Marine and Environmental Biology Section at the Department of Biological Sciences, University of Southern California, Los Angeles, California, USA.
Department of Zoology & Biodiversity Research Centre, University of British Columbia, Vancouver, British Columbia, Canada.
Ecology. 2022 Jun;103(6):e3683. doi: 10.1002/ecy.3683. Epub 2022 May 3.
In metacommunity ecology, a major focus has been on combining observational and analytical approaches to identify the role of critical assembly processes, such as dispersal limitation and environmental filtering, but this work has largely ignored temporal community dynamics. Here, we develop a "virtual ecologist" approach to evaluate assembly processes by simulating metacommunities varying in three main processes: density-independent responses to abiotic conditions, density-dependent biotic interactions, and dispersal. We then calculate a number of commonly used summary statistics of community structure in space and time and use random forests to evaluate their utility for inferring the strength of these three processes. We find that (i) both spatial and temporal data are necessary to disentangle metacommunity processes based on the summary statistics we test, and including statistics that are measured through time increases the explanatory power of random forests by up to 59% compared to cases where only spatial variation is considered; (ii) the three studied processes can be distinguished with different descriptors; and (iii) each summary statistic is differently sensitive to temporal and spatial sampling effort. Including repeated observations of metacommunities over time was essential for inferring the metacommunity processes, particularly dispersal. Some of the most useful statistics include the coefficient of variation of species abundances through time and metrics that incorporate variation in the relative abundances (evenness) of species. We conclude that a combination of methods and summary statistics is probably necessary to understand the processes that underlie metacommunity assembly through space and time, but we recognize that these results will be modified when other processes or summary statistics are used.
在集合生态学中，一个主要的焦点是结合观察和分析方法来确定关键集合过程的作用，例如扩散限制和环境过滤，但这项工作在很大程度上忽略了时间社区动态。在这里，我们开发了一种“虚拟生态学家”方法，通过模拟三种主要过程（对非生物条件的密度独立响应、密度依赖的生物相互作用和扩散）的集合来评估集合过程。然后，我们计算了一些常用的空间和时间社区结构的综合统计数据，并使用随机森林来评估它们用于推断这三个过程强度的效用。我们发现：（i）基于我们测试的综合统计数据，必须同时使用空间和时间数据来解开集合过程，并且通过时间测量的统计数据比仅考虑空间变化的情况增加了随机森林的解释能力最多可达 59%；（ii）这三个研究过程可以通过不同的描述符来区分；（iii）每个综合统计数据对时间和空间采样努力的敏感性不同。随着时间的推移对集合群落进行重复观测对于推断集合过程（特别是扩散）至关重要。一些最有用的统计数据包括物种丰度随时间的变异系数，以及包含物种相对丰度（均匀性）变化的度量标准。我们得出结论，可能需要结合方法和综合统计数据来理解通过空间和时间进行集合组装的过程，但我们认识到，当使用其他过程或综合统计数据时，这些结果将发生变化。
