Deep-sea biodiversity at the extremes of the Salas y Gómez and Nazca ridges with implications for conservation

where pi is the proportion of all individuals counted that were of species i. Pielou’s evenness was calculated as: J = H´/ln(S), where S is the total number of species present. A principal components analysis was used to examine environmental variables between subregions. Environmental variables in this analysis included depth, temperature, salinity, dis- solved oxygen, %O2 saturation, nitrate, silicate, phosphate, chlorophyll-a, broad-scale BPI, fine-scale BPI, POC, crust age, nearest seamount, and number of seamounts within 30 km of the sample sites. Data were Box-Cox transformed, centered, and standardized prior to analysis. The absolute values of fine-scale and broad-scale BPI metrics were used since peaks and troughs both represent measures of habitat complexity. A comparison of environmental vari- ables between subregions and habitats was conducted using permutation-based multivariate analysis of variance (PERMANOVA) with a Euclidean distance matrix. Deployment depth was examined as a predictor variable in explaining the variability in invertebrate taxa richness, fish taxa richness, and fish MaxN between subregions and habitats (i.e., hard and soft) using generalized linear models (GLMs) with a Poisson distribution with a log-link function. The GLMs for fish diversity and evenness used a Gamma distribution with a log-link function. Comparison of Amphipoda abundance between subregions was tested using a Wilcoxon rank sum test. Comparisons of invertebrate and fish assemblage structure between

subregions and habitats was also performed using PERMANOVA with Bray-Curtis similarity matrices. Similarity percentages analysis (SIMPER) was used to examine differences in inverte- brate and fish assemblages between subregions using Bray-Curtis similarity analysis of hierar- chical agglomerative group average clustering. Relationships between environmental variables and invertebrate and fish assemblages were analyzed using distance-based linear models (DistLM) and redundancy analysis (dbRDA). Linear models are appropriate for these data because a preliminary detrended correspondence analysis showed short gradient lengths (<2 SD) [84]. The dbRDA introduces a series of explanatory (environmental) variables and resembles the model of multivariate multiple regression, allowing us to determine what linear combinations of these environmental vari- ables determine the gradients. These analyses were conducted using Bray-Curtis similarity matrices on MaxN Class or Order-level invertebrate data and fish family-level data. MaxN invertebrate data were ln(x+1) transformed and MaxN fish data were 4th-root transformed prior to analyses to conform to the assumptions of normality and homogeneity of variances [85]. Normality was tested using a Shapiro-Wilk W test (p < 0.05) while a Bartlett’s test (p < 0.05) was used to examine homogeneity of variance. We determined the model with envi- ronmental drivers that best explained variation in invertebrate and fish assemblages using a stepwise routine that employed 9999 permutations based on Akaike’s Information Criterion

with second order correction (AICc) selection criterion. Environmental variables that had high correlations (r > 0.95) were removed prior to analysis. Those removed were temperature, silicate, %O2 saturation, and phosphate. All statistical analyses were conducted using JMP Pro 15 and Primer 6.0. Results Environmental variables Of the 20 camera deployments conducted around RN/SyG, 40% (N = 8) were located on hard bottom habitat. In contrast, only 15% (n = 4) of the 27 deployments at Desventuradas were located on this habitat type. Deployment depth was not significantly different between subre- gions, habitats, or their interaction (all GLMs p > 0.05). Subregions were well separated in ordination space based on environmental variables (Fig 4). Environmental variables were significantly different between subregions (PERMANOVA PLOS ONE Deep-sea biodiversity along the Salas y Go´ mez and Nazca ridges PLOS ONE | https://doi.org/10.1371/journal.pone.0253213 June 30, 2021 9 / 27