欢迎访问林业科学,今天是

林业科学 ›› 2016, Vol. 52 ›› Issue (6): 66-75.doi: 10.11707/j.1001-7488.20160608

• 论文与研究报告 • 上一篇    下一篇

应用最大熵模型模拟预测大尺度范围油松毛虫灾害

宋雄刚1, 王鸿斌1, 张真1, 孔祥波1, 苗振旺2, 刘随存3, 李永福4   

  1. 1. 中国林业科学研究院森林生态环境与保护研究所 国家林业局森林保护学重点实验室 北京 100091;
    2. 山西省森林病虫害防治检疫站 太原 030012;
    3. 山西省林业科学研究院 太原 030012;
    4. 山西省大同市灵丘县森林病虫害防治检疫站 灵丘 034400
  • 收稿日期:2014-07-30 修回日期:2015-01-27 出版日期:2016-06-25 发布日期:2016-07-04
  • 基金资助:
    科技支撑计划"适应气候变化的病虫害风险预警技术研发与应用"(2013BAC09B0202)。

Application of the Maximum Entropy Model (MaxEnt) to Simulation and Forecast of Large Scale Outbreaks of Dendrolimus tabulaeformis (Lepidoptera: Lasiocampidae)

Song Xionggang1, Wang Hongbin1, Zhang Zhen1, Kong Xiangbo1, Miao Zhenwang2, Liu Suicun3, Li Yongfu4   

  1. 1. Key Laboratory of Forest Protection of State Forest Administration Research Institute of Forest Ecology, Environment and Protection, CAF Beijing 100091;
    2. Forest Pest Control Station of Shanxi Province Taiyuan 030012;
    3. Shanxi Academy of Forestry Taiyuan 030012;
    4. Lingqiu Forest Pest Control Station of Shanxi Province Lingqiu 034400
  • Received:2014-07-30 Revised:2015-01-27 Online:2016-06-25 Published:2016-07-04
  • Contact: 王鸿斌

摘要: [目的] 探讨利用最大熵模型MaxEnt,基于油松毛虫暴发的历史灾情数据和相应的气象数据,对未来大尺度范围油松毛虫暴发区进行模拟和预测的可行性。[方法] 以山西省2002-2011年的油松毛虫灾情数据和山西省2002-2011年的地面气象数据为基础,结合油松毛虫完成生活史不同发育阶段对不同气候因子的响应衍生出与油松毛虫灾害发生潜在相关的物候因子80个,运用主成分分析和逐步回归法从中筛选出与油松毛虫灾害发生相关性最高的前8个物候因子,即X29(10月份日均温<5℃的天数)、X43(7月平均湿度>75%的天数)、X54(3月平均风速)、X55(4,5,6月分平均风速)、X56(7,8月均风速)、X62(10月日均风速>10 m·s-1的天数)、X63(9月最大日均风速)、X67(4,5,6月总降雨量)。[结果] 利用筛选出的8个物候因子,运用最大熵模型MaxEnt代入实际灾害发生数据进行训练模拟,并应用刀切法分析确立气象因子X43,X54X55为灾害模型应用的最佳因子,其接受者操作特性曲线(ROC)检验结果即曲线下面积(AUC)值为0.820,标准差为0.019;最终利用该模型参数,对气候变化背景不同外排模式下未来油松毛虫灾害趋势进行预测,2050年2种外排模式(RCP4.5与RCP6.0)下灾害发生趋势变化不同,其中RCP4.5模式灾害发生将集中于北京、河北及河北与内蒙古交界处, RCP6.0模式在山西中南部灾害将会加强。[结论] MaxEnt模型对未来气候变化条件下油松毛虫害虫暴发区的准确模拟与预测具有潜在应用价值。

关键词: 油松毛虫, MaxEnt, 物候因子, 气候变化, 灾害

Abstract: [Obiective] The Chinese pine caterpillar, Dendrolimus tabulaeformis, is a serious native pine defoliator with frequent outbreaks in northern China. The MaxEnt model is one of the most effective software packages available for modeling species' distributions. The main objective of the current study was to test and determine the possibility of using MaxEnt to simulate and predict future large-scale outbreaks of D. tabulaeformis based on county-level historical outbreak records (2002-2011), and daily meteorological data from 19 weather stations in Shanxi province. [Method] Using Principal Component Analysis and Step-wise Regression methods with actual pest outbreak data, the 8 most relevant factors were chosen from 80 outbreak-related bio-climate factors potentially affecting development of the insect. The key factors were X29 (days with mean temperature<5℃ in October), X43 (days with humidity >75% in July), X54 (mean monthly wind speed in March), X55 (mean monthly wind speed in April, May and June), X56 (mean monthly wind speed in July and August), X62 (days with wind speed >10 m·s-1 in October), X63 (maximum daily wind speed in September), X67 (precipitations in April, May and June). [Result] With the 8 screened phenological factors, the MaxEnt model was used to make the training simulation with the actual disaster data. The Jackknife test showed that X43, X54 and X55 were the three principle climatic factors that best simulated historical outbreaks using the MaxEnt model, and ROC (recevier operating characteristic curve) test showed an AUC (area uner the ROC curve) value of 0.82 with a STD(standard deviation) of 0.019. Based on data from the WorldClim database for future climate scenarios, pine caterpillar outbreak distribution maps for 2050 were generated via the MaxEnt model under RCP(representative concentration pathway)4.5 and RCP6.0. According to these maps, in the 2050s, Beijing and Hebei province, plus the southern border area of Inner Mongolia Autonomous Region with Hebei, will have a high risk of outbreaks under RCP4.5, while more serious outbreak area will be the south central region of Shanxi province under RCP6.0. [Conclusion] MaxEnt model is potentially useful for forecasting future pine caterpillar outbreaks under climate change.

Key words: Dendrolimus tabulaeformis, MaxEnt, bio-climatic variables, climate change, outbreak

中图分类号: