Understanding groundwater dynamics in structurally heterogeneous aquifers

Heterogeneous geologic systems often contain discrete structural elements such as faults, fractures, and sedimentary structures like coarse-grained channel deposits. These structures are characterized by hydraulic properties that are substantially different from the properties of surrounding material. Thus, the spatial arrangement and geometry of structures represents key heterogeneity that can strongly affect groundwater dynamics. By considering relevant field data and through the use of numerical modeling, this research explores the flow impact of subsurface structures in clastic sedimentary systems. Also investigated in this work is the importance of connections between structures.;In the first part of this dissertation, a theoretical study is presented to investigate subsurface fluid flow in the presence of high-permeability (high-k) structures. A new directed percolation model is developed to generate synthetic channel networks. The model is flexible and is capable of producing networks with broadly varying geometrical properties. The channels serve as high-k structures and are embedded within a uniform, less permeable matrix material. Two-dimensional, steady-state flow simulations are performed to explore the controls on fluid flow behavior in the channel-matrix systems. Numerous systems with differing network properties are considered. The main result of this effort is a new qualitative and quantitative understanding that relates the effective permeability tensor to geometrical properties of the channel network. It is demonstrated that, for the range of network geometries and hydraulic properties considered, the effective permeability is a predictable function of the channel tortuosity, channel network fractal dimension, and the channel-matrix permeability contrast.;The second major contribution of this dissertation involves the use of an inverse modeling technique to investigate field observations of non-ideal hydraulic response. Field data are from the Lawrence Livermore National Laboratory (LLNL). The structurally heterogeneous aquifer underlying LLNL is characterized by discrete channel deposits contained within finer-grained and less permeable floodplain deposits. Aquifer-test data from the site reveal a complex response to pumping that cannot be explained using standard analytical models. To investigate how this response behavior is influenced by the sedimentary architecture, a simulation-inversion methodology is developed. The method couples a multiple-point geostatistical model with a dynamic groundwater flow model. The geostatistical model simulates the distribution of channel and floodplain material, conditional to direct data (observed geologic material) at borehole locations. The dynamic flow model simulates the effects of pumping for a given geostatistical realization of the aquifer architecture. Application of the inverse technique leads to the identification of specific channel structures, which offer an explanation for the interesting aquifer-test observations. Shortest-path analysis demonstrates how long-range, high-k channel connections can have a critical impact on subsurface hydraulic response.