loading...
running...
The science
What the model solves. Lava is fed into the top of a slope at an effusion rate Q through a channel of width w. The flow is followed along the path in 200 cells of 20 m. Each cell has a thickness h, a core temperature T and a crust thickness δ. Every step the flow law gives the volume flux, mass is moved downhill (conservatively), then each cell cools. The 2D profile and the 3D Blender terrain draw exactly the same numbers.
Viscosity of the melt. A Vogel-Fulcher-Tammann law, with a silica-dependent constant B:
log10 ηmelt = -4.5 + B(SiO2) / (T[K] - 600)
B is tabulated so that basalt at 1200 C is about 102 Pa s, andesite at 1000 C about 105.5, dacite at 900 C about 108 and rhyolite at 850 C about 109.5 Pa s. A rule of thumb: roughly one decade of viscosity per 100-130 K for basalt. Crystals stiffen it (Einstein-Roscoe):
η = ηmelt (1 - φ/0.6)-2.5, φ(T) = φ0 + 0.002 (T0 - T)
so a cooling flow gets stiffer twice: through the melt and through new crystals. The yield strength is tied to the same stiffening, τy = τy0 (η/η0)1/2 (an empirical stand-in, not a measured law).
Three flow laws. Newtonian: τ = η γ̇. Bingham: τ = τy + η γ̇. Herschel-Bulkley: τ = τy + K γ̇n (n < 1 shear-thinning). The graph "Rheogram" draws all three for the current lava.
Flow down a slope (wide channel). The shear stress at height y above the bed is τ(y) = ρ g sinθ (h - y). Where it is below τy the lava does not shear and moves as a rigid plug of thickness hp = τy / (ρ g sinθ). Integrating the Herschel-Bulkley law gives the flux per unit width, with hs = h - hp and m = (n+1)/n:
q = [n/(n+1)] (ρ g sinθ / K)1/n hsm [hp + hs (n+1)/(2n+1)]
For τy = 0, n = 1 this is the Jeffreys equation, mean speed u = ρ g sinθ h² / (3 η). With a yield stress and n = 1 it is the Bingham result q = ρ g sinθ hs² (3h - hs) / (6 η). A flow thinner than hp does not move at all; it stops with a finite thickness (this is why lava levees and flow fronts have steep, thick edges).
Rigid crust lid. A crust of thickness δ on top cannot shear, so the sheared layer is cut off at y = h - δ; the lid and the lava beneath it ride along at that speed. The same closed form is used with the shear profile truncated, which slows a thin flow once its crust is a large part of its depth.
Cooling. The surface of the crust loses heat by radiation and wind; the crust conducts heat from the core:
k (T - Ts) / δ = ε σ (Ts4 - Ta4) + hc (Ts - Ta), ρ cp h dT/dt = - qsurf
(k = 1.5 W/m K, cp = 1200 J/kg K, the latent heat of crystallisation is not added). The crust grows as the cold solid-to-liquid isotherm moves into the lava; for a semi-infinite body with a cold surface T(x,t) = Ta + (T - Ta) erf(x / 2√κt), so the rigid-crust isotherm Tsol sits at
δ = 2 ξ √(κ t), erf(ξ) = (Tsol - Ta) / (T - Ta), dδ/dt = 2 ξ² κ / δ
with κ = 6e-7 m²/s. If the core falls to Tsol the whole column is solid and the cell stops.
Flow-front advance. The thickness obeys the kinematic-wave equation ∂h/∂t + ∂q/∂s = 0, solved with a conservative first-order upwind scheme (mass balance error stays at rounding level). For a constant supply into a dry slope the front is a shock moving at q/h = Q / (w hsteady), which for isothermal Newtonian lava is exactly the Jeffreys speed; the graph "Front position" draws both. When the lava cools the front slows and eventually stalls. The cooling-limited length uses the Pinkerton-Wilson idea: a flow runs until the Graetz number u h² / (κ L) falls to about 300.
Checks in the page. (1) numerical integration of the velocity profile against the closed-form flux with a rigid lid; (2) the Newtonian limit against Jeffreys; (3) the Bingham limit against its textbook formula; (4) the solved flow depth against the imposed Q/w; (5) the PDE flow front against the Jeffreys speed times time; (6) mass supplied against mass stored plus lost; (7) crust thickness from explicit finite-difference conduction against the erf solution.
Honest simplifications. A wide, straight, rectangular channel with fixed width (no overflow, levee building, lateral spreading or branching); the kinematic wave ignores the pressure-gradient term, so the front is steeper than a real one; first-order upwind smears fronts over a few cells; the viscosity law is a fit, not a measurement of any named eruption (hydrous melt, bubbles and real crystal-size effects are not modelled); the yield strength law is an empirical scaling; heat loss ignores latent heat, rain, lava tubes and viscous heating; the crust is a uniform lid and does not break or re-form. Treat it as a teaching tool for how composition, temperature, slope and cooling change the style and reach of a flow, not as a hazard forecast.