Artificial neural networks for rice yield prediction in mountainous regions

Abstract

Decision-making processes in agriculture often require reliable crop response models. The Fujian province of China is a mountainous region where weather aberrations such as typhoons, floods and droughts threaten rice production. Agricultural management specialists need simple and accurate estimation techniques to predict rice yields in the planning process. The objectives of the present study were to: (1) investigate whether artificial neural network (ANN) models could effectively predict Fujian rice yield for typical climatic conditions of the mountainous region, (2) evaluate ANN model performance relative to variations of developmental parameters and (3) compare the effectiveness of multiple linear regression models with ANN models. Models were developed using historical yield data at multiple locations throughout Fujian. Field-specific rainfall data and the weather variables (daily sunshine hours, daily solar radiation, daily temperature sum and daily wind speed) were used for each location. Adjusting ANN parameters such as learning rate and number of hidden nodes affected the accuracy of rice yield predictions. Optimal learning rates were between 0·71 and 0·90. Smaller data sets required fewer hidden nodes and lower learning rates in model optimization. ANN models consistently produced more accurate yield predictions than regression models. ANN rice grain yield models for Fujian resulted in R² and RMSE of 0·87 and 891 vs 0·52 and 1977 for linear regression, respectively. Although more time consuming to develop than multiple linear regression models, ANN models proved to be superior for accurately predicting rice yields under typical Fujian climatic conditions.