Author(s): Shreenivas N. Londhe; Madhura Kulkarni

Linked Author(s):

Keywords: Evapotranspiration; Neural networks; Penman-Monteith equation

Abstract: Evapotranspiration (ET) is a major component of hydrologic cycle. Information about evapotranspiration is essential in water resource management, irrigation system designs and crop productions. Traditional and empirical methods of estimating ET require measurement of many meteorological parameters which need special equipment and careful observations. There is no formula or methods which can estimate it to the fullest of accuracy. This study aims to develop Artificial Neural Network models using Feed Forward Back Propagation algorithm (FFBP) and Generalized Regression Neural Network (GRNN) in order to estimate evapotranspiration. Daily meteorological data provided by three stations in Florida obtained from USGS site are used in model formulation. Models were created using measured input variables such as net radiation, solar radiation, wind speed, minimum, maximum and average temperature, minimum and maximum relative humidity and precipitation. Output is evapotranspiration measured by Eddy Covariance. The second part of the study is comparing results of ANN models with each other and also with ET estimates of empirical formula Penman Monteith (PM). The results of ANN model seem to be better when compared with Penman Monteith method.

DOI:

Year: 2018