A new automated procedure for estimation of evapotranspiration for universal acceptance

Evapotranspiration (ET), a complex dynamic, nonlinear phenomenon, is composed of evaporation and transpiration. Numerous models for estimation of ET exist but none has performed equally well over all regions of the world. Besides, complex calculations, local calibrations and adjustments of parameters in some popular models render them unfit for field applications from non-expert users' point-of-view. This paper introduces, for the first time, a fully automated user-friendly procedure, requiring as input only climate data of the location. The procedure identifies the most dominant variables which influence ET for the location, recommends application of Artificial Neural Network (ANN)/Ridge Regression (RR), whichever is judged appropriate on a set of developed rules. It eliminates the need of several trials required in the ANN applications for choosing the best subset of variables. The procedure, tested on climate data of several regions of the world, is found to be worthy of applications in field.