Shifts in biodiversity and physical structure of seagrass beds across 5 decades at Carriacou, Grenadines

RESEARCH ARTICLE Shifts in biodiversity and physical structure of seagrass beds across 5 decades at Carriacou, Grenadines David Patriquin1, Robert E. Scheibling1, Karen Filbee-DexterID2,3* 1 Department of Biology, Dalhousie University, Halifax, Nova Scotia, Canada, 2 University of Western Australia, Perth, Australia, 3 Institute of Marine Research, His, Norway * karen.dexter@uwa.edu.au Abstract Caribbean seagrass beds are facing increasing anthropogenic stress, yet comprehensive ground-level monitoring programs that capture the structure of seagrass communities before the 1980s are rare. We measured the distribution of seagrass beds and species com- position and abundance of seagrass and associated macroalgae and macroinvertebrates in 3 years over a 47-year period (1969, 1994, 2016) at Carriacou, Granada, an area not heavily impacted by local human activity. Seagrass cover and physical parameters of fringing beds were measured in transects at high (HWE) and low wave energy (LWE) sites; frequency of occurrence of all species, and biomass and morphology of seagrasses, were measured at 100 m2 stations around the island. Losses in nearshore seagrass cover occurred at HWE but not LWE sites between 1969 and 2016 and were associated with increases in the sea- grass-free inshore zone (SFI) and erosional scarps within beds. Total biomass did not vary across years although there were progressive changes in seagrass composition: a decline in the dominant Thalassia testudinum and concomitant increase in Syringodium filiforme,

and establishment of invasive Halophila stipulacea in 2016 at LWE sites. Species richness and diversity of the seagrass community were highest in 1994, when 94% of macroalgae (excluding Caulerpa) were most abundant, and sea urchins were least abundant, compared to 1969 and 2016. Multivariate statistical analyses showed differences in community compo- sition across the 3 years that were consistent with trends in urchin abundance. Increases in SFI and scarp number in seagrass beds at HWE sites occurred mainly after 1994 and likely were related to increased wave forcing following degradation of offshore coral reefs between 1994 and 2016. Our observations suggest that landward migration of seagrass beds with rapidly rising sea level in future will not be realized in reef-protected seagrass beds at Carriacou barring reversal in the processes that have caused reef flattening. Introduction Seagrasses are major components of coastal ecosystems in the Tropical Atlantic Bioregion [1]. They colonize rubble, sand or muddy substrates on reef flats, back-reef lagoons, and in PLOS ONE PLOS ONE | https://doi.org/10.1371/journal.pone.0306897 August 1, 2024 1 / 25 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 OPEN ACCESS Citation: Patriquin D, Scheibling RE, Filbee-Dexter K (2024) Shifts in biodiversity and physical structure of seagrass beds across 5 decades at Carriacou, Grenadines. PLoS ONE 19(8): e0306897. https://doi.org/10.1371/journal. pone.0306897 Editor: Judi Hewitt, The University of Auckland - City Campus: University of Auckland, NEW ZEALAND Received: July 6, 2023

Accepted: June 25, 2024 Published: August 1, 2024 Copyright: © 2024 Patriquin et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability Statement: All data are available in the public iNaturalist Project https://www. inaturalist.org/projects/seagrass-bed-flora-fauna- of-carriacou-grenada-2016. Funding: In 1969 DP was supported by a McConnell Scholarship from McGill University, Montreal. RES was supported by a Discovery Grant (RGPIN-2016-04878) from the Natural Sciences and Engineering Research Council of Canada (https://www.nserc-crsng.gc.ca/). KFD was

sheltered bays and estuaries. Seagrass beds provide significant ecosystem services locally as nurseries and feeding areas for fish on coral reefs and mangroves [2–4], and as habitat for tra- ditionally fished sea urchins [5], green turtle (now protected) [6], conch [7] and lobster [8]. By dampening wave action and stabilizing sediments, seagrasses enhance coastal protection and water quality [9–11] and calcareous epiphytes and seaweeds in seagrass beds are significant contributors to sediments [12]. Interactions between seagrass habitats and nearby coral reefs and mangroves increase overall health of various habitat types [10, 13]. Seagrasses also coun- teract ocean acidification locally by raising pH leading to significantly increased calcification by corals downstream of seagrass beds [14] and by calcareous algae within beds [15]. Sea- grasses, like mangroves and saltmarshes, are significant contributors to blue carbon [16] due to their high productivity and belowground carbon storage [17]. However, there remains con- siderable uncertainty about the total area of seagrass beds globally [18], and carbon storage varies widely according to site conditions and seagrass species [19]. Recent estimates [20] indicate that seagrass cover globally has declined by 19% since the 1880s, with declines occurring in all seven of seagrass bioregions defined by [1]. Absolute losses were highest in the Tropical Atlantic Bioregion, although there has been a trend of increase in seagrass cover in this region since the late 1990s. Principal drivers of sustained

losses globally, and in the Tropical Atlantic Bioregion specifically, are coastal development and poor water quality [20–22]. Increases in seagrass cover have been observed in some areas with- out any management [23] and others after deliberate actions such as reduction of nutrient input [24]. In general, seagrass beds in the Caribbean have proved remarkably resilient to extreme storms, either being minimally impacted (especially those dominated by the climax species Thalassia testudinum) or recovering over time [11, 21]. However there have been shifts towards earlier successional seagrass species, and disruptions by intense storm activity appear to facilitate expansion of Halophila stipulacea [25], an invasive seagrass first noted in the Trop- ical Atlantic at Grenada in 2002 [26]. Such changes can affect ecosystem services such as coastal protection [27]. Sea level rise can have negative or positive effects on the area of seagrass and other coastal wetlands depending on their capacity to build vertically, sediment supply, and the lateral and vertical “accommodation space” for landward migration of wetlands, and extent to which this is realized through local coastal management [19, 28, 29]. Global mean sea level rise for the Carib- bean, estimated as 2.8 mm yr−1 for 1993–2018, is predicted to rise dramatically after 2050, increasing to as much as 30 mm yr−1 by 2100 [30]. Such increases could lead to disproportion- ally large increases in wave energy and negative impacts on seagrass beds in back-reef environ-

ments if vertical reef accretion does not keep pace [31, 32]. Reef “flattening”, the loss of hard coral and reef structural complexity, has been widespread in the Caribbean since the early 1980s [33]. Based on modelling studies [34], showed that a significant increase in mean wave height could occur at present sea level given sustained degradation of reef structural complexity. Observations conducted on seagrass beds at Carriacou, Grenada in the eastern Caribbean in 1969, 1994 and 2016 provide an opportunity to explore changes in seagrass distribution and community composition on a multidecadal scale in an area not heavily impacted by local human activity aside from subsistence fishing. These seagrass beds occur in windward lagoons behind barrier reefs as well as in leeward areas of the island [35, 36]. The barrier reefs have undergone the types of degradation [37] observed more generally for the Caribbean [38, 39]. The initial set of observations in 1969 was conducted as part of a broad characterization of T. testudinum and “turtle grass beds” at Barbados and Carriacou in the eastern Caribbean [35, 40]. Subsets of those observations were repeated at Carriacou, the less developed island, in 1994 and 2016. Scheibling et al. [36] reported on the distribution and abundance of the inva- sive seagrass H. stipulacea at these and other sites in Carriacou in 2016. PLOS ONE Seagrass bed biodiversity and structure and across 5 decades PLOS ONE | https://doi.org/10.1371/journal.pone.0306897 August 1, 2024 2 / 25 supported by the Australia Research Council

(DE190100692, DP220100650) and the Norwegian Blue Forest Network (nbfn.no). The funders had no role in study design, data collection and analysis, decision to publish, or reparation of the manuscript. Competing interests: The authors have declared that no competing interests exist.

The primary objectives of our study are: 1. To compare changes in abundance (cover, frequency of occurrence, biomass, shoot density) and morphology of T. testudinum, considered the dominant or climax species at the onset of our study, along with changes in abundance of other native or invasive seagrass species and of associated macroalgae and sessile invertebrates (corals, sponges) across this 47-year period. 2. To use a species-level measure of abundance (frequency of occurrence) to examine tempo- ral changes in biodiversity and community composition of seagrass beds. 3. To provide concurrent measures of physical parameters of seagrass beds (extent of the sea- grass-free inshore zone, depth profiles and frequency of erosional scarps) that can be affected by changing hydrodynamic forces, and to relate these to population and commu- nity level changes documented in objectives 1 and 2. At this time of accelerating climatic change and increasing levels of global and local human impact on coastal ecosystems, the importance of establishing baselines against which ecologi- cal change can be measured becomes increasingly apparent. Longitudinal studies are particu- larly valuable in establishing relevant baselines or in providing some semblance of a pristine state, that enable comparisons across geographic regions and inform conservation and man- agement initiatives [41]. Most comparisons of seagrass systems over periods of 40 years or more are based on aerial imagery [20, 23]. Comprehensive monitoring programs that capture

ground-level observations on the structure of tropical seagrass communities before the 1980s are rare [22]. Our study provides this insight for a relatively undeveloped location in the Carib- bean that has nonetheless been impacted by broad-scale changes in biotic and abiotic factors that have shaped the local seagrass ecosystem. Materials and methods Study sites and sampling methods To provide a basic description of “turtle grass beds” (seagrass beds dominated by Thalassia tes- tudinum) around Carriacou in 1969, areas putatively dominated by T. testudinum were identi- fied as dark patches in aerial photographs of the island taken in 1966 by the Directorate of Overseas Surveys, United Kingdom (Contract 85, print scale 1:12,900). Detailed tracings of these patches provided cursory maps of seagrass distribution that were used to select study sites representative of seagrass beds around the island. For logistical reasons, most sites were on the east, windward coast. Sites were located by reference to prominent geographic features (e.g., promontories, beaches, cays, reefs) and accessed directly from shore or with a small motorboat. As a measure of ground-truthing, T. testudinum was present at all putative turtle grass sites during the initial sampling in 1969. Sampling took place in March 1969, late January to early March 1994, and January to February 2016 (except 1 transect in February 2019). Sam- pling was mainly done by snorkeling; SCUBA was used to access deeper beds in some offshore regions in 1994. Transects

Eleven sites were selected in 1969 to sample line transects that spanned fringing seagrass beds around Carriacou (i.e., located along the shoreline) [35]. Ten sites were resampled in 1994 and 2016 (Fig 1, S1 Table); one of the 1969 line transects could not be relocated. Transects were run from shore (sea level), and perpendicular to it, across all or part of a fringing bed. Transect length in 1969 was variable [35]; transects extended to at least the point where no further PLOS ONE Seagrass bed biodiversity and structure and across 5 decades PLOS ONE | https://doi.org/10.1371/journal.pone.0306897 August 1, 2024 3 / 25

erosional scarps occurred, confirmed by observation well beyond the last scarp. Direction was maintained using a compass or by reference to two aligned markers on the shore. Transects were conducted within 1–2 h of low tide when low current and high visibility optimized sam- pling. Distances and depths along transects were measured with a calibrated 2-m pole sequen- tially flipped in the horizontal direction and vertically positioned at 10-m intervals, or less where there were abrupt changes in depth; a depth gauge was used for depths > 2 m. Depth Fig 1. Sites where seagrass beds were sampled at Carriacou. Sampling stations (S) were distributed across representative seagrass beds around the island (green circles); transects (T) were sampled at low wave energy (orange triangles) and high wave energy (red triangles) sites. See S1 Table for GPS coordinates and additional site description. Also shown is a diagrammatic representation of coral reefs around Carriacou and surrounding islands, after [37]. https://doi.org/10.1371/journal.pone.0306897.g001 PLOS ONE Seagrass bed biodiversity and structure and across 5 decades PLOS ONE | https://doi.org/10.1371/journal.pone.0306897 August 1, 2024 4 / 25

was adjusted to mean low water (MLW). The presence of different seagrass species, and of sea- grass-free bottom, were recorded at points (distances) along transects where changes in sea- grass composition occurred. The recorded substratum cover type was deemed constant from each observation point until the point where there was a change in type. Total cover per sub- stratum type was calculated as the sum of the distances (in m) represented by that type along a transect to 50 m offshore, common to all transects (and thus restricted to nearshore areas only). Cover values were converted to percent of the total distance (50 m) for analysis. We counted erosional scarps occurring on the inshore margin of seagrass beds and associ- ated with “blowouts” within the seagrass beds [35] along transects, including those beyond 50 m. Only actively eroding scarps (i.e. those with underlying seagrass bed sediments exposed) were counted; older, overgrown scarps were not included in these counts. We also measured the extent of the seagrass-free inshore zone (SFI) as the distance along a transect (in m) from the shore at MLW to the first occurrence of seagrass. This zone often was the subtidal exten- sion of a sandy beach; in some areas it contained scattered rubble, boulders or sparse macroal- gae but no coherent seagrass patch or bed. We limited our analysis of these erosional features to 80 m offshore, common to all transects. In 1994, due to logistical constraints, the full set of transect measurements was limited to 4

of the 10 transects (T6, T7, T9, T14); scarps however were counted on all 10 transects. In 2016, the full set of transect measurements was repeated on 9 the 10 transects. The remaining tran- sect (T9) was incorrectly positioned in 2016; the correct site was sampled in February 2019 and the observations included with the 2016 data. Based on observed wave conditions in the field, we classified 6 transects (T1, T5 –T9) as HWE (High Wave Energy) sites, and 4 transects T2, T3, T10 and T14 as LWE (Low Wave Energy) sites in 1969. Relative Exposure Index was calculated after [42, 43] for the shore posi- tion on each of the 10 transects in 2016 (2019 for T9) based on local wind data [44] and limit- ing ocean fetch to 200 km [45]. There was an order of magnitude difference in mean REI between HWE sites (T1, T5 –T9: 5327) and LWE sites (T2, T3, T10, T14: 421) in 2016 (S1 Fig), consistent with our qualitative classification. The minimum REI recorded at 1 HWE site (T1) was equivalent to the maximum recorded at 1 LWE Site (T2). However, our observations of wave conditions at T1 indicated it was a high wave energy site, which was not captured in the REI due to the reduction in fetch from surrounding islands. This HWE designation was sup- ported by the occurrence of coarser, better sorted sediments at T1 than T2 based on samples in 1969 [35]. Based on studies of reef sediment transport and distribution off the east coast of Carriacou, Watering Bay (where T2 is located) has been characterized as a low wave energy environment [46, 47]. Stations

In addition to line transect sites, in 1969, 24 sampling stations (10 x 10 m grids) were located at sites chosen to be representative of seagrass beds around Carriacou, as viewed in the aerial photographs (e.g., fringing bed, offshore patch, lagoonal patch, cobble bank). Of these 24 sta- tions, 17 were resampled in both 1994 and 2016 and form the basis of our spatial-temporal comparisons (Fig 1, S1 Table). At each station, 12 quadrats (0.25 m2) were randomly posi- tioned within the sampling grid. Presence of all readily visible macroflora (seagrasses, sea- weeds) and macrofauna (sponges, corals, echinoderms, conch) was recorded in each quadrat. In 1969 and 1994, macroalgal specimens that were not clearly identifiable in the field were col- lected for subsequent description and identification. Macroalgae were identified by reference to Taylor [48]. A collection of 14 sponges in 1969 was identified by Prof. George Hechtel (State University of New York) in 1971. In 2016, photographs of all species were taken with an PLOS ONE Seagrass bed biodiversity and structure and across 5 decades PLOS ONE | https://doi.org/10.1371/journal.pone.0306897 August 1, 2024 5 / 25

Olympus TG4 camera and placed on iNaturalist (www.inaturalist.org/projects/seagrass-bed- flora-fauna-of-carriacou-grenada-2016). In the final compilation of species occurrence for comparison between the 3 sampling years, 69 species or “species groups” could be reliably compared between the 3 sampling years. These include 3 seagrasses, 46 macroalgae, 2 echino- derms, 1 gastropod, 6 corals, 7 sponges and 4 gorgonians. Of these, 56 were identified to the species level. The remaining 13 were classified into “species groups”, which included closely related species that were not reliably or consistently distinguished in the field or for which the taxonomic distinctions are not well defined (e.g., the Porites furcata group refers to P. furcata and/or P. divaricata; Tripneustes ventricosus group refers to T. ventricosus and Lytechinus variegatus). In 1969, 1994 and 2016, samples of seagrass (cut at substrate level) were collected from 3 haphazardly placed 0.0625 m2 quadrats within each 10 x 10 m sampling station. Seagrasses were sorted by species, shaken to remove excess water, and weighed fresh as a measure of above ground biomass. For T. testudinum, leaf width was measured (0.5 mm precision) for 35 haphazardly sampled leaves. Maximum leaf length was calculated as L10, the average length (0.5 cm precision) of the longest 10 leaves for samples of 200 leaves and greater, or as L5%, the average length of the 5% of leaves for samples with less than 200 leaves [49]. In 1994 and 2016, we also counted the number of leaves of T. testudinum in our biomass samples, and the num-

ber of leaves per shoot for 25 shoots sampled in situ at each station. Shoot density was esti- mated by dividing the total number of leaves per unit area by the average number of leaves per shoot. We used a significant linear relationship (F1,30 = 179, P < 0.001) between mass per leaf and the product of maximum length and average width, based on our 1994 and 2016 samples (when all leaves were measured and counted), to estimate leaf number for the 1969 samples from the 1969 measures of maximum leaf length, average leaf width and total fresh weight. We used the average number of leaves per shoot for sites sampled in 1994 to estimate shoot densi- ties in 1969. This assumes that the number of leaves per shoot is similar between years, consis- tent with our observations in 1994 and 2016 (see Results). Ethics statement Aside from sponge samples collected for species identification in 1969, no other physical sam- ples of macrofauna were taken and no permits were required. Throughout the study, sites were accessed mainly by sea and not on privately owned or protected land. No protected spe- cies were sampled, and there was no field or laboratory experimentation. Statistical analyses Differences in frequency of occurrence and biomass of seagrass species across years were com- pared using 1-way ANOVA for each species and response variable. Shoot density and leaf length and width for T. testudinum were compared using 1-way ANOVA for each response variable. Differences in scarp number across wave exposure and year was compared using a

2-way fixed effects ANOVA (car package) with type III sums of squares for unbalanced design. We examined patterns of community diversity using community characteristics (species richness, Shannon diversity) calculated for each station from frequency of occurrence data (proportional frequency over the 12 quadrats) restricted to the 66 species that formed the ses- sile benthic assemblage; mobile species (Tripneustes ventricosus group, Oreaster reticulatus, and Aliger gigas) were excluded. Differences in these response variables across years were tested using 1-way ANOVA. Assumptions of homogeneity of residual variance were evaluated using Levene’s test and inspection of plotted residuals. Where we detected significant main effects, posthoc tests were performed using Tukey’s HSD. PLOS ONE Seagrass bed biodiversity and structure and across 5 decades PLOS ONE | https://doi.org/10.1371/journal.pone.0306897 August 1, 2024 6 / 25

We examined the assemblage structure for all three years and all stations using multivariate analysis. Difference between years were analyzed using a one factor permutational multivariate analysis of variance (PERMANOVA) based on a Bray–Curtis similarity matrix generated from arcsine sqrt transformed data [50]. Year was a fixed factor and analysis used type III sums of squares. Non-metric multidimensional scaling (nMDS) ordination, based on the same Bray– Curtis similarity matrix, was used to examine multivariate patterns in the composition of the benthic assemblage among years and stations. Spearman’s rank correlation was used to exam- ine relationships between ordination scores and the frequency of occurrence of main func- tional groups (green algae, red algae, brown algae, corals, and seagrass). To test the hypothesis that variation in the sessile benthic community was related to variation in sea urchin abun- dance, we also used Spearman’s rank correlation for the ordination scores and examined the frequency of occurrence of Tripneustes ventricosus at each station. All multivariate procedures reported here were performed using Primer 7 with the PERMANOVA add-on [51]. Results Seagrass cover on transects Total cover by seagrass along 50-m transects (Fig 2) was higher at LWE sites (mean: 69%) than at HWE sites (48.2%) in 1969. By 2016, seagrass cover had declined more than 2-fold at HWE sites (21.4%), but there was little change at LWE sites (64.7%). An extreme shift occurred at one HWE site (T5) where the elevated, sand and pebble-based seagrass bed (83% cover) that

existed in 1969 had been completely scoured down to bedrock by 2016, with only a thin cover of shifting, coarse sand and T. testudinum persisting (21% cover) in sand-filled crevices. 2-way ANOVA showed a significant effect of wave exposure on total seagrass cover within transects (F1,16 = 11.5, P = 0.004) but no effect of year (F1,16 = 3.9, P = 0.065) or interaction between year and wave exposure (F1,16 = 2.7, P = 0.118). A similar result was obtained for total cover of a transect by the seagrass bed per se (which includes seagrass-free patches within beds). Given the unbalanced design, high variability between transects at HWE sites and the marginal test result for the effect of year, we conducted a paired-sample t-test to compare cover of the fring- ing seagrass bed between years (1969, 2016) within each wave exposure level. This showed a highly significant effect of Year (higher in 1969) at HWE sites (P = 0.003), but no effect at LWE sites (P = 0.771). There were no differences in cover of seagrass-free patches within sea- grass beds (Fig 2) across years or wave exposures (2-way ANOVA, p > 0.47). This indicates the pronounced decrease in the extent of the bed and the total cover of seagrass at HWE sites over time is related to the expansion of the seagrass-free inshore zone (SFI), which was not evi- dent at LWE sites. Mean cover by monospecific T. testudinum (Fig 2) changed very little at LWE sites between 1969 (33.8%) and 2016 (34.2%) but declined by 4-fold at HWE sites (from 19.5% to 4.3%)

although this difference was not statistically significant (paired samples t test, P = 0.132). In 2016 the invasive seagrass H. stipulacea was observed within 3 of the 4 LWE transects (Fig 2); at the remaining transect (T2) H. stipulacea occurred beyond our 50-m limit for analysis (at 93–100 m over the interval 0–180 m). H. stipulacea was absent from the 6 HWE transects. Sea- grass cover in pure stands and mixtures was highly variable between transects (Fig 2). Seagrass abundance and morphology at stations Quadrat observations at 17 stations in 1969, 1994 and 2016 provide indicators of changing conditions within contiguous seagrass beds (Fig 1). Ten of the stations (S2, S7, S8, S9, S10, S13, S16, S19, S23) were located within fringing seagrass beds and 6 stations (S3, S5, S6, S14, S24) were in ‘patches’ or ‘streaks’ or ‘cobble banks’ seaward of fringing beds. The remaining station PLOS ONE Seagrass bed biodiversity and structure and across 5 decades PLOS ONE | https://doi.org/10.1371/journal.pone.0306897 August 1, 2024 7 / 25

PLOS ONE Seagrass bed biodiversity and structure and across 5 decades PLOS ONE | https://doi.org/10.1371/journal.pone.0306897 August 1, 2024 8 / 25

forme respectively. However, these changes in biomass are not statistically significant for either species (1-way ANOVA, square root transformation: T. testudinum, F2,48 = 1.63, P = 0.206; S. filiforme, F2,48 = 0.863, P = 0.448). Fig 2. Substratum cover in seagrass beds. Mean (+ SE) % cover of transects in 6 HWE sites 4 LWE sites in 1969 and 2016, and 3 HWE sites and 1 LWE site in 1994 (top row). Cover on individual transects at HWE and LWE sites in each year (subsequent 3 rows). Height of bars indicates total cover of seagrass bed; remainder is cover of seagrass-free inshore zone (SFI). The large component “Other” in 2016 consists mainly of a mixture of Thalassia testudinum, Syringodium filiforme, and Halodule wrightii on Transect T9. https://doi.org/10.1371/journal.pone.0306897.g002 Fig 3. Abundance of seagrass species/groups at stations (n = 17) in 1969, 1994 and 2016. Mean (+ SE) of a) frequency of occurrence (%) and b) biomass (kg m-2) of Thalassia testudinum, Syringodium filiforme and Halophila stipulacea. (Halodule wrightii is not included as it occurred at very low frequencies and biomass at stations). https://doi.org/10.1371/journal.pone.0306897.g003 PLOS ONE Seagrass bed biodiversity and structure and across 5 decades PLOS ONE | https://doi.org/10.1371/journal.pone.0306897 August 1, 2024 9 / 25

Mean shoot density of T. testudinum declined from 572 to 348 shoots m-2 between 1969 and 2016 (Fig 4A), a significant decrease to 61% of the 1969 value (1-way ANOVA, F2,48 = 3.7, P = 0.033; Tukey’s HSD test: P = 0.031). Mean maximum leaf length (range: 21–25 cm, (Fig 4B) did not vary significantly among years (F2,48 = 1.7, P = 0.201) and mean width (9.4– 9.5 mm, Fig 4C) was effectively constant (F2,48 = 0.132, P = 0.877). Mean number of leaves per shoot also varied little between 1994 (2.8) and 2016 (2.7); leaf number was not measured in 1969. Therefore, the possible decline in biomass of T. testudinum during our study likely was related to a decrease in shoot density, rather than changes in leaf number or dimensions. Macroalgae and macrofauna at stations The frequency of abundance of the more common sessile species and species groups, defined as those occurring at a mean frequency of � 0.75 per 12 quadrats (6.25%) in at least one of the 17 stations in at least one year, are shown in Fig 5 (a complete list of species including the less common forms is given in S2 Table). Among macroalgae, 19 of the 25 species/groups were more frequent in 1994 (grand mean frequency, 16.5%) than in 1969 (5.0%) or 2016 (2.9%). Only the red alga Amphiroa tribulus, the green alga Halimeda simulans, and 4 of 6 Caulerpa spp. did not conform to this pattern. The most common Caulerpa species was C. prolifera, which occurred at 8, 8 and 5 stations in 1969, 1994 and 2016 respectively. Caulerpa scapellifor- mis, an invasive Indo-Pacific species, was absent at our stations in 1969, occurred at 1 station

in 1994, and at 6 stations (5 with C. prolifera) in 2016. The 2 more common corals, Manacina areolata and the Porites furcata group (P. furcata / P. divaricata), also were most frequent in 1994. There was no consistent trend with time for sponges, which as a group were common in all years. Of the 3 mobile macrofaunal species or species groups observed, only sea urchins (Tripneustes ventricosus group) were more common. Unlike the corals and most macroalgae, sea urchins were less frequent in 1994 (mean fre- quency, 0.01%) compared to 1969 (10.3%) or 2016 (11.8%). The other 2 mobile species were recorded only at low frequency in quadrats and in only 1 of the 3 sampling years (Oreaster reti- culatus: 1.4% in 2016, Alger gigas: 0.98% in 1969). Seagrass community comparisons Taxonomic richness and diversity (Shannon Index) of seagrass and associated macroalgae and sessile macrofauna differed significantly across the three sample years (F2,48 = 8.4, P <0.001 Fig 4. Shoot density and leaf dimensions of Thalassia testudinum at stations (n = 17) in 1969, 1994 and 2016. Mean (+SE) of a) shoot density (shoots m- 2), b) maximum leaf length (cm), and c) leaf width (cm). https://doi.org/10.1371/journal.pone.0306897.g004 PLOS ONE Seagrass bed biodiversity and structure and across 5 decades PLOS ONE | https://doi.org/10.1371/journal.pone.0306897 August 1, 2024 10 / 25

for richness; F2,48 = 7.3, P <0.002 for diversity) (Fig 6). In 1994, richness and diversity were significantly greater (by 2- and 1.4-fold respectively; Tukey HSD test, P <0.01) than in 1969 or 2016, when both measures did not differ (P > 0.96). The overall composition of the benthic community also differed significantly across years (PERMANOVA, P = 0.001) with a similar pattern of between-year differences: community composition was less similar between 1994 Fig 5. Frequency of occurrence of more common species of macroalgae and sessile macroinvertebrates, and of sea urchins, at stations (n = 17) in 1969, 1994, 2016. Mean (+ SE) frequency (%) in 12 quadrats per station. Green algae: Avrainvillea nigricans (AvNi*), Avrainvillea rawsonii (AvRa*), Dictyosphaeria cavernosa (DiCa*), Halimeda incrassata (HaInc), Halimeda opuntia (HaOpu), Udotea cyathiformis (UdCya), Udotea flabellum (UdFla), Valonia ventricosa (VaVe*); brown algae: Canistrocarpus cervicornis (CaCer), Dictyopteris delicatula (DiDel), Dictyopteris justii (DiJu*), Dictyota ciliolata (DiCil), Dictyota dichotoma (DiDic), Dictyota implexa (DiImp), Dictyota mertensii (DiMer); Caulerpa: Caulerpa cupressoides (CaCup), Caulerpa cupressoides var. lycopodium (CaLyc), Caulerpa mexicana (CaMex), Caulerpa prolifera (CaPro), Caulerpa scalpelliformis (CaSca), Caulerpa sertularioides (CaSer); red algae: Amphiroa fragilissima (AmFra), Amphiroa rigida (AmRig), Amphiroa tribulus (AmTri), Hypnea cervicornis (HyCer), coral: Manicina areolata (MaAre), Porites Furcata

(PoFu*); sponges: Amphimedon erina (AmEri), Cliona varians (ClVar), Halichondria melanodocia (HaMe*), Hyrtios violaceus (HyVio), Tedania ignis (TeIgn); sea urchins: Tripneustes ventricosus (TrVe*). https://doi.org/10.1371/journal.pone.0306897.g005 PLOS ONE Seagrass bed biodiversity and structure and across 5 decades PLOS ONE | https://doi.org/10.1371/journal.pone.0306897 August 1, 2024 11 / 25

and 1969 or 2016, than between 1969 and 2016 (S3 Table). Ordination analysis (non-metric MDS) shows a broad cluster of stations in 1994 characterized by the absence (except at one site) of sea urchins (Tripneustes ventricosus group), in contrast to stations in both 1969 and 2016 where sea urchins occurred at 7 out of 17 stations, generally at moderate to high frequen- cies (Fig 7). Correlations between ordinations for benthic assemblages and frequencies of occurrence for functional groups were highest for seagrasses followed by the various macroal- gal groups, and lowest for sponges. Extent of the seagrass-free inshore zone and erosional scarps on transects The extent of the seagrass-free inshore zone (SFI) increased at all 6 HWE sites between 1969 and 2016, with a more than 2-fold increase in the average (from 16.9 to 35.9 m). In 2 of the 3 HWE transects where the SFI was measured in 1994 and SFI increased between 1969 and 2016 (T6, T7), most of this increase occurred between 1994 and 2016 (Fig 8, S2 Fig); at T9, SFI decreased between 1994 and 2016. At LWE sites, changes in SFI were generally smaller and more variable between 1969 and 2016: SFI increased measurably at one site (T2, from 0 to 14 m), decreased at another (T14, from 27 to 17 m), and was effectively stable (~ 4–5 m) at the remaining 2 sites (T3, T10), resulting in a minimal change on average (from 8.6 to 10.2 m). 2-way ANOVA showed a significant effect of year (greater in 2016) on SFI (F1,16 = 8.0, P = 0.012) but no effect of wave exposure (F1,16 = 1.2, P = 0.289) or interaction between year

and wave exposure (F1,16 = 2.7, P = 0.119). Given the unbalanced design and high variability between transects at both LWE and HWE sites, we conducted Welch 2-sample t-tests to com- pare SFI between wave exposure levels in 1969 and 2016. This showed a highly significant effect of wave exposure on SFI in 2016 (t7.2 = -3.53, P = 0.009) but no effect in 1969 (t4.8 = -1.2, P = 0.287). No scarps were observed on transects at LWE sites in 1969 and 1994 but 1–2 scarps were present at 3 of the 4 sites in 2016. In 2016 scarps were present at on all HWE transects, except T5 where the seagrass bed had been scoured down to the underlying bed rock (Fig 9, S2 Fig). Fig 6. Taxon richness and diversity of the seagrass bed community. Mean + SE richness and diversity (Shannon Index) at stations (n = 17) in 1969, 1994, 2016. https://doi.org/10.1371/journal.pone.0306897.g006 PLOS ONE Seagrass bed biodiversity and structure and across 5 decades PLOS ONE | https://doi.org/10.1371/journal.pone.0306897 August 1, 2024 12 / 25

On average the number of scarps had increased 1.6-fold at HWE sites (from 2.8 to 4.6) between 1969 and 2016 (we excluded T5 from analysis as there was no possibility for scarps to form by 2016). At 3 HWE sites (T1, T6, T7), scarp number increased 3-fold between 1969 and 2016, with most or all of the increase occurring between 1994 and 2016. At the other 2 sites (T8, T9), a small decline in scarp number occurred between 1994 and 2016/2019. 2-way ANOVA showed a significant effect of wave energy (F1,21 = 51.1, P <0.001) and year (F2,21 = 5.3, P = 0.014) on scarp number, but no interaction between wave energy and year (F2,21 = 0.36, P = 0.702). Tukey’s HSD post hoc test showed a significant increase in scarp number between 1969 and 2016 (P = 0.019) and 1994 and 2016 (p = 0.042), but not between 1969 and 1994 (P = 0.927). Fig 7. Non-metric MDS of seagrass bed community at each station and year. Points are arcsine transformed proportional frequency of occurrence measures using Bray Curtis dissimilarity matrix of square-root transformed data. Vectors (black) show correlations between ordinations for benthic communities and average frequencies of occurrence for functional groups (seagrass, brown algae, green algae and sponges) at each station and year. The size of each point indicates the frequency of sea urchins at each station and year. https://doi.org/10.1371/journal.pone.0306897.g007 PLOS ONE Seagrass bed biodiversity and structure and across 5 decades PLOS ONE | https://doi.org/10.1371/journal.pone.0306897 August 1, 2024 13 / 25

Discussion Temporal and spatial context of this study Longitudinal (multidecadal) studies of seagrass beds at individual sites in the Tropical Atlantic Bioregion have mapped changes in relative cover or extent of these beds, typically by aerial photography [22, 23, 52]. Our island-wide study, spanning 5 decades, to our knowledge, is the longest of those that record taxonomically detailed, replicated measures of abundance of sea- grass and associated macroalgae and sessile invertebrates (percent cover, frequency of occur- rence, biomass, density) as well as concurrent measures of seabed erosion (recession of inner bed margins, blowout frequency within beds) related to hydrodynamic forces. Although the temporal resolution is minimal, our study encompasses a period that spans a relatively undis- turbed state in 1969 to an increasingly perturbed state in 2016, with a midpoint in 1994 that appears to capture a period during which biodiversity is highest but erosional features— including those indirectly related to biogenic changes such structural degradation of barrier reefs due to coral mortality—are in transition from relative stasis to an increase in magnitude. Fig 8. Offshore extent of the seagrass-free inshore zone (SFI, m) on transects in 1969, 1994 and 2016. Mean (± SE) extent of SFI for 6 HWE sites and 4 LWE sites in 1969 and 2016, and 3 HWE sites and 1 LWE site in 1994. Also shown (right column) is cover on replicate transects at HWE and LWE sites in each year. https://doi.org/10.1371/journal.pone.0306897.g008 PLOS ONE

Seagrass bed biodiversity and structure and across 5 decades PLOS ONE | https://doi.org/10.1371/journal.pone.0306897 August 1, 2024 14 / 25

Importantly, our study spans a period of marked broad-scale environmental change in the Tropical Atlantic Bioregion, including increased sea temperature and frequency of heat waves [53], increased sea level and significant wave height [54, 55], and increased frequency and severity of strong storms [56], in addition to localized deterioration of the coastal seagrass hab- itat associated with eutrophication and reduced water transparency related to land based activ- ities [21, 22, 57], dieback of T. testudinum associated with drought and hypersaline conditions [58], and direct physical disturbance by recreational or commercial activities [59]. Carriacou is a relatively small island (34 km-2, population ~10,000) that remains largely free of many of the localized anthropogenic factors that have caused degradation and loss of sea- grass beds elsewhere in the Caribbean. It has retained a small-scale livelihood fishery (e.g., fish traps, spearfishing), in contrast to the more intensive fishing of reefs and lagoons recorded elsewhere in the Caribbean by 1985 [60]. There are no large freshwater inputs or restricted bays in Carriacou that might cause changes in seagrass beds associated with salinity fluctua- tions and sediment in runoff. Blooms of free-floating Sargassum seaweed, which had begun to impact coastal waters in Florida and much of the Caribbean after 2012 [61], had not been observed in Carriacou by end of our study in 2019. Dredging and infilling associated with

Fig 9. Number of scarps on transects in 1969, 1994 and 2016. Mean (± SE) scarp number for 6 HWE sites and 4 LWE sites in 1969 and 2016, and 3 HWE sites and 1 LWE site in 1994. Also shown is scarp number on replicate transects at HWE and LWE sites in each year. https://doi.org/10.1371/journal.pone.0306897.g009 PLOS ONE Seagrass bed biodiversity and structure and across 5 decades PLOS ONE | https://doi.org/10.1371/journal.pone.0306897 August 1, 2024 15 / 25

construction of a large marina in Tyrell Bay in 2015–2018, likely resulted in water quality issues which, along with increased shipping and motor vessel activity, likely impacted the sea- grass community of the bay and contributed to establishment of the invasive seagrass Halo- phila stipulacea [36]. However, Tyrell Bay, the most protected bay on the island, was not included in our multi-year comparisons. The changes in seagrass beds that we observed more likely are related to broader scale phenomena that affect seagrass directly or indirectly. Seagrass cover, composition, and morphology There were significant losses in nearshore seagrass cover between 1969 and 2016 at HWE sites, but not LWE sites. These losses were associated with an increase of the inshore area devoid of seagrass (SFI) which doubled in average extent in transects at HWE sites. The power to detect differences in species composition of seagrass in the nearshore between years and between HWE and LWE sites was low because of high variability in composition between transects at both HWE and LWE sites and the relatively low number of sites in an unbalanced sampling design. The only clearly defined change was the appearance of the invasive seagrass H. stipula- cea in 2016, exclusively at LWE sites. Based on a broader survey of H. stipulacea at Carriacou [36], which included our 2016 transect and station observations as well as those at additional sites, we found that H. stipulacea occurred primarily in large, sheltered bays along the leeward,

west coast of Carriacou where it formed extensive meadows. It also occurred in smaller patches at low wave energy sites in Watering Bay on the windward, east coast and in Manchioneal Bay on the south coast. Station records provided indicators of temporal change in species composition and abun- dance of seagrass, and in morphological characteristics of T. testudinum, within broadly dis- tributed seagrass beds. T. testudinum, was the dominant species in terms of both frequency of occurrence and biomass at our stations across all three sampling years. A gradual decline in average biomass of T. testudinum, and concomitant increase in biomass of S. filiforme (although not statistically significant for either species) suggests a shift towards earlier succes- sional stages in these beds [35, 62]. Similar successional shifts in other parts of the Caribbean have been attributed to environmental deterioration through localized anthropogenic impacts [55]. At Carriacou, however, increased intensity of wave-driven erosional processes, including increases in frequency of erosional scarps (blowouts) that re-initiate seagrass successional pro- cesses [35], may be the primary driver of such shifts. There was no change in leaf width or length of T. testudinum among sampling years, although there was a significant decrease in shoot density. This suggests that any decline in mean biomass was associated with a decrease in shoot density rather than changes in plant morphology or in the overall occurrence of T. testudinum at the 17 stations. Leaf length and

width of T. testudinum varies with environmental conditions locally and regionally, decreasing with reduced light [63, 64] or salinity [65]. In a study of T. testudinum at 3 sites across a sea- grass suitability gradient in the Gulf of Mexico, mean leaf width was 4 and 5 mm in the two lower suitability sites, and 8 mm in the highest suitability site [66]. By comparison, mean leaf width at our 17 stations in ranged from 9.4 to 9.6 mm over the 3 sampling years. The constancy of total seagrass biomass regardless of composition, and large leaf width of T. testudinum, sug- gests that high quality habitat for seagrass at Carriacou prevailed in all three sampling years. Benthic macroalgae and macrofauna in seagrass habitats In contrast to progressive changes in seagrass cover and species-specific biomass across the three sampling years, taxonomic richness and species diversity of the benthic community (sea- grass and associated macroalgae and sessile macrofauna) were greatest in 1994 compared to PLOS ONE Seagrass bed biodiversity and structure and across 5 decades PLOS ONE | https://doi.org/10.1371/journal.pone.0306897 August 1, 2024 16 / 25

1969 or 2016. Also, 94% of more common species/groupings of associated of macroalgae (excluding Caulerpa species) and both coral taxa were most frequent in 1994 compared to 1969 and 2016, and multivariate analysis showed that community composition was less similar between 1994 and 1969 or 2016 than between 1969 and 2016. Ordination analysis showed that these patterns in seagrass community composition across years may be related to an inverse pattern in the frequency of occurrence of sea urchins (Tripneustes ventricosus group), which was minimal in 1994 compared to 1969 and 2016. The importance of sea urchins as major grazers of seagrass and macroalgae is well established [67, 68] and some species are known to graze coral recruits [69–71]. Sea urchins were intensively harvested in Grenada (less so in Car- riacou) for export in the early 1990s, leading to the collapse and closure of the fishery by 1995 [72]. Disease or mass mortality from other causes, reported at other Tropical Atlantic locations in the 1990s [73, 74], may have contributed to the low abundance of sea urchins that we observed in 1994. Herbivorous fish also are important grazers of macroalgae in the seagrass beds [75] and increased fishing pressure [37] may have amplified the effect of decreased sea urchin abundance in seagrass beds. The frequency of occurrence of Caulerpa species did not follow the trend for other macroal- gal species: the highest frequency of occurrence of the more common 4 of 6 species of Caulerpa occurred in 1969 and 2016, when sea urchins were relatively common. Secondary metabolites

in Caulerpa species (notably caulerpenyne) that could limit urchin grazing [76] may account in part for this discrepancy from the general inverse pattern of macroalgal and sea urchin abundance across sampling years in our study. Caulerpa prolifera was the most common spe- cies, except in 2016 when C. scalpelliformis [77] was more frequent. C. scalpelliformis, an Indo- Pacific species first reported in the Caribbean at Barbados in 1966 [78] and Antigua in 1967 [79], was absent at Carriacou in 1969 and rare in 1994 but had expanded to over a third of the stations by 2016, including 5 stations where it co-occurred with C. prolifera. Erosional features and expansion of the inshore seagrass-free zone The greater extent of the seagrass-free inshore zone (SFI) and the higher frequency of erosional scarps and blowouts on transects across fringing seagrass beds at HWE sites compared to LWE sites are consistent with studies showing that, in the absence of significant anthropogenic disturbance, variation in structural features and cover by seagrass is related to differences in wave energy or current speed required to initiate sediment movement, combined with factors that affect erodibility of sediments [80–82]. At HWE sites, the average extent of the SFI more than doubled between 1969 and 2016 (an increase of 19 m) and for the 2 of 3 transects also measured in 1994 (T6, T7) most of the increase occurred between 1994 and 2016. As our transects began at the MLW mark, and not at a geographically fixed position, landward recession of the MLW contour due to erosion

could account, at least in part, for an increase in SFI. All HWE sites are on shorelines without any man-made structures (e.g., walls or dwellings) within 50 m of shore and where there has been no attempt to reduce shoreline erosion. Cambers [83] found that beach erosion on 5 vol- canic islands in the Eastern Caribbean averaged 0.65 m y-1 between 1985 and 2000. In Carria- cou, Fitzpatrick et al. [84] estimated an erosion rate of 1 m y-1 between 1999 and 2006 at a sandy beach in Grand Bay (between HWE transects T8 and T9) where sand mining contrib- uted to erosion. Assuming a conservative erosion rate of 0.5 m y-1 between 1969 and 2016, landward recession of the MLW contour at our HWE sites would be 23.5 m, suggesting it could account for most or all of the increase in SFI, as opposed to net loss of seagrass at the inshore margin of beds. Average landward recession of the MLW contour associated with sea level rise alone from 1969 to 2016 (9.6 cm, based on a mean of 2.04 cm y-1 for 1950–2010 [85]) PLOS ONE Seagrass bed biodiversity and structure and across 5 decades PLOS ONE | https://doi.org/10.1371/journal.pone.0306897 August 1, 2024 17 / 25