基于万维网大数据的农药场地土壤污染快速预测方法研究
作者:
作者单位:

1.中国科学院南京土壤研究所;2.土壤与农业可持续发展国家重点实验室中国科学院南京土壤研究所

作者简介:

通讯作者:

中图分类号:

基金项目:

国家重点研发计划专项(2018YFC1800104);科技基础性工作专项(2015FY110700-S2)


Research on the Method of Rapid Prediction of Soil Pollution in Pesticide Polluted-Sites Based on Network Big Data
Author:
Affiliation:

Institute of Soil Science,Chinese Academy of Sciences

Fund Project:

the Special Project of the National Key Research and Development Program (No.2018YFC1800104); Special Project of the National Science and Technology Basic Work (No.2015FY110700-S2)

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    及时高效预测和筛查潜在农药污染场地对环境污染风险管控具有重要意义。基于万维网公开的46个农药场地样本数据,利用五分制层次分析法建立农药场地土壤污染快速预测指标体系,包括产品特征、局部气候条件、土壤属性和场地生产特征4个因素及其相应的产品毒性、持久性、气温、降水、风速、光照、土壤质地、土壤pH、有机质含量、生产时间和闲置时间11个特征指标。其中,农药场地生产时间、产品毒性及其持久性指标五分制分级后与农药场地土壤污染均存在显著线性相关性,三个指标不同组合对场地土壤污染的线性综合预测精度小于65%,而基于11个指标的机器学习方法综合预测精度为82%,但存在污染场地严重漏判问题。以综合评价指数值P≥0.6作为农药场地土壤污染的预测阈值,五分制层次分析法综合预测精度达到91%,优于线性预测以及机器学习方法,具有关键数据需求少、预测快速高效特点,体现 “宁严勿漏”的预测原则,可用于各类型农药场地的土壤污染筛查。

    Abstract:

    【Objective】 Predicting and screening potential pesticide-contaminated sites timely and efficiently is important for controlling environmental pollution. 【Method】 Based on 46 pesticide sites samples published on the World Wide Web, the index system and method for rapid prediction of soil pollution in pesticide sites was established by a five-score analytic hierarchy process. 【Result】The predictive system was constituted with four factors: product characteristics, local climatic conditions, soil properties and site characteristics, including 11 characteristic indicators: product toxicity, persistence, temperature, precipitation, wind speed, light, soil texture, soil pH, organic matter content, production time and idle time. There is a significant linear correlation between the three indicators: production time level, product toxicity and durability level, and the soil pollution of the pesticide sit. The linear comprehensive prediction accuracy of the three indicators is less than 65%. Also, the comprehensive judgment accuracy of the machine learning method combining 11 indicators is 82%, but all of them have significant limitations as they missed classified the severity of the contaminated sites. 【Conclusion】The comprehensive evaluation index value p ≥ 0.6 is used as the prediction threshold of soil pollution in pesticide sites. The accuracy of the comprehensive prediction of the five component AHP is 91%, which is better than linear prediction and the machine learning method. It has the characteristics of low demand for key data, fast and efficient diagnosis, and reflects the principle of ‘Implemented to the strictest standards without leaving a contaminated site”. It can be used for pre-diagnosis of soil pollution in various types of pesticide sites.

    参考文献
    相似文献
    引证文献
引用本文

王 鑫,于东升,马利霞,陆晓松,陈 洋,冯凯月.基于万维网大数据的农药场地土壤污染快速预测方法研究[J].土壤学报,DOI:10.11766/trxb202012300343,[待发表]
WANG Xin, YU Dongsheng, MA Lixia, LU Xiaosong, CHEN Yang, FENG Kaiyue. Research on the Method of Rapid Prediction of Soil Pollution in Pesticide Polluted-Sites Based on Network Big Data[J]. Acta Pedologica Sinica, DOI:10.11766/trxb202012300343,[In Press]

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2020-06-30
  • 最后修改日期:2021-08-20
  • 录用日期:2021-08-27
  • 在线发布日期: 2021-09-01
  • 出版日期: