生物多样性 ›› 2007, Vol. 15 ›› Issue (4): 365-372.  DOI: 10.1360/biodiv.060280

• 论文 • 上一篇    下一篇

ROC曲线分析在评价入侵物种分布模型中的应用

王运生1,2,谢丙炎1*,万方浩3,肖启明2,戴良英2   

  1. 1 (中国农业科学院蔬菜花卉研究所, 北京 100081)
    2 (湖南农业大学生物安全科学技术学院, 长沙 410128)
    3 (中国农业科学院植物保护研究所, 北京 100081)
  • 收稿日期:2006-11-20 修回日期:2007-07-03 出版日期:2007-07-20 发布日期:2007-07-20

Application of ROC curve analysis in evaluating the performance of alien species’ potential distribution models

Yunsheng Wang1,2, Bingyan Xie1*, Fanghao Wan3, Qiming Xiao2,, Liangying Dai2   

  1. 1 Institute of Vegetables and Flowers, Chinese Academy of Agricultural Sciences, Beijing 100081
    2 College of Bio-safety Science and Technology, Hunan Agricultural University, Changsha 410128
    3 Institute of Plant Protection (South Section), Chinese Academy of Agricultural Sciences, Beijing 100081
  • Received:2006-11-20 Revised:2007-07-03 Online:2007-07-20 Published:2007-07-20

摘要: 生态位模型(ecological niche models, ENMs)已广泛应用于物种潜在分布区预测, ENMs的应用也为外来入侵物种的风险分析提供了重要的定量化分析工具, 但如何评价不同模型之间的预测效果成了当今研究的热点问题。本文介绍了受试者工作特征(ROC)曲线分析在评价不同生态位模型预测效果中的应用原理和分析方法, 并以一种植物病原线虫—相似穿孔线虫(Radopholus similis)为例, 应用ROC曲线分析法对其5种模型(BIOCLIM, CLIMEX, DOMAIN, GARP, MAXENT)的预测结果进行了比较分析。5种模型的ROC曲线下面积AUC(Area Under Curve)值分别为0.810, 0.758, 0.921, 0.903和0.950, 以MAXENT模型的AUC值最大, 表明其预测效果最好; 方差分析结果表明, 除GARP与DOMAIN模型之间AUC值差异不显著外, 其余各模型之间差异显著。

AbstractEcological niche models (ENMs), which are widely employed to predict the potential geographic distribution of species, provide an important tool to quantify the risks imposed by invasive alien species. The problem of how to evaluate the performance of different models has attracted more and more attention. In the present paper, we introduced the principle of the method of Receiver Operating Characteristic (ROC) curve analysis in assessing the accuracy of different ENMs. We predicted the suitable distribution area of Radopholus similis, an important banana toppling disease nematode, with five widely used ENMs and evaluated the performance of different models by ROC curve analysis. The area under ROC curve (AUC) for BIOCLIM, CLIMEX, DOMAIN, GARP, and MAXENT models was 0.810, 0.758, 0.921, 0.903, and 0.950, respectively. Among these, the biggest value of AUC was assigned to MAXENT, indicating that the result gained by MAXENT should be better than the other four models. According to the results of analysis of variance (ANOVA), there was a remarkable difference in AUC between each model except for DOMAIN and GARP.