四川农业大学学报 ›› 2019, Vol. 37 ›› Issue (04): 481-489.doi: 10.16036/j.issn.1000-2650.2019.04.008

• • 上一篇    下一篇

基于组合物种分布模型(Ensemble Model)的厚朴适宜生境分布模拟

朱妮   

  1. 1.安康学院 旅游与资源环境学院,陕西 安康 725000;
    2.秦巴国土资源利用与环境保护协同创新中心,陕西 安康 725000;
    3.安康市汉江水资源保护与利用工程技术研究中心,陕西 安康 725000
  • 收稿日期:2018-10-19 出版日期:2019-08-28 发布日期:2021-01-31
  • 作者简介:朱妮,硕士,讲师,主要从事空间数据建模分析与物种分布研究,E-mail:zhuni822611@163.com。
  • 基金资助:
    陕西省科技厅项目(2017NY-202); 陕西省教育厅项目(18JK0016); 安康学院校级大学生创新创业训练计划项目(2018akxy015)

Modelling the Suitable Habitat Distribution of Magnolia Officinalis Using Ensemble Model

ZHU Ni   

  1. 1. School of Tourism and Environment,Ankang University,Ankang 725000,Shaanxi,China;
    2. Collaborative Innovation Center of Qinba Land and Resources Utilization and Environment Protection,Ankang 725000,Shaanxi,China;
    3. Engineering Technology Research Center for Water Resource Protection and Utilization of Hanjiang River,Ankang 725000,Shaanxi,China
  • Received:2018-10-19 Online:2019-08-28 Published:2021-01-31

摘要: 【目的】建立组合物种分布模型,模拟厚朴在我国南方的分布格局,探讨影响其生长环境因子的阈值,为厚朴的人工种植,野生资源保护等提供理论依据。【方法】基于公开发表的文献中精确的厚朴分布数据与23个环境因子变量,利用BIOMOD2建模平台中的9个模型算法构建组合物种分布模型模拟厚朴的适宜生境分布。【结果】①厚朴在我国南方适宜生境面积为0.53×106 km2,主要分布在四川东部、陕西南部、重庆东部、湖北西部、贵州、福建、江西、湖南等地区。②在组成组合物种分布模型的9个单模型算法中,推进式回归树模型(GBM)和随机森林(RF)的模拟效果最好,表面分室模型(SRE)与分类树分析模型(CTA)结果较差。组合物种分布模型的TSS为0.905,ROC值为0.975,说明组合模型能够在一定程度上提高模型的精度。【结论】厚朴在我国南方分布较为广泛,但是考虑植被类型限制时,其生境较为破碎,应该在其适宜生境区域加强物种保护。

关键词: 厚朴, 物种分布模型, BIOMOD2, 组合物种分布模型

Abstract: 【Objective】 The aim of this study was to build an ensemble model (EM) to determine the potential geographic distribution of Magnolia officinalis in southern China, and to identify the thresholds of environmental factors affecting its distribution. In addition, this study also can provide a theoretical basis for the artificial cultivation and wild resource protection of M. officinalis. 【Method】 The species presence data in this study were obtained from the published literature and field survey reports. The potential geographic distribution of M.officinalis in southern China was simulated with EM which was formed by nine different algorithms in the statistical package BIOMOD2, based on species presence data and 23 climatic variables. 【Result】 The area of M. officinalis suitable habitat was about 0.53×106 km2, mainly distributing in eastern Sichuan, southern Shaanxi, eastern Chongqing, western Hubei, Guizhou, Fujian, Jiangxi, Hunan and other regions. Among 9 single model algorithms formed Ensemble Model, the generalized boosting models (GBM) and random forest (RF) had high precision and surface range envelope (SRE) and the classification tree analysis model (CTA) had low accuracy. The true skill statistic (TSS) value of the EM was 0.905 and the ROC value was 0.975, indicating that the EM may, to a certain extent, improve the accuracy of the model. 【Conclusions】 M. officinalis is widely distributed in southern China. its habitat is relatively fragmented when the limitation of vegetation types is considered. Therefore, species protection should be strengthened in its suitable habitat areas.

Key words: Magnolia officinalis, species distribution model, BIOMOD2, ensemble model

中图分类号: 

  • Q948