稀疏标准化是定量古生物工作中矫正多样性统计偏差的常用方法。相比基于样本大小的传统稀疏化, 基于采样充分度的改进能更忠实反映多样性信息。然而一些案例对于稀疏化的适用性不够重视, 尤其是改进的方法鲜有国内文献介绍。本文阐述了稀疏化的原理, 强调了应用的注意事项和改进方法的优势。稀疏化的原理是从大小不同的样本中二次抽样出彼此“公平”的子样本, 以比较其分类单元丰富度。传统方法据样本大小衡量公平, 改进的方法据采样充分度评估公平, 要求子样本在群落中代表的个体频率总和相等。两种思路均可通过计算机模拟多次重复二次抽样或公式推导来计算, 已有PAST和iNext等软件可以实现。采样是否充分代表了古生物群落是有效应用该方法的首要前提。

Taxonomic diversity of paleocommunities is a key metric for tracing the evolution of life and underlying geological events. However, the taxonomic richness of fossil collections or compiled data is easily biased by differences in sampling size. Rarefaction is a routine statistical method to mitigate such biases by reducing larger collections to a consistent sample size with the smaller ones. Traditional individual-based rarefaction has been increasingly superseded in the literature by coverage-based rarefaction (or SQS, shareholder quorum subsampling as named by some paleontologists). However, some case studies still show certain misunderstanding of this longstanding method, and coverage-based rarefaction has rarely been clarified in the Chinese literature. In order to better apply this method, this paper introduces the principle, details of calculation and suggestions for application of the rarefaction techniques. The core idea of rarefaction is to randomly resample from the original samples until the subsamples reach a consistent sample level, then the mathematical expectation of the taxonomic richness of these subsamples is calculated for comparison. Traditional rarefaction method evaluates such consistency by the same sample size, such as the number of specimens or fossil occurrences in literature. One major drawback of this traditional method is that the information of larger samples is often severely compressed. To address this problem, an updated method, i.e., coverage-based rarefaction, requires resampling until the equal sample coverage is achieved. The degree of coverage is measured by the sum of the individual frequencies in the community covered by the taxa in the subsamples. It has been well demonstrated that the updated method could more faithfully reflect the true ratio of taxonomic richness among communities. Both the traditional and updated rarefaction methods can be implemented by algorithmic simulation or analytical derivation, and software such as PAST or iNext is convenient for implementation. The primary requirement for applying rarefaction is that the samples at hand are as representative of the paleocommunity as possible. We also suggest several potential directions to further develop the rarefaction techniques in the field of quantitative paleontology.

国家自然科学基金面上项目(41872036)资助