HR: 12:05h
AN: H11I-08 [PDF]
TI: A Coupled Zonation-Kriging Method for Parameter Structure Identification in Groundwater Modeling: A
Case Study of a Seawater Intrusion Problem, the Alamitos Barrier in Southern California Coastal
Plain
AU: * Tsai, F T
EM: ftsai@lsu.edu
AF: Louisiana State University, Civil and Environmental Engineering Dept.
3507 CEBA, Baton Rouge, LA 70803
AU: Sim, Y
EM: ysim@ladpw.org
AF: LA County Dept. of Public Works, Water Resources Division, 900 S. Fremont Ave, Alhambra, CA 91803
AU: Yeh, W W
EM: williamy@ucla.edu
AF: University of California, Los Angeles, Civil and Environmental Engineering Dept.
5731 Boulder Hall, Los Angeles, CA 90095
AB:
In this research, we propose a coupled zonation-kriging method for parameter heterogeneity characterization as well as
parameter structure identification. With a set of distinct point measurements over a field, on one hand, a zonation structure
for the field can be created based on a Voronoi tessellation method. Each Voronoi cell contains one sampled point and has
the corresponding measured value. On the other hand, a kriging method can be used to characterize a continuous parameter
distribution. In this study, we only use local information for the kriging method to estimate the parameter value at an
unsampled site. First, we apply Voronoi tessellation and Delaunay triangulation to define a set of natural neighbors among
the sampled points for an unsampled site. Each site has its unique natural neighbors. Then the kriging method uses the local
information provided from these natural neighbors for the estimation. This is called the natural neighbor kriging (NNK)
method. We combine the zonation method and the NNK method together as one parameterization method by introducing a set of
weighting coefficients to the measurement points. The zonation-NNK method unifies zonation and kriging methods and generates
a distribution between a pure zone structure and a continuous structure over a set of weighting coefficients. It shows
greater flexibility in manipulating spatial distribution and spatial optimization. When a non-smooth field is investigated,
the zonation-NNK outperforms all other parameterization schemes. For the inverse problem, we identify parameter heterogeneity
with the zonation-NNK method by seeking the optimal weighting coefficients while minimizing the fitting residual of
observations. We demonstrate the developed methodology by a case study of the Alamitos barrier for the seawater intrusion
problem in Southern California Coastal Plain. The unknown distributed parameter is the hydraulic conductivity of five
aquifers. We adopted the FEMWATER as a model to simulate the density-dependent coupled flow and transport in the coastal
aquifers. For the given set of head and salinity concentration observations, we have identified the hydraulic conductivity
field in three dimensions.
DE: 1800 HYDROLOGY
DE: 1829 Groundwater hydrology
DE: 1831 Groundwater quality
DE: 1832 Groundwater transport
DE: 3260 Inverse theory
SC: Hydrology [H]
MN: 2003 Fall Meeting