USING ARTIFICIAL NEURAL NETWORK TO ESTIMATE REFERENCE EVAPOTRANSPIRATION

Irrigation, when rationally used, can contribute to the efficient performance of the agribusiness. Planning irrigation, monitoring the soil moisture, the rainfall and the reference evapotranspiration (ET 0 ) is necessary for a rational water management. The FAO Penman-Monteith (FAO PM) method is the standard method for estimating ET 0 , but in some cases, the use of this method is restricted due to missing some climatic variables. For this reason, methods with a lower number of meteorological variables, as temperature values, are quite often used. This study aims to propose an artificial neural network (ANN) to estimate the ET 0 as a function of maximum and minimum air temperatures for the city of Salinas, Minas Gerais State, Brazil. After training, validation and comparison with the Hargreaves methodology, it was observed the existence of a good correlation between the values estimated by the standard method and those estimated by ANN, with the performance index classified as optimal, better than the Hargreaves methodology one. The use of ANN proved to be an excellent alternative for ET 0 estimation, reducing the costs of acquiring climatic data.