Evapotranspiration Modelling using Artificial Neural Network and Multiple Linear Regression Approach in Semi-Humid Region of Sikkim

Abstract This study investigates superiority between multiple linear regression (MLR) and artificial neural network (ANN) approach for prediction of weekly reference evapotranspiration (ET0) in semi-humid region of Sikkim, India. Daily meteorological parameters (1985–2009) were used for the estimation of ET0 using Penman Monteith equation and various combinations of meteorological parameters was used for MLR and ANN model development. Predicted ET0 were compared with ET0 estimated using coefficient of determination (R2), root mean square error (RMSE), mean absolute error (MAE), mean absolute relative error (MARE) and Nash Sutcliffe efficiency (NSE). Among ANN models, ANN model 4 with one hidden layer and four neurons had smallest RMSE (0.62 & 0.49 mm/day) and highest R2 (0.75 & 0.82) for training and testing phase. ANN model 4 had highest NSE compared to MLR model 4. Comparison between MLR and ANN models revealed more accuracy of ANN to predict weekly ET0 for the sub humid region like Tadong in East Sikkim district of North-eastern India. Study show the effectiveness of the both MLR and ANN models for estimation of the ET0 with minimal meteorological variables and increased parameters increases the model prediction efficiency. The findings of the study could be helpful for similar studies in the data scarce northeast Indian part and other identical climatic regions.