Résumé
During the last deglaciation, the Bolivian Altiplano (15e23 S, 66e70 W) was occupied by paleolake Tauca covering, at least, ~51,000 km 2 at its maximum highstand between 16.5 and 15 ka. Twenty-five hundred years later, after a massive regression, a new transgressive phase, produced paleolake Coi-pasa, smaller than Tauca and restricted to the southern part of the basin. These paleolakes were overlooked at the west by the Sajama ice cap. The latter provides a continuous record of the oxygen isotopic composition of paleo-precipitation for the last 25 ka. Contemporaneously to the end of paleolake Tauca, around 14.3 ka, the Sajama ice cap recorded a significant increase in ice oxygen isotopic composition (d 18 O ice). This paper examines to what extent the disappearance of Lake Tauca contributed to precipitation on the Sajama summit and this specific isotopic variation. The water d 18 O values of paleolakes Tauca and Coipasa (d 18 O lake) were quantitatively reconstructed from 18.5 to 11.7 ka based on diatom isotopic composition (d 18 O diatoms) and ostracod isotopic composition (d 18 O carbonates) retrieved in lacus-trine sediments. At a centennial time scale, a strong trend appears: abrupt decreases of d 18 O lake during lake fillings are immediately followed by abrupt increases of d 18 O lake during lake level stable phases. The highest variation occurred at ~15.8 ka with a d 18 O lake decrease of about ~10‰, concomitant with the Lake Tauca highstand, followed ~400 years later by a 7‰ increase in d 18 O lake. A simple hydro-isotopic modeling approach reproduces consistently this rapid "decreaseeincrease" feature. Moreover, it suggests that this unexpected re-increase in d 18 O lake after filling phases can be partly explained by an equilibration of isotopic fluxes during the lake steady-state. Based on isotopic calculations during lake evaporation and a simple water stable isotopes balance between potential moisture sources at Sajama (advection versus lake evaporation), we show that total or partial evaporation (from 5 to 60%) of pale-olake Tauca during its major regression phase at 14.3 ka could explain the pronounced isotopic excursion at Sajama ice cap. These results suggest that perturbations of the local hydrological cycle in lacustrine areas may substantially affect the paleoclimatic interpretation of the nearby isotopic signals (e.g. ice core or speleothems).