AN ITERATIVE COKRIGING-LIKE TECHNIQUE FOR GROUNDWATER-FLOW MODELING

The importance of field heterogeneities in ground-water pollution prob lems has been widely recognized during the last few decades. To addres s the impact of field heterogeneities on ground-water flow and solute transport, many different stochastic methods have been developed. Amon g all these stochastic methods kriging is the most popular one used by many practitioners to interpolate and extrapolate measured transmissi vity data. However, hydraulic head measurements are generally more abu ndant than transmissivity data. Therefore, the cokriging technique whi ch utilizes both the head and transmissivity measurements to estimate transmissivity and/or hydraulic head distributions has also received m uch attention in recent years. Classical cokriging relies on a linear predictor approach and uses covariance and cross covariance functions derived from a first-order approximation. Consequently, it often resul ts in head and transmissivity fields that can produce unacceptable vel ocity distributions. In this paper, we develop an iterative method whi ch combines classical cokriging and a numerical flow model to obtain o ptimum estimates of transmissivity and head distributions and to allev iate the limitations of classical cokriging. Through several numerical examples, we demonstrate that this method is superior to the classica l cokriging method in terms of producing mass conservative velocity fi elds. In addition, results of the study also indicate that hydraulic h ead measurements can improve for our prediction of ground-water flow d irections and paths in aquifers significantly.