Risk prediction of pine wilt disease based on graphical convolutional network in China

Pine wilt disease is one of the most destructive conifer diseases affecting pine species worldwide. Since its introduction to China in 1982, it has infected more than one billion pine trees, leading to significant ecological and economic losses. To enable precise prevention and control of pine wilt disease, this study first conducted exploratory visual analysis of the spatiotemporal data of pine wilt disease outbreaks and environmental factors to identify high-risk areas. A risk prediction model for pine wilt disease was then developed using a graph convolutional network (GCN). After validating the model's accuracy, national-scale county-level risk prediction results were generated.

The results showed that a total of 755 counties across 24 provinces in China were predicted as risk areas, with 219 of these being classified as high-risk zones. The disease exhibits a 'northward expansion and westward progression' trend, with the southeastern region, Beijing-Tianjin-Hebei, and the three northeastern provinces being the main concentrations of risk areas. In addition, 28 cities with no reported outbreaks were also predicted as potential high-risk areas.

These findings suggest that GCN technology demonstrates high accuracy and potential in the spatiotemporal risk prediction of pine wilt disease, providing important technical support for relevant departments in early warning and the formulation of prevention and control strategies. © 2025 Society of Chemical Industry.