Dissimilarity of species interaction networks: how to partition rewiring and species turnover components
Dissimilarity of species interaction networks: how to partition rewiring and species turnover components
复制标题
DOI：
10.1002/ecs2.3653
复制
发表时间：
2021-07-01
期刊：
影响因子：
2.7
通讯作者：
Fruend, Jochen
中科院分区：
文献类型：
--
作者：
Fruend, Jochen
描述物种相互作用网络在时空上的变化，有助于更好地理解物种群落如何应对全球变化。为了理解这种差异，有人建议将网络差异划分为物种更替驱动的一个组件，即群落组成的变化，以及反映重新布线的另一个组件，即共享物种之间相互作用的灵活性。后者为投入大量精力记录互动提供了强有力的理由，而不是简单地基于社区数据建立网络。在这里，我提出了一个灵活的R函数（可在R包bipartite）来计算网络不相似性及其组成部分，包括二进制和定量网络。使用这个新工具，我比较了两种已发表的划分网络差异性的方法，使用概念示例、已发表的植物传粉者网络和一组模拟。这一比较突出表明，最受关注的方法高估了重新布线对整个网络差异性的重要性。相比之下，先前提出的方法是从相互作用集的加性划分中得出的，因此可以准确地表示两个不相似的成分。此外，我认为“重新布线”一词在网络生态学中没有得到很好的定义，这两种方法都高估了重新布线的重要性，尤其是在定量网络中。计算网络差异性多个方面的统一功能的可用性将促进其在表征网络动力学和识别潜在驱动因素方面的关键应用。对网络差异性和重新布线的研究在方法的选择和解释上需要更加谨慎。
Describing variation of species interaction networks across space and time promises a better understanding of how species communities respond to global change. To understand this variation, it has been suggested to partition network dissimilarity into one component driven by species turnover, that is, changes in community composition, and another component reflecting rewiring, that is, flexibility of interactions among shared species. The latter makes a strong case for investing the enormous effort in empirically recording interactions, instead of simply building networks based on community data. Here, I present a flexible R function (available in the R package bipartite) to calculate network dissimilarity and its components, with binary and quantitative networks. With this new tool, I compare two published methods for partitioning network dissimilarity, using conceptual examples, published plant-pollinator networks, and a set of simulations. This comparison highlights that the method that has received most attention overestimates the importance of rewiring for total network dissimilarity. In contrast, an earlier-proposed method is derived from additive partitioning of the sets of interactions and thus accurately represents the two dissimilarity components. Furthermore, I argue that the term rewiring is not well defined in network ecology and that there are reasons why both methods overestimate the importance of rewiring, in particular with quantitative networks. The availability of a unified function to calculate multiple aspects of network dissimilarity will foster its critical application to characterize network dynamics and to identify underlying drivers. Studies on network dissimilarity and rewiring will have to be more careful in the choice of method and its interpretation.