The Korean Journal of Crop Science ISSN:0252-9777(Print) 2287-8432(Online) 한국작물학회지

Rice is one of the world’s three major staple crops and serves as a primary food source for over half of the global population. Accurately diagnosing the growth stages of rice is essential for effective growth management. In particular, the heading date is influenced significantly not only by genetic factors but also by environmental factors such as temperature. Moreover, recent climate changes have increased the interannual variation in heading dates, making precise cultivation management more challenging. This study analyzes and predicts the heading date of Ilpum rice, a variety commonly cultivated in Jeongeup City, using deep learning techniques. The deep learning model employed is a gated recurrent unit (GRU) model. To reflect the time-series characteristics of the data, time-related variables are extracted and normalized. The model’s generalization performance and learning efficiency are enhanced with GRU layers, Dropout, and Batch Normalization. For optimization, techniques such as EarlyStopping and learning rate adjustment are applied to ensure stable and efficient training. The predicted temperature range for 2025 to 2034 is estimated to be between 27.46°C and 30.59°C, which is either within or close to the previous temperature range, suggesting relative stability in temperature during this period. When calculating the slope of the trend line, the equation is y=0.083x－139, indicating an annual temperature increase of approximately 0.08°C. While this may lead to long-term temperature increases, the predicted temperature range suggests that heading dates of rice will not undergo significant changes from 2025 to 2034. Additionally, this study demonstrates that time-series analysis and future temperature predictions can provide valuable insights for crop growth. Beyond predicting rice heading dates, this approach can be extended to various analyses, showcasing the potential of deep learning techniques in agriculture.