Ascertaining Predictability Cognizance for the Prediction of Reference Evapotranspiration

Abstract

Reference evapotranspiration (ET0) is a rudimental variable in the estimation of crop water requirement, and preparation of irrigation schedule. Prediction of ET0 is a necessitous one for estimation of crop water requirement in future time step. In this paper ET0 is predicted using Artificial Neural Network (ANN) by different inputs Like Temperature, Cloud cover, Vapor pressure, Precipitation and its combinations by various models. Before prediction, the predictability of all the input time series is calculated individually and the effect of predictability on prediction is analyzed in models having single predictor. In spite of inserting additional predictor in input, the reason for increase of Root mean squared error is justified in terms of predictability in the models having multiple predictors. Also it is seen that the performance of models with multiple predictors is better when compared to single predictor models in the estimation of ET0.