Reef manta rays (Mobula alfredi) are filter feeders that utilise their specialised cephalic lobes to funnel water into their mouths, primarily consuming pelagic and demersal zooplankton (Dewar et al., 2008; Last and Stevens, 2009; Couturier et al., 2012; Couturier et al., 2013; Peel et al., 2019a; Harris et al., 2023). Their foraging behaviour is closely associated with dense zooplankton aggregations (Armstrong et al., 2016; Armstrong et al., 2021), which are necessary to meet the energetic demands of this large species with its high metabolic rate (Meekan et al., 2015). To optimise feeding efficiency and balance energy expenditure, M. alfredi often target zooplankton blooms where prey densities surpass a critical threshold. These events are ephemeral and influenced by dynamic environmental processes (Couturier et al., 2011; Weeks et al., 2015; Armstrong et al., 2016; Armstrong et al., 2021).

Although research on M. alfredi has advanced the understanding of their feeding ecology (Couturier et al., 2013; Peel et al., 2019a; Harris et al., 2023), the dynamics of zooplankton availability, an important factor influencing the residency, movement, and foraging behaviour of manta rays, are not fully understood (Armstrong et al., 2021). A significant gap still exists in understanding their feeding ecology, including dietary composition, spatial and temporal patterns in feeding and the environmental conditions that drive aggregations. This knowledge is essential for designing effective conservation regulations for this threatened species and its habitats. Since this species has a wide circumglobally distribution (Marshall et al., 2022), further investigations across a broader range of locations, habitat types, and environmental conditions would enhance knowledge of its feeding ecology and support the development of targeted conservation strategies. Research conducted at Lady Elliot Island in Australia (Armstrong et al., 2016) and Hanifaru Bay in the Maldives (Armstrong et al., 2021) revealed that although M. alfredi tended to feed in areas with high zooplankton densities, however, zooplankton accumulation varied under different conditions at the two locations. In Australia, zooplankton densities were higher around the low tide, while in the Maldives, they peaked around high tide. The prey density threshold (PDT), a level beyond which the probability of feeding increases significantly, was identified as 11.2 mg/m³ in Australia and 53.7 mg/m³ in the Maldives. Analysis of the zooplankton composition in Australia showed that neither size nor species composition affected feeding behaviour, highlighting zooplankton accumulation as the primary factor influencing feeding behaviour rather than composition. However, in the Maldives, composition significantly influenced feeding behaviour, with M. alfredi showing a preference for calanoid copepods over gelatinous taxa such as chaetognaths or eggs (Armstrong et al., 2016; Armstrong et al., 2021). These variations could be attributable to differences in habitats across geographically distinct locations, driving scientific interest in various sites and habitat types.

D’Arros Island is utilised by M. alfredi year-round and hosts the largest known aggregation in the Seychelles. In contrast, the species is less frequently observed around other islands throughout the region (Peel et al., 2019a; Peel et al., 2019b; Peel et al., 2024). In 2023, the surrounding waters of D’Arros Island were officially gazetted as a Marine Protected Area (MPA) with the highest level of protection (Zone 1 – Marine National Park under the Seychelles Marine Spatial Plan Initiative (Official Gazette No 34 – Ministry of Agriculture, Climate Change and Environment, 2020)), encompassing the known foraging grounds located close to the island. Before finalising the site-specific management plan, science-based research is required to inform appropriate regulations for the effective management of both the species and its habitat. Understanding factors such as preferred feeding areas and the conditions that drive M. alfredi aggregations can help identify key feeding sites and guide regulations for recreational and commercial motorised activities. This study contributes to ongoing efforts and strengthens the existing knowledge of M. alfredi feeding ecology.

This study aimed to understand the relationship between plankton biomass and feeding behaviour of M. alfredi to provide insights into the key environmental drivers that influence its foraging activities. This knowledge will aid in optimizing research and conservation efforts. Quantifying the PDT for M. alfredi at this site provides valuable information in understanding the critical point at which plankton biomass levels start influencing feeding behaviour. By comparing this threshold to the only two others that have been calculated for M. alfredi in Australia (Armstrong et al., 2016) and the Maldives (Armstrong et al., 2021), regional differences and similarities in feeding behaviour, prey and habitat requirements can be assessed. Such comparisons are crucial for contextualizing study findings. Lastly, although M. alfredi are present year-round, observations of seasonal peaks in their abundance between July and December suggests potential seasonal variations (SOSF-DRC unpublished data). To explore this further, it is important to investigate whether these peaks in M. alfredi abundance are correlated with variations in prey availability. This approach will help to identify patterns of prey availability like those observed in the Maldives, where a more pronounced seasonality in both food supply and M. alfredi numbers has been documented (Harris et al., 2020). Understanding these seasonal dynamics will not only allow for more accurate predictions of M. alfredi visitations but also provide valuable insights into other life-history traits, such as reproduction and migratory patterns. Regarding its conservation value, this research will inform the development of rules and regulations for the D’Arros Island marine protected area, which are currently being drafted. Within the management plan, information acquired from this study on patterns in surface feeding of reef manta rays will directly influence the placement of speed restriction zones.

A total of 89 plankton samples were collected (Feeding - 39, Non-feeding – 10 and Absent 40) from November 2022 to March 2024 encompassing both the North-west monsoon and South-east monsoon. Supplementary Table S1 shows the dates of sampling, number of mantas sampled on each date, their behaviour, and the plankton biomass recorded on that date.

GLM for feeding behaviour showed that plankton biomass had a significant relationship with M. alfredi feeding behaviour, whereas all other variables did not (Table 1; Figure 2). Four models contributed to the top two AICc units and eight models contributed 95% evidence weight (see Supplementary Table S2). The model with the greatest support was that which included biomass, wind average and moon phase as predictors. A PDT of 26.9 mg/m³ was estimated (Figure 3), representing the minimum plankton density required for M. alfredi to feed efficiently. Plankton biomass exceeded this threshold value during surface feeding events compared to non-feeding instances, which did not (Figure 2). Mean biomass during feeding = 31.65 mg/m³ ± 4.48 SE whereas mean biomass when non-feeding =13.16 mg/m³ ± 2.35 SE.

The GLM with plankton biomass as a response variable showed a significant relationship between plankton biomass and tide (Table 2). Two models contributed to the top two AICc units and four models contributed 95% evidence weight. The model with the greatest support was that which included wind average, moon and tide as predictors (see Supplementary Table S3). Feeding was mostly observed on the northeast side of the island with the exception of two observations further to the west and south (see Figure 1). M. alfredi are infrequently found on the southern and western side of D’Arros, resulting in the limited sampling in that area. Consequently, a significant proportion of sampling was conducted along the northeastern coast of D’Arros with > 50% of all feeding behaviour observed in a 300 m stretch along the northern coastline (highlighted with yellow box in Figure 1). Biomass data (max values) from this location were presented across the tidal cycle for visualisation (Figure 4). Plankton biomass surpassed the critical PDT for M. alfredi feeding described above in the time between two hours before to five hours after the high tide. Some samples were collected during these tidal windows without any M. alfredi observed feeding. Moreover, no M. alfredi feeding was observed outside of this part of the tidal cycle.

Despite known seasonal fluctuations in M. alfredi abundance at the site and several instances of higher plankton biomass recorded during the south-east monsoon, season had no statistical effect on M. alfredi feeding and plankton biomass.

Findings from this study provide new insights into the feeding ecology of M. alfredi. The correlation between plankton biomass and manta feeding behaviour reported here aligns with previous studies that have indicated the requirement of dense plankton availability for M. alfredi feeding (Armstrong et al., 2016; Armstrong et al., 2021). Higher plankton biomass during feeding occurrence likely reflects M. alfredi’s need to target periods of increased prey abundance to meet its energetic demands (Meekan et al., 2015). Similarly, according to Armstrong (Armstrong et al., 2021), instances where plankton density was high, but feeding was not observed could be attributed to the composition of the zooplankton, with gelatinous species dominating certain samples, which was found to be avoided by M. alfredi in the Maldives (Armstrong et al., 2021). Due to time constraints, zooplankton composition, which requires extensive analysis, was excluded from this study; however, exploration of these data in future would help to address this.

In comparison to the PDT of 26.9 mg/m³ quantified around D’Arros Island, research conducted in Australia estimated a PDT of 11.2 mg/m³ (Armstrong et al., 2016), while research in the Maldives showed a substantially higher PDT of 53.7 mg/m³ (Armstrong et al., 2021). Such variations in PDT could be driven by adaptation to regional differences in food availability, related to habitat-specific features. PDT may therefore be linked to the size of a population using a given site, since food availability within a habitat can affect the population size of a species using it (Sibly and Hone, 2002; Edwards and Edwards, 2011). For example, the study in the Maldives, home to the largest recorded population of M. alfredi (~5,000 individuals), found higher PDT at Hanifaru Bay (53.7 mg/m³) than Seychelles (26.9 mg/m³, ~500 individuals) and Australia (11.2 mg/m³, ~1,000 individuals) (Armstrong et al., 2016; Armstrong et al., 2021). Alternatively, some sites may be preferentially selected as suitable not only for foraging but also for other key life history processes and that feeding a lower density thresholds allow M. alfredi to remain in areas important for these processes. Cleaning is an important behaviour in M. alfredi, and use of the cleaning stations at D’Arros has been shown to be an important driver of presence (Peel et al., 2024) with recent observations evidencing repetitive movements between feeding and cleaning (Newsome et al., 2024). Further research into the metabolic requirements of M. alfredi, potentially with the use of acceleration data loggers (Whitney et al., 2007; Armstrong et al., 2016), may also help to better define the underlying drivers of differences in prey density thresholds between different populations. However, these findings open opportunities for further research to better understand feeding thresholds both within and across geographically distinct populations.

Tide had the strongest correlation with zooplankton biomass, a previously known driver for plankton accumulation along coastlines (Alldredge and Hamner, 1980), and consequently M. alfredi feeding behaviour. The tidal trend reported here is similar to that reported from the Maldives (Armstrong et al., 2021), Indonesia (Dewar et al., 2008), Lady Elliot island eastern Australia (Armstrong et al., 2016), and the Chagos Archipelago (Harris et al., 2021). Tidal related changes in coastal currents, which transport zooplankton to nearshore areas, could influence the timing and locality of plankton accumulation. Slack periods over tidal cycles can create “quiet regions” which increases localised zooplankton abundance (Alldredge and Hamner, 1980). Studies from estuaries may suggest some explanations, especially the influence of slack tide on increased zooplankton abundance around the high tide. Vineetha et al. (2015) determined that periods of slack may allow zooplankton to carry out effective vertical migration in shallow areas, which they cannot do with tidal movements (Vineetha et al., 2015). Fine-scale spatial differences in the topography of D’Arros may influence surface accumulation of plankton. The shallow bank (<10m) on the northeastern side of D’Arros Island, where most feeding was recorded, may act as a buffer, with tidal movements and currents pushing zooplankton up the steep reef slopes adjacent to the shallow bank and to the surface. Not all samples collected during the peak around high tide observed M. alfredi feeding, suggesting that other factors such as plankton composition may be influencing surface abundance patterns. The topography and oceanic processes affecting the northeastern side of D’Arros, where M. alfredi tend to feed, could be further investigated to help better understand this spatial preference in comparison to the southwestern part of the island. For example, understanding how tides interact with other environmental drivers and geographical features to bring plankton to the surface, thereby supporting feeding in M. alfredi.

Since both GLM’s showed no significant seasonal differences in plankton biomass and feeding behaviour, this suggests that the variation in M. alfredi abundance throughout the year is not related to seasonal changes in food availability. However, the consistent year-round availability of prey may support the species throughout the year, highlighting the importance of this site for M. alfredi. This is supported by previous research which indicates that M. alfredi is present and feeding year-round at D’Arros Island (Peel et al., 2019b; Peel et al., 2020; Peel et al., 2024). By ruling out changes in plankton biomass throughout the year as a key factor for presence and absence of individuals, it creates opportunity for further research on the fluctuations in M. alfredi abundance and the habitat requirements for this species. For example, seasonal changes in M. alfredi numbers might be due to behavioural factors or life history traits, such as reproduction rather than feeding, as indicated by seasonal observations of courtship and mating behaviours at the site (SOSF-DRC unpublished data). The migration of females to other locations could also explain observed seasonal fluctuations (Peel et al., 2024). Average wind (which is stronger and more consistent in the south-east monsoon season) showed only a weak positive relationship to plankton biomass and no relationship with M. alfredi feeding behaviour which contrasts with findings from the Maldives where M. alfredi abundance has shown a strong relationship with wind (Harris et al., 2020). This could relate to the more pronounced seasonality in M. alfredi abundance at that study site, when compared to D’Arros.

These findings have direct conservation application in supporting the protection of feeding reef manta rays, which can be susceptible to boat strike injuries in coastal habitats (Strike et al., 2022). The D’Arros National Marine Park is a recently designated and gazetted Zone 1 marine protected area for high biodiversity protection that came into legal effect on 31st March 2025. The Save Our Seas Foundation represents the management authority for the D’Arros site and is responsible for developing site-specific management measures for this MPA. Data collected for this study should inform the placement of speed restriction zones, particularly in the the northeastern area of D’Arros in order to mitigate boat strike injuries.

There were several limitations to the study design which should be noted. Sampling efforts fluctuated across the year due to logistical constraints, such as unfavourable weather, meaning the sampling was not conducted throughout January and February 2024 and this could have hindered interpretation of seasonal effects on both M. alfredi behaviour and plankton biomass accumulation. GLM’s are useful to identify the drivers influencing plankton biomass and ultimately feeding behaviours. However, assessing the overall fine-scale interaction between a combination of drivers and factors, such as prey availability and their shifts, using traditional modelling approaches can be challenging and sometimes impractical. This is because important components such as topography and oceanographic processes, which are highly dynamic and can strongly influence how plankton accumulates (Richardson, 2008) are often excluded. It is recommended that future research better incorporates an understanding of spatial plankton dynamics at a fine-scale. To address this and integrate fine scale elements, a computer-based parameterization model could be developed to simulate plankton movement patterns around D’Arros Island and visually interpret the findings. Such approaches have previously been used to model oceanographic drivers of coral connectivity on a regional scale (Burt et al., 2024). This model would incorporate all additional suspected drivers and factors such as ocean currents, tides, wind patterns, water temperature, topography, and other environmental parameters collected over an extended period. Such a model could help predict plankton accumulation in specific areas and once calibrated to the known PDT at which M. alfredi typically begin feeding, generate predictions for M. alfredi visitation and visualize the interaction of these drivers. To date, no such model exists for this specific application, but might be worth pursuing and could be a valuable area of focus for future research.

Overall, this research showed that feeding behaviour in M. alfredi at D’Arros Island was significantly related to plankton biomass, which, in turn, was strongly influenced by tidal phases. A PDT was identified, which differs from those identified in two previous studies, and no seasonal shifts in feeding behaviour or plankton biomass were observed. The findings from this study will actively support management efforts for the D’Arros Island marine protected area by identifying key feeding areas and temporal patterns that drive M. alfredi surface feeding aggregations. From this it should be possible to demarcate specific coastal areas as a critical feeding ground within the MPA and to establish regulations, particularly regarding motorised activities that may disturb or injure individuals. The findings will also aid in predicting M. alfredi visitation patterns, improving research efficiency and long-term monitoring programs.