Groundwater parameter estimation via ensemble kalman filter with covariance localization
Ensemble Kalman Filter (EnKF) has been introduced to estimate groundwater parameters and examined by some numerical experiments in literature. However, it suffers from spurious correlations resulting from sampling noises when approximating covariances from a small-size ensemble due to limitation of computation cost. It is important to deal with such inaccuracy in order to maintain filter process robust and reasonable. Covariance localization is one of suggested methods to modify EnKF using a small ensemble, and used to rectify spurious correlations which weaken the EnKFs capability to estimate uncertainty correctly. In this study, covariance localization is implemented by adding distance-related weight to covariance in the way of Schur product. A two-dimensional synthetic example is constructed to demonstrate the effect of covariance localization. It is found that EnKF with localization can capture reliable both a mean and a variance of hydraulic conductivity field with improved efficiency, as well as avoiding filter divergence to a large extent.
T.C.Nan J.C.Wu
State Key Laboratory of Pollution Control and Resources Reuse,P.R.China Department of Hydro Sciences State Key Laboratory of ollution ontrol and Resources Reuse,P.R.China Department of HydroSciences, N
国际会议
武汉
英文
51-54
2009-09-20（万方平台首次上网日期，不代表论文的发表时间）