Vol. 12, Issue 5 (2023)

Author(s):
Vikas Kumar Singh, Manish Kumar, DK Singh, Shivam, Vijay Kumar Singh, Pramod Kumar Mishra and RJ Singh

Abstract:
Evapotranspiration requirement is a principal component in planning any crop and precise prediction of this component would reduce the squandering of huge quantities of water. The aim of this study was to develop an Artificial Neural Network (ANN) model for prediction of daily reference evapotranspiration (ET₀) in the sub-tropical regions of India. Feed-forward Back-propagation Neural Network model is employed in this study to evaluate the performance of Artificial Neural Networks in comparison with Empirical FAO Penman-Monteith (FAO-56) equation in predicting reference evapotranspiration. Daily climatic data were collected and used for analysis of best fit ANN model. After the network structure and parameters were determined reasonably, the network was trained with daily climatic data (maximum and minimum temperature, relative humidity, solar radiation and wind speed) as input and the FAO-56 estimated ET₀ as output. The ANN learning model recognized the evapotranspiration patterns with acceptable accuracy, with RMSE of 0.211 in comparison to the results of FAO-56, coefficient of determination of 0.910, and Linear Correlation Coefficient (r) of 0.948 demonstrating the best approximation for the 5-4-1 network architecture, with multilayers, back-propagation learning algorithm and learning rate of 0.01. The Mean Absolute Error (MAE) between ET₀ by ANN and FAO-56 equation was 0.308 mm. The result shows that the predictions of ET₀ using the developed ANN model were strongly correlated with the results of FAO-56 equation.