Cd accumulation varied between the two settings. In geogenically Cd-enriched systems, the main influences were sand content and topography. In contrast, in smelting pollution areas, pH levels and interactions between silt and organic matter became the dominant drivers. The Geodetector method quantified interactions between synergistic factors, with q-values ranging from 0.60 to 0.85, illuminating clay-slope coupling in natural systems, as opposed to the silt-cation exchange capacity synergy found in contaminated zones. This framework enables efficient and cost-effective large-scale monitoring of Cd levels while providing insights into the lithology-dependent mechanisms of contamination. The soil Cd spatial distribution prediction model in this work offers critical information for precise soil management in vulnerable karst ecosystems. [ABSTRACT FROM AUTHOR]
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=enr&AN=192481239
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1007/s10661-026-15062-1
    Languages:
      – Code: eng
        Text: English
    PhysicalDescription:
      Pagination:
        PageCount: 18
        StartPage: 1
    Subjects:
      – SubjectFull: X-ray fluorescence
        Type: general
      – SubjectFull: Machine learning
        Type: general
      – SubjectFull: Inorganic soil pollutants
        Type: general
      – SubjectFull: Soil pollution
        Type: general
      – SubjectFull: Pollution
        Type: general
      – SubjectFull: Karst
        Type: general
      – SubjectFull: China
        Type: general
    Titles:
      – TitleFull: Predicting soil cadmium spatial distribution with HDXRF coupling to multi-model machine learning in the karst area.
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Gu, Zichen
      – PersonEntity:
          Name:
            NameFull: Liu, Hongyan
      – PersonEntity:
          Name:
            NameFull: Mei, Xue
      – PersonEntity:
          Name:
            NameFull: Wu, Longhua
      – PersonEntity:
          Name:
            NameFull: Rasool, Ghulam
      – PersonEntity:
          Name:
            NameFull: Li, Xuexian
      – PersonEntity:
          Name:
            NameFull: Chen, Yonglin
      – PersonEntity:
          Name:
            NameFull: Zhang, Hai
      – PersonEntity:
          Name:
            NameFull: Li, Chunyan
      – PersonEntity:
          Name:
            NameFull: Song, Wei
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 03
              Text: Mar2026
              Type: published
              Y: 2026
          Identifiers:
            – Type: issn-print
              Value: 01676369
          Numbering:
            – Type: volume
              Value: 198
            – Type: issue
              Value: 3
          Titles:
            – TitleFull: Environmental Monitoring & Assessment
              Type: main
ResultId 1