Sunlight absorbed by the floe is Qabs = S·(1 − α), where S is incoming shortwave and α is the surface albedo. Each surface type reflects differently — fresh snow α≈0.84, bare sea ice α≈0.55, open ocean α≈0.06 — and a melt pond sits in between, darkening as it deepens:
α_pond(h) = α_water + (α_ice − α_water)·e^(−h/h₀)
Q_abs = S·(1 − α) h = pond depth, h₀ ≈ 40 cm
Each simulated day, absorbed energy above a small conductive-loss baseline melts the snowpack, exposing bare ice; part of that bare ice then floods with meltwater pooling in low spots (modeled with a fixed, ranked micro-topography so ponds always start in the same low-lying cells); pond depth grows with further melt and shrinks with drainage through cracks and seal holes; and once a pond has eroded deep enough it melts fully through to open water. Every one of those transitions lowers the area-averaged albedo, which raises Qabs, which speeds every transition again — a positive feedback loop with no built-in ceiling other than running out of ice.
- Incoming shortwave — solar forcing at the surface; higher values feed every step of the loop.
- Initial snow depth — a deeper snowpack delays bare-ice exposure and pond formation.
- Meltwater drainage — models ice permeability: high drainage caps pond depth (and keeps albedo higher) even as ponds spread; low drainage lets ponds deepen and darken fast.
This is the same mechanism sea-ice models (e.g. CICE's melt-pond schemes) parameterize to reproduce the sharp late-summer albedo drop observed over real Arctic floes — distinct from the whole-Arctic seasonal extent cycle or a zero-dimensional global energy balance.