Drying of a partially saturated rock matrix by fracture ventilation: Experiments and modeling in a single fracture-matrix system

We present an efficient approach for modeling drying resulting from airflow through fractures embedded within a rock matrix by coupling a nonlinear diffusion equation for moisture transport in the matrix with a depth-averaged quasi steady state advection-dispersion equation for the vapor transport in the fracture. Coupling between the moisture transport equation in the fracture and matrix is achieved by applying Kelvin's equation at the fracture-matrix interface. The proposed modeling approach was then validated based on a single fracture-matrix drying experiment involving airflow through a fracture sandwiched between two matrix blocks. The moisture diffusivity for the rock matrix was first estimated in a one-dimensional matrix drying experiment using a mass balance method. Numerical simulations based on our coupled modeling approach captured the important characteristics of the drying behavior in the single fracture-matrix experiment. Additional numerical simulations are presented to clarify the factors controlling drying behavior. The important features of drying in a single fracture-matrix system are as follows: (1) The total evaporation rate exhibits an initial constant value at early time, followed by a 1/t decrease with time. (2) At early time a "drying toe'' forms near the fracture entrance and propagates along the fracture length as the fracture-matrix interface becomes progressively drier. During this period, the moisture content distributions in the rock matrix are two-dimensional, with gradients both parallel and perpendicular to the fracture. (3) At late time, after the drying toe reaches the end of the fracture, a "drying front'' forms and propagates away from the fracture-matrix interface, leading to a largely one-dimensional (perpendicular to the fracture) variation of moisture content in the rock matrix. The experimental measurements could not fully resolve the early time two-dimensional moisture content variations due to the coarse resolution, precluding a rigorous comparison with simulated behavior. However, the later-time one-dimensional moisture content variation observed in the experimental data was reasonably reproduced in the numerical simulations.