Evapotranspiration estimation using feed-forward neural networks

Two different feed-forward neural network algorithms, Levenberg–Marquardt (LM) and conjugate gradient (CG), are used for estimation of daily reference evapotranspiration (ET) from climatic data. The performances of the LM and CG algorithms in estimating ET are analyzed and discussed and various combinations of wind speed, solar radiation, relative humidity, air and soil temperature data as inputs to the artificial neural network (ANN) models are examined in the study so as to evaluate the degree of the effect of each of these variables on ET. The LM and CG training algorithms are compared with each other according to their convergence velocities in training and estimation performances of ET. The results of the ANN models are compared with those of multi-linear regression (MLR) and the empirical models of Penman and Hargreaves. Based on the comparisons, it was found that the neural computing technique could be employed successfully in modelling evapotranspiration process from the avaliable climatic data.