Résumé : The evolution of the main physico-chemical properties of the unflooded 90-cm-thick first-year sea-ice cover at the Ice Station POLarstern (ISPOL) "clean site" is described. ISPOL was an international experiment of the German research icebreaker R.V. Polarstern. The vessel was anchored to an ice floe for an observation period of 5 weeks, during the early summer melt onset in the Western Weddell Sea. The "clean site" was specially designed and accessed so as to prevent any trace metal contamination of the sampling area. Observations were made at 5-day intervals during December 2004 in the central part of the main floe. Results show the succession of two contrasting phases in the behavior of the brine network (brine channels, pockets, and tubes). Initially, brine salinity was higher than that of sea-water, leading to brine migration and a decrease in the mean bulk salinity of the ice cover. This process is highly favored by the already high bulk porosity (14%), which ensures full connectivity of the brine network. Gravity drainage rather than convection seems to be the dominant brine transfer process. Half-way through the observation period, the brine salinity became lower than that of the sea-water throughout the ice column. The brine network therefore switched to a "stratified" regime in which exchange with sea-water was limited to molecular diffusion, strongly stabilizing the bulk mean sea-ice salinity. During the transition between the two regimes, and in areas closer to ridges, slush water (resulting from a mixture of snow meltwater and sea water accumulated at the snow-ice interface) penetrated through the growing "honeycomb-like structure" and replaced the downward draining brines. This resulted in a slight local replenishment of nutrients (as indicated by dissolved silicic acid). However, as a whole, the described decaying regime in this globally unflooded location with limited snow cover should be unfavorable to the development of healthy and active surface and internal microbial communities. The switch from gravity to diffusion controlled transport mechanisms within the ice column also should affect the efficiency of gas exchange across the sea-ice cover. The observed late build-up of a continuous, impermeable, superimposed ice layer should further significantly hamper gas exchange. Statistical estimates of the evolution of the ice thickness during the observation period and salinity trends of the under-ice water salinity down to 30 m corroborate model predictions of a moderate bottom melting (5-10 cm) from ocean heat fluxes. © 2008 Elsevier Ltd. All rights reserved.