Using Neuro-fuzzy and linear models to estimate reference Evapotranspiration in South region of Algeria (A comparative study)

In order to estimate daily reference evapotranspiration (ETo) in arid region of Algeria, Adaptive Neuro-Fuzzy Inference System (ANFIS) and regression methods as Robust Regression (RR), Bayesian Regression (BR) and Multiple Linear Regression (MLR) techniques were used to develop models based on four explanatory climatic factors: temperature, relative humidity, wind speed and sunshine duration. These factors have been used as inputs, and ETo values computed by the Penman-Monteith formula have been used as outputs. Determination coefficient (R-2), root mean square error (RMSE), Mean absolute error (MAE), mean absolute relative error (MARE) and Nash-Sutcliffe efficiency coefficient (NSE) were used to evaluate the performance of models developed with different input configurations. We concluded that RR, BR and MLR models were able to successfully estimate ETo, but ANFIS technique seems to be more powerful. Thus, the obtained results by the best ANFIS model, during the test phase are: 0.98, 0.27 (mm/day)(2), 0.36 (mm/day) and 5.52 % respectively for R, MAE, RMSE and MARE.