Comprehensive geostatistical based parameter optimization and inverse modeling of North Avra Valley, Arizona

Abstract

Geostatistical based optimization was applied to the North A vra Valley ground water model to estimate the transmissivity field and boundary conditions that minimize the difference of the modeled and measured head. The Sequential Self-Calibration (SSC) method was used for the inverse modeling and optimization. SSC is an iterative technique that combines geostatistics with an optimization routine to condition both transmissivity and head fields to measured data. Two calibration methodologies were compared. In the first, the inflow and outflow boundary conditions are adjusted to minimize head residuals, using the uniform geometric mean transmissivity field and the subsequent SSC calibrated transmissivity field is based on those initial boundary conditions. The second method ran the model independent optimization software PEST in series with SSC. This approach calibrates the inflow and outflow boundary conditions and transmissivity field iteratively against the head residuals. As a consequence, the inflow and outflow boundary conditions are optimized against the final geostatistical based transmissivity field used in the model. The serial PEST-SSC calibration method produces consistently better results with respect to head residuals, by an average of 27 .1 percent. The resulting calibrated transmissivity fields of both methods were compared using stochastic error analysis, showing similar results for both methods. A final model run was done employing the PEST-SSC method for a more detailed analysis. This resulted in a relative error ( O'head residuals I head-range) of only 1.5 percent.