Integrative risk assessment method via combining geostatistical analysis, random forest, and receptor models for potentially toxic elements in selenium-rich soil

被引：3

作者：

Wu, Hao ^{[1
]}

Cheng, Nan ^{[1
]}

Chen, Ping ^{[1
]}

Zhou, Fei ^{[1
]}

Fan, Yao ^{[1
]}

Qi, Mingxing ^{[1
]}

Shi, Jingyi ^{[1
]}

Zhang, Zhimin ^{[2
]}

Ren, Rui ^{[2
]}

Wang, Cheng ^{[1
]}

Liang, Dongli ^{[1
,3
]}

机构：

[1] Northwest A&F Univ, Coll Nat Resources & Environm, Yangling 712100, Shaanxi, Peoples R China

[2] Shaanxi Hydrogeolog Engn Geosci & Environm Geosci, Yangling, Peoples R China

[3] Minist Agr, Key Lab Plant Nutr & Agrienvironm Northwest China, Yangling 712100, Shaanxi, Peoples R China

来源：

ENVIRONMENTAL POLLUTION | 2023年 / 337卷

基金：

中国国家自然科学基金;

关键词：

Potentially toxic elements; Selenium -rich soil; Risk assessment; Source identification; Monte Carlo simulation; HEAVY-METALS; SOURCE APPORTIONMENT; MERCURY EMISSIONS; COAL; BIOAVAILABILITY; GEOCHEMISTRY; DEPOSITION; SPECIATION;

D O I：

10.1016/j.envpol.2023.122555

中图分类号：

X [环境科学、安全科学];

学科分类号：

08 ; 0830 ;

摘要：

Revealing the spatial features and source of associated potentially toxic elements (PTEs) is crucial for the safe use of selenium (Se)-rich soils. An integrative risk assessment (GRRRA) approach based on geostatistical analysis (GA), random forest (RF), and receptor models (RMs) was first established to investigate the spatial distribution, sources, and potential ecological risks (PER) of PTEs in 982 soils from Ziyang City, a typical natural Se-rich area in China. RF combined with multiple RMs supported the source apportionment derived from the RMs and provided accurate results for source identification. Then, quantified source contributions were introduced into the risk assessment. Eighty-three percent of the samples contain Cd at a high PER level in local Se-rich soils. GA based on spatial interpolation and spatial autocorrelation showed that soil PTEs have distinct spatial characteristics, and high values are primarily distributed in this research areas. Absolute principal component score/ multiple line regression (APCS/MLR) is more suitable than positive matrix factorization (PMF) for source apportionment in this study. RF combined with RMs more accurately and scientifically extracted four sources of soil PTEs: parent material (48.91%), mining (17.93%), agriculture (8.54%), and atmospheric deposition (24.63%). Monte Carlo simulation (MCS) demonstrates a 47.73% probability of a non-negligible risk (RI > 150) caused by parent material and 3.6% from industrial sources, respectively. Parent material (64.20%, RI = 229.56) and mining (16.49%, RI = 58.96) sources contribute to the highest PER of PTEs. In conclusion, the GRRRA method can comprehensively analyze the distribution and sources of soil PTEs and effectively quantify the source contribution to PER, thus providing the theoretical foundation for the secure utilization of Se-rich soils and environmental management and decision making.

引用

页数：10