The Azores archipelago is composed of nine islands, spanning 600 km in the Northeast Atlantic, between 23°W and 33°W and 37°N and 40°N ( Figure 1 ) and divided into three geographical groups. The Mid Atlantic Ridge (MAR) crosses the archipelago. Two islands are located west of the MAR, on the American plate, forming the Western Group (WG), while the other islands are located on the Eurasian plate with 5 islands positioned in the Central Group (CG) of the Archipelago and two islands situated further east (Eastern Group: EG).

Larvae were simulated as passive particles, being advected by the hydrodynamic currents. The particle density distribution (PDD) illustrates the larval dispersal along the study area, ( Figure 2 ), indicating the potential particle exchange between different locations. The results reveal a heterogeneous distribution depending on the spawning seasonality. Larvae from the MAR group can reach the Central Group (CG) only during the winter and autumn months (January, February, September, October, November and December). Larval dispersal plumes from the MAR can also reach the Western Group (WG), surrounding Flores and Corvo islands, during March, April, May and June months. Larvae exchange between CG and WG only occurs during July, November and December. However, the low PDD between these groups ( Figure 2 ) indicates a low probability of particle exchange.

In March an eddy is formed between the Western Group and the MAR ( Figure 3A ), contributing to particle retention in this area. Larvae from the CG mostly remain around the islands. Nevertheless, the low current velocity, ( Figure 3 ), is not sufficient to transport larvae from CG to the EG populations, given the PLD used in these scenarios.

Larvae from the Central Group populations, reach various sponge aggregations. Between the Central and Eastern Group, the hydrodynamic patterns go mainly in the eastward direction during March, Figure 3A , while in October there is a dominant current coming from the east and going westward ( Figure 3B ). In the northwest part of the study area, currents are stronger during October ( Figure 3B ) than during March ( Figure 3A ), and the major hydrodynamic fields present another pattern, where the higher velocity (>0.1m/s) matches the absence or reduced particle density, and the higher particle density occurs in locations with lower currents magnitude.

Larvae can travel from a few kilometers to a maximum travelling distance of over 400km ( Table 6 ). Travel distances are highly non-normal, confirmed by the Shapiro-Wilk Normality Test (using Rstudio 2022.02.3). Therefore, rather than considering the average values, the median and the 95th percentile are used for this analysis (Phelps, 2015). Maximum total travel distances are an order of magnitude larger than median distances. Across all March and October scenarios, the longer PLD causes higher travel distances. Median total travel distances for a PLD of 30 days are double the distances with a PLD of 15 days ( Table 6 ). Larval travel distances vary both seasonally and spatially. During March, larvae from the MAR locations: L#1, and L#2, have higher travel distances than those from other locations (with a 95th percentile of 207 and 220 km respectively).

In the Central Group (CG, locations L#3 to L#8), the overall median particle distance in all tested scenarios is lower than in other groups, ranging from 8 to 20 km, which may contribute to higher self-recruitment levels. This lower travel distance is due to the lower current velocities in the region ( Figure 3 ), in contrast to the MAR populations, which are situated in areas with stronger currents.

Simulated larvae mainly remain close to the bottom at the released bathymetry ( Figures 4 , 5 ), despite some travelling through different bathymetric ranges. Histograms of particles’ depth along their trajectory are represented in Figure 4 for the March release and in Figure 5 for the October release. In all the locations, larvae remain mostly within the bathymetric range of the release (>40%). A longer PLD contributes to a wider distribution of particles along the water column. The results show no significant difference in particle depth between the March and October releases. This demersal larval dispersal can also lead to higher levels of particle retention at their source locations.

Results show variation in the level of population connectivity depending on the PLD of 15 or 30 days. Connectivity between the different populations in the study is illustrated by connectivity matrices, and also by connectivity arrows represented on the maps ( Figure 6 ). The matrices depict the percentage of larvae exchanged between different populations, indicating connectivity, or in the case of self-recruitment, the percentage of larvae that settled in the source population after the PLD. A PLD of 30 days enables more connectivity between the different sponge aggregations than a PLD of 15 days ( Figure 6 ). Conversely, a shorter PLD, results in higher self-recruitment levels. There is no connectivity between the EG and the CG or WG, in both PLD scenarios. In the WG, connectivity between Formigas (L#10) and Mar da Prata Seamount (L#9) is unidirectional, with a PLD of 15 days, and bidirectional with a PLD of 30 days. Between the CG and the EG, the larvae exchange is minimal, with no significant larvae exchange among the different populations, as is represented in the connectivity matrices.

This yearly release allows for the analysis of different spawning and hydrodynamic scenarios. The hypothesis of a seasonal spawning release was studied, considering releases of one month in March and one month in October to cover spring and fall scenarios, respectively. The results indicate different connectivity patterns in terms of seasonality. The October release ( Figure 7 ), generates lower larvae exchange between the different sponge aggregations. Population connectivity between MAR and CG only occurs in the March_PLD_30_2018 scenario, namely between the Azores Bank (L#4) and the Cavala (L#1) and Gigante (L#2) Seamounts. In the other scenarios, there is no connectivity between the different groups. In the CG, the connectivity between different sponge aggregations reveals seasonality, with a higher number of connectivity relations in the March scenarios. The percentage of self-recruitment ( Supplementary Tables A3–A6 in the Supplementary Material ) is different from March to October, with the releases from CG revealing the highest self-recruitment percentage in October. In different situations, there is neither self-recruitment, nor connectivity between Gigante Seamount (L#2) and any other populations. The same occurs in the Mar da Prata Seamount (L#9), which shows no self-recruitment in October_PLD30_2017.

Besides the sponge aggregations in the study, larvae can also settle in other regions of the domain. This study was performed within the defined bathymetric range of 600 to 1000 meters depth. Seasonal dynamics can induce different settlement positions. The yearly release considers all the monthly scenarios, and consequently more settled larvae along the Azores region ( Figure 8 ). Considering the bathymetric range from 600m to 1000m, larvae can settle around all the islands except Terceira island. Larvae settle along the MAR and on seamounts situated on its western and eastern flanks but also at different locations of the CG region’. The Central Group is an important recruitment area. This result highlights the potential connectivity between the populations in the Central Group. Major results show that during the yearly release, larvae from Cavala and Gigante seamounts, settle along the MAR, around Flores and Corvo islands, and along the west slope of the Central Group, but they hardly reach the East part of the Central Group. Gigante Seamount can receive larvae from different locations (Cavala (L#1), Gigante(L#2), Açores Bank (L#4), Condor (L#5), and South of Faial (L#6)), contrasting with the seasonal releases where larvae exchange occurs only during the October scenario, between Gigante (L#2) and Cavala (L#1) seamounts. The larvae from South of Faial (L#6) are the ones with a wider range of settling locations; they can settle along MAR, in the WG, in the CG, in João de Castro bank, and in the EG, in the North of São Miguel ( Figure 9F ).

There are no connectivity studies of benthic organisms in the Azores region. A retention particle study between 0 and 150 meters showed the potential of the different island groups of the Azores to retain particles, and their dependence on the current regime and topography (Sala et al., 2013). Close to the Azores a study of the physical connectivity between the NE Atlantic Seamounts showed self-recruitment abilities and great retention of particles on the seamounts (Lima et al., 2020).

This study shows larval dispersal between different sponge aggregations in the Azores, (Figure 9A–J) namely between the different sponge aggregations located in the same group (MAR, CG and EG), as well as between aggregations located in the MAR and the CG. However, there is no particle exchange between the CG and the EG. Moreover, these Eastern Group sponge aggregations do not exchange larvae with the other populations, and are the most isolated. Therefore, these populations may be spatially fragmented, similar to what was identified in other larval dispersal model studies for deep-sea populations (Hilário et al., 2015).

The results show that the local ocean currents shape the larval dispersal and connectivity, creating sub-regions not connected, given the studied scenarios. These kind of hydrodynamic barriers were also identified in other modelling approaches for deep-sea larval sponge dispersal on the east coast of North America (Wang et al., 2021). Moreover, this modelling approach also shows that the local eddies can supply and induce local retention of larvae, promoting connectivity and self-recruitment (Wang et al., 2021), as occurs in this case study in the Azores, namely between the WG and the MAR populations. A retention particle study in the Azores at 150 meters also showed the Eastern group isolation and the large capacity of the central area in the Azores to retain particles (Sala et al., 2016), which is coherent with the observed pattern in this study.

Connectivity among different benthic populations promotes the increase of their genetic diversity (Busch et al., 2021; Wang et al., 2021) and the resilience of the species (Bracco et al., 2019). It is an essential aspect in the development of management and conservation plans for marine ecosystems (Combes et al., 2021). Detailed knowledge of the hydrodynamic patterns and biological interactions that drive the transport of planktonic phases is crucial for improving the effectiveness of the MPAs (Stratoudakis et al., 2019). The model results of larval dispersal and larvae settlement positions for this target species in the Azores were analyzed considering the MPAs from the Azores Marine Park. When MPAs are isolated from each other, they are more vulnerable to local extinction as they cannot be replenished by organisms or larvae from other locations (Stratoudakis et al., 2019). This is especially true for sessile organisms such as deep-sea sponges, which rely on larval dispersal to colonize new habitats (Metaxas and Saunders, 2009), and to support remote populations (Gary et al., 2020). Networks of marine protected areas arranged considering the larval dispersal patterns are considered ideal for the protection of marine species (Steneck et al., 2009). Therefore, studying larval dispersal patterns and identifying the main larval source and sink locations can contribute to the better implementation and management of conservation plans (Combes et al., 2021).

The study of larval dispersal and connectivity of these P. carpenteri aggregations serves as a pilot study for the connectivity of deep-sea benthic organisms in the Azores. The current regime and topography have been shown in other studies to be the main drivers of physical connectivity, explaining the genetic connectivity patterns (Busch et al., 2021; Taboada et al., 2023). Using P. Carpenteri as a proxy, we can discuss how effective the network of MPA from the Azores Marine Park is from the physical connectivity point of view. Among the 10 deep-sea sponge aggregations in the study, two are located in MPAs: Condor and Formigas populations. The Princesa Alice aggregation is located on the southern slope of the Princesa Alice Seamount, outside but adjacent to the MPA delimitation. Larval dispersal and connectivity results, indicate that Condor is an important source population, exchanging larvae with 8 of the 10 aggregations in the study ( Figure 6 ). In addition, its larvae can achieve different locations in the Central Group, in the Western Group (around the islands of Flores and Corvo), and along the MAR: in the north, at 40.3°N (seamount south of the Kurchatov Fracture Zone); at 39°N in the Gigante Seamount; and in the Voador Seamount at 37.5°N ( Figure 9E ). The MPAs of Princesa Alice and Condor may act as sink locations, once that larvae from different aggregations in the study achieve these locations. These results demonstrate the potential connectivity that sponge populations from Condor may have with different populations in the Azores region.

The Princesa Alice sponge aggregation in the study, adjacent to the limits of the MPA, demonstrates the importance of extending the limits of this MPA as it has the potential to retain larvae from multiple locations in the study: Cavala (Figure 9A), Gigante (Figure 9B) and Condor (Figure 9E) seamounts, and South of Faial (Figure 9F), aggregations.

The main results show a lack of larvae exchange between the EG and the other aggregations in the study (Figures 9I, J), mainly due to its distance from the remaining sponge grounds, but also as a consequence of the hydrodynamic patterns between the CG and the EG, impeding the larvae advection in the westward direction. No larvae exchange was identified between EG and the remaining sponge grounds. Therefore, the protection of an area in Mar da Prata will favor the maintenance of the sponge aggregations on the plateau south of São Miguel island, and both Mar da Prata and Formigas populations, which can be isolated from the remaining archipelago, due to their distance, but also restricted by the hydrodynamic patterns. The vulnerability of Formigas sponge grounds, which face several scenarios with no larvae exchange and the absence of self-recruitment, also reinforces the need to maintain the MPA of Formigas.

The João de Castro Bank MPA can receive larvae from the CG and from the Mar da Prata (in the EG) being an steppingstone location between the Central and the Eastern parts of the archipelago. Therefore, this site is very important for the longer-term protection of these aggregations, by promoting the establishment of a linkage between the CG and the EG. The Gigante Seamount, in the MAR, is a recruitment area for deep-sea sponges, retaining larvae from different sources from the CG (Condor, Açores Bank and South of Faial), as well as from Cavala and Gigante Seamount. The results of larval dispersal also show that larvae from the Gigante seamount can achieve the MAR region, up to seamounts located South of the Kurchatov Fracture Zone (north of the study area), and in seamounts west of the MAR like the Buchanon Seamount. On the other hand, larvae from aggregations of Voador Seamount will favor resilience along the MAR, and South of the Azores EEZ.

This study does not take into account the density of these sponge aggregations. Research by Graves et al. (2023) on the drivers of P. carpenteri density suggests that the available environmental data, combined with the scarcity of long-term time series for deep-sea environmental data, do not capture the key factors influencing this species’ density. This limitation restricts the ability to model and identify areas where aggregations are likely to occur, particularly those with dense aggregations that could be considered VMEs. Increasing to a multi-year analysis would provide the opportunity to examine variability over the years and test other hypotheses, such as the average temperature variability, which has been identified as an important driver of P. carpenteri distribution (Graves et al., 2023).

The application of biological larvae traits and their parameterization in biophysical models presents a challenge (Hilário et al., 2015) even in well studied shallow-water systems (Metaxas and Saunders, 2009). When regarding these biophysical models for deep-sea dispersal, the application of biological traits is even more challenging due to the scarcity of data, and difficulty data collection, observation and model validation (Adams et al., 2012; Ross et al., 2016). So, these mode results can be used as a first approach to study the potential larval dispersal in the Azores, but always considering the associated uncertainties. However, models are always a simplification of the reality and despite all the uncertainty regarding the biological parameters (Simons et al., 2013), the validation of the physical circulation model used in this case study (Viegas, 2022) provides us with the guarantee of the correct physical model component, providing a baseline to study different scenarios, and also to be used to study climate change scenarios hypothesis.

Furthermore, this modelling approach could be used in combination with the species distribution models (Beazley et al., 2021) to construct explicit spatial hypotheses for deep-sea populations connectivity following a seascape framework approach (Kenchington et al., 2006; Kavanaugh et al., 2014; Zeng et al., 2020; Swanborn et al., 2022), in this case for the Azores region. In the future, the possibility of collecting physical samples and performing a population genetic study, will contribute to a deeper knowledge of these populations connectivity (Kool et al., 2011; Selkoe and Toonen, 2011; Antonio Baeza et al., 2019; Wang et al., 2021), and will help to validate our hypothesis.