Introduction

Globally, many flowering plant species rely on fruit-eating animals to swallow and disperse their seeds (endozoochory) (Herrera 2002). Such removal and movement of seeds away from the parent plant is a key process in the plant life cycle and vital for both ecosystem resilience and functioning (Howe and Smallwood 1982; Wang and Smith 2002). Seed dispersal is vital for plant fitness and population persistence as seeds that fall under parent plants are unlikely to be successful (Howe and Smallwood 1982; Wang and Smith 2002; Kendrick et al. 2012). Additionally, animal-mediated seed dispersal can impact the ability of a plant population to cope with or escape local environmental changes (Travis et al. 2013; Beckman et al. 2019). Thus, the disruption of such mutualisms, through the loss of animal seed dispersers, could have far-reaching consequences for plant species reproduction and survival, plant community structure, and the overall healthy functioning of ecosystems (Dunn et al. 2009; Aslan et al. 2013).

For large-seeded plant species, the risk of losing dispersal services may be particularly dire, due to the increased extinction risks faced by many large-bodied animals (Dirzo et al. 2014; Pérez-Méndez et al. 2014). These large-bodied frugivores play a key role in the dispersal of large-seeded plant species, as seed size restricts who can swallow seeds for endozoochorous dispersal. The size-biased loss of frugivore mutualist in defaunated sites has been linked with reductions in fruit removal (Holdbrook and Loiselle 2009; Wotton and Kelly 2011), dispersal distances (Pérez-Méndez et al. 2016), and recruitment (Nunez-Iturri et al. 2008; Albert et al. 2021). These size-biased losses may also have associated effects on the genetic composition and spatial structure of large-seeded plant populations (Pérez-Méndez et al. 2016; Carvalho et al. 2016), and on the rapid evolution of smaller seeds in a plant community (Galetti et al. 2013). Even when multiple dispersal mechanisms exist, the loss or reduced abundance of an animal mutualist may mean the loss of its unique contribution to the removal, handling, and movement of seeds, altering dispersal patterns and outcomes (Peres and Roosmalen 2002). Thus, identifying and characterizing the dispersal mechanisms that contribute to the movement of large seeds can provide important insights into the long-term dynamics of plant populations, allow us to assess the risks posed by mutualist extinction, and develop strategies for conservation and management.

While it is certain that endozoochory by large-bodied vertebrates plays a substantial role in the ecology and evolution of large-seeded plant species, it is unlikely to be the sole dispersal mechanism. Other mechanisms may include primary dispersal by dispersers that carry instead of swallowing seeds (synzoochory & stomatochory), secondary dispersal by ground-dwelling animals, or dispersal by strong winds, runoff or flooding associated with cyclones (Guimarães et al. 2008; Loayza et al. 2014; Federman et al. 2016). For example, seasonal runoff and flooding have been found to play an important role in the dispersal of some plant species in tropical ecosystems (Ganzhorn 1995; Guimarães et al. 2008). The extreme winds associated with tropical storms could also carry and disperse large fruits, despite a lack of specific adaptations for wind dispersal (i.e., wings, plumes, and low fruit mass) (Howe and Smallwood 1982). Beyond these abiotic dispersal mechanisms, frugivores like fruit bats are known to carry seeds much larger than they can swallow, while native or introduced scatter-hoarding rodents may support secondary dispersal of seed fallen under tree canopies (i.e., spatially scattering single seeds, buried in soil or under a pile of leaf litter) (Forget 1990; Wilson et al. 2007). In a Neotropical rainforest system, for example, rodents were found to be responsible for the dispersal of large-sized palm seeds through scatter-hoarding (Jansen et al. 2012). However, for many large-seeded plant species, the contributions of these dispersal mechanisms remain a hypothesis and further study is necessary to resolve their involvement and quality.

Madagascar’s ecosystems present a compelling avenue to investigate the contribution of multiple mechanisms to the dispersal of large seeds because (i) the majority of plant species in Madagascar have traits adapted for seed dispersal by animals (Razafindratsima 2014; Albert-Daviaud et al. 2018); (ii) many of these large-seeded plant species are presumed to be at risk from the local loss of large-bodied frugivorous lemurs (Federman et al. 2016; Albert-Daviaud et al. 2020); and (iii) Madagascar’s conservation challenges place the island’s largest seed dispersers, the lemurs, under high extinction risks (Razafindratsima et al. 2022). In this study, our overarching goal is to investigate the dispersal mechanisms that contribute to the reproduction of a large-seeded tree genus in Madagascar’s rainforests, Canarium spp. Engl. (Burseraceae). Canarium spp. are long-lived canopy trees and bear single-seeded fruits. With large seed sizes across the genus (fruit length approximately 2–5 cm; Federman et al. 2016), it is thought that many Canarium species in Madagascar may have only been dispersed by now-extinct megafauna (i.e. Babakotia, Hadropithecus, Pachylemur and Paleopropithecus species) and some of the largest of extant lemur species, namely those of the Varecia genus (Federman et al. 2016). However, observations of feeding ecology on extant frugivores have reported that the fruits of Canarium species are consumed and dispersed by other lemur species, namely the red-fronted lemur (Eulemur rufifrons), black lemur (E. macaco), Sanford’s brown lemur (E. sanfordi), black-and-white ruffed lemurs (Varecia variegata editorum), and red-ruffed lemurs (V. rubra) (Birkinshaw 1999; Bollen et al. 2004; Donati et al. 2007; Moses and Semple 2011; Wright et al. 2011; Razafindratsima and Martinez 2012; Razafindratsima et al. 2014; Chen et al. 2016). Nevertheless, the endangered status of most lemurs brings the question of which other dispersal mechanisms are available to this genus in Madagascar, and how these mechanisms contribute to its regeneration, as compared to the services that may be lost. In other tropical systems, Canarium species were found to be consumed and dispersed by hornbills, squirrels, macaques, and mouse-deer, and similar dispersal mechanisms may be present in Madagascar (Yasuda et al. 2005; Naniwadekar et al. 2021; Gopal et al. 2021).

To address our goal, we investigated the dispersal of Canarium trees in rainforest sites within and adjacent to Ranomafana National Park in Madagascar. Individuals of this genus are present within habitat patches that include different compositions of their most likely local dispersers (V.v. editorum, E. rubriventer, and E. rufifrons). Varecia variegata editorum, the largest of the three species, is not present in some of our study sites due to habitat degradation and hunting pressure. Other vertebrates that may act as seed dispersers in this ecosystem include the greater dwarf lemur (Cheirogaleus major), the brown mouse lemur (Microcebus rufus), the milne-edward’s sifaka (Propithecus edwardsi), the greater and lesser vasa parrots (Coracopsis vasa and C. nigra), red forest rat species (Nesomys sp), the crested drongo (Dicrurus forficatus), the blue coua (Coua caerulea), the Malagasy bulbul (Hypsipetes madagascariensis), the Chabert vanga (Leptopterus Chabert), and the velvet asity (Philepitta castanea) (Rakotomanana et al. 2003; Razafindratsima 2017; Razafindratsima et al. 2022). The role of these species in the dispersal of Canarium seeds has not been confirmed. This ecosystem is also exposed to frequent cyclones, which may facilitate seed movement through strong winds and flooding. To clarify the role of various dispersal mechanisms in the movement of Canarium seeds and assess the risks it might face from the threatened loss of its largest-bodied local disperser, we sought to answer the following research questions: Q1. How might winds and flooding, associated with cyclones, contribute to Canarium dispersal? Q2. What animal species could effectively contribute to their primary and secondary seed dispersal? Q3. How do non-lemur animal and abiotic dispersal mechanisms compare to the imperiled function of extant lemur primary dispersers?

We expected that strong winds could remove fruits from the canopy and flood waters may carry fallen fruits away from the parent trees. Although V. v. editorum is the only known disperser of Canarium in this forest (Razafindratsima et al. 2014), records of Eulemur species eating the fruit at other sites indicate that it may be the case here too (Birkinshaw 1999; Chen et al. 2016). In addition, scatter-hoarding native and non-native rodents and other ground-dwelling vertebrates (e.g., granivores and seed predators) could also disperse the fruits fallen from the parent trees. We also expected that the dispersal by these secondary dispersers would be effective in ensuring long-distance movement away from the parent tree. We expect that non-lemur dispersal mechanisms will be present, though the quality of these dispersal services is likely to differ from that of lemur frugivores.

Fig. 1
figure 1

Map of field sites within (red) and adjacent (white) to Ranomafana National Park in Madagascar where we studied the dispersal mechanisms of plants of the Canarium genus. These sites varied in the presence (circles) and absence (triangles) of the largest seed disperser in this ecosystem, Varecia variegata editorum due to habitat degradation

Full size image

Materials and methods

Study system

We carried out this research within and adjacent to Ranomafana National Park (RNP) in the southeastern part of Madagascar. Close to 41,601 hectares of cloud forests (with an additional mix-use 3 km buffer zone) were designated as the RNP protected area in 1991 with a particular focus on the protection of the newly discovered golden bamboo lemur (Hapalemur aureus)(Wright et al. 2012). This park, named a UNSECO World Heritage Site in 2007, hosts highly biodiverse communities of plants, lemurs, frogs, and chameleons (Rothman et al. 2022). The park’s elevation extends over a low to medium elevation zone ranging from 600 to 1,500 m (Wright and Andriamihaja 2002) and receives an average annual precipitation of 3,635 mm (Rothman et al. 2022). The forest fragments around the park are vulnerable areas in terms of biodiversity due to major pressures, such as land clearing and bushfire practices (Wright et al. 2012).

We selected 11 sites of varying habitat quality, which included two locations within the protected RNP (Vatoharanana and Valohoaka) and nine locations in forest fragments outside of the park limits. These fragment locations are separated from the protected forest by low-intensity agriculture, such as banana plantations, that maintain some canopy connectivity (Fig. 1). These sites also varied in the presence of V. v. editorum populations (Fig. 1). It is worth noting that although five species of Canarium spp. were identified by local names within these sites (Ramy, Ramy lavaravina, Ramy ravimboanjo, Ramy boriravina, Ramy vaventy ravina), samples sent for identification by the Parc Botanique et Zoologique de Tsimbazaza and the Missouri Botanical Graden office in Antananarivo could only be identified to the genus level. For our research, we focused on individuals of the Canarium genus, which were fruiting during our study period (July-September of 2021). These had fruits with an average length of 4.8 cm (2.9 –6.7 cm) and width of 2.5 cm (2.0 –2.7 cm) with thin pulp surrounding a single seed of 4.0 cm x 1.6 cm on average (length range: 2.4–5.4 cm; length width: 1.2–2.7 cm).

Abiotic dispersal

To test whether the large-sized seeds of Canarium could be dispersed by wind, we created a simulation model to examine the potential of wind to move seeds away from the parent plants. To inform the model, we first measured the height of fruits in the canopy (H) of 27 Canarium sp. trees in the field across nine sites (1 to 5 trees per site). For each tree, we then performed fruit fall experiments at the corresponding fruit heights to record fall speed (fruit terminal velocity, F). To do so, we climbed the trees up to the branches where fruits were located (at about 10–25 m high). From there, we dropped fruits one by one (for a total of 10–65 fruits per site) and recorded the time it took to drop to the ground using a timer. The model also utilized weather data from Meteoblue (https://www.meteoblue.com/historyplus accessed on Sept 15, 2021), which included wind speed values in Ranomafana for a period spanning from 2015 to 2020 (U). The core of the model was the calculation of wind-mediated dispersal distances first formulated by (Schmidt 1918):

$$D=\frac{H \text{*} U}{F}$$

where D is horizontal dispersal distance, H is the height of seed release, U is horizontal wind speed and F is the fruits’ terminal velocity. We performed the above calculations of dispersal distance for 1000 fruits, sampling from recorded values of wind speed, terminal velocity, and fruit release height. All models were coded in R-4.2.3 software (R Core Team 2022).

To examine the likelihood of dispersal by flood or runoff, we conducted an experiment to assess the ability of Canarium fruits to be carried by water. Although flooding is uncommon in this high-elevation rugged terrain, extreme rainfall events can cause runoff with the potential to wash seeds down-slope where they may be carried along ephemeral streams to larger waterways. To examine this possibility, we assessed if Canarium fruits and seeds could float, and measured their movement speed along a relatively straight segment of two streams, one in the forest interior and one within a forest fragment. The streams measured in the fragment and forest interior sites were both between 14 and 25 cm in depth and 2–5 m in width with a speed of 0.26 and 0.37 m/second (measured by the movement of a floating bamboo leaf) along a 22 and 23 m section respectively. We placed 30 fruits and seeds (one at a time) in each stream at the beginning of the 22 and 23 m segments to record if they floated, and recorded the time it took to travel the chosen segment. As measurements were taken during a dry period (August-September 2022), streams may be expected to flood more at other parts of the year providing a higher capacity to displace seeds.

Primary dispersers

To complete our knowledge of which primary dispersers may contribute to the movement of Canarium seeds, we conducted animal surveys in 9 of the 11 sites. We excluded the sites within the national park, Vatoharanana, and Valohoaka, as data were available from similar surveys conducted during previous studies in these sites (DeSisto et al. 2020). To assess the presence of potential primary dispersers, we set 3 to 4 linear transects of 500 m per site to survey for frugivorous animals. We walked each transect four times a day (6:00 am, 10:00 am, 3:00 pm, and 6:00 pm) for 10 days, during which we recorded the species and activity of any fruit-eating animals seen. We specifically looked for lemurs, birds, bats, and rodents during each visit. To identify Canarium dispersers, we also conducted direct observations of individual Canarium trees within each site to assess the proportion of visits and number of seeds swallowed by different species. The observations were conducted at consistent time intervals over two consecutive days from 6:00 am to 1:00 pm and from 1:00 pm to 8:00 pm the next day by two observers positioned at about 10 m away from the tree. Surveys of disperser presence included a total of 28 trees observed for 2234 h across all sites. During observations, we recorded the identity of any animal visitors, the time spent consuming fruits, and the number of fruits consumed.

As no other frugivory visitors were recorded during direct observations, we focus on dispersal by lemurs V.v.editorum, E. rubriventer and E. rufifrons. We combined the observation data with estimates of seed dispersal distance available in Razafindratsima et al. (2014) to parameterize a model of seed movement under different scenarios. Our models ran for 1000 observation minutes, the proportion of this time for which each lemur visited was based on the observed proportion of each species visit durations per total observation time across our study sites. In the first model step, we determine total visitation time per lemur species (out of 1000 min) using the proportion of visitation by each lemur species as the probability in a binomial distribution. We then sampled observed feeding rate values (fruits/minute) using a normal distribution generated with the observed feeding rate mean and standard deviation for each lemur species. The number of fruits consumed was calculated by multiplying the sampled feeding rate and visit duration. For each fruit consumed, we then determined the dispersal distance for each seed by randomly sampling from an empirically informed sample of dispersal distances. Using data on seed dispersal distances from Razafindratsima et al. (2014) we first divided dispersal distances into 25 m bins for each lemur and estimated the proportion of dispersal to each bin. We then generated a dataset of 1000 dispersal events by sampling from distance bins using these proportions as probabilities and then determined the final dispersal distance through a uniform random sample of distances within the selected bin.

Secondary dispersal

To determine which ground-dwelling small mammals can act as secondary dispersers of Canarium seeds that have dropped from the trees, we deployed 10 camera traps during the fruiting season to observe the visitors under the canopy of 10 trees (> 100 m apart) with abundant fruits on the ground at six of our field sites. Cameras were deployed at four of the forest fragments (Piepieky, Bevoahazo, Vohiparara, and Andafirano) and two forest interior sites (Vatoharanana and Valohoaka). These motion-activated cameras took pictures of animals visiting the area immediately below the crown of Canarium individuals. Each camera was deployed for 15 days with data downloads and camera checks occurring once per week. These pictures were used to identify visiting animal species and to define their seed-handling behavior as seed dispersers, fruit-pulp consumers, or seed predators (Donatti et al. 2007). We used an existing field guide of Malagasy mammals (Soarimalala and Goodman 2011) to aid in species identification.

To test the impacts of secondary dispersers on seed fate, we conducted experiments with marked seeds using a traditional approach for tracking secondary dispersal. This approach consists of marking each seed by tying and gluing ~ 75 cm of non-conspicuous string on its coat; a flagging tape of 15 cm with a unique identification number was added at the distal end of each string (as in Razafindratsima 2017; Forget 1990 and Wenny 1999). For this experiment, we set up ten stations at each of the above-mentioned six sites. Each station consisted of an area in a 2-m radius around the base of a fruiting Canarium tree. We cleared the 2-m radius station from any fruits or seeds before the experiment so that animal visitors would focus on our marked fruits/seeds. We placed 5 marked seeds on the ground at each station and monitored them for 15 days, replacing any seeds that were predated or moved. A total of 157 seeds were used and their fates were classified as (following (Chang and Zhang 2014): predated (41 seeds) or intact (71 seeds) and remained where found (45 seeds). The distance of each seed from the station was also measured with a measuring tape. With these data we then examined how dispersal away from the parent tree, and dispersal distance, impacted the survival of dispersed seeds. We assessed the proportion of seeds removed and remaining that were predated, and then used a generalized linear model with a binomial family to test how dispersal distance as a fixed variable (including no dispersal as 0) impacted predation as a binary response variable (predated or intact).

Results

Abiotic dispersal

We found that both wind dispersal and run-off dispersal, specifically associated with cyclones, can successfully move seeds away from the parent plants. For wind dispersal, we found a mean movement distance of 30.5 m, with a maximum distance of 147.7 m in conditions of 14 km/h (Fig. 2). Most seeds were moved less than 50 m. Runoff or flood dispersal showed potential for dispersal as 74% of seeds and 45% of fruits floated and moved along a stream at an average speed of 0.30 m/s and 0.29 m/s, respectively.

Fig. 2
figure 2

Seed dispersal distances for Canarium fruits based on model estimates of seed horizontal movement by wind. Estimates were based on fruit terminal velocity, average windspeeds at Ranomafana National Park, and fruit release height

Full size image

Primary dispersal

During our direct observations, we recorded only the lemur species Varecia variegata editorum, Eulemur rufifrons, and Eulemur rubriventer consuming Canarium fruits. V. variegata editorum was only present, and active as a Canarium frugivore, in four of the six sites (Sahavoaemba, Mangevo, Vatoharanana, and Valohoaka). The two Eulemur species were present at all sites but were not observed at Canarium trees in Mangevo or Bevohazo. We did not observe any other frugivores consuming Canarium fruits, either in the canopy or the ground, despite the presence of multiple frugivore species at all sites (Supplementary material Table S1). The three lemur species varied in their fruit consumption, with fruit consumption rate (number of fruits swallowed per minute) for Eulemur rubriventer 1 (SD = 1.3) fruits per minute, 0.43 (SD = 0.37) fruits per minute for Eulemur rufifrons and 0.49 (SD = 0.45) fruits minute for V. variegata editorum. The proportion of observation times across all sites during which each lemur species was observed was 0.035 for V. v. editorum, 0.012 for E. rubriventer, and 0.023 for E. rufifrons foraging. Our models show dispersal distances of up to 600 m, averaging 121 m (SD = 101 m) for E. rubriventer, 97 m (SD = 72 m) for E. rufifrons, and 118 (SD = 93) for V.v. variegata. Seed dispersal simulations show dispersal distances up to 624 m with average dispersal distances at 118 m (SD = 99 m, max = 624 m) for V. variegata editorum, 121 m (SD = 102 m, max = 375 m) for Eulemur rubriventer and 97 m (SD = 72 m, max = 418 m)(Fig. 3).

Fig. 3
figure 3

Seed dispersal distances for Canarium seeds modeled based on observed fruit consumption rates and seed dispersal distances by primary dispersing lemurs (Eulemur rubriventer, Eulemur rufifrons, and Varecia variegata editorum)

Full size image

Secondary dispersal

From our camera trap survey data, we found that rodents were primarily involved in visiting and consuming Canarium seeds, though we also found one bird visitor. Of 480 records of animal visitation, only 77 involved feeding on Canarium fruits, the rest involved birds, rodents, and carnivores traveling but not interacting with fruits. We captured images of rodents of the genus Eliurus (62 visits) and Nesomys (12 visits), both endemic to Madagascar, and the Madagascar magpie robin Copsychus albospecularis (1 visit) feeding on fallen Canarium fruits. The lack of invasive rodents in our camera trap data was surprising, as we detected them during our preliminary surveys (see Supplementary Material Table S1).

Secondary dispersal experiments resulted in a 47% removal rate of seeds, with a 0.19 rate of predation across removed seeds and 0.001 across non-dispersed seeds. Seeds were moved at an average distance of 22.58 m, with the longest dispersal distance at 575 m. Recovered seeds were found on the forest floor with no evidence of hoarding. When examining the effect of secondary dispersal, we found significant (estimate = 0.009, P < 0.001, z-value = 3.77, df = 392, Nagelkerke pseudo-R2 = 0.09) and positive effect of movement distance on predation probability. This means that secondarily dispersed seeds had a higher rate of predation than those remaining under the parent plants. This likely resulted from rodents carrying seeds to different locations for consumption.

Discussion

Our investigations of the dispersal of the large-seeded Canarium in the rainforest of Madagascar revealed the potential contribution of multiple mechanisms, namely winds, flooding, and animal dispersers through primary and secondary dispersal. In Madagascar, plants of the Canarium genus are thought to be at risk of being left without animal dispersers. This is due to the plant’s seed sizes, which seem too large for endozoochorous dispersal by all but now-extinct Malagasy megafauna and the largest extant lemur dispersers, many of which are now extirpated from their once-shared ranges (Federman et al. 2016; Albert-Daviaud et al. 2020). However, this study demonstrates how a few extant frugivores, including lemurs, rodents, and birds, as well as abiotic dispersal by wind and water, may sustain Canarium seed movement despite the threatened loss of the largest lemur dispersers, such as those of the Varecia genus. Nevertheless, the quality of dispersal provided by these mechanisms differs from that provided by primary dispersers, resulting in a lower frequency of long dispersal distances or high predation rates away from the parent plants. Such disparities in the quality of the dispersal services could impact plant reproduction if one dispersal mechanism, and its distinct contribution to seed movement, is lost.

In our study sites, lemurs of the genus Varecia and Eulemur were the only recorded endozoochorous dispersers of Canarium plants, transporting seeds long distances (mean = 114 m) away from the parent plants. The loss of one or even all lemur dispersers, in part or the whole of their range, may not lead to Canarium extinction, just as the theorized loss of megafaunal dispersers did not lead to the extinction of even the largest-seeded Malagasy species within this genus. On the other hand, it is still possible that the role of extant lemurs is distinct from that of their other dispersal mechanisms, beyond seed dispersal distances (Morales et al. 2013). Body size and behavior differences could lead to distinct outcomes for the dispersed seeds, affecting the treatment of seeds in frugivore guts (or eliminating gut passage altogether), the distance and quality of the deposition sites they reach, and even their exposure to natural seed predators (Portela and Dirzo 2020; Carvalho et al. 2021). The mismatch in the geographic range of Canarium and extant frugivore dispersers and the continued decline of lemur populations require further study into the impacts of primary, secondary, and abiotic dispersal mechanisms on seed movement and regeneration.

Many plant species exhibit traits directly for dispersal by wind or water, but plants typically dispersed by animals may also be moved by strong winds and flood waters (Nathan 2006; Parolin et al. 2013). In our study, we found that these animal-dispersed fruits may be carried away by strong winds and can float along a stream. Modeled wind dispersal was able to carry seeds up to ~ 150 m, effectively moving them away from the parent crown, although most dispersal events occurred at less than 50 m. These dispersal distances may be sufficient to allow seeds to escape the density-dependent factors that reduce survival near the parent plant (Comita et al. 2014), though movement by wind was not able to move seeds as far as dispersal by primary or secondary dispersers. Thus, dispersal by animals might be the most effective at the rare long-distance dispersal events that could facilitate colonization of new sites or connectivity across fragments. It is also important to note that dispersal distance is only one aspect of dispersal quality, and it is most likely that wind dispersal could not replicate how frugivore foraging behavior determines seed deposition patterns (Clark et al. 2004; Cousens et al. 2010; Razafindratsima and Dunham 2016) The importance of deposition site quality may also play a role in the effectiveness of water seed dispersal for non-riparian plants. While seeds of Canarium were able to float, it is uncertain what the impact of prolonged water exposure would be on seed germination, or the suitability of the final arrival sites for seed germination and growth (Fadini and Castro 2013). Thus, further work will be needed to quantify the quality of dispersal services provided by water to plants that have animal-dispersal syndromes. Despite potential issues and shortfalls associated with abiotic dispersal of a zoochorous plant, any movement of seeds away from the parent plants can lead to increased recruitment and thus survival of a species (Comita et al. 2014), enabling persistence when biotic dispersal is infrequent or missing.

Biotic secondary dispersal by scatter-hoarding rodents proved to be another effective mechanism for Canarium seed movement. Although granivorous rodents can move seeds with the intent to consume them, many factors may lead a seed to be left unscathed and allowed to grow (Velho et al. 2009; Hirsch et al. 2012; Loayza et al. 2014). In our study, we found that secondary dispersers move about half of the seeds provided in our experiments away from the parent plants, transporting them to distances like those achieved by primary lemur dispersers. However, we also found that dispersal distance by secondary dispersal had a positive relationship with predation, reducing the effectiveness of long-distance dispersal events. Despite these risks, secondary dispersal by native rodents provides a viable mechanism for moving seeds considerable distances. Additionally, though we did not observe invasive rodents interacting with Canarium seeds, they may also play a role in the predation and removal of seeds in Madagascar (Goodman and Sterling 1996). Further work is needed to resolve how movement by secondary dispersal, abiotic factors, and primary dispersal compare in their impact (e.g. predation risk) on Canarium recruitment. It is also necessary to assess how these dispersal mechanisms may interact (e.g. dispersal by lemurs followed by secondary dispersal by rodents) to shape dispersal distances and deposition patterns (Landim et al. 2022). Additionally, as our study was limited, an in-depth study that takes into account seasonal variations and spans the entire fruiting period of Canarium could provide more insights regarding the existence and effectiveness of other dispersal mechanisms.

The survival of many plant species may depend on their ability to utilize the multiple dispersal mechanisms available to them and their tolerance to low recruitment (Godfrey et al. 2008; Dirzo et al. 2014). While the loss of a primary disperser might not doom a plant species to extinction, it may place them in a precarious position, where lower incidence and altered quality of dispersal lowers regeneration rates and can curtail range expansion (Galetti et al. 2006; Carvalho et al. 2016; Onstein et al. 2018). With ongoing environmental changes and disruption of ecological interactions it is essential that we identify at-risk mutualisms, and seek to understand the extent to which co-occurring or novel interactions provide redundant or complementary services (Harper et al. 2008; Barlow et al. 2016; Portela and Dirzo 2020; Carvalho et al. 2021).