康学远, 施小清, 邓亚平, 廖凯华, 吴吉春. 基于EnKF融合地球物理数据刻画含水层非均质性[J]. 水科学进展, 2018, 29(1): 40-49. DOI: 10.14042/j.cnki.32.1309.2018.01.005
引用本文: 康学远, 施小清, 邓亚平, 廖凯华, 吴吉春. 基于EnKF融合地球物理数据刻画含水层非均质性[J]. 水科学进展, 2018, 29(1): 40-49. DOI: 10.14042/j.cnki.32.1309.2018.01.005
KANG Xueyuan, SHI Xiaoqing, DENG Yaping, LIAO Kaihua, WU Jichun. Assimilation of hydrogeophysical data for the characterization of subsurface heterogeneity using Ensemble Kalman Filter (EnKF)[J]. Advances in Water Science, 2018, 29(1): 40-49. DOI: 10.14042/j.cnki.32.1309.2018.01.005
Citation: KANG Xueyuan, SHI Xiaoqing, DENG Yaping, LIAO Kaihua, WU Jichun. Assimilation of hydrogeophysical data for the characterization of subsurface heterogeneity using Ensemble Kalman Filter (EnKF)[J]. Advances in Water Science, 2018, 29(1): 40-49. DOI: 10.14042/j.cnki.32.1309.2018.01.005

基于EnKF融合地球物理数据刻画含水层非均质性

Assimilation of hydrogeophysical data for the characterization of subsurface heterogeneity using Ensemble Kalman Filter (EnKF)

  • 摘要: 含水层非均质性的刻画是模拟地下水中污染物运移的关键。以渗透系数为研究对象,构建了综合集合卡尔曼滤波方法、有效电阻率模型与地下水运移模型的同化框架,通过融合地球物理观测数据与污染物浓度观测数据来推估渗透系数的空间分布。基于理想算例,验证了该同化框架刻画含水层非均质渗透系数场的有效性,并针对不同初始参数信息与观测类型对比了耦合与非耦合水文地球物理方法的适用性。研究结果表明:基于集合卡尔曼滤波方法同化多种类型的观测数据,可有效地推估非均质参数空间分布。当初始信息较准确时,耦合方法的参数推估精度更高;初始信息存在偏差时,非耦合方法有更好的同化效果。由于非耦合方法计算成本较低且对初始信息缺失时适用性更强,在实际应用中可先基于非耦合方法初步估计参数,再利用耦合方法进一步提高参数推估精度。融合多种类型观测数据可有效提高参数推估效果。

     

    Abstract: Characterization of spatial variability of hydrogeologic properties is the key to simulate and predict the fate and transport of contaminants in the subsurface. In this study, we present a sequential data assimilation framework to estimate the heterogeneous saturated hydraulic conductivity fields through the assimilation of Electrical Resistivity Tomography (ERT)-monitored data and groundwater flow/transport observation data. This framework is integrated Ensemble Kalman Filter (EnKF), groundwater flow/transport models and effective medium resistivity model. To test the performance of the framework, synthetic cases of contaminant transport are reconstructed. We compare the performance of the coupled and uncoupled methods. The factors to control the performance of coupled and uncoupled methods are also discussed in a number of different scenarios. Results showed that both methods can effectively estimate the spatial distribution of hydraulic conductivity via time-lapse ERT-monitored data. The coupled method performs better than the uncoupled one when the prior statistics are close to real field. Meanwhile, the uncoupled method is more robust when the prior statistics is biased. The accuracy of estimated heterogeneous parameter field could be improved when integrating of multiple type observations including ERT-monitored data and a few observations of groundwater flow/transport model (i. e., concentration). As the uncoupled method requires a small computational effort compared to the coupled one, it is suggested to use the uncoupled method as a preliminary inversion before refining the results with a fully coupled method. We conclude that integrating multiple types of observations is recommended to improve the ability to delineate subsurface heterogeneity.

     

/

返回文章
返回