A long-standing problem of great interest among the meteorological and hydrological communities has been how to represent the spatial distribution of precipitation at small scales in regions without radar coverage and with only a sparse rain gauge network. In this case, satellite derived Quantitative Precipitation Estimates are an extremely powerful tool for obtaining rainfall pattern that can be used for by distributed hydrologic models to produce forecasts of discharge. The main purpose of this paper is to present the performance of the South American version of NOAA/NESDIS Auto-Estimator satellite rainfall estimation technique in selected regions of the Del Plata basin (60º - 47º W, 35º - 22º S). Two major events (more than 52 mm in 24 hours) during May and June 2001 were chosen to evaluate the algorithm. The version used in this work includes a new screening technique to separate raining and non raining pixels according to the difference between the pixel brightness temperature and the mean value of surrounding pixels. The rain-rate curve is also corrected according to the moisture environment of the storm (expressed as a product between relative humidity and precipitable water). Preliminary evaluation of rainfall cases suggested that the Auto-Estimator generated too little rainfall compared to 24-hour measurements from the existing rain gauge network. Scarce availability of southern hemisphere GOES 8 images (with gaps of up to 90 minutes between images) may be responsible for this. A new integration algorithm was tested in order to reduce this difference, and it produced a significant reduction in dry bias and improvement of other statistical parameters.