Calibration of distributed parameter groundwater flow models viewed as a multiple‐objective decision process under uncertainty

Abstract

Owing to the deterministic nature of most groundwater flow models there has been a tendency in the past to overlook the strong element of uncertainty that invariably enters into the problem of parameter identification. It is shown that because of this uncertainty an approach based on the minimization of a single error functional does not in general lead to satisfactory results. A multiple‐objective decision process is postulated taking into account all the available information on the aquifer flow system as well as the range of environmental conditions under which the system is expected to operate in the future. According to this new approach a continuous or discrete set of alternative solutions to the identification problem is generated with the aid of mathematical programing techniques, and the decision maker is asked to apply his own value judgment in selecting a particular model structure. The method is illustrated by applying parametric linear programing to a finite element model of steady state flow in a locally anisotropic aquifer. The reliability of each parameter estimate is ascertained with the aid of a postoptimal sensitivity analysis.