Application of artificial intelligence to estimate the reference evapotranspiration in sub-humid Doon valley

Abstract

Estimation of evapotranspiration (ET) is an essential component of the hydrologic cycle, which is also requisite for efficient irrigation water management planning and hydro-meteorological studies at both the basin and catchment scales. There are about twenty wellestablished methods available for ET estimation which depends upon various meteorological parameters and assumptions. Most of these methods are physically based and need a variety of input data. The FAO-56 Penman– Monteith method (PM) for estimating reference evapotranspiration (ET0) is recommend for irrigation scheduling worldwide, because PM generally yields the best results under various climatic conditions. This study investigates the abilities of artificial neural networks (ANN) to improve the accuracy of monthly evaporation estimation in sub-humid climatic region of Dehradun. In the first part of the study, different ANN models, comprising various combinations of training function and number of neutrons were developed to estimate the ET0 and it has been compared with the Penman–Monteith (PM) ET0 as the ideal (observed) ET0. Various statistical approaches were considered to estimate the model performance, i.e. Coefficient of Correlation (r), Sum of Squared Errors, Root Mean Square Error, Nash–Sutcliffe Efficiency Index (NSE) and Mean Absolute Error. The ANN model with Levenberg–Marquardt training algorithm, single hidden layer and nine number of neutron schema was found the best predicting capabilities for the study station with Coefficient of Correlation (r) and NSE value of 0.996 and 0.991 for calibration period and 0.990 and 0.980 for validation period, respectively. In the subsequent part of the study, the trend analysis of ET0 time series revealed a rising trend in the month of March, and a falling trend during June to November, except August, with more than 90% significance level and the annual declining rate was found to 1.49 mm per year.