小岛屿效应(small-island effect, SIE)描述了种-面积关系(species-area relationship, SAR)的一种特殊现象: 在面积低于某个阈值时, 物种数不随岛屿面积的增加而增加或以一种比大岛屿低的速率增加的现象。由于小岛屿效应的面积阈值可能是多种生物地理格局和生态学过程在空间尺度上的转折点, 另外该现象在生物多样性保护领域具有重要指导意义, 因此小岛屿效应已经成为岛屿生物地理学和生物多样性研究领域的一种重要格局。目前主要有5种分析和检验小岛屿效应的方法, 包括种-面积关系形状比较法(SAR shape comparison)、断点回归法(breakpoint/piecewise regression)、零模型法(null model)、路径分析法(path analysis)和树模型法(tree-based model)。本文首先简要介绍了小岛屿效应与种-面积关系的关联, 然后重点总结了文献中记载的小岛屿效应检测的5种方法的优点和不足。在SAR形状比较法中, 受大岛屿离群值效应的影响, SAR的形状往往很难呈现出“S”形曲线。在断点回归法中, 数据的对数转换使检测到的SIE可能只是一种假象。在零模型法中, 随机化过程忽略了物种之间生态特征的差异, 降低了岛屿物种丰富度预期值的可信度。在路径分析法中, 生境多样性难以量化以及SIE范围内如果SAR具有斜率则会降低该方法的适用性。在树模型法中, 如果面积不是物种丰富度变异的最佳预测因子, 树模型不会首先选择面积对样本进行拆分; 另外如果SAR存在两个面积阈值, 树模型可能不会从SIE阈值处进行拆分。因此, 为了避免因某种方法的自身不足而造成的错误判断, 我们建议: 首先应同时使用多种方法来进行SIE检测, 当至少有两种方法同时检测出SIE时, 方可认为系统中存在SIE; 其次, 针对现有检测方法中存在的缺陷加以改进也是今后重要的发展方向。最后, 本文针对国内学者从事较多的生境岛屿中SIE的特征以及导致全球变化的人类活动如何影响SIE的出现等问题给出了一些启发性建议。本文将为小岛屿效应的准确检测提供参考依据并为完善小岛屿效应的理论框架起到推动作用。

Background & Aim: The small-island effect (SIE) describes a phenomenon when below a certain threshold area, species richness varies independently of island size or increases at a lesser rate than for larger islands. The SIE has become a fundamental framework in biodiversity science and island biogeography because the area threshold of an SIE may be the turning point for many biogeographic patterns and ecological processes across spatial scales, and such phenomenon is significant in the research field of biodiversity conservation. In this paper, we first briefly introduce the relationship between the SIE and species-area relationships (SARs). Secondly, we summarize the five detection methods of SIEs found in the literature. Finally, we discuss the advantages and shortcomings of each detection method. We provide feasible suggestions for the accurate detection of an SIE to improve the theoretical framework in this field of research.

Progress: Currently, five SIE detection methods have been published, including SAR shape comparison, breakpoint/piecewise regression, null model, path analysis, and a tree-based model. The following shortcomings for each method are presented. In the SAR shape comparison method, large islands may be outliers, causing the shape of the SAR curve to be poorly represented by a sigmoidal curve. In the piecewise regression method, the logarithmic data transformations may make the SIE just a statistical artifact. In the null model method, the randomization process ignores the differences in ecological characteristics between species, causing the island species richness expected values to have a reduced credibility. In the path analysis method, the habitat diversity is difficult to quantify and the SAR slope within the limits of the SIE may lower this method’s applicability. In the method of tree-based model, it is questionable whether the model split first with the independent variable of area. Moreover, an SAR may actually have two area thresholds which brings into question which threshold is best for model splitting.

Perspective: To avoid the inadequacies of a specific method, we propose using multiple methods for SIE detection. An SIE in the system can only be considered when at least two methods simultaneously detect the SIE. In addition, improvements can be made to address the defects in the existing methods. For example, for the existing null model method, some restrictive conditions can be added to its randomization processes to make the expected results more reasonable.