An efficient numerical model for incompressible two-phase flow in fractured media

Various numerical methods have been used in the literature to simulate single and multiphase flow in fractured media. A promising approach is the use of the discrete-fracture model where the fracture entities in the permeable media are described explicitly in the computational grid. In this work, we present a critical review of the main conventional methods for multiphase flow in fractured media including the finite difference (FD), finite volume (FV), and finite element (FE) methods, that are coupled with the discrete-fracture model. All the conventional methods have inherent limitations in accuracy and applications. The FD method, for example, is restricted to horizontal and vertical fractures. The accuracy of the vertex-centered FV method depends on the size of the matrix gridcells next to the fractures; for an acceptable accuracy the matrix gridcells next to the fractures should be small. The FE method cannot describe properly the saturation discontinuity at the matrix–fracture interface. In this work, we introduce a new approach that is free from the limitations of the conventional methods. Our proposed approach is applicable in 2D and 3D unstructured griddings with low mesh orientation effect; it captures the saturation discontinuity from the contrast in capillary pressure between the rock matrix and fractures. The matrix–fracture and fracture–fracture fluxes are calculated based on powerful features of the mixed finite element (MFE) method which provides, in addition to the gridcell pressures, the pressures at the gridcell interfaces and can readily model the pressure discontinuities at impermeable faults in a simple way. To reduce the numerical dispersion, we use the discontinuous Galerkin (DG) method to approximate the saturation equation. We take advantage of a hybrid time scheme to alleviate the restrictions on the size of the time step in the fracture network. Several numerical examples in 2D and 3D demonstrate the robustness of the proposed model. Results show the significance of capillary pressure and orders of magnitude increase in computational speed compared to previous works.     

An equivalent medium model for wave simulation in fractured porous rocks

Seismic wave propagation in reservoir rocks is often strongly affected by fractures and micropores. Elastic properties of fractured reservoirs are studied using a fractured porous rock model, in which fractures are considered to be embedded in a homogeneous porous background. The paper presents an equivalent media model for fractured porous rocks. Fractures are described in a stress‐strain relationship in terms of fracture‐induced anisotropy. The equations of poroelasticity are used to describe the background porous matrix and the contents of the fractures are inserted into a matrix. Based on the fractured equivalent‐medium theory and Biot's equations of poroelasticity, two sets of porosity are considered in a constitutive equation. The porous matrix permeability and fracture permeability are analysed by using the continuum media seepage theory in equations of motion. We then design a fractured porous equivalent medium and derive the modified effective constants for low‐frequency elastic constants due to the presence of fractures. The expressions of elastic constants are concise and are directly related to the properties of the main porous matrix, the inserted fractures and the pore fluid. The phase velocity and attenuation of the fractured porous equivalent media are investigated based on this model. Numerical simulations are performed. We show that the fractures and pores strongly influence wave propagation, induce anisotropy and cause poroelastic behaviour in the wavefields. We observe that the presence of fractures gives rise to changes in phase velocity and attenuation, especially for the slow P‐wave in the direction parallel to the fracture plane.     

Matrix–fracture transfer shape factor for modeling flow of a compressible fluid in dual-porosity media

The matrix–fracture transfer shape factor is one of the important parameters in the modeling of fluid flow in fractured porous media using a dual-porosity concept. Warren and Root [36] introduced the dual-porosity concept and suggested a relation for the shape factor. There is no general relationship for determining the shape factor for a single-phase flow of slightly compressible fluids. Therefore, different studies reported different values for this parameter, as an input into the flow models. Several investigations have been reported on the shape factor for slightly compressible fluids. However, the case of compressible fluids has not been investigated in the past. The focus of this study is, therefore, to find the shape factor for the single-phase flow of compressible fluids (gases) in fractured porous media. In this study, a model for the determination of the shape factor for compressible fluids is presented; and, the solution of nonlinear gas diffusivity equation is used to derive the shape factor. The integral method and the method of moments are used to solve the nonlinear governing equation by considering the pressure dependency of the viscosity and isothermal compressibility of the fluid. The approximate semi-analytical model for the shape factor presented in this study is verified using single-porosity, fine-grid, numerical simulations. The dependency of the shape factor on the gas specific gravity, pressure and temperature are also investigated. The theoretical analysis presented improves our understanding of fluid flow in fractured porous media. In addition, the developed matrix–fracture transfer shape factor can be used as an input for modeling flow of compressible fluids in dual-porosity systems, such as naturally fractured gas reservoirs, coalbed methane reservoirs and fractured tight gas reservoirs.     

Unified fractional differential approach for transient interporosity flow in naturally fractured media

A unified approach to modeling flows of slightly compressible fluids through naturally fractured media is presented. The unified fractional differential model is derived by combining the flow at micro scale for matrix blocks and macro scale for fractures, using the transient interporosity flow behavior at the interface between matrix blocks and fractures. The derived model is able to unify existing transient interporosity flow models formulated for different shapes of matrix blocks in any medium dimensions. The model is formulated in the form of a fractional order partial differential equation that involves Caputo derivative of order 1/2 with respect to time. Explicit solutions for the unified model are derived for different axisymmetrical spatial domains using Hankel or Hankel–Weber finite or infinite transforms. Comparisons between the predictions of the unified model and those obtained from existing transient interporosity flow models for matrix blocks in the form of slabs, spheres and cylinders are presented. It is shown that the unified fractional derivative model leads to solutions that are very close to those of transient interporosity flow models for fracture-dominant and transitional fracture-to-matrix dominant flow regimes. An analysis of the results of the unified model reveals that the pressure varies linearly with the logarithm of time for different flow regimes, with half slope for the transitional fracture-to-matrix dominant flow regime vs. the fracture and matrix dominant flow regimes. In addition, a new re-scaling that involves the characteristic length in the form of matrix block volume to surface area ratio is derived for the transient interporosity flow models for matrix blocks of different shapes. It is shown that the re-scaled transient interporosity flow models are governed by two dimensionless parameters Θ and Λ compared to only one dimensionless parameter Θ for the unified model. It is shown that the solutions of the transient interporosity flow models for different shapes of matrix blocks are almost identical for the re-scaled variables. Furthermore, the driving parameters for solution behavior are identified based on asymptotic approximations for different flow regimes. It is found that the matrix diffusion and the matrix area-to-volume ratio affect the solution behavior only for the transitional fracture-to-matrix dominant flow regime, that the capacitance ratio affects the solution behavior only for transitional and matrix dominant flow regimes and that the fracture diffusion is involved in all three flow regimes. Similar identification of the driving parameters is also presented in the re-scaled case.     

FracKfinder: A MATLAB Toolbox for Computing Three-Dimensional Hydraulic Conductivity Tensors for Fractured Porous Media

Fractures in porous media have been documented extensively. However, they are often omitted from groundwater flow and mass transport models due to a lack of data on fracture hydraulic properties and the computational burden of simulating fractures explicitly in large model domains. We present a MATLAB toolbox, FracKfinder, that automates HydroGeoSphere (HGS), a variably saturated, control volume finite-element model, to simulate an ensemble of discrete fracture network (DFN) flow experiments on a single cubic model mesh containing a stochastically generated fracture network. Because DFN simulations in HGS can simulate flow in both a porous media and a fracture domain, this toolbox computes tensors for both the matrix and fractures of a porous medium. Each model in the ensemble represents a different orientation of the hydraulic gradient, thus minimizing the likelihood that a single hydraulic gradient orientation will dominate the tensor computation. Linear regression on matrices containing the computed three-dimensional hydraulic conductivity (K) values from each rotation of the hydraulic gradient is used to compute the K tensors. This approach shows that the hydraulic behavior of fracture networks can be simulated where fracture hydraulic data are limited. Simulation of a bromide tracer experiment using K tensors computed with FracKfinder in HGS demonstrates good agreement with a previous large-column, laboratory study. The toolbox provides a potential pathway to upscale groundwater flow and mass transport processes in fractured media to larger scales.     

Effect of fracture pressure depletion regimes on the dual-porosity shape factor for flow of compressible fluids in fractured porous media

A precise value of the matrix-fracture transfer shape factor is essential for modeling fluid flow in fractured porous media by a dual-porosity approach. The slightly compressible fluid shape factor has been widely investigated in the literature. In a recent study, we have developed a transfer function for flow of a compressible fluid using a constant fracture pressure boundary condition [Ranjbar E, Hassanzadeh H, Matrix-fracture transfer shape factor for modeling flow of a compressible fluid in dual-porosity media. Adv Water Res 2011;34(5):627-39. doi:10.1016/j.advwatres.2011.02.012]. However, for a compressible fluid, the consequence of a pressure depletion boundary condition on the shape factor has not been investigated in the previous studies. The main purpose of this paper is, therefore, to investigate the effect of the fracture pressure depletion regime on the shape factor for single-phase flow of a compressible fluid. In the current study, a model for evaluation of the shape factor is derived using solutions of a nonlinear diffusivity equation subject to different pressure depletion regimes. A combination of the heat integral method, the method of moments and Duhamel’s theorem is used to solve this nonlinear equation. The developed solution is validated by fine-grid numerical simulations. The presented model can recover the shape factor of slightly compressible fluids reported in the literature. This study demonstrates that in the case of a single-phase flow of compressible fluid, the shape factor is a function of the imposed boundary condition in the fracture and its variability with time. It is shown that such dependence can be described by an exponentially declining fracture pressure with different decline exponents. These findings improve our understanding of fluid flow in fractured porous media.     

Variable-density groundwater flow and solute transport in porous media containing nonuniform discrete fractures

Variations in fluid density can greatly affect fluid flow and solute transport in the subsurface. Heterogeneities such as fractures play a major role for the migration of variable-density fluids. Earlier modeling studies of density effects in fractured media were restricted to orthogonal fracture networks, consisting of only vertical and horizontal fractures. The present study addresses the phenomenon of 3D variable-density flow and transport in fractured porous media, where fractures of an arbitrary incline can occur. A general formulation of the body force vector is derived, which accounts for variable-density flow and transport in fractures of any orientation. Simulation results are presented that show the verification of the new model formulation, for the porous matrix and for inclined fractures. Simulations of variable-density flow and solute transport are then conducted for a single fracture, embedded in a porous matrix. The simulations show that density-driven flow in the fracture causes convective flow within the porous matrix and that the high-permeability fracture acts as a barrier for convection. Other simulations were run to investigate the influence of fracture incline on plume migration. Finally, tabular data of the tracer breakthrough curve in the inclined fracture is given to facilitate the verification of other codes.     

The significance of flow in the matrix of the Chalk unsaturated zone

The significance of flow in the matrix of the Chalk unsaturated zone, in comparison with flow in fractures, has been the subject of much debate. In this article, important elements of the literature are discussed in detail and several simple modelling analyses based on steady-state flow are presented. A study of the sensitivity of solute spreading to fracture spacing in models that ignore matrix flow shows that this latter assumption is generally incompatible with observed solute profiles, unless unrealistically small fracture spacings are assumed. The effect of air phase continuities (e.g. bedding planes) on matrix flow has also been examined. These discontinuities are frequently interrupted by points of connectivity between matrix blocks. An issue therefore is the relationship between connectivity and its effect on inter-block conductance. A simple analysis of the Laplace equation shows that just 1% connectivity represents an effective pathway equivalent to 18% of the local rock hydraulic conductivity. Obviously, when there is no fracture flow, solute spreading is significantly reduced. However, dual permeability model simulations show that matrix flow reduces solute spreading in the presence of persistent fracture flow. All of the above studies suggest that flow in the matrix of the Chalk unsaturated zone is significant and that ignoring it may result in a serious misunderstanding of the system.     

Application of implicit sub-time stepping to simulate flow and transport in fractured porous media

In general, the accuracy of numerical simulations is determined by spatial and temporal discretization levels. In fractured porous media, the time step size is a key factor in controlling the solution accuracy for a given spatial discretization. If the time step size is restricted by the relatively rapid responses in the fracture domain to maintain an acceptable level of accuracy in the entire simulation domain, the matrix tends to be temporally over-discretized. Implicit sub-time stepping applies smaller sub-time steps only to the sub-domain where the accuracy requirements are less tolerant and is most suitable for problems where the response is high in only a small portion of the domain, such as within and near the fractures in fractured porous media. It is demonstrated with illustrative examples that implicit sub-time stepping can significantly improve the simulation efficiency with minimal loss in accuracy when simulating flow and transport in fractured porous media. The methodology is successfully applied to density-dependent flow and transport simulations in a Canadian Shield environment, where the flow and transport is dominated by discrete, highly conductive fracture zones.     

An efficient numerical model for multicomponent compressible flow in fractured porous media

An efficient and accurate numerical model for multicomponent compressible single-phase flow in fractured media is presented. The discrete-fracture approach is used to model the fractures where the fracture entities are described explicitly in the computational domain. We use the concept of cross flow equilibrium in the fractures. This will allow large matrix elements in the neighborhood of the fractures and considerable speed up of the algorithm. We use an implicit finite volume (FV) scheme to solve the species mass balance equation in the fractures. This step avoids the use of Courant–Freidricks–Levy (CFL) condition and contributes to significant speed up of the code. The hybrid mixed finite element method (MFE) is used to solve for the velocity in both the matrix and the fractures coupled with the discontinuous Galerkin (DG) method to solve the species transport equations in the matrix. Four numerical examples are presented to demonstrate the robustness and efficiency of the proposed model. We show that the combination of the fracture cross-flow equilibrium and the implicit composition calculation in the fractures increase the computational speed 20–130 times in 2D. In 3D, one may expect even a higher computational efficiency.     

Multiple Well-Shutdown Tests and Site-Scale Flow Simulation in Fractured Rocks

A new method was developed for conducting aquifer tests in fractured-rock flow systems that have a pump-and-treat (P&T) operation for containing and removing groundwater contaminants. The method involves temporary shutdown of individual pumps in wells of the P&T system. Conducting aquifer tests in this manner has several advantages, including (1) no additional contaminated water is withdrawn, and (2) hydraulic containment of contaminants remains largely intact because pumping continues at most wells. The well-shutdown test method was applied at the former Naval Air Warfare Center (NAWC), West Trenton, New Jersey, where a P&T operation is designed to contain and remove trichloroethene and its daughter products in the dipping fractured sedimentary rocks underlying the site. The detailed site-scale subsurface geologic stratigraphy, a three-dimensional MODFLOW model, and inverse methods in UCODE_2005 were used to analyze the shutdown tests. In the model, a deterministic method was used for representing the highly heterogeneous hydraulic conductivity distribution and simulations were conducted using an equivalent porous media method. This approach was very successful for simulating the shutdown tests, contrary to a common perception that flow in fractured rocks must be simulated using a stochastic or discrete fracture representation of heterogeneity. Use of inverse methods to simultaneously calibrate the model to the multiple shutdown tests was integral to the effectiveness of the approach.     

Application of ERT,Saline Tracer and Numerical Studies to Delineate Preferential Paths in Fractured Granites

Accurate quantification of in situ heterogeneity and flow processes through fractured geologic media remains elusive for hydrogeologists due to the complexity in fracture characterization and its multiscale behavior. In this research, we demonstrated the efficacy of tracer-electrical resistivity tomography (ERT) experiments combined with numerical simulations to characterize heterogeneity and delineate preferential flow paths in a fractured granite aquifer. A series of natural gradient saline tracer experiments were conducted from a depth window of 18 to 22 m in an injection well (IW) located inside the Indian Institute of Technology Hyderabad campus. Tracer migration was monitored in a time-lapse mode using two cross-sectional surface ERT profiles placed in the direction of flow gradient. ERT data quality was improved by considering stacking, reciprocal measurements, resolution indicators, and geophysical logs. Dynamic changes in subsurface electrical properties inferred via resistivity anomalies were used to highlight preferential flow paths of the study area. Temporal changes in electrical resistivity and tracer concentration were monitored along the vertical in an observation well located at 48 m to the east of the IW. ERT-derived tracer breakthrough curves were in agreement with geochemical sample measurements. Fracture geometry and hydraulic properties derived from ERT and pumping tests were further used to evaluate two mathematical conceptualizations that are relevant to fractured aquifers. Results of numerical analysis conclude that dual continuum model that combines matrix and fracture systems through a flow exchange term has outperformed equivalent continuum model in reproducing tracer concentrations at the monitoring wells (evident by a decrease in RMSE from 199 to 65 mg/L). A sensitivity analysis on model simulations conclude that spatial variability in hydraulic conductivity, local-scale dispersion, and flow exchange at fracture-matrix interface have a profound effect on model simulations.     

Lanczos method for the solution of groundwater flow in discretely fractured porous media

One of the more advanced approaches for simulating groundwater flow in fractured porous media is the discrete-fracture approach. This approach is limited by the large computational overheads associated with traditional modeling methods. In this work, we apply the Lanczos reduction method to the modeling of groundwater flow in fractured porous media using the discrete-fracture approach. The Lanczos reduction method reduces a finite element equation system to a much smaller tridiagonal system of first-order differential equations. The reduced system can be solved by a standard tridiagonal algorithm with little computational effort. Because solving the reduced system is more efficient compared to solving the original system, the simulation of groundwater flow in discretely fractured media using the reduction method is very efficient. The proposed method is especially suitable for the problem of large-scale and long-term simulation. In this paper, we develop an iterative version of Lanczos algorithm, in which the preconditioned conjugate gradient solver based on ORTHOMIN acceleration is employed within the Lanczos reduction process. Additional efficiency for the Lanczos method is achieved by applying an eigenvalue shift technique. The “shift” method can improve the Lanczos system convergence, by requiring fewer modes to achieve the same level of accuracy over the unshifted case. The developed model is verified by comparison with dual-porosity approach. The efficiency and accuracy of the method are demonstrated on a field-scale problem and compared to the performance of classic time marching method using an iterative solver on the original system. In spite of the advances, more theoretical work needs to be carried out to determine the optimal value of the shift before computations are actually carried out.     

A numerical method for two-phase flow in fractured porous media with non-matching grids

We propose a novel computational method for the efficient simulation of two-phase flow in fractured porous media. Instead of refining the grid to capture the flow along the faults or fractures, we represent the latter as immersed interfaces, using a reduced model for the flow and suitable coupling conditions. We allow for non matching grids between the porous matrix and the fractures to increase the flexibility of the method in realistic cases. We employ the extended finite element method for the Darcy problem and a finite volume method that is able to handle cut cells and matrix-fracture interactions for the saturation equation. Moreover, we address through numerical experiments the problem of the choice of a suitable numerical flux in the case of a discontinuous flux function at the interface between the fracture and the porous matrix. A wrong approximate solution of the Riemann problem can yield unphysical solutions even in simple cases.     

基于数字岩心的含裂隙储层砂岩介电性质研究

了解储层岩石的介电特性在石油工业的各个方面都有重要的应用.小尺度裂隙是影响岩石介电性质的地质因素之一,获得裂隙对含裂隙岩石介电性质影响的定量关系具有重要的理论和实践意义.以含裂隙人造砂岩的三维微观数字结构为基础,通过基于三维有限差分算法计算的岩石介电性质与实验数据的对比验证数值计算方法的有效性.在此基础上,通过理论模型获得不同孔隙度基质的介电性质,并在不含裂隙人造砂岩的三维微观数字结构中人为添加以裂隙密度和纵横比为定量表征参数的裂隙,应用验证后的数值算法模拟随频率变化的含裂隙砂岩的介电性质,分析和研究不同孔隙度基质中定向排列裂隙对砂岩介电性质的影响.结果表明,当裂隙孔隙度随裂隙纵横比或裂隙密度发生改变时,含裂隙砂岩的介电性质与裂隙密度以及裂隙纵横比呈正相关关系,而当裂隙孔隙度保持不变时,含裂隙砂岩的介电性质随裂隙纵横比的减小而增大;裂隙参数的改变对不同基质孔隙度的含裂隙砂岩的介电性质的影响趋势较为一致,但随着基质孔隙度的减小,裂隙对砂岩介电性质的影响逐渐增大.裂隙参数和基质孔隙度对含裂隙砂岩介电性质影响的研究结果为基于介电特性的裂缝性油气储层的定量表征提供了依据,在油气勘探开发中具有重要的应用前景.     

Weathering and erosion of fractured bedrock systems

We explore the contribution of fractures (joints) in controlling the rate of weathering advance for a low‐porosity rock by using methods of homogenization to create averaged weathering equations. The rate of advance of the weathering front can be expressed as the same rate observed in non‐fractured media (or in an individual block) divided by the volume fraction of non‐fractured blocks in the fractured parent material. In the model, the parent has fractures that are filled with a more porous material that contains only inert or completely weathered material. The low‐porosity rock weathers by reaction‐transport processes. As observed in field systems, the model shows that the weathering advance rate is greater for the fractured as compared to the analogous non‐fractured system because the volume fraction of blocks is < 1. The increase in advance rate is attributed both to the increase in weathered material that accompanies higher fracture density, and to the increase in exposure of surface of low‐porosity rock to reaction‐transport. For constant fracture aperture, the weathering advance rate increases when the fracture spacing decreases. Equations describing weathering advance rate are summarized in the ‘List of selected equations’. If erosion is imposed at a constant rate, the weathering systems with fracture‐bounded bedrock blocks attain a steady state. In the erosional transport‐limited regime, bedrock blocks no longer emerge at the air‐regolith boundary because they weather away. In the weathering‐limited (or kinetic) regime, blocks of various size become exhumed at the surface and the average size of these exposed blocks increases with the erosion rate. For convex hillslopes, the block size exposed at the surface increases downslope. This model can explain observations of exhumed rocks weathering in the Luquillo mountains of Puerto Rico. Published 2017. This article is a U.S. Government work and is in the public domain in the USA     

Matrix diffusion-derived plume attenuation in fractured bedrock

Matrix diffusion can attenuate the rate of plume migration in fractured bedrock relative to the rate of ground water flow for both conservative and nonconservative solutes of interest. In a system of parallel, equally spaced constant aperture fractures subject to steady-state ground water flow and an infinite source width, the degree of plume attenuation increases with time and travel distance, eventually reaching an asymptotic level. The asymptotic degree of plume attenuation in the absence of degradation can be predicted by a plume attenuation factor, beta, which is readily estimated as R' (phi(m)/phi(f)), where R' is the retardation factor in the matrix, phi(m) is the matrix porosity, and phi(f) is the fracture porosity. This dual-porosity relationship can also be thought of as the ratio of primary to secondary porosity. Beta represents the rate of ground water flow in fractures relative to the rate of plume advance. For the conditions examined in this study, beta increases with greater matrix porosity, greater matrix fraction organic carbon, larger fracture spacing, and smaller fracture aperture. These concepts are illustrated using a case study where dense nonaqueous phase liquid in fractured sandstone produced a dissolved-phase trichloroethylene (TCE) plume approximately 300 m in length. Transport parameters such as matrix porosity, fracture porosity, hydraulic gradient, and the matrix retardation factor were characterized at the site through field investigations. In the fractured sandstone bedrock examined in this study, the asymptotic plume attenuation factors (beta values) for conservative and nonconservative solutes (i.e., chloride and TCE) were predicted to be approximately 800 and 12,210, respectively. Quantitative analyses demonstrate that a porous media (single-porosity) solute transport model is not appropriate for simulating contaminant transport in fractured sandstone where matrix diffusion occurs. Rather, simulations need to be conducted with either a discrete fracture model that explicitly incorporates matrix diffusion, or a dual-continuum model that accounts for mass transfer between mobile and immobile zones. Simulations also demonstrate that back diffusion from the matrix to fractures will likely be the time-limiting factor in reaching ground water cleanup goals in some fractured bedrock environments.     

Specification of Matrix Cleanup Goals in Fractured Porous Media

Semianalytical transient solutions have been developed to evaluate what level of fractured porous media (e.g., bedrock or clay) matrix cleanup must be achieved in order to achieve compliance of fracture pore water concentrations within a specified time at specified locations of interest. The developed mathematical solutions account for forward and backward diffusion in a fractured porous medium where the initial condition comprises a spatially uniform, nonzero matrix concentration throughout the domain. Illustrative simulations incorporating the properties of mudstone fractured bedrock demonstrate that the time required to reach a desired fracture pore water concentration is a function of the distance between the point of compliance and the upgradient face of the domain where clean groundwater is inflowing. Shorter distances correspond to reduced times required to reach compliance, implying that shorter treatment zones will respond more favorably to remediation than longer treatment zones in which back‐diffusion dominates the fracture pore water response. For a specified matrix cleanup goal, compliance of fracture pore water concentrations will be reached sooner for decreased fracture spacing, increased fracture aperture, higher matrix fraction organic carbon, lower matrix porosity, shorter aqueous phase decay half‐life, and a higher hydraulic gradient. The parameters dominating the response of the system can be measured using standard field and laboratory techniques.     

Impact of fracture network geometry on free convective flow patterns

The effect of fracture network geometry on free convection in fractured rock is studied using numerical simulations. We examine the structural properties of fracture networks that control the onset and strength of free convection and the patterns of density-dependent flow. Applicability of the equivalent porous medium approach (EPM) is also tested, and recommendations are given, for which situations the EPM approach is valid. To date, the structural properties of fracture networks that determine free convective flow are examined only in few, predominantly simplified regular fracture networks. We consider fracture networks containing continuous, discontinuous, orthogonal and/or inclined discrete fractures embedded in a low-permeability rock matrix. The results indicate that bulk permeability is not adequate to infer the occurrence and magnitude of free convection in fractured rock. Fracture networks can inhibit or promote convection depending on the fracture network geometry. Continuous fracture circuits are the crucial geometrical feature of fracture networks, because large continuous fracture circuits with a large vertical extent promote convection. The likelihood of continuous fracture circuits and thus of free convection increases with increasing fracture density and fracture length, but individual fracture locations may result in great deviances in strength of convection between statistically equivalent fracture networks such that prediction remains subject to large uncertainty.