摘要
正定矩阵因子分解模型(PMF)可用于污染源未知情况下的土壤中重金属来源解析,但对数据样本敏感,结果波动大。为探究PMF模型对土壤重金属源解析的适用性,本研究以湖南省水口山铅锌矿周边农田土壤为研究对象,考察成分谱元素种类和异常值剔除两个因素对PMF模型解析结果的影响。根据成分谱中有无地壳元素和是否剔除异常值,建立4种数据样本,对PMF模型结果进行比较分析。结果表明:根据散点图剔除2个异常样本后未改变分类结果,仅改变各源贡献率;引入6种地壳元素后PMF模型的源轮廓(源数量和贡献元素)均发生变化。在成分谱中增加地壳元素后,源解析结果受异常值影响小,结果更稳定且容易解释。因此,应将地壳元素引入成分谱并对数据进行预处理,可较好地保证源解析结果的稳定性和可信度。结合文献和该地区实际情况对模型结果进行解读,最终确定5个来源:Pb、Zn、Cd和Sb主要来自铅锌矿的采选及冶炼等工业活动源(26.81%),As和Hg主要来自污水灌溉和农药化肥施用等农业活动源(14.68%),Cr、Ni、Co和Mo主要来自土壤母质源(24.41%),Mn和Fe主要来自铁矿石开采和交通运输源(16.39%),Al和Ca主要来自矿石风化源(17.72%)。
Positive matrix factorization （PMF）is widely used to apportion the sources of heavy metals in soils even when the sources of the pollution are unknown. However, PMF is sensitive to the data of receptor samples; thus, the results may vary significantly. To evaluate the availability of PMF in classifying heavy metal sources in soil, this study investigated two factors：The composition of elements and anomalous data-using soil samples collected from Shuikoushan lead-zinc ore farmland in Hunan Province. By changing the composition of the elements （whether crustal elements were added or not） and the composition of samples （whether the anomalous data were removed or not）, four datasets were produced and used to compare the differences in the results. When two samples were removed from the dataset based on the detection of anomalies, the source profiles did not change but the contribution rates of each source to each element varied significantly. After six species of crustal elements of the samples were added to the statistical analysis, both the source profiles and contribution rates changed. The anomalous data had a much smaller influence, and the results of PMF were more stable and easy to explain when the crustal elements were included. Therefore, this study suggested that crustal elements, in addition to the eight species of heavy metals, should be determined for soil samples. Based on the documents and investigations on site, five sources were identified：Pb, Zn, Cd, and Sb came mainly from industrial activities, such as lead-zinc ore beneficiation and smelting （contribution rate of 26.81%）; As and Hg were mainly from agricultural activities, such as sewage irrigation and chemical fertilizer application （14.68%）; Cr, Ni, Co, and Mo were found mainly in soil parent material （24.41%）; Mn and Fe came mainly from iron ore mining and transportation （16.39%）; and Al and Ca were mainly from the weathering of ore （17.72%）.
作者
魏迎辉
李国琛
王颜红
张琪
李波
王世成
崔杰华
张红
周强
WEI Ying-hui;LI Guo-chen;WANG Yan-hong;ZHANG Qi;LI Bo;WANG Shi-cheng;CUI Jie-hua;ZHANG Hong;ZHOU Qiang(Institute of Applied Ecology,Chinese Academy of Sciences,Shenyang 110016,China;University of Chinese Academy of Sciences,Beijing 100049,China;Liaoning Province Engineering Research Center for Agro-products Environment and Quality Control Technology,Shenyang 110016,China)
出处
《农业环境科学学报》
CAS
CSCD
北大核心
2018年第11期2549-2559,共11页
Journal of Agro-Environment Science
基金
国家重点研发计划项目(2016YFD0800303)
农业部国家风险评估国家农产品质量安全风险评估项目(GJFP201601306)
沈阳市科技局科技创新平台建设计划项目(17-194-1-00)~~
关键词
重金属
源解析
受体模型
地壳元素
异常值
heavy metals
source apportionment
receptor model
crustal elements
anomalous data
