Location: 9° 50’N on the East Pacific Rise, depth 2500m.

Functional traits were gathered for fifty-eight invertebrate species or higher taxa (when species ID was uncertain) collected from settlement surfaces deployed at deep-sea hydrothermal vents (https://doi.org/10.26008/1912/bco-dmo.733173.2). The identifier columns “scientificNameID” and “AphiaID” in this trait dataset are standardized to World Register of Marine Species (WoRMS), and are an exact subset of the of the same columns in the dataset of colonists from experimental settlement surfaces. Species names are in some cases generic and are not standardized to WoRMS.

Eight traits were chosen to test ecological hypotheses relevant to recovery and succession at deep-sea hydrothermal vents. We define “trait” as a feature or behavior of a species that affects or responds to its environment, and “modality” as a scoring level reflecting how the organism expresses a given trait. Four traits were taken from the sFDvent Database: maximum adult body size, habitat complexity, trophic mode (i.e., trophic level), and relative adult mobility. Four additional traits were added due to their ecological relevance and common use in functional analyses for aquatic invertebrates. These are external protection, feeding method (i.e., how a species feeds), reproductive type, and larval development (Bolam et al. 2016, Greenfield et al. 2016, Veríssimo et al. 2017). For each trait, a modality was assigned for every species based on literature or personal observation. For the traits from sFDvent, modalities were assigned from the recommended data set, except in cases where we suggested updates (Dykman et al., 2021). When species identity was uncertain and individuals were identified to a higher taxonomic level (e.g., amphipods), modalities were assigned from a likely species that is found at our site and included in sFDvent. For the four traits not included in sFDvent, we either provided a citation or cited “expert opinion,” meaning the modality choice was based on direct observation by one of the co-authors. Although the majority of specimens in our study were juveniles, most modality assignments were based on the characteristics of adult organisms due to the availability of data. Modality assignments were fixed for a given species rather than specific to life stages or individuals.

For the column “FUNCTIONAL_GUILD,” species were clustered into guilds based on the similarity of their modalities for all eight traits. The pairwise dissimilarity of species was calculated using the function gowdis in the R package FD (Laliberté et al. 2015). We chose Gower dissimilarity because it accepts both numerical and categorical data and handles missing values (Gower 1971). Podani’s extension was implemented to include ordinal variables (Podani 1999). Clusters were computed from the Gower dissimilarity matrix using the hclust function in the R package cluster (Maechler 2018) and plotted as a dendrogram. The cutoff for assigning functional guilds was chosen by optimizing the tradeoff between minimizing within-group distance and maximizing between-group distance. Code for this and all subsequent analyses are available at Zenodo (Dykman, 2021, DOI: 10.5281/zenodo.4625160).