Altered juvenile fish communities associated with invasive Halophila stipulacea seagrass habitats in the U.S. Virgin Islands

Caribbean seagrass habitats provide food and protection for reef-associated juvenile fish. The invasive seagrass Halophila stipulacea is rapidly altering these seascapes. Since its arrival in the Caribbean in 2002, H. stipulacea has colonized and displaced native seagrasses, but the function of this invasive seagrass as a juvenile fish habitat remains unknown. To compare diversity, community structure, and abundance of juvenile fish between H. stipulacea and native seagrass beds, fish traps were deployed in four nearshore bays around St. Thomas, U.S. Virgin Islands. Traps were deployed in Frenchman, Lindbergh, and Sprat Bays for 24 h intervals in patches of bare sand, patches of H. stipulacea and patches of the native Caribbean seagrasses Thalassia testudinum and Syringodium filiforme. Traps were then deployed in Brewers Bay for 12 h intervals in stands of H. stipulacea and S. filiforme. Relative and total abundances of juvenile fish, identified at least to family, were compared across treatment habitats for each trap deployment period. The catch from H. stipulacea, compared to native seagrasses, comprised a greater abundance of nocturnal carnivores Lutjanus synagris (family Lutjanidae) and Haemulon flavolineatum (family Haemulidae). Additionally, the herbivore species Sparisoma aurofrenatum (family Labridae) and Acanthurus bahianus (family Acanthuridae) and the diurnal carnivore species Pseudopeneus maculatus (family Mullidae) were relatively scarce in H. stipulacea. The catch from sand was much smaller, compared to vegetated habitats, and comprised only L. synagris, H. flavolineatum, and H. aurolineatum. These results provide evidence of reduced family diversity and altered juvenile fish assemblages in H. stipulacea, driven by an abundance of some nocturnal carnivores and scarcity of herbivores and diurnal carnivores. The findings from the present work underpin the need for further investigation and mitigation of this invasion, particularly where H. stipulacea is driving seascape-alterations of key juvenile fish habitats.


Introduction
The invasion of H. stipulacea in the USVI and wider Caribbean often corresponds to displacement of native seagrasses, but the impact of this invasion on the juvenile fish that rely on these affected habitats is poorly understood [28]. Alteration of juvenile fish assemblages have occurred elsewhere as a result of invasion by habitat-forming species such as the common reed Phragmites australis [29], and the green alga Caulerpa taxifolia [30]. One recent study conducted in Dominica found that H. stipulacea supported greater abundances of epibiont food resources and larger overall fish (inclusive of all life stages), compared to the native seagrass S. filiforme. Fish abundance, however, did not differ between the native and invasive seagrass; it was cautioned that this could be due to low experimental power (27% power to detect a 50% difference at alpha = .05). Interestingly, juveniles were twice as abundant in S. filiforme compared to H. stipulacea, suggesting that S. filiforme habitats in Dominica may be a better juvenile habitat [22].
The present study investigated how juvenile fish diversity, community structure, and abundance differ between habitats composed of H. stipulacea, native seagrass S. filiforme, native seagrass Thalassia testudinum, and sand in four shallow bays of St. Thomas, USVI. The exclusive focus on juveniles in the present work is because, aside from the value of seagrasses to juvenile fish, little is known about seagrass-associated juvenile assemblages in St. Thomas; most research has concerned reef-associated adult populations around St. Thomas (e.g., [31,32]). To explore broad-scale patterns in habitat use, fish traps were deployed at three sites for 24 h periods within patches of H. stipulacea, S. filiforme, T. testudinum, and sand. To investigate diel variation in habitat use, traps were next deployed at a fourth site for 12 h periods within patches of H. stipulacea and S. filiforme.
The first hypothesis in this study was that there would be less diverse and less abundant juvenile fish associated with H. stipulacea compared to native seagrass species, as the invasive may not provide adequate shelter or food resources [28,33]. The second hypothesis was that there would be more diverse and abundant juvenile carnivores in H. stipulacea, due to recent findings that H. stipulacea supports abundant carnivore food resources [22]. The third hypothesis in the present study was that there would be less diverse and less abundant juvenile herbivores in H. stipulacea, as the invasive likely does not provide adequate shelter and may be associated with fewer or suboptimal herbivore food resources. Though not studied in the Caribbean, Mediterranean H. stipulacea was associated with less diverse and abundant algal epiphytes compared to other seagrasses [34]. Hypotheses were tested equally during 24 h and 12 h deployments.

Ethics statement
This study was carried out in strict accordance with the American Fisheries Society Guidelines for the Use of Fishes in Research. A passive capture method was used, specimens were handled briefly, and no invasive techniques were used. This protocol was approved by the University of the Virgin Islands IACUC (IRB #747807-1). No endangered or protected species were involved in this study. All study sites are publicly-owned waterways, and fish capture was conducted with permission from the Department of Planning and Natural Resources Division of Fish and Wildlife (permit #DFW15053T).

Site selection
Data collection for this study occurred during February 2016 across four sites located along the south side of St. Thomas, USVI (Fig 1). Frenchman, Lindbergh, and Sprat Bays were selected for 24 h trap deployment sites (Fig 1A-1C) because each bay contained distinct monotypic stands of H. stipulacea, S. filiforme, and T. testudinum. Brewers Bay (Fig 1D) was selected for 12 h trap deployments because it contained particularly large monotypic stands of H. stipulacea and S. filiforme.
The calmest conditions for southern St. Thomas bays occur during winter months (December to March) because they are sheltered on the leeward side of the island from seasonal northeast winds. Between April and November, prevailing winds are out of the southeast and expose easternmost bays (Frenchman and Sprat) to high winds and wave action. Westernmost bays (Brewers and Lindbergh) are the calmest year-round due to their leeward orientation.
At every site, prospective seagrass patches were identified, and their size, depth, distance from shore, and distance from other patches were noted. Habitat characteristics were then surveyed in each prospective patch. Five 0.25m 2 quadrats were placed randomly in each patch, and percent cover of the target seagrass species in that quadrat was estimated. Then the canopy height was recorded by measuring the length of one shoot on the bottom right of the quadrat.
Only monotypic stands were chosen as treatment habitats. Monotypic stands were defined in this study as patches that were larger than 10 m 2 with a greater than 75% cover of the target seagrass species. Additionally, only patches at similar depths (2-4 m), close to shore (<100 m), and relatively isolated from other seagrass species (>10 m) were chosen as trap deployment locations. Sand habitats with no structure or vegetation over at least a 10 m 2 area were also identified to serve as treatment habitats for 24 h trap deployments. A small threshold area (10m 2 ) could have led to the detection of edge effects, in which fish community structure is influenced by migration, predation, and competition at seagrass-sand and seagrass-reef interfaces [35,36]. To minimize edge effects, seagrass or sand patches in the visible range (within  ~25 m) of well-developed aggregate reef were excluded from this study, and traps were deployed in the center of each patch.

Experimental methods
Fish were collected using unbaited traps (53 x 53 x 53 cm) with 1 cm square mesh made from vinyl-coated wire. Traps had a double funnel opening and resembled rectangular Antillean fish traps. Traps were used because they non-selectively target juvenile fish and do not affect fish behavior, unlike other capture techniques such as hook and line fishing or trawling [37,38]. Some drawbacks to using traps include risks of conspecific attraction [39], high variation across traps [40], repeatedly catching the same individuals [38], and predation [41]. Traps permitted capture of medium and large juveniles, but the 1 cm mesh precluded capture of the smallest juveniles. Exclusion of the smallest juveniles was determined to be permissible. It has been argued that, because mortality is high among the smallest juveniles, it is more reliable to characterize assemblages of medium and large juveniles that have spent more time in their nursery habitat [42].
To investigate broad-scale patterns in juvenile assemblages across habitats and sites, 24 h trap deployments were conducted at three sites (Frenchman, Lindbergh, and Sprat Bays) between February 6-12, 2016. Four traps were assigned to each site (Fig 1A-1C); in each site, one trap was assigned to and deployed in either H. stipulacea, S. filiforme, T. testudinum, or sand habitats daily at 7 am (± 0.5 h). Each trap was hauled five times between February 6-12, for a total of five trap samplings over the duration of 24 h deployments.
To investigate fine-scale diel variation in juvenile assemblages across native and invasive seagrasses, 12 h trap deployments were conducted in Brewers Bay between February 17-20, 2016. Five traps each were allocated to H. stipulacea and S. filiforme beds in Brewers Bay ( Fig  1D) and deployed at least 10 m apart within the treatment habitat. Twice daily at 7 am and 7 pm (± 0.5 h), traps were collected and redeployed in both habitats. Each of the five traps in both treatment habitats was hauled for three consecutive nights and days between February 17-20, for a total of three nighttime trap samplings and three daytime trap samplings over the duration of 12 h deployments.
In the field, each trapped fish was identified to either species or family, its total length was measured (to the nearest mm), and it was photographed before being released. Fish that could not be identified at least to the taxonomic level of family in the field or from photographs were excluded from all analyses. Fish that exceeded the length at maturity [43][44][45][46][47][48][49] for their species were excluded from all analyses. When an individual could be identified only to family, the smallest length at maturity for all species in its family was used to provide the most conservative cutoff ( Table 1). All families were categorized as either herbivores, diurnal carnivores or nocturnal carnivores [17,[50][51][52].

Statistical analysis
Pielou's evenness indices (J) and Simpson's diversity indices (D) were calculated to compare family diversity between treatment habitats (H. stipulacea, S. filiforme, T. testudinum, and sand). For a more representative sample size, catch data from 24 h trap deployments were pooled across sites and days, and catch data from 12 h trap deployments were pooled across traps and days. Simpson's D is defined in the present study as one minus the probability of randomly picking two individuals from the population that are the same species: where S is the number of species in a sample or population, and p i is the fraction of a sample of individuals belonging to species i. This index was chosen because it is one of the simplest calculations of species diversity, and it is appropriate for aquatic systems [53]. Indices of family diversity are only nebulous descriptions of communities, but they can provide practical summaries of community structure when implemented as part of a more detailed analysis [54].
Non-metric multidimensional scaling (NMDS) biplots, based on the Bray-Curtis similarity measure, were generated to discern differences in juvenile assemblages across treatment habitats in each trap deployment period. Total abundance data were converted to relative abundances, and traps with only one fish were excluded when the family of that fish was unique. Ellipses, generated from the standard error of the weighted average of the scores in each treatment habitat, were superimposed on each biplot to visualize the dispersion of each treatment habitat group. A permutational analysis of variance (PERMANOVA) was conducted to test the multivariate response of fish family relative abundance to treatment habitats. When PER-MANOVAs were significant (p < 0.05), a similarity percentage analysis (SIMPER) was performed to evaluate the contribution of each family to the difference between treatment habitats.
Univariate analyses were then used to explore differences in total abundance of juveniles across treatments and strata. Two-way analyses of variance (ANOVA) were conducted on the mean total abundance of juvenile fish (identified at least to family) across treatment habitats and sites in 24 h deployments and across treatment habitats and soak types (day vs. night) in 12 h deployments. To meet assumptions of homoscedasticity and equal variance, all data were square-root transformed prior to conducting ANOVAs.
Total abundance data were then separated by guild, and a variety of analyses were used to explore responses of each guild to treatment habitats and strata. For carnivores in 24 h deployments, data were square-root transformed, and a two-way ANOVA was used to test for a significant response of mean carnivore abundance to treatment habitats and sites. The data for each guild in 12 h deployments were zero-inflated. Generalized linear models (GLMs) were therefore used to test for significant responses in mean abundance of nocturnal carnivores, L m = length of maturity, defined as the total length at which 50% of the population of that species becomes mature for the first time [43]. The smallest L m (in bold) in each family was applied as the cutoff length for individuals that were identified only to family. https://doi.org/10.1371/journal.pone.0188386.t001 The invasive seagrass H. stipulacea as a juvenile fish habitat diurnal carnivores, and herbivores to treatment habitats and soak types. GLMs with a Poisson distribution were rejected after examination of the ratio of the residual deviance to the degrees of freedom suggested overdispersion. Alternatively, GLMs with negative binomial error structures and log-link functions were used. Starting at the null model, model fit was assessed incrementally by adding a predictor and conducting a X 2 goodness of fit test. The best model was determined when, compared to a smaller model, the null hypothesis that the model with fewer predictors is adequate was rejected at the p < 0.05 level. For each guild, the predictor from the best fit model was determined to have a significant effect on the magnitude of the response at p < 0.05, and estimates (describing the response in S. filiforme relative to H. stipulacea for the seagrass predictor and describing the response in night relative to day for the soak type predictor) were used to assess the direction of the response between levels of the predictor. All statistical analyses were performed in R [55].

Juvenile assemblages
During 24 h deployments, a total of 117 fish were caught. Seven fish were not identified to the family level, and 27 fish were unable to be confirmed as juveniles. All analyses of 24 h deployments were completed on the remaining 83 juveniles. During 12 h deployments, a total of 188 fish were caught. Five fish were not identified to the family level, and 14 fish were unable to be confirmed as juveniles. All analyses of 12 h deployments were completed on the remaining 169 juveniles. Diversity. The juvenile fish assemblage associated with H. stipulacea was consistently less diverse than that of native seagrasses. During 24 h deployments, H. stipulacea was associated The invasive seagrass H. stipulacea as a juvenile fish habitat with greater family richness but lower family diversity (n = 3 families, D = 0.21) than sand habitats (n = 2 families, D = 0.50); the native T. testudinum was associated with the greatest family richness and diversity (n = 5 families, D = 0.77) during this 24 h deployments. During 12 h deployments, H. stipulacea supported the lowest overall family richness and diversity during both nighttime (n = 3 families, D = 0.50) and daytime (n = 5 families, D = 0.61) soaks. Juvenile family richness and diversity were consistently greater in native S. filiforme during nighttime (n = 6 families, D = 0.67) and daytime (n = 5 families, D = 0.71) soaks (Table 3). Community structure. During 24 h trap deployments, the juvenile fish communities associated with H. stipulacea, S. filiforme, and sand were relatively similar to each other but distinct from T. testudinum (Fig 2). The comparable communities of H. stipulacea, S. filiforme, and sand were driven by a considerable relative abundance of Lutjanidae and Haemulidae in each habitat. Communities in T. testudinum were uniquely shaped by a high relative abundance of Labridae (n = 5) and the presence of Sciaenidae (n = 1), as these families were absent from all other treatment habitats (  Table 3) during the daytime in S. filiforme. The daytime catch from S. filiforme also included herbivores (n = 3 Acanthuridae, n = 6 Labridae).  The invasive seagrass H. stipulacea as a juvenile fish habitat Nocturnal carnivores, diurnal carnivores, and herbivores were associated with S. filiforme, while the assemblage associated with H. stipulacea was comprised almost entirely of nocturnal carnivores. Despite greater family richness associated with S. filiforme, the community structure associated with S. filiforme overlapped with H. stipulacea (Fig 3) and community structure did not differ significantly between the two seagrass habitats (PERMANOVA: F1, 46 = 2.400, p = 0.070). The communities in each seagrass were pooled over soak types (daytime and nighttime), and similar nighttime abundances of Lutjanidae and Haemulidae drove statistically indistinguishable differences in community structure across seagrass habitats in 12 h deployments.
Abundance. Across 24 h trap deployments, average juvenile fish abundance differed significantly across habitats (p = 0.005), and sites (p = 0.029), with a significant interaction (p = 0.007) (Table 4A) (Fig 4). Juveniles were significantly more abundant in H. stipulacea than in sand habitats (Tukey test p = 0.002) and in Lindbergh Bay than in Sprat Bay (Tukey test p = 0.039). A significant interaction was due to a greater juvenile abundance in Lindbergh Bay H. stipulacea compared to Sprat Bay H. stipulacea (Tukey test p = 0.0013) (Fig 4).
Carnivores (both nocturnal and diurnal) comprised 94% of the total catch of 83 juvenile fish in 24 h deployments. The nocturnal carnivore family Lutjanidae were the most abundant family (n = 54); other carnivores included nocturnal families Haemulidae, Holocentridae, and Sciaenidae (n = 18) as well as the diurnal family Mullidae (n = 2) ( Table 3). Five herbivores (Labridae) represented the remaining 6% of total fish abundance, and all Labridae were caught in T. testudinum beds in either Sprat Bay (n = 2) or Lindbergh Bay (n = 3).
Because carnivores comprised the majority of the total catch, results from two-way ANO-VAs on carnivore abundance were similar to two-way ANOVAs on total abundance. Carnivore abundance differed significantly across habitats (p = 0.002) and sites (p = 0.043), with an interaction (p = 0.017) (Table 4B). Carnivores were more abundant in H. stipulacea than in T. testudinum (Tukey test p = 0.008) and sand habitats (Tukey test p = 0.003). Carnivores were also significantly more abundant in Lindbergh Bay than in Sprat Bay (Tukey test p = 0.038). A significant interaction was driven by a greater abundance of carnivores in Lindbergh Bay H. stipulacea than in Sprat Bay H. stipulacea (Tukey test p = 0.044). The invasive seagrass H. stipulacea as a juvenile fish habitat Across 12 h trap deployment periods in Brewers Bay, mean juvenile fish abundance did not significantly differ between seagrass beds (p = 0.403) or soak type (night vs. day) (p = 0.060) ( Table 4) (Fig 5A). Significant differences emerged when analyses were conducted on separate guilds.
GLMs revealed that nocturnal carnivore abundance differed across soak types and diurnal carnivore and herbivore abundance differed across seagrass habitats. Nocturnal carnivores were more abundant at night, but their abundance did not differ significantly between H.  The invasive seagrass H. stipulacea as a juvenile fish habitat stipulacea and S. filiforme habitats ( Table 5) (Fig 5B). Haemulidae comprised the majority of the catch in both seagrasses at night (n = 52 in H. stipulacea, n = 32 in S. filiforme), but much fewer were caught during the day in either seagrass (n = 17 in H. stipulacea, n = 19 in The invasive seagrass H. stipulacea as a juvenile fish habitat S. filiforme) (Fig 5A). Diurnal carnivores and herbivores were significantly more abundant in S. filiforme than in H. stipulacea (Table 5) (Fig 5C and 5D). While diurnal carnivores were abundant in S. filiforme (n = 24 Mullidae, n = 2 Carangidae), only two (both Mullidae) were caught in H. stipulacea over the duration of the experiment (Fig 5A). Similarly, herbivores were abundant in S. filiforme (n = 6 Labridae, n = 13 Acanthuridae), but only one Acanthuridae was trapped in H. stipulacea over the duration of the experiment (Fig 5A).

Discussion
This study is the first to characterize juvenile fish assemblages in St. Thomas  Comparison between this study and previous reports is important yet challenging due to differences in sampling technique, habitats, seasons, geographic localities, and environmental conditions [42]. The catch data from this study reflects a relatively low diversity compared to elsewhere in the Caribbean, but this could be partly explained by an exclusive focus on the juvenile life stage. Compared to studies that report catch-data inclusive of all life stages, a focus on juveniles in the current study led to the omission of families when individuals exceeded L m ; for example, adults belonging to the family Tetraodontidae were caught on multiple occasions during 24 h deployments. Juvenile fish diversity may also be reduced from historical levels [56] due to overfishing of reproductive adults, and therefore diminished larval supply, of many commercially important species in St. Thomas waters [31].
The identity of habitats surrounding seagrass beds is an important driver of seagrass-associated juvenile assemblages. In both the present study and previous reports from Dominica, mangrove habitats were absent surrounding seagrass study sites [22]. In contrast, studies from Curacao and Guadeloupe were conducted near large red mangrove forests in shallow (<2 m) seagrass habitats [3,57]. The catch of Lutjanidae from both the present work and Dominica  [3,57,58]. Though less abundant, herbivore assemblages in the present study were also more similar to the catch from Dominica. Labridae were absent from seagrasses adjacent to mangroves in Guadeloupe, and Acanthurus bahianus was notably absent from seagrasses adjacent to mangroves in both Curacao and Guadeloupe [3,22,57]. Moreover, species such as Haemulon flavolineatum (family Haemulidae) and Pseudopeneus maculatus (family Mullidae) were relatively abundant in the current study, but absent from mangrove-associated seagrass habitats in Guadeloupe [3]. Species and lifestage specific habitat preference can explain discrepancies in juvenile assemblages across these different habitats. Some reef-associated species (e.g. A. bahianus) may remain close to the reef and avoid seagrass-mangrove habitats, some (e.g. O. chrysurus) may initially inhabit sheltered seagrass and mangrove habitats but gradually migrate to seagrass habitats adjacent to reef as they grow, and others (e.g. L. griseus and L. apodus) may remain in seagrass-mangrove habitats throughout the majority of their juvenile life-stage [3,57]. Physical characteristics, particularly canopy height, determine what protection and food resources a seagrass habitat may provide [59]. Canopy heights reported from this study are similar to those previously reported from Dominica (20 cm for S. filiforme, 18 cm for T. testudinum, and 4 cm for H. stipulacea) [24]. Varying canopy heights among the three seagrass species may partly explain why fewer families were present in H. stipulacea than in native seagrasses. In the present study, the native seagrasses T. testudinum and S. filiforme had the tallest canopies and also supported the most diverse assemblages. The patterns of juvenile fish diversity reported in the present work corroborate previous studies that linked tall canopies and greater fish diversity [59]. Moreover, the canopy height of H. stipulacea was recently attributed to reduced diversity of associated fish assemblages in Bonaire [33]. Juveniles may be deterred from using H. stipulacea habitats if the invasive species' short canopy provides less protection, increases predation risk, and reduces probability of survival [60].
Contrary to the association between seagrass canopy height and increased juvenile fish diversity, diversity indices indicate that sand supported a more diverse assemblage than H. stipulacea. The Simpson's D for sand habitats was inflated, however, as a result of a small catch and incidentally perfect species evenness. In general, juvenile fish prefer vegetated habitats over unvegetated habitats [61], and this is in line with the present findings of greater abundance in H. stipulacea and greater diversity in native seagrasses, compared to sand. H. stipulacea opportunistically expands into unvegetated substrates (S1 Fig) and has driven elimination of sand patches and reef halos in Dominica [62]. Though not preferred by juvenile fish, sand habitats are important for schooling fish and large mobile individuals [61]. While the invasive seagrass, by its vegetative nature, may support a greater abundance of juvenile fish, species that depend on sand habitats may suffer habitat loss due to expansion of H. stipulacea.
Nocturnal carnivores were more abundant in H. stipulacea than in native seagrasses during 24 h deployments, implying that the juveniles of some species belonging to this group may prefer H. stipulacea. Preference for H. stipulacea may be due to an abundance of preferred food resources associated with the invasive seagrass. In Dominica, amphipod epibionts were more abundant on H. stipulacea than in S. filiforme [22], and amphipods are preferred food items for juvenile L. synagris [63], the most abundant nocturnal carnivore in 24 h deployments. It has been suggested that epibiont abundance in H. stipulacea is a result of its densely matted structure [22], and H. stipulacea treatment habitats in the present study were composed of dense mats of the invasive seagrass.
Contradictory to results from 24 h deployments, nocturnal carnivore abundance in 12 h trap deployments did not differ across treatment habitats. This inconsistency may be a product of limited temporal extent and sample size, but other explanations are proposed. Conflicting results could be due to dissimilar abundances of Lutjanidae and Haemulidae across trap deployment periods, as species belonging to these families exhibit a diverse range of habitat preferences [2]. The catch from 24 h deployments was dominated by L. synagris, but H. aurolineatum, H. flavolineatum, and L. synagris were similarly abundant in 12 h deployments. H. aurolineatum, which were almost exclusively caught in S. filiforme in 12 h deployments, may prefer the native seagrass. Previous gut content analyses revealed that juvenile Haemulidae consume mainly tanaid prey [64], while juvenile L. synagris prefer amphipod prey [63]. Amphipods were the only crustaceans previously quantified in H. stipulacea [22], however, so it remains unknown whether tanaid abundance differs between H. stipulacea and S. filiforme. H. aurolineatum may also forage in S. filiforme as a result of competitive exclusion from H. stipulacea habitats; populations that migrate daily from large coral reefs to smaller seagrass habitats likely face heightened competition for food and space [64].
The orientation of H. stipulacea and S. filiforme in Brewers Bay may also explain inconsistent findings in nocturnal carnivore abundance across 24 and 12 h deployment periods. Juvenile Lutjanidae and Haemulidae are highly mobile while foraging and thus may migrate haphazardly between seagrass patches [5]. The habitat configuration in Brewers Bay may facilitate such migration; while treatment habitats in the 24 h deployment sites were separated by sand, the border between the two treatment habitats in Brewers Bay was comprised of a mixed stand of S. filiforme and H. stipulacea. A sheltered corridor between the two treatment habitats (provided by continuous vegetation) may permit juveniles to use the Brewers Bay seascape as a spatial continuum instead of discrete patches. Haphazard use of H. stipulacea and S. filiforme habitats has been discussed previously [22], and facilitation of this indiscriminate behavior (by a vegetated corridor) may have led to indistinguishable abundances across Brewers Bay treatment habitats.
Nocturnal carnivores were more abundant in nighttime soaks than in daytime soaks, reflecting this group's daily transition between habitats. Previous studies suggest that Lutjanidae and Haemulidae use seagrasses as a foraging habitat at night and nearby coral reefs as sheltering habitats during the day [64]. Diel variation in habitat use is a way to take advantage of available resources through balancing high growth rates (abundant nighttime food resources provided by seagrass beds) with low predation-induced mortality (ample daytime shelter provided by coral reefs) [65]. The juveniles abundant in these seagrasses at night likely migrated to nearby coral reefs during daylight hours.
Unlike nocturnal carnivores, juvenile diurnal carnivores belonging to the Mullidae family may avoid habitats composed of H. stipulacea. Diurnal carnivores were most abundant in Brewers Bay S. filiforme habitats during 12 h deployments, and P. maculatus was the most common species identified. P. maculatus, like most Mullidae, are benthic carnivores that forage in and ingest sediments to consume their preferred food, and their specialized diets drive their requirement for sparsely vegetated foraging habitats [66,67]. Habitat alterations resulting from the invasion of the marine alga Caulerpa taxifolia to the French Mediterranean coast were recently linked to restricted foraging of the striped red mullet Mullus surmuletus (family Mullidae). It was suggested that the invasion of C. taxifolia could increase intraspecific competition and restructure M. surmuletus populations [68]. H. stipulacea is structurally similar to the densely matted invasive macroalgae, and its displacement of S. filiforme in Dominica led to a loss of space under seagrass canopies and between seagrass shoots [24]. Similar to the response of M. surmuletus to invasion by C. taxifolia, heightened intraspecific competition among P. maculatus populations may result from invasion of H. stipulacea and loss of preferential foraging habitat.
Juvenile herbivores may prefer native seagrasses T. testudinum and S. filiforme habitats over H. stipulacea. Labridae was the most abundant family and shaped the unique community associated with T. testudinum. Three out of five herbivores caught in T. testudinum were Sparisoma aurofrenatum. The association between S. aurofrenatum and this seagrass corroborates a study from the Florida Keys, in which patch reefs and surrounding T. testudinum beds were identified to be the home range of S. aurofrenatum [69]. The absence of herbivores in S. filiforme during 24 h deployments, however, is inconsistent with previous findings that selective grazing on S. filiforme by S. aurofrenatum and other herbivores plays a role in zonation of S. filiforme and T. testudinum [70]. The absence of herbivores in S. filiforme during 24 h deployments is not too notable, however, considering that a total of only five juvenile herbivores were trapped during this period.
Proximity to reef in Brewers Bay may explain why herbivores were more abundant in 12 h deployments than in 24 h deployments. Herbivores tend to be more abundant in closer proximity to reefs [3], and seagrass patches in Brewers Bay were relatively close to a well-developed reef, unlike the treatment habitats in Frenchman, Lindbergh, and Sprat Bays. During daytime trap deployments, herbivores were absent from H. stipulacea, perhaps implying that food resources provided by the invasive are sub-optimal. If juveniles graze on seagrass blades, the nutritional content of H. stipulacea may be inferior to S. filiforme, as previous studies suggest that minor changes in nutritional content can sway herbivore preference for one seagrass over another [71]. If juveniles instead graze on epiphytic algae, the algal community of H. stipulacea may be inferior to S. filiforme. In the Mediterranean, leaves of H. stipulacea were found to support a poor variety of epiphytic flora and a distinct absence of encrusting coralline algae, compared to other seagrasses [34].
The catch of Acanthuridae during 12 h deployments was unexpected; the nighttime catch abundance of this diurnal group was double that of the daytime abundance. Fewer Acanthuridae during daytime soaks may be explained by the highly mobile behavior of this family; individuals been observed frequently entering and exiting traps during active daytime hours [41]. The nighttime catch of Acanthuridae in the present study was also inconsistent with previous observations that this family migrates from seagrass beds to nearby reefs at dusk to seek shelter [2]. It is possible that traps provided an incidental source of nighttime shelter and attracted more juvenile Acanthuridae than S. filiforme would otherwise harbor at night.
This study is one of only a few to characterize juvenile fish communities at multiple sites around the same island [42]. Uniform sampling permitted robust investigation of juvenile assemblages in native and invasive seagrasses across four sites. Patterns of nocturnal carnivore abundance across treatment habitats varied with site, with Lindbergh Bay H. stipulacea habitats supporting the greatest abundance. Though environmental quality data were not collected in the present study, an interaction between treatment habitat and site may be due to site-specific water temperature and salinity. Lindbergh Bay is subjected to high levels of terrestrial runoff due to urban development, and the bay receives inputs of thermohaline effluent from a nearby power plant [72]. Seagrass-associated juvenile fish assemblages in non-estuarine embayments can be influenced by salinity and temperature changes, driving increased density of a select few juvenile species but reduced density of most others [73]. Such patterns have only been linked to seasonal drops in salinity and temperature due to rainfall, however, and no reports could be found linking terrestrial runoff to altered juvenile communities in tropical seagrass ecosystems.
The present study is limited in sample size and temporal extent, compared to similar studies that carried out more than 20 trap pulls per seagrass habitat over multiple months [22]. Replication in the present study was limited because few traps were deployed in each treatment habitat (n = 1 and 5 habitat -1 site -1 in 24 and 12 h deployments, respectively), and trap samplings took place within a brief period (Feb 6-20, 2016). Because juvenile assemblages vary greatly with seasonal changes in water temperatures as well as wind speed and direction, a thorough characterization of juvenile assemblages requires repeated field surveys across seasons [74,75]. Limited replication and temporal extent prevents broad conclusions from being drawn from the present study, and such limitations could be responsible for conflicting results in nocturnal carnivore abundance across trap deployment periods. Consistent differences in assemblage structure between native and invasive seagrasses identified in the present work, however, represent a potentially concerning consequence of the invasion of H. stipulacea.

Future work and management implications
Despite the rapid takeover of H. stipulacea in many parts of the Caribbean, few studies explore potential trophic consequences of this invasion. Further trapping studies that address a broader temporal or geographic extent would be valuable to establish the degree of seasonal and cross-island variability in juvenile assemblages associated with H. stipulacea. Additionally, combining stable isotope and fatty acid analyses with trapping experiments would be instrumental in resolving uncertainties about species-specific foraging in H. stipulacea; previous work has demonstrated the use of these techniques in determining where fish are foraging and what they are consuming [9,76]. Dietary analyses will be more informative once the nutritional content and algal communities of Caribbean H. stipulacea are characterized, similar to how the crustacean epibiota associated with the invasive have been described [22]. If future studies corroborate the results of the present work, the abundance, growth, and mortality of juvenile herbivores and diurnal carnivores should be monitored in bays where H. stipulacea is displacing preferable nursery habitat. Because large-scale habitat characteristics (on the scale of hundreds of meters) can play a considerable role in structuring reef fish populations [77], future research should also address the impacts to fish communities as a result of the homogenizing effect of H. stipulacea. Lastly, the expansion of H. stipulacea should be monitored closely, with particular regard to its ongoing spread to Caribbean islands, colonization following disturbance, and displacement of native seagrasses.
Special management priority should be given to seagrass habitats known to be key juvenile fish nurseries because alterations to these seascapes could disproportionately affect the surrounding habitats to which juveniles eventually recruit [1]. In Marine Protected Areas (MPAs) where H. stipulacea invasions have already occurred, such as the Virgin Islands National Park [23], special assessments should be made to determine the magnitude of threat that the invading seagrass poses to juvenile fish assemblages. Key juvenile habitats not presently within MPA boundaries should be identified using improved spatial prediction of juvenile fish species richness and abundance and prioritized in regards to MPA candidacy [74]. The present study demonstrates that H. stipulacea may be a suboptimal nursery habitat for herbivores and diurnal carnivores, and further monitoring should be carried out to identify species-specific patterns in juvenile use of habitats composed of the invasive seagrass. Functional redundancy of juvenile habitats is already reduced around the urbanized island of St. Thomas; intense coastal development has resulted in the loss of most mangrove habitats [78], another key juvenile fish habitat. In St. Thomas, therefore, native seagrasses may be the only remaining preferable habitats for some juvenile species, and displacement of these seagrasses by H. stipulacea may reduce survivorship if juveniles are obligated to seek suboptimal nursery habitat.
Efforts should be made to inhibit the spread of H. stipulacea to other Caribbean islands (e.g. Puerto Rico) and Florida through enforcement of proper recreational and commercial boating and anchoring practices. This will help minimize fragmentation and vegetative propagation, the principle means of H. stipulacea dispersal throughout the Caribbean [23]. Once H. stipulacea has invaded, a number of actions may help mitigate the negative impacts of its introduction. While some suggest that physical removal of the invasive will be ineffective [28], others have proposed dredging H. stipulacea beds and seeding areas with native seagrasses [79,80]. One alternative to complete removal is selective extraction of dense mats of H. stipulacea, which may prevent sediment anoxia and liberate foraging habitat for benthic carnivores [27,68]. Working to minimize the spread of H. stipulacea in tropical nearshore ecosystems will be vital to preserving landscape complexity, habitat quality, and ultimately biodiversity in these productive environments.