Harnessing a mesopelagic predator as a biological sampler reveals taxonomic and vertical resource partitioning among three poorly known deep-sea fishes

Pelagic predators are effective biological samplers of midtrophic taxa and are especially useful in deep-sea habitats where relatively mobile taxa frequently avoid observation with conventional methods. We examined specimens sampled from the stomachs of longnose lancetfish, Alepisaurus ferox, to describe the diets and foraging behaviors of three common, but poorly known deep-sea fishes: the hammerjaw (Omosudis lowii, n = 79, 0.3–92 g), juvenile common fangtooth (Anoplogaster cornuta, n = 91, 0.6–22 g), and juvenile Al. ferox (n = 138, 0.3–744 g). Diet overlap among the three species was high, with five shared prey families accounting for 63 ± 11% of the total prey mass per species. However, distinct differences in foraging strategies and prey sizes were evident. Resource partitioning was greatest between An. cornuta that specialized on small (mean = 0.13 ± 0.11 g), shallow-living hyperiid amphipods and O. lowii that specialized on large (mean = 0.97 ± 0.45 g), deep-dwelling hatchetfishes. Juvenile Al. ferox foraged on a high diversity of prey from both shallow and deep habitats. We describe the foraging ecologies of three midtrophic fish competitors and demonstrate the potential for biological samplers to improve our understanding of deep-sea food webs.

Sternoptychidae (e)).Axes are on the log10-scale.ANCOVA results for each model are given in Table S6.Anoplogaster cornuta, and Omosudis lowii are given for each predator as the total number (n), mean proportional abundance (N ̄), total mass (m), mean proportional mass (M ̄), frequency (FO) and percent frequency of occurrence (% FO).These metrics are also summarized at the prey type level for each species.See "Supplementary_TableS2.xlsx".
Table S3: Observed Shannon Diversity ( 1 D) and estimated true diversity ( 1 Dex (± se)) quantified using family-level diversity accumulation curves (see Fig. S2).The sample size included in the analysis (n) and the average sample size required to observe 95% of the estimated true diversity (95% coverage, t95%) are also given.

Fig. S1 :
Fig. S1: Number of specimens dissected by collection year for each predator species.In all panels, darker shading indicates stomachs with prey and lighter shading indicates empty stomachs.
Fig. S2: Diversity accumulation curves describing rate of family-level prey discovery quantified as Shannon Diversity (hill number of order q =1).Solid lines indicate interpolated diversity up to the sample size (circles) and dashed lines indicate extrapolated diversity to two-times the sample size.

Fig. S3 :
Fig. S3: Partial-effects plots from generalized additive models describing changes in the proportional mass (M ̄) of broad prey types with predator size (a, b, c) and the effects of predator species and mass on estimated foraging depth (d, e).Plots describe the relationship between each covariate and its parametric contribution ("f(x)") or the contribution of its smoother ("s(x)") to the model's fitted values.See Table2for full model summaries.

Fig. S4 :
Fig. S4: Linear models describing the effects of predator species and mass on individual prey mass for broad prey types (crustaceans (a), fishes (b), molluscs (c)) and the most abundant prey families (Phrosinidae (d),

Fig. S5 :
Fig. S5: Median depth assignments for all unique prey taxa colored by prey type.Some taxa are represented by multiple size classes with different depth habitats, see TableS1.Overlaying density plot summarizes the bimodal distribution of median depths of occurrence.

Table S1 :
Median depth assignments for each prey taxa (per size class when appropriate).See READ_ME and data tables in "Supplementary_TableS1.xlsx".

Table S2 :
Contributions of each unique prey taxa to the diets of juvenile Alepisaurus ferox, juvenile

Table S4 :
Mean pairwise Morisita-Horn Similarity for all within-and between group comparisons are given in the lower half of table (gray shading).Test statistics from PERMANOVA and PERMDISP are given in the upper half (t-value PERMANOA, t-value PERMDISP).The p-values for all comparisons are 0.001 unless otherwise indicated ( "*" = 0.001< p <0.01, "**" = 0.01< p < 0.05).

Table S5 :
Summaries for multilinear models describing the effects of predator species and mass on individual prey mass and the total count and mass of prey per stomach (see Fig.3).Model results are given as the Adjusted R 2 (adj.R 2 ), F-statistic, an p-value.The estimate, standard error (se), t-values (t), and p-values (p) are given for all terms and their interactions (:) for each model.

Table S6 :
Model results and ANCOVA outputs for all linear models in Fig.S3describing effect of predator species and mass on prey mass for each main prey group (a-c) and the two most abundant prey families (d,e).The sum of squares (SS), degrees of freedom (df), F-statistics (F) and p-values (p) are given for terms and their interactions (:) for each model.