Gradual conditioning of non-Gaussian transmissivity fields to flow and mass transport data: 3. Application to the Macrodispersion Experiment (MADE-2) site, on Columbus Air Force Base in Mississippi (USA)

SummaryA large-scale natural-gradient tracer experiment conducted in a highly heterogeneous aquifer at the Macrodispersion Experiment (MADE-2) site on Columbus Air Force Base in Mississippi (USA) is simulated using the gradual conditioning (GC) method. This methodology allows the stochastic inversion of hydraulic conductivity data (K), and transient piezometric (h) and solute concentration (c) measurements in a non-Gaussian framework, including soft and secondary data. Results show (i) that the GC method allows the reproduction of the heavy tailing of the tracer plume as observed in the field by using a dual-domain mass transfer approach together with conditioning to K, h and c data, in a non-Gaussian framework, (ii) a good agreement between data and simulated mass distribution at time 328 days, including the non-Gaussian plume behaviour, (iii) the necessity of using a dual-domain mass transfer approach – or other transport equation different to the advection–dispersion equation (ADE) – when treating with upscaled models regardless of what random function is used to generate the K distribution, (iv) the reduction of uncertainty results when conditioning to all available information and not only to K data, and (v) the importance of preferential flow paths on the anomalous tracer plume spreading at the MADE site. Besides, the viability of the GC method in a highly heterogeneous 3D aquifer is proven, and also its contribution to the state-of-the-art in stochastic inverse modelling.