Artificial Neural Networking on Crop Water Stress Index of Paddy Crop

Abstract

This study compares the performance of two artificial neural network models, the feed-forward back-propagation neural network (FF-BP-ANN) and the self-organizing map (SOM), in predicting the crop water stress index (CWSI) of rice. The CWSI is an important indicator of water stress in crops and is derived from key climatic variables, including air temperature (Ta), canopy temperature (Tc), and relative humidity (RH), which impact crop water consumption and irrigation requirements. The FF-BP-ANN and SOM models were evaluated by comparing their predicted CWSI values to empirically determined CWSI values. Statistical errors and x-y plots were used for performance assessment. Results showed that the SOM model achieved higher prediction accuracy (R2 = 0.97) compared to the FF-BP model (R2 = 0.86). Consequently, the SOM model is recommended as a reliable approach for estimating the CWSI of rice and similar crops. The findings of this study have implications for crop water management, as accurate estimation of CWSI is crucial for optimizing irrigation practices and water allocation. By utilizing the SOM model, farmers and agronomists can make informed decisions regarding irrigation scheduling, ultimately improving crop productivity and resource efficiency. In conclusion, this study demonstrates the superiority of the SOM model in predicting the crop water stress index of rice compared to the FF-BP model. However, further research is necessary to validate the applicability of these neural network models, particularly, for estimating CWSI across diverse agro-climatic regions. Accurate CWSI estimation models have the potential to enhance water management strategies and crop productivity in agriculture.