Geostatistical analysis and inverse modeling of the upper Santa Cruz Basin, Arizona

Abstract

Data from the Upper Santa Cruz basin are analyzed to determine parameters for input into a groundwater model. The geostatistical interpolation and averaging technique of kriging are used to obtain prior estimates of log- transmissivity. These estimates are enhanced with specific-capacity data by means of regression and co-kriging. Parameters of the log- transmissivity semi-variogram model are improved using a maximum likelihood technique. Hydraulic head measurements are kriged in two ways, by universal kriging and by using an iterative generalized least squares method of semi-variogram and drift estimation. An inverse method for steady state conditions is used to estimate optimum transmissivity and mountain front recharge for modeling purposes. Input includes the prior transmissivity estimates, their covariance, and steady state head data. Results are judged on six criteria: (1) log-likelihood function, (2) head objective function, (3) mean weighted head residual, (4) mean squared weighted head residual, (5) residual coefficient of variation, and (6) a Chi-square fit. None of these criteria is sufficient to define a best estimate, but taken together, they point toward a possible optimum set of parameters.