To account for the variable range of adult densities observed by fishery-independent surveys at each jurisdiction, we conducted an uncertainty analysis. From the above-mentioned jurisdiction-specific distributions of adult conch densities, we randomly drew 100 draws for each jurisdiction, and then translated the randomly drawn densities of adults to reproductive output as with previous simulations. We then estimated larval exchange between spawning and settlement jurisdictions by multiplying the probabilistic connectivity matrix of the unexploited release case by each reproductive output (egg production * connectivity, Kool et al., 2010; Kool et al., 2011). The results of the 100 transition matrices were aggregated for two extremes: minimum and maximum estimated connectivity. The matrices for each scenario were based on locating the minimum (or maximum) number of larvae exchanged between each spawning-settlement pair for the 100 different possible connectivity results.

To identify clusters of highly isolated spawning and settlement sites of queen conch within the Wider Caribbean, we applied the graph theory method described in Jacobi et al. (2012). This method is a minimization algorithm that recursively seeks solutions that minimize connections between clusters, while maximizing connections within clusters (Jacobi et al. (2012), Garavelli et al., 2014). Briefly, it minimizes an objective function consisting of the connectivity between populations and a penalty parameter to avoid aggregating sites i.e., to avoid the optimal minimization solution of the connectivity matrix, which would be achieved considering only one cluster composed of all spawning and settlement sites. We used the Jacobi et al. (2012) minimization algorithm from the open-source R-package ConnMatTools (Kaplan et al., 2017). We did 10 different iterations of the algorithm using different penalty parameters (Jacobi et al., 2012). From these iterations, we selected the solution yielding the least amount of connectivity between clusters while resulting in the highest self-recruitment within the cluster. For this, we selected the solution where the connectivity between clusters normalized by the self-recruitment resulted in the smallest values, a value termed “quality” (Jacobi et al., 2012).

Given the distribution of habitat (patches of habitat on coasts, islands, and banks) within the queen conch distribution area, we assumed that each one of these clusters, as characterized by the minimization algorithm and estimated larval dispersal, acts as a metapopulation (sensu Wells and Richmond, 1995). Thus, metapopulations were the results of the clustering analyses, populations were defined by jurisdictional divisions, and distinct banks within a population were considered subpopulations.

From the probability matrices, we also derived the connectivity distance between each spawning-settlement pair for several generations. We did this by a forward projection of the matrix using a modified model of migration based on Bodmer and Cavalli-Sforza (1968) as described in (Kool et al. 2010; Kool et al. 2011) and Foster et al. (2012). In the unexploited scenario, we assumed that all habitats in our model have the same carrying capacity and reproductive output, and thus, the forward matrix with the allele frequencies expected at time t (Qt) is given by Q t = M + I t ¯ , where M is the connectivity matrix, I is an identity matrix of the same dimension as M, and the bar denotes row-normalization (Kool et al., 2010). While the connectivity matrix reflects single dispersal events between populations during one generation, the forward projection of the matrix captures indirect connections over multiple generations. We then estimated the Nei’s D_A genetic distances (Nei’s angular distance, as defined in Nei et al., 1983) between sites using the forward matrices, following Kool et al. (2010). The estimated Nei’s D_As are used for comparisons to the linearized G_ST (G_ST/(1- G_ST)) of a Caribbean-wide genetic study (Truelove et al., 2017) by means of a Mantel-r test (Mantel, 1967; Legendre and Legendre, 2012) to test for isolation by distance using the R package adegent v2.1.1 (Jombart et al., 2008).

The probabilities of connectivity for both hydrodynamic fields (HYCOM and GLORYS, Figure 4 ) show that larval exchange tends to follow the main pattern of currents in the Caribbean, Gulf of Mexico, and North Atlantic ( Figure 1 ). In the unexploited scenario, all network nodes (jurisdictions) were generally well connected, with multiple pathways of dispersal for both exporting and receiving larval subsidies - the median number of exporting connections (considering all jurisdictions and both hydrodynamic models) was 9, while the median number of incoming connections was 7. The connectivity from both hydrodynamic products also presented a high likelihood of self-settlement. However, the level of self-retention varied for some jurisdictions by model, e.g., Turks and Caicos Islands, which exhibited a higher likelihood of self-settlement when dispersed with the HYCOM than the GLORYS. Moreover, both networks showed that the only completely isolated node in the network is Bermuda, which did not receive nor provide larval subsidies to any other jurisdiction.

Some key differences between the two ocean circulation models were confined to regional scale connections. For instance, the HYCOM estimated a higher likelihood of exchange 1) between and within reefs in Cuba and The Bahamas, and 2) from Cuba to Mexico and the Greater Antilles, while the GLORYS simulations estimated a higher likelihood of export from Cuba to the Bahamas. Transport within the Windward and Leeward Islands also differed slightly between the two sets of simulations. For the HYCOM, larvae spawned on the Windward and Leeward Islands were estimated to have a higher likelihood of settlement in Puerto Rico and the Dominican Republic when compared to the high likelihood of local settlement observed in the GLORYS for this region.

When we introduced realistic reproductive output levels based on exploitation patterns ( Figures 4B, D ), connectivity was largely reduced among regions for both models, as expected, and differences in connectivity patterns between the two models became less apparent. Notably, when considering exploited densities ( Figures 4 , 6 ), there was a reduction in the likelihood of self-settlement. Indeed, even the highest probability of self-settlement in the exploited scenario (The Bahamas, Jamaica, and Cuba) was smaller than observed in the historical scenario. Other jurisdictions in the network are likely to export most of their larval subsidies. By contrasting the connectivity networks for unexploited and exploited densities, there was a clear reduction of connections between the eastern and western Caribbean. It was also clear that the lack of spawning in Belize and Mexico, and the reduced spawning in Nicaragua, further break the transport to upstream sites, particularly the USA, The Bahamas, and northern Cuba. The regions that remained mostly inter-connected were the Central American jurisdictions (Costa Rica, Honduras, Nicaragua), the westernmost Greater Antilles Islands (Jamaica, Cuba, Turks and Caicos Islands), and the offshore Serrana Bank. Larvae released in these regions were exported to other regional reefs (Belize, Mexico, Panama, Colombia, Haiti, Serrana, and Quitasueño Banks) for both models.

Generally, the pattern of export or import of larvae for the unexploited homogenous spawning scenario matched the experiments with both hydrodynamic fields ( Figure 6A , “unexploited”). Mexico, Costa Rica, British Virgin Islands, Anguilla, Saint Martin, and Saint Barthelemy presented a higher likelihood of acting as relative sinks, while the Windward Islands, Nicaragua, and Colombia’s Mesoamerican offshore banks (Quitasueño, Serrana, and Serranilla) appeared as (relative) sources. Once reproductive output was scaled by adult conch densities ( Figure 6A , “exploited”), the relative importance of regions as a source and a sink also changed. Saba, Saint Kitts and Nevis, Nicaragua, Serrana Bank, and Turks and Caicos Islands were sources on both models, while Puerto Rico, Honduras, and Cayman Islands became sinks. Two countries with the largest habitat areas, Cuba and Jamaica, appeared as neutral, which is expected since these countries also showed high self-recruitment. The probability of The Bahamas acting as a relative sink slightly increased in the exploited simulations.

The betweenness centrality (BC) for the unexploited scenario ( Figure 6B , “unexploited”) showed that despite differences in connectivity patterns between the simulations using different hydrodynamic products, the jurisdictions showing the highest BC were mostly consistent among the two simulations. These were the Dominican Republic, Puerto Rico, Jamaica, and Cuba. Following these were also Honduras, Colombia, and the British Virgin Islands. When considering exploited densities to scale the spawning ( Figure 6B , “exploited”), Jamaica and Cuba emerged as central nodes for both hydrodynamic products. The Bahamas, Nicaragua, and Honduras appeared as secondary nodes. A notable change between the unexploited and exploited scenarios was Puerto Rico - when exploited densities were considered, the local reefs were unlikely to serve as an ecological corridor between Caribbean regions.

For the unexploited spawning scenario, local settlers comprised most of the settlers in several jurisdictions, including the Greater Antilles Islands, some of the Windward Islands ( Figure 6C , “unexploited”), The Bahamas, Florida, and Bermuda, and results obtained with the two hydrodynamic products largely agreed. When considering exploited distributions ( Figure 6C , “exploited”), local settlers comprised most of the settlers in Nicaragua, Jamaica, Cuba, and Turks and Caicos Islands, while largely important for The Bahamas and Saint Kitts and Nevis. Many of the jurisdictions that appeared to mostly rely on their own settlers in the unexploited experiment did not exhibit spawning (e.g., Haiti, Dominican Republic, Barbados, Trinidad and Tobago, Colombia, Belize, US, and Bermuda).

Based on the minimum exchange scenario for each jurisdiction ( Figures 5A, C ), connectivity was further reduced from the exploited scenario, with more countries not producing viable settling larvae (or not spawning), including Honduras, Turks and Caicos Islands, Serrana Bank, St. Kitts and Nevis, and Sint Eustatius. However, the reduction in spawning did not further alter the probability of connectivity with other jurisdictions which were spawning, either for the HYCOM or GLORYS simulations. However, there were differences in the strength of connections and the number of larvae settling on their natal reefs. There was a higher likelihood of self-settlement in Jamaica, Cuba, and The Bahamas for both models. The connectivity between the eastern and western Caribbean regions here was further fragmented.

On the other hand, for the scenario considering the maximum connectivity between two countries ( Figures 5B, D ), spawning occurred in nine additional jurisdictions from the sixteen in the exploited scenario, including the British and US Virgin Islands, Guadeloupe, Colombia, Serranilla Bank, Belize, Mexico, and The Bahamas (Western Great Bahama Bank and Central Great Bahama Bank). As expected, this scenario presented higher connectivity than the exploited scenario (based on median values of densities), although a break in larval transport between the western and eastern Caribbean still emerged, since the centrally located Dominican Republic and Haiti only served as settlement sites and did not support reproductively active populations in this scenario. Several countries also presented a higher likelihood of local settlement in this case, particularly Jamaica, Cuba, The Bahamas, and Colombia. Patterns of transport and local settlement differed between the two hydrodynamic models for the Windward and Leeward Islands, but generally there was a higher likelihood of local settlement for the jurisdictions still reproductively active in this case. The probability of connectivity was still similar for the countries where conch were spawning in the two scenarios. However, in this case, there was much more exchange between the Western Caribbean jurisdictions and a higher likelihood of export of larvae downstream of the local currents, particularly to countries located downstream, such as Mexico and Florida.

Regional patterns of connectivity between distinct Caribbean and North Atlantic regions showed changes between hydrodynamic models and scenarios ( Figure 7 and SI Table 3 ). The major differences in the unexploited scenario between models were estimated in Cuba, south Florida, the Leeward Antilles, southern Mesoamerica (Costa Rica and Panama), and the southern Leeward Islands for the unexploited scenario and both hydrodynamic models, with six metapopulations. For the HYCOM simulations there were: 1) the Lesser Antilles and Puerto Rico; 2) Venezuela, the Leeward Antilles, northwestern Caribbean, Florida; 3) Haiti, Dominican Republic, The Bahamas, Turks and Caicos Islands; 4) Cuba and Cayman Islands; and 6) Bermuda. For the GLORYS: 1) Western Caribbean, Jamaica, and Cuba; 2) Florida; 3) Puerto Rico and Northern Lesser Antilles; 4) Southern Lesser Antilles; 5) The Bahamas, Turks and Caicos Islands, and Venezuela; and 6) Bermuda.

When spawning was reduced according to the relative exploited densities of conch, connections were decreased by the lack of spawning in exploited jurisdictions, as particularly evident in Florida, Mexico, Haiti, and the Dominican Republic. However, the entire region was still connected by larval dispersal through connections within the central regions of the network. When exploited densities were considered, there was an increase of metapopulations for the HYCOM (from 6 to 7 metapopulations, SI Table 3 ) and a decrease for the GLORYS (from 6 to 4 metapopulations). Remarkably, for both models, the jurisdictions harboring the largest conch populations (Cuba, Jamaica, and The Bahamas) became metapopulations by themselves (Jamaica) or only with their nearest neighbors (Cuba and Cayman Islands, The Bahamas, and Turks and Caicos Islands). Both models also diverged in the number of connections exhibited by each node.

The metapopulations obtained with the uncertainty analyses ( SI Figure 6 ) suggested that the metapopulation organization of queen conch is highly dependent on the spatial patterns of reproductive output. The metapopulations obtained with the minimum connectivity ( SI Figures 6A, C ) were highly distinct for the two hydrodynamic models. While six connected metapopulations were identified for the HYCOM, only two isolated metapopulations emerged for the GLORYS - one encompassing the western Caribbean, and another for the easternmost region (Windward and Leeward Islands and Puerto Rico). For the maximum exchange, the number of metapopulations remained the same for the HYCOM, although with changes to its spatial structure. The number of metapopulations greatly increased for the GLORYS, which showed eight distinct clusters of mostly self-contained habitat. Metapopulations were more similar among the models, i.e., Panama, The Bahamas, and Jamaica were individual metapopulations for both models. Similarly, output from both models resulted in a cluster formed by jurisdictions not exhibiting reproductive activity and that showed a low probability of receiving larval subsidies from other metapopulations.

The origin of settlers at each discrete habitat unit was grouped by the genetic divisions characterized by Truelove et al. (2017, Figure 8 ). The unexploited spawning scenarios ( Figures 8A, B ) corresponded well to the patterns observed by Truelove et al. (2017) given that larvae within each region identified by the authors were most likely locally originated. The exception was the high probability of larval exchange between The Bahamas and Turks and Caicos Islands and the Greater Antilles, which occurred for both hydrodynamic models. In the exploited reproductive output scenario, we considered that six of the twelve jurisdictions sampled by Truelove et al. (2017) were not reproductively active. Due to the lack of spawning, it was expected that not all connectivity patterns could be reproduced. Indeed, in this case, the high self-settlement observed for Mexico, Belize, and Florida ( Figures 8C, D ) was absent due to the lack of reproductive activity. The genetic connectivity analyses subsequently focused only on the results of the unexploited simulations since the results of the exploited density scenario were insignificant due to the reduced number of data points.

The results of the Mantel-r test between G_ST values from Truelove et al. (2017) and the Nei’s D_A genetic distance derived from our connectivity matrices for 10, 50, and 100 generations ( SI Table 2 ) showed that the Mantel-r increased with the number of generations considered. By considering the averages of the Nei’s D_A genetic distances calculated for each individual hydrodynamic product, the Mantel-r increased for 50 and 100 generations, while the p decreased. This approach captured the uncertainties of the hydrodynamic field estimated by each individual model and improved our estimations ( Figure 9 and SI Figure 7 ). Results of both the Mantel-r tests ( SI Table 2 ), the scatter plot between the estimated Nei’s D_A genetic distances and linearized G_STs from Truelove et al., (2017, Figure 9 ), and plots of distances between sites ( SI Figure 7 ) suggested that queen conch populations exhibit a distance-by-isolation pattern. Our model explained part of the variability and the consideration of conch specific biological traits improved the results obtained solely by considering oceanographic distances between sites.

The connectivity network ( Figure 4 ) of queen conch for both hydrodynamic models used generally follows the main surface currents in the region - Caribbean Current, Yucatán Current, Loop Current, and Florida Current ( Figure 1 ) - flowing in a clockwise direction from the Lesser Antilles to the Florida Straits. In the northern part of the Greater Antilles, the Antilles Current flowing northwestward favors connections within the Windward and Leeward Islands with the Greater Antilles and Bahamas. Despite the remarkable footprint of this current system, the circulation in the Inter-American Sea is highly variable. The intensity and extension of the currents change at seasonal and inter-annual scales and numerous sub-mesoscale and mesoscale features facilitate recirculation, settlement of larvae close to their spawning sites, and larval exchange with neighboring jurisdictions (Paris and Cowen, 2004; Paris et al., 2005; Kough et al., 2013; Kough et al., 2016; Lopera et al., 2020). Indeed, numerous recurrent gyres are known to occur in the region, notably the semi-permanent cyclonic Panama-Colombia Gyre (Lopera et al., 2020), the transient Honduras and Cozumel Gyres (Carrillo et al., 2015), and the anti-cyclonic gyre around Cayman Islands (Molinari, 1980). Thus, differences in connectivity between the hydrodynamic products reflect the variability of these surface and subsurface flow patterns, particularly around the Lesser and Greater Antilles, which affect the connectivity patterns between The Bahamas, Cuba, and the Windward and Leeward Islands. However, the most apparent differences in connectivity are between our exploitation scenarios, and not between our hydrodynamic products.

The most apparent differences between the unexploited and exploited connectivity emerged from the fact that most jurisdictions were estimated to currently hold conch densities well below the critical threshold for reproduction ( Figure 3 , Horn et al., 2022), and thus were here considered to be reproductively non-functional ( SI Figure 4 ). These jurisdictions, therefore, could only act as sinks for larvae, not providing larval subsidies for other locations or themselves. The connectivity patterns emerging from the exploited scenarios, considering any range of densities (maximum, minimum, or median Figures 4 , 5 ), are thus drastically different than expected from uniform spawning in the unexploited scenario. For example, due to their up-current position, the Leeward and Windward Islands (i.e., Lesser Antilles) were estimated to be historically important for contributing larval input to other jurisdictions, particularly the Leeward Antilles, Venezuela, and the Greater Antilles. However, due to overfishing, most of these jurisdictions are no longer likely to provide larval subsidies in the exploited scenarios and are relatively isolated from outside replenishment.

Other patterns in comparing the unexploited and exploited simulations were more subtle, but locally significant. For example, historically the Turks and Caicos Islands were estimated to have received subsidies from the Dominican Republic and Haiti. However, due to overfishing in these source jurisdictions, the exploited scenario suggests that the Turks and Caicos Islands are now dependent on reduced local production, which is less likely to be locally retained, with a large percentage of larvae being exported to The Bahamas. Likewise, the unexploited simulation suggests that Florida was dependent on relatively high local retention, with other significant input coming from Mesoamerica (particularly Nicaragua, Belize, and Mexico). These jurisdictions now have estimated densities below the critical reproductive threshold; thus, Florida currently has no significant sources of larvae, except a small likelihood of receiving larvae from Cuba. This could explain why, despite a moratorium on fishing for several decades, Florida’s conch population has not recovered (Delgado and Glazer, 2020).

Heterogeneous fishing pressure and localized depletion also significantly influenced the role of each jurisdiction in the conch network ( Figure 6 ), as exemplified by the changes in the betweenness centrality of the Dominican Republic and Puerto Rico ( Figure 6B ). These jurisdictions presented some of the highest centrality measures in the unexploited scenario yet no longer function as important connectors in the exploited scenario. An almost complete break in the network was apparent based on the exploited scenario, with the Dominican Republic receiving limited larvae from Cuba, Turks and Caicos Islands, and from a deep mesophotic reef off the west coast of Puerto Rico. The channel between Puerto Rico and the Dominican Republic, the Mona Passage, has been previously shown to act as a transport barrier or filter between these two jurisdictions (Baums et al., 2006; Beltrán et al., 2017). However, the flow in the channel is highly variable (Metcalf et al., 1977), allowing for a small likelihood of periodic connections between populations across the passage (Baums et al., 2006; Beltrán et al., 2017; Prakash et al., 2018, Willis et al., 2022). Since our putative mesophotic spawning site was located at the Abrir La Sierra reef (Garcia-Sais et al., 2012), on the eastern edge of the Mona Passage, it is expected that probabilities of cross-channel connections might be observed in the explored scenario.

While the Dominican Republic and Puerto Rico nodes have been removed from the chain of larval supply, Jamaica and Cuba remained as important connectors in the western portion of the range, and some of the offshore banks in Colombia remained as functional connectors. Although connections have been lost in other locations due to localized overfishing, there were still connection points, albeit reduced, that would allow some exchange of larvae and maintenance of some genetic diversity over time.

Localized patterns of conch overfishing also affected the conch network in the Wider Caribbean Region and could influence both conch demographics and genetic population structure. As shown by the clustering analyses of sub-populations under unexploited and exploited reproductive output ( Figure 7 and SI Figure 6 ), changing the reproductive output dramatically changed metapopulation clusters within the distribution range of queen conch. Comparisons between the G_ST (from Truelove et al., 2017) and Nei’s D_A estimated from the larval dispersal simulations showed that the inclusion of conch life cycle characteristics improved the Mantel-r compared to purely oceanographic distance (Truelove et al., 2017), indicating the importance of considering species-specific biological traits and spatial patterns of fecundity in larval simulation models. By aggregating the estimated Nei’s D_A for both hydrodynamic models into an ensemble value, we reduced the uncertainties of the hydrodynamic field improving our connectivity estimations.

In the unexploited scenario, individual settlement habitat units (i.e., polygons) in Mexico and Belize showed a high level of self-recruitment ( Figure 6C ), which corroborated the genetic patterns observed by Truelove et al. (2017). However, when considering larval exchange probabilities at a jurisdiction level, this pattern was not clear and the populations were not isolated, as also demonstrated by the metapopulations emerging from the metapopulation analyses ( Figure 7 and SI Figure 6 ). Furthermore, these patterns disappear when considering exploited contemporary density estimates of adult conch, highlighting a disconnect between genetic and ecological (or demographic) connectivity of this population.

This disconnect is not unexpected since genetic and ecological connectivity explore processes spanning different spatiotemporal scales. The ecological connectivity covers the shorter scales of early planktonic stages, examining the exchange of individuals between populations (Kough et al., 2017). Typical studies of genetic connectivity unveil patterns occurring through multiple generations or even evolutionary timescales and characterize the gene flow between these populations (Thomas et al., 2014; Brown et al., 2017; Carr et al., 2017). Whereas simulations of larval dispersal can inform the evolutionary connectivity of organisms, characterizing, for instance, transport barriers and isolated populations (Cowen et al., 2006; Baums et al., 2006; Truelove et al., 2017), dispersal models are particularly useful to help elucidate patterns of ecological connectivity (Jones et al., 2009). Indeed, our results reflect the different spatial-temporal scales unveiled by the genetic and ecological connectivity. By comparing similarities for multiple generations obtained from the unexploited scenarios to genetic divergence (linearized G_ST), we found these to be significantly correlated. However, once we considered localized depletion patterns, these correlations disappeared, suggesting that the genetic results reported by Truelove et al. (2017) could have captured patterns of connectivity that no longer exist under the (conservative) depletion scenarios considered in our study.

The objective of this study was to explore how localized harvest patterns affect the probabilistic ecological connectivity of the queen conch throughout its range. The lack of detailed demographic and habitat information (e.g., population distribution, age structure, adult densities, fecundity, carrying capacities) precludes a direct estimation of the strength of larval recruitment. Regional cooperative management efforts have advocated for a more consistent and uniform collection of these data across the WCR (Prada et al., 2017). These data would allow us to calculate the persistence and replenishment rates of individual populations (Leis, 2006; Burgess et al., 2014; Lowerre-Barbieri et al., 2017; Hamilton et al., 2021). Moreover, the interpretation of ecological and genetic connectivity has important spatial management implications. While small and sporadic exchanges of a few individuals might lead to panmixia of distant sub-populations, a population requires a steady and strong source of larval subsidies for its support (Wright, 1931; Botsford et al., 2003; Lowe and Allendorf, 2010; Paris-Limouzy, 2011; Hawkins et al., 2017). Therefore, ecological connectivity projections provide vital information for management by delineating areas that are self-reliant on their recruits to support their populations and in identifying relative sources and sinks (Botsford et al., 2003; Botsford et al., 2009; Hawkins et al., 2016).

Here, we characterized distinct metapopulation delineations for queen conch for the scenarios considered in our study (distinct hydrodynamic models and assumptions of adult densities) throughout its range. The existence of these self-sustained metapopulations implies the need for effective management measures within these regions. Here we showed that configuration of such metapopulations is strongly affected by localized depletion patterns, more so than by oceanographic patterns, given the changes in the metapopulation configurations in the homogenous and depleted scenarios ( Figure 7 ). As these metapopulations cover multiple jurisdictions, their management would thus require international cooperation (Kough et al., 2017).

Connectivity processes have been explored in many studies using larval dispersal simulations in the last decades; however, studies still consider uniform spawning populations, despite the noted importance of the effects of variable spatio-temporal fecundity on connectivity patterns (Hughes and Tanner, 2000; Karnauskas et al., 2011; Lowerre-Barbieri et al., 2017; Castorani et al., 2017; Johnston et al., 2018; Hamilton et al., 2021). We found that the inclusion of realistic present-day patterns in reproductive output changes connectivity patterns and the structuring of queen conch sub-populations and metapopulations across their range, which must be considered in future studies.

Overharvest not only can change the patterns of connectivity, as addressed in this study, but also lower the genetic diversity of fisheries populations (Pinsky and Palumbi, 2014). A reduction in diversity due to conch overharvest has been suggested by Zamora-Bustillos et al. (2011) investigating genetic diversity and structure of two populations along the Yucatán peninsula in Mexico, and by Kitson-Walters et al. (2018) studying the genetic differentiation within the Jamaican EEZ. Truelove et al. (2017) also observed a reduction in allele frequencies in Mexico and Anguilla, however, it is unclear whether these changes are due to isolation by distance, regional overharvest, local overharvest, or their combined effect. The results of our theoretical “virgin stock” simulations, however, suggested that there are no complete barriers to conch larval transport in our study area and thus, exploited patterns of isolation and inbreeding could be related to anthropogenic actions, such as habitat destruction and overharvest. This is a topic which warrants further investigation.

The simulations relied on several assumptions regarding the spatial structure and demographics of conch populations in the Caribbean. As illustrated by our results ( Figures 4 - 7 ), estimated ecological connectivity patterns are sensitive to assumptions regarding both the distributions of queen conch populations and their reproductive output. Thus, we highlight the need for updated demographic and habitat information for queen conch, strongly corroborating needs identified in previous studies (Prada et al., 2017; Kough et al., 2017; Truelove et al., 2017; Kitson-Walters et al., 2018; Boman et al., 2021). To refine larval connectivity estimates and fully understand conch demographics – at both broad and smaller spatial scales within jurisdictions – the following are needed: 1) representative surveys of conch population densities across a range of habitats, along with a precise understanding of the areas over which these densities can be extrapolated (Prada et al., 2017; Kough et al., 2017); 2) densities of individuals categorized as juveniles and adults and/or detailed stock size structure reported; 3) detailed maps and/or areal estimates of conch habitat (i.e., areas where the species currently or has historically occurred) in each jurisdiction; and 4) surveys of conch populations located in depths below the typical range of fishing (i.e., >30m) to assess the contribution of mesophotic populations that might serve as deep refugia for shallow populations and which remain largely excluded from conch surveys (Boman et al., 2021).

Overall, our results support that estimates of contemporary demographic rates, and thus of fecundity, need to be considered for ecological connectivity studies (Lowe and Allendorf, 2010; Castorani et al., 2017; Johnston et al., 2018). Moreover, our results strongly corroborate advice for international management and conservation of queen conch, as well as research needs identified in previous conch studies (Prada et al., 2017; Kough et al., 2017; Stoner et al., 2019; Delgado and Glazer 2020; Boman et al., 2021). However, given the high likelihood of local settlement and distribution of distinct metapopulations, replenishment processes through larval dispersal are likely to occur only within the metapopulation spatial scales. Our simulations showed that local sub-population success would depend on effective local management, but also on the management of specific upstream source populations. The combination of local self-seeding and distant larval transport are both required for success throughout the species’ range. Regional management efforts are working to implement many of the management plans and measures detailed by Prada and collaborators (2017). Local measures of conservation and management are thus critical to protect individual reproductively active sub-populations and nursery habitat across multiple jurisdictions.