A deep learning-based model for forecasting rice leaf folder Cnaphalocrocis medinalis

The rice leaf folder (Cnaphalocrocis medinalis) is one of the worldwide pests that pose a significant threat to rice crops. Severe infestations can lead to substantial yield reductions in rice production. Effective and timely prediction of this pest at a regional scale is essential for controlling and preventing its damage. Based on the daily field observations of moth population data from 2010 to 2022 at the Hunan Plant Protection and Quarantine Station in China, a composite model combining Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) called LSTM-GRU was proposed to predict the occurrence of the rice leaf folder. To validate the advantages of this model, the predictive results of LSTMGRU were compared with those of single models and other hybrid models. The results indicated that the LSTM-GRU composite model demonstrated greater stability and accuracy compared to the single models and other hybrid models. This provides a more precise method for predicting the occurrence of the rice leaf folder.