Seasonal variation in food availability and relative importance of dietary items in the Gambian epauletted fruit bat (Epomophorus gambianus)

Abstract The Gambian epauletted fruit bat (Epomophorus gambianus) is very common across a variety of West African habitats, but very little information is available on its feeding ecology or its contribution to ecosystem function. We investigated seasonal variation in food availability and the relative importance of dietary items used by this species in a forest‐savannah transitional ecosystem. Dietary items were identified from 1,470 samples of fecal and ejecta pellets which had been collected under day roosts or from captured bats over a 2‐year period (2014–2015). Plant phenology studies illustrated strong seasonal correlations between fruiting and flowering and rainfall patterns: Fruits were available throughout the year but with peaks of abundance during the rainy season, while flowers were mostly abundant during the dry season. Epomophorus gambianus bats utilized fruit and flower resources from 30 plant species. Although the plant species used depended on seasonal availability, there were clear preferences for certain species. Flowers were an important food source for this fruit bat species especially during the dry season, contributing up to 79% of dietary items when fruit abundance was low. Ficus fruits were also important food item for E. gambianus, constituting over 40% of all dietary samples identified. Policy implications. Our results show the importance of flowers in the diet of E. gambianus and highlight this species as an important pollinator and seed disperser, including for economically and ecologically important plant species. These findings contribute to an improved understanding of the ecological importance and potential role of this species in the forest‐savannah transition ecosystem for the development of fruit bat conservation management strategies.


| INTRODUC TI ON
Dietary studies can increase our understanding of the interrelationships between animals and their environment and how individual species affect and contribute to their ecosystems (Stier & Mildenstein, 2005).
Fruit bats (Pteropodidae) are almost exclusively phytophagous, relying extensively on fruits, flowers, and leaves (Aziz, Olival, Bumrungsri, Richards, & Racey, 2016;Marshall, 1985). Through their feeding interactions with plants, many fruit bat species provide vital ecosystem services and are, therefore, regarded as keystone species. The biology of bats, particularly their reproduction and migration, is influenced by the distribution and timing of food availability (Marshall, 1985), which varies across landscapes and seasons (Cumming & Bernard, 1997). The timing of fruiting/flowering is often irregular across landscapes, which makes it difficult to generalize findings (such as reproductive synchrony with food abundance) obtained from one landscape to others (Cumming & Bernard, 1997). The utilization of food resources is also not uniform among fruit bats, and some studies, for example Baker and Harris (1957), Marshall and McWilliam (1982) and Barclay and Jacobs (2011), have shown inter-and intraspecific variations in food utilization in some African members of the Pteropodidae.
Very few studies have explored the relative use and importance of different food items in the diet of fruit bats (Marshall, 1985;Stier & Mildenstein, 2005). Most studies that explored the dietary resources of fruit bats failed to quantify the relative use of identified dietary resources, primarily providing lists of food items and apparently assuming equal use (Stier & Mildenstein, 2005). The current lack of knowledge on the relative use of food items limits efforts to assign trophic roles to fruit bats (Marshall, 1985), resulting in inadequate characterization of the importance of their roles in the functioning of ecosystems. This limits our understanding of how bats impact ecosystems at local or larger scales and how changing land use and habitat modifications affect bat populations (Stier & Mildenstein, 2005;Wood et al., 2012).
Epomophorus gambianus (family Pteropodidae) is a medium sized fruit bat that is widespread throughout much of West Africa ( Figure 1). The species is commonly described as a generalist and opportunistic fruit feeder that thrives well in degraded forests and a variety of human modified habitats. We studied the diet of this widespread species in a West African forest-savannah transition ecological zone, and here, we describe the seasonal availability and relative use of food items for this species. Our findings contribute to an improved understanding of the ecological importance and potential role of this species in the forest-savannah transition ecosystem for the development of fruit bat conservation management strategies.

| Study area
Data for this study were collected in Golokuati, a town located in the Volta region of Ghana (N 06°59.851′ E 000°26.218′) from January 2014 to December 2015. A large E. gambianus colony (ca. 5,000 bats) that roosts in trees within the town was the focus of this study.
The vegetation of the area is a transition between semi-deciduous forest and Guinea savannah woodland, but it has undergone significant changes mainly due to agricultural use, timber exploitation, and human settlement, resulting in a heterogeneous landscape with remnant pockets of the original forest, regenerating secondary forest and farm bush. The area experiences the wet semi-equatorial climate where rainfall occurs in a double maxima pattern (annual range 1,016-1,210 mm),but this is changing gradually to a single extended rainy season (late April-October) and a 4-5 months dry season between November and April (Ghana Statistical Service, 2014). The mean annual temperature is 29°C, ranging between 26 and 32°C.

| Diet identification by fecal and ejecta sample analysis
The dietary components of E. gambianus were identified from collections of fecal and ejecta pellets under roosting and feeding trees and from captured bats. Fecal samples were collected by placing 1.5 × 1 m plastic sheets directly under day roosts of E. gambianus during the early hours of the day (0500-0900 hr).
To avoid repeated collection of fecal samples from the same individuals that could arise from splatters or discontinuous deposition, feces that were within 5 cm and had same characteristics (color, texture) of already collected material were ignored (Stier & Mildenstein, 2005). Fresh ejecta pellets were collected under feeding roosts within the study area. Fecal/ejecta sample collection was done on two to three consecutive days in each sampling month for 21 months.
Bats captured through mist-netting for other studies also provided opportunities to collect additional samples and data on dietary sources for E. gambianus; trapping was done between 1900 F I G U R E 1 A Gambian epauletted fruit bat (Epomophorus gambianus). Credit: Kofi Amponsah-Mensah and 0500 hr using ground mist nets. Any whole or remnants of fruits or feces or ejecta pellets expelled by E. gambianus bats during the process were collected opportunistically for the dietary study. Bats trapped were checked also for the presence of fresh pollen on their nostrils or wings, or for remnants of flower parts in their mouths.
The occurrence of pollen/flower parts on the body or in the mouth of a bat was used as a proxy for feeding on flower resources by that bat and was counted as a "flower resource" sample.
We conducted informal interviews with local farmers to help locate plants known to be utilized by bats in West Africa (Marshall, 1985) and also to identify plants which they knew to be utilized by bats. Plants identified by the local farmers were monitored for visits by E. gambianus to confirm their use as a food source. Other fruiting and flowering trees in the study area were also monitored for visits and feeding activity by E. gambianus bats.
Each ejecta and fecal sample was collected separately into a clean, new plastic bag using a wooden spatula. Latex gloves were worn during fecal and ejecta pellet collection process. Each sample collected was washed through a 0.3 mm sieve using tap water in the field before examining for seeds, flower parts, and other food particles with the aid of a handheld magnifying glass. A reference collection based on fruits and seeds, and their characteristics (color, smell, texture) of fruiting plants occurring in the study area was developed and used to identify the fruits and seeds collected from the fecal and ejecta samples (Djossa, Fahr, Kalko, & Sinsin, 2008;Picot, Jenkins, Ramilijaona, Racey, & Carriere, 2007;Stier & Mildenstein, 2005). Seeds from up to five different species of fig that occurred at the study site were very similar and could not be identified to the species level. Hence, all fig seeds from ejecta/fecal samples were pooled together as "Ficus spp." Each dietary material identified in each fecal or ejecta sample was recorded to occur just once in that sample (i.e., as presence only without quantification within a sample) as it was impossible to quantify the number or quantity of particular food items within each fecal/ejecta sample collected. Both fecal and ejecta samples were treated as equal and dietary items identified from both were treated the same and pooled together.
Dietary materials identified from samples were expressed as relative abundance (percentage) of total monthly samples collected, with the assumption that each sample collected was from an individual bat.
All dietary materials were assumed to be equal; the calorimetric or energetic content of dietary items was not determined in this study.

| Estimation of monthly food resource availability and abundance, and the timing of food resource abundance in relation to rainfall
To estimate availability and seasonal variations in food resources, plants identified as being utilized by E. gambianus were monitored monthly for fruiting and flowering abundance. Fruit/flower abundance was determined through visual estimation (Chapman et al., 1992) by a single observer (KAM) throughout the study to maintain consistency and to increase internal validity. Fruit/flower abundance was estimated by a modification of the method described by Devineau (1999), which entailed categorization of flowering and fruiting phases into four stages (0-3) with corresponding phenology scores of 0, 0.09, 0.5, and 1 (Table 1) We tested for an association between food availability and dietary use using spearman ranked correlation coefficient (ρ). We used the estimated percentage of fruits/flowers that are matured on each monitored plant (as described in Table 1) as a measure of food abundance rather than the estimated monthly total number of fruits on each plant. We chose this measure of abundance because early fruiting/flowering stages of plants produce high numbers of fruits/flowers which are immature and not readily available for use by bats, therefore, using these estimates will produce false-negative correlations.
To determine the timing of flower and fruit abundance in relation to rainfall within the study area, we related mean monthly fruit and flower abundance to the mean monthly rainfall for the study area. Rainfall data for the study area were obtained from the Ghana Meteorological Agency (www.meteo.gov.gh).  Thirty species of plant, belonging to at least 16 families, were identified in the E. gambianus fecal and ejecta samples examined (  and Solanum sp. however was weakly and negatively correlated to their abundance. The use of Ficus spp. was mostly relatively higher than that of other food items with similar monthly abundances.

| Timing of food resource abundance in relation to rainfall
Twenty-two plant species were monitored for the timing and abundance of fruits and/or flowers. Fruits of many species, for example, Psidium guajava, Solanum sp., A. vogelii, and Ficus spp., were available throughout the sampling period but with varying fruit abundances Mean monthly rainfall in the study area was high over a continuous 7-to 8-month period with peaks in June and October. These two peaks in rainfall were separated by a short period of reduced rain in August and a 5-month-long dry season following the October peak ( Figure 5). The main period of flower abundance occurred during the dry season, November to February, with a peak in December. A lower incidence of flowering occurred from April through to August with minimum flower abundance around September-October. Compared to flowers, fruits were relatively more available throughout the year, with a major peak in fruit abundance occurring in March and a minor peak in October.

| Dietary composition
The diet of E. gambianus within our study area in Ghana, West Africa, was found to consist of 30 plant species from at least 16 families.
These results confirm the varied diet of this species (Boulay & Robbins, 1989;Marshall, 1985;Marshall & McWilliam, 1982). We were unable to identify only 42 samples from 1,470 that did not match our reference collection, possibly because they were eaten outside our sampling area; therefore, our results provide a conservative estimate of the dietary sources available to E. gambianus in this forest-savannah transition zone.
With the exception of the family Cucurbitaceae, all of the plant families we identified were already known to be eaten by fruit bats in West Africa (Marshall, 1985;Marshall & McWilliam, 1982;Mickleburgh, Hutson, & Racey, 1992;Rosevear, 1965). Most food items identified as part of the diet for E. gambianus appeared generally to be linked to the seasonal availability of these food items within the study area, thus supporting the description of this bat species as an opportunistic feeder (Boulay & Robbins, 1989;Marshall, 1985;Marshall & McWilliam, 1982). Although seasonal availability may dictate the use of food resources in fruit bats, some species can also exhibit preferential diet selection. For example, Baker and Harris (1957)  E. gambianus and most other fruit bats could be spatio-temporal "sequential specialists," with a preference for some food items among those available, rather than true generalists (Marshall, 1985).  Marshall (1985); Marshall and McWilliam (1982); c: indigenous knowledge; *Unidentified trees.      (Thomas & Marshall, 1984). Thus, these two plant species together with Ficus spp. could be important dietary items during such vital stages in the reproduction of E. gambianus.

| Relative importance of food of different plant items in the diet of Epomophorus gambianus
While fruit bats usually feed mostly on fruits, flowers also constitute important dietary items (Marshall, 1985). However, very few studies have assessed the extent of flowers in the diet of fruit bats.
In our study, flower resources contributed over 9% of the dietary items identified. In the dry ("lean" fruiting) seasons, flowers contributed up to 79% of dietary items, although our analyses did not allow us to estimate the proportion of calorific intake in different months.
We identified the flowers of at least 10 plant species as being utilized by E. gambianus, all of which have been reported previously as food F I G U R E 2 Correlations between the abundance and use of dietary items of Epomophorus gambianus. Spearman ranked correlation coefficient (ρ) values are indicated beside species/food items. Abundance is expressed as the percentage of the estimated fruits/flowers that are matured per tree. The use of dietary items is expressed as the number of times that dietary item is recorded in total monthly fecal collections. Only dietary items that were detected in the diet of the bats in three or more sampling months were used. One outlier value for Ficus spp. (use = 288, abundance = 37.5) not shown on graph F I G U R E 3 Estimated mean monthly fruiting abundance of plant species whose fruits were identified to be consumed by Epomophorus gambianus. Abundance is estimated as the total number of fruits × monthly phenology score per tree sources for other fruit bat species in West Africa (Baker & Harris, 1957;Marshall, 1985;Marshall & McWilliam, 1982;Mickleburgh et al., 1992).  (Andriafidison et al., 2006), where flowers from a single tree could sustain a large chiropteran community over a short period (Gribel, Gibbs, & Queiroz, 1999). Flowers can contribute considerable amounts of dietary protein (Law, 1992a;Long & Racey, 2007;Nelson, Miller, Heske, & Fahey, 2000;Ruby et al., 2000), supplementing the low protein content of most fruits (Barclay & Jacobs, 2011;Marshall, 1985;Ruby et al., 2000;Stier & Mildenstein, 2005).
Although E. gambianus was observed to visit flowering trees and several trapped bats were covered in pollen, we could not confirm if these bats were actively eating pollen, nectar, or both.
The importance of pollen in the diet of E. gambianus is unclear. Boulay and Robbins (1989) stated that there was no evidence to support E. gambianus feeding on pollen, based on the absence of pollen in analyzed gut contents Harris (1957). Happold (2013) also suggests that this species visits flowers for their nectar but not the pollen. However, Pteropodidae bats may lick pollen directly from anthers during feeding (Marshall, 1985) or ingest it during grooming of fur dusted with pollen after visiting plants to feed on nectar (Law, 1992b). Pteropus spp. may actually feed and depend on pollen as an important food source (Long & Racey, 2007;Marshall, 1985;Mickleburgh et al., 1992). Andriafidison et al. (2006) report that in Madagascar, over 40 percent of the 118 plant taxa that have been identified as part of the diet of Pteropus rufus and Eidolon dupreanum were identified from pollen in feces.
They suggested that the contribution of pollen and nectar to the diet of fruit bats may be largely underestimated and the low reporting of pollen in the diet of bats could be due to a surveillance bias or methodological constraints. We were not equipped to carry out microscopic analysis of pollen; therefore, further studies of the pollen content of feces are needed to confirm its importance as a dietary material in fruit bats.
F I G U R E 4 Estimated mean monthly flowering abundance of plants whose flowers were identified to be utilized by Epomophorus gambianus. Abundance is estimated as total number of flowers × monthly phenology score per tree F I G U R E 5 Abundance and the timing of flowering and fruiting in relation to mean monthly rainfall. Combined mean monthly fruit and flower abundance over 2 years were used

| Timing of food resources
Food availability for African fruit bats is constrained by rainfall as fluctuations in fruit abundance are tied to rainfall patterns (Cumming & Bernard, 1997;Happold & Happold, 1990;Rautenbach, Kemp, & Scholtz, 1988). Fruit phenology studies show an abundance of fruits during the rainy season, while flowering occurs in the preceding dry season (Frankie, Baker, & Opler, 1974;Janzen, 1967). Our findings are consistent with this pattern; however, our findings demonstrate that fruits are available throughout the year, but with peaks in fruit abundance occurring at the onset of, or during, the rainy season. Each fruiting period occurred a month or two after the flowering period.
In drier areas of Africa, with stronger seasonal variations in rainfall, seasonality in fruiting and flowering may be more pronounced, causing fruits to be less available throughout much of the year (Cumming & Bernard, 1997). In such areas, peaks in flower abundance might play an even more important role in the diet of frugivorous bats during periods of low fruit abundance.
The timing and seasonality of food abundance are particularly important to the timing of reproduction in fruit bats (Cumming & Bernard, 1997;Happold & Happold, 1990), where birthing is timed such that food is available for juveniles after they are weaned (Cumming & Bernard, 1997;Fleming, Hooper, & Wilson, 1972). In West Africa, annual bimodal peaks in parturition for E. gambianus occur in April/May and in October/November (Thomas & Marshall, 1984). Lactation lasts about 7-13 weeks (Nowak, 1999;Thomas & Marshall, 1984), implying that the first annual postweaning period for E. gambianus occurs around June/July, coinciding with a period of availability of both flowers and fruits. The second postweaning period occurs around December/January, which coincides with the observed major peak in flower abundance. These observations suggest that flowers might have an important role in supporting the maintenance and growth of newly weaned bats.  (Hawthorne & Gyakari, 2006). These plants contribute substantially to ecosystem biomass, and fruits of plants like Ficus spp.
By spreading the seeds of these species, therefore, E. gambianus likely plays an important role in the maintenance and regeneration of forest vegetation, and the persistence of original forest plant species. This role could be especially important when considering current levels of forest degradation and loss.
In addition to helping to maintain forest ecosystem function, the  (Hawthorne & Gyakari, 2006). In particular, M. excelsa is heavily exploited to the extent that it is now threatened by overexploitation in Ghana (Hawthorne & Gyakari, 2006;Taylor, Kankam, & Wagner, 2000).
Fruits of plants, such as V. doniana, S. mombin, Syzygium sp., and P. guajava, are sold and consumed locally and contribute to the diet and income of people, especially rural dwellers across West Africa.
Even though plants like M. indica, P. guajava, and Anacardium occidentale are cultivated on large scales and may no longer rely on bats for their dispersal or pollination, fruit bats such as E. gambianus remain relevant for the maintenance of the genetic diversity of their wild types (Kunz et al., 2011). The services provided by E. gambianus through its foraging highlight the importance of this species (and fruit bats in general) to the ecosystems in which they occur. Evidence of these services, such as those presented in this study, therefore, should be used to inform the public and policymakers to promote the conservation of E. gambianus and other fruit bat species.

ACK N OWLED G M ENTS
This work was part-funded by the Carnegie Corporation of New