Modeling in agriculture represents an important tool to understand processes as water and nutrient losses by drainage, or to test different conditions and scenarios of soil and crop management. Among the existing computational models to describe hydrological processes, SWAP (Soil, Water, Atmosphere and Plant model) has been successfully used under several conditions. This model was originally developed to simulate short cycle crops and its use also to cover longer cycles, e.g. perennial crops, is a new application. This report shows a SWAP application to a mature coffee crop over one-production cycle, focusing on deep drainage losses in a typical soil-plant-atmosphere system of the Brazilian savanna (Cerrado). The estimated annual deep drainage Q=. 1019. mm obtained by SWAP was within 99% of the value determined by the climatologic water balance of 1010. mm. Monthly results of SWAP for Q compared to the estimative using the climatological method presented a determination coefficient of 0.77. A variety of coffee fertigation scenarios were simulated using SWAP and compared to farmer's management scenario, leading to the conclusion that larger irrigation intervals result in lower Q losses, better water productivity and higher crop yield.