文章摘要
郭新蕾,赵玉杰,刘潇威,周其文,王夏晖,李志涛,朱智伟,张铁亮,王祖光,张璠,孙扬.基于空间聚类与随机森林的稻米富集镉影响因素筛选研究[J].农业环境科学学报,2019,38(8):1794-1801.
基于空间聚类与随机森林的稻米富集镉影响因素筛选研究
Screening for factors affecting rice uptake of cadmium based on spatial clustering and random forests
投稿时间:2019-01-23  
DOI:10.11654/jaes.2019-0096
中文关键词: 稻米    空间聚类  随机森林  相关性分析
英文关键词: rice  cadmium  spatial cluster  random forest  correlation analysis
基金项目:国家重点研发计划项目(2016YFD0800307-4);国家科技支撑计划项目(2015BAD06B03-1);国家农产品质量安全风险评估重大专项(GJFP2019012)
作者单位E-mail
郭新蕾 农业农村部农产品质量安全环境因子控制重点实验室, 天津 300191
农业农村部环境保护科研监测所, 天津 300191 
 
赵玉杰 农业农村部农产品质量安全环境因子控制重点实验室, 天津 300191 yujiezhao@126.com 
刘潇威 农业农村部农产品质量安全环境因子控制重点实验室, 天津 300191 xwliu2006@163.com 
周其文 农业农村部农产品质量安全环境因子控制重点实验室, 天津 300191  
王夏晖 生态环境部环境规划院, 北京 100012  
李志涛 生态环境部环境规划院, 北京 100012  
朱智伟 中国水稻研究所, 杭州 310006  
张铁亮 农业农村部农产品质量安全环境因子控制重点实验室, 天津 300191  
王祖光 农业农村部农产品质量安全环境因子控制重点实验室, 天津 300191  
张璠 农业农村部农产品质量安全环境因子控制重点实验室, 天津 300191  
孙扬 农业农村部环境保护科研监测所, 天津 300191  
摘要点击次数: 2142
全文下载次数: 1971
中文摘要:
      以湖南省某县为例,通过空间聚类筛选研究热点区,并对热点区筛选的土壤样品进行多参数检测获取分析数据,通过随机森林回归锁定影响稻米富集Cd的主控因子。结果表明:研究区高高、高低聚类区稻米Cd富集系数差异达9倍,低低、低高聚类区稻米Cd富集系数差异达16倍;影响稻米富集Cd的最主要因素为Ca、pH、Mn,其次是Fe、Si,再次是Zn、DOC、Cl、K、P、Mg、S、Cd、Na、Cu,影响程度逐渐减小,最后是SOM影响程度最小。通过调控Ca含量可显著影响土壤pH,进而影响稻米对Cd的吸收。随着土壤Fe、Mn的升高稻米Cd含量呈指数下降。土壤Si含量与土壤Fe、pH等均呈负相关,Si含量增加,稻米Cd含量相应升高。有机质对稻米富集Cd的影响不显著。研究表明,双变量空间聚类与随机森林相结合可以有效筛选出稻米富集Cd的土壤主控因子,从而为稻米Cd污染的修复治理及污染源解析热点筛查提供基础支撑。
英文摘要:
      Understanding the main soil-related factors affecting the accumulation of Cd in rice samples is key to regulating rice Cd concentrations. Regression predictions based on point data often do not reflect regional differences, and most of the main controlling factors show clustering and spatial differences. Based on this principle, a county in Hunan Province was selected as an example and spatial clustering was used to screen hotspots for various soil factors. Soil samples screened in the hotspots were analyzed using multi-parameter detection and the main controlling factors affecting the accumulation of Cd in rice samples were determined through random forest regression. There was up to a 10 fold difference between the Cd enrichment coefficients of rice samples in the high-high and high-low clustering areas, and up to a 17 fold difference between the low-low and low-high clustering areas. In the study area, the critical factors affecting Cd accumulation in rice samples were Ca, pH, and Mn, followed by Fe and Si, and then Zn, DOC, Cl, K, P, Mg, S, Cd, Na, and Cu with a gradually decreasing degree of influence; the influence of SOM was lowest. Regulating the Ca concentration in the soil had a significant impact on soil pH values and subsequently on the uptake of Cd by rice. With an increase in Fe and Mn concentrations in the soil, the rice Cd concentrations decreased exponentially. The Si concentration in the soil was negatively correlated with soil pH values and Fe concentration, and positively correlated with Cd concentration in rice. Organic matter had no significant effect on Cd accumulation in rice. This study confirmed that the combination of bivariate spatial clustering and random forest regression can effectively determine the main factors of soils controlling Cd accumulation in rice samples. Such studies can provide fundamental support for the remediation and control of Cd accumulation in rice and hotspot screening of pollution sources.
HTML    查看全文   查看/发表评论  下载PDF阅读器