We explore the reliability of large-scale climate variables, namely precipitation and temperature, as inputs for a basin-lake hydrological model in central Argentina. We used data from two regions in NCEP/NCAR reanalyses and three regions from LMDZ model simulations forced with observed sea surface temperature (HadISST) for the last 50 years. Reanalyses data cover part of the geographical area of the Sali-Dulce Basin (region A) and a zone at lower latitudes (region B). The LMDZ selected regions represent the geographical area of the Sali-Dulce Basin (box A), and two areas outside of the basin at lower latitudes (boxes B and C). A statistical downscaling method is used to connect the large-scale climate variables inferred from LMDZ and the reanalyses, with the hydrological Soil Water Assessment Tool (SWAT) model in order to simulate the Rio Sali-Dulce discharge during 1950-2005. The SWAT simulations are then used to force the water balance of Laguna Mar Chiquita, which experienced an abrupt level rise in the 1970's attributed to the increase in Rio Sali-Dulce discharge. Despite that the lowstand in the 1970's is not well reproduced in either simulation, the key hydrological cycles in the lake level are accurately captured. Even though satisfying results are obtained with the SWAT simulations using downscaled reanalyses, the lake level are more realistically simulated with the SWAT simulations using downscaled LMDZ data in boxes B and C, showing a strong climate influence from the tropics on lake level fluctuations. Our results highlight the ability of downscaled climatic data to reproduce regional climate features. Laguna Mar Chiquita can therefore be considered as an integrator of large-scale climate changes since the forcing scenarios giving best results are those relying on global climate simulations forced with observed sea surface temperature. This climate-basin-lake model is a promising approach for understanding and simulating long-term lake level variations.