地表水水质监测时序数据插补方法比较及应用  

Comparison and Application of Imputation Methods for Temporal Data in Surface Water Quality Monitoring

作　　者：高海燕 刘畅 马文娟 GAO Haiyan;LIU Chang;MA Wenjuan(School of Statistics and Data Science,Lanzhou University of Finance and Economics,Lanzhou 730020,China;Key Laboratory of Digital Economy and Social Computing Science,Gansu Province,Lanzhou 730020,China)

机构地区：[1]兰州财经大学统计与数据科学学院,甘肃兰州730020 [2]甘肃省数字经济与社会计算科学重点实验室,甘肃兰州730020

出　　处：《水文》2024年第5期61-67,共7页Journal of China Hydrology

基　　金：国家社会科学基金项目(19XTJ002);甘肃省自然科学基金项目(23JRRA1186);兰州财经大学科研项目(Lzufe2023C-005)。

摘　　要：国控地表水水质监测数据的完整性和准确性对于保障公众健康、保护生态环境、支持水资源管理等具有重要意义。对比分析2种单一插补(均值插补、KNN)、7种多重插补(MF、MICE、blasso、norm、norm.boot、norm.nob、ri)等9种方法在地表水水质监测数据中的适用性和有效性,针对2020—2022年天津市武清北运河土门楼断面的7个地表水水质指标进行以上插补方法的性能评估,并对相同指标的实际缺失数据进行实证分析。结果表明:在不同缺失率下,blasso多重插补方法的插补效果更优,它能够最大程度地利用各指标的辅助变量以及先验信息提高插补精度,且收敛速度快,插补时间可控。The integrity and accuracy of national surface water quality monitoring data are of great significance for safeguarding public health,protecting ecological environment and supporting water resources management.Comparative analysis of the applicability and effectiveness of nine methods including two kinds of single imputation(Mean Imputation and KNN) and seven kinds of multiple imputation methods(MF,MICE,blasso,norm,norm.boot,norm.nob,ri) in surface water quality monitoring data.The imputation methods performance of 7 surface water quality indicators in Tumenlou section of Beiyun River,Wuqing District,Tianjin from 2020 to 2022 was evaluated by above 9 imputation methods,and the actual missing data of the same indicators were analyzed empirically.The results showed that the blasso multiple imputation method produced superior imputation results.It maximizes the utilization of auxiliary variables and prior information of various indicators to improve imputation accuracy.Additionally,Bayesian Lasso has a fast convergence speed and controllable imputation time.

关 键 词：地表水 水质监测数据 多重插补 blasso 

分 类 号：P333[天文地球—水文科学]