Stochastic estimation of facies using ground penetrating radar data

Explicitly defining large-scale heterogeneity is a necessary step of groundwater model calibration if accurate estimates of flow and transport are to be made. In this work, neural networks are used to estimate radar facies probabilities from ground penetrating radar (GPR) images, yielding stochastic facies-based models that honour the large-scale architecture of the subsurface. For synthetic GPR images, a neural network was able to correctly identify radar facies with an accuracy of approximately 90%. Manual interpretation of a set of 450 MHz GPR field data from the Borden aquifer resulted in the identification of four radar facies. Of these, a neural network was able to identify two facies with an accuracy of near 80% and one with an accuracy of 44%. The neural network was not able to identify the fourth facies, likely due to the choice of defining facies characteristics. Sequential indicator simulation was used to generate facies realizations conditioned to the radar facies probabilities. Numerical simulations indicate that significant improvements in the prediction of solute transport are possible when GPR is used to constrain the facies model compared to using well data alone, especially when data are sparse.