Spawning events

The spawning events from which the suite of larval dispersal models were created from are published in Danylchuk et al. [44] and Lombardo et al. [49], which are the only detailed documented instances of bonefish spawning movements. Offshore spawning migrations from the PSA location in South Abaco were actively tracked with acoustic telemetry in November 2013, 2018, and 2019 [See 44, 49]. In 2013, five bonefish were gastrically tagged with Vemco continuous acoustic transmitters with pressure sensors rated to a depth of 50 m (V9P-2H, 9 mm diameter, 21 mm in length, 1.6 g in air, 2000 ms transmission period). In 2018 and 2019, six and four bonefish, respectfully, were surgically implanted with Vemco continuous pressure and temperature tags with a 250 m depth limit (V9TP, 9 mm diameter, 31 mm long, 4.9 h in air, period 1000 ms; Innovasea Systems Inc., Massachusetts). Not all tagged fish were successfully tracked offshore, as some fish emigrated, or were lost due to mortality or tag failure. The numbers of fish successfully tracked in 2013, 2018, and 2019 were one, three, and one, respectfully, and in 2019 the entire aggregation was also observed by sonar during the offshore migration. Active acoustic telemetry allowed for continuous monitoring of the migration into offshore waters that exceed 1000 m, and the use of pressure sensor tags logged depth information. The spawning events were described in four-dimensions: longitude, latitude, depth, and time. In 2019, CTD casts were taken over the three days preceding the spawning event and immediately upon observation of spawning behavior using a Castaway CTD rated to 100 m (SonTek YSI, Xylem Inc., New York).

In 2013 and 2019, spawning rush behaviors—a rapid ascent rate greater than the average moving rate—and mixing movements—balling and swirling behaviors when spawning [56]—were observed in the depth sensor data, indicative of the spawning events. However, gamete release over the full duration of the ascent or solely at the apex has not been confirmed. In 2018, the tag signals were lost during the initial descent in rough seas. Observations from the 2019 spawning event provided the most complete and detailed information about bonefish spawning ecology. The spawning event in 2019 revealed that bonefish spawning depth is likely associated with the pycnocline and thermocline boundary layer, as evidenced by the ejection of the acoustic telemetry tag [49]. Assuming an obligate relationship, thermocline depth was used to inform the parameterization of the 2013 and 2018 spawning events where the extent of the spawning rush below 50 m and the movement beyond the initial descent, respectively, were not observed (See Larval Dispersal Model below).

Hydrodynamic model and grid

The hydrodynamic model chosen for this study is the Navy Coastal Ocean Model American Seas (NCOM AmSeas) [27] regional model from the Naval Oceanographic Office (NAVOCEANO). NCOM AmSeas is capable of resolving mesoscale and sub-mesoscale eddies, lending itself to applications in biophysical models [57–59]. The global NCOM model was developed by the Naval Research Laboratory and is based upon the Princeton Ocean Model with time invariant hybrid (sigma over z) vertical coordinates, allowing for higher resolution terrain following (sigma) and computational efficiency at deeper depths (z). NCOM AmSeas is a high-resolution derivative of global NCOM created in response to prediction and monitoring needs following the Deepwater Horizon incident [60]. The extent of the regional model covers the Gulf of Mexico and the Caribbean at 1/36° (~3.1 km) horizontal resolution and 40 depth levels. NCOM AmSeas ingests quality-controlled observations from the Navy Coupled Ocean Data Assimilation System (NCODA) [61], consisting of satellite sea surface temperature and altimetry as well as surface and profile temperature and salinity data. Boundary conditions are applied from operational global HYCOM (1/12° resolution) [62] and atmospheric forcings are provided by the Navy Global Environmental Model (NAVGEM) [63]. Tidal forcing is applied through two different methods. At the boundary, tides are imposed from the TPXO global inverse tidal prediction model [64]. Within the model domain, a tide-generating force is applied by incorporating astronomical forcing, ocean loading, and ocean self-attraction consistent with the TPXO model [see 65 for more detail]. NCOM AmSeas simulation archives consist of 3-hourly ocean state variables aggregated in NetCDF files including ocean temperature, salinity, sea surface height, east and northward flow velocity, and atmospheric forcing fields. Data were downloaded for each spawning season (October–April) and years (2013, 2018, 2019) observed with a spatial extent inclusive of potential dispersal to viable bonefish fisheries in The Bahamas, the northern extent of Cuba, and Florida and the Florida Keys.

All hydrography data transformations were done in MATLAB. To meet the hydrography data format and naming scheme requirements of Ichthyop, the NCOM AmSeas data were translated to a ROMS-like sigma coordinate grid. The bathymetry mask was created using the ROMS_TOOLS package [66] with bathymetry sourced from a subset of the ETOPO1 DEM [67], and clipped to only include depths greater than 1 m to eliminate land from the DEM. The shoreline mask was created from the Global Self-consistent, Hierarchical, High-resolution Geography database (GSHHG) [68]. Grid cells were masked when land comprised > 50% of the cell, single cell embayments, and single cell width edges along the domain boundaries. Spatial interpolation of the NCOM AmSeas data within the grid file was conducted using the ROMS_TOOLS package [66], which enforces mass conservation on the ROMS grid.

Larval dispersal model

We selected Icthyop [32] as our biophysical LDM program due to its frequent co-application with NCOM AmSeas data in 3-dimensional particle dispersal modeling within the Caribbean [59, 69, 70]. A 3 h computational timestep was used, matching the NCOM AmSeas output time interval. The computational timestep did not exceed the fine-scale limit of the Courant-Friedrichs-Lewy criterion, approximated by , where dt is the timestep, dGrid is the average length of the grid cells, and Umax is the order of magnitude of the fastest current velocities in the hydrodynamic model [24, 32, 71]. A Forward Euler advection scheme was implemented with a horizontal dispersion rate (ɛ) set following equations (2 & 3) of Pelíz et al. [72]. Okubo [73] equation (4) was used to calculate the explicit Lagrangian horizontal diffusion (K_h) for a 3.1 km grid. The horizontal dispersion is randomized at each timestep along a uniform distribution (δ ∈[−1,1]) [72], and the vertical dispersion is randomized using a vertical dispersion model [74] with cubic spline interpolation of the vertical diffusivity fields of the hydrodynamic model [32]. The randomized dispersion acts as a 3-D random-walk across multiple simulations, thus producing unique dispersal for each LDM iteration. To incorporate the stochastic variability of larval dispersal into the assessment of development environments, dispersal pathways, and connectivity, the LDM was run 100 times for each observed year, releasing 10000 larvae per iteration. The number of model iterations and larvae released were selected to minimize error in the potential or realized dispersal kernel, allow for representative sampling of the distributions governing the randomized directionality of the dispersion parameter enacted at each timestep, buffer against potentially significant mortality events (see Biology Parameterization), and to allow for within-year and across-year comparisons [24, 75, 76]. Furthermore, computational costs were considered when assessing the upper limit of how many larvae to release, as post hoc inspection would assess parameters governing every larva, at every time step, in each LDM iteration. Parameterizations for the simulation time, extent, stain (spawning event), transport, biology, and recruitment were set in the configuration XML file (see S1 File for 2019 model configuration XML).

Spawning event parameterization

The simulation start-time and location for each spawning event began at the time and location of the spawning observations in 2013 and 2019 or the last detection in 2018. Spawning events were observed on 17 November 2013 at 20:17 EST and 12 November 2019 at 06:00 EST, while on 20 November 2018 at 06:18 EST the final detections were received during the descent before being lost in rough seas. Spawning in 2013 and 2019 occurred west of the southern tip of Abaco, while in 2018 the fish were tracked south beyond the southern tip of Abaco (Fig 1). The simulation end-time of 71 d post-release coincided with the upper limit of the A. vulpes PLD and settlement competency period [53], allowing for the full spatial extent of potential dispersal. The range of spawning depths for each year were informed by observations in Lombardo et al. [49], where bonefish pelagic spawning habitat was coupled to the thermocline. In situ measurements of the water column were only taken by Lombardo et al. [49] in 2019. The in situ thermocline depth and the NCOM AmSeas thermocline depth were not equal, at 75.9 m and 50 m, respectively (See S1 Fig for NCOM AmSeas and CTD comparison). Therefore, the observed movement and spawning behaviors were translated into proportional depths in relation to the in situ thermocline. From these proportions, the spawning movements and behaviors could be related to the NCOM AmSeas thermocline depth so that the simulated spawning event depths would be at an equally proportional depth distance from the thermocline as the observed spawning event. In 2019, the upper observed spawning depth was 63.7 m, which was 84% of the thermocline depth (75.9 m). Therefore, the upper spawning depth for each simulated spawning event would be set at 84% of the NCOM AmSeas thermocline depth. The bottom extent of the spawning depth was determined by using the observed bottom extent of the 2019 spawn and the temperature data from the expelled acoustic tag, which functioned similar to an expendable CTD capable of returning data at depths greater than the depth rating of the Castaway CTD [See 49]. The bottom extent of the 2019 spawn coincided with a deviation from the consistent temperature-depth gradient, where the rate of cooling increased with depth (i.e., the maximum first-order derivative below the thermocline). Therefore, the bottom extent of spawning for each year was placed at the depth below the thermocline where the first-order derivative was maximized in the NCOM AmSeas data. These relationships resulted in simulated spawning depth ranges of 50.36–100 m, 50.36–125 m, 41.96–150 m for 2013, 2018, and 2019, respectively. Each spawning event released 10000 particles randomly distributed throughout the water column between the bottom and peak spawning depth with a horizontal radius of 150 m; equivalent to the 300 m detection range of the acoustic telemetry tags. Larval position, salinity, temperature, and vital status were recorded every 3 h.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Settlement zone polygons and A. vulpes spawning locations.

The domain of the hydrodynamic model was bounded within the gray box, inclusive of The Bahamas, Florida, and Cuba. Settlement zone polygons were classified per island group and cropped to the extent of the hydrodynamic model space. Island groups are color coded with settlement zones designated within the one to two grid cells (3.1–4.2 km) of the periphery, which has been extend from the shoreline of the land mask. Settlement zones shaded gray were not settled in any larval dispersal models. Inset are the putative spawning locations which correspond with the particle release locations. Prominent oceanographic features are labeled. Established protected areas are denoted with solid gray bars and teal outline, while proposed (2022) protected areas have broken gray bars and a violet outline. Protected areas were sourced from The Nature Conservancy Caribbean Protected Areas 05052022 (https://www.arcgis.com/home/item.html?id=6d15ba7e11a84278a13b40c8424db850t). The Esri “World Imagery” basemap layer and TNC Caribbean Protected Areas 05052022 layer are licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 License (http://creativecommons.org/licenses/by-nc-sa/3.0/).

https://doi.org/10.1371/journal.pone.0276528.g001

Biology parameterization

Little is known about bonefish leptocephalus biology and behavior throughout the pelagic duration of their development. While captively reared bonefish leptocephali displayed a vertical orientation and sporadic short burst swimming motions [77], these observations were only during the first < 10 days of the 71-day larval period, so volitional swimming behavior throughout development is not understood well enough to inform parameterization. Although burst and critical swimming velocities of other Elopomorph leptocephali have been experimentally measured and could be used in proxy, the ecology governing sustained swimming behavior throughout ontogeny is still not known [78]. Non-volitional movement may be imparted through specific gravity, or a particle’s relative density to seawater, which is likely to be important in leptocephalus dispersal as leptocephalus are uniquely buoyant [78, 79]. Larval density has not been measured for bonefish to parameterize buoyancy over development. However, density measurements over development have been taken for another Elopomorph, the Japanese eel Anguilla japonica [79], which we applied to bonefish by creating an equivalency between development time and the ratio of larval density to seawater density. The measured densities of individual A. japonica were converted to a proportion in relation to the density of water at the upper thermocline. Furthermore, the length of each individual A. japonica can be converted into a proportion of the time duration of larval development. Therefore, specific gravity measures for each pre-metamorphic larva by Tsukamoto et al. [79] were converted from a length:density ratio to , where L is the larva length, L_max is the maximum observed length of the pre-metamorphic larvae, D_l is the density of the larva at length L, and D_s is the seawater density at the thermocline. These proportions were matched to the larval duration of bonefish (See S2 File). Diel vertical migration (DVM) in bonefish is not well documented; however, most settlement competent larvae have been caught in the top 1 m of water when light traps were set in channels exceeding 4 m deep [53]. DVM was parameterized to maintain larval position within the top 2 m of water during the settlement competency period [41–71 days post spawn (DPS)]. The DVM programming is independent of spatial proximity to settlement zones, as there is no spatial component to the DVM parameters in Ichthyop. Lethal temperatures were parameterized with a lower bound at 14°C and upper bound at 32°C. Congener bonefish larvae have been captured in the Gulf of California at 15°C [80, 81] and in waters up to 30°C [21], while congener juveniles in the Indian River Lagoon, Florida have been observed in cold-snap mortality events at 14°C [82]. Transport behavior was parameterized to a beaching particle—the computational process of particle mortality due to contacting the shore before settlement competency. The biological basis for the beaching particle was not inclusive of larvae physically contacting the shore, as coastal hydrodynamic processes would deter such occurrences. Therefore, the beaching particle parameterization would be more appropriately described as onshore mortality. Onshore mortality encompasses mortality events driven by life stage-habitat mismatches (e.g., starvation, predation, inhospitable habitat) [83, 84], for which nearshore habitats inside of fringing reefs were presumed to be mismatched to pre-settlement competent bonefish larvae [53]. Additionally, the onshore mortality parameterization reduces the influence of unresolved nearshore processes over the full duration of dispersal that could not be captured at the grid resolution. Herein, the biological process of onshore mortality is synonymous with the computational process of the beaching particle parameterization.

Settlement zones parameterization

Bonefish nursery habitats are associated with littoral habitats [54]. Settlement zones (Fig 1)—generalized settlement and nursery habitat equivalents that result in successful settlement if reached within the settlement competency period (41–71 DPS)—were created using ArcMap 10.6.1 (ESRI, Redlands, CA). Grid centroids were extracted from the hydrodynamic grid file in MATLAB and imported into ArcMap using the Make NetCDF Feature Layer tool. Grid points for each settlement zone were selected and converted into polygons using a 500 m buffer around the GSHHG shoreline polyline shapefile [68]. As grid points were centroids and not vertices, vertices were manually adjusted for the settlement zones to extend one to two grid cells beyond the shoreline of the land mask. Settlement zones were extended from shore to account for the potential ability of larval bonefish to maintain their position until biological and environmental conditions are suitable for settlement [85]; an ability that could not be parameterized into the biology of the larvae. Furthermore, the extension of the settlement zones ameliorates possible deficiencies in the hydrodynamic resolving of fine-scale shorelines processes. Shoreline composition was not discriminated against, as larval behavior and ability to navigate sub-optimal settlement habitats is unknown. Settlement zones from the shapefile created in ArcMap were converted to an XML file compatible with Ichthyop using MATLAB.

Assessment of dispersal

Dispersal patterns and connectivity

Discrete gridded larval densities of all 100 LDM iterations for each of the observed spawning years throughout the full extent of the PLD reveal the diversity in larval dispersal pathways across years (Fig 2). In 2013, the eastern current flow through the Northeast Providence Channel forced most pre-settlement competent larvae into the east-facing southern shore of South Abaco, resulting in onshore mortality and leaving a highly reduced number of larvae to be dispersed. Remaining larvae were transported eastward into the Atlantic where most larvae were transported south. Limited larvae successfully dispersed onto Great Bahama Bank at the southern extent by transport through Exuma Sound, the Crooked Island Passage, and Mayaguana Passage. Anomalous dispersal towards Cuba was observed via the Crooked Island and Mayaguana Passages. Gyre formations north and northeast of Abaco retained relatively few larvae without indication of settlement feasibility. Larvae that left the domain did so in the southeast direction.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Larval density per grid cell at all time steps.

Larval density per grid cell (natural log scaled) at all time steps. Data were aggregated across all LDM iterations. Panels correspond to (A) 2013, (B) 2018, and (C) 2019. Land-sea mask used for visualization is extracted from the NGA Global Shoreline Dataset. (https://shoreline.noaa.gov/data/datasheets/pgs.html). This work is licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 License (http://creativecommons.org/licenses/by-nc-sa/3.0/).

https://doi.org/10.1371/journal.pone.0276528.g002

In 2018, larvae exhibited more diverse dispersal patterns with strong pathway consistency exhibited by high gridded densities. Again, pre-settlement competent larvae experienced onshore mortality along the east-facing southern shore of South Abaco, though magnitudes less than in 2013, but also along the south-facing shore of South Abaco, resulting in onshore mortality. Larvae that did not experience onshore mortality proximal to the spawning location were mostly retained by eddies within Northeast Providence Channel and the large ephemeral gyre within Northwest Providence Channel. Strong northern transport from both Providence Channels was observed, with a substantial abundance of larvae exiting the model domain to the north-northwest, most of which originated from Northeast Providence Channel. Increased abundances on the north side of Grand Bahama indicate the presence of a larvae retaining eddy feature. Lesser southern transport was observed into the Tongue of the Ocean (TOTO), and fewer into the Exuma Sound. Eddy features retained larvae at both the northern and southern extents of the dispersal region, though it did not appear that larvae returned to the Bahamian Archipelago once circulated beyond its extent. Westward dispersal extended out to Bimini and was limited by the Gulf Stream, while eastward dispersal was controlled by Antilles Current dynamics. Few larvae were dispersed to the northern shore of Cuba via TOTO and traversing over the shallows of Great Bahama Bank.

The 2019 dispersal patterns were most similar to the patterns seen in 2018, with strong retention within the Providence Channels and large numbers of pre-settlement competent larvae experiencing onshore mortality along the south-facing shore in South Abaco. The majority dispersal pattern interacted with the large ephemeral gyre positioned within Northwest Providence Channel, with larvae transported westward out of Northwest Providence Channel and north along the Gulf Stream, exiting the model domain to the north. Larvae exited the domain over a vast area bounded to the west by the Gulf Stream and uninhibited to the east. Larvae did disperse south into TOTO and south along the Antilles Current, though their southern extent was limited to the eastern nearshore region of Andros and to the eastern nearshore region of Long Island. The western extent of dispersal reached Bimini and continued further west beyond its shore, and larvae uniquely occupied the area between Bimini and Andros. Similar to 2018, abundances north of Grand Bahama indicate an established eddy. In contrast to 2018, transport northward out of Northeast Providence Channel was not as significant as that from Northwest Providence Channel. Larval transport out of the domain was also more diffuse, and a gyre northeast of Eleuthera reduced the strength of transport south along the eastern shore of Eleuthera into Exuma Sound.

The final position of all larvae across aggregated LDM iterations in 2013 shows the highest density of larvae were found along the southern shore of Abaco, with 95% of larvae within one grid cell (Fig 3) as a product of pre-settlement competent onshore mortality (Fig 4). Larvae dispersed eastward beyond the model domain’s 72.5°W limit, with most leaving the domain southeastward. No apparent retention features were visible given the absence of clustering. In 2018, 46% of larvae were found within one grid cell along the southern shore of Abaco as a result of pre-settlement competent onshore mortality. Many larvae remained within the Northeast Providence Channel after 71 d, with larvae retained by the eddies within Northeast Providence Channel and the ephemeral gyre within Northwest Providence Channel. Larvae continued to be projected from the Northeast Providence Channel eddies in a northward trajectory, and a cluster of larvae were positioned within the shallow waters of Great Bahama Bank north of Andros. The majority of larvae left the model domain through the northern boundary. In 2019, 14% of larvae were found within one grid cell along the south-facing shore of South Abaco, mostly as a result of pre-settlement competent onshore mortality. Similar to 2018, larvae were retained within the Northwest Providence Channel ephemeral gyre, though in much greater numbers. Larvae continued to be projected from the gyre northward along the Gulf Stream. Larvae were also similarly clustered north of Andros, but over a much larger area spanning 100 km across. Large numbers of larvae were diffusely distributed throughout the sea northeast of the Bahamian Archipelago, and larvae exited the model domain over the full expanse to the north and northeast. The final positions of larvae across years were mostly determined by pre-settlement competent onshore mortality events (Fig 4). All years experienced ≥ 50% pre-settlement competent onshore mortality within the first day, with 95% pre-settlement competent onshore mortality in 2013. No heat mortalities were experienced in any year and cold mortalities were observed in 2018 and 2019, accounting for 4.7% and 4.6% of respective mortalities, with the majority occurring at the initial interaction with northeast edge of the Northwest Providence Channel ephemeral gyre. Larvae began to leave the domain earlier in 2019 than in 2013 or 2018, with the earliest departures always driven by the Gulf Stream.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Larval density per grid cell at end of pelagic larval duration.

Larval density per grid cell (natural log scaled) at final time step of the 71 d dispersal period. Data were aggregated across all LDM iterations. Density curves for larvae per grid are on the top and right axes. Panels correspond to (A) 2013, (B) 2018, and (C) 2019. Land-sea mask used for visualization is extracted from the NGA Global Shoreline Dataset. (https://shoreline.noaa.gov/data/datasheets/pgs.html). This work is licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 License (http://creativecommons.org/licenses/by-nc-sa/3.0/).

https://doi.org/10.1371/journal.pone.0276528.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Mean proportion of survival and mortality fates throughout pelagic larval duration.

Mean proportions per day of assigned fates—alive, out of domain, dead beached, dead cold, and dead hot—throughout the full 71 d dispersal period. Proportions on each day are an average of the proportion of larvae (n = 10000) assigned a given fate across all LDM iterations (N = 100). The settlement competency period is denoted by the gray box.

https://doi.org/10.1371/journal.pone.0276528.g004

Larval settlement was successful for < 0.01% of larvae in the aggregated 2013 LDM iterations, with five island groups settled upon. Settlement KDEs indicate that the core settlement distribution was located at the southern extent of The Bahamas, at Acklins/Semana Cay (Fig 5). In the aggregated 2018 LDM iterations, 0.12% of larvae successfully settled upon nine island groups with core settlement distributions located at the north end of Andros and west end of Grand Bahama. In the aggregated 2019 LDM iterations, 0.38% of larvae successfully settled upon eight island groups with core settlement distributions located at the southern end of Abaco, the Berry Islands, and mid-longitude of Grand Bahama.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Kernel Density Estimates (KDE) for settled larvae.

Kernel Density Estimates (KDE) for settled larvae for aggregated LDM iterations. Larvae are colored by settlement days post spawn (DPS), and KDE density estimates (Level) are standardized to a maximum value within a bandwidth of one. Core settlement areas colored yellow. Panels correspond to (a) 2013, (b) 2018, and (c) 2019. Land-sea mask used for visualization is extracted from the NGA Global Shoreline Dataset. (https://shoreline.noaa.gov/data/datasheets/pgs.html). This work is licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 License (http://creativecommons.org/licenses/by-nc-sa/3.0/).

https://doi.org/10.1371/journal.pone.0276528.g005

Connectivity between Abaco and other islands varied across years, with the strongest connectivity showing island-level self-recruitment to Abaco and strong larval supply to Grand Bahama (Fig 6). The relationship between the number of larvae settled and the number of LDM iterations settled per island was positively monotonic. Connectivity in 2013 was especially poor given the high onshore mortality. The percent of 2013 LDM iterations which yielded successfully settled larvae was low, with larvae settling upon Acklins/Semana Cay most frequently at 14% of LDM iterations, though larval abundance was low. The most common 2013 LDM iteration settlement footprint—combinations of islands settled—occurred in 14% of the iterations and consisted of only Acklins/Semana Cay (S2A Fig). In 2013, no LDM iteration yielded more than 0.02% successfully settled larvae, which occurred in 4% of the LDM iterations, and the median percent of successfully settled larvae per LDM iteration was 0%.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Larval connectivity matrix.

Settled larvae (natural log scaled) per island over the full 71 d larval dispersal period. Count data were aggregated across all LDM iterations. The percent of LDM iterations that resulted in settled larvae are annotated.

https://doi.org/10.1371/journal.pone.0276528.g006

Connectivity in 2018 was substantially stronger, with more consistent and abundant larval sourcing. The percent of LDM iterations settled in 2018 was nearly 100%, with Abaco settled most frequently at 96% of LDM iterations. Successful settlement in 2018 decreased with latitudinal distance from Abaco, as did abundances, though Eleuthera and New Providence had much lower settlement success than Andros. However, this could be attributed to differences in shoreline extents. Islands south of New Providence showed connectivity of little importance. The most common settlement footprint, at 22% occurrence, was Abaco, Andros, and the Berry Islands. In 2018, the median percent of successfully settled larvae was 0.1%, which occurred in 5% of LDM iterations, while the minimum of 0.03% occurred in 3% of iterations and the maximum of 0.23% occurred in 1% of iterations.

The 2019 spawn supplied the highest abundances and consistent sourcing of larvae. Four islands—Abaco, Grand Bahama, the Berry Islands, and Eleuthera—had larvae settle in 100% of LDM iterations, with both Andros and New Providence above 50%. The most common settlement footprint occurred in 32% of LDM iterations, included Abaco, Grand Bahama, the Berry Islands, Eleuthera, Andros, and New Providence. In 2019, the median abundance of successfully settled larvae was 0.37% and occurred in 3% of LDM iterations, while the minimum of 0.17% successfully settled larvae occurred in 1% of iterations, and the maximum of 0.58% successfully settled larvae occurred in 2% of iterations.

Larval development environment

Larvae were grouped by year and island settled to quantify the variability in development environments experienced and to assess whether relative relationships are consistent across years. Due to the nearly complete onshore mortality event of larvae from the 2013 spawn, statistical comparisons were only possible between the 2018 and 2019 spawns. Across years, larvae that settled more proximal to the spawning location had lower coefficients of dispersal, indicating retention in circulation features for longer distances (Fig 7). Successfully settled larvae ranged in their coefficient of dispersal from < 0.0001 to 0.620, equivalent to dispersal paths that were >1000 times to approximately 1.5 times the straight-line distance from the spawning location to the settlement location. The median coefficients of dispersal across years were 0.47, 0.13, and 0.07, respectively, with median dispersal path distances of 986.1 km, 1050.5 km, and 1057.5 km. Island specific coefficient of dispersal distributions were similar in shape and range in 2018 and 2019, which is indicative of similar or prevailing hydrodynamic conditions. However, there were differences in the coefficient of dispersal distributions for Abaco and Grand Bahama between years, as the gyre and eddy structures near the spawning locations were different in position, size, and numbers (S2E and S2F Fig). Larvae exhibited within year island level differences in coefficients of dispersal (S1A and S1B Table), with effect size differences in 2018 smaller than those in 2019 (Fig 8). Of the total 43 contrasts made within 2018 and within 2019, 36 had large differences with the difference in means ≥ 0.8 of the pooled standard deviation (sd) per comparison. Relative relationships between islands were largely conserved between the 2018 and 2019 spawns, with 73% of relationships maintained, indicating stable dispersal pathways.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Larval development environment distributions.

Larval development environment distributions by year, settled island group, and environmental parameter. Data were aggregated across all LDM iterations. Panels are detailed left to right, and top to bottom. (1) Coefficient of dispersal distributions for settled larvae per island. Coefficients of Dispersal tending towards zero indicate a more circuitous dispersal path. (2) Distributions of settlement time (days post spawn–DPS) for settled larvae per island. (3) Mean temperature history distributions over the duration of dispersal for settled larvae per island. (4) Release depth distributions over the duration of dispersal for settled larvae per island. Data were aggregated across all LDM iterations. Top and bottom spawning extents are indicated by black horizontal lines, and the NCOM AmSeas thermocline is indicated by the horizontal hashed gray line. (5) Mean dispersal depth distributions over the duration of dispersal for settled larvae per island. Data were aggregated across all LDM iterations. Top and bottom spawning extents are indicated by black horizontal lines, and the NCOM AmSeas thermocline is indicated by the horizontal hashed gray line.

https://doi.org/10.1371/journal.pone.0276528.g007

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. Larval development environment effect sizes.

Effect size measures (Hedges’ g) for within year, paired settlement island comparisons of larval development environment characteristics. Effect sizes are classified according to Cohen [89] with exclusive bounds: Negligible (< 0.2), Small (< 0.5), Medium (< 0.8), Large (≥ 0.8). Comparisons are in relation to the first island listed, with negative values indicating less than. 95% confidence intervals are provided for each comparison.

https://doi.org/10.1371/journal.pone.0276528.g008

Settlement most frequently occurred within the first third of the settlement competency period for all years, with settlement occurring throughout the 41–71 d settlement competency period (Fig 7). Settlement dates in 2013 ranged from 29 Dec 2013 to 24 Jan 2014 (42.5–68.75 DPS), nearly spanning the full settlement competency period despite low settlement success. The overall median settlement date occurred on 4 Jan 2014 (48.13 DPS) and was the earliest among the three years. In 2018, the settlement dates ranged from 23 Dec 2018 to 22 Jan 2019 (41.38–71.12 DPS), with the median date of 30 Dec 2018 (48.38). In 2019, the settlement dates ranged from 23 Dec 2019 to 22 Jan 2019 (41.38–71.0 DPS), with the latest median settlement date of 2 Jan 2020 (51.38 DPS). Median settlement dates among islands within years varied between 22, 25, and 15 d, respectively. Islands proximal to Abaco tended towards earlier peaks in settlement, and islands extending below the southern extent of the Berry Islands at 25.4°N reached the 50^th percentile of their total LDM aggregated abundance later than those above this latitude. Larvae exhibited within year island level differences in settlement DPS (S1C and S1D Table), with effect size differences in 2018 smaller than those in 2019 (Fig 8). Large differences in effect size were only seen within the 2019 spawn, with 10 comparisons having a difference in means ≥ 0.8 of the pooled sd per comparison. The large differences in 2019 were mostly comprised of comparisons with Bimini, where larvae dispersed to Bimini settled earlier than other islands. Relative relationships between islands were mostly conserved between the 2018 and 2019 spawns, with 53% of relationships maintained.

Larvae were released randomly throughout the parameterized spawning depths for each year and successfully settled larvae were produced from nearly the full range of depths. Across all years, larvae were more likely to successfully settle if released at a deeper depth (Fig 7). However, successful settlement did diminish at the bottommost extent of the spawning depths, which was parameterized to coincide with the depth below the thermocline where the temperature-depth gradient grew. Very few larvae released above the thermocline successfully settled in any year. Island was not significantly associated with release depths within either year (S1E and S1F Table), and no comparative effect sizes exceeded a Hedges’ g of 0.8 (Fig 8). The mean release depth of larvae dispersed to each island within years was statistically equivalent, with only seven of the 43 contrasts having confidence intervals that did not overlap with zero, most of which were comparisons with larvae that settled upon Bimini.

Larvae successfully dispersed over a wide range of depths across years, apart from the few larvae that settled from the 2013 spawn (Fig 7). Most successfully settled larvae were dispersed at a mean depth at or near the thermocline, though having been released below the thermocline and often much deeper. The distribution of mean dispersal depths in 2018 was broader than the distribution for the 2019 spawn, in part due to the heavily skewed shallow mean dispersal depths for larvae settling upon Andros in 2018. While mean dispersal depths were statistically different in within year, across islands comparisons (S1G and S1H Table), the mean dispersal depths only differed in effect size in 2018 when comparisons were made with Andros (Fig 8). As such, mean dispersal depths exhibited stable relative relationships, in that mean dispersal depths did not differ.

Mean salinities experienced by larvae over their dispersal did not vary more than approximately 0.6 PSU. Minimum and maximum salinities were also examined, though they collectively varied by only 2 PSU. Such minute differences given the euryhaline classification of Albulid larvae [92] did not warrant further investigation of differences in salinity exposure during the dispersal process or potential implications. Mean temperatures experienced by successfully settled larvae ranged from 21.73°C to 26.40°C (Fig 7), the coldest minimum temperature approached within 0.02°C of the 14°C cold mortality limit, and the warmest temperature was 28.95°C. Across all years and all larvae, the average range in temperatures experienced was 9°C, with a minimum and maximum deviation of 3.5°C and 14.4°C, respectively. The coldest mean temperatures were experienced by larvae dispersed to Abaco and Grand Bahama, with larvae experiencing warmer waters when dispersed to more southern locations. Mean temperatures were similarly ranged in 2018 and 2019, with the medians approximately at 25°C. Within year island comparisons were statistically different (S1I and S1J Table), and 15 of 43 contrasts having large differences in effect size (Fig 8). Differences were larger in 2018, though 67% of relative relationships maintained their directionality despite smaller differences among islands.

Dispersal patterns and connectivity

Dispersal pathways.

Larvae that did not experience mortality shortly after hatching circulated up to 71 DPS, allowing for inspection of unique and prevailing hydrodynamic conditions, including those associated with increased settlement success. Dispersal patterns across the years can be classified into two paradigms: Northern dominant and Southern dominant (Fig 9). Northern dominant dispersal is supported as the prevailing dispersal pattern given the consistency between the 2018 and 2019 LDMs. Furthermore, the underlying hydrodynamics follow known prevailing seasonal circulation patterns [98], and regional LDMs from other studies and species show dispersal patterns consistent with the Northern dominant paradigm [55, 99].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 9. Dispersal paradigms.

Two larval dispersal paradigms were evident across the LDMs: the prevailing Northern dominant and the anomalous Southern dominant. Each dispersal paradigm is promoted by a suite of hydrodynamic patterns depicted by flow vectors derived from the NCOM AmSeas hydrodynamic model. Hydrodynamic features are identified within the white hashed boxes. NwPC—Northwest Providence Channel, NePC—Northeast Providence Channel, AC—Antilles Current.

https://doi.org/10.1371/journal.pone.0276528.g009

We did observe variability within the Northern dominant dispersal paradigm. In 2018, the presence of clustering mesoscale eddies within Northeast Providence Channel forced diffuse dispersal as portions of the larvae were divided among the periphery of a cyclonic eddy against the west facing shore of South Abaco and two anticyclonic eddies to the south and east of South Abaco. The anticyclonic eddy to the east subsequently passed a majority of the larvae to the north-flowing Antilles Current that flows along the eastern shores of Abaco, which was the main source of larval dispersal out of the Bahamian Archipelago and the model domain. Larvae within the southern anticyclonic eddy were moved south into TOTO and dispersed across New Providence and eastern Andros. The cyclonic eddy pushed larvae into the Northwest Providence Channel ephemeral gyre from the south, as the eddy and south flowing currents prevented larvae from dispersing westward immediately post spawn like what was observed in 2019. The 2019 spawn yielded the most successfully settled larvae and had the smaller of the two settlement footprints. A majority of the larvae were immediately drawn into the Northwest Providence Channel ephemeral gyre, mediated by the presence of an anticyclonic eddy positioned more proximally to the Little Bahama Bank between Abaco and Grand Bahama. Undisrupted inclusion into the Northwest Providence Channel gyre generated a slow and consistent dispersal of larvae to islands that interface with both Providence Channels. This also passed larvae westward into the Gulf Stream where larvae were lost from the northern domain.

The presence of the ephemeral gyre within Northwest Providence Channel undoubtedly is the most important circulation feature in determining the larval dispersal dynamics of bonefish sourced from the South Abaco spawn. Its presence provides a prevailing NW dispersal distribution and promotes larval retention within the northern extent of the Bahamian Archipelago. The regularity in which the gyre forms over winter [98] provides another predictable hydrodynamic feature in addition to the predictably stable thermocline that bonefish can rely upon to maximize larval survival [100]. Predictable and prevailing environments are central to the migratory spawning strategy, as significant energy investments are made prior to undergoing migrations to distant locations where conditions are unknown [11, 101].

The Southern dominant dispersal paradigm appears to be anomalous, only strictly occurring in 2013. Successful settlement was low given this paradigm, providing more support for the Northern dominant dispersal paradigm as the prevailing condition. However, the difference among the two paradigms might not be as large as presented here if not for the near total onshore mortality event in 2013. Conversely, there is the potential for complete mortality in the absence of prey availability if sufficient prey densities for larvae are reliant upon the presence of aggregating features, such as gyres and eddies [102], that were notably absent in the 2013 dispersal paths. The Southern dominant dispersal paradigm is characterized by an interaction with the Antilles Current, in particular the deep waters and counter current. The surface currents of the Antilles Current flow in a northward direction, while the counter current and deep waters flow south [98]. Evidence of a sustained southward flow along the eastern boundary of the Bahamian Archipelago can been seen across all years, with larval deposition varying between Eleuthera and Acklins. Strict adherence to the Southern dominant dispersal paradigm is anomalous, but the underlying currents do appear to be regularly involved in shaping the full dispersal footprint of the Northern dominant dispersal paradigm as well given the presence of southern dispersal along the eastern margin of the Bahamian Archipelago in 2018 and 2019.

These LDMs were uniquely constructed to incorporate new empirical information to simulate outcomes from observed spawning events. However, this effort was not the first to examine bonefish larval dispersal through LDMs, as Zeng et al. [55] did so to estimate regional connectivity of the Caribbean bonefish metapopulation. Comparisons between the two model structures and time periods allows for a more longitudinal assessment of the dispersal pathways, though parameterizations and metrics of success do vary considerably (Table 1). Zeng et al. [55] summarized dispersal as the final position for larvae at day 53, which is comparable to this study (Figs 2 and 3), as their final positions reported are not truly settlement. The density of final positions appears to capture the overall variability in dispersal observed in the LDMs within this study, with dispersal throughout the entire Bahamian Archipelago and Cuba; however, vagrants from the Bahamian Archipelago to the north are fewer in relative abundance to the rest of the domain and dispersal to Cuba is substantially greater. These differences can be attributed to contrasting programming in addition to hydrodynamic variation, inclusive of yearly, monthly, dispersal depth, and resolution differences. The density of final dispersal locations from 2009 to 2015 most closely resembles dispersal in 2018, with larval retention strongest within Northeast Providence Channel, though diffusive losses of larvae to the northeast most closely resemble 2019. It may be inferred from the combined timeseries of both Zeng et al. [55] and this study that the prevailing dispersal pattern is that of the Northern dominant paradigm and that larvae are most consistently distributed—independent of settlement—within Northeast Providence Channel.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Comparison of LDM structure to Zeng et al. 2019.

https://doi.org/10.1371/journal.pone.0276528.t001

Larval development environment

Environments exert influence on physical performance characteristics during larval dispersal and development, as larvae exhibit phenotypic plasticity in traits such as growth rate, thermal tolerance, osmoregulation, and swimming performance [109]. Variability in phenotypic expression does influence the true measure of connectivity as phenotype-environment mismatches can reduce individual fitness, preventing the full cycle of connectivity where an individual further contributes offspring to the population [110]. Accordingly, assessing the larval development environments experienced by larvae as they disperse throughout The Bahamian Archipelago was important for gaining insight into significant variability that could confer advantages or deficiencies to larvae dispersed to certain islands.

The relative relationships of larval development environments for successfully settled larvae were largely consistent between 2018 and 2019, as both years followed the Northern dominant dispersal paradigm. However, the magnitude of differences was not equal between years, which can be attributed to variation in the hydrodynamics. Conversely, many comparisons indicated no differences between islands. Collectively, consistent relative relationships and null comparisons are indicative of prevailing conditions and consistency in the larval development environments across space and time.

The strongest contrasts in larval development environments per island were in the coefficients of dispersal and settlement DPS (Fig 7). The two metrics were not correlated, though both could potentially have strong influences on larval condition and success in adapting to littoral nursery habitats. Many of the islands in which larvae successfully settled had small coefficients of dispersal, implying that their dispersal pathways were heavily influenced by interacting with circulation features. Beyond their larval retention capacities, gyres and eddies create environments that have unique physical properties that strongly influence both biotic-abiotic and biotic-biotic interactions [111].

The primary circulation feature in the northern extent of the Bahamian Archipelago is the mesoscale gyre found within Northwest Providence Channel. The gyre forms in winter and is created by an influx of cool Atlantic water from the east through Northeast Providence Channel and warm Gulf Stream water from the west through Northwest Providence Channel [98]. Inputs from the Atlantic to the east and the Gulf Stream to the west are not equal or always consistent in their volume transport [98]. In 2013, this resulted in a deeper mixed layer depth. In 2018, a warm core anticyclonic eddy was present due to elevated velocity intrusion of cold water from Little Bahama Bank [98]. And in 2019, a slower velocity warm core anticyclonic eddy was present (S2E-S2G Fig). Gyre presence under prevailing conditions places larvae into highly productive conditions, where both upwelling from cyclonic eddies within the > 1000 m deep Providence Channels and the downwelling anticyclonic gyre can supply nutrients for phytoplankton and zooplankton to thrive in an otherwise nutrient deficient environment [112, 113].

The stable thermal stratification of the region in winter and eddy presence appears to be a larval development environment that is tightly coupled to bonefish evolutionary history as congeners in the Gulf of California have been found in their highest abundances within the same environment [21]. Accordingly, larvae with higher coefficients of dispersal may have increased access to prey, and subsequently have improved body condition and higher survival upon settlement. The positive relationship between body condition and settlement survival has been documented in reef fish [114], and body condition can even offset deficiencies in competitive ability brought on by late settlement (i.e., priority effects) [115–117]. Bonefish larvae settling in The Bahamas are likely to be less dependent upon the compensatory competitive advantage of settling with higher body condition than reef fish due to nursery habitat composition and obligations. Bonefish exploit unstructured littoral nursery habitats that are not constrained by competitive or predatory priority effects, as such habitats in The Bahamas are plentiful and the success of their recruitment appears to be facilitated by inter-species schooling with gregarious small-bodied fish [118, 119]. More research is needed to understand how body condition in settlement stage bonefish effects recruitment.

Differences between the other successfully settled larval development environment measures—depth released, mean dispersal depth, and mean temperature—were much smaller and more inconsistent in their relative relationships (Figs 7 and 8). However, consistency across islands can support the presence of predictable and stable environments that are vital to larval survival. The peak distributions of mean dispersal depths were shallower than release depths across all islands and were closely associated with the thermocline. Larvae are universally found more frequently associated with stratified oceanographic features and have both biological and behavioral means to maintain close association [112]. The boundary layer within the stratification feature provides insular protection from the larger stochastic processes that occur within the larger water masses, providing reduced turbulence and flow velocities [120] over large amplitude orbital motions [100], more predictable dispersal [121], and prey availability [112]. The strong association of the deep chlorophyll maximum and the thermocline within the eastern Caribbean exemplifies the nutritional benefits of thermocline associations for larval fish within The Bahamas [122]. The buoyancy of bonefish eggs and larvae [123, 124] are most likely to have imparted the consistent relationship between the thermocline and mean dispersal depth. An instantaneous buoyancy rate of 8.5 m·d^-1 at 0 h post-hatch (25 h post-fertilization) was calculated using the standard Stokes’ Law following the methods in McDonnell and Buesseler [125] with a larval density of 1.02 × 10⁻⁵ kg·m^-3, head diameter of 2 × 10⁻⁴ m, a length of 0.005 m [123], a dynamic viscosity of 0.959 kg·m^-1·s [126], and a water density of 1024.4 kg·m^-3. The buoyancy rate allowed for larvae to rise towards the thermocline, and the rate of decreasing buoyancy over development reached a nearly neutral specific gravity in close temporal proximity to the larvae reaching the thermocline.

Temperature also significantly impacts larval development, growth, and survival [109]. While temperature effects on bonefish larvae are unknown, Japanese eels in captivity experienced higher hatching and survival rates, as well as faster hatching, prefeeding growth, and onset of feeding in waters progressively warmer from 19–28°C [127]. Similar results were observed in tropical reef fish [128]. Consistency in the positive relationship between temperature and development suggests warmer conditions would benefit bonefish larval development; however, the lower and upper range of beneficial temperatures cannot be assumed. Future research into how temperature impacts bonefish larval development is needed in order to fully assess the implications of variable temperatures during dispersal.

Future management and bonefish research suggestions

Zeng et al. [55] concluded that management plans effectuated at scales pertaining to isolated stocks were not sufficient to effectively conserve bonefish populations within the Caribbean. The same issue of scaling can be applied at an intra-jurisdictional scale within The Bahamas. While the harvest of bonefish is restricted to subsistence fishing within The Bahamas [33], loss of PSA and nursery habitats remains a present threat to bonefish populations [38]. The Bahamas has established 77 marine protected areas (MPAs), marine managed areas (MMAs), fisheries reserves, ecological reserves, and national parks that protect marine and shore habitats [129], with bonefish as a focal species for habitat protection in some of these areas. Tagging studies focusing on the movements of bonefish from home flats to PSA locations [14, 48] helped to inform park inclusion of bonefish habitats. Data on nursery habitats and settlement patterns did not exist for consultation and incorporation into spatial management plans. The results of our study show an alignment of core settlement areas and park protected areas throughout Abaco, Grand Bahama, the Berry Islands, and Northern Andros. Within these locations, protections from development and water quality degradation should be sufficient to protect nursery habitats and larvae sourced from spawning events in Abaco. Our models did not discern potential bonefish nursery habitats through remote sensing or in situ assessments. All shorelines were included as potential settlement zones, as the capabilities of bonefish larvae to move through habitats is unknown. Furthermore, methods to delineate bonefish nursery habitats are still being refined and are labor intensive [54, 104]. Accordingly, suggestions to include other locations with high concentrations of larval influx to the Bahamas’ parks and spatial protections programs requires assessments of habitats to delineate nursery status. We suggest further examination and consideration of potential nursery habitats in the core settlement areas of Moore’s Island Abaco (northwest of the spawning locations), the central most eastern islands of the Berry Islands, and North Eleuthera that would result in the creation of new parks. Expanding and conjoining of Marls of Abaco National Park (west central Abaco) to Cross Harbor National Park (South Abaco) would encompass the 23 km of catchment to the north that contributes to the South Abaco PSA [14], and thus would be inclusive of self-recruiting larvae. Larval supply was strong along the northern shore of Grand Bahama, and westward expansion of North Shore/The Gap National Park (north central Grand Bahama) to conjoin with West End MPA (northwest Grand Bahama) would protect recruitment to this region. A core settlement area was observed upon the northern tip of Andros, where Joulter Cays National Park encompasses the sand flats habitats that North Andros bonefish occupy. However, Joulter Cays National Park does not extend to the creeks and marls between the sand flats and mainland of North Andros—habitats that are likely to support juvenile bonefish recruitment. Expansion of Joulter Cays National Park southward to include the creek and marl habitats is recommended. Core settlement areas were also observed in East New Providence and West Grand Bahama; however, both regions are heavily developed by nearby Nassau and Freeport, which have already contributed to the destruction of nursery habitats and low bonefish densities found in those regions.

As new biological information is discovered on the spawning and early life history of bonefish, the construct of bonefish LDMs should be further refined to increase accuracy in representing the dispersal of larvae. We suggest future research subjects that will contribute to bonefish LDMs. First, prioritize describing ontogenetic changes in larval swimming capabilities, in both the pelagic and nearshore environments, since larval behavior can significantly influence dispersal and connectivity [76]. Second, evaluate the relationship between bonefish spawning site selection and physical ocean conditions. Additional bonefish spawning tracks should be conducted while implementing the use of a CTD with a depth rating > 100 m to capture the bottom extent of the offshore migration in instances where fish move below 100 m. The required equipment will be a logistical constraint, as the CTD and retrieval equipment needed to sample such depths increase in size and mass. Finally, future bonefish LDM studies should incorporate ground truthing using juvenile habitat surveys [see 105] and/or larval collections [see 53] and evaluate genetic relatedness. Collectively, these research objectives will fill knowledge gaps in bonefish early life history that have persisted for decades, and ultimately help to inform conservation measures across the Greater Caribbean.