A new two-step particle tracking and channels model method for preferential 3-D flow and transport in large DFNs of fractured aquifers

Abstract

A new two-step method has been developed for predicting three-dimensional (3-D) pollutant plumes and breakthrough curves (BTC) in fractured aquifers. Predicting fluid motion fields and pollutant concentrations in groundwater is a challenging task, due to the difficulties in building a 3-D discrete fracture network (DFN) with the same geometry of fractures and interconnections as in the studied aquifer and the computational complexity of the problem being modeled. To improve the representation of DFN, geological characterization of fracture apertures was performed using field well-pumping data as the prerequisite for the modeling approach. The modeling was performed in two steps: first, 3-D particle tracking following streamline (PTFS) simulations in DFN backbones, and second, a 3-D channels model (CM) analytical solution. The PTFS simulations captured the mean properties of the 3-D velocity field relevant to spatial and temporal transport along preferential flow pathways of a fractured aquifer. The 3-D CM solution yielded the exact BTC of concentrations accounting for the slowest flow pathways typically excluded from DFN backbones extraction methods. The PTFS/CM simulations were initially performed at the lab scale on a 3-D 20 m rock block from the Bari (South Italy) fractured aquifer. Results were consistent with the chlorophyll-A BTC obtained from a well-to-well tracer experiment. Subsequently, the capability of the new method was tested by increasing the scale of simulations to a larger 3-D DFN of 200 m with 2,200 fractures. The resultant BTC was obtained in 32 min, showing the proposed PTFS/CM technique can rapidly determine complex 3-D solutions in heterogeneous fracture-dominated aquifers. The applied Lagrangian and Eulerian techniques can provide quasi-deterministic BTCs and 3-D maps of pollutant concentrations, avoiding extensive computations required for conforming mesh methods. Moreover, the solution does not require multiple randomly arranged DFN realizations, as in particle tracking-based stochastic methods. These results have significant implications in modeling transport at the aquifer scale, where DFNs are composed of thousands of fractures.