基于机器学习的区域土壤重金属污染风险诊断

钱江波; 陈涤; 王夏晖; 李喜林; 黄国鑫

doi:10.13205/j.hjgc.202312037

基于机器学习的区域土壤重金属污染风险诊断

doi: 10.13205/j.hjgc.202312037

1. 辽宁工程技术大学土木工程学院, 辽宁阜新 123000;
2. 生态环境部环境规划院, 北京 100144

基金项目:

国家重点研发计划项目(2018YFC1800205)

辽宁省教育厅科学研究面上项目(LJKZ0344)

详细信息

作者简介:
钱江波(1999-),男,硕士研究生,主要研究方向为土壤和地下水污染风险管控与修复。2325364514@qq.com

通讯作者:
黄国鑫(1980-),男,研究员,主要研究方向为地下水和土壤污染防治。huanggx@caep.org.cn

计量
- 文章访问数: 146
- HTML全文浏览量: 9
- PDF下载量: 3
- 被引次数: 0
出版历程
- 收稿日期: 2022-10-26
- 网络出版日期: 2024-03-08

RISK DIAGNOSIS OF HEAVY METAL POLLUTION IN REGIONAL SOIL BASED ON MACHINE LEARNING

1. School of Civil Engineering, Liaoning Technical University, Fuxin 123000, China;
2. Chinese Academy of Environmental Planning, Beijing 100144, China

摘要

摘要: 准确描述土壤重金属的空间分布并合理划分土壤污染风险区,是土壤重金属污染风险管控的重要前提。以广东省某工业化区域为研究区,基于随机森林(RF)和模糊c均值(FCM),采集577个采样点的土壤重金属数据和12个环境协变量数据,分析了土壤重金属(Cr、Pb、Cu和Zn)浓度,构建了土壤重金属浓度预测模型,开展了数据驱动的区域土壤重金属污染风险分类,并提出了相应的风险分类管控策略。结果表明:Cr、Pb、Cu和Zn的实测浓度分别为4.00~885.60,9.39~2588.11,2.20~475.00,11.05~8162.42 mg/kg。除Pb外,Cr、Cu和Zn的平均值是当地土壤背景值的1.10~1.29倍。4种重金属变异系数均在86%~319%,属中上强度的变异水平,土壤重金属受人为活动影响较大。当RF节点变量数(m_try)值为5,决策数个数(n_tree)值为800时展现了较好的预测性能,预测性能R²的最大值分别达到0.78(Cr)、0.79(Pb)、0.85(Cu)和0.76(Zn),RMSE的最小值分别达到21.57(Cr)、59.88(Pb)、62.38(Cu)和105.88(Zn)。Cr的高值区多位于研究区的东北部和中北部,而Pb、Cu和Zn的高值区则集中于中北部和中南部。在制定土壤重金属污染风险管控策略时,应优先关注中部地区。研究区被划分为2个风险区,其中A区为高风险区,应严格进行重金属污染风险管控;B区为低风险区,应加强监控预警,采取源头控制措施,防止新增污染。
- 土壤 /
- 重金属 /
- 随机森林 /
- 模糊c聚类 /
- 风险诊断
Abstract: Accurately describing the spatial distribution of soil heavy metals and reasonably dividing soil pollution risk areas is an important prerequisite for soil heavy metal pollution risk control. In this paper, an industrialized area in Guangdong Province was taken as the study area. Based on random forest (RF) and fuzzy-c-means (FCM), the soil heavy metal data of 577 sampling points and 12 environmental covariate data were used to analyze the soil Cr, Pb, Cu and Zn concentrations. The prediction models of soil heavy metal concentrations were constructed, the data-driven regional soil heavy metal pollution risk classification was carried out, and the corresponding risk classification control strategy was proposed. The results showed that the measured concentrations of Cr, Pb, Cu and Zn were 4.00 to 885.60, 9.39 to 2588.11, 2.20 to 475.00 and 11.05 to 8162.42 mg/kg, respectively. Except for Pb concentration, the average values of Cr, Cu and Zn concentrations were 1.10~1.29 times higher than the local soil background values. The variation coefficients of the four heavy metals were between 86% and 319%, which belonged to the variation level of medium or upper intensity. Soil heavy metals were greatly affected by human activities. The random forest had a good prediction performance when the number of node variables (m_try) was 5 and the number of decision trees (n_tree) was 800. The maximum values of R² were 0.78, 0.79, 0.85 and 0.76, respectively, and the minimum values of RMSE were 21.57, 59.88, 62.38 and 105.88, respectively. The high-value areas of Cr were mostly located in the northeastern and north-central of the study area, while the high-value areas of Pb, Cu and Zn were concentrated in the north-central and south-central parts. In the development of soil heavy metal pollution risk control strategies, priority should be given to the central region. The study area was divided into two risk areas, of which, area A was a high-risk area, and the risk control of heavy metal pollution should be strictly carried out. Area B was a low-risk area. It is also necessary to strengthen monitoring and early warning and take source control measures to prevent new pollution.
- soil /
- heavy metals /
- random forest /
- fuzzy c-means /
- risk diagnosis

HTML全文

参考文献(47)

[1]	金昭, 吕建树. 基于机器学习模型的区域土壤重金属空间预测精度比较研究[J]. 地理研究, 2022, 41(6): 1731-1747.
[2]	安文超, 孙立娥, 马立科, 等. 某典型工业聚集区遗留地土壤重金属污染特征及健康风险评价[J]. 湖南师范大学自然科学学报, 2022, 45(5): 108-116.
[3]	孟令华, 杜小亮, 刘乾, 等. 泰安市城区土壤重金属污染特征及风险评价[J]. 中国无机分析化学, 2022, 12(5): 41-49.
[4]	HAN I, WHITWORTH K W, CHRISTENSEN B, et al. Heavy metal pollution of soils and risk assessment in Houston, Texas following Hurricane Harvey[J]. Environ. Pollut, 2022, 296: 118717.
[5]	FERNANDO M, ASIM B. Are heavy metals in urban garden soils linked to vulnerable populations? a case study from Guelph, Canada[J]. Scientific Reports, 2021, 11(1): 11286.
[6]	SERGEEV A P, BUEVICH A G, BAGLAEVA E M, et al. Combining spatial autocorrelation with machine learning increases prediction accuracy of soil heavy metals[J]. Catena, 2019, 174:425-435.
[7]	TALUKDER R, RABBI M H, BAHARIM N B, et al. Source identification and ecological risk assessment of heavy metal pollution in sediments of Setiu wetland, Malaysia[J]. Environmental Forensic, 2022, 23(1/2):241-254.
[8]	汪峰, 黄言欢, 李如忠, 等. 有色金属矿业城市典型村镇土壤重金属污染评价及来源解析[J]. 环境科学, 2022, 43(9): 4800-4809.
[9]	李喜林, 于晓婉, 刘玲, 等. 复合制剂修复铬污染土的条件优化及微观特性[J]. 长江科学院院报, 2021, 38(7): 80-87.
[10]	黄国鑫, 刘瑞平, 杨瑞杰, 等. 我国农用地土壤重金属污染风险管控研究进展与实践要求[J]. 环境工程, 2022, 40(1): 216-223.
[11]	周颖, 王雪梅, 蒋玉琢, 等. 北京市平谷区金矿区周边土壤砷、汞赋存形态特征及生态风险评价[J]. 环境工程, 2021, 39(8): 203-210 ,164.
[12]	WANG Z, LUO Y F, ZHENG C L, et al. Spatial distribution, source identification, and risk assessment of heavy metals in the soils from a mining region: a case study of Bayan Obo in northwestern China[J]. Human and Ecological Risk Assessment: An International Journal, 2020, 27(5): 1276-1295.
[13]	杨杰, 董静, 宋洲, 等. 鄂西铜铅锌尾矿库周边农田土壤-水稻重金属污染状况及风险评价[J]. 岩矿测试, 2022, 41(5): 867-879.
[14]	CHEN J, ZNANG J L, QU M K, et al. Pollution characteristics and risk assessment of soil heavy metals in the areas affected by the mining of metal-bearing minerals in southwest China[J]. Bulletin of Environmental Contamination and Toxicology, 2021, 107(6): 1070-1079.
[15]	韩存亮, 罗炳圣, 常春英, 等. 基于多种方法的区域农业土壤重金属污染成因分析研究[J]. 生态与农村环境学报, 2022, 38(2): 176-183.
[16]	尚婷婷, 张亚群, 周静, 等. 多元统计分析在农田土壤重金属污染源解析中的应用[J]. 环境生态学, 2022, 4(4): 93-97.
[17]	焦思佳, 吴田军, 董世英, 等. 基于反距离加权随机森林的空间推测方法研究[J]. 昆明理工大学学报(自然科学版), 2022, 47(4): 46-54.
[18]	卢月明, 王亮, 仇阿根, 等. 局部加权线性回归模型的PM_2.5空间插值方法[J]. 测绘科学, 2018, 43(11): 79-84 ,91.
[19]	盛红坤, 徐泽, 王佳楠, 等. 天津市某校园土壤中重金属污染研究及其评价[J]. 应用化工, 2021, 50(6): 1529-1532.
[20]	刘雪松, 王雨山, 尹德超, 等. 白洋淀内不同土地利用类型土壤重金属分布特征与污染评价[J]. 土壤通报, 2022, 53(3): 710-717.
[21]	JIANG Y F, YE Y C, GUO X, et al. Spatiotemporal variation of soil heavy metals in farmland influenced by human activities in the Poyang Lake region, China[J]. Catena, 2019, 176: 279-288.
[22]	OBIRI-NYARKO F, DUAH A A, KARIKARI A Y, et al. Assessment of heavy metal contamination in soils at the Kpone landfill site, Ghana: implication for ecological and health risk assessment[J]. Chemosphere, 2021, 282:131007.
[23]	HOU D Y, O'CONNOR D, NATHANAIL P, et al. Integrated GIS and multivariate statistical analysis for regional scale assessment of heavy metal soil contamination: a critical review[J]. Environmental Pollution, 2017, 231:1188-1200.
[24]	LEUNG H M, DUZGOREN-AYDIN N S, AU C K, et al. Monitoring and assessment of heavy metal contamination in a constructed wetland in Shaoguan(Guangdong Province, China): bioaccumulation of Pb, Zn, Cu and Cd in aquatic and terrestrial components[J]. Environmental Science and Pollution Research, 2016,24(10):9079-9088.
[25]	ZHOU M, LIAO B, SHU W, et al. Pollution assessment and potential sources of heavy metals in agricultural soils around four Pb/Zn mines of Shaoguan City, China[J]. Soil and Sediment Contamination. 2015, 24(1): 76-89.
[26]	贾建丽. 环境土壤学[M] 北京: 化学工业出版社, 2016.
[27]	于雷, 洪永胜, 耿雷, 等. 基于偏最小二乘回归的土壤有机质含量高光谱估算[J]. 农业工程学报, 2015, 31(14): 103-109.
[28]	孙小丽, 阿不都艾尼·阿不里, 哈力旦·艾赛都力, 等. 基于PMF模型的五彩湾矿区土壤重金属污染空间分布与来源解析[J]. 中国矿业, 2022, 31(11): 62-70.
[29]	XIE T, LU F, WANG M, et al. The application of urban anthropogenic background to pollution evaluation and source identification of soil contaminants in Macau, China[J]. Science Total Environment, 2021, 778: 146263.
[30]	WU J, TENG Y G, CHEN H Y, et al. Machine-learning models for on-site estimation of background concentrations of arsenic in soils using soil formation factors[J]. Jounral of Soils and Sediments, 2016, 16(6): 1787-1797.
[31]	王腾军, 方珂, 杨耘, 等. 随机森林回归模型用于土壤重金属含量多光谱遥感反演[J]. 测绘通报, 2021(11): 92-95.
[32]	毛丽丽, 于静洁, 张一驰, 等. 模糊c均值聚类方法在黑河下游土壤属性制图中的初步应用研究[J]. 干旱区资源与环境, 2013, 27(1): 195-201.
[33]	MORAL F J, TERRÓN J M, MARQUES DA SILVA J R, et al. Delineation of management zones using mobile measurements of soil apparent electrical conductivity and multivariate geostatistical techniques[J]. Soil Tillage Research, 2010, 106(2): 335-343.
[34]	MOURA-BUENO J M, DALMOLIN R S D, HORST-HEINEN T Z, et al. Environmental covariates improve the spectral predictions of organic carbon in subtropical soils in southern Brazil[J]. Geoderma, 2021, 393: 114981.
[35]	CHEN S C, LIANG Z Z, WEBSTER R, et al. A high-resolution map of soil pH in China made by hybrid modelling of sparse soil data and environmental covariates and its implications for pollution[J]. Science of the Total Environment, 2018, 665(10): 273-283.
[36]	JIA X L, FU T T, HU B F, et al. Identification of the potential risk areas for soil heavy metal pollution based on the source-sink theory[J]. Journal of Hazardous Materials, 2020, 393: 122424.
[37]	BREUNING F M, GALVÃO L S, DALAGNOL R, et al. Assessing the effect of spatial resolution on the delineation of management zones for smallholder farming in southern Brazil[J]. Remote Sensing Applications: Society and Environment, 2020, 19: 100325.
[38]	WANG H Y, LU S G. Spatial dstribution, source identification and affecting factors of heavy metals contamination in urban-suburban soils of Lishui city, China[J]. Environmental Earth Science, 2011, 64(7): 1921-1929.
[39]	李懿. 区域土壤重金属污染风险评价、驱动因子与管控策略研究[D]. 杭州:浙江大学, 2022.
[40]	王佳昱. 基于地统计和数据挖掘技术的土壤重金属空间分异与源解析研究[D]. 杭州:浙江大学, 2018.
[41]	生态环境部. 土壤环境质量农用地土壤污染风险管控标准(试行): GB 15618—2018[S]. 北京: 中国标准出版社, 2018.
[42]	郑堃, 任宗玲, 覃小泉, 等.韶关工矿区水稻土和稻米中重金属污染状况及风险评价[J]. 农业环境科学报, 2018, 37(5): 915-925.
[43]	王其枫, 王富华, 孙芳芳, 等. 广东韶关主要矿区周边农田土壤铅、镉的形态分布及生物有效性研究[J]. 农业环境科学学报, 2012, 31(6): 1097-1103.
[44]	罗莹华. 韶关某冶炼厂周边土壤重金属污染调查与生态风险评价[J]. 安徽农业科学, 2016, 44(19): 133-136.
[45]	奉大博, 董树义, 杨棣, 等. 广东韶关乐昌铅锌矿土壤重金属污染特征及评价[J]. 矿物岩石, 2022, 42(3): 123-133.
[46]	许超, 夏北成, 秦建桥, 等. 广东大宝山矿山下游地区稻田土壤的重金属污染状况的分析与评价[J]. 农业环境科学学报, 2007,26(增刊2): 549-553.
[47]	孟令华, 杜小亮, 刘乾, 等. 泰安市城区土壤重金属污染特征及风险评价[J]. 中国无机分析化学, 2022, 12(5): 41-49.