Identification of Parameter Structure in Groundwater Inverse Problem

This paper develops an optimization method that can be used to identify parameter structure as well as its values in connection with the inverse problem in groundwater modeling. The complexity of the parameter structure is determined by the quantity and quality of observation data. Output least square error is used as the criterion of identification. Sensitivity coefficients at finite element nodes are evaluated by the variational method. The problem of identifying the parameter structure is formulated as a problem of combinatorial optimization. A new parameterization technique is presented for describing the parameter structure. It can represent various distributions of parameters from piecewise constant to continuous variability. Based on this representation, the combinatorial optimization problem for determining the parameter structure is then solved by a simple one‐at‐a‐time searching procedure. Numerical experimentations are used to demonstrate the efficiency and practicality of the proposed method.