ESTIMATIVA DA EVAPOTRANSPIRAÇÃO DE REFERÊNCIA ATRAVÉS DE REDES NEURAIS ARTIFICIAIS

Abstract

This work was carried out with the daily weather data of the maximum and minimum air temperature, solar radiation, wind speed, and relative humidity of maximum and minimum air concerning the city Areia - Paraíba, collected from the database of the database Instituto Nacional de Meteorologia (INMET) after collection was estimated reference evapotranspiration by the standard Penman-Monteith in the period between 2008 and 2012, with the objective use of artificial neural networks to estimate ET0 and evaluate their performance, we used the MATLAB to create the RNA template with architecture [6-7-1]. The training phase was used climate variables from 2008 to 2011 as input data and ET0 as output data, and to test the network was used the 2012 climate data as input to the prediction of ET0, all input data network were normalized in the range [-1, 1]. Excellent results were obtained by comparing the obtained ET0 the network from ET0 by the standard method with the relative error rate of 0.5%, 0.003 SSE, MSE 0.054, the coefficient of determination r² was 0.9966, the d accuracy coefficient of 0.955 and performance coefficient c of 0.953 which is classified as excellent by CAMARGO and Sentelhas (1997), showing that it is possible to obtain good quality results using the RNAs to estimate the reference evapotranspiration.