Presets
Playback
Water supply
Meltwater input, a moulin surge, water pressure and ice softness; the conduit re-adjusts continuously.
Forcing
How the discharge changes with time.
Drawing
Only the picture is exaggerated, not the numbers.
Camera
Quality
Layers
Model checks
The science
A tunnel held open by friction heat
Meltwater that reaches the glacier bed flows in a channel of cross-section S (Rothlisberger, 1972). Water losing potential energy as it flows down the hydraulic-potential gradient ψ warms and melts the channel wall: the opening rate per unit length is Q ψ / (ρi L) (L = 3.34e5 J/kg). Ice creep closes the channel at 2 A (N/n)n S (Glen's law, n = 3, A = 2.4e-24 Pa-3 s-1 for temperate ice), driven by the effective pressure N = pice − pwater = (1 − λ) ρi g H. Thick ice squeezes a conduit shut fast; thin ice near the terminus lets it stay wide, which is why ice caves form there.
The gradient depends on the flow: Manning's law for a semicircular channel gives Q = k S4/3 ψ1/2, so a full (pressurised) conduit has ψ = Q2 / (k2 S8/3). When the conduit is big enough that this falls below the open-channel gradient ρw g tan θ, the water runs as a free-surface stream and ψ stays at that value (the model uses the larger of the two). With that, each node obeys dS/dt = Q ψ / (ρi L) − c S.
Closed forms used as the check
Pressurised branch: with u = S11/3 the equation becomes linear, du/dt = (11/3)(a Q3 − c u) with a = 1/(k2 ρi L), so for constant Q Seq = (a Q3/c)3/11 and the approach is exp(−(11/3) c t). Open branch: Seq = Q ρw g tanθ / (ρi L c), relaxing as exp(−c t). The page integrates the full time-dependent model (Runge-Kutta, adaptive sub-steps) and tests it against both exact solutions, against the relaxed steady state, and against the log-log slope of S versus Q (9/11 = 0.818 pressurised, 1 open).
Honest simplifications: the effective pressure is prescribed from the ice thickness and the slider λ instead of being solved from water mass conservation and the pressure gradient; discharge is routed instantly down the profile (no flood-wave delay); one channel, no tributary arborescence, no ice flow or sliding, no heat exchange with the air in the cave; a fixed bed slope and a fixed thickness profile (170 m at the head, 55 m at the mouth); Manning roughness 0.05 s m-1/3. The conduit in the pictures is drawn exaggerated (slider) and capped inside the ice; the numbers in the read-outs are the real values. Typical-literature constants only; check against a glaciology text before citing.
What to try
- Start with Melt season: the conduit starts small (winter closure) and widens as discharge grows; compare the cyan section with the yellow outline of where it is heading.
- Open the third graph: closure time is days near the head, where ice is thick, but hundreds of days at the mouth, so the cave mouth lags behind the water supply.
- Try Moulin surge: extra water entering at the moulin enlarges the channel downstream of it only.
- Stiff cold ice closes the same conduit more slowly than soft ice; compare the two presets.
- Switch to the Cave mouth camera to see the opening cut into the terminus face.