The role of mesoscale-driven connectivity patterns in coral recovery around Moorea and Tahiti, French Polynesia

Abstract

Coral reefs are declining due to anthropogenic warming. Nonetheless, some have recovered quickly from repeated bleaching events. Coral recovery depends on adaptation capabilities, fishing pressure, overall number of stressors, reef conditions before the event, and degree of connectivity. Coral reefs that are connected to many others can receive viable larvae and regain coverage faster. Around Moorea and Tahiti, within the Society Islands of French Polynesia, coral cover has regained its previous levels rapidly, despite several mass bleaching events over the past three decades. Here it is explored whether the connectivity with distant reefs may support such recovery by modeling the transport of coral larvae around the islands over 28 years. Ocean currents enable connectivity with the Tuamotu Islands, ~ 250 km to the northeast, that act as sources to Moorea and Tahiti for pelagic larval durations of three weeks or longer. The circulation around Moorea and Tahiti is very dynamic; mesoscale eddies can also halt the connectivity with the Tuamotu Islands and sporadically transport larvae from reefs to the west and southeast instead. With many undisturbed coral reefs within a 300 km radius and strong mesoscale variability, a dynamic, long-range connectivity may explain the recovery of reefs around Moorea and Tahiti.

Introduction

Sea surface temperatures (SST) exceeding 1.5 °C warming on the global average can trigger multiple tipping points in the climate system, including mass mortality of warm-water coral communities at tropical latitudes^1,2 due to more frequent bleaching events and overgrowth of algae. State-of-the-art climate models project SSTs to raise by an additional 1–4 °C by 2100. Coral reefs are stressed by other factors in addition to warming, from overfishing to nitrogen pollution, predator outbreaks, and ocean acidification^3–9. Nonetheless, many reefs, especially in the Pacific and Indian Ocean, have recovered, relatively quickly, the overall percentage of coral coverage following intense bleaching events. Their recovery ability and possible resilience have raised questions as to what may be the contributing factors^10–12.

Among examples in the tropical Pacific, there is French Polynesia, which consists of 118 islands and atolls with five main archipelagos. The Society archipelago is centered around the islands Moorea and Tahiti (Fig. 1). Moorea is the most studied island within French Polynesia due to the presence of the Moorea Coral Reef Long Term Ecological Research Network (MCR LTER) (http://mcr.lternet.edu/)¹³ and of the Centre de Recherches Insulaires et Observatoire de l’Environnement (CRIOBE), a French field station established in the 1970s. Long term monitoring has helped achieve enhanced understandings of the ecosystem dynamics in the reefs around Moorea and their changes over time. Moreover, reefs in Moorea have a well-documented history of noted periods of dramatic coral death due to bleaching or to Acanthaster planci, also known as crown-of-thorns starfish (COTS), outbreaks, commonly followed by rapid recovery of coral coverage^5–7,9. Over the past three decades, there have been five main warming events that have caused mass bleaching around Moorea and Tahiti, in 1994, 2002, 2007, 2016, and 2019. Despite bleaching levels up to 100% for some coral species, reefs experienced as high as ~76% recovery following each event^8,14–18. Fast recovery of coral coverage has even followed COTS outbreaks, which, however, tend to be more localized. For those, we have good quality historic recordings at reefs around Moorea, but not around Tahiti.

Fig. 1.

Chosen study site centered around the French Polynesian Society Archipelago, Moorea and Tahiti and SST trend (°C/yr). The grey box represents the area used for variability analysis.

It is currently unknown what controls the ability of coral coverage to recover quickly at these locations. It has been suggested that reefs may develop an increased tolerance to higher SSTs following each bleaching event, and that the increased resilience would allow for a shorter recovery period with less die-off under subsequent SST extremes¹⁹. However, the two most recent bleaching events in 2016 and 2019 caused some of the worst damage across multiple reef species²⁰. It was also noted that reefs with reduced land-based impacts and healthy herbivorous fish or sea-cucumber populations, have better survival or recovery trajectories^21,22. Moorea and Tahiti benefit from an abundance of herbivorous fish that consume algae, preventing them from overtaking degraded reefs and aiding larval settlement (http://mcr.lternet.edu/)²³. The richness of fish abundance results from multiple well established Marine Protected Areas (MPAs) and non-fishing zones around the islands²⁴, which in turn are due to the framework of conservation put forward by French Polynesia, further making it a marine sanctuary from pollution and other disturbances¹³.

Another factor that may contribute to recovery potential and maintenance of biodiversity in Moorea and Tahiti is marine connectivity amongst different, potentially distant reefs^25–27. Corals rely on connectivity across reefs to maintain gene flow, which inherently promotes biodiversity and contributes to the entire ecosystem (coral, fish, invertebrates, etc.) health and resilience^28,29. Coral connectivity results from strong, local recruitment that can lead to hundreds to thousands of successful settlements per year³⁰, classified as ecological connectivity in³¹, or from rare, weaker events arising from few immigrants from further away over multiple generations³², classified as evolutionary connectivity in³¹. In this study we will focus on the latter, characterizing coral larval connectivity at Moorea and Tahiti that occurs through spatial recruitment among spatially distant locations.

Reefs that are located hundreds of kilometers apart may not undergo thermal stress at coinciding times. If genetic material may be exchanged, then the least impacted reefs could support the recovery of bleached reefs. Connectivity for the reefs of Moorea and Tahiti has been studied so far only focusing on reefs surrounding the two islands. Even under those spatial constrains, connectivity has been recognized as extremely important for recovery rates after disturbances³³.

The ocean circulation around Moorea and Tahiti is highly dynamic, with significant mesoscale variability in the form of eddies and other coherent circulations at scales of tens to hundreds of kilometers (the Rossby deformation radius in the region is 60–70 km). These eddies have a life span relevant to the time-scales of larval development and settling, and they impact the overall horizontal dispersal of genetic material. Horizontal dispersion at the ocean mesoscales (10–500 km) is indeed determined by mesoscale motions³⁴, which have a non-local range, while kilometer scale circulations have a local impact and are relevant only at short time scales, from hours to few days (much shorter than Lagrangian autocorrelation time), modulating mixing³⁵.

Using ocean currents from a mesoscale-resolving reanalysis dataset that assimilates satellite observations from 1993 through 2020 (GLORYS12V1,³⁶), we ask if the ocean circulation may aid in multispecies coral larval transport to Moorea and Tahiti from islands further away on time-scales of 3 to 5 weeks, and explore the possibility of enhanced recovery rates through mesoscale evolutionary connectivity. Furthermore, we verify that the simulated long-distance larval connectivity is robust to increasing temporal and spatial resolution of the ocean currents over a two-year period using a regional ocean model run at 3 km horizontal resolution.

Results

Average circulation and variability

Many more studies have focused on Moorea’s reefs compared to Tahiti, given the research conducted by CRIOBE and the MCR LTER Network. As a result, we have better observations of the bleaching events and the effects of nutrients and biodiversity in Moorea^20,37,38. However, bleaching in Moorea coincides with bleaching in Tahiti¹⁵, and due to the close proximity of the two islands, the temperature variability around Moorea is similar to that of Tahiti. Given the resolution of the GLORYS12V1 reanalysis and the goal of this work, we consider the two islands as belonging to the same ecoregion (Fig. 1), and we focus on bleaching events neglecting the more localized COTS invasions.

The yearly mean SST spatial pattern for our chosen study site (Fig. S1) consists in a latitudinal gradient with slightly higher temperatures in the western portion of the domain, especially north of 16°S, where the Pacific Intertropical Convergence Zone (ITCZ) exerts a large influence. The SST linear trend around the two islands is approximately 0.007 °C/yr on average over the 28 years considered. The pattern is not uniform, with stronger warming, up to 0.02 °C/yr, to the west of Moorea, and smaller changes (as low as 0.005 °C/yr) to the east (Fig. 1). The yearly mean currents (Fig. S1) are dominated by the presence of the South Equatorial Current³⁹ with mean surface transport from the north east to the south west⁴⁰. This initial analysis indicates that on average, the northern portion in our study domain is exposed to higher SST than regions below 16°S, suggesting that genetic material from northern reefs may be more resilient to high temperatures, since they are inherently exposed to them. Additionally, the average current flow is consistent with a transport pathway from the northeast towards the southwest, providing a direct route from the Tuamotu Islands to Moorea and Tahiti.

At interannual scales, the El Niño Southern Oscillation (ENSO) is the most influential climate mode in the Tropical Pacific. The link between bleaching in Moorea and Tahiti and ENSO, however, is tenuous, as the five bleaching events do not all correspond with a specific phase. We verified that indeed ENSO -on average- exerts only a weak influence on our sites by performing an empirical orthogonal function and a regression analysis on monthly SST values. The time evolution of ENSO can be indexed by the SST anomalies averaged over different regions spanning the equatorial Pacific. For our analysis, we chose the Nino3.4, which is calculated as the average SST anomalies between $5^{\circ }$N-$5^{\circ }$S and ${170}^{\circ }$W-${120}^{\circ }$W, as it best captures the overall impacts of ENSO on the Tropical Pacific⁴¹. We found no influence of ENSO around French Polynesia, with a regression coefficient as low as 0.1 (Fig. S1).

We next explored the variability in the current system around Moorea and Tahiti (black box in Fig. 1), by calculating the eddy kinetic energy (EKE) (Eq. 2) which quantifies the variability in ocean currents due to mesoscale circulation patterns (Fig. 2). This analysis showed that surface currents around Moorea and Tahiti are highly dynamic at all times as EKE peaks throughout the period investigated with large amplitude excursions and no clear seasonality. Looking at past bleaching events, which typically occur in the warmest season, usually March or April and are indicated in Fig. 2, there is no apparent correlation between EKE and the likelihood of bleaching. In 1994 and 2007, the EKE was low both before and during beaching events, while the opposite was recorded before the 2002 event, but with reduced mesoscale activity during the bleaching period. In contrast, in 2016 the EKE was elevated at the start and end of the bleaching period with low values in between, and in 2019 the EKE was above the mean throughout the period of elevated SST. These bleaching events are not only characterized by large contrasts in EKE levels, but also by a range of SST behaviors¹⁵.

Fig. 2.

Time series of EKE and residual ($KE-EKE$) averaged over the area within the box in Fig. 1 with a 10 day smoothing; grey dashed bars represent the five documented bleaching events during this time period, the solid red line represents mean EKE (0.0068 ${\text{m}}^2/{\text{s}}^2$), and the dashed red line represents EKE variance (0.004 ${\text{m}}^2/{\text{s}}^2$).

Figure 3A shows the evolution of SST anomalies from mid February to the end of April during all major bleaching events recorded around Moorea in the past three decades. Each of these events caused extreme bleaching and mortality to main reef species. While SST alone often cannot explain intensity or duration of specific bleaching episodes, and bleaching is often less or more than that predicted using indices based on surface temperatures^42,43. These maps highlight that each bleaching episode had unique SST anomaly signals (Fig. 3B–F). Two commonly used bleaching thresholds for tropical reefs are defined as anomalies exceeding the monthly mean by $1^{\circ }$C⁴⁴ which measures intensity alone, or as accumulated anomalies exceeding the monthly mean by $1_{}^{\circ }$C over the past 12 weeks (the so-called degree heating weeks or DHW in °C-weeks)⁴⁵, which accounts for intensity and duration. The French Polynesian region is subequatorial and both indices are somewhat extreme. Indeed, bleaching occurred when SSTa were below the intensity-only threshold. For example, in 2002 and 2007 the anomalies were consistently positive through the entire 2.5 months period, however, they never exceed 1 °C. Nonetheless, the 2002 bleaching event resulted in average bleaching levels of 55 ± 15% across all colonies¹⁴, and the 2007 event resulted in bleaching levels as high as 64%¹⁵. In 2007 the impact of elevated SSTs over a long period was likely exacerbated by a COTS outbreak that began in 2006 and expanded island-wide in 2008 just after the bleaching event^6,7. Since COTS are corallivores, it is probable that the corals were already stressed and more susceptible to bleaching¹⁵. In contrast, in 1994 and 2019 SST anomalies were elevated throughout the season but exceeded the 1 °C bleaching threshold for short periods in time. The 1994 event resulted in 72.4% of corals bleached in the northeast region of Moorea and 39.6% in the northwest region^5,46, while the 2019 event resulted in the die-off of 76% and 65% of Pocillopora and Acropora respectively²⁰. Finally, in 2016 negative and positive anomalies alternated, with values exceeding the 1 °C threshold for about three weeks, exposing reefs to extreme temperatures which caused the death of Pocillopora and Acropora colonies⁸.

Fig. 3.

(A) Sea surface temperature anomalies calculated using GLORYS12V1 reanalysis and averaged over the study site area (the boxes over Moorea and Tahiti in B–F) for each bleaching season. Each solid line represents one of the years when coral bleaching occurred. Temporally averaged sea surface currents plotted over SST anomalies during the five mass bleaching events in (B) 1994, (C) 2002, (D) 2007, (E) 2016, and (F) 2019 within the study region.

The spatial distribution of SST anomalies, along with surface current anomalies, further supports the notion that each bleaching event was indeed unique (Figure 3). For example, let us consider 2002 and 2019, shown in Fig. 3C and F respectively. In 2002, SST anomalies show only slightly elevated temperatures propagating from the west towards the southeast and occupying the southern portion of the domain. A strong current originates in the area of high SST anomalies with weaker currents elsewhere and a series of mesoscale eddies in the south and around Moorea and Tahiti. In contrast, in 2019 coral bleaching co-occurred with extremely high SST anomalies that moved over French Polynesia from the north, with a more energetic eddy field and a coherent cyclonic-anticyclonic dipole sitting over Moorea and Tahiti. This mesoscale structure has been identified as responsible for the prolonged and severe bleaching in that year, with subsurface temperatures remaining substantially higher than normal for longer than at the ocean surface⁴⁷.

Ichthyop: Lagrangian simulations of larvae transport

To evaluate the connectivity potential of coral larvae around Moorea and Tahiti, we selected a study site that includes 31 islands or atolls centered around the Society Archipelago and western region of the Tuamotu Archipelago of French Polynesia (Fig. 4). Sites are grouped in ecoregions defined by boxes that extend by a one-grid point distance from the N, S, E, and W edges of each island/atoll; from these boxes, areas that have superimposed grids are grouped into larger ecoregions (i.e., W2, E4, and NE1).

Fig. 4.

Ecoregions defined by one-grid point from the N, S, E, and W edges of each island and used in the connectivity analysis. Ecoregions that overlap are grouped together (i.e. W1, E4, and NE1). The white area surrounding Moorea and Tahiti represents the release location for 10,000 virtual larvae.

To establish the connectivity patterns around Moorea and Tahiti focusing on the long-range evolutionary connectivity, we analyzed Lagrangian integrations, performed using the Ichthyop model both backward and forward in time (see “Methods”). The backward runs indicate that most larvae arriving to Moorea and Tahiti originate in the northeastern island ecoregions, the Tuamotu Islands (Fig. 5). The dominant flow during spawning periods is indeed westward/southwestward. This flow pattern is further confirmed by the forward runs which show that most of the virtual larvae released from Moorea and Tahiti move into the western portion of our domain. As run time decreases, in both backward and forward runs, particles can disperse less, decreasing the probability of potential connectivity to distant islands and atolls (Figs. S5 and S6 vs Fig. 5). However, the connectivity potential from the northeastern islands is greater than zero even for a 3-week pelagic larval duration (PLD). Most of the potential connectivity for both backward and forward runs is amongst Moorea and Tahiti themselves as well as ecoregion C1, which includes the island Tetiaroa (Fig. 5, Fig. S6). Due to Fig. 5 being the average potential connectivity of the 28 years of simulations, some of the connectivity is minimized. Along with strong self-recruitment, however, there is significant recruitment from further ecoregions such as C2 (Mai’ao), E1 (Makatea), E2 (Mataiva), and SE1 (Mehetia). When analyzing potential connections for backwards runs on a yearly basis, we found that some specific years exhibit enhanced connectivity reaching ecoregions more than 250 km away from Moorea and Tahiti (e.g. NE1 and NE2 in Fig. 6C) throughout all PLD values explored (Figs. S4, S5).

Fig. 5.

Network plots of the probability of potential connectivity between Moorea and Tahiti and the surrounding islands considering a pelagic larval duration (PLD) of five weeks; network plots show where larvae that settled in the reefs around (A) Moorea and (B) Tahiti could have originated/spawned (backwards runs). Blue circles represent the various ecoregions within the given study site, dark green lines show the connectivity to Moorea, brown lines show the connectivity to Tahiti, solid lines represent the probability of potential connectivity amongst Moorea and Tahiti and the surrounding ecoregions, and dashed lines represent a probability of potential connectivity less than 1%. Legend quantifies the relation between line width and percentage of potential connectivity.

Figure 6.

(A) Four highest (solid lines) and lowest (dashed lines) absolute dispersion years for backward runs considering a dispersal time periods of 5 weeks. Absolute dispersion is calculated using Eq. (5). (B) Final particle dispersal pattern in 2006 (white particles) and 2020 (green particles). (C) Probability of potential connectivity for the four highest and lowest absolute dispersion years.

Furthermore, we calculated the potential connectivity for the western region islands W1 (Huahine), W2 (Raiatea, Taha’a, Bora Bora, and Tupai), and W3 (Maupiti). Given the minimal connectivity potential amongst the western region islands in W4 (Maupihaa), W5 (Manuae), and W6 (Motu One) and Moorea and Tahiti, we investigated if W1, W2, or W3 could act as a bridge for the further islands by advecting virtual larvae forward in time from these given ecoregions. Indeed, we found a positive probability of potential connectivity, comprised between ~ 2.8 and 5.6% (Fig. S6). Although the probability is low, it is still significant and indicates that these western ecoregions can act as a bridge for larval and genetic material exchanges between W2, W3 and W4 and Moorea and Tahiti.

Next we analyzed which years exhibited maximum and minimum absolute dispersion in the backward runs considering fixed PLDs of 5, 4, and 3 weeks (Fig. S5). In the backward integrations, the maximum absolute dispersal reached ~ 250 km within 3 weeks, ~ 300 km after 4 weeks, and ~ 400 km for 5 weeks. The four years with highest dispersion in the simulations with a 5 weeks PLD (1999, 2006, 2008 and 2012, Fig. 6A) all displayed strong potential connectivity to C1, E1, E2, NE1, and NE2 (Fig. 6C). However, 2012 was somewhat different from the others, with a diminished probability of potential connectivity with these ecoregions and instead elevated potential connectivity with SE1. Rather than flowing from NE to SW, virtual larvae in 2012 arrived in Moorea and Tahiti from the SE region and followed a northwestard path. In contrast, when we focused on years of minimum absolute dispersion, most of the potential connectivity was with C1, due to the inability for particles in these years to disperse more than 150 km (Fig. 6B).

Given that the highly variable current system impacts not only mass bleaching, but also probability of potential connectivity, we analyzed the correlation of mesoscale circulations with the connectivity potential. By considering the temporally averaged surface currents along with values of enstrophy, we can visualize the relation between eddy activity and dispersal potential.

First, we compared the years with minimum (2020) and maximum (2006) absolute dispersion for simulations ran backward in time for five weeks (Fig. 6A, B). 2020 exhibited minimum absolute dispersal of about 119 km to the north of Moorea and Tahiti, in contrast to 2006, when the absolute dispersal was 397 km to the northeast island region (Fig. 6A, B). The spatial extent of enstrophy and currents suggests that larvae in 2020 were trapped by the elevated eddy activity. The eddies around Moorea and Tahiti impeded dispersion and resulted in decreased long-distance connectivity, likely facilitating self-recruitment among reefs in the two islands. In contrast, in 2006 eddy activity around Moorea and Tahiti and between the islands and the eastern ecoregions was at a minimum, allowing for the enhanced dispersion and long-range reef connectivity (Fig. 7A, B).

Figure 7.

Surface currents plotted over enstrophy and averaged over larval release periods in (A) 2006 and (B) 2020. (C) Absolute dispersion (“o” markers) and KE (“x” markers) plotted with respect to total count of particles that reached the eastern ecoregions (E1, E2, NE1, NE2, SE1, and SE2), divided by 10,000 giving a total probability of potential connectivity for the eastern ecoregions. The black lines represent the best fits for absolute dispersion (solid) and KE (dashed) with respect to connectivity potential.

Absolute dispersion and enstrophy (which quantifies the strength and presence of vortices) are anticorrelated, but the correlation coefficient is not statistically significant (c.c. = − 0.21) (Fig. S7). Nonetheless, the plot suggests that most (but not all) years with minimum absolute dispersion had high enstrophy and vice versa. The correlation between absolute dispersion and connectivity amongst the eastern islands in the backward simulations is, on the other hand, positive and statistically significant (Fig. 7C). In this correlation analysis we summed up the total amount of particles released from both Moorea and Tahiti that reached ecoregions E1, E2, NE1, NE2, SE1, and SE2 and divided by the total number of particles released, to obtain the overall percentage of potential connectivity with the eastern region of our study site. This percentage was then correlated with the maximum absolute dispersion value for each year, resulting in an $R^2$ value of 0.77 and a p-value of 1.7e-6. Additionally, we found a smaller but significant correlation between KE and absolute dispersion ($R^2$ = 0.55) which further indicates that the strength and direction of the mean total currents controls the distance of larval dispersal during spawning periods.

Over 2006 and 2007 we also compared the overall potential connectivity simulated using GLORYS currents at 9 km resolution and CROCO currents at 3 km (Fig. S8). CROCO does not assimilate any observational data, and because of this the mesoscale variability in this model differs from GLORYS, but the mesoscale statistics should be similar, and therefore the existence of long-range (> 200 km) connectivity. This test was performed to quantify the role of smaller circulations and high frequency variability in the ocean currents, while validating the role of mesoscale advection. The major difference is that simulations using CROCO show additional connectivity beyond the main islands. Most notably, increasing the spatial resolution and the temporal frequency at which currents are saved results in higher variability and increases the probability of potential connectivity of Moorea and Tahiti with western ecoregions (as seen in the two supplementary animations of CROCO and GLORYS runs for 2006 simulations). This additional connectivity indicates that enhanced spatial and temporal resolution would expose circulations that are five km and smaller in size that would further enhance connectivity to the southwestern region.

We stress that despite some differences, the particle pathways in CROCO and GLORYS are comparable, and so is the eddy statistics in the reanalysis and model simulation (Fig. S9), validating our hypothesis that at the latitudes of French Polynesia, mesoscale advection controls particle dispersal, with the Rossby deformation radius that sets the scale for the eddies being resolved by both datasets.

For the same two years (2006 and 2007) we also investigated if results were sensitive to the choice of fertilization dates by adding or subtracting two days from our chosen release period. We found that the exact date of fertilization following spawning had minimal impact on the overall potential connectivity.

Validation was also conducted by implementing an exponentially decaying PLD on the base of the maximum PLD recorded for coral species common around Moorea and Tahiti⁴⁸, instead of using conservative fixed values. The implementation of a decaying PLD results, as to be expected, in a decrease of potential probability of connectivity corresponding to the percentage of larvae removed by the fixed values considered. For example, in our case for a maximum PLD of 100 days, the total amount of viable larvae decreased to from 10,000 to 6,000 by week 5. The overall connectivity patterns, however, remained unchanged (Tables S2 and S3).

Discussion

Through modeling approaches, we investigated the connectivity potential of the coral reefs around Moorea and Tahiti to explore if their ability to regain coral coverage, which has been high and fast during the past 5 major bleaching events, could be linked to their ability to receive larvae from reefs distant enough to be less or not impacted by the same weather conditions, mesoscale eddies, or degree of warming. While water temperatures and other anthropogenic stressors impact bleaching levels, coral recruitment affects recovery potential. Even in areas that have been damaged by cyclones or stressed by overfishing or invasive species, increased coral recruitment improves recovery rates³³. Potential coral larval connectivity is also likely to promote re-population and maintain biodiversity²⁵.

By combining analysis of the mean variability around Moorea and Tahiti along with results from potential coral larval connectivity simulations, we quantified the relationship between the potential for larval dispersal and surface ocean currents. Mean currents are southwestward throughout the year with a latitudinal temperature gradient and higher temperatures in the northern regions, however, mesoscale circulations generate a substantial variability in both time and space. As the more northern atolls of the Tuamotu Islands tend to experience higher temperatures on average, they may have developed stronger resilience to warmer waters. This resilience could transfer, through the larvae, to the reefs of Moorea and Tahiti. In addition, the reefs in the northeastern atolls have experienced a less intense warming trend in the period considered (Fig. 1), despite being subject at times to temperature excursions similar to those recorded around Moorea and Tahiti (Fig. 3). These same reefs are less disturbed, low lying, and subject to minimal or no human influence.

We found that larval dispersal around Moorea and Tahiti can be fairly long-range (up to 250 km over just 21 days) and follows preferentially a southwestward path. However, transport to the two islands is highly variable and impacted by abundant eddies. As seen in our EKE and KE analysis, the ocean around Moorea and Tahiti is characterized by the presence of mesoscale eddies at all times. The eddy variability, which has not changed over the 28 years considered, modulates the probability of potential connectivity. The potential for larval dispersal appears to be impeded whenever eddy activity around Moorea and Tahiti is elevated, and vice versa, aided by fewer or weaker eddies (Figure 7A,B). Most importantly, there is a positive and strong correlation between probability of potential connectivity and kinetic energy. Enhanced KE, meaning stronger mean currents, leads to longer and stronger dispersal potential and greater connectivity with the northeastern atolls.

If most long-range virtual coral larvae that reach Moorea and Tahiti originate from islands to the northeast, mesoscale variability ensures that both the atolls to the southeast and those in the west may contribute potential larval transport in some years. Atolls in the far west may also have a non-zero connectivity by exchanging larvae with Moorea and Tahiti using the islands in Huahine (W1), Tupai, Bora Bora, Taha’a, Tumaraa (W2), and Maupiti (W3) as a bridge. This very dynamical setting further improves the recovery capabilities of Moorea and Tahiti by supplying diverse and far apart sources of larvae to re-populate the reefs. This potential long-range connectivity spans all decades considered and may have allowed for genetic similarities amongst islands and atolls throughout time. Sharing environmental conditions and genetic material also contributes to larval ability to settle and survive to maturity in distant reefs.

We focused on bleaching events specifically due to the records available and the fact that SST variations have a broad spatial footprint and show comparable impacts in Moorea and Tahiti. COTS outbreaks are also a periodic occurrence at both islands, but they can be localized to specific reefs, and we mostly have historic recordings around Moorea, thanks to the LTER network and the work at the CRIOBE research station. However, the dynamic and long-range potential connectivity identified in this work may have the ability to aid in recovery following COTS outbreaks as well.

The model set-up and resolution of the ocean currents used in this study were selected with the specific goal of capturing mesoscale advection and verifying whether long-range (i.e. evolutionary) connectivity may be important to maintain coral cover and genetic flow around Moorea and Tahiti^31,48. Self-seeding, on the other hand, is influenced by dynamics that are not sufficiently well resolved in our simulations, even in CROCO at 3 km horizontal resolution, and is likely underestimated in the connectivity calculations, especially in the backward runs.

It is worth noting that while connectivity across hundreds of kilometers may have played an important role in the corals reef recovery around Moorea and Tahiti to date, not all coral species have a PLD long enough to ensure connectivity over these distances^30,49. With frequent stressors and bleaching events occurring every 4–7 years¹⁵, the recovery of all species is near impossible, so coral coverage may return close to previous values, but biodiversity may be lost nonetheless. Additionally, elevated connectivity potential is necessary but not sufficient to ensure re-population of reefs. Even if coral larvae can reach Moorea and Tahiti, settlement happens and leads to survival only under favorable conditions, and chances will decrease in highly stressed or frequently bleached reefs.

Lastly, our study has limitations worth noting. First, the PLD in this work was considered constant, between 3 and 5 weeks, but in reality the PLD decreases when water temperatures increase, because larvae tend to grow faster⁵⁰. The same behavior (faster growth) affects the COTS, with the end results that warming may decrease coral connectivity while promoting COTS outbreaks⁵¹. These scenarios were not explored. Second, we adopted a reanalysis product to explore connectivity over decadal time-scales. Its advantages include a reliable representation of the mesoscale variability and the assimilation of all satellite and in-situ data. However, it has a coarse horizontal and temporal resolution relative to the size of reefs and the atolls. Higher resolution (3 km) and frequency (2 h) dynamics do not eliminate the possibility of long-distance connectivity, but models at even higher resolution should be considered when focusing on specific reefs instead of regional connectivity (e.g.^52,53). Higher resolution simulations will also allow for exploring larval behavior, neglected here given the approximations made in the physical fields and the space and time scales considered. Finally, we did not consider adaptation among the contributions that may aid recovery. Research suggests that some coral species around Moorea may be developing more heat-tolerant characteristics. For example, by comparing bleaching events in the early 2000’s, a study found a slight decrease in the levels of coral bleaching in later years¹⁹.

In conclusions, we show that long-range (order hundreds of kilometers) and highly variable—in both time and space—larval connectivity might be possible around Moorea and Tahiti, and may have contributed to relatively rapid repopulation of the reefs after each bleaching event. Genetic analysis among same-species corals within the domain considered are needed to validate our simulations . If the long-range and highly dynamic connectivity is confirmed, it could be enhanced by accounting for preferential routes of exchange of genetic material while expanding existing, or establishing, new MPAs. For example, it would be wise to protect reefs that are sources of larvae to many others rather than those that are sinks, have more strict levels of protections for regions that are unlikely to benefit from inflows of diverse larvae from far away but have to rely on self or nearby recruiting strategies instead, and ease restrictions for those that benefit from a higher biodiversity score as indicated by their connectivity potential. While this is not an easy solution, as effective MPAs are monitored and most atolls in French Polynesia are very remote, it is still a possible approach worth exploring^54–56. By enhancing protections on reefs that may act as sources for other regions, repopulation can further be promoted in distant reefs that may rely on these sources for recovery. These adjustments could enhance the genetic material shared amongst reefs and ease resilience and restoration efforts within stressed and degraded reef ecosystems.

Methods

Data and analysis

Potential connectivity of coral reefs around Moorea and Tahiti is analyzed through Lagrangian simulations of larval dispersal using ocean currents from reanalysis data. The reanalysis chosen for this study is the Global Ocean Physics Reanalysis (GLORYS12V1, accessed through Marine Copernicus Services,³⁶). GLORYS12V1 assimilates available satellite information (along-track altimeter sea level anomaly, satellite sea surface temperature, and sea ice concentration), in addition to in-situ temperature and salinity profiles. Both SST and surface current have a spatial resolution of $1/{12}^o$ ($\sim$9 km) and daily temporal resolution. The period considered is from 1993 through 2020.

Initial analysis of our study site included a spatial and temporal investigation of SST and current anomalies. SST and current anomalies were calculated by removing the monthly mean seasonal cycle from each data point. Sea surface currents were also used in calculating kinetic energy (KE) and eddy kinetic energy (EKE) using Eqs. (1) and (2) respectively, where $v^{\prime }$ and $u^{\prime }$ represent meridional and zonal velocity anomalies and v and u represent the total meridional and zonal velocities.

$$\begin{array}{cc}\hfill KE=& \frac{1}{2}(v^2+u^2)\hfill \end{array}$$ 1

$$\begin{array}{cc}\hfill EKE=& \frac{1}{2}(v^{\prime 2}+u^{\prime 2})\hfill \end{array}$$ 2

$$\begin{array}{cc}\hfill \zeta =& \left(\frac{\delta v}{\delta x},-,\frac{\delta u}{\delta y}\right)/f\hfill \end{array}$$ 3

$$\begin{array}{cc}\hfill E=& \frac{1}{2}{(\zeta )}^2\hfill \end{array}$$ 4

EKE quantifies the variability in ocean currents due to mesoscale circulations, while KE measures the overall amount of energy near the ocean surface. EKE uses the perturbations in the u and v components of current velocity to help visualize the amount of variability seen in mesoscale eddies over time. This measurement is then used to see if there is any seasonality or pattern in the timeseries of eddies for our given region.

Additionally, vorticity and enstrophy were evaluated during the spawning season for the 28-year period considered. Vorticity, $\zeta$, and enstrophy, E, are calculated using Eqs. (3) and (4) respectively, where f is the Coriolis parameter, and x and y represent longitude and latitude. Vorticity is a vector and describes the local spinning motion of water parcels, while enstrophy is a scalar quantity that quantifies the strength of vorticity⁵⁷ by considering its square. Vorticity concentrates in eddies and mesoscale fronts, and enstrophy identifies these structures without a vector implication. In this study, positive vorticity is indicative of clockwise rotation. In connecting enstrophy to potential connectivity, enstrophy was quantified in the smaller region around Moorea and Tahiti shown in Fig. 1.

Connectivity simulations

In order to simulate larval transport, we used Ichthyop, a publicly available Langrangian tool to model the physical (and biological) effects of a fluid environment on plankton and larvae (https://github.com/ichthyop). Here, we perform simulations covering the Acropora coral spawning season, loosely defined as a three to five week window from the middle of October into November (Fig. S2B,D). Acropora, likely the most abundant sessile coral genera around Moorea and Tahiti, was found to spawn about 7 days after the (end of) September to November full moons⁵⁸. Another coral genera abundant in French Polynesia is Pocillopora, which only very recently has been found to spawn 2–3 days after the full moon in November and December 2022⁵⁹, so within or a month after the interval considered. Simulations were run based on mass spawning dates which were calculated as being 7–10 days following a full moon in October or November of each respective year. We chose to run simulations on these dates since this is when the largest number of coral species synchronously mass spawn^58,60–62. However, our investigation is not seasonal or month dependent and could apply to the dispersal of other biological material during other periods of the year, given that the seasonality of currents in the area is extremely small. Simulated coral larvae were released near the ocean surface and integrated both forward and backward in time. We therefore considered both Moorea and Tahiti as a source (forward runs) and sink (backward runs) of coral larvae and genetic material.

Additionally, simulations assumed a fixed PLD of three, four and five weeks following research by^31,62–68 and, most recently⁴⁸. Results from these works find that varying species of Scleractinia and the genus of Acropora, which are commonly found in the south tropical Pacific, have enhanced PLD up to 209 days, but with the highest settlement success prior to 45 days following spawning. Additionally, various Indo-Pacific broadcast spawning species have been found to be competent to settle beyond 70 days showing a more complex temporal settlement behavior than previously believed⁴⁸. Our choice of fixed PLD is therefore conservative. For all simulations, 10,000 virtual larvae were released in a randomized fashion around the coastlines of Moorea and Tahiti (Fig. 2). The use of a fixed PLD is common in the modeling community (e.g.⁶³), because the exact functional form is not known and is likely to be species and time dependent (it is likely to vary with water temperatures, for example). While a fixed PLD will not fully capture the complete complexity of coral larval settlement, it still allows us to answer if long-range connectivity is possible for reefs around Moorea and Tahiti for reasonable larval durations. We validated our results by also adopting a more sophisticated function to describe the decrease of viable larvae over time (see Validation further below).

Several approximations are made in this study. First of all, given the kilometer-scale resolution of our advective fields, a virtual larva cannot exactly portray an actual larva. It can, however, be interpreted as a region of water, a few hundreds of meters by a few hundreds of meters in scale, containing a large number of larvae because of the distances considered and models used. Secondly, coral larvae are characterized by different behaviors and developmental stages depending on the species and habitat considered (e.g.⁶⁹), and are subject to dilution. However, little is currently known about these behaviors and dilution effects in the study area at the scales considered. Consequently, following on from previous literature, we did not introduce specific behaviors such as diel vertical migration, random walk subgrid-scale diffusivity, or dilution. The present analysis focuses on the larval transport under the assumption that coral larvae are passively transported by near-surface oceanic currents. Our approach should be viewed as a first step that sets the basis for future research.

Forward runs simulate the dispersal of larvae on the date of each spawning event, therefore considering Moorea and Tahiti as larval sources. In contrast, backward runs release particles at the same location, but starting 21, 28, and 35 days after the spawning event and advect them backward in time (Table S1). For example, if the spawning date is October 10th, the three week release starts on October 31st, the four week release on November 7th, and the five week release on November 14th, with the virtual larvae being integrated backward in time and ending always on October 10th. The ecoregions that particles reach throughout the backward simulations indicates potential larval sources for Moorea and Tahiti.

We also performed forward simulations in which we released virtual larvae in the region to the west of Moorea and Tahiti indicated as W1, W2, and W3 in Fig. 2. These runs had a similar structure to the other simulations. Their purpose was to see if these western ecoregions may act as a bridge connecting Moorea and Tahiti to the ecoregions further to the west (i.e., W4, W5, and W6).

From all backward and forward runs, connectivity was then calculated by counting the number of virtual larvae that reached different ecoregions during their life span. Average probability of potential connectivity was calculated for each case by averaging the total amount of particles released at Moorea and Tahiti that reached a different ecoregion and dividing by the number of particles released from the two original islands. Finally, absolute dispersion D was calculated using Eq. (5), where $x_i$ and $x_o$ are the final and initial longitudinal positions, $y_i$ and $y_o$ are the final and initial latitudinal positions, and N is the total number of particles. Absolute dispersion was used to determine the correlations between dispersion and potential connectivity, enstrophy (E, quantifying the importance of eddies and front as trapping and advecting features), and KE.

$$\begin{array}{c}\hfill D=\left(\sum _{n=i}^N,\sqrt{{(x_i-x_o)}^2+{(y_i-y_o)}^2}\right)/N\end{array}$$ 5

Validation

In this work our key hypothesis is that mesoscale currents may be responsible for long distance connectivity at Moorea and Tahiti. We must therefore validate the representation of these currents in the selected dataset. Validation of the GLORYS12V1 currents was conducted using the MEaSUREs Gridded L4 Sea Surface Height Anomalies accessed through the cloud enabled data services⁷⁰. We used these observed sea surface heights (SSH) to calculate velocity maps to compare with GLORYS12V1 velocity maps. We found that GLORYS12V1 accurately represent the mesoscale variability contained in the MEaSUREs data for various years within our study period (Fig. S9)

Regional simulations

The time and spatial scales relevant to mesoscale advection are well captured by GLORYS12V1 resolution (daily and about 9 km, respectively). At the time scales of interest for larval connectivity (several weeks) horizontal dispersion is controlled by mesoscale currents, with a dominant nonlocal, long range impact due to mesoscale eddies when present^71,72. Circulations at smaller scales (kilometer scale or smaller, so-called submesoscales) do not play a significant role on the overall range of lateral transport, while being important locally for horizontal mixing and at all times for vertical exchanges (e.g.³⁵). To confirm that this applies to our region of interest and validate the ability for GLORYS to capture mesoscale variability and its impacts on particle dispersal, we modeled currents at higher resolution using the Coastal and Regional Ocean Community (CROCO) model⁷³, which is a publicly available ocean modeling system build upon ROMS (Regional Ocean Modeling System). CROCO, in its hydrostatic version, was configured within the domain spanning from ${13}^{\circ }$S to ${26}^{\circ }$S and from $156.5^{\circ }$W to ${130}^{\circ }$W, with open boundaries on all four sides. The rectangular grid, with a horizontal resolution of 3 km and 35 terrain-following levels in the vertical axis with most of the vertical resolution concentrated in the upper 300 m of the water column, was built using the General Bathymetric Chart of the Oceans (GEBCO) with a Shapiro smoother of 0.3 applied to limit pressure gradient errors. CROCO was forced using wind stress, heat, and water flux data from ERA5⁷⁴, while the boundaries were nudged to the GLORYS12V1 Reanalysis³⁶. The model was initialized on January 1st 2006 using GLORYS12V1 data and run for two years, discarding the first six months. The model configuration included the most important tidal components from the TPXO-7 global tidal model and the 3D velocity fields were stored every two hours resolving the diurnal cycling⁷⁵.

The currents modeled by CROCO were then used to backtrack the Lagrangian tracers using Ichthyop in a similar fashion to what done with GLORYS12V1 data in the fall of 2006 and 2007. Five runs were conducted for PLDs of 3, 4, and 5 weeks in each year varying their starting date by 2 days (and the same was done for GLORYS) in order to estimate possible sources of larvae and genetic material for Moorea and Tahiti. The ensemble of five runs was performed to partially account for the differences in mesoscale circulation between CROCO and GLORYS. Potential connectivity was also calculated in the same way as done in the original analysis.

The outcome of these simulations is discussed in the Results section.

PLD representation

We tested a more realistic functional form for the PLD assuming a maximum PLD ($PD{L}_{\mathit{Max}}$) of 100 days based on⁴⁸ analysis, and an exponential decay of the number of viable larvae (i.e., still alive) over time following the modelling of broadcast spawning coral in⁷⁶. In this case the percentage of surviving and competent larvae P at time t is given by:

$$\begin{array}{c}\hfill P(t+1)=P(t)e^{-\lambda t}\end{array}$$ 6

where $\lambda =ln(2)/(PL{D}_{\mathit{Max}}/2)$ is the mortality decay rate. Survival was assumed to be 100% for the initial 10 days of the simulation with mortality beginning to affect larvae on day 11. Ichthyop simulations of larval transport in 2006 and 2007 (as used in our regional validation) were then post-processed using this exponentially decaying PLD. Prior to adding a mortality decay rate, particle positions were randomized to ensure that decay in larval viability would not depend on the initial particle positions. Following post-processing, the potential probability of connectivity was then calculated as for the other cases.

Author contributions

SJL performed the CROCO forcing-data preprocessing, Lagrangian simulations, and conducted the analysis and validation for this study. LL set-up and performed the CROCO simulations. SJL and AB drafted and contributed to each aspect of the writing and interpretation of results for the manuscript in full. All authors read and approved the submitted manuscript.

Data availability

The dataset analyzed for this study can be found at Copernicus Marine Services (https://data.marine.copernicus.eu/product/GLOBAL_MULTIYEAR_PHY_001_030). Additionally, the Lagrangian modeling tool, Ichthyop, used for simulations can be found at https://ichthyop.org/. CROCO data used for validation can be accessed through DropBox at https://www.dropbox.com/scl/fo/qe5ayos7otyqpf2f2l5kq/AHyoCpiJZI_sKhTSkx3b6xM?rlkey=u1vrwnoog26xmconp7wibyltr&dl=0. Further inquiries regarding data availability can be directed to the corresponding author, SJL.

Competing interests

The authors declare no competing interests.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-024-73185-2.