Geostatistical Co-Estimation of Ground Water Flow Parameters.

摘要

The parameter estimation is a crucial phase of any ground water flow investigation, which determines the success or failure of the prediction process. The commonly used parameter-estimation approaches are based on various inverse procedures that in many instances prove to be inadequate. The report presents an advanced alternative approach which is based on the geostatistical coestimation technique, known as co-kriging. The investigated parameters are: aquifer transmissivity, piezometric head, and residual, which is defined as the computed minus the measured water level at a specified location. Values of computed head were obtained by using the MODular Finite Element model (MODFE) of two-dimensional ground water flow. The data set was derived from a regional study of the Upper Floridan aquifer in southern Georgia. The initial results indicate that the log-transformed transmissivity and the uncalibrated residual are suitable for co-estimation. Direct kriging provides estimated maps for these variables, which are unbiased linear estimates with minimum variance of estimation. Considering the high cost of field tests in ground water investigations, the above more efficient co-estimation process offers an attractive alternative for parameter evaluation than previously used inverse procedures.