基于元胞自动机模型模拟松材线虫病的传播
Simulating the spread of pine wilt disease based on cellular automaton model
[目的]在之前的松材线虫病预测研究中，使用的影响因素往往不够全面，只考虑了自然因素的作用，缺少对人为影响因素的利用，同时在建模过程中忽略了传播扩散过程和对空间信息的利用。为了探究影响松材线虫病传播扩散的主要影响因素，结合人类活动因素和空间特征，围绕松材线虫“传入-定殖-扩散”的生态入侵过程，构建一种适用于更小空间尺度数据的预测模型，实现对高风险发生地区的精准预测和早期预警。[方法]本研究使用国家林业草原局获得的江苏省松材线虫病小班本底历史发生数据，基于松材线虫病的生态和地理分布特点，选取自然气候、地理信息以及人类活动因素作为影响因子数据，通过主成分分析技术(PCA)进行预处理，结合贝叶斯估计方法进行对影响因子数据进行特征增强，探究不同影响因子对于松材线虫病发生的作用关系。建立灰狼优化算法—元胞自动机模型来模拟松材线虫病的入侵扩散过程。同时与支持向量机模型(SVM)，随机森林模型(RF)，分类与回归树(CART)，广义线性模型(GLM)，人工神经网络模型(ANN)五种主流机器学习模型预测结果进行对比，对江苏省2022至2023的松材线虫病发生情况进行了模拟，计算召回率和精确率等评价指标对模型的性能进行比较。[结果]本研究所构建的灰狼优化算法—元胞自动机模型的召回率达到了80.4％，明显优于其他五种机器学习模型，同时F1分数达到了45.3％，显示模型在保持预测结果准确性的同时，还能有效地对松材线虫病新发小班进行预测。通过对比分析，本研究不仅还强调空间特征在松材线虫病传播预测研究中的重要性，同时证实元胞自动机模型在处理复杂时空数据和更精细尺度空间数据预测方面的高度适用性。[结论]本研究证实了温度、降水以及人口和GDP是影响是松材线虫病传播扩散的主要因素，松材线虫的发生主要集中在低海拔的地区。元胞自动机模型作为一种结合空间信息和时间序列的模型，在处理复杂的生态数据和入侵物种风险评估中具有更高的适用性，元胞自动机模型可以作为一种有效的工具为精准高效地采取预防措施提供技术支持。
[Objective]Previous studies on the prediction of pine wood nematode(PWN)disease often lacked comprehensiveness,focusing solely on natural factors and neglecting human influences and the process of spread and diffusion.In response,this study aims to investigate the major factors influencing the spread of PWN by integrating human activity and spatial characteristics.Centered around the ecological invasion process of"introduction-settlement-diffusion,"a predictive model suitable for smaller spatial scale data was developed to enable precise forecasting and early warning in high-risk areas.[Method]This study utilized historical incidence data of PWN in Jiangsu Province obtained from the National Forestry and Grassland Administration.Based on the ecological and geographical distribution characteristics of the disease,natural climate,geographical information,and human activity factors were selected as the data variables.The data was preprocessed using Principal Component Analysis(PCA)to reduce dimensions and extract significant features,coupled with Bayesian estimation methods to enhance the features of the impact factors.A Grey Wolf Optimizer-Cellular Automata model was developed to simulate the invasion and spread of PWN.This model was compared against five mainstream machine learning models:Support Vector Machine(SVM),Random Forest(RF),Classification and Regression Trees(CART),Generalized Linear Model(GLM),and Artificial Neural Network(ANN).The models were used to simulate the PWN occurrence in Jiangsu Province for the years 2022 to 2023,evaluating their performance through metrics such as recall and precision rates.[Result]The Grey Wolf Optimizer-Cellular Automata model achieved a recall rate of 80.4%,significantly outperforming the other models,with an F1 score of 45.3%.This demonstrates the model's accuracy in predicting new occurrences of PWN while maintaining high predictive precision.The study not only highlights the importance of spatial features in predicting the spread of PWN but also confirms the high applicability of the Cellular Automata model in managing complex spatio-temporal data and fine-scale spatial predictions.[Conclusion]This study confirms that temperature,precipitation,population,and GDP are primary factors influencing the spread of PWN,with occurrences mainly concentrated in low-altitude areas.The Cellular Automata model,integrating spatial information and time series data,proves highly suitable for handling complex ecological data and assessing risks of invasive species.As an effective tool,the Cellular Automata model provides technical support for taking precise and efficient preventative measures against PWN disease.
周宏威;李永正;陈怡帆;胡浩昌;张思岩;崔迪;
东北林业大学 黑龙江 哈尔滨 150040;东北林业大学 黑龙江 哈尔滨 150040;东北林业大学 黑龙江 哈尔滨 150040;东北林业大学 黑龙江 哈尔滨 150040;东北林业大学 黑龙江 哈尔滨 150040;东北林业大学 黑龙江 哈尔滨 150040;
Q958.15;S154.5
松材线虫病 传播预测模型 贝叶斯估计 元胞自动机模型 灰狼优化算法
Pine Wilt Disease Spread Prediction Model Bayesian Estimation Cellular Automaton Model Grey Wolf Optimizer
79-92 / 14
评论