Impact of Data Density and Geostatistical Simulation Technique on the Estimation of Residence Times in a Synthetic Two-dimensional Aquifer

Connectivity patterns of heterogeneous porous media are important in the estimation of groundwater residence time distributions (RTDs). Understanding the connectivity patterns of a hydraulic conductivity ( \(K\) ) field often requires knowledge of the entire aquifer, which is not practical. As such, the method used to estimate unknown \(K\) values using known \(K\) values is important. This study investigates how varying levels of conditioning data and four simulation techniques, one multi-Gaussian and three multi-point, are able to recreate key \(K\) field features and connectivity patterns of a synthetic two-dimensional bimodal distributed ln( \(K\) ) field with highly connected high \(K\) features. These techniques are then assessed in the context of RTD estimation. It was found that the multi-Gaussian technique presented a bias towards earlier travel times with increased conditioning data. This was due to the inability of the method to recreate multiple scales of connecting features. Of the multi-point methods investigated, the facies method was unable to predict early arrival times. The use of a continuous variable training image produced good fits to the observed residence time distribution with a high number of conditioning points. The ability of the methods to predict the shape of residence time distributions appears to be related to their ability to reproduce the connection patterns of higher \(K\) features.