基于ANN-CA的外来入侵植物互花米草的扩散趋势预测研究

Prediction of Diffusion Trend of Invasive Plant Spartina Alterniflora Based on ANN-CA

采用遥感技术手段对九段沙互花米草(Spartina alterniflora)信息提取与动态监测基础上,开展滨海地区互花米草入侵时空特征分析、入侵地区景观演变规律以及预测互花米草扩散趋势等研究对于沿海城市生态环境协调发展具有重要的意义。选取邻域因子、生态属性和地理属性等相关变量因子,耦合人工神经网络(ANN)与元胞自动机(CA),基于2015-2019年间九段沙互花米草与其他湿地景观类型数据构建ANN-CA模型,采用3层BP神经网络,在不断训练与校正与验证模型精度的基础上,对2019年九段沙区域互花米草进行模拟,采用Lee-Sallee形状指数来对比模拟分类结果与实际分类结果在空间分布层面的相似度,并进行相应的模型校正,在此基础上预测2025年九段沙互花米草扩散与群落分布格局,为互花米草的扩张趋势、景观演变以及分布扩散预测提供理论依据和科学基础。结果表明,1)1995-2025年间互花米草自从引种定居到不断扩张,所呈现的面积变化图近乎S型曲线,其中2015-2019年面积增长了847 hm^(2),2019-2025年面积增长了646 hm^(2),因此互花米草在九段沙自然保护区的动态扩散,分别为定居阶段、滞缓阶段、快速增长阶段以及缓慢增长阶段。2)2025年九段沙区域互花米草的扩张速率从趋势看已进入扩张的缓慢期,与前几年相比尽管有所增加,但趋势有所放缓。3)构建的ANN-CA模型能够较好地对九段沙自然保护区的植被演变进行模拟和预测。

Dynamic monitoring and information extraction of Spartina alterniflora in Jiuduansha were conducted using remote-sensing technology.The analysis of the spatiotemporal characteristics of Spartina alterniflora invasion in coastal areas,landscape evolution patterns in invaded areas,and prediction of Spartina alterniflora diffusion trends are of great significance for the coordinated development of the ecological environment in coastal cities.By selecting relevant variable factors,such as neighborhood factors,ecological attributes,and geographical attributes,coupled with artificial neural networks(ANN)and Cellular Automata(CA),an ANN-CA model was constructed based on the data of Spartina alterniflora in the Jiuduansha area and other wetland landscape types from 2015 to 2019.A three-layer BP neural network was used to simulate Spartina alterniflora in the Jiuduansha area in 2019,and the Lee-Sallee shape index was used to compare the spatial similarity between simulated classification results and actual classification results,and to perform corresponding model corrections based on continuous training,correction,and validation of model accuracy.In this study,the diffusion and community distribution patterns of Spartina alterniflora in Jiuduansha were predicted for 2025,thereby providing a theoretical and scientific basis for predicting the expansion trends,landscape evolution,and distribution of Spartina alterniflora.The results showed that 1)from 1995 to 2025,the area change of Spartina alterniflora showed an almost S-shaped curve from its introduction and settlement to continuous expansion.The area increased by 847 hm^(2) from 2015 to 2019 and 646 hm^(2) from 2019 to 2025.Therefore,the dynamic diffusion of Spartina alterniflora in the Jiuduansha Nature Reserve could be divided into three stages:settlement,stagnation,rapid growth,and slow growth.2)The expansion rate of Spartina alterniflora in the Jiuduansha area in 2025 has entered a slow period of expansion;although it has increased compared to previous years,the trend has slowed.3)The constructed ANN-CA model could effectively simulate and predict vegetation evolution in the Jiuduansha Nature Reserve.The simulation results provide a theoretical and scientific basis for predicting the expansion trend of Spartina alterniflora,landscape evolution,and future distribution dynamics of Spartina alterniflora.