Paini Dean R, Mokany Karel
Ecol Appl. 2016 Mar;26(2):415-23. doi: 10.1890/15-0798.
Estimating extinction vulnerability for a large number of species presents significant challenges for researchers, but is of high importance considering the large number of species that are currently unassessed. We present a method using a type of artificial neural network (self organizing map; SOM), which utilizes the co-occurrence patterns of species to estimate each species' vulnerability to extinction. We use this method on Australian bird assemblages and compare the SOM-generated rankings for vulnerability with assessments from the IUCN Red List for those species in which populations have actually been estimated. For species that have had their populations estimated, the SOM performed well in distinguishing those species ranked of least concern by IUCN from those species in one of the other IUCN categories. Further, 19 species that were identified as highly vulnerable by the SOM analysis have never had their populations estimated and have been ranked by the IUCN of least concern. We show how the SOM can identify spatial variation in vulnerability for a species, and identify those regions in Australia in which the resident species have the greatest levels of vulnerability (central Australia). We conclude that the SOM provides a useful tool for researchers and agencies dealing with conservation strategies focused on large numbers of species and we urge a more detailed assessment of the 19 bird species identified by this analysis as vulnerable to extinction.
对大量物种的灭绝脆弱性进行评估给研究人员带来了巨大挑战，但鉴于目前大量未被评估的物种，这一工作具有高度重要性。我们提出了一种使用人工神经网络（自组织映射；SOM）的方法，该方法利用物种的共现模式来估计每个物种的灭绝脆弱性。我们将此方法应用于澳大利亚鸟类群落，并将SOM生成的脆弱性排名与国际自然保护联盟（IUCN）红色名录中对实际已评估种群的那些物种的评估进行比较。对于已评估种群的物种，SOM在区分IUCN列为最不关注的物种与IUCN其他类别中的物种方面表现良好。此外，通过SOM分析被确定为高度脆弱的19个物种从未进行过种群评估，且被IUCN列为最不关注物种。我们展示了SOM如何识别一个物种脆弱性的空间变化，并确定澳大利亚境内常住物种脆弱性最高的那些区域（澳大利亚中部）。我们得出结论，SOM为研究人员和致力于针对大量物种的保护策略的机构提供了一个有用的工具，并且我们敦促对通过该分析确定为易灭绝的19种鸟类进行更详细的评估。
