At the Surface Heat Budget of the Arctic Ocean (SHEBA) program's field site in the northern Chukchi Sea, snow and ice meltwater flow was found to have a strong impact on the heat and mass balance of sea ice during the summer of 1998. Pathways and rates of meltwater transport were derived from tracer studies (H218O, 7Be, and release of fluorescent dyes), complemented by in situ sea-ice permeability measurements. It was shown that the balance between meltwater supply at the surface (averaging between 3.5 and 10.5 mm d−1) and ice permeability (between <10−11 and >10−9 m2) determines the retention and pooling of meltwater, which in turn controls ice albedo. We found that the seasonal evolution of first-year and multiyear ice permeability and surface morphology determine four distinct stages of melt. At the start of the ablation season (stage 1), ponding is widespread and lateral melt flow dominates. Several tens of cubic meters of meltwater per day were found to drain hundreds to thousands of square meters of ice through flaws and permeable zones. Significant formation of underwater ice, composed between <30 and >50% of meteoric water, formed at these drainage sites. Complete removal of snow cover, increase in ice permeability, and reductions in hydraulic gradients driving fluid flow mark stage 2, concurrent with a reduction in pond coverage and albedo. During stage 3, maximum permeabilities were measured, with surface meltwater penetrating to 1 m depth in the ice and convective overturning and desalination found to dominate the lower layers of first-year and thin multiyear ice. Enhanced fluid flow into flaws and permeable zones was observed to promote ice floe breakup and disintegration, concurrent with increases in pond salinities and 7Be. Advective heat flows of several tens of watts per square meter were derived, promoting widening of ponds and increases in pond coverage. Stage 4 corresponds to freeze-up. Roughly 40% of the total surface melt was retained by the ice cover within the ice matrix as well as in surface and under-ice ponds (with a total net retention of 15%). Based on this work, areas of improvement for fully prognostic simulations of ice albedo are identified, calling for parameterizations of sea-ice permeability and the integration of ice topography and refined ablation schemes into atmosphere-ice-ocean models.