摘要
【目的】基于随机森林模型原理,对当前气候环境以及21世纪50年代的美国白蛾在中国的空间分布、环境因子重要性、发生面积及迁移情况进行预测和分析,为美国白蛾有效防控提供理论参考。【方法】获取2011-2017年美国白蛾发生的县市级数据,利用地理信息系统ArcGIS中随机点生成工具生成未发生点数据。采用随机森林模型原理,选择19个气候因子以及海拔、坡度、坡向、植被覆盖率、有效光合辐射等5个环境因子,通过ArcGIS中的提取值工具提取发生点和未发生点的24个环境变量的值,将海拔、坡度、坡向进行离散化。使用R语言模拟2011-2030年美国白蛾的潜在生境分布模型,以ROC曲线验证模型的精度;利用构建好的模型分析环境因子重要性并排序,预测在2050s(2041-2060年)时期两种气候情景(RCP2.6、RCP4.5)下全中国范围内美国白蛾的潜在生境分布。【结果】ROC曲线分析表明,随机森林模型预测美国白蛾潜生境分布的训练数据和测试数据的AUC值分别为0.997和0.963,模型精度较高;当前时期下美国白蛾潜在生境(适生区)占中国国土面积的8.74%,其中低适生区、中适生区、高适生区和极高适生区面积分别占适生区总面积的41.47%、20.85%、18.90%和18.78%,适生区主要集中在华北地区东南部、中南地区北部、华东地区北部和东北地区南部。对美国白蛾潜在生境分布影响较大的环境因子依次是:海拔、植被覆盖率、最湿季平均气温、最暖季平均气温。在2050 s时期RCP2.6气候情景下,美国白蛾适生区总面积占中国国土面积的14.38%,其中低适生区、中适生区、高适生区和极高适生区面积分别占适生区总面积的51.87%、20.37%、16.49%和12.27%;在RCP4.5气候情景下,美国白蛾适生区总面积占中国国土面积的19.06%,其中低适生区、中适生区、高适生区和极高适生区面积分别占适生区总面积的51.14%、15.11%、20.
【Objective】 Hyphantria cunea is extremely harmful to plant that are highly susceptible to outbreaks of infection. Predicting the potential habitat of H. cunea is essential for its prevention and control, and this prediction is based on a random forest model, which predicts and analyzes spatial distribution, the importance of environmental factors, occurrence area and migration situation of H. cunea under the current climate and based on the data from the 1950 s;overall, these data can provide a theoretical basis for effective prevention and control for this pest. 【Method】 County and municipal data on the occurrence of H. cunea from 2011 to 2017 were obtained, and non-occurrence points were made with the create random points tool of ArcGIS. By adopting the principle of the random forest model, 19 climate and 5 environmental factors(altitude, slope, aspect, vegetation coverage and effective photosynthetic radiation) were selected, and the environmental variables of occurrence and non-occurrence points were extracted by extracting values using the points tool of ArcGIS. Then, the altitude, slope and aspect were discretized. This study used R to simulate a potential habitat distribution model for H. cunea from 2011 to 2030, and the ROC curve was used to check the accuracy of the model. The order of importance of environmental factors was determined using this model. The future habitat distribution of H. cunea in China was also predicted under two climate scenarios(RCP2.6 and RCP4.5) for 2041-2060(2050 s). 【Result】 ROC curve analysis indicated that the use of the random forest model to predict the potential habitat distribution of H. cunea achieves high precision;the AUC of training and testing data was 0.997 and 0.963, respectively. In the current period, the potential distribution(suitable areas) of H. cunea accounted for 8.74% of the total study area;the areas of low, medium, high and extremely high suitable accounted for 41.47%, 20.85%, 18.90% and 18.78%, respectively. The suitable areas were mainly co
作者
纪烨琳
苏喜友
于治军
JI Yelin;SU Xiyou;YU Zhijun(School of Information Science and Technology,Beijing Forestry University,Beijing 100083,China;General Station of Forest Pest Management,State Forestry and Grassland Administration,Shenyang 110034,China)
出处
《南京林业大学学报（自然科学版）》
CAS
CSCD
北大核心
2019年第6期121-128,共8页
Journal of Nanjing Forestry University：Natural Sciences Edition
基金
北京农业信息技术研究中心开放课题(KF2018W004)
关键词
美国白蛾
随机森林模型
潜在生境
气候情景
空间分布
Hyphantria cunea
random forest model
potential habitat
climate scenarios
spatial distribution
