aceEcological risk zoning of soil heavy metal pollution using biomarker-enhanced indexing and machine learning
Abstract: Traditional soil heavy metal risk assessments suffer from fragmented indices and poor logical integration. Compared to single-chemical indicators, biomarkers more sensitively reflect biological effects and early ecological risks. To address these issues, a method for constructing a comprehensive index integrating biological and non-biological multi-indexes was proposed in this study. First, the Criteria Importance Through Intercriteria Correlation (CRITIC)-Fuzzy Biomarker Response Index (CFBRI) is developed by i…
Effects of amendments on heavy metal immobilization and uptake by Rhizoma chuanxiong on copper and cadmium contaminated soil by Xie, Yanluo; Xiao, Kemeng; Sun, Yang; Gao, Yufeng; Yang, Han; Xu, Heng. CC BY 4.0 · https://creativecommons.org/licenses/by/4.0
Researchers proposed a framework that integrates biological and non-biological indicators for soil heavy metal risk assessment. They developed a CRITIC-Fuzzy Biomarker Response Index to weight multi-timepoint biomarker data, then combined it with four abiotic pollution indices via PCA to create agricultural and construction land comprehensive indexes, which were used as labels to train five machine learning classifiers.
By the September 2026 publication date, the Random Forest models showed the highest cross-validation accuracies and produced zoning consistent with land use functions, suggesting a transferable decision-support tool for regional land management. What remains uncertain from the text is geographic generalizability beyond the reported case study and long-term field validation of management outcomes.
- CFBRI was developed by integrating CRITIC method and fuzzy comprehensive evaluation into BRI to dynamically weight multi-timepoint biomarker data and resolve non-monotonic responses.
- Land use-specific comprehensive indexes AL-CI and CL-CI were built via Principal Component Analysis integrating Pollution Load Index, Nemerow Index, Potential Ecological Risk Index, Improved Geo-accumulation Index, and CFBRI.
- Five machine learning algorithms were tested for multi-classification prediction using the CIs as risk labels, with Random Forest selected as best performer.
Machine learning-based risk zoning using integrated biological and abiotic indexes improved fit and classification, with CFBRI raising R2 to 0.62 and Random Forest models reaching cross-validation accuracies of 0.888 for agricultural land and 0.905 for construction land.
The rundown
Traditional assessments were described as suffering from fragmented indices and poor logical integration, while biomarkers were noted to more sensitively reflect biological effects and early ecological risks.
The study constructed comprehensive indexes by PCA and then built multi-classification predictive models with five machine learning algorithms, using AL-CI and CL-CI as reliable risk labels for a regional case study.
Sources
- Peer-reviewedEnvironmental Geochemistry and Health2026-09-12
How should this claim be treated?
The debate