Abstract:

Objective: In this study, multi-source data comprising natural climate variables, anthropogenic activity indicators, and geospatial features is used to analyze the key factors influencing the spread and expansion of pine wilt disease (PWD). Focusing on the ecological invasion process of PWD of ‘introduction-colonization-expansion’, a predictive model applicable at a fine spatial scale is constructed, aiming to achieve precise identification and early warning of high-risk outbreak areas of pine wilt disease. Method: This study utilized subcompartment-level outbreak records of PWD in Jiangsu Province published by the National Forestry and Grassland Administration of China. Based on the ecological characteristics and spatial distribution patterns of PWD, a total of 25 influencing variables were selected, covering natural climate conditions, human activity, and spatial features. Principal Component Analysis (PCA) was used for dimensionality reduction, and Spearman correlation analysis and the Apriori data mining algorithm were applied to examine the interactions between each influencing factor and the occurrence of PWD. Bayesian estimation was employed to enhance the feature of the variables. A Grey Wolf Optimizer-Cellular Automata (GWO-CA) model was constructed to simulate the spatiotemporal spread of PWD. The model’s predictive performance was further evaluated through horizontal comparison with five mainstream machine learning models, with precision, recall, and AUC as evaluation metrics. Result: The Grey Wolf Optimizer-Cellular Automata model developed in this study exhibited excellent performance in predicting the new occurrence of pine wilt disease in subcompartment. The model achieved a recall rate of 78.5%, and significantly outperformed the other five mainstream machine learning models. Additionally, the model yielded an AUC value of 89.0%, indicating a high level of predictive accuracy and discriminative ability in identifying new outbreak locations. This study also underscored the critical role of geospatial features in forecasting the spread of pine wilt disease, and confirmed the strong suitability of cellular automata for modeling complex spatiotemporal data, especially at fine spatial scales. Conclusion: This study has identified timber transportation as a key driver of the spread of pine wood nematode, and temperature and precipitation differences also exert significant influence on outbreak risk. As a modeling approach that integrates spatial heterogeneity and temporal dynamics, the Cellular Automata model has proven to be highly adaptable and effective for complex ecological data analysis and invasive species risk assessment. It offers robust technical support for the precise prevention and efficient management of pine wilt disease.

Key words: pine wilt disease, spread prediction model, big data, cellular automata, data mining