Pollination by the locally endangered island flying fox (Pteropus hypomelanus) enhances fruit production of the economically important durian (Durio zibethinus)

Abstract Fruit bats provide valuable pollination services to humans through a unique coevolutionary relationship with chiropterophilous plants. However, chiropterophily in the Old World and the pollination roles of large bats, such as flying foxes (Pteropus spp., Acerodon spp., Desmalopex spp.), are still poorly understood and require further elucidation. Efforts to protect these bats have been hampered by a lack of basic quantitative information on their role as ecosystem service providers. Here, we investigate the role of the locally endangered island flying fox Pteropus hypomelanus in the pollination ecology of durian (Durio zibethinus), an economically important crop in Southeast Asia. On Tioman Island, Peninsular Malaysia, we deployed 19 stations of paired infrared camera and video traps across varying heights at four individual flowering trees in a durian orchard. We detected at least nine species of animal visitors, but only bats had mutualistic interactions with durian flowers. There was a clear vertical stratification in the feeding niches of flying foxes and nectar bats, with flying foxes feeding at greater heights in the trees. Flying foxes had a positive effect on mature fruit set and therefore serve as important pollinators for durian trees. As such, semi‐wild durian trees—particularly tall ones—may be dependent on flying foxes for enhancing reproductive success. Our study is the first to quantify the role of flying foxes in durian pollination, demonstrating that these giant fruit bats may have far more important ecological, evolutionary, and economic roles than previously thought. This has important implications and can aid efforts to promote flying fox conservation, especially in Southeast Asian countries.

In Malaysia, where the species is thought to be native (Morton, 1987;Subhadrabandhu & Ketsa, 2001), exports of durian fruits keep increasing annually and now contribute more than USD 17.9 million to the national economy (United Nations, 2016). This is probably just a tiny fraction of the economic value of the domestic trade within the country, where the popular fruit has high cultural importance and is considered the "king of fruits." Flying foxes and other pteropodids such as the cave nectar bat (Eonycteris spelaea) visit flowering durian orchards to feed (Aziz et al., 2017a;Bumrungsri et al., 2013;Gould, 1977Gould, , 1978Soepadmo & Eow, 1976), leading to a perception among farmers that bats cause damage and negatively affect fruit production (Aziz, Olival, Bumrungsri, Richards, & Racey, 2016;Bumrungsri et al., 2009). Recent pollination experiments have shown that rather than being destructive, the cave nectar bat is actually a major pollinator for semi-wild durian in southern Thailand (Acharya et al., 2015;Bumrungsri et al., 2009)-an example of chiropterophilous pollination syndrome playing an important role in both culture and economy. However, the role of larger fruit bats such as flying foxes in pollination ecology is still poorly understood and requires further elucidation. Early literature postulated that flying foxes likely have a negative impact in durian orchards, because they were believed to consume the whole flower or to destroy it through chewing (Lee, Norsham Suhaina, Boon, & Chua, 2002;Soepadmo & Eow, 1976;Start, 1974). However, these claims were not based on actual observations or empirical studies. In contrast, Gould (1977Gould ( , 1978 reported, based on direct observations in an orchard, that P. vampyrus feeding in flowering durian trees only licked nectar from flowers and did not destroy or damage the flowers. Until now, however, no attempt has been made to test the role of flying foxes in durian reproductive ecology. While exclusion experiments have been a successful approach for studying the role of smaller pteropodids in pollination (Bumrungsri et al., , 2009Srithongchuay et al., 2008), the comparatively greater sizes and different feeding behavior of flying foxes (Nathan, Karuppudurai, Raghuram, & Marimuthu, 2009;Nathan, Raghuram, Elangovan, Karuppudurai, & Marimuthu, 2005) are difficult to account for when using such a study design; it is challenging to design a specific treatment that can exclude flying foxes while still allowing access to smaller pteropodids. Consequently, exclusion experiments that included both flying foxes and smaller pteropodids have simply grouped them together as "bats" (e.g., Nathan et al., 2005Nathan et al., , 2009).
Here, we use camera traps as a novel approach to investigate the role of the island flying fox (Pteropus hypomelanus) in the pollination ecology of durian trees on Tioman Island, Peninsular Malaysia, and to ascertain whether flying foxes have an effect on durian reproduction. Specifically, we asked the following questions related to the nature of the relationship between flying foxes and durian: (1) What animals visit durian flowers? (2) How do these animals interact with durian flowers over time and space? (3) How do bat-flower interactions affect pollination and reproductive success?
As flying foxes are still frequently hunted, persecuted and even legally killed as pests (Epstein et al., 2009;Fujita, 1988;Mildenstein, Tanshi, & Racey, 2016), such crucial information can improve our understanding of how their declines may impact the survival of chiropterophilous plants in the Palaeotropics. This will also help us better understand the roles of flying foxes as ecosystem service providers, which can then be used to justify their protection and conservation (Scanlon et al., 2014;Vincenot, Florens, & Kingston, 2017).

| METHODS
This study conforms to the research ethics criteria stipulated by The University of Nottingham Malaysia Campus.
Much of the island inland is still covered by tropical rainforest, which has been designated as Pulau Tioman Wildlife Reserve (82.96 km 2 ). It has a hilly topography, with flat areas only along the coast (Abdul, 1999). The area designated as a wildlife reserve is composed of lowland mixed dipterocarp forest and hill dipterocarp forest. Most forested areas are still inaccessible because of the rugged topography, with many steep slopes and rocky outcrops (Latiff et al., 1999). The climate is tropical, uniformly warm, and humid throughout the year (Hasan Basyri et al., 2001), but the northeast monsoon takes place from November to March (Bullock & Medway, 1966).
There are currently seven villages (Air Batang, Genting, Juara, Mukut, Paya, Salang, and Tekek) on the island, situated along the coastline ( Figure 1). The majority of the local people are Muslim, and therefore, due to religious dietary restrictions do not hunt the bats for food or medicine. As the island's marine area is also a designated Marine Park and a popular tourist destination, many local people are involved in the tourism industry (Abdul, 1999). Our study was conducted during 21 April-29 July 2015 in one durian orchard in Juara, located next to the main road in the village. The durian trees ranged in height from 10 to 25 m; the exact ages of the trees were unknown, but the largest and tallest tree (D3) was said by the orchard owner to be around 90 years old. Only four durian trees in the orchard were flowering during the time of the study, while another four did not flower and could not be included in the study. The orchard generally had plenty of open space with no other cultivated fruit trees.

| Pteropus hypomelanus
The island flying fox (Figure 2a) is also known as the variable flying fox and, less commonly, as the small flying fox (Francis et al., 2008).
It has a wingspan of more than 1 m and exhibits sexual size dimorphism, with males weighing around 570 g, and females around 470 g (Ouillette, 2006). It roosts gregariously, forming colonies of up to 5,000 individuals. It is a widespread insular species and considered to be abundant throughout its range, which extends from the Maldives and Indian islands in the west to Melanesia in the east. Because of this, it is considered to be Least Concern on a global scale by the IUCN Red List; however, its population trend is noted to be decreasing (Francis et al., 2008;Olival, 2008 (Olival, 2008) has shown the east coast populations off Peninsular Malaysia to be a subspecies-P. hypomelanus lepidus-genetically distinct from the west coast populations of P. hypomelanus robinsoni. The species is listed as Endangered on the Malaysian Red List (DWNP, 2010), reflecting its precarious situation in the country.
On Tioman, P. hypomelanus can be found roosting permanently in two villages ( Figure 1): Tekek, the main and biggest village (~1,260 people), located on the west coast, and Juara, the second largest village (~350 people) and the only one located on the east coast.
Monthly roost counts conducted during March-October 2015 yielded estimated ranges of 675-1,033 individuals in Juara, and 1,503-4,352 individuals for Tekek.
Following Kingston (2010) and the Southeast Asian Bat Conservation Research Unit (SEABCRU; http://www.seabcru. org/?portfolio=flying-foxes), the common term "flying fox" is used here to refer only to the genera Pteropus and Acerodon; recent taxonomic revisions, however, have revalidated the genus Desmalopex (Almeida, Giannini, Simmons, & Helgen, 2014), which can now also be considered under this term. Thus, within the context and description of this specific study, "flying fox" is used to refer to P. hypomelanus, whereas the nectarivorous E. spelaea is referred to as "nectar bat."

| Durio zibethinus
Durian (Family Malvaceae, previously Bombacaceae) is a tree likely native to Borneo, Sumatra and Peninsular Malaysia (Morton, 1987;Subhadrabandhu & Ketsa, 2001). Semi-wild durian grown from seed and exposed to little or no artificial management is commonly planted in rural areas of southern Thailand, Malaysia, and Indonesia. Its seed results from open pollination, and its genetic diversity is reflected in a variety of taste and aril characters (Bumrungsri et al., 2009). Its flowers ( Figure 2b) display typical chiropterophilous traits: large, strong, and wide-mouthed, whitish or creamy in color, cauliflorous brush inflorescence, nocturnal anthesis lasting only for one night, and emitting a strong and distinctive odor (Marshall, 1983;Yumoto, 2000). A few durian cultivars are planted on a commercial scale. Hand-crossed pollination is sometimes carried out in such commercial plantations.
Durian in Juara and Tekek has three distinct flowering seasons: April-May, July-August, and October-November. However, each individual tree only flowers once a year, and the trees in this study all flowered from late April to early May. Due to constraints on accessibility (e.g., height of inflorescences, the presence of aggressive honeybees), we were unable to ascertain the full floral biology of the study trees, although flowers were observed to open and secrete nectar around 16:15 hr, and flower corollas dropped naturally from the trees between 01:00 hr and 02:00 hr. It is reasonable to assume that they possess floral biology characteristics similar to those reported by Soepadmo and Eow (1976)   at durian inflorescences at a distance of within 2 m to document the animal visitors and their feeding behavior. All cameras were removed on 29 July 2015.

| Camera-trapping
An interaction was defined as physical contact between an animal and an inflorescence, regardless of whether this involved feeding or not. Both camera and video traps were set to allow a one-second interval in between captures. The duration of each video capture was set for 10 s. Consecutive captures that depicted the same continuous physical contact were pooled together as one interaction.
Spatiotemporal patterns were visualized in density plots using packages overlap, reshape2, and ggplot2 in R statistical environment 3.2.2 (R Development Core Team, 2015). All mammal visitors were identified to species. Wherever possible, insects were identified to groups such as moths, stingless bees, and to species in the case of the Asian giant honeybee (Apis dorsata; Figure 2b). The frequency and duration of interactions between visitors and flowers were quantified, and the feeding behavior of visitors at inflorescences was also noted.

| Bat sampling at flowering trees
In order to confirm the identity of the smaller pteropodids, mistnetting was conducted in the durian orchard on 6 May 2015. To maximize the likelihood of capturing bats that had already fed in the durian trees, we avoided mist-netting earlier in the evening. One mist net (2.6 × 12 m) was set up across a flyway between two durian trees, at a height of ~6 m. The net was manned and monitored directly from 20:30 hr, when it was first put up, until midnight when it was taken down and was never left unattended. It was checked every 15 min.
Captured bats were identified to species following Kingston, Lim, and Zubaid (2006).

| Effects of bat-flower interactions on pollination and reproductive success
We investigated the effects of flying fox [FF] and nectar bat [NB] interactions with flowers on fruit set by constructing generalized linear mixed-effect models (GLMMs) that included all possible subsets using a multimodel inference framework (Burnham & Anderson, 2003). We as initial fruit set after 20 days, and reproductive success, measured as mature fruit set after 60 days (Figure 2d). Initial fruit set was estimated from camera-and video-trap footage 20 days after the first bat-flower interaction, as the late-acting self-incompatibility of the D. zibethinus breeding system causes unfertilized fruit to be aborted within this period (Bumrungsri et al., 2009;Honsho, Yonemori, Somsri, Subhadrabandhu, & Sugiura, 2004). As Bos et al. (2007) recommended that mature fruit set should be used as the metric for the economic role of pollinators, fruit set was quantified 60 days after the first batflower interaction, following Bumrungsri et al. (2009).
We included durian tree characteristic [DUR] as an additional covariate to distinguish one taller (~25 m) and isolated durian tree (D3) from the other three shorter (~10 m) and spatially clumped durian trees (D1, D2, and D4). To account for possible nonindependence due to other individual tree characteristics that we cannot account for (e.g., genetic variation), we also allowed model intercepts to vary across a random effect (TRE). We used Poisson's (log-link) GLMMs to model the continuous response variables (i.e., fruit set). Before running the GLMMs, we first assessed whether the covariates were correlated (coefficient values >|0.5|) in order to obtain more stable and interpretable parameter estimates. We used sample size corrected Akaike Information Criterion (AICc) to determine the best candidate model, Akaike weights (wAICc) to quantify the probability by which a given model is the best within the candidate models set, and the sum of Akaike weights (SW) to estimate relative variable importance (Burnham & Anderson, 2003;Giam & Olden, 2016

| Animal visitors to durian flowers
From a survey spanning 54 days, we obtained 2,733 10-s video clips and 3,367 still photographs of animal visitors from 13 camera stations (data from six camera stations could not be used due to malfunctions or inappropriate positioning). The number of inflorescences included within a camera's range varied from two to 19 (mean = 8; mode = 5).
Our camera traps revealed six vertebrate taxa visiting durian flowers (Table 1). Two pteropodid bat species were photocaptured: The larger pteropodid was identified to be P. hypomelanus, while the smaller was identified as E. spelaea, based on Lim et al. (1999) and Yong et al. (2012). Invertebrate taxa, mainly A. dorsata and moths (Lepidoptera), were also photocaptured along with vertebrates in video clips, but only photocaptured independently in still photographs. Insect visitors could thus be quantified only from still photos, which were also used for quantifying flower abundance and fruit set; videos were more effective overall at identifying animal visitors, quantifying vertebrateflower interactions, and observing feeding behavior.
Bats were the most abundant overall of vertebrate visitors in the video-trap footage, with E. spelaea being the most abundant species (Table 1). Based on the video footage, E. spelaea landed on inflorescences directly, head up and occasionally horizontally, thumb claws holding opened flowers, and inserting their muzzle into the corolla tubes of flowers. These visits lasted anywhere from less than a second to over one minute before the nectar bat would fly away again; the mean duration of visits was 2 s.
In contrast, P. hypomelanus hung from the branch next to an inflorescence using their hind claws, occasionally using thumb claws to hold and move opened flowers closer toward them, and inserting their muzzle into the corolla tubes of flowers ( Figure 3). Some P. hypomelanus occasionally hung from the inflorescence stalks using their hind claws while feeding. These feeding bouts could last for more than 1.5 min (mean duration 21 s). P. hypomelanus would either fly off after feeding or continue to hang from branches or inflorescence stalks T A B L E 1 Durian tree characteristics and six vertebrate-flower interactions from camera traps at four durian trees in Juara, Tioman, 6 May -29 July 2015  even when not feeding, but would also move by crawling quadrupedally along a branch. Thus, P. hypomelanus was observed crawling from one branch to another, and also crawling from one inflorescence to another along the same branch to continue feeding.  (Table 1) showed this squirrel species biting into the flowers directly, often at the base, tearing off pieces, and even whole flowers. We found that more than a quarter (27%, n = 669) of the dropped corollas found under two durian trees (D1 and D2; D3 was not in flower yet) were damaged, with holes in their bases ( Table 2). The only tree without these damaged corollas on the ground was the tallest and oldest tree (D3), which was surrounded by open space. This tree was also the only one without any squirrel detections, probably due to its isolation from the rest of the trees. Squirrels were observed during the day in the other three trees, occasionally appearing to feed on durian flowers by nibbling at the bases.
Videos also showed that long-tailed macaques (Macaca fascicularis) fed destructively, plucking flowers off with their hands and consuming these whole. However, M. fascicularis were only recorded feeding on two occasions. Sunbirds/spiderhunters (Family: Nectariniidae) were photocaptured feeding nondestructively during the day on four occasions, but before presumed full anthesis had occurred. One Sunda colugo (Galeopterus variegatus) was photocaptured during the study.
Although it was briefly recorded brushing its face against flowers, the exact nature of the interaction could not be determined.

| Bat sampling at flowering trees
Four bats were captured during 3.5 net hours. Three were identified as E. spelaea, and one as Cynopterus sp. Pollen was found on the body of all E. spelaea, but none was found on the Cynopterus individual.

| Animal-flower interactions and durian fruit set
Initial fruit set (pollination success) could be recorded from camera stations at three durian trees, but mature fruit set (reproductive success) could only be recorded at two durian trees, D2 and D3 ( although the two trees without fruit had the highest number of animal-flower interactions overall, including P. hypomelanus interactions, this also included the highest number of antagonistic interactions (e.g., removing or damaging flowers rather than pollinating them). In contrast, although the two trees with fruit had comparatively fewer P. hypomelanus interactions, they also had comparatively fewer antagonistic interactions overall.

| Effects of bat-flower interactions on pollination and reproductive success
We

| DISCUSSION
This study has yielded important preliminary insights into the role of flying foxes in durian pollination. We obtained photographic and video evidence that P. hypomelanus feeds on the nectar produced by durian flowers. Camera-trapping indicated vertical stratification in feeding niches among different bat species. Most importantly, we show that P. hypomelanus has a positive effect on durian reproductive success, suggesting a mutualistic relationship that developed through coevolution. However, our understanding of the strength of this relationship was compromised by the small sample size of trees and needs to be tested further by expanding the study into a more in-depth and extensive investigation involving both isolated and nonisolated durian trees of varying heights from other orchard sites, and preferably over several durian seasons. The entire durian pollination network should ideally be studied at the community level (Memmott, 1999), taking into account mutualisms, antagonisms, and the dynamics of various inter-species interactions within the pollination complex.
The evidence we have obtained disproves earlier assertions that flying foxes feed destructively on durian flowers (Soepadmo & Eow, 1976;Lee et al., 2002;Start, 1974), and instead supports Gould's (1977Gould's ( , 1978 observations of nondestructive feeding. Such nondestructive feeding behavior has also been reported for flying foxes in kapok trees (Ceiba pentandra) in southern India (Nathan et al., 2005;Singaravelan & Marimuthu, 2004) and Madagascar (Andriafidison et al., 2006). Studies elsewhere corroborate our findings that demonstrate the potential of flying foxes to increase pollination success; for example, Elmqvist et al. (1992) showed through exclusion experiments that the kapok tree in Western Samoa depends entirely on flying foxes as pollinators. In subtropical Japan, flying foxes are the primary pollinators of the native plant Mucuna macrocarpa (Nakamoto et al., 2009;Toyama, Kobayashi, Denda, Nakamoto, & Izawa, 2012). Birt, Hall, and Smith (1997) examined pteropodid tongue ecomorphology and found that the structure of tongues and papillae of flying foxes support a role as pollinators. All these provide strong evidence that flying foxes are also important agents in chiropterophily.

| Mutualistic and antagonistic network interactions in durian pollination ecology
Our study found the same type of animal groups as those reported visiting durian in southern Thailand (Bumrungsri et al., 2009) and kapok in southern India (Nathan et al., 2005). In addition, we discovered three further species of animal that also feed on durian flowers: squirrels (C. notatus), macaques (M. fascicularis), and a colugo (G. variegatus). The first two species fed destructively. The colugo has been reported before from durian orchards (Ketol, Abdullah, & Tedong, 2006), and although we could not ascertain its actual feeding behavior in this study, we may yet have obtained the first visual evidence showing that this animal uses durian trees as a food resource.
Like Bumrungsri et al. (2009) andStart (1974), we also found that a few sunbirds/spiderhunters (Nectariniidae) occasionally fed on durian flowers that had opened in the afternoon. However, as these interactions happened between 16:00 hr and 19:00 hr before full anthesis occurred, it was unlikely to result in pollination success (Soepadmo & Eow, 1976;Start & Marshall, 1976)-although this still needs to be verified. Stingless bees (Apidae) and A. dorsata also fed in the afternoon, with A. dorsata also feeding at night. Wayo and Bumrungsri (2017) have shown that bees can contribute slightly to pollination at least for the "Mon Thong" commercial cultivars in southern Thailand. However, our observations of bee feeding behavior at the semi-wild durian trees in our study suggest that in this particular instance this animal group acted largely as pollen robbers, not pollinators. Moths, which feed only at night, could also play a role as pollinators, but in our study their impact was likely to be low as they were not camera-trapped frequently.
Also, Start (1974) observed moths feeding on durian nectar without actually coming into contact with either the anthers or stigma. Thus, in our study only the pteropodid bats seem to have a truly mutualist relationship with the durian tree.
The results of the GLMM analyses must be treated with caution.
The inconclusive results for preliminary fruit set suggest an insufficient sample size and a need to replicate this study using more trees, more orchards, and more sites. However, the analysis for mature fruit set provides some useful clues for further investigation. For example, the taller and isolated characteristic of D3 tree appeared to have a positive effect on mature fruit set (Table 3). This positive effect could be due to fewer antagonistic interactions occurring at greater heights. In our study, fewer A. dorsata and C. notatus interactions occurred at the higher levels, where P. hypomelanus interactions were more numerous.
Additionally, C. notatus was not detected in D3, possibly because it was surrounded by open space and the orchard owner had wrapped linoleum around the trunk to prevent access from the ground. The other trees had also received the same protective treatment; however, they were in close proximity to each other as well as other trees in the vicinity, essentially connected through a network of branches. C. notatus could thus easily cross over from one to tree to another using closely positioned branches as bridges. Damaged flower corollas, which were likely caused by C. notatus nibbling holes into the flower bases to access nectar, were not found under D3. Therefore, it is likely that this tree did not suffer any flower damage from C. notatus, which may be one reason why it produced the most fruits. In any case, we could not use GLMMs to determine whether C. notatus had any effect on fruit set because of zero-inflated data (camera traps did not detect any C. notatus on D3). Durian produces copious numbers of flowers, and low levels of flower damage/loss may be negligible or even beneficial-too many flowers can lead to resource limitation, resulting in decreased fruit abundance and/or quality (Yumoto, 2000;S. Bumrungsri, unpublished). Thus, removal or nonpollination of excess flowers can actually improve fruit production. However, observations of C. notatus feeding behavior showed that their interactions with durian flowers were extremely destructive, and therefore, it is possible that a certain threshold number of individuals/visits could begin to have a detrimental effect on the tree's reproductive success. Trees in our study that did not produce fruits had more such antagonistic interactions than the trees that did produce fruit, and therefore, it is possible that taller and isolated trees may enjoy greater reproductive success; indeed, the potential importance of flying foxes for facilitating long-distance pollen transfer among tall trees has already been observed in Australia (Bacles et al., 2009).

| Implications of niche partitioning and feeding behavior on pollinator effectiveness
Clear temporal differentiation in visits has been observed elsewhere between flying foxes and other pteropodids feeding on flowering kapok in southern India (Nathan et al., 2005;Singaravelan & Marimuthu, 2004) and Madagascar (Andriafidison et al., 2006), involving differences of several hours. This temporal differentiation may be due to roost locations influencing commuting times and energetic requirements, but is likely also due to resource partitioning (Andriafidison et al., 2006;Nathan et al., 2009). In contrast, our study found only very slight temporal partitioning between flying foxes and nectar bats. Although P. hypomelanus arrived and left slightly earlier, the difference in timing was only 30-40 min, and peak activity for both bat species occurred at almost the same time. Nathan et al. (2009) found an even smaller time difference, of only ~15 min, between P. giganteus and Cynopterus sphinx arriving to feed on madhuca (Madhuca latifolia) flowers in southern India, but peak activity timings were 1 hr apart. In that study, however, flying foxes arrived later and left earlier. It is notable that the durian orchard in our study is located only ~300 m from the nearest P. hypomelanus roosts, which are situated on the beach in Juara, presumably allowing quick and early access by the larger pteropodid. Roost locations for E. spelaea were unknown, but the large cave roosts required by this nectarivorous species (Start & Marshall, 1976) were not observed anywhere in close proximity to the orchard. Nathan et al. (2009) have suggested that smaller bats may gain sufficient energy from a single flower, enabling longer commuting flights from more distant roosts (Start & Marshall, 1976). These factors may account for there being only a slight temporal difference between the two bat species.
Our study revealed a definite vertical stratification in the foraging heights of the two pteropodid species. Similar to findings from the above-mentioned studies on kapok (Andriafidison et al., 2006;Nathan et al., 2005;Singaravelan & Marimuthu, 2004) and madhuca flowers (Nathan et al., 2009), flying foxes preferred the upper levels of a tree, whereas the smaller bats were more likely to feed in the lower levels.
This vertical stratification was also observed for different pteropodid species feeding in fruit trees in southern India (Sudhakaran & Doss, 2012), and pteropodids caught in Fijian rainforest (Scanlon & Petit, 2016), and likely helps to avoid inter-specific competition (Fischer, 1992;Fleming, 1979;Thomas & Fenton, 1978). Again, this spatial partitioning corroborates findings by Gould (1977Gould ( , 1978, who reported that flying foxes visited most flowers in the upper canopy of durian trees, and that only small bats fed in the lower canopy. In our study, this spatial partitioning also appeared to influence the species of bat found feeding in durian trees of differing heights; comparatively more P. hypomelanus were found in D3, the tallest tree (~25 m), whereas comparatively more E. spelaea visited the shorter D1 (~10 m). This height differentiation suggests that semi-wild durian, which tends to be taller than commercial cultivars, could be particularly dependent on flying foxes for reproductive success. As it is used as grafting and breeding stock for commercial cultivars, its continued survival is important for the commercial durian plantation industry.
Height differentiation may also possibly explain the surprising slight negative effect of E. spelaea interactions with durian flowers (Table 3). This negative effect appears to contradict all previous studies showing that this bat species is an effective and even principal pollinator promoting cross-pollination of durian (Acharya et al., 2015;Bumrungsri et al., 2009). It is unclear why E. spelaea may have had a negative effect on reproductive success in our study; however, comparatively more mature fruit set was observed in the higher levels of trees, which coincides with higher P. hypomelanus interactions. In our study, fruit set was not correlated with flower abundance. Tree D1 had the highest number of flowers and highest overall number of E. spelaea interactions, yet no fruit set. This inverse relationship could be related to tree height and number of antagonistic interactions as mentioned above, but could also result from resource limitation (Yumoto, 2000), or the health and/or age of the tree. Pollination experiments have found that even hand-crossed pollination conducted on 10-year-old durian around 10 m in height produced very few fruits; in addition, even older trees only set fruit in the higher branches (S. Bumrungsri, unpublished). If durian trees characteristically produce more fruits at greater heights, then P. hypomelanus may have served as a more important pollinator than E. spelaea in this particular study, because the former feeds in the higher levels of the trees.
It is also possible that perhaps E. spelaea feeding behavior simply does not transfer pollen as effectively as P. hypomelanus. Interestingly, Tschapka (2003) found that the perching behavior of frugivorous bats in Costa Rica feeding on flowers of the Neotropical palm Calyptrogyne ghiesbreghtiana facilitated better pollen transfer than the hovering behavior of nectarivorous bats. Although Sritongchuay and Bumrungsri (2016) have shown that nectarivorous bats have higher network strength than small frugi-nectarivorous bats in mixed-fruit orchards of southern Thailand, such differences in feeding behavior may also be a factor and should also be taken into account when assessing pollination effectiveness.

| Implications of feeding behavior and pollinator dynamics on durian reproductive success
An alternative explanation for the apparent negative impact of E. spelaea on mature durian fruit set could be that excessive visits by pollinators might have a negative effect resulting in low fruit production. Such a scenario was observed by Wilmott and Búrquez (1996) for the self-incompatible desert climbing vine Merremia palmeri of Mexico, where more than five visits by its primary pollinator actually resulted in lower fruit set. A similar effect has been postulated by Avila, Pinheiro, and Sazima (2015) for the generalist forest tree Inga subnuda luschnathiana in Brazil, where visitations by animals to flowers resulted in decreased fixed polyads in stigmas. If a similar scenario occurs for durian, P. hypomelanus may not have had this effect in our study since it fed solitarily, occupied the same branch for extended lengths of time even when not feeding, and due to its territorial feeding behavior would have defended its floral resources against other visitors. On the other hand, smaller bats, which do not defend feeding territories, may be more likely to congregate on flowers in larger numbers, resulting in more overall visits to individual flowers. It is thus possible that once the number of visits exceeds a certain threshold, pollination success may become less likely.
Differences in feeding behavior between pteropodid species may also be a factor influencing to what extent cross-pollination occurs. Acharya et al. (2015) have shown how visits by nectar bats help to promote cross-pollination by depositing conspecific pollen on stigmas.
It would be interesting to investigate whether both inter-specific and intra-specific bat feeding interactions may exert a pressure that promotes cross-pollination. McConkey and Drake (2006) have shown that high densities of flying foxes lead to more aggressive feeding interactions, which then facilitates more effective seed dispersal; such territorial feeding behavior by flying foxes has also been reported by Brooke (2001), Gould (1977Gould ( , 1978, and Wiles, Engbring, and Falanruw (1991). Elmqvist et al. (1992) suggested that flying fox feeding behavior may also lead to density-dependent pollen dispersal, transferring conspecific pollen both within and among trees, a theory also postulated by Gould (1977). In our study, only one aggressive feeding interaction was recorded between flying foxes. There were more incidents of flying foxes physically defending inflorescences against nectar bats than against other flying foxes; comparatively less aggressive feeding interactions took place between flying foxes, as almost all flying foxes fed solitarily without physical intra-specific interference. This lack of interaction contrasts with Gould's (1977Gould's ( , 1978 observations of P. vampyrus feeding in durian trees, which involved many close agonistic encounters. However, given that our study recorded fewer flying foxes than expected given our population counts for Juara and the limited number of flowering durian trees during this period, it could be that alternative food sources were also available at the same time, thereby reducing the need for flying foxes to compete for durian flowers. Also, Gould (1977Gould ( , 1978 observed flying foxes defending durian flowers from conspecifics using spread-wing displays; this caused flying bats to veer away at a distance of 30 m without physical interactions taking place. Additionally, Brooke (2001), Trewhella, Rodriguez-Clark, Davies, Reason, and Wray (2001), and Wiles and Conry (1990) all reported wing-clapping similar to that which we observed, where wings were quickly brought together so that the forearms hit and created loud claps in order to threaten conspecifics during feeding. In our study, these spread-wing and wing-clapping threat displays, along with frequent vocalizations, may have been sufficient to deter conspecific competitors; Trewhella et al. (2001) note that only when such behavior was insufficient did more physical interactions take place.
Although we observed some loss of floral components in our study, the amount was small relative to the number of flowers in an inflorescence. This loss seemed far less substantial than the significant destruction caused by flying foxes to flowers and immature fruits of kapok in Samoa, where the tree appears to be completely dependent on a dystrophic pollination system in which its sole pollinator also destroys up to 50% of its flowers and fruit (Elmqvist et al., 1992). However, flying foxes in Madagascar were observed to feed on kapok nectar without causing significant damage (Andriafidison et al., 2006). In our study, bats never ate the durian flowers or fruits. The large numbers of flowers produced by durian probably means that the overall pollination benefits from flying fox visits outweigh the low amounts of occasional flower loss-an observation also made by Gould (1977). Ultimately, this study provides valuable clues regarding the implications of flying fox declines throughout their range. It is possible that coevolution has produced interrelated population dynamics between "apex pollinators" (e.g., flying foxes) and "mesopollinators" (e.g., nectar bats), resulting in an effect similar to that of mesopredator release (Crooks & Soulé, 1999). In this case, "mesopollinator" visits, when no longer suppressed by the presence of a larger pollinator, may increase, leading to reduced pollination success, and negatively affecting a plant's reproductive ability. Cox and Elmqvist (2000) also proposed a similar "keystone pollinator" role for Samoan flying foxes that potentially shape ecosystem structures in a manner analogous to that of predators. Boulter, Kitching, Howlett, and Goodall (2005) suggested that plants with a generalist pollination system such as Syzygium sayeri can compensate for the absence of certain pollinators with the presence of other pollinators, in which case, the loss of a particular pollinating animal may not represent a significant difference. Plant species such as durian, however, clearly rely on a highly specialized pollination system that depends entirely on the nocturnal pteropodids that the plant has coevolved with over millions of years (Marshall, 1983). In this kind of system, the loss of an "apex pollinator" could well have reproductive consequences, even when another pollinator is still available to do the job.

| Conservation implications
Given the potential importance of flying foxes in ensuring the continued reproductive success of durian trees, the economic implications for the durian fruit industry should not be ignored. The conservation value of such an economic role is obvious. It is particularly significant given that some commercial durian farmers, such as in southern Thailand, have resorted to artificial cross-pollination by hand in the absence of natural pollinators-a laborious, time-consuming, costly, and dangerous method (Wayo & Bumrungsri, 2017). On Tioman, local people hold negative perceptions and misconceptions of P. hypomelanus in addition to low awareness of bat ecosystem services, which can be an obstacle to conservation (Aziz, Clements, Giam, Forget, & Campos-Arceiz, 2017b). Showing how P. hypomelanus is in fact an important durian pollinator provides yet another example of a batserviced plant that has high value to humans-such case studies of ecosystem services can be used to develop a tailored approach to promote bat conservation among local communities, and overcome negative attitudes toward flying foxes (Scanlon et al., 2014).

| CONCLUSION
This study is the first attempt to assess the specific role of flying foxes in durian pollination ecology. We have demonstrated that this can be successfully conducted using camera traps that provide both photographs and video footage-a novel approach for studying the ecological function of flying foxes, and which should be combined with exclusion experiments where possible. The results of our study have shown that flying foxes do visit flowering durian orchards to feed on nectar-and importantly, without causing damage to flowers. We also show that there is a greater density of flying foxes feeding at higher levels in the trees, which allows the use of camera-trap data to determine the effect of flying foxes on reproductive success. Pollination by flying foxes appears to have a positive effect on mature fruit set-this expands the current scant body of knowledge on chiropterophily in the Old World.
Due to small sample size, however, caution must be exercised in interpreting the data, and further replication with larger datasets is needed in order to test the validity of the results in this study. However, the ecological, evolutionary, economic, and cultural importance of this bat species should not be underestimated and warrants further exploration. Ching for providing helpful literature, and Jamie Wadey for fruitful discussions and ideas. We thank two anonymous reviewers for providing constructive feedback to improve the manuscript. Last but not least, we are deeply grateful to Sanjay and Ryan Ramarao for spotting and counting the damaged flower corollas!

AUTHOR CONTRIBUTIONS
Sheema Abdul Aziz conceived and designed the study, collected the data, analyzed the data, contributed equipment/materials/analysis tools, wrote the manuscript, prepared figures and/or tables, and reviewed drafts of the manuscript. Sara Bumrungsri conceived and designed the study, provided technical input and advice, and reviewed drafts of the manuscript. Gopalasamy Reuben Clements collected the data, analyzed the data, contributed equipment/materials/analysis tools, wrote the manuscript, prepared figures and/or tables, and reviewed drafts of the manuscript. Kim R McConkey provided technical input, collected the data, and reviewed drafts of the manuscript. Tuanjit Sritongchuay provided training, advice and input on study design, and reviewed drafts of the manuscript. Saifful Pathil and Muhammad Nur Hafizi Abu Yazid contributed equipment/materials and collected the data. Ahimsa Campos-Arceiz contributed equipment/materials/analysis tools, provided input on data analysis, and reviewed drafts of the manuscript. Pierre-Michel Forget provided input on study design, contributed equipment, and reviewed drafts of the manuscript.

SUPPORTING INFORMATION
Additional Supporting Information may be found online in the supporting information tab for this article.