Developing a Water Management Model for Paddy Growth Using Artificial Neural Networks

Water management plays a crucial role in paddy rice cultivation. Inappropriate use of irrigation water can hinder optimal plant growth and lead to wastage. Therefore, it is essential to optimize the irrigation water delivery system. This optimization can be achieved after developing a model that identifies the relationship between the irrigation system and plant growth. This research aims to develop a plant growth model influenced by the irrigation system. An artificial neural network model with a backpropagation algorithm is employed to predict plant growth under different irrigation treatments. The model development is based on a lab-scale rice cultivation experiment conducted over two growing seasons in 2021 and 2022, comparing a flooded system (CFI) and a more water-efficient intermittent irrigation system (IIS). The first growing season was used for model training, while the second season was for model validation. The developed model incorporates three inputs: soil moisture, plant height, and the number of tillers from the previous week. The outputs are the plant height and the number of tillers for the following week. The results of the model training indicate that the neural network model accurately predicts plant height and the number of tillers, with R 2 values of 0.99 and 0.93, respectively. For validation, the R 2 values are slightly lower, at 0.97 and 0.65. These results suggest that the model can effectively predict plant height and the number of tillers.