1/22
Looks like no tags are added yet.
Heterogeneity
Properties of the porous medium vary in space (with position). • Homogeneous: K is independent of position → K(x,y,z) = c (constant) • Heterogeneous: K depends on position → K(x,y,z) ≠ c • Warning: these definitions depend on the scale at which the system is considered, modelled or observed
Heterogeneity & scale (REV)
Another example of the REV (Representative Elementary Volume) concept. • Averaging K over a large enough volume gives a similar value at different locations (same rock formation) • The same material can look very different in a 10 × 10 m window compared with a 150 × 150 m window. Only the window changes. (Schulze-Makuch et al., 1999)
Heterogeneity example: North Bay, Ontario
Unconfined sand aquifer (Sudicky et al., 2010) • K across a single sand aquifer ranges from ~15 m/d down to 4 mm/d • K data are strongly skewed. ln K is approximately normal (symmetric). • Arithmetic mean = 4.4 × 10⁻⁴ m/s • Geometric mean = 6.4 × 10⁻⁵ m/s
Isotropic vs anisotropic
Properties can vary with direction. • Isotropic: K ≠ f(θ), independent of direction • Anisotropic: K = f(θ), dependent on direction • θ = angle between the horizontal and the direction of measurement
Transversely isotropic
A layered formation where K only has horizontal (or sub-horizontal) and vertical (or sub-vertical) components • Kx = Ky ≠ Kz • Example: Grand Canyon, Arizona (horizontal rock beds)
Heterogeneous vs anisotropic (difference)
Heterogeneous = many materials jumbled together • K depends on WHERE you are: K = K(x,y,z) • Move a metre sideways → different material, different K • Within each patch, direction doesn't matter
Anisotropic = one material laid down in beds • K depends on WHICH WAY you go: Kx ≠ Kz • Kx easy (along the beds), Kz hard (across them) • Same material everywhere. The fabric makes flow along the laminae easier than across them.
The two are independent. A formation can be either, neither or both (homog + iso, homog + aniso, heterog + iso, heterog + aniso).
Darcy's law with hydraulic conductivity tensor
q = −K∇h (K = hydraulic conductivity tensor) • qx = −Kxx ∂h/∂x − Kxy ∂h/∂y − Kxz ∂h/∂z (horizontal x flow) • qy = −Kyx ∂h/∂x − Kyy ∂h/∂y − Kyz ∂h/∂z (horizontal y flow) • qz = −Kzx ∂h/∂x − Kzy ∂h/∂y − Kzz ∂h/∂z (vertical z flow)
K tensor: why 9 terms
A head gradient in one direction can drive flow in another direction. • Dipping beds: a purely horizontal gradient pushes water along the dip (partly downwards), because that is the easy direction • The off-diagonal terms (Kxz, Kzx, …) carry this coupling
K tensor: matrix form
q = [qx, qy, qz]ᵀ, ∇h = [∂h/∂x, ∂h/∂y, ∂h/∂z]ᵀ K = [Kxx Kxy Kxz / Kyx Kyy Kyz / Kzx Kzy Kzz] • Symmetric: Kxy = Kyx, Kxz = Kzx, Kyz = Kzy • If the axes of max and min K are NOT aligned with x, y, z → off-diagonal terms ≠ 0 (e.g. dipping beds: Kmax along bedding, Kmin across it, rotated from x and z) • If aligned → diagonal only: K = diag(Kxx, Kyy, Kzz) • Transversely isotropic: Kxx = Kyy ≠ Kzz • In modelling, the Cartesian axes are usually aligned with the principal axes of anisotropy
Layered formation → equivalent single layer
A sequence of discrete layers, each homogeneous and isotropic (conductivity Ki, thickness di, total thickness d) → can be treated as ONE homogeneous anisotropic layer (strictly, transversely isotropic) with Kx and Kz
Vertical hydraulic conductivity Kz (flow across layers)
• Continuity: q_in = q_out ≡ qz (SAME flux through every layer) • Head losses ADD: Δh = Δh1 + Δh2 + … + ΔhN • Darcy: qz = K1 Δh1/d1 = K2 Δh2/d2 = … = KN ΔhN/dN = Kz Δh/d • Rearrange: Δhi = qz di / Ki • Kz = qz d/Δh = d / (d1/K1 + d2/K2 + … + dN/KN) → Weighted HARMONIC mean: Kz = d [Σ di/Ki]⁻¹
Horizontal hydraulic conductivity Kx (flow along layers)
• Head loss Δh over horizontal distance L (SAME gradient in every layer) • Q′ = discharge per unit thickness = sum of the discharges through the layers (discharges ADD) • qx = Q′/d = (1/d) Σ Q′i = (1/d) Σ Ki di Δh/L = Kx Δh/L → Weighted ARITHMETIC mean: Kx = (1/d) Σ Ki di
Kx vs Kz for layered systems
• Horizontal: gradient same in each layer, discharges add → arithmetic mean • Vertical: flux same in each layer, head losses add → harmonic mean • For any set of layers with differing Ki: Kx > Kz → Isotropic layers make an anisotropic formation
Arithmetic, geometric & harmonic means
• Arithmetic: KA = (1/N) Σ Ki • Geometric: KG = exp[(1/N) Σ ln Ki] • Harmonic: KH = [(1/N) Σ 1/Ki]⁻¹ • Always: KH ≤ KG ≤ KA
Which mean to use
• Arithmetic → flow PARALLEL to layering • Harmonic → flow PERPENDICULAR to layering • Geometric → heterogeneous but not systematically layered, K log-normal • KG is often used as the effective parameter when averaging over the REV in heterogeneous systems
Means example (log-normal K)
Generate 10,000 log-normally distributed random numbers (Python: K = 10**np.random.randn(N)) and compute the means: • KA = 12.6983 • KG = 0.9591 • KH = 0.0776 → The means can differ by 2 orders of magnitude or more!
No-slip condition
However smooth the solid surface or however small the fluid viscosity, fluid particles immediately next to the surface do NOT move relative to it. • Velocity increases away from the boundary towards an asymptote u′ = u′m
Viscosity (Newton)
Two layers of fluid travelling at different velocities (u′ and u′ + Δu′) • Newton: shear stress τ = F/A ∝ du′/dz • τ = μ du′/dz • μ = dynamic viscosity, units Pa s [M L⁻¹ T⁻¹] • Pure water at 20 °C: μ = 1.002 × 10⁻³ Pa s
Pipe flow: force balance
Steady laminar flow, pipe radius R, control volume radius r and length Δx • Steady state → force equilibrium: πr²p + 2πrΔx τ = πr²(p + Δp) • Reduces to τ = (r/2)(Δp/Δx) → infinitesimal: τ = (r/2) dp/dx • Substitute viscosity: μ du′/dr = (r/2) dp/dx • Full derivatives are used because pressure doesn't vary across the pipe (valid when flow is steady and fully developed)
Pipe flow: velocity profile
• du′/dr = (r/2μ) dp/dx • Integrate: u′ = (1/2μ)(dp/dx) ∫ r dr = (r²/4μ) dp/dx + C • Boundary condition: at r = R, u′ = 0 → C = −(R²/4μ) dp/dx • u′ = −(1/4μ)(dp/dx)(R² − r²) → PARABOLIC velocity profile (max at centre, 0 at wall)
Poiseuille's law
Total flow through the pipe: • Q′ = 2π ∫₀ᴿ r u′(r) dr = −(π/2μ)(dp/dx) ∫₀ᴿ r(R² − r²) dr = −(π/2μ)(dp/dx) [R²r²/2 − r⁴/4]₀ᴿ • Q′ = −(πR⁴/8μ) dp/dx (Massey, 1995, Mechanics of Fluids)
Flow through a porous block (pipe model)
Block L × B × H containing N pipes of radius R • Total flow: Q = NQ′ = −(πNR⁴/8μ) dp/dx • Effective porosity: ne = vol. of voids accepting water / total vol. = πLNR²/(LHB) = πNR²/(HB) • Cross-sectional area: A = BH • → Q = −(A ne R²/8μ) dp/dx
Hydraulic conductivity from the pipe model
• Hydraulic pressure p = ρg(h − z), where h = hydraulic head →