Use of machine-learning and receptor models for prediction and source apportionment of heavy metals in coastal reclaimed soils

Ecological Indicators

Shuangshuang Shao¹  Aijing Yin²  Xiaohui Yang³  Pengbao Wu⁴  Jingtao Wu⁴  Huan Zhang⁴  Ming Zhang⁴  Chao Gao⁵  Manman Fan⁶ 

[1] Corresponding authors.;Key Laboratory of Coastal Zone Exploitation and Protection, Ministry of Natural Resources, Institute of Land Surveying and Planning of Jiangsu, Nanjing 210096, China;Nanjing Institute of Environmental Sciences, Ministry of Ecology and Environment, Nanjing 210042, China;School of Geography and Ocean Science, Nanjing University, Nanjing 210023, China;School of Geography and Tourism, Huizhou University, Huizhou 516007, Guangdong, China;Shandong Provincial Key Laboratory of Water and Soil Conservation and Environmental Protection, College of Resources and Environment, Linyi University, Linyi 276000, China;

关键词: Source apportionment;    Spatial prediction;    Heavy metal;    Random forest;    Positive matrix factorization;   

DOI  :  

来源: DOAJ

【 摘 要 】

Quantitative estimations of sources and spatial distribution of soil heavy metals (HMs) is essential for strategizing policies for soil protection and remediation. As a special soil ecosystem, the intensified human activities on coastal reclaimed lands generally causes soil contamination with HMs. This study aimed to apportion sources of HMs and predict their spatial distributions on coastal reclaimed lands. A total of 241 surface (0–20 cm) soil and sediment samples were collected from a reclamation zone following intensive agricultural use of eastern China. The concentrations of soil and sediment As, Cr, Cu, Ni, Pb, Zn, Cd, and Hg were measured along with organic carbon, nitrogen, phosphorus, pH, Cl, clay, silt, sand, CaO, Fe2O3, Al2O3, and SiO2. The potential sources of HMs were identified and apportioned using random forest (RF) and positive matrix factorization (PMF) models. According to the models, natural and a portion of anthropogenic sources, agricultural activities, and human emission from solar power generation and vehicle exhaust contributed 42.9%, 28.9%, and 28.2% of the total HMs, respectively. Separately, 65.0% of As, 36.6% of Cr, 49.1% of Cu, 46.4% of Ni, 39.5% of Pb, and 44.0% of Zn were originated from natural and some anthropogenic sources. Agricultural activities contributed 54.9% of Cd and 46.4% of Hg to the reclaimed soils. Emissions from solar power generation and vehicle exhaust had significant influences on Cr and Pb, with contributions of 39.0% and 28.0%, respectively. Furthermore, the RF model yielded satisfying results in predicting HM distributions based on the measurement of soil variables. When only considering independent variables, the RF model revealed slightly lower but still satisfactory abilities in HMs prediction. In reclaimed soils, the temporal increase and close relationship between soil Cd and phosphorus signified the potential threats of Cd contamination in coastal reclaimed soils. Therefore, the applications of Cd-rich phosphoric fertilizers should be considered with high concern.

【 授权许可】

Unknown