Neuro-Fuzzy Modelling of Reference Evapo-transpiration

Abstract

Reference evapo-transpiration (ETo) can be either directly estimated using lysimeter or water balance approach, or estimated indirectly using the climatological data. However, it is not always possible to obtain ETo value using lysimeter, as it is a time consuming method and needs precise and carefully planned experiments. Owing to the difficulty of obtaining accurate field measurements, reference evapo-transpiration is generally estimated from weather parameters. The FAO Penman-Monteith method has now been accepted as standard method to estimate reference evapo-transpiration. However, it requires several weather parameters. The ensuing study is an attempt to predict ETo using adaptive neuro-fuzzy inference system with fewer and simple weather data for Nagini watershed, located near Chamba-Ranichauri, on Rishikesh- Uttarkashi route, in Tehri Garhwal district of Uttarakhand, India. The ETO predicted hy adaptive neuro-fuzzy inference system was found to be comparable with the ETo estimated by FAO Penman-Monteith method. The developed model was validated by testing its performance using correlation coefficient (R² = 0.931, root mean square error (RMSE = 0.48) and coefficient of variation of the residual error (CVRE = 0.17). It shows the applicability of developed neuro-fuzzy model to predict ETo for future events. Apart from this, developed model requires lesser input variables than the input required by different empirical methods.