| 基于自组织神经网络的污染场地多监测指标相关性分析 |
| |
| 引用本文： | 马春龙,施小清,许伟伟,任静华,王佩,吴吉春.基于自组织神经网络的污染场地多监测指标相关性分析[J].水文地质工程地质,2021,48(3):191-202. |
| |
| 作者姓名： | 马春龙 施小清 许伟伟 任静华 王佩 吴吉春 |
| |
| 作者单位： | 1.表生地球化学教育部重点实验室/南京大学地球科学与工程学院，江苏 南京 210023 |
| |
| 基金项目： | 自然资源部国土（耕地）生态监测与修复工程技术创新中心开放课题；国家自然科学基金项目（41672229） |
| |
| 摘 要： | 为查明场地污染分布特征，需对场地土壤和地下水进行钻探取样，按规范的检测指标进行逐一测试。在初查和详查阶段将获得大量的土壤和地下水污染数据，数据样本数量大、监测指标多，数据结构复杂，如何从场地大数据中提取价值信息已成为研究热点。以某有机污染场地为例，基于自组织映射神经网络（SOM）和K均值算法开展大数据分析，深入探讨地下水和土壤中各污染指标间的相关性。结果表明:（1）基于自组织映射神经网络的大数据分析可快速挖掘复杂多维的污染场地监测数据，有效完成关键信息的提取；（2）地下水中污染检出指标存在显著的聚类特征，同一聚类中的污染指标具备相似的空间分布特征。对场地污染物检测采取先分类后分级的优化筛选策略，减少污染物检测指标数目，从而有效降低场地检测费用；（3）土壤和地下水中污染检出指标存在良好的空间相关性，这与该污染场地地下水渗流速度缓慢有关。土壤和地下水污染检出指标空间分布的相关性，有助于场地污染源的追溯。
|
| 关 键 词： | 自组织映射神经网络 污染场地 多监测指标 相关性分析 土壤 地下水 |
| 收稿时间： | 2020-08-01 |
Correlation analysis of multiple monitoring indicators of contaminated site based on self-organizing map |
| |
| Institution: | 1.Key Laboratory of Surficial Geochemistry, Ministry of Education/School of Earth Sciences and Engineering, Nanjing University, Nanjing, Jiangsu 210023, China2.Engineering Technology Innovation Center for Ecological Monitoring and Restoration in Arable Land, Ministry of Natural Resources/Geological Survey of Jiangsu Province, Nanjing, Jiangsu 210018, China3.Changzhou Research Academy of Environment Sciences, Changzhou, Jiangsu 213022, China |
| |
| Abstract: | In order to investigate the distribution characteristics of pollutants at contaminated sites, it is necessary to collect soil and groundwater samples by drilling and test them by the standard procedure. In the preliminary and detailed investigation, a large amount of data of soil and groundwater pollution will be obtained. These data are often characterized by large sample size, multiple monitoring indicators and complex data structures, and how to extract valuable information from the big data has become an important research issue. This study takes an organic contaminated site as an example, and carries out big data analytics by using self-organizing map (SOM) and k-means algorithm to explore the correlation between each organic pollution indicator of groundwater and soil. The results show that (1) the big data analytics based on self-organizing map can rapidly mine the complicated multi-dimensional monitoring data of contaminated site, and extract key information effectively. (2) The pollution indicators in groundwater are characterized by significant clustering, and the indicators in the same cluster are of similar spatial distribution characteristics. In view of this, a screening strategy may classify the indicators first and then rank them, and can be adopted at contaminated site to reduce the number of pollution indicators detected and finally save the cost of site detection. (3) The pollution indicators in soil and groundwater also have strong spatial correlation, which is mainly due to the slow seepage velocity of groundwater. According to the correlation of the spatial distribution of pollution indicators in soil and groundwater, it is helpful to trace the pollution sources at contaminated sites. |
| |
| Keywords: | |
|
| 点击此处可从《水文地质工程地质》浏览原始摘要信息 |
| 点击此处可从《水文地质工程地质》下载免费的PDF全文 |
|
