Determination of Evapotranspiration from Stream Flow with the Help of Classified Neurogenetic Model

Abstract

The efficiency of an adaptive neuro-fuzzy computing technique in estimation of reference evapotranspiration (ET0) from limited stream flow is investigated in this chapter. Forty years of stream flow data are clustered into nine groups, with the help of 11 basic rules where a ranking mechanism is used to sort the stream flows, and from these sorted data, top 10 stream flows and the same for that year are applied to the model as input. Corresponding ET0, derived from Penman equation, is also classified into nine groups with the help of the same 11 basic rules and applied as output of the model. The stream flow dataset predicted by an HECHMS model is also clustered and applied into the same model to verify model flexibility and efficiency. Correct classification rate, coefficient of relationship, standard deviation, and coefficient of efficiency are used as the measure of the applicability of neural model in predicting ET0 from stream flow. These are compared with the ET0 derived from the Penman model. The comparison results reveal that the neuro-fuzzy models could be employed successfully in modeling ET0 from stream flow, if the datasets of both ET0 and stream flow are clustered according to the basic rules developed as per the stability of the catchment.