TY - JOUR

T1 - Using a bias aware EnKF to account for unresolved structure in an unsaturated zone model

AU - Erdal, D.

AU - Neuweiler, I.

AU - Wollschläger, U.

PY - 2013/11/27

Y1 - 2013/11/27

N2 - When predicting flow in the unsaturated zone, any method for modeling the flow will have to define how, and to what level, the subsurface structure is resolved. In this paper, we use the Ensemble Kalman Filter to assimilate local soil water content observations from both a synthetic layered lysimeter and a real field experiment in layered soil in an unsaturated water flow model. We investigate the use of colored noise bias corrections to account for unresolved subsurface layering in a homogeneous model and compare this approach with a fully resolved model. In both models, we use a simplified model parameterization in the Ensemble Kalman Filter. The results show that the use of bias corrections can increase the predictive capability of a simplified homogeneous flow model if the bias corrections are applied to the model states. If correct knowledge of the layering structure is available, the fully resolved model performs best. However, if no, or erroneous, layering is used in the model, the use of a homogeneous model with bias corrections can be the better choice for modeling the behavior of the system. Key Points Accounting for unresolved subsurface structures using bias aware EnFK Predicting average system behavior using local observations Using EnKF for data assimilation with a nonlinear unsaturated zone model

AB - When predicting flow in the unsaturated zone, any method for modeling the flow will have to define how, and to what level, the subsurface structure is resolved. In this paper, we use the Ensemble Kalman Filter to assimilate local soil water content observations from both a synthetic layered lysimeter and a real field experiment in layered soil in an unsaturated water flow model. We investigate the use of colored noise bias corrections to account for unresolved subsurface layering in a homogeneous model and compare this approach with a fully resolved model. In both models, we use a simplified model parameterization in the Ensemble Kalman Filter. The results show that the use of bias corrections can increase the predictive capability of a simplified homogeneous flow model if the bias corrections are applied to the model states. If correct knowledge of the layering structure is available, the fully resolved model performs best. However, if no, or erroneous, layering is used in the model, the use of a homogeneous model with bias corrections can be the better choice for modeling the behavior of the system. Key Points Accounting for unresolved subsurface structures using bias aware EnFK Predicting average system behavior using local observations Using EnKF for data assimilation with a nonlinear unsaturated zone model

KW - bias correction

KW - EnKF

KW - model error

KW - parameter estimation

KW - unresolved structure

KW - unsaturated zone

UR - http://www.scopus.com/inward/record.url?scp=84896710135&partnerID=8YFLogxK

U2 - 10.1002/2012WR013443

DO - 10.1002/2012WR013443

M3 - Article

AN - SCOPUS:84896710135

VL - 50

SP - 132

EP - 147

JO - Water resources research

JF - Water resources research

SN - 0043-1397

IS - 1

ER -