Diet composition of the African manatee: Spatial and temporal variation within the Sanaga River Watershed, Cameroon

Abstract The present study aimed to investigate the diet of African manatees in Cameroon to better inform conservation decisions within protected areas. A large knowledge gap on diet and seasonal changes in forage availability limits the ability to develop informed local management plans for the African manatee in much of its range. This research took place in the Sanaga River Watershed, which includes two protected areas in the Littoral Region of Cameroon: the Douala‐Edea National Park and the Lake Ossa Wildlife Reserve. We analyzed 113 manatee fecal samples and surveyed shoreline emergent and submerged vegetation within the Sanaga River Watershed. We used microhistological analyses to determine the relative contribution of each plant species to African manatee diets and compared across locations and across seasons (wet vs. dry season). We found that the shoreline vegetation is diverse with over 160 plant species, unevenly distributed across space and season, and dominated by emergent vegetation mostly represented by the antelope grass (Echinochloa pyramidalis). We recorded a total of 36 plant species from fecal samples with a spatial and temporal distribution mostly reflecting that of the corresponding shoreline vegetation. African manatees appear to be primarily opportunistically feeding on available vegetation across the seasons and habitat. This work documents the current, but changing, state of plant availability in the Sanaga River Watershed and reports the African manatee diet in Cameroon for the first time. This information can play a critical role in successfully managing the species and these protected areas. If we wish to protect the African manatee and the aquatic ecosystems within the Sanaga River Watershed, we must understand how forage availability changes over time, especially as its waters become nutrient enriched, eutrophic, and exposed to invasive species of plants in a changing world.


| INTRODUC TI ON
The Order Sirenia, also known as sea cows, consists of four extant aquatic and hindgut fermenter herbivorous species: the West Indian manatee, the Amazonian manatee, the dugong, and the African manatee. The African manatee is the least studied of all sirenians (Marsh et al., 2011). They are threatened by poaching, accidental catches in fishing nets, entrapment by dams, and habitat degradation-despite legal protection in all 21 countries within their range, as well as international laws (Powell, 1996). They are Red-listed by the International Union for Conservation of Nature (IUCN) as "Vulnerable" and belong to Appendix I of the Convention on International Trade in Endangered Species (CITES) and of the Convention on Migratory Species (CMS).
Understanding the influences on a species' distribution and its seasonal diet can be crucial for effectively managing wildlife populations. The availability of food resources is an essential component of a species' ecological niche and contributes to the species' distribution patterns (Cox & Moore, 2000). A large knowledge gap on diet and seasonal changes in forage availability limits the ability to develop informed local management plans for the African manatee in much of its range. Plant species composition and feeding strategies of the African manatee differ widely across habitats (Keith-Diagne, 2014;Powell, 1996); thus, management plans must consider local food availability and dietary needs of local manatee populations. Akoi (2004) examined 35 African manatee fecal samples from the Ivory Coast and corroborated earlier findings of Best (1983) that the African manatee diet is composed predominantly of grasses but also includes fruit and deposited organic materialparticularly during the dry season when decreased water levels limit access to emergent vegetation. In later work, Keith-Diagne (2014) determined the lifetime diet composition of 24 African manatees via carbon and nitrogen stable isotope analysis of ear bones from carcasses recovered in Senegal and Gabon. The stable isotope signatures recorded from samples collected in Gabon indicated that their diet was made up of 90-94% of plants and 6-7% of hermit crabs. The diets from Senegal, however, differed substantially and were composed of 46-57% of plants, 24-27% of fish, and 19-24% of mollusks. This omnivorous behavior, which has been anecdotally reported by local fishermen in many countries (Dodman et al., 2008;Mayaka et al., 2019;Powell, 1996;Reeves et al., 1988;Takoukam Kamla, 2012), had been previously considered opportunistic, only when plants are not available, but these results from Senegal suggest otherwise.
In Cameroon, manatees are present in the lower reaches of most rivers and lakes with a direct connection to the Atlantic Ocean. Lake Ossa and the reaches of the Sanaga River downstream of the hydroelectric dam of Edea provide a dry season sanctuary for manatees within the watershed (Powell, 1996) (Figure 1). These areas encompass two adjacent protected areas including the Lake Ossa Wildlife Reserve and the National Park of Douala-Edea, respectively, which were set aside for manatee conservation. Yet, the African manatee is an elusive and cryptic species, living in the turbid waters of the Sanaga River Basin where water transparency in most places is less than a meter . Due to the poor visibility, it is difficult to conduct direct observational studies on the feeding behavior and diet of the African manatee in these areas. To establish an efficient management plan, information on manatee use of the habitat and diet composition is needed for the two protected areas. Results of this study aim to identify important manatee use areas within the protected areas, areas for habitat restoration, and to help inform conservation efforts to ensure the long-term survival of manatees in this region.
The purpose of this study is to contribute to the conservation of the African manatee in the Sanaga River Watershed by determining manatee diet composition and characterizing the aquatic vegetation present in the area. More specifically, the goals of this study were to (a) document aquatic plants present in the Sanaga River Watershed, (b) document the diet composition of African manatees in the Sanaga River Watershed, and (c) assess differences and similarities in diet composition by locations or habitats and water levels (seasonal variation).

| Study area
The Sanaga River Watershed encompasses two protected areas in the Littoral Region of Cameroon ( Figure 1). The Douala-Edea National Park (3°14′ 3°50′N; 9°34′-10°03′E) was created in 1932 and covers approximately 160,000 ha of land and water (Blaikie & Simo, 2000). The park stretches along both sides of the lower reaches of the Sanaga River and 100 km along the Atlantic coastline of Cameroon (Feka et al., 2009). The park surface is predominantly a tropical lowland equatorial forest (80%) and covered by about 6.4% of mangrove, which is seriously threatened by deforestation. The local fishing community utilizes mangrove wood to smoke fish, which is the major economic activity in the area.
The Lake Ossa Wildlife Reserve (3°45′-3°52′N; 9°45′-10°4′E) is a complex of lakes located about 13 km from Edea, Cameroon (Wirrmann & Elouga, 1998). The water surface is estimated to be 4000 ha, which represents about 90% of the Lake Ossa Wildlife Reserve. The reserve was established to provide a refuge for the protection of the African manatee.

| Sampling design
We used boats to survey submerged and emergent vegetation along the shoreline of four areas within the Sanaga River Watershed (Figure 1) twice between June and September 2016 to characterize the vegetation. They were also frequently visited between June 2015 and November 2017 in search of free-floating manatee fecal samples. These areas represented three habitat types: lake (Lakes Tissongo and Ossa), river (Sanaga River), and estuary (Sanaga Estuary; Figure 1). Plant surveys and manatee fecal collection were conducted during the high-water season (average depth >2 m) in Lake Tissongo, the Sanaga River, and the Estuary, but were collected during both the high-and the low-water seasons (average depth <2 m) in Lake Ossa. The low-water season extends from November to May or June, and the high-water season extends from July to October or November (Nguetsop et al., 2004).

| Fecal sample collection
A total of 113 fecal samples (112 free-floating manatee boluses and one hindgut sample taken from the lower intestine of a stranded calf carcass) were examined using microhistological analysis techniques, described below. The distribution of fecal collection sites in our study area is presented in Table 1 and Figure S1 in the Appendix S1. In Lake Ossa, 29 fecal samples were collected during the high-water season and 31 during the low-water season. Samples were preserved in 70% ethanol before examination (Allen et al., 2018;Hurst & Beck, 1988

| Habitat characterization and plant library collection
In order to understand the variety of plant species accessible to the African manatee in each location, the shoreline vegetation was surveyed, and representative samples were collected. A total of 238 100-m transects were evenly spaced along the shoreline of the study area. Each transect consisted of two to six 1 × 1-m plots for a total of 958 plots distributed across the four study locations (Table 2; Finally, the percent coverage by plants in the plot was estimated.

| Reference Plant slide preparation
Reference slide preparation was done following the protocol described in Hurst and Beck (1988)

| Microhistological analysis
Fecal samples were examined using microhistological features of undigested plant fragments and were analyzed using the techniques described by Hurst and Beck (1988 could not be identified but with distinctive microscopic features, a generic name was assigned, for example, "Unidentified 1." The same generic name was given to fragments with the same features.

| Quantification of percent food plant species occurrence
Each of the five slides for each fecal sample was examined using 20 different fields of view indicated by the vertical and horizontal graduations of the mechanical stage as described by Hurst and Beck (1988

TA B L E 2 Survey effort and diversity index of shoreline vegetation by location in the lower Sanaga River Watershed
The occurrence of each plant species on a slide was counted, then the percent occurrence in a fecal sample was obtained by tallying occurrence for each species from each of the five slides. Thus, each fecal sample was examined at 500 intercept points (5 slides × 20 fields of view × 5 intercepts). The percent occurrence of each food plant species was averaged by location, season, and a categorical feces size (i.e., small, medium, large).

| Habitat shoreline species characterization
All data were analyzed using EXCEL Data Analysis tools and XLSTAT-Ecology (De Levie, 2004;Guerrero, 2019;Middleton, 1996) and R (R Core Team, 2017). Plot locations were mapped using ArcGIS (version 10.6). For each location, the number of unique plant species encountered, and the relative abundance of each species, was computed by averaging the percent occurrence per plot by the total number of plots for that location. The relative abundance of the plant families and types was also computed. The species diversity of the shoreline vegetation was estimated using the Shannon diversity index H (Shannon, 1948).
where n is the total number of species, P i the proportion of each species, i.
The difference in species composition among locations was measured using the Bray-Curtis dissimilarity matrix (Bray & Curtis, 1957). In order to buffer the influence of strongly dominate species, the relative abundance of each species was standardized by first taking the logarithm of the relative abundance before calculating the distance matrix (Kindt & Coe, 2005).
The analysis of similarities (ANOSIM) was performed on the PAST 3.24 software (Hammer et al., 2001) to test for the significance of the difference in species composition between location and sites. ANOSIM is a non-parametric statistical test similar to an ANOVA test which uses a permutation and randomization methods from a ranked similarity matrix to generate the Rstatistics that determines whether the similarity between groups is greater than or equal to the similarity within groups (Clarke, 1993). All ANOSIM tests in this study were performed with 10,000 permutations.
The similarity of percentages (SIMPER) computed the contribution of each species in the dissimilarities between locations and seasons using the Bray-Curtis similarity index of the most frequent species. The Bonferroni correction was applied to compensate for errors due to the multiple comparisons test (Clarke, 1993). Shoreline species composition and diversity were also compared between seasons for Lake Ossa as it was the only location surveyed during both the low-and high-water seasons.  (Kruskal & Wallis, 1952).
A Venn diagram was built through web-based software (Heberle et al., 2015) to analyze the inclusivity and exclusivity between diet and potential food plants by location.
The dominant plant species in Lake Ossa and the Upper Sanaga was E. pyramidalis, while the dominant plant species in Lake Tissongo and the Sanaga Estuary were E. macrocarpa and Rhizophora racemosa, respectively (Figure 2b). The Venn diagram in Figure S3   species encountered during the high-water season (101) was 2.6 times higher than in the low-water season (39). Echinochloa pyramidalis, D. erecta, and D. falcipila were more abundant during the lowwater than during the high-water season (Figure 3a). The Shannon diversity index was higher (H = 2.74) during the high-water than during the low-water (H = 2.01) season.

| Diet composition from African manatee fecal samples in the Sanaga River watershed
The number of unique plant species recorded in each fecal sample ranged from one to nine, with an average of four. Most plant fragments (89.2%) were identified to at least the genus level. Among the 113 fecal samples analyzed, vegetative material from 31 plant species (from nine plant families) consumed by the African manatee were observed (Table 5), in addition to fruits from six species and one emergent macrophyte. The greatest number of forage plant species was found in Lake Ossa (24) and the Sanaga Estuary (24), and the lowest number was found in the Upper Sanaga (15) and Lake Tissongo (16). molesta, Leptochla sp., Panicum sp., Ficus sp., and Macaranga sp. (Akoi, 2004;Husar, 1978;Keith-Diagne, 2014;Powell, 1996;Reeves et al., 1988 Note: Lightly shaded rows represent plant Families, whereas darkly shaded rows represent larger taxonomic groups. n = the number of fecal samples that contained the focal plant. a Newly reported African manatee food plants. b Fruit pieces in feces were not observed using the microhistological analysis protocol described in the methodology. Each fecal sample (including those that were not selected for microhistological analysis in this study) was systematically visually examined to identify fruits, scales, and other debris present. Thus, we did not provide percent frequency and n value for the fruits identified. c Plant species already listed once above within Arecaceae.

TA B L E 5 (Continued)
F I G U R E 3 Relative abundance of top 15 plant species identified along the shorelines of Lake Ossa (a) and within African manatee feces (b) by season (water level) across all locations (Table S2). Echinochloa pyramidalis was the dominant species in all locations but was more abundant in feces from Lake Ossa (mean = 63.2%, SD = 33.34) and Upper Sanaga (mean = 59.7%, SD = 20.0); while E. macrocarpa was abundant only in Lake Tissongo (mean = 29.7%, SD = 35.3); Cyperus sp. was very abundant in the Sanaga Estuary (mean = 29.9%, SD = 44.12). Unidentified 13, a mostly filamentous plant, was only present and abundant in the Sanaga Estuary (mean = 12.1%, SD = 27.9), and Rynchospora corymbosa was most abundant in Lake Ossa (mean = 8.3%, SD = 12.9).

| Manatee diet by season in Lake Ossa
A total of 60 fecal samples collected in Lake Ossa were analyzed, including 29 collected during the high-water season and 31 during the low-water season. The analysis of diet composition in manatee feces indicated that there was a moderate but significant difference between seasons in Lake Ossa (R = 0.18, p = .00001). A total of 22 and 18 different plants were recorded in the fecal samples collected in Lake Ossa during the high-and low-water seasons, respectively.
The major contributors to the dissimilarity between seasons in Lake Ossa were E. pyramidalis ( (Table   S3). Echinochloa pyramidalis fragments were two times more abundant in manatee feces during the low-water season (mean = 81.1%, SD = 21.6) than the in high-water season (mean = 44.2%, SD = 33.1).
Cyperus sp. was 40 times more abundant in manatee feces during the high-water season (mean = 15.3%, SD = 24.8) than during the lowwater season (mean = 0.4%, SD = 1.48; Figure 2c). Figure S4 gives a clear visualization of the manatee diet composition profile between the low-and high-water seasons.

| D ISCUSS I ON
The high plant diversity along the shorelines (H = 3.52, 160 morphospecies) of the Sanaga River Watershed is characteristic of tropical areas of southern Cameroon that are considered to have the richest flora in continental tropical Africa (Myers, 1988). This high diversity of aquatic plant species likely provides a broad array of food options for the African manatee. Yet, plant species composition was highly variable among the four study locations, reflecting the difference in habitat type, water quality and salinity among those locations.
Lake Ossa and the Sanaga Estuary were the most dissimilar in plant composition, which is unsurprising because the distance between the two is the greatest (40 km) among locations. Also, Lake Ossa is purely freshwater while the Sanaga Estuary is brackish water that is under the influence of the tides. The manatee diet composition in these two locations was the most dissimilar. This suggests that manatees are opportunistically feeding on the most available plants in each location. Similar observations on the influence of the local vegetation on manatee diet have been reported among African manatee populations in the Ivory Coast (Akoi, 2004), Amazonian manatees (Guterres-Pazin et al., 2014), Antillean manatee in Belize (Allen et al., 2018), and the Florida manatee (Alves-Stanley et al., 2010).
Each manatee fecal sample contained between one and nine unique plant species. Similar results have been observed in Antillean manatees (Allen et al., 2018) and Amazonian manatees (Colares & Colares, 2002;Guterres-Pazin et al., 2014). This further suggests that the African manatee is likely opportunistic in their diet, similar to the Amazonian and the West Indian manatee (Hartman, 1979;Marsh et al., 2011). This also might reflect the relatively high plant diversity of the area, as manatees often visit multiple sites in a single feeding bout (Akoi, 2004 (Hurst & Beck, 1988). Differential digestion also could render some plant species unidentifiable. More comprehensive stable isotope studies from the Sanaga River will be crucial to our understanding of the level of omnivory in the diet of African manatees in this location.
Despite the abundance of mangrove in the Sanaga Estuary, no Rhizophora sp. fragment was recorded in the feces, whereas the plant has been reported as an African manatee diet item in several countries including Cameroon (Husar, 1978;Keith-Diagne, 2014;Powell, 1996). The absence of Rhizophora sp. in the diet of manatee in this study could be due to the abundance of grasses in the mangrove area which are more available and perhaps more nutrient-rich for the manatees than the mangrove leaves hanging over the estuary edges. Manatees are, in general, grazers rather than browsers (Marsh et al., 2011). It is also possible that our sampling period and effort were not representative enough to capture seasonal manatee feeding on mangrove.
Manatee diets here consisted of 96% emergent grasses, which occurred in all fecal samples analyzed-consistent with previous work in the Ivory Coast (Akoi, 2004). Floating plant species represented an insignificant fraction of our fecal samples, and no submerged plant fragments were detected. The absence of submerged aquatic plants in the entire study area could be associated with the high turbidity and low transparency of water, preventing light from reaching the bottom to sustain plant growth. This may also be due to recent increases in nutrient enrichment of the lake that might have been caused by the construction of the largest reservoir dam of the country (Lom-Pangar), upstream on the Sanaga River.
The seasonal effect on the manatee diet was explored only in Lake Ossa because of the low fecal sample size for the low-water season in the other locations, further highlighting that Lake Ossa is a refuge for manatees during the low-water season (Takoukam Kamla, 2019). There was a significant difference in manatee diet composition between the low-and high-water seasons in Lake Ossa. This result is unsurprising as the shoreline plant survey showed that the number of species available during the high season was nearly threefold greater than that of the low-water season. Indeed, manatee diets reflected this difference as more plant species were observed in feces during the high-water season.
The abundance of food plants during the high-water season may be partially responsible for the seasonal manatee breeding activity.
African manatee mating herd behavior primarily starts during the beginning of the rainy season in June/July (Cadenat, 1957;Dodman et al., 2008;Powell, 1996). The African manatee may have adapted to synchronize their breeding period with the seasonal variation in the availability of food so that calves are born during rising water when females have access to a greater amount and better quality of forage necessary to increase fat reserves (or energy reserves) to support gestation and lactation.
As water levels rise and the grass plains of the lake are flooded, manatees gain access to more diverse and abundant plant species.
When the water level drops during the dry season, manatee food availability is reduced and limited to floating or semi-floating species.
Floating plant species were scarce in Lake Ossa, and E. pyramidalis was the major emergent species. Therefore, it is not surprising that the manatee diet in Lake Ossa was highly dominated by E. pyramidalis (81%) during the low-water season whereas it only represented 44% of the diet during the high-water season.
Due to this lower food availability, relative to the high-water season, the manatee population in Lake Ossa may leave the lake and move to the Sanaga Estuary where they can take advantage of the high tide for access to shoreline plants. Indeed, interview surveys among 144 fishermen in the Sanaga River Watershed indicated that manatees are more frequently seen in the Sanaga River estuary during the low-water (dry) season than during the high-water (rainy) season (Takoukam Kamla, 2012).

| CON CLUS IONS
This work documents the current, but changing, state of plant availability in the Sanaga River Watershed and the African manatee diet in Cameroon for the first time. This information fills an existing knowledge gap and can play a critical role in successfully informing management of the species and these protected areas. For successful long-term protection of the African manatee and the aquatic ecosystems within the Sanaga River watershed, it is critical to limit nutrient enrichment, and subsequent eutrophication, and to stop the spread of invasive plants species.

ACK N OWLED G EM ENTS
We thank Visikol®, Inc. for providing the clearing and mounting

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.