Li Jiuhui, Wu Zhengfang, He Hongshi, Lu Wenxi
Key Laboratory of Geographical Processes and Ecological Security in Changbai Mountains, Ministry of Education, School of Geographical Sciences, Northeast Normal University, Changchun, 130024, China.
School of Natural Resource, University of Missouri, Columbia, MO, USA.
Environ Sci Pollut Res Int. 2022 Dec;29(60):90081-90097. doi: 10.1007/s11356-022-21974-5. Epub 2022 Jul 21.
The location and release history of groundwater contaminant sources (GCSs) are usually unknown after groundwater contamination is detected, thereby greatly hindering the design of contamination remediation schemes and contamination risk assessments. Many previous studies have used prior information such as the observed contaminant concentrations (OCC) to obtain information of GCSs, and various methods have been proposed for identifying GCSs, including simulation optimization (S/O) and ensemble Kalman filter (EnKF) methods. For the first time, the present study compared the suitability of the S/O and EnKF methods for GCSs identification based on two case studies by specifically considering the calculation time and effectiveness of GCS identification. The results showed that EnKF could reduce the calculation time required by more than 62% compared with S/O. However, the time saved did not compensate for the poor accuracy of the GCSs identification results. When the simulated contaminant concentrations (SCC) were used for GCSs identification, the MRE of the identification results with the S/O and EnKF methods were 2.79% and 5.09% in case one, respectively, and were 4.75% and 6.72% in case two. When the OCC were used for GCSs identification, the MRE of the identification results with the S/O and EnKF methods were 27.77% and 110.74% in case one, respectively, and 27.53% and 60.61% in case two. The identification results obtained using the EnKF method were not credible and the superior performance of the S/O method was obvious, thereby indicating that the EnKF method is much less suitable for actual GCSs identification compared with the S/O method.
在检测到地下水污染后，通常不清楚地下水污染源（GCSs）的位置和释放历史，这极大地阻碍了污染修复方案的设计和污染风险评估。许多先前的研究利用诸如观测到的污染物浓度（OCC）等先验信息来获取GCSs的信息，并提出了各种识别GCSs的方法，包括模拟优化（S/O）和集合卡尔曼滤波器（EnKF）方法。本研究首次通过具体考虑GCSs识别的计算时间和有效性，基于两个案例研究比较了S/O和EnKF方法对GCSs识别的适用性。结果表明，与S/O相比，EnKF可将所需的计算时间减少62%以上。然而，节省的时间并不能弥补GCSs识别结果准确性差的问题。当使用模拟污染物浓度（SCC）进行GCSs识别时，在案例一中，S/O和EnKF方法识别结果的平均相对误差（MRE）分别为2.79%和5.09%，在案例二中分别为4.75%和6.72%。当使用OCC进行GCSs识别时，在案例一中，S/O和EnKF方法识别结果的MRE分别为27.77%和110.74%，在案例二中分别为27.53%和60.61%。使用EnKF方法获得的识别结果不可信，S/O方法的优越性能明显，这表明与S/O方法相比，EnKF方法不太适合实际的GCSs识别。
