Estimation of daily reference evapotranspiration using artificial neural network for Raipur region in Chhattisgarh

Abstract

This study presents estimation of daily reference evapotranspiration using artificial neural network (ANN) for Raipur region of Chhattisgarh. An ANN is a flexible mathematical structure, which is capable of identifying complex nonlinear relationships between input and output data sets. To test artificial neural networks (ANN) for estimating reference evapotranspiration with limited climatic data, comparison has been made with Penman Montieth (PM) method. Data on daily maximum temperature (Tmax), minimum temperature (Tmin), relative humidity (RH), wind speed (WS) and sunshine hours (SSH) of 31 years were collected from the observatory of Department of Agrometeorology, IGKV, Raipur. Four different models have been developed based on input combination viz: M1 (Tmax, Tmin, RH (I), RH (II), WS and SSH), M2 (Tmax, RH (I), RH (II), WS and SSH), M3 (Tmax, WS and SSH) and M4 (Tmax and SSH). The analysis was carried out in MATLAB R2014b software. Performance evaluation of the models have been carried out by calculating mean absolute deviation (MAD), root mean square error (RMSE), absolute Prediction Error (APE), coefficient of correlation (CC), coefficient of efficiency (CE) and Index of Agreement (IOA). Upon comparison the Model M1R showed the highest CC, CE and IOA value as 99.7, 99.4 and 99.8 during training and highest CC, CE and IOA value as 99.7, 99.5 and 99.6 during testing period. The developed ANN models may therefore be adopted for estimating ET0 in the region with reasonable degree of accuracy.