Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Artículo científico |
| Language: | en |
| Published: |
Water research
2026
|
| Subjects: | |
| Online Access: | https://pubmed.ncbi.nlm.nih.gov/41547211/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1868266097716756480 |
|---|---|
| author | Wu, Nian-Nian Liu, Yuan Liu, Shan Chen, Ji-Xun Zhou, Qin-Ge Xu, Xiang-Rong Yang, Qing-Song Ling, Juan Zhao, Jian-Liang |
| author_facet | Wu, Nian-Nian Liu, Yuan Liu, Shan Chen, Ji-Xun Zhou, Qin-Ge Xu, Xiang-Rong Yang, Qing-Song Ling, Juan Zhao, Jian-Liang Wu, Nian-Nian Liu, Yuan Liu, Shan Chen, Ji-Xun Zhou, Qin-Ge Xu, Xiang-Rong Yang, Qing-Song Ling, Juan Zhao, Jian-Liang |
| collection | PubMed - marine biology |
| contents | Integrating machine learning and prioritization to decipher the partitioning and risk of tire-derived chemicals in highly populated coastal regions. Wu, Nian-Nian Liu, Yuan Liu, Shan Chen, Ji-Xun Zhou, Qin-Ge Xu, Xiang-Rong Yang, Qing-Song Ling, Juan Zhao, Jian-Liang Machine Learning Water Pollutants, Chemical Geologic Sediments Environmental Monitoring Random Forest Tire-derived chemicals (TDCs) are emerging contaminants of escalating ecological concern. This study presents a comprehensive investigation into the occurrence, partitioning behavior, and environmental risks of TDCs in the Greater Bay Area, one of the world's four major bay areas. Seven and nine TDCs were identified in water and sediments, respectively, with total concentrations reaching 33,080 ng/L and 723 ng/g. 2-Methylthio-benzothiazole was established as a robust, phase-independent indicator for ΣTDCs, showing strong correlations across compartments. An interpretable machine learning framework was developed by integrating molecular descriptors with sediment-related predictors, water-quality/hydrodynamic parameters, and spatial drivers to elucidate the key drivers of pseudo partitioning coefficient (log K), with the Random Forest model achieving superior performance (test R² = 0.95). Shapley additive explanations (SHAP) analysis identified descriptors (e.g., GATS1dv, BCUTc-1l) as dominant predictors, highlighting the critical roles of molecular topology and charge distribution. Sediment pH emerged as an important environmental parameter of partitioning, whereas water salinity contributed comparatively less across models. Ecological risk was high (Hazard Index > 1) across all sites, largely due to hexamethoxymethylmelamine (HMMM) and 1H-benzotriazole, with HMMM further prioritized as a high-priority contaminant. This monitoring-modeling-interpretation workflow provides mechanistic insights into the environmental behavior of TDCs and establishes a transferable framework for risk-based chemical management in urbanized coastal environments worldwide. |
| format | Artículo científico |
| id | pubmed_41547211 |
| institution | PubMed |
| language | en |
| publishDate | 2026 |
| publisher | Water research |
| record_format | pubmed |
| spellingShingle | Integrating machine learning and prioritization to decipher the partitioning and risk of tire-derived chemicals in highly populated coastal regions. Wu, Nian-Nian Liu, Yuan Liu, Shan Chen, Ji-Xun Zhou, Qin-Ge Xu, Xiang-Rong Yang, Qing-Song Ling, Juan Zhao, Jian-Liang Machine Learning Water Pollutants, Chemical Geologic Sediments Environmental Monitoring Random Forest Integrating machine learning and prioritization to decipher the partitioning and risk of tire-derived chemicals in highly populated coastal regions. Wu, Nian-Nian Liu, Yuan Liu, Shan Chen, Ji-Xun Zhou, Qin-Ge Xu, Xiang-Rong Yang, Qing-Song Ling, Juan Zhao, Jian-Liang Machine Learning Water Pollutants, Chemical Geologic Sediments Environmental Monitoring Random Forest Tire-derived chemicals (TDCs) are emerging contaminants of escalating ecological concern. This study presents a comprehensive investigation into the occurrence, partitioning behavior, and environmental risks of TDCs in the Greater Bay Area, one of the world's four major bay areas. Seven and nine TDCs were identified in water and sediments, respectively, with total concentrations reaching 33,080 ng/L and 723 ng/g. 2-Methylthio-benzothiazole was established as a robust, phase-independent indicator for ΣTDCs, showing strong correlations across compartments. An interpretable machine learning framework was developed by integrating molecular descriptors with sediment-related predictors, water-quality/hydrodynamic parameters, and spatial drivers to elucidate the key drivers of pseudo partitioning coefficient (log K), with the Random Forest model achieving superior performance (test R² = 0.95). Shapley additive explanations (SHAP) analysis identified descriptors (e.g., GATS1dv, BCUTc-1l) as dominant predictors, highlighting the critical roles of molecular topology and charge distribution. Sediment pH emerged as an important environmental parameter of partitioning, whereas water salinity contributed comparatively less across models. Ecological risk was high (Hazard Index > 1) across all sites, largely due to hexamethoxymethylmelamine (HMMM) and 1H-benzotriazole, with HMMM further prioritized as a high-priority contaminant. This monitoring-modeling-interpretation workflow provides mechanistic insights into the environmental behavior of TDCs and establishes a transferable framework for risk-based chemical management in urbanized coastal environments worldwide. |
| title | Integrating machine learning and prioritization to decipher the partitioning and risk of tire-derived chemicals in highly populated coastal regions. |
| topic | Machine Learning Water Pollutants, Chemical Geologic Sediments Environmental Monitoring Random Forest |
| url | https://pubmed.ncbi.nlm.nih.gov/41547211/ |