loading...
running...
How to use
The coloured block is a vertical slice of ground, 60 cm wide and 30 cm deep, between a high water level on the left and a low one on the right. White lines join points of equal head; dark lines are flow lines drawn so that equal flux passes between neighbours; white dots are tracer particles moving with the water. Choose how the layers lie (across the flow, along it, dipping or as a lens), set each layer's grain size and porosity, and watch the numbers: the effective conductivity of the section is always between the harmonic and arithmetic means. Inject dye releases a tracer at the left edge so you can see the fast path win. Hover over the section to read the local head, conductivity and speed.
Try: (1) Layers across the flow, then thin the silt: K_eff climbs back toward the sand; (2) Layers along the flow: the gravel seam takes nearly all the flux and the dye shoots through it; (3) Silt lens in sand, colour by speed; (4) Dipping beds with different angles; (5) Loose vs dense side by side. Space pauses, R rewinds the particles.
The sand textures in the section and the picture of the laboratory permeameter below were rendered in Blender (Cycles) from the same scene as the 3D companion; the flow net, tracer paths and all numbers are computed live in the page.
The science
Underground water moves through the gaps between grains, slowly enough that its average motion follows Darcy's law: the flux per unit area is proportional to the gradient of the hydraulic head h (elevation plus pressure head),
q = − K ∇h, ∇·q = 0 ⇒ ∇·(K ∇h) = 0
The second equation says water is neither created nor stored in a steady, saturated soil. Where K is uniform this is Laplace's equation, and the lines of constant head and the flow lines cross at right angles, drawing the familiar flow net of curvilinear squares. Across a boundary between two materials the head is continuous but its slope jumps, so equipotentials and flow lines refract: the flow line bends toward the normal when it enters a coarser material, as in the dipping and lens presets. Here the equation is solved for the 60 cm by 30 cm section on a 64 by 32 grid with harmonic-mean face conductances and a preconditioned conjugate-gradient solver; the left edge is held at the head you set, the right edge at zero, top and bottom are sealed.
Conductivity of each material. Each layer's K comes from its grain size d and porosity n by the Kozeny-Carman relation, with the viscosity of water at the chosen temperature:
k = d² n³ / [180 (1 − n)²], K = k ρ g / μ
Two bounds for the effective conductivity. If the layers lie across the flow (water passes through them in turn), the flux is the same in each and Keff is the thickness-weighted harmonic mean, dominated by the tightest layer. If they lie along the flow (water takes parallel paths), the head drop is the same and Keff is the arithmetic mean, dominated by the most permeable layer:
Kseries = Σ fi / Σ (fi/Ki), Kparallel = Σ fi Ki
These are the Wiener bounds: any arrangement of the same materials in the same proportions has an effective conductivity between them. Tilted beds and lenses fall in between, and the readout reports where (0 = harmonic, 1 = arithmetic). With a conductivity contrast of 100 or more the two bounds differ by a factor of ten or more: the same grains, laid differently, give very different wells and drains.
Travel time. The pore velocity is the Darcy flux divided by porosity, v = q/n. Tracer particles here are followed through the solved flux field with Pollock's semi-analytic method (velocity varies linearly across each cell, so the exit time has a closed form). Because the particles are released across the inflow face in proportion to the flux, the flux-weighted mean travel time must equal the pore volume of the section divided by its discharge, a statement of volume conservation that the page tests. The breakthrough curve shows that in parallel layers the first tracer arrives orders of magnitude before the median, which is why a gravel seam makes a contaminant arrive early.
Numerical checks run in the page. (1) The 1D finite-volume column against the harmonic mean, flux in against flux out, the head budget, an RK4 falling-head test against its closed form and the K recovered from it, a Monte Carlo tracer against Σ L n / q, and Kozeny-Carman against Hazen and the viscosity of water; (2) this 2D solver on layers across the flow against the harmonic mean and on layers along the flow against the arithmetic mean (grid-snapped layer thickness, relative error about 1e-7); (3) the 2D mass balance; (4) for your current section, the flux-weighted mean tracer time against pore volume over discharge, the Wiener bounds, and the continuity of flux through three vertical cuts.
Simplifications. Steady, fully saturated, two-dimensional flow with isotropic materials (real sediments are often two to ten times more conductive along bedding than across it); one effective grain size and porosity per layer; no dispersion in the drawn tracer, so fronts stay sharp; no free surface, so the section is a confined strip rather than an unconfined aquifer with a water table; the grid is 64 by 32, so thin features are rounded to whole cells; and Darcy's law is only valid for slow flow (grain Reynolds number below about 1).