1 INTRODUCTION

Standardized field surveys provide critical data on rare and endangered species, hotspots of species richness, and the spread of invasive species (Hallmann et al., 2017; Verheyen et al., 2017). Coupling contemporary records with historical archives provides an essential approach for addressing the effects of climate or land-use change on the distribution and abundance of species, and for helping to identify the ‘winners’ and ‘losers’ in changing environments (Alfonsi et al., 2017; Hedl et al., 2017; Kelemen & Rehan, 2021; Moritz et al., 2008; Socolar et al., 2017; Tingley & Beissinger, 2009).

However, relevant field data are not always available because field surveys are labour-intensive, technically demanding, and logistically challenging (Lawton et al., 1998). Moreover, field survey data are patchy even for well-studied regions and taxa (Dornelas et al., 2018). This void is partly filled by dated, geo-referenced specimens in museum collections around the world. Data from museum collections have been used successfully to examine changes in species ranges (Farnsworth & Ogurcak, 2006; Loiselle et al., 2008; Pardi et al., 2020), declines (Case et al., 2007; Habel et al., 2019; Mathiasson & Rehan, 2019; Rowe, 2007; Shaffer et al., 1998) and possible extinctions (Gotelli et al., 2012; Jacobson et al., 2018) of endangered species, expanding distributions of invasive species (Bradley et al., 2015), phenological shifts due to climate change (Burkle et al., 2013; Miller-Rushing et al., 2006), and changes in body size and condition of animals over time (Johnson et al., 2003). At the community level, museum collections have been used to estimate regional species richness (Rahbek & Graves, 2001), the frequency of species associations (Lyons et al., 2016), catastrophic species losses in the fossil record (Raup & Sepkoski, 1982), and the evolutionary diversification and spread of novel traits (Holmes et al., 2016).

Quantitative analysis of museum specimen records, however, poses its own set of challenges. There is uneven geographical and temporal coverage of global biodiversity (Daru et al., 2018; Meyer et al., 2016), as well as taxonomic collecting biases (Prather et al., 2004), such as a preference for species that are easy to collect, process, identify and store. Perhaps the most general source of bias is the tendency for species that are rarely encountered in the field to be sought after and therefore over-represented in museum collections relative to their true abundances. This is a specific manifestation of the general ‘rarity-seeking syndrome’ prevalent in taxonomy and systematics (Kruckeberg & Rabinowitz, 1985).

The degree to which museum collections reflect the natural abundance of species in the wild likely varies by taxon along a spectrum from random standardized sampling to highly selective cherry-picking of prized species. But where do different taxa fall along this spectrum? If there is a strong correlation between the number of museum specimens of a species or higher taxon and its abundance in the field, then museum records may justifiably serve as proxy variables to estimate their relative abundance in nature. To our knowledge, this relationship has not been empirically tested.

We assembled 17 coupled field and museum datasets representing a diverse array of plant and animal assemblages. Datasets range from 100-year-old botanical records from the area around Concord's Walden Pond to recent citizen science surveys of butterflies from North Carolina and museum studies of springtails from Germany. These analyses verify for the first time a strong, general relationship between the abundance of species in field surveys and the number of museum records in all of the test cases.

However, this relationship, by itself, is not useful for quantitative analysis because the units—number of records—are not meaningful for comparisons within or between studies. Instead, the raw counts of field or museum records need to be converted to measures of relative abundance, which can be meaningfully compared. To address this issue, we employed a novel application of the Dirichlet distribution to estimate the relative abundance of each species in both the historical and contemporary collections. The Dirichlet distribution is often used as a prior distribution in Bayesian analysis, and is appropriate for multinomial data, such as integer counts of individuals classified into species or other taxonomic groups.

When comparable field and museum data are available, the Dirichlet distribution can be used to construct a realistic calibration curve (Figure 1) so that estimates of relative abundance from museum records can be converted to approximate estimates that would have been obtained from field samples. We call this procedure of estimation and validation by the acronym FAMA (field abundance–museum abundance).

中文翻译：

---

从博物馆记录中估计物种相对丰度

1 简介

标准化实地调查提供了有关稀有和濒危物种、物种丰富度热点和入侵物种扩散的关键数据（Hallmann 等人，2017 年；Verheyen 等人，2017 年）。将当代记录与历史档案相结合，为解决气候或土地利用变化对物种分布和丰度的影响，以及帮助确定不断变化的环境中的“赢家”和“输家”提供了一种重要方法（Alfonsi 等人，2017 年）。 ，2017 年；Hedl 等人，2017 年；Kelemen 和 Rehan，2021 年；Moritz 等人，2008 年；Socolar 等人，2017 年；Tingley 和 Beissinger，  2009 年）。

然而，相关的实地数据并不总是可用，因为实地调查是劳动密集型、技术要求高且逻辑上具有挑战性的（Lawton 等人，1998）。此外，即使对于研究充分的地区和分类群，实地调查数据也不完整（Dornelas 等人，2018 年）。这个空白部分被世界各地博物馆收藏的标本标注了日期和地理参考。来自博物馆藏品的数据已成功用于检查物种范围的变化（Farnsworth 和 Ogurcak，  2006 年；Loiselle 等人，  2008 年；Pardi 等人，  2020 年），减少（Case 等人，2007 年；Habel 等人，  2019 年；马蒂亚森和雷汉，  2019 年；罗， 2007年；Shaffer et al., 1998 ) 和可能的灭绝 (Gotelli et al., 2012 ; Jacobson et al.,  2018 ) 濒危物种、入侵物种分布扩大 (Bradley et al., 2015 )、气候变化导致的物候变化 ( Burkle 等人，  2013 年；Miller-Rushing 等人，2006 年），以及动物体型和状况随时间的变化（Johnson 等人，2003 年）。在社区层面，博物馆藏品已被用于估计区域物种丰富度（Rahbek 和 Graves，  2001 年）、物种关联的频率（Lyons 等人，2016 年）)、化石记录中的灾难性物种损失 (Raup & Sepkoski,  1982 )，以及新特征的进化多样化和传播 (Holmes et al., 2016 )。

然而，博物馆标本记录的定量分析也带来了一系列挑战。全球生物多样性的地理和时间覆盖面不均衡（Daru 等人，2018 年；Meyer 等人，2016 年），以及分类收集偏差（Prather 等人，2004 年），例如偏爱容易获得的物种收集、处理、识别和存储。也许最普遍的偏见来源是在野外很少遇到的物种往往会受到追捧，因此在博物馆收藏中相对于它们的真实丰度而言数量过多。这是分类学和系统学中普遍存在的一般“稀有性寻求综合症”的具体表现（Kruckeberg & Rabinowitz，  1985）。

博物馆藏品反映野外物种自然丰度的程度可能因分类单元而异，从随机标准化抽样到珍贵物种的高度选择性樱桃采摘。但是不同的类群属于这个范围内的什么地方？如果一个物种或更高分类单元的博物馆标本数量与其在野外的丰度之间存在很强的相关性，那么博物馆记录就可以合理地作为代理变量来估计它们在自然界中的相对丰度。据我们所知，这种关系尚未经过实证检验。

我们组装了 17 个耦合的野外和博物馆数据集，代表了各种各样的植物和动物组合。数据集的范围从康科德瓦尔登湖周围地区 100 年历史的植物学记录到最近对北卡罗来纳州蝴蝶的公民科学调查和对德国跳虫的博物馆研究。这些分析首次证实了实地调查中的物种丰度与所有测试案例中的博物馆记录数量之间存在强有力的一般关系。

然而，这种关系本身对定量分析没有用，因为单位（记录数）对于研究内部或研究之间的比较没有意义。相反，田野或博物馆记录的原始计数需要转换为相对丰度的测量值，以便进行有意义的比较。为了解决这个问题，我们采用狄利克雷分布的新应用来估计历史和当代收藏品中每个物种的相对丰度。Dirichlet 分布通常用作贝叶斯分析中的先验分布，适用于多项式数据，例如分类为物种或其他分类组的个体的整数计数。

当可比较的野外和博物馆数据可用时，狄利克雷分布可用于构建现实的校准曲线（图 1），以便将博物馆记录中的相对丰度估计值转换为从野外样本中获得的近似估计值。我们用首字母缩略词 FAMA（野外丰度-博物馆丰度）来称呼这个估计和验证过程。