Geostatistical analysis and stochastic modeling of the Tajo basin aquifer, Spain

Abstract

The Tajo basin, Spain, contains an extensive alluvial aquifer. Kriging and condition simulation have been used to estimate logtransmissivity, hydraulic head, and source terms and their estimation errors given measured specific capacity, transmissivity, and waterlevel data from this aquifer. The geostatistical technique of kriging was used to calculate estimates of spatially-averaged log-transmissivities and point estimates of hydraulic head, and to compute the covariance matrices of estimation error. An iterative generalized least squares procedure allowed for simultaneous determination of the trend and head residual semi-variogram prior to kriging of the heads. A new application of conditional simulation was used to generate independent realizations of hydraulic head and log-transmissivity. Source terms were calculated from these realizations using the flow equation. Mean source terms and their estimation uncertainty were computed after 600. realizations, This application of conditional simulation appears to be useful in estimating source terms in areas where they are unmeasured and not well known.